linux/kernel/bpf/btf.c

9266 lines
237 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0
/* Copyright (c) 2018 Facebook */
#include <uapi/linux/btf.h>
#include <uapi/linux/bpf.h>
#include <uapi/linux/bpf_perf_event.h>
#include <uapi/linux/types.h>
#include <linux/seq_file.h>
#include <linux/compiler.h>
#include <linux/ctype.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/anon_inodes.h>
#include <linux/file.h>
#include <linux/uaccess.h>
#include <linux/kernel.h>
#include <linux/idr.h>
#include <linux/sort.h>
#include <linux/bpf_verifier.h>
#include <linux/btf.h>
#include <linux/btf_ids.h>
#include <linux/bpf.h>
#include <linux/bpf_lsm.h>
#include <linux/skmsg.h>
#include <linux/perf_event.h>
#include <linux/bsearch.h>
#include <linux/kobject.h>
#include <linux/sysfs.h>
#include <net/netfilter/nf_bpf_link.h>
#include <net/sock.h>
#include <net/xdp.h>
#include "../tools/lib/bpf/relo_core.h"
/* BTF (BPF Type Format) is the meta data format which describes
* the data types of BPF program/map. Hence, it basically focus
* on the C programming language which the modern BPF is primary
* using.
*
* ELF Section:
* ~~~~~~~~~~~
* The BTF data is stored under the ".BTF" ELF section
*
* struct btf_type:
* ~~~~~~~~~~~~~~~
* Each 'struct btf_type' object describes a C data type.
* Depending on the type it is describing, a 'struct btf_type'
* object may be followed by more data. F.e.
* To describe an array, 'struct btf_type' is followed by
* 'struct btf_array'.
*
* 'struct btf_type' and any extra data following it are
* 4 bytes aligned.
*
* Type section:
* ~~~~~~~~~~~~~
* The BTF type section contains a list of 'struct btf_type' objects.
* Each one describes a C type. Recall from the above section
* that a 'struct btf_type' object could be immediately followed by extra
* data in order to describe some particular C types.
*
* type_id:
* ~~~~~~~
* Each btf_type object is identified by a type_id. The type_id
* is implicitly implied by the location of the btf_type object in
* the BTF type section. The first one has type_id 1. The second
* one has type_id 2...etc. Hence, an earlier btf_type has
* a smaller type_id.
*
* A btf_type object may refer to another btf_type object by using
* type_id (i.e. the "type" in the "struct btf_type").
*
* NOTE that we cannot assume any reference-order.
* A btf_type object can refer to an earlier btf_type object
* but it can also refer to a later btf_type object.
*
* For example, to describe "const void *". A btf_type
* object describing "const" may refer to another btf_type
* object describing "void *". This type-reference is done
* by specifying type_id:
*
* [1] CONST (anon) type_id=2
* [2] PTR (anon) type_id=0
*
* The above is the btf_verifier debug log:
* - Each line started with "[?]" is a btf_type object
* - [?] is the type_id of the btf_type object.
* - CONST/PTR is the BTF_KIND_XXX
* - "(anon)" is the name of the type. It just
* happens that CONST and PTR has no name.
* - type_id=XXX is the 'u32 type' in btf_type
*
* NOTE: "void" has type_id 0
*
* String section:
* ~~~~~~~~~~~~~~
* The BTF string section contains the names used by the type section.
* Each string is referred by an "offset" from the beginning of the
* string section.
*
* Each string is '\0' terminated.
*
* The first character in the string section must be '\0'
* which is used to mean 'anonymous'. Some btf_type may not
* have a name.
*/
/* BTF verification:
*
* To verify BTF data, two passes are needed.
*
* Pass #1
* ~~~~~~~
* The first pass is to collect all btf_type objects to
* an array: "btf->types".
*
* Depending on the C type that a btf_type is describing,
* a btf_type may be followed by extra data. We don't know
* how many btf_type is there, and more importantly we don't
* know where each btf_type is located in the type section.
*
* Without knowing the location of each type_id, most verifications
* cannot be done. e.g. an earlier btf_type may refer to a later
* btf_type (recall the "const void *" above), so we cannot
* check this type-reference in the first pass.
*
* In the first pass, it still does some verifications (e.g.
* checking the name is a valid offset to the string section).
*
* Pass #2
* ~~~~~~~
* The main focus is to resolve a btf_type that is referring
* to another type.
*
* We have to ensure the referring type:
* 1) does exist in the BTF (i.e. in btf->types[])
* 2) does not cause a loop:
* struct A {
* struct B b;
* };
*
* struct B {
* struct A a;
* };
*
* btf_type_needs_resolve() decides if a btf_type needs
* to be resolved.
*
* The needs_resolve type implements the "resolve()" ops which
* essentially does a DFS and detects backedge.
*
* During resolve (or DFS), different C types have different
* "RESOLVED" conditions.
*
* When resolving a BTF_KIND_STRUCT, we need to resolve all its
* members because a member is always referring to another
* type. A struct's member can be treated as "RESOLVED" if
* it is referring to a BTF_KIND_PTR. Otherwise, the
* following valid C struct would be rejected:
*
* struct A {
* int m;
* struct A *a;
* };
*
* When resolving a BTF_KIND_PTR, it needs to keep resolving if
* it is referring to another BTF_KIND_PTR. Otherwise, we cannot
* detect a pointer loop, e.g.:
* BTF_KIND_CONST -> BTF_KIND_PTR -> BTF_KIND_CONST -> BTF_KIND_PTR +
* ^ |
* +-----------------------------------------+
*
*/
#define BITS_PER_U128 (sizeof(u64) * BITS_PER_BYTE * 2)
#define BITS_PER_BYTE_MASK (BITS_PER_BYTE - 1)
#define BITS_PER_BYTE_MASKED(bits) ((bits) & BITS_PER_BYTE_MASK)
#define BITS_ROUNDDOWN_BYTES(bits) ((bits) >> 3)
#define BITS_ROUNDUP_BYTES(bits) \
(BITS_ROUNDDOWN_BYTES(bits) + !!BITS_PER_BYTE_MASKED(bits))
#define BTF_INFO_MASK 0x9f00ffff
#define BTF_INT_MASK 0x0fffffff
#define BTF_TYPE_ID_VALID(type_id) ((type_id) <= BTF_MAX_TYPE)
#define BTF_STR_OFFSET_VALID(name_off) ((name_off) <= BTF_MAX_NAME_OFFSET)
/* 16MB for 64k structs and each has 16 members and
* a few MB spaces for the string section.
* The hard limit is S32_MAX.
*/
#define BTF_MAX_SIZE (16 * 1024 * 1024)
#define for_each_member_from(i, from, struct_type, member) \
for (i = from, member = btf_type_member(struct_type) + from; \
i < btf_type_vlen(struct_type); \
i++, member++)
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
#define for_each_vsi_from(i, from, struct_type, member) \
for (i = from, member = btf_type_var_secinfo(struct_type) + from; \
i < btf_type_vlen(struct_type); \
i++, member++)
DEFINE_IDR(btf_idr);
DEFINE_SPINLOCK(btf_idr_lock);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
enum btf_kfunc_hook {
BTF_KFUNC_HOOK_COMMON,
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
BTF_KFUNC_HOOK_XDP,
BTF_KFUNC_HOOK_TC,
BTF_KFUNC_HOOK_STRUCT_OPS,
BTF_KFUNC_HOOK_TRACING,
BTF_KFUNC_HOOK_SYSCALL,
BTF_KFUNC_HOOK_FMODRET,
BTF_KFUNC_HOOK_CGROUP,
BTF_KFUNC_HOOK_SCHED_ACT,
BTF_KFUNC_HOOK_SK_SKB,
BTF_KFUNC_HOOK_SOCKET_FILTER,
BTF_KFUNC_HOOK_LWT,
BTF_KFUNC_HOOK_NETFILTER,
BTF_KFUNC_HOOK_KPROBE,
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
BTF_KFUNC_HOOK_MAX,
};
enum {
BTF_KFUNC_SET_MAX_CNT = 256,
BTF_DTOR_KFUNC_MAX_CNT = 256,
BTF_KFUNC_FILTER_MAX_CNT = 16,
};
struct btf_kfunc_hook_filter {
btf_kfunc_filter_t filters[BTF_KFUNC_FILTER_MAX_CNT];
u32 nr_filters;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
};
struct btf_kfunc_set_tab {
struct btf_id_set8 *sets[BTF_KFUNC_HOOK_MAX];
struct btf_kfunc_hook_filter hook_filters[BTF_KFUNC_HOOK_MAX];
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
};
struct btf_id_dtor_kfunc_tab {
u32 cnt;
struct btf_id_dtor_kfunc dtors[];
};
struct btf_struct_ops_tab {
u32 cnt;
u32 capacity;
struct bpf_struct_ops_desc ops[];
};
struct btf {
void *data;
struct btf_type **types;
u32 *resolved_ids;
u32 *resolved_sizes;
const char *strings;
void *nohdr_data;
struct btf_header hdr;
u32 nr_types; /* includes VOID for base BTF */
u32 types_size;
u32 data_size;
refcount_t refcnt;
u32 id;
struct rcu_head rcu;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
struct btf_kfunc_set_tab *kfunc_set_tab;
struct btf_id_dtor_kfunc_tab *dtor_kfunc_tab;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
struct btf_struct_metas *struct_meta_tab;
struct btf_struct_ops_tab *struct_ops_tab;
/* split BTF support */
struct btf *base_btf;
u32 start_id; /* first type ID in this BTF (0 for base BTF) */
u32 start_str_off; /* first string offset (0 for base BTF) */
char name[MODULE_NAME_LEN];
bool kernel_btf;
__u32 *base_id_map; /* map from distilled base BTF -> vmlinux BTF ids */
};
enum verifier_phase {
CHECK_META,
CHECK_TYPE,
};
struct resolve_vertex {
const struct btf_type *t;
u32 type_id;
u16 next_member;
};
enum visit_state {
NOT_VISITED,
VISITED,
RESOLVED,
};
enum resolve_mode {
RESOLVE_TBD, /* To Be Determined */
RESOLVE_PTR, /* Resolving for Pointer */
RESOLVE_STRUCT_OR_ARRAY, /* Resolving for struct/union
* or array
*/
};
#define MAX_RESOLVE_DEPTH 32
struct btf_sec_info {
u32 off;
u32 len;
};
struct btf_verifier_env {
struct btf *btf;
u8 *visit_states;
struct resolve_vertex stack[MAX_RESOLVE_DEPTH];
struct bpf_verifier_log log;
u32 log_type_id;
u32 top_stack;
enum verifier_phase phase;
enum resolve_mode resolve_mode;
};
static const char * const btf_kind_str[NR_BTF_KINDS] = {
[BTF_KIND_UNKN] = "UNKNOWN",
[BTF_KIND_INT] = "INT",
[BTF_KIND_PTR] = "PTR",
[BTF_KIND_ARRAY] = "ARRAY",
[BTF_KIND_STRUCT] = "STRUCT",
[BTF_KIND_UNION] = "UNION",
[BTF_KIND_ENUM] = "ENUM",
[BTF_KIND_FWD] = "FWD",
[BTF_KIND_TYPEDEF] = "TYPEDEF",
[BTF_KIND_VOLATILE] = "VOLATILE",
[BTF_KIND_CONST] = "CONST",
[BTF_KIND_RESTRICT] = "RESTRICT",
[BTF_KIND_FUNC] = "FUNC",
[BTF_KIND_FUNC_PROTO] = "FUNC_PROTO",
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
[BTF_KIND_VAR] = "VAR",
[BTF_KIND_DATASEC] = "DATASEC",
[BTF_KIND_FLOAT] = "FLOAT",
[BTF_KIND_DECL_TAG] = "DECL_TAG",
[BTF_KIND_TYPE_TAG] = "TYPE_TAG",
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
[BTF_KIND_ENUM64] = "ENUM64",
};
bpf: Support bpf program calling kernel function This patch adds support to BPF verifier to allow bpf program calling kernel function directly. The use case included in this set is to allow bpf-tcp-cc to directly call some tcp-cc helper functions (e.g. "tcp_cong_avoid_ai()"). Those functions have already been used by some kernel tcp-cc implementations. This set will also allow the bpf-tcp-cc program to directly call the kernel tcp-cc implementation, For example, a bpf_dctcp may only want to implement its own dctcp_cwnd_event() and reuse other dctcp_*() directly from the kernel tcp_dctcp.c instead of reimplementing (or copy-and-pasting) them. The tcp-cc kernel functions mentioned above will be white listed for the struct_ops bpf-tcp-cc programs to use in a later patch. The white listed functions are not bounded to a fixed ABI contract. Those functions have already been used by the existing kernel tcp-cc. If any of them has changed, both in-tree and out-of-tree kernel tcp-cc implementations have to be changed. The same goes for the struct_ops bpf-tcp-cc programs which have to be adjusted accordingly. This patch is to make the required changes in the bpf verifier. First change is in btf.c, it adds a case in "btf_check_func_arg_match()". When the passed in "btf->kernel_btf == true", it means matching the verifier regs' states with a kernel function. This will handle the PTR_TO_BTF_ID reg. It also maps PTR_TO_SOCK_COMMON, PTR_TO_SOCKET, and PTR_TO_TCP_SOCK to its kernel's btf_id. In the later libbpf patch, the insn calling a kernel function will look like: insn->code == (BPF_JMP | BPF_CALL) insn->src_reg == BPF_PSEUDO_KFUNC_CALL /* <- new in this patch */ insn->imm == func_btf_id /* btf_id of the running kernel */ [ For the future calling function-in-kernel-module support, an array of module btf_fds can be passed at the load time and insn->off can be used to index into this array. ] At the early stage of verifier, the verifier will collect all kernel function calls into "struct bpf_kfunc_desc". Those descriptors are stored in "prog->aux->kfunc_tab" and will be available to the JIT. Since this "add" operation is similar to the current "add_subprog()" and looking for the same insn->code, they are done together in the new "add_subprog_and_kfunc()". In the "do_check()" stage, the new "check_kfunc_call()" is added to verify the kernel function call instruction: 1. Ensure the kernel function can be used by a particular BPF_PROG_TYPE. A new bpf_verifier_ops "check_kfunc_call" is added to do that. The bpf-tcp-cc struct_ops program will implement this function in a later patch. 2. Call "btf_check_kfunc_args_match()" to ensure the regs can be used as the args of a kernel function. 3. Mark the regs' type, subreg_def, and zext_dst. At the later do_misc_fixups() stage, the new fixup_kfunc_call() will replace the insn->imm with the function address (relative to __bpf_call_base). If needed, the jit can find the btf_func_model by calling the new bpf_jit_find_kfunc_model(prog, insn). With the imm set to the function address, "bpftool prog dump xlated" will be able to display the kernel function calls the same way as it displays other bpf helper calls. gpl_compatible program is required to call kernel function. This feature currently requires JIT. The verifier selftests are adjusted because of the changes in the verbose log in add_subprog_and_kfunc(). Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20210325015142.1544736-1-kafai@fb.com
2021-03-25 09:51:42 +08:00
const char *btf_type_str(const struct btf_type *t)
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
{
return btf_kind_str[BTF_INFO_KIND(t->info)];
}
/* Chunk size we use in safe copy of data to be shown. */
#define BTF_SHOW_OBJ_SAFE_SIZE 32
/*
* This is the maximum size of a base type value (equivalent to a
* 128-bit int); if we are at the end of our safe buffer and have
* less than 16 bytes space we can't be assured of being able
* to copy the next type safely, so in such cases we will initiate
* a new copy.
*/
#define BTF_SHOW_OBJ_BASE_TYPE_SIZE 16
/* Type name size */
#define BTF_SHOW_NAME_SIZE 80
bpf: Allow trusted args to walk struct when checking BTF IDs When validating BTF types for KF_TRUSTED_ARGS kfuncs, the verifier currently enforces that the top-level type must match when calling the kfunc. In other words, the verifier does not allow the BPF program to pass a bitwise equivalent struct, despite it being allowed according to the C standard. For example, if you have the following type: struct nf_conn___init { struct nf_conn ct; }; The C standard stipulates that it would be safe to pass a struct nf_conn___init to a kfunc expecting a struct nf_conn. The verifier currently disallows this, however, as semantically kfuncs may want to enforce that structs that have equivalent types according to the C standard, but have different BTF IDs, are not able to be passed to kfuncs expecting one or the other. For example, struct nf_conn___init may not be queried / looked up, as it is allocated but may not yet be fully initialized. On the other hand, being able to pass types that are equivalent according to the C standard will be useful for other types of kfunc / kptrs enabled by BPF. For example, in a follow-on patch, a series of kfuncs will be added which allow programs to do bitwise queries on cpumasks that are either allocated by the program (in which case they'll be a 'struct bpf_cpumask' type that wraps a cpumask_t as its first element), or a cpumask that was allocated by the main kernel (in which case it will just be a straight cpumask_t, as in task->cpus_ptr). Having the two types of cpumasks allows us to distinguish between the two for when a cpumask is read-only vs. mutatable. A struct bpf_cpumask can be mutated by e.g. bpf_cpumask_clear(), whereas a regular cpumask_t cannot be. On the other hand, a struct bpf_cpumask can of course be queried in the exact same manner as a cpumask_t, with e.g. bpf_cpumask_test_cpu(). If we were to enforce that top level types match, then a user that's passing a struct bpf_cpumask to a read-only cpumask_t argument would have to cast with something like bpf_cast_to_kern_ctx() (which itself would need to be updated to expect the alias, and currently it only accommodates a single alias per prog type). Additionally, not specifying KF_TRUSTED_ARGS is not an option, as some kfuncs take one argument as a struct bpf_cpumask *, and another as a struct cpumask * (i.e. cpumask_t). In order to enable this, this patch relaxes the constraint that a KF_TRUSTED_ARGS kfunc must have strict type matching, and instead only enforces strict type matching if a type is observed to be a "no-cast alias" (i.e., that the type names are equivalent, but one is suffixed with ___init). Additionally, in order to try and be conservative and match existing behavior / expectations, this patch also enforces strict type checking for acquire kfuncs. We were already enforcing it for release kfuncs, so this should also improve the consistency of the semantics for kfuncs. Signed-off-by: David Vernet <void@manifault.com> Link: https://lore.kernel.org/r/20230120192523.3650503-3-void@manifault.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-01-21 03:25:16 +08:00
/*
* The suffix of a type that indicates it cannot alias another type when
* comparing BTF IDs for kfunc invocations.
*/
#define NOCAST_ALIAS_SUFFIX "___init"
/*
* Common data to all BTF show operations. Private show functions can add
* their own data to a structure containing a struct btf_show and consult it
* in the show callback. See btf_type_show() below.
*
* One challenge with showing nested data is we want to skip 0-valued
* data, but in order to figure out whether a nested object is all zeros
* we need to walk through it. As a result, we need to make two passes
* when handling structs, unions and arrays; the first path simply looks
* for nonzero data, while the second actually does the display. The first
* pass is signalled by show->state.depth_check being set, and if we
* encounter a non-zero value we set show->state.depth_to_show to
* the depth at which we encountered it. When we have completed the
* first pass, we will know if anything needs to be displayed if
* depth_to_show > depth. See btf_[struct,array]_show() for the
* implementation of this.
*
* Another problem is we want to ensure the data for display is safe to
* access. To support this, the anonymous "struct {} obj" tracks the data
* object and our safe copy of it. We copy portions of the data needed
* to the object "copy" buffer, but because its size is limited to
* BTF_SHOW_OBJ_COPY_LEN bytes, multiple copies may be required as we
* traverse larger objects for display.
*
* The various data type show functions all start with a call to
* btf_show_start_type() which returns a pointer to the safe copy
* of the data needed (or if BTF_SHOW_UNSAFE is specified, to the
* raw data itself). btf_show_obj_safe() is responsible for
* using copy_from_kernel_nofault() to update the safe data if necessary
* as we traverse the object's data. skbuff-like semantics are
* used:
*
* - obj.head points to the start of the toplevel object for display
* - obj.size is the size of the toplevel object
* - obj.data points to the current point in the original data at
* which our safe data starts. obj.data will advance as we copy
* portions of the data.
*
* In most cases a single copy will suffice, but larger data structures
* such as "struct task_struct" will require many copies. The logic in
* btf_show_obj_safe() handles the logic that determines if a new
* copy_from_kernel_nofault() is needed.
*/
struct btf_show {
u64 flags;
void *target; /* target of show operation (seq file, buffer) */
__printf(2, 0) void (*showfn)(struct btf_show *show, const char *fmt, va_list args);
const struct btf *btf;
/* below are used during iteration */
struct {
u8 depth;
u8 depth_to_show;
u8 depth_check;
u8 array_member:1,
array_terminated:1;
u16 array_encoding;
u32 type_id;
int status; /* non-zero for error */
const struct btf_type *type;
const struct btf_member *member;
char name[BTF_SHOW_NAME_SIZE]; /* space for member name/type */
} state;
struct {
u32 size;
void *head;
void *data;
u8 safe[BTF_SHOW_OBJ_SAFE_SIZE];
} obj;
};
struct btf_kind_operations {
s32 (*check_meta)(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left);
int (*resolve)(struct btf_verifier_env *env,
const struct resolve_vertex *v);
int (*check_member)(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type);
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
int (*check_kflag_member)(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type);
void (*log_details)(struct btf_verifier_env *env,
const struct btf_type *t);
void (*show)(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offsets,
struct btf_show *show);
};
static const struct btf_kind_operations * const kind_ops[NR_BTF_KINDS];
static struct btf_type btf_void;
static int btf_resolve(struct btf_verifier_env *env,
const struct btf_type *t, u32 type_id);
bpf: Fix a btf decl_tag bug when tagging a function syzbot reported a btf decl_tag bug with stack trace below: general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007] CPU: 0 PID: 3592 Comm: syz-executor914 Not tainted 5.16.0-syzkaller-11424-gb7892f7d5cb2 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:btf_type_vlen include/linux/btf.h:231 [inline] RIP: 0010:btf_decl_tag_resolve+0x83e/0xaa0 kernel/bpf/btf.c:3910 ... Call Trace: <TASK> btf_resolve+0x251/0x1020 kernel/bpf/btf.c:4198 btf_check_all_types kernel/bpf/btf.c:4239 [inline] btf_parse_type_sec kernel/bpf/btf.c:4280 [inline] btf_parse kernel/bpf/btf.c:4513 [inline] btf_new_fd+0x19fe/0x2370 kernel/bpf/btf.c:6047 bpf_btf_load kernel/bpf/syscall.c:4039 [inline] __sys_bpf+0x1cbb/0x5970 kernel/bpf/syscall.c:4679 __do_sys_bpf kernel/bpf/syscall.c:4738 [inline] __se_sys_bpf kernel/bpf/syscall.c:4736 [inline] __x64_sys_bpf+0x75/0xb0 kernel/bpf/syscall.c:4736 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae The kasan error is triggered with an illegal BTF like below: type 0: void type 1: int type 2: decl_tag to func type 3 type 3: func to func_proto type 8 The total number of types is 4 and the type 3 is illegal since its func_proto type is out of range. Currently, the target type of decl_tag can be struct/union, var or func. Both struct/union and var implemented their own 'resolve' callback functions and hence handled properly in kernel. But func type doesn't have 'resolve' callback function. When btf_decl_tag_resolve() tries to check func type, it tries to get vlen of its func_proto type, which triggered the above kasan error. To fix the issue, btf_decl_tag_resolve() needs to do btf_func_check() before trying to accessing func_proto type. In the current implementation, func type is checked with btf_func_check() in the main checking function btf_check_all_types(). To fix the above kasan issue, let us implement 'resolve' callback func type properly. The 'resolve' callback will be also called in btf_check_all_types() for func types. Fixes: b5ea834dde6b ("bpf: Support for new btf kind BTF_KIND_TAG") Reported-by: syzbot+53619be9444215e785ed@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20220203191727.741862-1-yhs@fb.com
2022-02-04 03:17:27 +08:00
static int btf_func_check(struct btf_verifier_env *env,
const struct btf_type *t);
static bool btf_type_is_modifier(const struct btf_type *t)
{
/* Some of them is not strictly a C modifier
* but they are grouped into the same bucket
* for BTF concern:
* A type (t) that refers to another
* type through t->type AND its size cannot
* be determined without following the t->type.
*
* ptr does not fall into this bucket
* because its size is always sizeof(void *).
*/
switch (BTF_INFO_KIND(t->info)) {
case BTF_KIND_TYPEDEF:
case BTF_KIND_VOLATILE:
case BTF_KIND_CONST:
case BTF_KIND_RESTRICT:
case BTF_KIND_TYPE_TAG:
return true;
}
return false;
}
bool btf_type_is_void(const struct btf_type *t)
{
return t == &btf_void;
}
static bool btf_type_is_fwd(const struct btf_type *t)
{
return BTF_INFO_KIND(t->info) == BTF_KIND_FWD;
}
bpf: Silence a warning in btf_type_id_size() syzbot reported a warning in [1] with the following stacktrace: WARNING: CPU: 0 PID: 5005 at kernel/bpf/btf.c:1988 btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... RIP: 0010:btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... Call Trace: <TASK> map_check_btf kernel/bpf/syscall.c:1024 [inline] map_create+0x1157/0x1860 kernel/bpf/syscall.c:1198 __sys_bpf+0x127f/0x5420 kernel/bpf/syscall.c:5040 __do_sys_bpf kernel/bpf/syscall.c:5162 [inline] __se_sys_bpf kernel/bpf/syscall.c:5160 [inline] __x64_sys_bpf+0x79/0xc0 kernel/bpf/syscall.c:5160 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd With the following btf [1] DECL_TAG 'a' type_id=4 component_idx=-1 [2] PTR '(anon)' type_id=0 [3] TYPE_TAG 'a' type_id=2 [4] VAR 'a' type_id=3, linkage=static and when the bpf_attr.btf_key_type_id = 1 (DECL_TAG), the following WARN_ON_ONCE in btf_type_id_size() is triggered: if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) && !btf_type_is_var(size_type))) return NULL; Note that 'return NULL' is the correct behavior as we don't want a DECL_TAG type to be used as a btf_{key,value}_type_id even for the case like 'DECL_TAG -> STRUCT'. So there is no correctness issue here, we just want to silence warning. To silence the warning, I added DECL_TAG as one of kinds in btf_type_nosize() which will cause btf_type_id_size() returning NULL earlier without the warning. [1] https://lore.kernel.org/bpf/000000000000e0df8d05fc75ba86@google.com/ Reported-by: syzbot+958967f249155967d42a@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230530205029.264910-1-yhs@fb.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-05-31 04:50:29 +08:00
static bool btf_type_is_datasec(const struct btf_type *t)
{
bpf: Silence a warning in btf_type_id_size() syzbot reported a warning in [1] with the following stacktrace: WARNING: CPU: 0 PID: 5005 at kernel/bpf/btf.c:1988 btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... RIP: 0010:btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... Call Trace: <TASK> map_check_btf kernel/bpf/syscall.c:1024 [inline] map_create+0x1157/0x1860 kernel/bpf/syscall.c:1198 __sys_bpf+0x127f/0x5420 kernel/bpf/syscall.c:5040 __do_sys_bpf kernel/bpf/syscall.c:5162 [inline] __se_sys_bpf kernel/bpf/syscall.c:5160 [inline] __x64_sys_bpf+0x79/0xc0 kernel/bpf/syscall.c:5160 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd With the following btf [1] DECL_TAG 'a' type_id=4 component_idx=-1 [2] PTR '(anon)' type_id=0 [3] TYPE_TAG 'a' type_id=2 [4] VAR 'a' type_id=3, linkage=static and when the bpf_attr.btf_key_type_id = 1 (DECL_TAG), the following WARN_ON_ONCE in btf_type_id_size() is triggered: if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) && !btf_type_is_var(size_type))) return NULL; Note that 'return NULL' is the correct behavior as we don't want a DECL_TAG type to be used as a btf_{key,value}_type_id even for the case like 'DECL_TAG -> STRUCT'. So there is no correctness issue here, we just want to silence warning. To silence the warning, I added DECL_TAG as one of kinds in btf_type_nosize() which will cause btf_type_id_size() returning NULL earlier without the warning. [1] https://lore.kernel.org/bpf/000000000000e0df8d05fc75ba86@google.com/ Reported-by: syzbot+958967f249155967d42a@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230530205029.264910-1-yhs@fb.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-05-31 04:50:29 +08:00
return BTF_INFO_KIND(t->info) == BTF_KIND_DATASEC;
}
bpf: Silence a warning in btf_type_id_size() syzbot reported a warning in [1] with the following stacktrace: WARNING: CPU: 0 PID: 5005 at kernel/bpf/btf.c:1988 btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... RIP: 0010:btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... Call Trace: <TASK> map_check_btf kernel/bpf/syscall.c:1024 [inline] map_create+0x1157/0x1860 kernel/bpf/syscall.c:1198 __sys_bpf+0x127f/0x5420 kernel/bpf/syscall.c:5040 __do_sys_bpf kernel/bpf/syscall.c:5162 [inline] __se_sys_bpf kernel/bpf/syscall.c:5160 [inline] __x64_sys_bpf+0x79/0xc0 kernel/bpf/syscall.c:5160 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd With the following btf [1] DECL_TAG 'a' type_id=4 component_idx=-1 [2] PTR '(anon)' type_id=0 [3] TYPE_TAG 'a' type_id=2 [4] VAR 'a' type_id=3, linkage=static and when the bpf_attr.btf_key_type_id = 1 (DECL_TAG), the following WARN_ON_ONCE in btf_type_id_size() is triggered: if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) && !btf_type_is_var(size_type))) return NULL; Note that 'return NULL' is the correct behavior as we don't want a DECL_TAG type to be used as a btf_{key,value}_type_id even for the case like 'DECL_TAG -> STRUCT'. So there is no correctness issue here, we just want to silence warning. To silence the warning, I added DECL_TAG as one of kinds in btf_type_nosize() which will cause btf_type_id_size() returning NULL earlier without the warning. [1] https://lore.kernel.org/bpf/000000000000e0df8d05fc75ba86@google.com/ Reported-by: syzbot+958967f249155967d42a@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230530205029.264910-1-yhs@fb.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-05-31 04:50:29 +08:00
static bool btf_type_is_decl_tag(const struct btf_type *t)
{
bpf: Silence a warning in btf_type_id_size() syzbot reported a warning in [1] with the following stacktrace: WARNING: CPU: 0 PID: 5005 at kernel/bpf/btf.c:1988 btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... RIP: 0010:btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... Call Trace: <TASK> map_check_btf kernel/bpf/syscall.c:1024 [inline] map_create+0x1157/0x1860 kernel/bpf/syscall.c:1198 __sys_bpf+0x127f/0x5420 kernel/bpf/syscall.c:5040 __do_sys_bpf kernel/bpf/syscall.c:5162 [inline] __se_sys_bpf kernel/bpf/syscall.c:5160 [inline] __x64_sys_bpf+0x79/0xc0 kernel/bpf/syscall.c:5160 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd With the following btf [1] DECL_TAG 'a' type_id=4 component_idx=-1 [2] PTR '(anon)' type_id=0 [3] TYPE_TAG 'a' type_id=2 [4] VAR 'a' type_id=3, linkage=static and when the bpf_attr.btf_key_type_id = 1 (DECL_TAG), the following WARN_ON_ONCE in btf_type_id_size() is triggered: if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) && !btf_type_is_var(size_type))) return NULL; Note that 'return NULL' is the correct behavior as we don't want a DECL_TAG type to be used as a btf_{key,value}_type_id even for the case like 'DECL_TAG -> STRUCT'. So there is no correctness issue here, we just want to silence warning. To silence the warning, I added DECL_TAG as one of kinds in btf_type_nosize() which will cause btf_type_id_size() returning NULL earlier without the warning. [1] https://lore.kernel.org/bpf/000000000000e0df8d05fc75ba86@google.com/ Reported-by: syzbot+958967f249155967d42a@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230530205029.264910-1-yhs@fb.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-05-31 04:50:29 +08:00
return BTF_INFO_KIND(t->info) == BTF_KIND_DECL_TAG;
}
bpf: Silence a warning in btf_type_id_size() syzbot reported a warning in [1] with the following stacktrace: WARNING: CPU: 0 PID: 5005 at kernel/bpf/btf.c:1988 btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... RIP: 0010:btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... Call Trace: <TASK> map_check_btf kernel/bpf/syscall.c:1024 [inline] map_create+0x1157/0x1860 kernel/bpf/syscall.c:1198 __sys_bpf+0x127f/0x5420 kernel/bpf/syscall.c:5040 __do_sys_bpf kernel/bpf/syscall.c:5162 [inline] __se_sys_bpf kernel/bpf/syscall.c:5160 [inline] __x64_sys_bpf+0x79/0xc0 kernel/bpf/syscall.c:5160 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd With the following btf [1] DECL_TAG 'a' type_id=4 component_idx=-1 [2] PTR '(anon)' type_id=0 [3] TYPE_TAG 'a' type_id=2 [4] VAR 'a' type_id=3, linkage=static and when the bpf_attr.btf_key_type_id = 1 (DECL_TAG), the following WARN_ON_ONCE in btf_type_id_size() is triggered: if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) && !btf_type_is_var(size_type))) return NULL; Note that 'return NULL' is the correct behavior as we don't want a DECL_TAG type to be used as a btf_{key,value}_type_id even for the case like 'DECL_TAG -> STRUCT'. So there is no correctness issue here, we just want to silence warning. To silence the warning, I added DECL_TAG as one of kinds in btf_type_nosize() which will cause btf_type_id_size() returning NULL earlier without the warning. [1] https://lore.kernel.org/bpf/000000000000e0df8d05fc75ba86@google.com/ Reported-by: syzbot+958967f249155967d42a@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230530205029.264910-1-yhs@fb.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-05-31 04:50:29 +08:00
static bool btf_type_nosize(const struct btf_type *t)
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
{
bpf: Silence a warning in btf_type_id_size() syzbot reported a warning in [1] with the following stacktrace: WARNING: CPU: 0 PID: 5005 at kernel/bpf/btf.c:1988 btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... RIP: 0010:btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... Call Trace: <TASK> map_check_btf kernel/bpf/syscall.c:1024 [inline] map_create+0x1157/0x1860 kernel/bpf/syscall.c:1198 __sys_bpf+0x127f/0x5420 kernel/bpf/syscall.c:5040 __do_sys_bpf kernel/bpf/syscall.c:5162 [inline] __se_sys_bpf kernel/bpf/syscall.c:5160 [inline] __x64_sys_bpf+0x79/0xc0 kernel/bpf/syscall.c:5160 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd With the following btf [1] DECL_TAG 'a' type_id=4 component_idx=-1 [2] PTR '(anon)' type_id=0 [3] TYPE_TAG 'a' type_id=2 [4] VAR 'a' type_id=3, linkage=static and when the bpf_attr.btf_key_type_id = 1 (DECL_TAG), the following WARN_ON_ONCE in btf_type_id_size() is triggered: if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) && !btf_type_is_var(size_type))) return NULL; Note that 'return NULL' is the correct behavior as we don't want a DECL_TAG type to be used as a btf_{key,value}_type_id even for the case like 'DECL_TAG -> STRUCT'. So there is no correctness issue here, we just want to silence warning. To silence the warning, I added DECL_TAG as one of kinds in btf_type_nosize() which will cause btf_type_id_size() returning NULL earlier without the warning. [1] https://lore.kernel.org/bpf/000000000000e0df8d05fc75ba86@google.com/ Reported-by: syzbot+958967f249155967d42a@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230530205029.264910-1-yhs@fb.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-05-31 04:50:29 +08:00
return btf_type_is_void(t) || btf_type_is_fwd(t) ||
btf_type_is_func(t) || btf_type_is_func_proto(t) ||
btf_type_is_decl_tag(t);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
}
bpf: Silence a warning in btf_type_id_size() syzbot reported a warning in [1] with the following stacktrace: WARNING: CPU: 0 PID: 5005 at kernel/bpf/btf.c:1988 btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... RIP: 0010:btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... Call Trace: <TASK> map_check_btf kernel/bpf/syscall.c:1024 [inline] map_create+0x1157/0x1860 kernel/bpf/syscall.c:1198 __sys_bpf+0x127f/0x5420 kernel/bpf/syscall.c:5040 __do_sys_bpf kernel/bpf/syscall.c:5162 [inline] __se_sys_bpf kernel/bpf/syscall.c:5160 [inline] __x64_sys_bpf+0x79/0xc0 kernel/bpf/syscall.c:5160 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd With the following btf [1] DECL_TAG 'a' type_id=4 component_idx=-1 [2] PTR '(anon)' type_id=0 [3] TYPE_TAG 'a' type_id=2 [4] VAR 'a' type_id=3, linkage=static and when the bpf_attr.btf_key_type_id = 1 (DECL_TAG), the following WARN_ON_ONCE in btf_type_id_size() is triggered: if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) && !btf_type_is_var(size_type))) return NULL; Note that 'return NULL' is the correct behavior as we don't want a DECL_TAG type to be used as a btf_{key,value}_type_id even for the case like 'DECL_TAG -> STRUCT'. So there is no correctness issue here, we just want to silence warning. To silence the warning, I added DECL_TAG as one of kinds in btf_type_nosize() which will cause btf_type_id_size() returning NULL earlier without the warning. [1] https://lore.kernel.org/bpf/000000000000e0df8d05fc75ba86@google.com/ Reported-by: syzbot+958967f249155967d42a@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230530205029.264910-1-yhs@fb.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-05-31 04:50:29 +08:00
static bool btf_type_nosize_or_null(const struct btf_type *t)
{
bpf: Silence a warning in btf_type_id_size() syzbot reported a warning in [1] with the following stacktrace: WARNING: CPU: 0 PID: 5005 at kernel/bpf/btf.c:1988 btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... RIP: 0010:btf_type_id_size+0x2d9/0x9d0 kernel/bpf/btf.c:1988 ... Call Trace: <TASK> map_check_btf kernel/bpf/syscall.c:1024 [inline] map_create+0x1157/0x1860 kernel/bpf/syscall.c:1198 __sys_bpf+0x127f/0x5420 kernel/bpf/syscall.c:5040 __do_sys_bpf kernel/bpf/syscall.c:5162 [inline] __se_sys_bpf kernel/bpf/syscall.c:5160 [inline] __x64_sys_bpf+0x79/0xc0 kernel/bpf/syscall.c:5160 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd With the following btf [1] DECL_TAG 'a' type_id=4 component_idx=-1 [2] PTR '(anon)' type_id=0 [3] TYPE_TAG 'a' type_id=2 [4] VAR 'a' type_id=3, linkage=static and when the bpf_attr.btf_key_type_id = 1 (DECL_TAG), the following WARN_ON_ONCE in btf_type_id_size() is triggered: if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) && !btf_type_is_var(size_type))) return NULL; Note that 'return NULL' is the correct behavior as we don't want a DECL_TAG type to be used as a btf_{key,value}_type_id even for the case like 'DECL_TAG -> STRUCT'. So there is no correctness issue here, we just want to silence warning. To silence the warning, I added DECL_TAG as one of kinds in btf_type_nosize() which will cause btf_type_id_size() returning NULL earlier without the warning. [1] https://lore.kernel.org/bpf/000000000000e0df8d05fc75ba86@google.com/ Reported-by: syzbot+958967f249155967d42a@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230530205029.264910-1-yhs@fb.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-05-31 04:50:29 +08:00
return !t || btf_type_nosize(t);
}
static bool btf_type_is_decl_tag_target(const struct btf_type *t)
{
return btf_type_is_func(t) || btf_type_is_struct(t) ||
btf_type_is_var(t) || btf_type_is_typedef(t);
}
bool btf_is_vmlinux(const struct btf *btf)
{
return btf->kernel_btf && !btf->base_btf;
}
u32 btf_nr_types(const struct btf *btf)
{
u32 total = 0;
while (btf) {
total += btf->nr_types;
btf = btf->base_btf;
}
return total;
}
bpf: Introduce BPF_PROG_TYPE_STRUCT_OPS This patch allows the kernel's struct ops (i.e. func ptr) to be implemented in BPF. The first use case in this series is the "struct tcp_congestion_ops" which will be introduced in a latter patch. This patch introduces a new prog type BPF_PROG_TYPE_STRUCT_OPS. The BPF_PROG_TYPE_STRUCT_OPS prog is verified against a particular func ptr of a kernel struct. The attr->attach_btf_id is the btf id of a kernel struct. The attr->expected_attach_type is the member "index" of that kernel struct. The first member of a struct starts with member index 0. That will avoid ambiguity when a kernel struct has multiple func ptrs with the same func signature. For example, a BPF_PROG_TYPE_STRUCT_OPS prog is written to implement the "init" func ptr of the "struct tcp_congestion_ops". The attr->attach_btf_id is the btf id of the "struct tcp_congestion_ops" of the _running_ kernel. The attr->expected_attach_type is 3. The ctx of BPF_PROG_TYPE_STRUCT_OPS is an array of u64 args saved by arch_prepare_bpf_trampoline that will be done in the next patch when introducing BPF_MAP_TYPE_STRUCT_OPS. "struct bpf_struct_ops" is introduced as a common interface for the kernel struct that supports BPF_PROG_TYPE_STRUCT_OPS prog. The supporting kernel struct will need to implement an instance of the "struct bpf_struct_ops". The supporting kernel struct also needs to implement a bpf_verifier_ops. During BPF_PROG_LOAD, bpf_struct_ops_find() will find the right bpf_verifier_ops by searching the attr->attach_btf_id. A new "btf_struct_access" is also added to the bpf_verifier_ops such that the supporting kernel struct can optionally provide its own specific check on accessing the func arg (e.g. provide limited write access). After btf_vmlinux is parsed, the new bpf_struct_ops_init() is called to initialize some values (e.g. the btf id of the supporting kernel struct) and it can only be done once the btf_vmlinux is available. The R0 checks at BPF_EXIT is excluded for the BPF_PROG_TYPE_STRUCT_OPS prog if the return type of the prog->aux->attach_func_proto is "void". Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200109003503.3855825-1-kafai@fb.com
2020-01-09 08:35:03 +08:00
s32 btf_find_by_name_kind(const struct btf *btf, const char *name, u8 kind)
{
const struct btf_type *t;
const char *tname;
u32 i, total;
bpf: Introduce BPF_PROG_TYPE_STRUCT_OPS This patch allows the kernel's struct ops (i.e. func ptr) to be implemented in BPF. The first use case in this series is the "struct tcp_congestion_ops" which will be introduced in a latter patch. This patch introduces a new prog type BPF_PROG_TYPE_STRUCT_OPS. The BPF_PROG_TYPE_STRUCT_OPS prog is verified against a particular func ptr of a kernel struct. The attr->attach_btf_id is the btf id of a kernel struct. The attr->expected_attach_type is the member "index" of that kernel struct. The first member of a struct starts with member index 0. That will avoid ambiguity when a kernel struct has multiple func ptrs with the same func signature. For example, a BPF_PROG_TYPE_STRUCT_OPS prog is written to implement the "init" func ptr of the "struct tcp_congestion_ops". The attr->attach_btf_id is the btf id of the "struct tcp_congestion_ops" of the _running_ kernel. The attr->expected_attach_type is 3. The ctx of BPF_PROG_TYPE_STRUCT_OPS is an array of u64 args saved by arch_prepare_bpf_trampoline that will be done in the next patch when introducing BPF_MAP_TYPE_STRUCT_OPS. "struct bpf_struct_ops" is introduced as a common interface for the kernel struct that supports BPF_PROG_TYPE_STRUCT_OPS prog. The supporting kernel struct will need to implement an instance of the "struct bpf_struct_ops". The supporting kernel struct also needs to implement a bpf_verifier_ops. During BPF_PROG_LOAD, bpf_struct_ops_find() will find the right bpf_verifier_ops by searching the attr->attach_btf_id. A new "btf_struct_access" is also added to the bpf_verifier_ops such that the supporting kernel struct can optionally provide its own specific check on accessing the func arg (e.g. provide limited write access). After btf_vmlinux is parsed, the new bpf_struct_ops_init() is called to initialize some values (e.g. the btf id of the supporting kernel struct) and it can only be done once the btf_vmlinux is available. The R0 checks at BPF_EXIT is excluded for the BPF_PROG_TYPE_STRUCT_OPS prog if the return type of the prog->aux->attach_func_proto is "void". Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200109003503.3855825-1-kafai@fb.com
2020-01-09 08:35:03 +08:00
total = btf_nr_types(btf);
for (i = 1; i < total; i++) {
t = btf_type_by_id(btf, i);
bpf: Introduce BPF_PROG_TYPE_STRUCT_OPS This patch allows the kernel's struct ops (i.e. func ptr) to be implemented in BPF. The first use case in this series is the "struct tcp_congestion_ops" which will be introduced in a latter patch. This patch introduces a new prog type BPF_PROG_TYPE_STRUCT_OPS. The BPF_PROG_TYPE_STRUCT_OPS prog is verified against a particular func ptr of a kernel struct. The attr->attach_btf_id is the btf id of a kernel struct. The attr->expected_attach_type is the member "index" of that kernel struct. The first member of a struct starts with member index 0. That will avoid ambiguity when a kernel struct has multiple func ptrs with the same func signature. For example, a BPF_PROG_TYPE_STRUCT_OPS prog is written to implement the "init" func ptr of the "struct tcp_congestion_ops". The attr->attach_btf_id is the btf id of the "struct tcp_congestion_ops" of the _running_ kernel. The attr->expected_attach_type is 3. The ctx of BPF_PROG_TYPE_STRUCT_OPS is an array of u64 args saved by arch_prepare_bpf_trampoline that will be done in the next patch when introducing BPF_MAP_TYPE_STRUCT_OPS. "struct bpf_struct_ops" is introduced as a common interface for the kernel struct that supports BPF_PROG_TYPE_STRUCT_OPS prog. The supporting kernel struct will need to implement an instance of the "struct bpf_struct_ops". The supporting kernel struct also needs to implement a bpf_verifier_ops. During BPF_PROG_LOAD, bpf_struct_ops_find() will find the right bpf_verifier_ops by searching the attr->attach_btf_id. A new "btf_struct_access" is also added to the bpf_verifier_ops such that the supporting kernel struct can optionally provide its own specific check on accessing the func arg (e.g. provide limited write access). After btf_vmlinux is parsed, the new bpf_struct_ops_init() is called to initialize some values (e.g. the btf id of the supporting kernel struct) and it can only be done once the btf_vmlinux is available. The R0 checks at BPF_EXIT is excluded for the BPF_PROG_TYPE_STRUCT_OPS prog if the return type of the prog->aux->attach_func_proto is "void". Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200109003503.3855825-1-kafai@fb.com
2020-01-09 08:35:03 +08:00
if (BTF_INFO_KIND(t->info) != kind)
continue;
tname = btf_name_by_offset(btf, t->name_off);
if (!strcmp(tname, name))
return i;
}
return -ENOENT;
}
s32 bpf_find_btf_id(const char *name, u32 kind, struct btf **btf_p)
{
struct btf *btf;
s32 ret;
int id;
btf = bpf_get_btf_vmlinux();
if (IS_ERR(btf))
return PTR_ERR(btf);
if (!btf)
return -EINVAL;
ret = btf_find_by_name_kind(btf, name, kind);
/* ret is never zero, since btf_find_by_name_kind returns
* positive btf_id or negative error.
*/
if (ret > 0) {
btf_get(btf);
*btf_p = btf;
return ret;
}
/* If name is not found in vmlinux's BTF then search in module's BTFs */
spin_lock_bh(&btf_idr_lock);
idr_for_each_entry(&btf_idr, btf, id) {
if (!btf_is_module(btf))
continue;
/* linear search could be slow hence unlock/lock
* the IDR to avoiding holding it for too long
*/
btf_get(btf);
spin_unlock_bh(&btf_idr_lock);
ret = btf_find_by_name_kind(btf, name, kind);
if (ret > 0) {
*btf_p = btf;
return ret;
}
btf_put(btf);
spin_lock_bh(&btf_idr_lock);
}
spin_unlock_bh(&btf_idr_lock);
return ret;
}
bpf: Introduce BPF_PROG_TYPE_STRUCT_OPS This patch allows the kernel's struct ops (i.e. func ptr) to be implemented in BPF. The first use case in this series is the "struct tcp_congestion_ops" which will be introduced in a latter patch. This patch introduces a new prog type BPF_PROG_TYPE_STRUCT_OPS. The BPF_PROG_TYPE_STRUCT_OPS prog is verified against a particular func ptr of a kernel struct. The attr->attach_btf_id is the btf id of a kernel struct. The attr->expected_attach_type is the member "index" of that kernel struct. The first member of a struct starts with member index 0. That will avoid ambiguity when a kernel struct has multiple func ptrs with the same func signature. For example, a BPF_PROG_TYPE_STRUCT_OPS prog is written to implement the "init" func ptr of the "struct tcp_congestion_ops". The attr->attach_btf_id is the btf id of the "struct tcp_congestion_ops" of the _running_ kernel. The attr->expected_attach_type is 3. The ctx of BPF_PROG_TYPE_STRUCT_OPS is an array of u64 args saved by arch_prepare_bpf_trampoline that will be done in the next patch when introducing BPF_MAP_TYPE_STRUCT_OPS. "struct bpf_struct_ops" is introduced as a common interface for the kernel struct that supports BPF_PROG_TYPE_STRUCT_OPS prog. The supporting kernel struct will need to implement an instance of the "struct bpf_struct_ops". The supporting kernel struct also needs to implement a bpf_verifier_ops. During BPF_PROG_LOAD, bpf_struct_ops_find() will find the right bpf_verifier_ops by searching the attr->attach_btf_id. A new "btf_struct_access" is also added to the bpf_verifier_ops such that the supporting kernel struct can optionally provide its own specific check on accessing the func arg (e.g. provide limited write access). After btf_vmlinux is parsed, the new bpf_struct_ops_init() is called to initialize some values (e.g. the btf id of the supporting kernel struct) and it can only be done once the btf_vmlinux is available. The R0 checks at BPF_EXIT is excluded for the BPF_PROG_TYPE_STRUCT_OPS prog if the return type of the prog->aux->attach_func_proto is "void". Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200109003503.3855825-1-kafai@fb.com
2020-01-09 08:35:03 +08:00
const struct btf_type *btf_type_skip_modifiers(const struct btf *btf,
u32 id, u32 *res_id)
{
const struct btf_type *t = btf_type_by_id(btf, id);
while (btf_type_is_modifier(t)) {
id = t->type;
t = btf_type_by_id(btf, t->type);
}
if (res_id)
*res_id = id;
return t;
}
const struct btf_type *btf_type_resolve_ptr(const struct btf *btf,
u32 id, u32 *res_id)
{
const struct btf_type *t;
t = btf_type_skip_modifiers(btf, id, NULL);
if (!btf_type_is_ptr(t))
return NULL;
return btf_type_skip_modifiers(btf, t->type, res_id);
}
const struct btf_type *btf_type_resolve_func_ptr(const struct btf *btf,
u32 id, u32 *res_id)
{
const struct btf_type *ptype;
ptype = btf_type_resolve_ptr(btf, id, res_id);
if (ptype && btf_type_is_func_proto(ptype))
return ptype;
return NULL;
}
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
/* Types that act only as a source, not sink or intermediate
* type when resolving.
*/
static bool btf_type_is_resolve_source_only(const struct btf_type *t)
{
return btf_type_is_var(t) ||
btf_type_is_decl_tag(t) ||
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
btf_type_is_datasec(t);
}
/* What types need to be resolved?
*
* btf_type_is_modifier() is an obvious one.
*
* btf_type_is_struct() because its member refers to
* another type (through member->type).
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
*
* btf_type_is_var() because the variable refers to
* another type. btf_type_is_datasec() holds multiple
* btf_type_is_var() types that need resolving.
*
* btf_type_is_array() because its element (array->type)
* refers to another type. Array can be thought of a
* special case of struct while array just has the same
* member-type repeated by array->nelems of times.
*/
static bool btf_type_needs_resolve(const struct btf_type *t)
{
return btf_type_is_modifier(t) ||
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
btf_type_is_ptr(t) ||
btf_type_is_struct(t) ||
btf_type_is_array(t) ||
btf_type_is_var(t) ||
bpf: Fix a btf decl_tag bug when tagging a function syzbot reported a btf decl_tag bug with stack trace below: general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007] CPU: 0 PID: 3592 Comm: syz-executor914 Not tainted 5.16.0-syzkaller-11424-gb7892f7d5cb2 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:btf_type_vlen include/linux/btf.h:231 [inline] RIP: 0010:btf_decl_tag_resolve+0x83e/0xaa0 kernel/bpf/btf.c:3910 ... Call Trace: <TASK> btf_resolve+0x251/0x1020 kernel/bpf/btf.c:4198 btf_check_all_types kernel/bpf/btf.c:4239 [inline] btf_parse_type_sec kernel/bpf/btf.c:4280 [inline] btf_parse kernel/bpf/btf.c:4513 [inline] btf_new_fd+0x19fe/0x2370 kernel/bpf/btf.c:6047 bpf_btf_load kernel/bpf/syscall.c:4039 [inline] __sys_bpf+0x1cbb/0x5970 kernel/bpf/syscall.c:4679 __do_sys_bpf kernel/bpf/syscall.c:4738 [inline] __se_sys_bpf kernel/bpf/syscall.c:4736 [inline] __x64_sys_bpf+0x75/0xb0 kernel/bpf/syscall.c:4736 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae The kasan error is triggered with an illegal BTF like below: type 0: void type 1: int type 2: decl_tag to func type 3 type 3: func to func_proto type 8 The total number of types is 4 and the type 3 is illegal since its func_proto type is out of range. Currently, the target type of decl_tag can be struct/union, var or func. Both struct/union and var implemented their own 'resolve' callback functions and hence handled properly in kernel. But func type doesn't have 'resolve' callback function. When btf_decl_tag_resolve() tries to check func type, it tries to get vlen of its func_proto type, which triggered the above kasan error. To fix the issue, btf_decl_tag_resolve() needs to do btf_func_check() before trying to accessing func_proto type. In the current implementation, func type is checked with btf_func_check() in the main checking function btf_check_all_types(). To fix the above kasan issue, let us implement 'resolve' callback func type properly. The 'resolve' callback will be also called in btf_check_all_types() for func types. Fixes: b5ea834dde6b ("bpf: Support for new btf kind BTF_KIND_TAG") Reported-by: syzbot+53619be9444215e785ed@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20220203191727.741862-1-yhs@fb.com
2022-02-04 03:17:27 +08:00
btf_type_is_func(t) ||
btf_type_is_decl_tag(t) ||
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
btf_type_is_datasec(t);
}
/* t->size can be used */
static bool btf_type_has_size(const struct btf_type *t)
{
switch (BTF_INFO_KIND(t->info)) {
case BTF_KIND_INT:
case BTF_KIND_STRUCT:
case BTF_KIND_UNION:
case BTF_KIND_ENUM:
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
case BTF_KIND_DATASEC:
case BTF_KIND_FLOAT:
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
case BTF_KIND_ENUM64:
return true;
}
return false;
}
static const char *btf_int_encoding_str(u8 encoding)
{
if (encoding == 0)
return "(none)";
else if (encoding == BTF_INT_SIGNED)
return "SIGNED";
else if (encoding == BTF_INT_CHAR)
return "CHAR";
else if (encoding == BTF_INT_BOOL)
return "BOOL";
else
return "UNKN";
}
static u32 btf_type_int(const struct btf_type *t)
{
return *(u32 *)(t + 1);
}
static const struct btf_array *btf_type_array(const struct btf_type *t)
{
return (const struct btf_array *)(t + 1);
}
static const struct btf_enum *btf_type_enum(const struct btf_type *t)
{
return (const struct btf_enum *)(t + 1);
}
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
static const struct btf_var *btf_type_var(const struct btf_type *t)
{
return (const struct btf_var *)(t + 1);
}
static const struct btf_decl_tag *btf_type_decl_tag(const struct btf_type *t)
{
return (const struct btf_decl_tag *)(t + 1);
}
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
static const struct btf_enum64 *btf_type_enum64(const struct btf_type *t)
{
return (const struct btf_enum64 *)(t + 1);
}
static const struct btf_kind_operations *btf_type_ops(const struct btf_type *t)
{
return kind_ops[BTF_INFO_KIND(t->info)];
}
static bool btf_name_offset_valid(const struct btf *btf, u32 offset)
{
if (!BTF_STR_OFFSET_VALID(offset))
return false;
while (offset < btf->start_str_off)
btf = btf->base_btf;
offset -= btf->start_str_off;
return offset < btf->hdr.str_len;
}
bpf/btf: Accept function names that contain dots When building a kernel with LLVM=1, LLVM_IAS=0 and CONFIG_KASAN=y, LLVM leaves DWARF tags for the "asan.module_ctor" & co symbols. In turn, pahole creates BTF_KIND_FUNC entries for these and this makes the BTF metadata validation fail because they contain a dot. In a dramatic turn of event, this BTF verification failure can cause the netfilter_bpf initialization to fail, causing netfilter_core to free the netfilter_helper hashmap and netfilter_ftp to trigger a use-after-free. The risk of u-a-f in netfilter will be addressed separately but the existence of "asan.module_ctor" debug info under some build conditions sounds like a good enough reason to accept functions that contain dots in BTF. Although using only LLVM=1 is the recommended way to compile clang-based kernels, users can certainly do LLVM=1, LLVM_IAS=0 as well and we still try to support that combination according to Nick. To clarify: - > v5.10 kernel, LLVM=1 (LLVM_IAS=0 is not the default) is recommended, but user can still have LLVM=1, LLVM_IAS=0 to trigger the issue - <= 5.10 kernel, LLVM=1 (LLVM_IAS=0 is the default) is recommended in which case GNU as will be used Fixes: 1dc92851849c ("bpf: kernel side support for BTF Var and DataSec") Signed-off-by: Florent Revest <revest@chromium.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andrii@kernel.org> Cc: Yonghong Song <yhs@meta.com> Cc: Nick Desaulniers <ndesaulniers@google.com> Link: https://lore.kernel.org/bpf/20230615145607.3469985-1-revest@chromium.org
2023-06-15 22:56:07 +08:00
static bool __btf_name_char_ok(char c, bool first)
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
{
if ((first ? !isalpha(c) :
!isalnum(c)) &&
c != '_' &&
bpf/btf: Accept function names that contain dots When building a kernel with LLVM=1, LLVM_IAS=0 and CONFIG_KASAN=y, LLVM leaves DWARF tags for the "asan.module_ctor" & co symbols. In turn, pahole creates BTF_KIND_FUNC entries for these and this makes the BTF metadata validation fail because they contain a dot. In a dramatic turn of event, this BTF verification failure can cause the netfilter_bpf initialization to fail, causing netfilter_core to free the netfilter_helper hashmap and netfilter_ftp to trigger a use-after-free. The risk of u-a-f in netfilter will be addressed separately but the existence of "asan.module_ctor" debug info under some build conditions sounds like a good enough reason to accept functions that contain dots in BTF. Although using only LLVM=1 is the recommended way to compile clang-based kernels, users can certainly do LLVM=1, LLVM_IAS=0 as well and we still try to support that combination according to Nick. To clarify: - > v5.10 kernel, LLVM=1 (LLVM_IAS=0 is not the default) is recommended, but user can still have LLVM=1, LLVM_IAS=0 to trigger the issue - <= 5.10 kernel, LLVM=1 (LLVM_IAS=0 is the default) is recommended in which case GNU as will be used Fixes: 1dc92851849c ("bpf: kernel side support for BTF Var and DataSec") Signed-off-by: Florent Revest <revest@chromium.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andrii@kernel.org> Cc: Yonghong Song <yhs@meta.com> Cc: Nick Desaulniers <ndesaulniers@google.com> Link: https://lore.kernel.org/bpf/20230615145607.3469985-1-revest@chromium.org
2023-06-15 22:56:07 +08:00
c != '.')
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
return false;
return true;
}
const char *btf_str_by_offset(const struct btf *btf, u32 offset)
{
while (offset < btf->start_str_off)
btf = btf->base_btf;
offset -= btf->start_str_off;
if (offset < btf->hdr.str_len)
return &btf->strings[offset];
return NULL;
}
static bool btf_name_valid_identifier(const struct btf *btf, u32 offset)
{
/* offset must be valid */
const char *src = btf_str_by_offset(btf, offset);
const char *src_limit;
bpf/btf: Accept function names that contain dots When building a kernel with LLVM=1, LLVM_IAS=0 and CONFIG_KASAN=y, LLVM leaves DWARF tags for the "asan.module_ctor" & co symbols. In turn, pahole creates BTF_KIND_FUNC entries for these and this makes the BTF metadata validation fail because they contain a dot. In a dramatic turn of event, this BTF verification failure can cause the netfilter_bpf initialization to fail, causing netfilter_core to free the netfilter_helper hashmap and netfilter_ftp to trigger a use-after-free. The risk of u-a-f in netfilter will be addressed separately but the existence of "asan.module_ctor" debug info under some build conditions sounds like a good enough reason to accept functions that contain dots in BTF. Although using only LLVM=1 is the recommended way to compile clang-based kernels, users can certainly do LLVM=1, LLVM_IAS=0 as well and we still try to support that combination according to Nick. To clarify: - > v5.10 kernel, LLVM=1 (LLVM_IAS=0 is not the default) is recommended, but user can still have LLVM=1, LLVM_IAS=0 to trigger the issue - <= 5.10 kernel, LLVM=1 (LLVM_IAS=0 is the default) is recommended in which case GNU as will be used Fixes: 1dc92851849c ("bpf: kernel side support for BTF Var and DataSec") Signed-off-by: Florent Revest <revest@chromium.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andrii@kernel.org> Cc: Yonghong Song <yhs@meta.com> Cc: Nick Desaulniers <ndesaulniers@google.com> Link: https://lore.kernel.org/bpf/20230615145607.3469985-1-revest@chromium.org
2023-06-15 22:56:07 +08:00
if (!__btf_name_char_ok(*src, true))
return false;
/* set a limit on identifier length */
src_limit = src + KSYM_NAME_LEN;
src++;
while (*src && src < src_limit) {
bpf/btf: Accept function names that contain dots When building a kernel with LLVM=1, LLVM_IAS=0 and CONFIG_KASAN=y, LLVM leaves DWARF tags for the "asan.module_ctor" & co symbols. In turn, pahole creates BTF_KIND_FUNC entries for these and this makes the BTF metadata validation fail because they contain a dot. In a dramatic turn of event, this BTF verification failure can cause the netfilter_bpf initialization to fail, causing netfilter_core to free the netfilter_helper hashmap and netfilter_ftp to trigger a use-after-free. The risk of u-a-f in netfilter will be addressed separately but the existence of "asan.module_ctor" debug info under some build conditions sounds like a good enough reason to accept functions that contain dots in BTF. Although using only LLVM=1 is the recommended way to compile clang-based kernels, users can certainly do LLVM=1, LLVM_IAS=0 as well and we still try to support that combination according to Nick. To clarify: - > v5.10 kernel, LLVM=1 (LLVM_IAS=0 is not the default) is recommended, but user can still have LLVM=1, LLVM_IAS=0 to trigger the issue - <= 5.10 kernel, LLVM=1 (LLVM_IAS=0 is the default) is recommended in which case GNU as will be used Fixes: 1dc92851849c ("bpf: kernel side support for BTF Var and DataSec") Signed-off-by: Florent Revest <revest@chromium.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andrii@kernel.org> Cc: Yonghong Song <yhs@meta.com> Cc: Nick Desaulniers <ndesaulniers@google.com> Link: https://lore.kernel.org/bpf/20230615145607.3469985-1-revest@chromium.org
2023-06-15 22:56:07 +08:00
if (!__btf_name_char_ok(*src, false))
return false;
src++;
}
return !*src;
}
/* Allow any printable character in DATASEC names */
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
static bool btf_name_valid_section(const struct btf *btf, u32 offset)
{
/* offset must be valid */
const char *src = btf_str_by_offset(btf, offset);
const char *src_limit;
if (!*src)
return false;
/* set a limit on identifier length */
src_limit = src + KSYM_NAME_LEN;
while (*src && src < src_limit) {
if (!isprint(*src))
return false;
src++;
}
return !*src;
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
}
2018-12-14 02:41:46 +08:00
static const char *__btf_name_by_offset(const struct btf *btf, u32 offset)
{
const char *name;
if (!offset)
return "(anon)";
name = btf_str_by_offset(btf, offset);
return name ?: "(invalid-name-offset)";
}
2018-12-14 02:41:46 +08:00
const char *btf_name_by_offset(const struct btf *btf, u32 offset)
{
return btf_str_by_offset(btf, offset);
2018-12-14 02:41:46 +08:00
}
bpf: Introduce bpf_func_info This patch added interface to load a program with the following additional information: . prog_btf_fd . func_info, func_info_rec_size and func_info_cnt where func_info will provide function range and type_id corresponding to each function. The func_info_rec_size is introduced in the UAPI to specify struct bpf_func_info size passed from user space. This intends to make bpf_func_info structure growable in the future. If the kernel gets a different bpf_func_info size from userspace, it will try to handle user request with part of bpf_func_info it can understand. In this patch, kernel can understand struct bpf_func_info { __u32 insn_offset; __u32 type_id; }; If user passed a bpf func_info record size of 16 bytes, the kernel can still handle part of records with the above definition. If verifier agrees with function range provided by the user, the bpf_prog ksym for each function will use the func name provided in the type_id, which is supposed to provide better encoding as it is not limited by 16 bytes program name limitation and this is better for bpf program which contains multiple subprograms. The bpf_prog_info interface is also extended to return btf_id, func_info, func_info_rec_size and func_info_cnt to userspace, so userspace can print out the function prototype for each xlated function. The insn_offset in the returned func_info corresponds to the insn offset for xlated functions. With other jit related fields in bpf_prog_info, userspace can also print out function prototypes for each jited function. Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2018-11-20 07:29:11 +08:00
const struct btf_type *btf_type_by_id(const struct btf *btf, u32 type_id)
{
while (type_id < btf->start_id)
btf = btf->base_btf;
type_id -= btf->start_id;
if (type_id >= btf->nr_types)
return NULL;
return btf->types[type_id];
}
EXPORT_SYMBOL_GPL(btf_type_by_id);
/*
* Regular int is not a bit field and it must be either
* u8/u16/u32/u64 or __int128.
*/
static bool btf_type_int_is_regular(const struct btf_type *t)
{
u8 nr_bits, nr_bytes;
u32 int_data;
int_data = btf_type_int(t);
nr_bits = BTF_INT_BITS(int_data);
nr_bytes = BITS_ROUNDUP_BYTES(nr_bits);
if (BITS_PER_BYTE_MASKED(nr_bits) ||
BTF_INT_OFFSET(int_data) ||
(nr_bytes != sizeof(u8) && nr_bytes != sizeof(u16) &&
nr_bytes != sizeof(u32) && nr_bytes != sizeof(u64) &&
nr_bytes != (2 * sizeof(u64)))) {
return false;
}
return true;
}
/*
* Check that given struct member is a regular int with expected
* offset and size.
*/
bool btf_member_is_reg_int(const struct btf *btf, const struct btf_type *s,
const struct btf_member *m,
u32 expected_offset, u32 expected_size)
{
const struct btf_type *t;
u32 id, int_data;
u8 nr_bits;
id = m->type;
t = btf_type_id_size(btf, &id, NULL);
if (!t || !btf_type_is_int(t))
return false;
int_data = btf_type_int(t);
nr_bits = BTF_INT_BITS(int_data);
if (btf_type_kflag(s)) {
u32 bitfield_size = BTF_MEMBER_BITFIELD_SIZE(m->offset);
u32 bit_offset = BTF_MEMBER_BIT_OFFSET(m->offset);
/* if kflag set, int should be a regular int and
* bit offset should be at byte boundary.
*/
return !bitfield_size &&
BITS_ROUNDUP_BYTES(bit_offset) == expected_offset &&
BITS_ROUNDUP_BYTES(nr_bits) == expected_size;
}
if (BTF_INT_OFFSET(int_data) ||
BITS_PER_BYTE_MASKED(m->offset) ||
BITS_ROUNDUP_BYTES(m->offset) != expected_offset ||
BITS_PER_BYTE_MASKED(nr_bits) ||
BITS_ROUNDUP_BYTES(nr_bits) != expected_size)
return false;
return true;
}
/* Similar to btf_type_skip_modifiers() but does not skip typedefs. */
static const struct btf_type *btf_type_skip_qualifiers(const struct btf *btf,
u32 id)
{
const struct btf_type *t = btf_type_by_id(btf, id);
while (btf_type_is_modifier(t) &&
BTF_INFO_KIND(t->info) != BTF_KIND_TYPEDEF) {
t = btf_type_by_id(btf, t->type);
}
return t;
}
#define BTF_SHOW_MAX_ITER 10
#define BTF_KIND_BIT(kind) (1ULL << kind)
/*
* Populate show->state.name with type name information.
* Format of type name is
*
* [.member_name = ] (type_name)
*/
static const char *btf_show_name(struct btf_show *show)
{
/* BTF_MAX_ITER array suffixes "[]" */
const char *array_suffixes = "[][][][][][][][][][]";
const char *array_suffix = &array_suffixes[strlen(array_suffixes)];
/* BTF_MAX_ITER pointer suffixes "*" */
const char *ptr_suffixes = "**********";
const char *ptr_suffix = &ptr_suffixes[strlen(ptr_suffixes)];
const char *name = NULL, *prefix = "", *parens = "";
const struct btf_member *m = show->state.member;
const struct btf_type *t;
const struct btf_array *array;
u32 id = show->state.type_id;
const char *member = NULL;
bool show_member = false;
u64 kinds = 0;
int i;
show->state.name[0] = '\0';
/*
* Don't show type name if we're showing an array member;
* in that case we show the array type so don't need to repeat
* ourselves for each member.
*/
if (show->state.array_member)
return "";
/* Retrieve member name, if any. */
if (m) {
member = btf_name_by_offset(show->btf, m->name_off);
show_member = strlen(member) > 0;
id = m->type;
}
/*
* Start with type_id, as we have resolved the struct btf_type *
* via btf_modifier_show() past the parent typedef to the child
* struct, int etc it is defined as. In such cases, the type_id
* still represents the starting type while the struct btf_type *
* in our show->state points at the resolved type of the typedef.
*/
t = btf_type_by_id(show->btf, id);
if (!t)
return "";
/*
* The goal here is to build up the right number of pointer and
* array suffixes while ensuring the type name for a typedef
* is represented. Along the way we accumulate a list of
* BTF kinds we have encountered, since these will inform later
* display; for example, pointer types will not require an
* opening "{" for struct, we will just display the pointer value.
*
* We also want to accumulate the right number of pointer or array
* indices in the format string while iterating until we get to
* the typedef/pointee/array member target type.
*
* We start by pointing at the end of pointer and array suffix
* strings; as we accumulate pointers and arrays we move the pointer
* or array string backwards so it will show the expected number of
* '*' or '[]' for the type. BTF_SHOW_MAX_ITER of nesting of pointers
* and/or arrays and typedefs are supported as a precaution.
*
* We also want to get typedef name while proceeding to resolve
* type it points to so that we can add parentheses if it is a
* "typedef struct" etc.
*/
for (i = 0; i < BTF_SHOW_MAX_ITER; i++) {
switch (BTF_INFO_KIND(t->info)) {
case BTF_KIND_TYPEDEF:
if (!name)
name = btf_name_by_offset(show->btf,
t->name_off);
kinds |= BTF_KIND_BIT(BTF_KIND_TYPEDEF);
id = t->type;
break;
case BTF_KIND_ARRAY:
kinds |= BTF_KIND_BIT(BTF_KIND_ARRAY);
parens = "[";
if (!t)
return "";
array = btf_type_array(t);
if (array_suffix > array_suffixes)
array_suffix -= 2;
id = array->type;
break;
case BTF_KIND_PTR:
kinds |= BTF_KIND_BIT(BTF_KIND_PTR);
if (ptr_suffix > ptr_suffixes)
ptr_suffix -= 1;
id = t->type;
break;
default:
id = 0;
break;
}
if (!id)
break;
t = btf_type_skip_qualifiers(show->btf, id);
}
/* We may not be able to represent this type; bail to be safe */
if (i == BTF_SHOW_MAX_ITER)
return "";
if (!name)
name = btf_name_by_offset(show->btf, t->name_off);
switch (BTF_INFO_KIND(t->info)) {
case BTF_KIND_STRUCT:
case BTF_KIND_UNION:
prefix = BTF_INFO_KIND(t->info) == BTF_KIND_STRUCT ?
"struct" : "union";
/* if it's an array of struct/union, parens is already set */
if (!(kinds & (BTF_KIND_BIT(BTF_KIND_ARRAY))))
parens = "{";
break;
case BTF_KIND_ENUM:
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
case BTF_KIND_ENUM64:
prefix = "enum";
break;
default:
break;
}
/* pointer does not require parens */
if (kinds & BTF_KIND_BIT(BTF_KIND_PTR))
parens = "";
/* typedef does not require struct/union/enum prefix */
if (kinds & BTF_KIND_BIT(BTF_KIND_TYPEDEF))
prefix = "";
if (!name)
name = "";
/* Even if we don't want type name info, we want parentheses etc */
if (show->flags & BTF_SHOW_NONAME)
snprintf(show->state.name, sizeof(show->state.name), "%s",
parens);
else
snprintf(show->state.name, sizeof(show->state.name),
"%s%s%s(%s%s%s%s%s%s)%s",
/* first 3 strings comprise ".member = " */
show_member ? "." : "",
show_member ? member : "",
show_member ? " = " : "",
/* ...next is our prefix (struct, enum, etc) */
prefix,
strlen(prefix) > 0 && strlen(name) > 0 ? " " : "",
/* ...this is the type name itself */
name,
/* ...suffixed by the appropriate '*', '[]' suffixes */
strlen(ptr_suffix) > 0 ? " " : "", ptr_suffix,
array_suffix, parens);
return show->state.name;
}
static const char *__btf_show_indent(struct btf_show *show)
{
const char *indents = " ";
const char *indent = &indents[strlen(indents)];
if ((indent - show->state.depth) >= indents)
return indent - show->state.depth;
return indents;
}
static const char *btf_show_indent(struct btf_show *show)
{
return show->flags & BTF_SHOW_COMPACT ? "" : __btf_show_indent(show);
}
static const char *btf_show_newline(struct btf_show *show)
{
return show->flags & BTF_SHOW_COMPACT ? "" : "\n";
}
static const char *btf_show_delim(struct btf_show *show)
{
if (show->state.depth == 0)
return "";
if ((show->flags & BTF_SHOW_COMPACT) && show->state.type &&
BTF_INFO_KIND(show->state.type->info) == BTF_KIND_UNION)
return "|";
return ",";
}
__printf(2, 3) static void btf_show(struct btf_show *show, const char *fmt, ...)
{
va_list args;
if (!show->state.depth_check) {
va_start(args, fmt);
show->showfn(show, fmt, args);
va_end(args);
}
}
/* Macros are used here as btf_show_type_value[s]() prepends and appends
* format specifiers to the format specifier passed in; these do the work of
* adding indentation, delimiters etc while the caller simply has to specify
* the type value(s) in the format specifier + value(s).
*/
#define btf_show_type_value(show, fmt, value) \
do { \
if ((value) != (__typeof__(value))0 || \
(show->flags & BTF_SHOW_ZERO) || \
show->state.depth == 0) { \
btf_show(show, "%s%s" fmt "%s%s", \
btf_show_indent(show), \
btf_show_name(show), \
value, btf_show_delim(show), \
btf_show_newline(show)); \
if (show->state.depth > show->state.depth_to_show) \
show->state.depth_to_show = show->state.depth; \
} \
} while (0)
#define btf_show_type_values(show, fmt, ...) \
do { \
btf_show(show, "%s%s" fmt "%s%s", btf_show_indent(show), \
btf_show_name(show), \
__VA_ARGS__, btf_show_delim(show), \
btf_show_newline(show)); \
if (show->state.depth > show->state.depth_to_show) \
show->state.depth_to_show = show->state.depth; \
} while (0)
/* How much is left to copy to safe buffer after @data? */
static int btf_show_obj_size_left(struct btf_show *show, void *data)
{
return show->obj.head + show->obj.size - data;
}
/* Is object pointed to by @data of @size already copied to our safe buffer? */
static bool btf_show_obj_is_safe(struct btf_show *show, void *data, int size)
{
return data >= show->obj.data &&
(data + size) < (show->obj.data + BTF_SHOW_OBJ_SAFE_SIZE);
}
/*
* If object pointed to by @data of @size falls within our safe buffer, return
* the equivalent pointer to the same safe data. Assumes
* copy_from_kernel_nofault() has already happened and our safe buffer is
* populated.
*/
static void *__btf_show_obj_safe(struct btf_show *show, void *data, int size)
{
if (btf_show_obj_is_safe(show, data, size))
return show->obj.safe + (data - show->obj.data);
return NULL;
}
/*
* Return a safe-to-access version of data pointed to by @data.
* We do this by copying the relevant amount of information
* to the struct btf_show obj.safe buffer using copy_from_kernel_nofault().
*
* If BTF_SHOW_UNSAFE is specified, just return data as-is; no
* safe copy is needed.
*
* Otherwise we need to determine if we have the required amount
* of data (determined by the @data pointer and the size of the
* largest base type we can encounter (represented by
* BTF_SHOW_OBJ_BASE_TYPE_SIZE). Having that much data ensures
* that we will be able to print some of the current object,
* and if more is needed a copy will be triggered.
* Some objects such as structs will not fit into the buffer;
* in such cases additional copies when we iterate over their
* members may be needed.
*
* btf_show_obj_safe() is used to return a safe buffer for
* btf_show_start_type(); this ensures that as we recurse into
* nested types we always have safe data for the given type.
* This approach is somewhat wasteful; it's possible for example
* that when iterating over a large union we'll end up copying the
* same data repeatedly, but the goal is safety not performance.
* We use stack data as opposed to per-CPU buffers because the
* iteration over a type can take some time, and preemption handling
* would greatly complicate use of the safe buffer.
*/
static void *btf_show_obj_safe(struct btf_show *show,
const struct btf_type *t,
void *data)
{
const struct btf_type *rt;
int size_left, size;
void *safe = NULL;
if (show->flags & BTF_SHOW_UNSAFE)
return data;
rt = btf_resolve_size(show->btf, t, &size);
if (IS_ERR(rt)) {
show->state.status = PTR_ERR(rt);
return NULL;
}
/*
* Is this toplevel object? If so, set total object size and
* initialize pointers. Otherwise check if we still fall within
* our safe object data.
*/
if (show->state.depth == 0) {
show->obj.size = size;
show->obj.head = data;
} else {
/*
* If the size of the current object is > our remaining
* safe buffer we _may_ need to do a new copy. However
* consider the case of a nested struct; it's size pushes
* us over the safe buffer limit, but showing any individual
* struct members does not. In such cases, we don't need
* to initiate a fresh copy yet; however we definitely need
* at least BTF_SHOW_OBJ_BASE_TYPE_SIZE bytes left
* in our buffer, regardless of the current object size.
* The logic here is that as we resolve types we will
* hit a base type at some point, and we need to be sure
* the next chunk of data is safely available to display
* that type info safely. We cannot rely on the size of
* the current object here because it may be much larger
* than our current buffer (e.g. task_struct is 8k).
* All we want to do here is ensure that we can print the
* next basic type, which we can if either
* - the current type size is within the safe buffer; or
* - at least BTF_SHOW_OBJ_BASE_TYPE_SIZE bytes are left in
* the safe buffer.
*/
safe = __btf_show_obj_safe(show, data,
min(size,
BTF_SHOW_OBJ_BASE_TYPE_SIZE));
}
/*
* We need a new copy to our safe object, either because we haven't
* yet copied and are initializing safe data, or because the data
* we want falls outside the boundaries of the safe object.
*/
if (!safe) {
size_left = btf_show_obj_size_left(show, data);
if (size_left > BTF_SHOW_OBJ_SAFE_SIZE)
size_left = BTF_SHOW_OBJ_SAFE_SIZE;
show->state.status = copy_from_kernel_nofault(show->obj.safe,
data, size_left);
if (!show->state.status) {
show->obj.data = data;
safe = show->obj.safe;
}
}
return safe;
}
/*
* Set the type we are starting to show and return a safe data pointer
* to be used for showing the associated data.
*/
static void *btf_show_start_type(struct btf_show *show,
const struct btf_type *t,
u32 type_id, void *data)
{
show->state.type = t;
show->state.type_id = type_id;
show->state.name[0] = '\0';
return btf_show_obj_safe(show, t, data);
}
static void btf_show_end_type(struct btf_show *show)
{
show->state.type = NULL;
show->state.type_id = 0;
show->state.name[0] = '\0';
}
static void *btf_show_start_aggr_type(struct btf_show *show,
const struct btf_type *t,
u32 type_id, void *data)
{
void *safe_data = btf_show_start_type(show, t, type_id, data);
if (!safe_data)
return safe_data;
btf_show(show, "%s%s%s", btf_show_indent(show),
btf_show_name(show),
btf_show_newline(show));
show->state.depth++;
return safe_data;
}
static void btf_show_end_aggr_type(struct btf_show *show,
const char *suffix)
{
show->state.depth--;
btf_show(show, "%s%s%s%s", btf_show_indent(show), suffix,
btf_show_delim(show), btf_show_newline(show));
btf_show_end_type(show);
}
static void btf_show_start_member(struct btf_show *show,
const struct btf_member *m)
{
show->state.member = m;
}
static void btf_show_start_array_member(struct btf_show *show)
{
show->state.array_member = 1;
btf_show_start_member(show, NULL);
}
static void btf_show_end_member(struct btf_show *show)
{
show->state.member = NULL;
}
static void btf_show_end_array_member(struct btf_show *show)
{
show->state.array_member = 0;
btf_show_end_member(show);
}
static void *btf_show_start_array_type(struct btf_show *show,
const struct btf_type *t,
u32 type_id,
u16 array_encoding,
void *data)
{
show->state.array_encoding = array_encoding;
show->state.array_terminated = 0;
return btf_show_start_aggr_type(show, t, type_id, data);
}
static void btf_show_end_array_type(struct btf_show *show)
{
show->state.array_encoding = 0;
show->state.array_terminated = 0;
btf_show_end_aggr_type(show, "]");
}
static void *btf_show_start_struct_type(struct btf_show *show,
const struct btf_type *t,
u32 type_id,
void *data)
{
return btf_show_start_aggr_type(show, t, type_id, data);
}
static void btf_show_end_struct_type(struct btf_show *show)
{
btf_show_end_aggr_type(show, "}");
}
__printf(2, 3) static void __btf_verifier_log(struct bpf_verifier_log *log,
const char *fmt, ...)
{
va_list args;
va_start(args, fmt);
bpf_verifier_vlog(log, fmt, args);
va_end(args);
}
__printf(2, 3) static void btf_verifier_log(struct btf_verifier_env *env,
const char *fmt, ...)
{
struct bpf_verifier_log *log = &env->log;
va_list args;
if (!bpf_verifier_log_needed(log))
return;
va_start(args, fmt);
bpf_verifier_vlog(log, fmt, args);
va_end(args);
}
__printf(4, 5) static void __btf_verifier_log_type(struct btf_verifier_env *env,
const struct btf_type *t,
bool log_details,
const char *fmt, ...)
{
struct bpf_verifier_log *log = &env->log;
struct btf *btf = env->btf;
va_list args;
if (!bpf_verifier_log_needed(log))
return;
if (log->level == BPF_LOG_KERNEL) {
/* btf verifier prints all types it is processing via
* btf_verifier_log_type(..., fmt = NULL).
* Skip those prints for in-kernel BTF verification.
*/
if (!fmt)
return;
/* Skip logging when loading module BTF with mismatches permitted */
if (env->btf->base_btf && IS_ENABLED(CONFIG_MODULE_ALLOW_BTF_MISMATCH))
return;
}
__btf_verifier_log(log, "[%u] %s %s%s",
env->log_type_id,
btf_type_str(t),
2018-12-14 02:41:46 +08:00
__btf_name_by_offset(btf, t->name_off),
log_details ? " " : "");
if (log_details)
btf_type_ops(t)->log_details(env, t);
if (fmt && *fmt) {
__btf_verifier_log(log, " ");
va_start(args, fmt);
bpf_verifier_vlog(log, fmt, args);
va_end(args);
}
__btf_verifier_log(log, "\n");
}
#define btf_verifier_log_type(env, t, ...) \
__btf_verifier_log_type((env), (t), true, __VA_ARGS__)
#define btf_verifier_log_basic(env, t, ...) \
__btf_verifier_log_type((env), (t), false, __VA_ARGS__)
__printf(4, 5)
static void btf_verifier_log_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const char *fmt, ...)
{
struct bpf_verifier_log *log = &env->log;
struct btf *btf = env->btf;
va_list args;
if (!bpf_verifier_log_needed(log))
return;
if (log->level == BPF_LOG_KERNEL) {
if (!fmt)
return;
/* Skip logging when loading module BTF with mismatches permitted */
if (env->btf->base_btf && IS_ENABLED(CONFIG_MODULE_ALLOW_BTF_MISMATCH))
return;
}
/* The CHECK_META phase already did a btf dump.
*
* If member is logged again, it must hit an error in
* parsing this member. It is useful to print out which
* struct this member belongs to.
*/
if (env->phase != CHECK_META)
btf_verifier_log_type(env, struct_type, NULL);
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (btf_type_kflag(struct_type))
__btf_verifier_log(log,
"\t%s type_id=%u bitfield_size=%u bits_offset=%u",
__btf_name_by_offset(btf, member->name_off),
member->type,
BTF_MEMBER_BITFIELD_SIZE(member->offset),
BTF_MEMBER_BIT_OFFSET(member->offset));
else
__btf_verifier_log(log, "\t%s type_id=%u bits_offset=%u",
__btf_name_by_offset(btf, member->name_off),
member->type, member->offset);
if (fmt && *fmt) {
__btf_verifier_log(log, " ");
va_start(args, fmt);
bpf_verifier_vlog(log, fmt, args);
va_end(args);
}
__btf_verifier_log(log, "\n");
}
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
__printf(4, 5)
static void btf_verifier_log_vsi(struct btf_verifier_env *env,
const struct btf_type *datasec_type,
const struct btf_var_secinfo *vsi,
const char *fmt, ...)
{
struct bpf_verifier_log *log = &env->log;
va_list args;
if (!bpf_verifier_log_needed(log))
return;
if (log->level == BPF_LOG_KERNEL && !fmt)
return;
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
if (env->phase != CHECK_META)
btf_verifier_log_type(env, datasec_type, NULL);
__btf_verifier_log(log, "\t type_id=%u offset=%u size=%u",
vsi->type, vsi->offset, vsi->size);
if (fmt && *fmt) {
__btf_verifier_log(log, " ");
va_start(args, fmt);
bpf_verifier_vlog(log, fmt, args);
va_end(args);
}
__btf_verifier_log(log, "\n");
}
static void btf_verifier_log_hdr(struct btf_verifier_env *env,
u32 btf_data_size)
{
struct bpf_verifier_log *log = &env->log;
const struct btf *btf = env->btf;
const struct btf_header *hdr;
if (!bpf_verifier_log_needed(log))
return;
if (log->level == BPF_LOG_KERNEL)
return;
hdr = &btf->hdr;
__btf_verifier_log(log, "magic: 0x%x\n", hdr->magic);
__btf_verifier_log(log, "version: %u\n", hdr->version);
__btf_verifier_log(log, "flags: 0x%x\n", hdr->flags);
__btf_verifier_log(log, "hdr_len: %u\n", hdr->hdr_len);
__btf_verifier_log(log, "type_off: %u\n", hdr->type_off);
__btf_verifier_log(log, "type_len: %u\n", hdr->type_len);
__btf_verifier_log(log, "str_off: %u\n", hdr->str_off);
__btf_verifier_log(log, "str_len: %u\n", hdr->str_len);
__btf_verifier_log(log, "btf_total_size: %u\n", btf_data_size);
}
static int btf_add_type(struct btf_verifier_env *env, struct btf_type *t)
{
struct btf *btf = env->btf;
if (btf->types_size == btf->nr_types) {
/* Expand 'types' array */
struct btf_type **new_types;
u32 expand_by, new_size;
if (btf->start_id + btf->types_size == BTF_MAX_TYPE) {
btf_verifier_log(env, "Exceeded max num of types");
return -E2BIG;
}
expand_by = max_t(u32, btf->types_size >> 2, 16);
new_size = min_t(u32, BTF_MAX_TYPE,
btf->types_size + expand_by);
treewide: kvzalloc() -> kvcalloc() The kvzalloc() function has a 2-factor argument form, kvcalloc(). This patch replaces cases of: kvzalloc(a * b, gfp) with: kvcalloc(a * b, gfp) as well as handling cases of: kvzalloc(a * b * c, gfp) with: kvzalloc(array3_size(a, b, c), gfp) as it's slightly less ugly than: kvcalloc(array_size(a, b), c, gfp) This does, however, attempt to ignore constant size factors like: kvzalloc(4 * 1024, gfp) though any constants defined via macros get caught up in the conversion. Any factors with a sizeof() of "unsigned char", "char", and "u8" were dropped, since they're redundant. The Coccinelle script used for this was: // Fix redundant parens around sizeof(). @@ type TYPE; expression THING, E; @@ ( kvzalloc( - (sizeof(TYPE)) * E + sizeof(TYPE) * E , ...) | kvzalloc( - (sizeof(THING)) * E + sizeof(THING) * E , ...) ) // Drop single-byte sizes and redundant parens. @@ expression COUNT; typedef u8; typedef __u8; @@ ( kvzalloc( - sizeof(u8) * (COUNT) + COUNT , ...) | kvzalloc( - sizeof(__u8) * (COUNT) + COUNT , ...) | kvzalloc( - sizeof(char) * (COUNT) + COUNT , ...) | kvzalloc( - sizeof(unsigned char) * (COUNT) + COUNT , ...) | kvzalloc( - sizeof(u8) * COUNT + COUNT , ...) | kvzalloc( - sizeof(__u8) * COUNT + COUNT , ...) | kvzalloc( - sizeof(char) * COUNT + COUNT , ...) | kvzalloc( - sizeof(unsigned char) * COUNT + COUNT , ...) ) // 2-factor product with sizeof(type/expression) and identifier or constant. @@ type TYPE; expression THING; identifier COUNT_ID; constant COUNT_CONST; @@ ( - kvzalloc + kvcalloc ( - sizeof(TYPE) * (COUNT_ID) + COUNT_ID, sizeof(TYPE) , ...) | - kvzalloc + kvcalloc ( - sizeof(TYPE) * COUNT_ID + COUNT_ID, sizeof(TYPE) , ...) | - kvzalloc + kvcalloc ( - sizeof(TYPE) * (COUNT_CONST) + COUNT_CONST, sizeof(TYPE) , ...) | - kvzalloc + kvcalloc ( - sizeof(TYPE) * COUNT_CONST + COUNT_CONST, sizeof(TYPE) , ...) | - kvzalloc + kvcalloc ( - sizeof(THING) * (COUNT_ID) + COUNT_ID, sizeof(THING) , ...) | - kvzalloc + kvcalloc ( - sizeof(THING) * COUNT_ID + COUNT_ID, sizeof(THING) , ...) | - kvzalloc + kvcalloc ( - sizeof(THING) * (COUNT_CONST) + COUNT_CONST, sizeof(THING) , ...) | - kvzalloc + kvcalloc ( - sizeof(THING) * COUNT_CONST + COUNT_CONST, sizeof(THING) , ...) ) // 2-factor product, only identifiers. @@ identifier SIZE, COUNT; @@ - kvzalloc + kvcalloc ( - SIZE * COUNT + COUNT, SIZE , ...) // 3-factor product with 1 sizeof(type) or sizeof(expression), with // redundant parens removed. @@ expression THING; identifier STRIDE, COUNT; type TYPE; @@ ( kvzalloc( - sizeof(TYPE) * (COUNT) * (STRIDE) + array3_size(COUNT, STRIDE, sizeof(TYPE)) , ...) | kvzalloc( - sizeof(TYPE) * (COUNT) * STRIDE + array3_size(COUNT, STRIDE, sizeof(TYPE)) , ...) | kvzalloc( - sizeof(TYPE) * COUNT * (STRIDE) + array3_size(COUNT, STRIDE, sizeof(TYPE)) , ...) | kvzalloc( - sizeof(TYPE) * COUNT * STRIDE + array3_size(COUNT, STRIDE, sizeof(TYPE)) , ...) | kvzalloc( - sizeof(THING) * (COUNT) * (STRIDE) + array3_size(COUNT, STRIDE, sizeof(THING)) , ...) | kvzalloc( - sizeof(THING) * (COUNT) * STRIDE + array3_size(COUNT, STRIDE, sizeof(THING)) , ...) | kvzalloc( - sizeof(THING) * COUNT * (STRIDE) + array3_size(COUNT, STRIDE, sizeof(THING)) , ...) | kvzalloc( - sizeof(THING) * COUNT * STRIDE + array3_size(COUNT, STRIDE, sizeof(THING)) , ...) ) // 3-factor product with 2 sizeof(variable), with redundant parens removed. @@ expression THING1, THING2; identifier COUNT; type TYPE1, TYPE2; @@ ( kvzalloc( - sizeof(TYPE1) * sizeof(TYPE2) * COUNT + array3_size(COUNT, sizeof(TYPE1), sizeof(TYPE2)) , ...) | kvzalloc( - sizeof(TYPE1) * sizeof(THING2) * (COUNT) + array3_size(COUNT, sizeof(TYPE1), sizeof(TYPE2)) , ...) | kvzalloc( - sizeof(THING1) * sizeof(THING2) * COUNT + array3_size(COUNT, sizeof(THING1), sizeof(THING2)) , ...) | kvzalloc( - sizeof(THING1) * sizeof(THING2) * (COUNT) + array3_size(COUNT, sizeof(THING1), sizeof(THING2)) , ...) | kvzalloc( - sizeof(TYPE1) * sizeof(THING2) * COUNT + array3_size(COUNT, sizeof(TYPE1), sizeof(THING2)) , ...) | kvzalloc( - sizeof(TYPE1) * sizeof(THING2) * (COUNT) + array3_size(COUNT, sizeof(TYPE1), sizeof(THING2)) , ...) ) // 3-factor product, only identifiers, with redundant parens removed. @@ identifier STRIDE, SIZE, COUNT; @@ ( kvzalloc( - (COUNT) * STRIDE * SIZE + array3_size(COUNT, STRIDE, SIZE) , ...) | kvzalloc( - COUNT * (STRIDE) * SIZE + array3_size(COUNT, STRIDE, SIZE) , ...) | kvzalloc( - COUNT * STRIDE * (SIZE) + array3_size(COUNT, STRIDE, SIZE) , ...) | kvzalloc( - (COUNT) * (STRIDE) * SIZE + array3_size(COUNT, STRIDE, SIZE) , ...) | kvzalloc( - COUNT * (STRIDE) * (SIZE) + array3_size(COUNT, STRIDE, SIZE) , ...) | kvzalloc( - (COUNT) * STRIDE * (SIZE) + array3_size(COUNT, STRIDE, SIZE) , ...) | kvzalloc( - (COUNT) * (STRIDE) * (SIZE) + array3_size(COUNT, STRIDE, SIZE) , ...) | kvzalloc( - COUNT * STRIDE * SIZE + array3_size(COUNT, STRIDE, SIZE) , ...) ) // Any remaining multi-factor products, first at least 3-factor products, // when they're not all constants... @@ expression E1, E2, E3; constant C1, C2, C3; @@ ( kvzalloc(C1 * C2 * C3, ...) | kvzalloc( - (E1) * E2 * E3 + array3_size(E1, E2, E3) , ...) | kvzalloc( - (E1) * (E2) * E3 + array3_size(E1, E2, E3) , ...) | kvzalloc( - (E1) * (E2) * (E3) + array3_size(E1, E2, E3) , ...) | kvzalloc( - E1 * E2 * E3 + array3_size(E1, E2, E3) , ...) ) // And then all remaining 2 factors products when they're not all constants, // keeping sizeof() as the second factor argument. @@ expression THING, E1, E2; type TYPE; constant C1, C2, C3; @@ ( kvzalloc(sizeof(THING) * C2, ...) | kvzalloc(sizeof(TYPE) * C2, ...) | kvzalloc(C1 * C2 * C3, ...) | kvzalloc(C1 * C2, ...) | - kvzalloc + kvcalloc ( - sizeof(TYPE) * (E2) + E2, sizeof(TYPE) , ...) | - kvzalloc + kvcalloc ( - sizeof(TYPE) * E2 + E2, sizeof(TYPE) , ...) | - kvzalloc + kvcalloc ( - sizeof(THING) * (E2) + E2, sizeof(THING) , ...) | - kvzalloc + kvcalloc ( - sizeof(THING) * E2 + E2, sizeof(THING) , ...) | - kvzalloc + kvcalloc ( - (E1) * E2 + E1, E2 , ...) | - kvzalloc + kvcalloc ( - (E1) * (E2) + E1, E2 , ...) | - kvzalloc + kvcalloc ( - E1 * E2 + E1, E2 , ...) ) Signed-off-by: Kees Cook <keescook@chromium.org>
2018-06-13 05:04:48 +08:00
new_types = kvcalloc(new_size, sizeof(*new_types),
GFP_KERNEL | __GFP_NOWARN);
if (!new_types)
return -ENOMEM;
if (btf->nr_types == 0) {
if (!btf->base_btf) {
/* lazily init VOID type */
new_types[0] = &btf_void;
btf->nr_types++;
}
} else {
memcpy(new_types, btf->types,
sizeof(*btf->types) * btf->nr_types);
}
kvfree(btf->types);
btf->types = new_types;
btf->types_size = new_size;
}
btf->types[btf->nr_types++] = t;
return 0;
}
static int btf_alloc_id(struct btf *btf)
{
int id;
idr_preload(GFP_KERNEL);
spin_lock_bh(&btf_idr_lock);
id = idr_alloc_cyclic(&btf_idr, btf, 1, INT_MAX, GFP_ATOMIC);
if (id > 0)
btf->id = id;
spin_unlock_bh(&btf_idr_lock);
idr_preload_end();
if (WARN_ON_ONCE(!id))
return -ENOSPC;
return id > 0 ? 0 : id;
}
static void btf_free_id(struct btf *btf)
{
unsigned long flags;
/*
* In map-in-map, calling map_delete_elem() on outer
* map will call bpf_map_put on the inner map.
* It will then eventually call btf_free_id()
* on the inner map. Some of the map_delete_elem()
* implementation may have irq disabled, so
* we need to use the _irqsave() version instead
* of the _bh() version.
*/
spin_lock_irqsave(&btf_idr_lock, flags);
idr_remove(&btf_idr, btf->id);
spin_unlock_irqrestore(&btf_idr_lock, flags);
}
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
static void btf_free_kfunc_set_tab(struct btf *btf)
{
struct btf_kfunc_set_tab *tab = btf->kfunc_set_tab;
int hook;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
if (!tab)
return;
for (hook = 0; hook < ARRAY_SIZE(tab->sets); hook++)
kfree(tab->sets[hook]);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
kfree(tab);
btf->kfunc_set_tab = NULL;
}
static void btf_free_dtor_kfunc_tab(struct btf *btf)
{
struct btf_id_dtor_kfunc_tab *tab = btf->dtor_kfunc_tab;
if (!tab)
return;
kfree(tab);
btf->dtor_kfunc_tab = NULL;
}
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
static void btf_struct_metas_free(struct btf_struct_metas *tab)
{
int i;
if (!tab)
return;
bpf: Remove btf_field_offs, use btf_record's fields instead The btf_field_offs struct contains (offset, size) for btf_record fields, sorted by offset. btf_field_offs is always used in conjunction with btf_record, which has btf_field 'fields' array with (offset, type), the latter of which btf_field_offs' size is derived from via btf_field_type_size. This patch adds a size field to struct btf_field and sorts btf_record's fields by offset, making it possible to get rid of btf_field_offs. Less data duplication and less code complexity results. Since btf_field_offs' lifetime closely followed the btf_record used to populate it, most complexity wins are from removal of initialization code like: if (btf_record_successfully_initialized) { foffs = btf_parse_field_offs(rec); if (IS_ERR_OR_NULL(foffs)) // free the btf_record and return err } Other changes in this patch are pretty mechanical: * foffs->field_off[i] -> rec->fields[i].offset * foffs->field_sz[i] -> rec->fields[i].size * Sort rec->fields in btf_parse_fields before returning * It's possible that this is necessary independently of other changes in this patch. btf_record_find in syscall.c expects btf_record's fields to be sorted by offset, yet there's no explicit sorting of them before this patch, record's fields are populated in the order they're read from BTF struct definition. BTF docs don't say anything about the sortedness of struct fields. * All functions taking struct btf_field_offs * input now instead take struct btf_record *. All callsites of these functions already have access to the correct btf_record. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230415201811.343116-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-16 04:18:03 +08:00
for (i = 0; i < tab->cnt; i++)
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
btf_record_free(tab->types[i].record);
kfree(tab);
}
static void btf_free_struct_meta_tab(struct btf *btf)
{
struct btf_struct_metas *tab = btf->struct_meta_tab;
btf_struct_metas_free(tab);
btf->struct_meta_tab = NULL;
}
static void btf_free_struct_ops_tab(struct btf *btf)
{
struct btf_struct_ops_tab *tab = btf->struct_ops_tab;
bpf: Create argument information for nullable arguments. Collect argument information from the type information of stub functions to mark arguments of BPF struct_ops programs with PTR_MAYBE_NULL if they are nullable. A nullable argument is annotated by suffixing "__nullable" at the argument name of stub function. For nullable arguments, this patch sets a struct bpf_ctx_arg_aux to label their reg_type with PTR_TO_BTF_ID | PTR_TRUSTED | PTR_MAYBE_NULL. This makes the verifier to check programs and ensure that they properly check the pointer. The programs should check if the pointer is null before accessing the pointed memory. The implementer of a struct_ops type should annotate the arguments that can be null. The implementer should define a stub function (empty) as a placeholder for each defined operator. The name of a stub function should be in the pattern "<st_op_type>__<operator name>". For example, for test_maybe_null of struct bpf_testmod_ops, it's stub function name should be "bpf_testmod_ops__test_maybe_null". You mark an argument nullable by suffixing the argument name with "__nullable" at the stub function. Since we already has stub functions for kCFI, we just reuse these stub functions with the naming convention mentioned earlier. These stub functions with the naming convention is only required if there are nullable arguments to annotate. For functions having not nullable arguments, stub functions are not necessary for the purpose of this patch. This patch will prepare a list of struct bpf_ctx_arg_aux, aka arg_info, for each member field of a struct_ops type. "arg_info" will be assigned to "prog->aux->ctx_arg_info" of BPF struct_ops programs in check_struct_ops_btf_id() so that it can be used by btf_ctx_access() later to set reg_type properly for the verifier. Signed-off-by: Kui-Feng Lee <thinker.li@gmail.com> Link: https://lore.kernel.org/r/20240209023750.1153905-4-thinker.li@gmail.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2024-02-09 10:37:49 +08:00
u32 i;
if (!tab)
return;
for (i = 0; i < tab->cnt; i++)
bpf_struct_ops_desc_release(&tab->ops[i]);
kfree(tab);
btf->struct_ops_tab = NULL;
}
static void btf_free(struct btf *btf)
{
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
btf_free_struct_meta_tab(btf);
btf_free_dtor_kfunc_tab(btf);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
btf_free_kfunc_set_tab(btf);
btf_free_struct_ops_tab(btf);
kvfree(btf->types);
kvfree(btf->resolved_sizes);
kvfree(btf->resolved_ids);
/* vmlinux does not allocate btf->data, it simply points it at
* __start_BTF.
*/
if (!btf_is_vmlinux(btf))
kvfree(btf->data);
kvfree(btf->base_id_map);
kfree(btf);
}
static void btf_free_rcu(struct rcu_head *rcu)
{
struct btf *btf = container_of(rcu, struct btf, rcu);
btf_free(btf);
}
const char *btf_get_name(const struct btf *btf)
{
return btf->name;
}
void btf_get(struct btf *btf)
{
refcount_inc(&btf->refcnt);
}
void btf_put(struct btf *btf)
{
if (btf && refcount_dec_and_test(&btf->refcnt)) {
btf_free_id(btf);
call_rcu(&btf->rcu, btf_free_rcu);
}
}
struct btf *btf_base_btf(const struct btf *btf)
{
return btf->base_btf;
}
const struct btf_header *btf_header(const struct btf *btf)
{
return &btf->hdr;
}
void btf_set_base_btf(struct btf *btf, const struct btf *base_btf)
{
btf->base_btf = (struct btf *)base_btf;
btf->start_id = btf_nr_types(base_btf);
btf->start_str_off = base_btf->hdr.str_len;
}
static int env_resolve_init(struct btf_verifier_env *env)
{
struct btf *btf = env->btf;
u32 nr_types = btf->nr_types;
u32 *resolved_sizes = NULL;
u32 *resolved_ids = NULL;
u8 *visit_states = NULL;
resolved_sizes = kvcalloc(nr_types, sizeof(*resolved_sizes),
GFP_KERNEL | __GFP_NOWARN);
if (!resolved_sizes)
goto nomem;
resolved_ids = kvcalloc(nr_types, sizeof(*resolved_ids),
GFP_KERNEL | __GFP_NOWARN);
if (!resolved_ids)
goto nomem;
visit_states = kvcalloc(nr_types, sizeof(*visit_states),
GFP_KERNEL | __GFP_NOWARN);
if (!visit_states)
goto nomem;
btf->resolved_sizes = resolved_sizes;
btf->resolved_ids = resolved_ids;
env->visit_states = visit_states;
return 0;
nomem:
kvfree(resolved_sizes);
kvfree(resolved_ids);
kvfree(visit_states);
return -ENOMEM;
}
static void btf_verifier_env_free(struct btf_verifier_env *env)
{
kvfree(env->visit_states);
kfree(env);
}
static bool env_type_is_resolve_sink(const struct btf_verifier_env *env,
const struct btf_type *next_type)
{
switch (env->resolve_mode) {
case RESOLVE_TBD:
/* int, enum or void is a sink */
return !btf_type_needs_resolve(next_type);
case RESOLVE_PTR:
/* int, enum, void, struct, array, func or func_proto is a sink
* for ptr
*/
return !btf_type_is_modifier(next_type) &&
!btf_type_is_ptr(next_type);
case RESOLVE_STRUCT_OR_ARRAY:
/* int, enum, void, ptr, func or func_proto is a sink
* for struct and array
*/
return !btf_type_is_modifier(next_type) &&
!btf_type_is_array(next_type) &&
!btf_type_is_struct(next_type);
default:
BUG();
}
}
static bool env_type_is_resolved(const struct btf_verifier_env *env,
u32 type_id)
{
/* base BTF types should be resolved by now */
if (type_id < env->btf->start_id)
return true;
return env->visit_states[type_id - env->btf->start_id] == RESOLVED;
}
static int env_stack_push(struct btf_verifier_env *env,
const struct btf_type *t, u32 type_id)
{
const struct btf *btf = env->btf;
struct resolve_vertex *v;
if (env->top_stack == MAX_RESOLVE_DEPTH)
return -E2BIG;
if (type_id < btf->start_id
|| env->visit_states[type_id - btf->start_id] != NOT_VISITED)
return -EEXIST;
env->visit_states[type_id - btf->start_id] = VISITED;
v = &env->stack[env->top_stack++];
v->t = t;
v->type_id = type_id;
v->next_member = 0;
if (env->resolve_mode == RESOLVE_TBD) {
if (btf_type_is_ptr(t))
env->resolve_mode = RESOLVE_PTR;
else if (btf_type_is_struct(t) || btf_type_is_array(t))
env->resolve_mode = RESOLVE_STRUCT_OR_ARRAY;
}
return 0;
}
static void env_stack_set_next_member(struct btf_verifier_env *env,
u16 next_member)
{
env->stack[env->top_stack - 1].next_member = next_member;
}
static void env_stack_pop_resolved(struct btf_verifier_env *env,
u32 resolved_type_id,
u32 resolved_size)
{
u32 type_id = env->stack[--(env->top_stack)].type_id;
struct btf *btf = env->btf;
type_id -= btf->start_id; /* adjust to local type id */
btf->resolved_sizes[type_id] = resolved_size;
btf->resolved_ids[type_id] = resolved_type_id;
env->visit_states[type_id] = RESOLVED;
}
static const struct resolve_vertex *env_stack_peak(struct btf_verifier_env *env)
{
return env->top_stack ? &env->stack[env->top_stack - 1] : NULL;
}
/* Resolve the size of a passed-in "type"
*
* type: is an array (e.g. u32 array[x][y])
* return type: type "u32[x][y]", i.e. BTF_KIND_ARRAY,
* *type_size: (x * y * sizeof(u32)). Hence, *type_size always
* corresponds to the return type.
* *elem_type: u32
* *elem_id: id of u32
* *total_nelems: (x * y). Hence, individual elem size is
* (*type_size / *total_nelems)
* *type_id: id of type if it's changed within the function, 0 if not
*
* type: is not an array (e.g. const struct X)
* return type: type "struct X"
* *type_size: sizeof(struct X)
* *elem_type: same as return type ("struct X")
* *elem_id: 0
* *total_nelems: 1
* *type_id: id of type if it's changed within the function, 0 if not
*/
static const struct btf_type *
__btf_resolve_size(const struct btf *btf, const struct btf_type *type,
u32 *type_size, const struct btf_type **elem_type,
u32 *elem_id, u32 *total_nelems, u32 *type_id)
{
const struct btf_type *array_type = NULL;
const struct btf_array *array = NULL;
u32 i, size, nelems = 1, id = 0;
for (i = 0; i < MAX_RESOLVE_DEPTH; i++) {
switch (BTF_INFO_KIND(type->info)) {
/* type->size can be used */
case BTF_KIND_INT:
case BTF_KIND_STRUCT:
case BTF_KIND_UNION:
case BTF_KIND_ENUM:
case BTF_KIND_FLOAT:
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
case BTF_KIND_ENUM64:
size = type->size;
goto resolved;
case BTF_KIND_PTR:
size = sizeof(void *);
goto resolved;
/* Modifiers */
case BTF_KIND_TYPEDEF:
case BTF_KIND_VOLATILE:
case BTF_KIND_CONST:
case BTF_KIND_RESTRICT:
case BTF_KIND_TYPE_TAG:
id = type->type;
type = btf_type_by_id(btf, type->type);
break;
case BTF_KIND_ARRAY:
if (!array_type)
array_type = type;
array = btf_type_array(type);
if (nelems && array->nelems > U32_MAX / nelems)
return ERR_PTR(-EINVAL);
nelems *= array->nelems;
type = btf_type_by_id(btf, array->type);
break;
/* type without size */
default:
return ERR_PTR(-EINVAL);
}
}
return ERR_PTR(-EINVAL);
resolved:
if (nelems && size > U32_MAX / nelems)
return ERR_PTR(-EINVAL);
*type_size = nelems * size;
bpf: Introduce BPF_MAP_TYPE_STRUCT_OPS The patch introduces BPF_MAP_TYPE_STRUCT_OPS. The map value is a kernel struct with its func ptr implemented in bpf prog. This new map is the interface to register/unregister/introspect a bpf implemented kernel struct. The kernel struct is actually embedded inside another new struct (or called the "value" struct in the code). For example, "struct tcp_congestion_ops" is embbeded in: struct bpf_struct_ops_tcp_congestion_ops { refcount_t refcnt; enum bpf_struct_ops_state state; struct tcp_congestion_ops data; /* <-- kernel subsystem struct here */ } The map value is "struct bpf_struct_ops_tcp_congestion_ops". The "bpftool map dump" will then be able to show the state ("inuse"/"tobefree") and the number of subsystem's refcnt (e.g. number of tcp_sock in the tcp_congestion_ops case). This "value" struct is created automatically by a macro. Having a separate "value" struct will also make extending "struct bpf_struct_ops_XYZ" easier (e.g. adding "void (*init)(void)" to "struct bpf_struct_ops_XYZ" to do some initialization works before registering the struct_ops to the kernel subsystem). The libbpf will take care of finding and populating the "struct bpf_struct_ops_XYZ" from "struct XYZ". Register a struct_ops to a kernel subsystem: 1. Load all needed BPF_PROG_TYPE_STRUCT_OPS prog(s) 2. Create a BPF_MAP_TYPE_STRUCT_OPS with attr->btf_vmlinux_value_type_id set to the btf id "struct bpf_struct_ops_tcp_congestion_ops" of the running kernel. Instead of reusing the attr->btf_value_type_id, btf_vmlinux_value_type_id s added such that attr->btf_fd can still be used as the "user" btf which could store other useful sysadmin/debug info that may be introduced in the furture, e.g. creation-date/compiler-details/map-creator...etc. 3. Create a "struct bpf_struct_ops_tcp_congestion_ops" object as described in the running kernel btf. Populate the value of this object. The function ptr should be populated with the prog fds. 4. Call BPF_MAP_UPDATE with the object created in (3) as the map value. The key is always "0". During BPF_MAP_UPDATE, the code that saves the kernel-func-ptr's args as an array of u64 is generated. BPF_MAP_UPDATE also allows the specific struct_ops to do some final checks in "st_ops->init_member()" (e.g. ensure all mandatory func ptrs are implemented). If everything looks good, it will register this kernel struct to the kernel subsystem. The map will not allow further update from this point. Unregister a struct_ops from the kernel subsystem: BPF_MAP_DELETE with key "0". Introspect a struct_ops: BPF_MAP_LOOKUP_ELEM with key "0". The map value returned will have the prog _id_ populated as the func ptr. The map value state (enum bpf_struct_ops_state) will transit from: INIT (map created) => INUSE (map updated, i.e. reg) => TOBEFREE (map value deleted, i.e. unreg) The kernel subsystem needs to call bpf_struct_ops_get() and bpf_struct_ops_put() to manage the "refcnt" in the "struct bpf_struct_ops_XYZ". This patch uses a separate refcnt for the purose of tracking the subsystem usage. Another approach is to reuse the map->refcnt and then "show" (i.e. during map_lookup) the subsystem's usage by doing map->refcnt - map->usercnt to filter out the map-fd/pinned-map usage. However, that will also tie down the future semantics of map->refcnt and map->usercnt. The very first subsystem's refcnt (during reg()) holds one count to map->refcnt. When the very last subsystem's refcnt is gone, it will also release the map->refcnt. All bpf_prog will be freed when the map->refcnt reaches 0 (i.e. during map_free()). Here is how the bpftool map command will look like: [root@arch-fb-vm1 bpf]# bpftool map show 6: struct_ops name dctcp flags 0x0 key 4B value 256B max_entries 1 memlock 4096B btf_id 6 [root@arch-fb-vm1 bpf]# bpftool map dump id 6 [{ "value": { "refcnt": { "refs": { "counter": 1 } }, "state": 1, "data": { "list": { "next": 0, "prev": 0 }, "key": 0, "flags": 2, "init": 24, "release": 0, "ssthresh": 25, "cong_avoid": 30, "set_state": 27, "cwnd_event": 28, "in_ack_event": 26, "undo_cwnd": 29, "pkts_acked": 0, "min_tso_segs": 0, "sndbuf_expand": 0, "cong_control": 0, "get_info": 0, "name": [98,112,102,95,100,99,116,99,112,0,0,0,0,0,0,0 ], "owner": 0 } } } ] Misc Notes: * bpf_struct_ops_map_sys_lookup_elem() is added for syscall lookup. It does an inplace update on "*value" instead returning a pointer to syscall.c. Otherwise, it needs a separate copy of "zero" value for the BPF_STRUCT_OPS_STATE_INIT to avoid races. * The bpf_struct_ops_map_delete_elem() is also called without preempt_disable() from map_delete_elem(). It is because the "->unreg()" may requires sleepable context, e.g. the "tcp_unregister_congestion_control()". * "const" is added to some of the existing "struct btf_func_model *" function arg to avoid a compiler warning caused by this patch. Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200109003505.3855919-1-kafai@fb.com
2020-01-09 08:35:05 +08:00
if (total_nelems)
*total_nelems = nelems;
if (elem_type)
*elem_type = type;
if (elem_id)
*elem_id = array ? array->type : 0;
if (type_id && id)
*type_id = id;
return array_type ? : type;
}
const struct btf_type *
btf_resolve_size(const struct btf *btf, const struct btf_type *type,
u32 *type_size)
{
return __btf_resolve_size(btf, type, type_size, NULL, NULL, NULL, NULL);
}
static u32 btf_resolved_type_id(const struct btf *btf, u32 type_id)
{
while (type_id < btf->start_id)
btf = btf->base_btf;
return btf->resolved_ids[type_id - btf->start_id];
}
/* The input param "type_id" must point to a needs_resolve type */
static const struct btf_type *btf_type_id_resolve(const struct btf *btf,
u32 *type_id)
{
*type_id = btf_resolved_type_id(btf, *type_id);
return btf_type_by_id(btf, *type_id);
}
static u32 btf_resolved_type_size(const struct btf *btf, u32 type_id)
{
while (type_id < btf->start_id)
btf = btf->base_btf;
return btf->resolved_sizes[type_id - btf->start_id];
}
const struct btf_type *btf_type_id_size(const struct btf *btf,
u32 *type_id, u32 *ret_size)
{
const struct btf_type *size_type;
u32 size_type_id = *type_id;
u32 size = 0;
size_type = btf_type_by_id(btf, size_type_id);
if (btf_type_nosize_or_null(size_type))
return NULL;
if (btf_type_has_size(size_type)) {
size = size_type->size;
} else if (btf_type_is_array(size_type)) {
size = btf_resolved_type_size(btf, size_type_id);
} else if (btf_type_is_ptr(size_type)) {
size = sizeof(void *);
} else {
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
if (WARN_ON_ONCE(!btf_type_is_modifier(size_type) &&
!btf_type_is_var(size_type)))
return NULL;
size_type_id = btf_resolved_type_id(btf, size_type_id);
size_type = btf_type_by_id(btf, size_type_id);
if (btf_type_nosize_or_null(size_type))
return NULL;
bpf: fix BTF verifier size resolution logic BTF verifier has a size resolution bug which in some circumstances leads to invalid size resolution for, e.g., TYPEDEF modifier. This happens if we have [1] PTR -> [2] TYPEDEF -> [3] ARRAY, in which case due to being in pointer context ARRAY size won't be resolved (because for pointer it doesn't matter, so it's a sink in pointer context), but it will be permanently remembered as zero for TYPEDEF and TYPEDEF will be marked as RESOLVED. Eventually ARRAY size will be resolved correctly, but TYPEDEF resolved_size won't be updated anymore. This, subsequently, will lead to erroneous map creation failure, if that TYPEDEF is specified as either key or value, as key_size/value_size won't correspond to resolved size of TYPEDEF (kernel will believe it's zero). Note, that if BTF was ordered as [1] ARRAY <- [2] TYPEDEF <- [3] PTR, this won't be a problem, as by the time we get to TYPEDEF, ARRAY's size is already calculated and stored. This bug manifests itself in rejecting BTF-defined maps that use array typedef as a value type: typedef int array_t[16]; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(value, array_t); /* i.e., array_t *value; */ } test_map SEC(".maps"); The fix consists on not relying on modifier's resolved_size and instead using modifier's resolved_id (type ID for "concrete" type to which modifier eventually resolves) and doing size determination for that resolved type. This allow to preserve existing "early DFS termination" logic for PTR or STRUCT_OR_ARRAY contexts, but still do correct size determination for modifier types. Fixes: eb3f595dab40 ("bpf: btf: Validate type reference") Cc: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-07-13 01:25:55 +08:00
else if (btf_type_has_size(size_type))
size = size_type->size;
else if (btf_type_is_array(size_type))
size = btf_resolved_type_size(btf, size_type_id);
bpf: fix BTF verifier size resolution logic BTF verifier has a size resolution bug which in some circumstances leads to invalid size resolution for, e.g., TYPEDEF modifier. This happens if we have [1] PTR -> [2] TYPEDEF -> [3] ARRAY, in which case due to being in pointer context ARRAY size won't be resolved (because for pointer it doesn't matter, so it's a sink in pointer context), but it will be permanently remembered as zero for TYPEDEF and TYPEDEF will be marked as RESOLVED. Eventually ARRAY size will be resolved correctly, but TYPEDEF resolved_size won't be updated anymore. This, subsequently, will lead to erroneous map creation failure, if that TYPEDEF is specified as either key or value, as key_size/value_size won't correspond to resolved size of TYPEDEF (kernel will believe it's zero). Note, that if BTF was ordered as [1] ARRAY <- [2] TYPEDEF <- [3] PTR, this won't be a problem, as by the time we get to TYPEDEF, ARRAY's size is already calculated and stored. This bug manifests itself in rejecting BTF-defined maps that use array typedef as a value type: typedef int array_t[16]; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(value, array_t); /* i.e., array_t *value; */ } test_map SEC(".maps"); The fix consists on not relying on modifier's resolved_size and instead using modifier's resolved_id (type ID for "concrete" type to which modifier eventually resolves) and doing size determination for that resolved type. This allow to preserve existing "early DFS termination" logic for PTR or STRUCT_OR_ARRAY contexts, but still do correct size determination for modifier types. Fixes: eb3f595dab40 ("bpf: btf: Validate type reference") Cc: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-07-13 01:25:55 +08:00
else if (btf_type_is_ptr(size_type))
size = sizeof(void *);
else
return NULL;
}
*type_id = size_type_id;
if (ret_size)
*ret_size = size;
return size_type;
}
static int btf_df_check_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
btf_verifier_log_basic(env, struct_type,
"Unsupported check_member");
return -EINVAL;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
static int btf_df_check_kflag_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
btf_verifier_log_basic(env, struct_type,
"Unsupported check_kflag_member");
return -EINVAL;
}
/* Used for ptr, array struct/union and float type members.
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
* int, enum and modifier types have their specific callback functions.
*/
static int btf_generic_check_kflag_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
if (BTF_MEMBER_BITFIELD_SIZE(member->offset)) {
btf_verifier_log_member(env, struct_type, member,
"Invalid member bitfield_size");
return -EINVAL;
}
/* bitfield size is 0, so member->offset represents bit offset only.
* It is safe to call non kflag check_member variants.
*/
return btf_type_ops(member_type)->check_member(env, struct_type,
member,
member_type);
}
static int btf_df_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
btf_verifier_log_basic(env, v->t, "Unsupported resolve");
return -EINVAL;
}
static void btf_df_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offsets,
struct btf_show *show)
{
btf_show(show, "<unsupported kind:%u>", BTF_INFO_KIND(t->info));
}
static int btf_int_check_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
u32 int_data = btf_type_int(member_type);
u32 struct_bits_off = member->offset;
u32 struct_size = struct_type->size;
u32 nr_copy_bits;
u32 bytes_offset;
if (U32_MAX - struct_bits_off < BTF_INT_OFFSET(int_data)) {
btf_verifier_log_member(env, struct_type, member,
"bits_offset exceeds U32_MAX");
return -EINVAL;
}
struct_bits_off += BTF_INT_OFFSET(int_data);
bytes_offset = BITS_ROUNDDOWN_BYTES(struct_bits_off);
nr_copy_bits = BTF_INT_BITS(int_data) +
BITS_PER_BYTE_MASKED(struct_bits_off);
if (nr_copy_bits > BITS_PER_U128) {
btf_verifier_log_member(env, struct_type, member,
"nr_copy_bits exceeds 128");
return -EINVAL;
}
if (struct_size < bytes_offset ||
struct_size - bytes_offset < BITS_ROUNDUP_BYTES(nr_copy_bits)) {
btf_verifier_log_member(env, struct_type, member,
"Member exceeds struct_size");
return -EINVAL;
}
return 0;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
static int btf_int_check_kflag_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
u32 struct_bits_off, nr_bits, nr_int_data_bits, bytes_offset;
u32 int_data = btf_type_int(member_type);
u32 struct_size = struct_type->size;
u32 nr_copy_bits;
/* a regular int type is required for the kflag int member */
if (!btf_type_int_is_regular(member_type)) {
btf_verifier_log_member(env, struct_type, member,
"Invalid member base type");
return -EINVAL;
}
/* check sanity of bitfield size */
nr_bits = BTF_MEMBER_BITFIELD_SIZE(member->offset);
struct_bits_off = BTF_MEMBER_BIT_OFFSET(member->offset);
nr_int_data_bits = BTF_INT_BITS(int_data);
if (!nr_bits) {
/* Not a bitfield member, member offset must be at byte
* boundary.
*/
if (BITS_PER_BYTE_MASKED(struct_bits_off)) {
btf_verifier_log_member(env, struct_type, member,
"Invalid member offset");
return -EINVAL;
}
nr_bits = nr_int_data_bits;
} else if (nr_bits > nr_int_data_bits) {
btf_verifier_log_member(env, struct_type, member,
"Invalid member bitfield_size");
return -EINVAL;
}
bytes_offset = BITS_ROUNDDOWN_BYTES(struct_bits_off);
nr_copy_bits = nr_bits + BITS_PER_BYTE_MASKED(struct_bits_off);
if (nr_copy_bits > BITS_PER_U128) {
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
btf_verifier_log_member(env, struct_type, member,
"nr_copy_bits exceeds 128");
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
return -EINVAL;
}
if (struct_size < bytes_offset ||
struct_size - bytes_offset < BITS_ROUNDUP_BYTES(nr_copy_bits)) {
btf_verifier_log_member(env, struct_type, member,
"Member exceeds struct_size");
return -EINVAL;
}
return 0;
}
static s32 btf_int_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
u32 int_data, nr_bits, meta_needed = sizeof(int_data);
u16 encoding;
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
if (btf_type_vlen(t)) {
btf_verifier_log_type(env, t, "vlen != 0");
return -EINVAL;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
int_data = btf_type_int(t);
if (int_data & ~BTF_INT_MASK) {
btf_verifier_log_basic(env, t, "Invalid int_data:%x",
int_data);
return -EINVAL;
}
nr_bits = BTF_INT_BITS(int_data) + BTF_INT_OFFSET(int_data);
if (nr_bits > BITS_PER_U128) {
btf_verifier_log_type(env, t, "nr_bits exceeds %zu",
BITS_PER_U128);
return -EINVAL;
}
if (BITS_ROUNDUP_BYTES(nr_bits) > t->size) {
btf_verifier_log_type(env, t, "nr_bits exceeds type_size");
return -EINVAL;
}
/*
* Only one of the encoding bits is allowed and it
* should be sufficient for the pretty print purpose (i.e. decoding).
* Multiple bits can be allowed later if it is found
* to be insufficient.
*/
encoding = BTF_INT_ENCODING(int_data);
if (encoding &&
encoding != BTF_INT_SIGNED &&
encoding != BTF_INT_CHAR &&
encoding != BTF_INT_BOOL) {
btf_verifier_log_type(env, t, "Unsupported encoding");
return -ENOTSUPP;
}
btf_verifier_log_type(env, t, NULL);
return meta_needed;
}
static void btf_int_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
int int_data = btf_type_int(t);
btf_verifier_log(env,
"size=%u bits_offset=%u nr_bits=%u encoding=%s",
t->size, BTF_INT_OFFSET(int_data),
BTF_INT_BITS(int_data),
btf_int_encoding_str(BTF_INT_ENCODING(int_data)));
}
static void btf_int128_print(struct btf_show *show, void *data)
{
/* data points to a __int128 number.
* Suppose
* int128_num = *(__int128 *)data;
* The below formulas shows what upper_num and lower_num represents:
* upper_num = int128_num >> 64;
* lower_num = int128_num & 0xffffffffFFFFFFFFULL;
*/
u64 upper_num, lower_num;
#ifdef __BIG_ENDIAN_BITFIELD
upper_num = *(u64 *)data;
lower_num = *(u64 *)(data + 8);
#else
upper_num = *(u64 *)(data + 8);
lower_num = *(u64 *)data;
#endif
if (upper_num == 0)
btf_show_type_value(show, "0x%llx", lower_num);
else
btf_show_type_values(show, "0x%llx%016llx", upper_num,
lower_num);
}
static void btf_int128_shift(u64 *print_num, u16 left_shift_bits,
u16 right_shift_bits)
{
u64 upper_num, lower_num;
#ifdef __BIG_ENDIAN_BITFIELD
upper_num = print_num[0];
lower_num = print_num[1];
#else
upper_num = print_num[1];
lower_num = print_num[0];
#endif
/* shake out un-needed bits by shift/or operations */
if (left_shift_bits >= 64) {
upper_num = lower_num << (left_shift_bits - 64);
lower_num = 0;
} else {
upper_num = (upper_num << left_shift_bits) |
(lower_num >> (64 - left_shift_bits));
lower_num = lower_num << left_shift_bits;
}
if (right_shift_bits >= 64) {
lower_num = upper_num >> (right_shift_bits - 64);
upper_num = 0;
} else {
lower_num = (lower_num >> right_shift_bits) |
(upper_num << (64 - right_shift_bits));
upper_num = upper_num >> right_shift_bits;
}
#ifdef __BIG_ENDIAN_BITFIELD
print_num[0] = upper_num;
print_num[1] = lower_num;
#else
print_num[0] = lower_num;
print_num[1] = upper_num;
#endif
}
static void btf_bitfield_show(void *data, u8 bits_offset,
u8 nr_bits, struct btf_show *show)
{
u16 left_shift_bits, right_shift_bits;
u8 nr_copy_bytes;
u8 nr_copy_bits;
u64 print_num[2] = {};
nr_copy_bits = nr_bits + bits_offset;
nr_copy_bytes = BITS_ROUNDUP_BYTES(nr_copy_bits);
memcpy(print_num, data, nr_copy_bytes);
#ifdef __BIG_ENDIAN_BITFIELD
left_shift_bits = bits_offset;
#else
left_shift_bits = BITS_PER_U128 - nr_copy_bits;
#endif
right_shift_bits = BITS_PER_U128 - nr_bits;
btf_int128_shift(print_num, left_shift_bits, right_shift_bits);
btf_int128_print(show, print_num);
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
static void btf_int_bits_show(const struct btf *btf,
const struct btf_type *t,
void *data, u8 bits_offset,
struct btf_show *show)
{
u32 int_data = btf_type_int(t);
u8 nr_bits = BTF_INT_BITS(int_data);
u8 total_bits_offset;
/*
* bits_offset is at most 7.
* BTF_INT_OFFSET() cannot exceed 128 bits.
*/
total_bits_offset = bits_offset + BTF_INT_OFFSET(int_data);
data += BITS_ROUNDDOWN_BYTES(total_bits_offset);
bits_offset = BITS_PER_BYTE_MASKED(total_bits_offset);
btf_bitfield_show(data, bits_offset, nr_bits, show);
}
static void btf_int_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
{
u32 int_data = btf_type_int(t);
u8 encoding = BTF_INT_ENCODING(int_data);
bool sign = encoding & BTF_INT_SIGNED;
u8 nr_bits = BTF_INT_BITS(int_data);
void *safe_data;
safe_data = btf_show_start_type(show, t, type_id, data);
if (!safe_data)
return;
if (bits_offset || BTF_INT_OFFSET(int_data) ||
BITS_PER_BYTE_MASKED(nr_bits)) {
btf_int_bits_show(btf, t, safe_data, bits_offset, show);
goto out;
}
switch (nr_bits) {
case 128:
btf_int128_print(show, safe_data);
break;
case 64:
if (sign)
btf_show_type_value(show, "%lld", *(s64 *)safe_data);
else
btf_show_type_value(show, "%llu", *(u64 *)safe_data);
break;
case 32:
if (sign)
btf_show_type_value(show, "%d", *(s32 *)safe_data);
else
btf_show_type_value(show, "%u", *(u32 *)safe_data);
break;
case 16:
if (sign)
btf_show_type_value(show, "%d", *(s16 *)safe_data);
else
btf_show_type_value(show, "%u", *(u16 *)safe_data);
break;
case 8:
if (show->state.array_encoding == BTF_INT_CHAR) {
/* check for null terminator */
if (show->state.array_terminated)
break;
if (*(char *)data == '\0') {
show->state.array_terminated = 1;
break;
}
if (isprint(*(char *)data)) {
btf_show_type_value(show, "'%c'",
*(char *)safe_data);
break;
}
}
if (sign)
btf_show_type_value(show, "%d", *(s8 *)safe_data);
else
btf_show_type_value(show, "%u", *(u8 *)safe_data);
break;
default:
btf_int_bits_show(btf, t, safe_data, bits_offset, show);
break;
}
out:
btf_show_end_type(show);
}
static const struct btf_kind_operations int_ops = {
.check_meta = btf_int_check_meta,
.resolve = btf_df_resolve,
.check_member = btf_int_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_int_check_kflag_member,
.log_details = btf_int_log,
.show = btf_int_show,
};
static int btf_modifier_check_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
const struct btf_type *resolved_type;
u32 resolved_type_id = member->type;
struct btf_member resolved_member;
struct btf *btf = env->btf;
resolved_type = btf_type_id_size(btf, &resolved_type_id, NULL);
if (!resolved_type) {
btf_verifier_log_member(env, struct_type, member,
"Invalid member");
return -EINVAL;
}
resolved_member = *member;
resolved_member.type = resolved_type_id;
return btf_type_ops(resolved_type)->check_member(env, struct_type,
&resolved_member,
resolved_type);
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
static int btf_modifier_check_kflag_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
const struct btf_type *resolved_type;
u32 resolved_type_id = member->type;
struct btf_member resolved_member;
struct btf *btf = env->btf;
resolved_type = btf_type_id_size(btf, &resolved_type_id, NULL);
if (!resolved_type) {
btf_verifier_log_member(env, struct_type, member,
"Invalid member");
return -EINVAL;
}
resolved_member = *member;
resolved_member.type = resolved_type_id;
return btf_type_ops(resolved_type)->check_kflag_member(env, struct_type,
&resolved_member,
resolved_type);
}
static int btf_ptr_check_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
u32 struct_size, struct_bits_off, bytes_offset;
struct_size = struct_type->size;
struct_bits_off = member->offset;
bytes_offset = BITS_ROUNDDOWN_BYTES(struct_bits_off);
if (BITS_PER_BYTE_MASKED(struct_bits_off)) {
btf_verifier_log_member(env, struct_type, member,
"Member is not byte aligned");
return -EINVAL;
}
if (struct_size - bytes_offset < sizeof(void *)) {
btf_verifier_log_member(env, struct_type, member,
"Member exceeds struct_size");
return -EINVAL;
}
return 0;
}
static int btf_ref_type_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
const char *value;
if (btf_type_vlen(t)) {
btf_verifier_log_type(env, t, "vlen != 0");
return -EINVAL;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
if (!BTF_TYPE_ID_VALID(t->type)) {
btf_verifier_log_type(env, t, "Invalid type_id");
return -EINVAL;
}
/* typedef/type_tag type must have a valid name, and other ref types,
* volatile, const, restrict, should have a null name.
*/
if (BTF_INFO_KIND(t->info) == BTF_KIND_TYPEDEF) {
if (!t->name_off ||
!btf_name_valid_identifier(env->btf, t->name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
} else if (BTF_INFO_KIND(t->info) == BTF_KIND_TYPE_TAG) {
value = btf_name_by_offset(env->btf, t->name_off);
if (!value || !value[0]) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
} else {
if (t->name_off) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
}
btf_verifier_log_type(env, t, NULL);
return 0;
}
static int btf_modifier_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
const struct btf_type *t = v->t;
const struct btf_type *next_type;
u32 next_type_id = t->type;
struct btf *btf = env->btf;
next_type = btf_type_by_id(btf, next_type_id);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
if (!next_type || btf_type_is_resolve_source_only(next_type)) {
btf_verifier_log_type(env, v->t, "Invalid type_id");
return -EINVAL;
}
if (!env_type_is_resolve_sink(env, next_type) &&
!env_type_is_resolved(env, next_type_id))
return env_stack_push(env, next_type, next_type_id);
/* Figure out the resolved next_type_id with size.
* They will be stored in the current modifier's
* resolved_ids and resolved_sizes such that it can
* save us a few type-following when we use it later (e.g. in
* pretty print).
*/
bpf: fix BTF verifier size resolution logic BTF verifier has a size resolution bug which in some circumstances leads to invalid size resolution for, e.g., TYPEDEF modifier. This happens if we have [1] PTR -> [2] TYPEDEF -> [3] ARRAY, in which case due to being in pointer context ARRAY size won't be resolved (because for pointer it doesn't matter, so it's a sink in pointer context), but it will be permanently remembered as zero for TYPEDEF and TYPEDEF will be marked as RESOLVED. Eventually ARRAY size will be resolved correctly, but TYPEDEF resolved_size won't be updated anymore. This, subsequently, will lead to erroneous map creation failure, if that TYPEDEF is specified as either key or value, as key_size/value_size won't correspond to resolved size of TYPEDEF (kernel will believe it's zero). Note, that if BTF was ordered as [1] ARRAY <- [2] TYPEDEF <- [3] PTR, this won't be a problem, as by the time we get to TYPEDEF, ARRAY's size is already calculated and stored. This bug manifests itself in rejecting BTF-defined maps that use array typedef as a value type: typedef int array_t[16]; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(value, array_t); /* i.e., array_t *value; */ } test_map SEC(".maps"); The fix consists on not relying on modifier's resolved_size and instead using modifier's resolved_id (type ID for "concrete" type to which modifier eventually resolves) and doing size determination for that resolved type. This allow to preserve existing "early DFS termination" logic for PTR or STRUCT_OR_ARRAY contexts, but still do correct size determination for modifier types. Fixes: eb3f595dab40 ("bpf: btf: Validate type reference") Cc: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-07-13 01:25:55 +08:00
if (!btf_type_id_size(btf, &next_type_id, NULL)) {
if (env_type_is_resolved(env, next_type_id))
next_type = btf_type_id_resolve(btf, &next_type_id);
/* "typedef void new_void", "const void"...etc */
if (!btf_type_is_void(next_type) &&
!btf_type_is_fwd(next_type) &&
!btf_type_is_func_proto(next_type)) {
btf_verifier_log_type(env, v->t, "Invalid type_id");
return -EINVAL;
}
}
bpf: fix BTF verifier size resolution logic BTF verifier has a size resolution bug which in some circumstances leads to invalid size resolution for, e.g., TYPEDEF modifier. This happens if we have [1] PTR -> [2] TYPEDEF -> [3] ARRAY, in which case due to being in pointer context ARRAY size won't be resolved (because for pointer it doesn't matter, so it's a sink in pointer context), but it will be permanently remembered as zero for TYPEDEF and TYPEDEF will be marked as RESOLVED. Eventually ARRAY size will be resolved correctly, but TYPEDEF resolved_size won't be updated anymore. This, subsequently, will lead to erroneous map creation failure, if that TYPEDEF is specified as either key or value, as key_size/value_size won't correspond to resolved size of TYPEDEF (kernel will believe it's zero). Note, that if BTF was ordered as [1] ARRAY <- [2] TYPEDEF <- [3] PTR, this won't be a problem, as by the time we get to TYPEDEF, ARRAY's size is already calculated and stored. This bug manifests itself in rejecting BTF-defined maps that use array typedef as a value type: typedef int array_t[16]; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(value, array_t); /* i.e., array_t *value; */ } test_map SEC(".maps"); The fix consists on not relying on modifier's resolved_size and instead using modifier's resolved_id (type ID for "concrete" type to which modifier eventually resolves) and doing size determination for that resolved type. This allow to preserve existing "early DFS termination" logic for PTR or STRUCT_OR_ARRAY contexts, but still do correct size determination for modifier types. Fixes: eb3f595dab40 ("bpf: btf: Validate type reference") Cc: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-07-13 01:25:55 +08:00
env_stack_pop_resolved(env, next_type_id, 0);
return 0;
}
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
static int btf_var_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
const struct btf_type *next_type;
const struct btf_type *t = v->t;
u32 next_type_id = t->type;
struct btf *btf = env->btf;
next_type = btf_type_by_id(btf, next_type_id);
if (!next_type || btf_type_is_resolve_source_only(next_type)) {
btf_verifier_log_type(env, v->t, "Invalid type_id");
return -EINVAL;
}
if (!env_type_is_resolve_sink(env, next_type) &&
!env_type_is_resolved(env, next_type_id))
return env_stack_push(env, next_type, next_type_id);
if (btf_type_is_modifier(next_type)) {
const struct btf_type *resolved_type;
u32 resolved_type_id;
resolved_type_id = next_type_id;
resolved_type = btf_type_id_resolve(btf, &resolved_type_id);
if (btf_type_is_ptr(resolved_type) &&
!env_type_is_resolve_sink(env, resolved_type) &&
!env_type_is_resolved(env, resolved_type_id))
return env_stack_push(env, resolved_type,
resolved_type_id);
}
/* We must resolve to something concrete at this point, no
* forward types or similar that would resolve to size of
* zero is allowed.
*/
bpf: fix BTF verifier size resolution logic BTF verifier has a size resolution bug which in some circumstances leads to invalid size resolution for, e.g., TYPEDEF modifier. This happens if we have [1] PTR -> [2] TYPEDEF -> [3] ARRAY, in which case due to being in pointer context ARRAY size won't be resolved (because for pointer it doesn't matter, so it's a sink in pointer context), but it will be permanently remembered as zero for TYPEDEF and TYPEDEF will be marked as RESOLVED. Eventually ARRAY size will be resolved correctly, but TYPEDEF resolved_size won't be updated anymore. This, subsequently, will lead to erroneous map creation failure, if that TYPEDEF is specified as either key or value, as key_size/value_size won't correspond to resolved size of TYPEDEF (kernel will believe it's zero). Note, that if BTF was ordered as [1] ARRAY <- [2] TYPEDEF <- [3] PTR, this won't be a problem, as by the time we get to TYPEDEF, ARRAY's size is already calculated and stored. This bug manifests itself in rejecting BTF-defined maps that use array typedef as a value type: typedef int array_t[16]; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(value, array_t); /* i.e., array_t *value; */ } test_map SEC(".maps"); The fix consists on not relying on modifier's resolved_size and instead using modifier's resolved_id (type ID for "concrete" type to which modifier eventually resolves) and doing size determination for that resolved type. This allow to preserve existing "early DFS termination" logic for PTR or STRUCT_OR_ARRAY contexts, but still do correct size determination for modifier types. Fixes: eb3f595dab40 ("bpf: btf: Validate type reference") Cc: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-07-13 01:25:55 +08:00
if (!btf_type_id_size(btf, &next_type_id, NULL)) {
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
btf_verifier_log_type(env, v->t, "Invalid type_id");
return -EINVAL;
}
bpf: fix BTF verifier size resolution logic BTF verifier has a size resolution bug which in some circumstances leads to invalid size resolution for, e.g., TYPEDEF modifier. This happens if we have [1] PTR -> [2] TYPEDEF -> [3] ARRAY, in which case due to being in pointer context ARRAY size won't be resolved (because for pointer it doesn't matter, so it's a sink in pointer context), but it will be permanently remembered as zero for TYPEDEF and TYPEDEF will be marked as RESOLVED. Eventually ARRAY size will be resolved correctly, but TYPEDEF resolved_size won't be updated anymore. This, subsequently, will lead to erroneous map creation failure, if that TYPEDEF is specified as either key or value, as key_size/value_size won't correspond to resolved size of TYPEDEF (kernel will believe it's zero). Note, that if BTF was ordered as [1] ARRAY <- [2] TYPEDEF <- [3] PTR, this won't be a problem, as by the time we get to TYPEDEF, ARRAY's size is already calculated and stored. This bug manifests itself in rejecting BTF-defined maps that use array typedef as a value type: typedef int array_t[16]; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(value, array_t); /* i.e., array_t *value; */ } test_map SEC(".maps"); The fix consists on not relying on modifier's resolved_size and instead using modifier's resolved_id (type ID for "concrete" type to which modifier eventually resolves) and doing size determination for that resolved type. This allow to preserve existing "early DFS termination" logic for PTR or STRUCT_OR_ARRAY contexts, but still do correct size determination for modifier types. Fixes: eb3f595dab40 ("bpf: btf: Validate type reference") Cc: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-07-13 01:25:55 +08:00
env_stack_pop_resolved(env, next_type_id, 0);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
return 0;
}
static int btf_ptr_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
const struct btf_type *next_type;
const struct btf_type *t = v->t;
u32 next_type_id = t->type;
struct btf *btf = env->btf;
next_type = btf_type_by_id(btf, next_type_id);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
if (!next_type || btf_type_is_resolve_source_only(next_type)) {
btf_verifier_log_type(env, v->t, "Invalid type_id");
return -EINVAL;
}
if (!env_type_is_resolve_sink(env, next_type) &&
!env_type_is_resolved(env, next_type_id))
return env_stack_push(env, next_type, next_type_id);
/* If the modifier was RESOLVED during RESOLVE_STRUCT_OR_ARRAY,
* the modifier may have stopped resolving when it was resolved
* to a ptr (last-resolved-ptr).
*
* We now need to continue from the last-resolved-ptr to
* ensure the last-resolved-ptr will not referring back to
* the current ptr (t).
*/
if (btf_type_is_modifier(next_type)) {
const struct btf_type *resolved_type;
u32 resolved_type_id;
resolved_type_id = next_type_id;
resolved_type = btf_type_id_resolve(btf, &resolved_type_id);
if (btf_type_is_ptr(resolved_type) &&
!env_type_is_resolve_sink(env, resolved_type) &&
!env_type_is_resolved(env, resolved_type_id))
return env_stack_push(env, resolved_type,
resolved_type_id);
}
if (!btf_type_id_size(btf, &next_type_id, NULL)) {
if (env_type_is_resolved(env, next_type_id))
next_type = btf_type_id_resolve(btf, &next_type_id);
if (!btf_type_is_void(next_type) &&
!btf_type_is_fwd(next_type) &&
!btf_type_is_func_proto(next_type)) {
btf_verifier_log_type(env, v->t, "Invalid type_id");
return -EINVAL;
}
}
env_stack_pop_resolved(env, next_type_id, 0);
return 0;
}
static void btf_modifier_show(const struct btf *btf,
const struct btf_type *t,
u32 type_id, void *data,
u8 bits_offset, struct btf_show *show)
{
bpf: Introduce BPF_MAP_TYPE_STRUCT_OPS The patch introduces BPF_MAP_TYPE_STRUCT_OPS. The map value is a kernel struct with its func ptr implemented in bpf prog. This new map is the interface to register/unregister/introspect a bpf implemented kernel struct. The kernel struct is actually embedded inside another new struct (or called the "value" struct in the code). For example, "struct tcp_congestion_ops" is embbeded in: struct bpf_struct_ops_tcp_congestion_ops { refcount_t refcnt; enum bpf_struct_ops_state state; struct tcp_congestion_ops data; /* <-- kernel subsystem struct here */ } The map value is "struct bpf_struct_ops_tcp_congestion_ops". The "bpftool map dump" will then be able to show the state ("inuse"/"tobefree") and the number of subsystem's refcnt (e.g. number of tcp_sock in the tcp_congestion_ops case). This "value" struct is created automatically by a macro. Having a separate "value" struct will also make extending "struct bpf_struct_ops_XYZ" easier (e.g. adding "void (*init)(void)" to "struct bpf_struct_ops_XYZ" to do some initialization works before registering the struct_ops to the kernel subsystem). The libbpf will take care of finding and populating the "struct bpf_struct_ops_XYZ" from "struct XYZ". Register a struct_ops to a kernel subsystem: 1. Load all needed BPF_PROG_TYPE_STRUCT_OPS prog(s) 2. Create a BPF_MAP_TYPE_STRUCT_OPS with attr->btf_vmlinux_value_type_id set to the btf id "struct bpf_struct_ops_tcp_congestion_ops" of the running kernel. Instead of reusing the attr->btf_value_type_id, btf_vmlinux_value_type_id s added such that attr->btf_fd can still be used as the "user" btf which could store other useful sysadmin/debug info that may be introduced in the furture, e.g. creation-date/compiler-details/map-creator...etc. 3. Create a "struct bpf_struct_ops_tcp_congestion_ops" object as described in the running kernel btf. Populate the value of this object. The function ptr should be populated with the prog fds. 4. Call BPF_MAP_UPDATE with the object created in (3) as the map value. The key is always "0". During BPF_MAP_UPDATE, the code that saves the kernel-func-ptr's args as an array of u64 is generated. BPF_MAP_UPDATE also allows the specific struct_ops to do some final checks in "st_ops->init_member()" (e.g. ensure all mandatory func ptrs are implemented). If everything looks good, it will register this kernel struct to the kernel subsystem. The map will not allow further update from this point. Unregister a struct_ops from the kernel subsystem: BPF_MAP_DELETE with key "0". Introspect a struct_ops: BPF_MAP_LOOKUP_ELEM with key "0". The map value returned will have the prog _id_ populated as the func ptr. The map value state (enum bpf_struct_ops_state) will transit from: INIT (map created) => INUSE (map updated, i.e. reg) => TOBEFREE (map value deleted, i.e. unreg) The kernel subsystem needs to call bpf_struct_ops_get() and bpf_struct_ops_put() to manage the "refcnt" in the "struct bpf_struct_ops_XYZ". This patch uses a separate refcnt for the purose of tracking the subsystem usage. Another approach is to reuse the map->refcnt and then "show" (i.e. during map_lookup) the subsystem's usage by doing map->refcnt - map->usercnt to filter out the map-fd/pinned-map usage. However, that will also tie down the future semantics of map->refcnt and map->usercnt. The very first subsystem's refcnt (during reg()) holds one count to map->refcnt. When the very last subsystem's refcnt is gone, it will also release the map->refcnt. All bpf_prog will be freed when the map->refcnt reaches 0 (i.e. during map_free()). Here is how the bpftool map command will look like: [root@arch-fb-vm1 bpf]# bpftool map show 6: struct_ops name dctcp flags 0x0 key 4B value 256B max_entries 1 memlock 4096B btf_id 6 [root@arch-fb-vm1 bpf]# bpftool map dump id 6 [{ "value": { "refcnt": { "refs": { "counter": 1 } }, "state": 1, "data": { "list": { "next": 0, "prev": 0 }, "key": 0, "flags": 2, "init": 24, "release": 0, "ssthresh": 25, "cong_avoid": 30, "set_state": 27, "cwnd_event": 28, "in_ack_event": 26, "undo_cwnd": 29, "pkts_acked": 0, "min_tso_segs": 0, "sndbuf_expand": 0, "cong_control": 0, "get_info": 0, "name": [98,112,102,95,100,99,116,99,112,0,0,0,0,0,0,0 ], "owner": 0 } } } ] Misc Notes: * bpf_struct_ops_map_sys_lookup_elem() is added for syscall lookup. It does an inplace update on "*value" instead returning a pointer to syscall.c. Otherwise, it needs a separate copy of "zero" value for the BPF_STRUCT_OPS_STATE_INIT to avoid races. * The bpf_struct_ops_map_delete_elem() is also called without preempt_disable() from map_delete_elem(). It is because the "->unreg()" may requires sleepable context, e.g. the "tcp_unregister_congestion_control()". * "const" is added to some of the existing "struct btf_func_model *" function arg to avoid a compiler warning caused by this patch. Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200109003505.3855919-1-kafai@fb.com
2020-01-09 08:35:05 +08:00
if (btf->resolved_ids)
t = btf_type_id_resolve(btf, &type_id);
else
t = btf_type_skip_modifiers(btf, type_id, NULL);
btf_type_ops(t)->show(btf, t, type_id, data, bits_offset, show);
}
static void btf_var_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
{
t = btf_type_id_resolve(btf, &type_id);
btf_type_ops(t)->show(btf, t, type_id, data, bits_offset, show);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
}
static void btf_ptr_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
{
void *safe_data;
safe_data = btf_show_start_type(show, t, type_id, data);
if (!safe_data)
return;
/* It is a hashed value unless BTF_SHOW_PTR_RAW is specified */
if (show->flags & BTF_SHOW_PTR_RAW)
btf_show_type_value(show, "0x%px", *(void **)safe_data);
else
btf_show_type_value(show, "0x%p", *(void **)safe_data);
btf_show_end_type(show);
}
static void btf_ref_type_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
btf_verifier_log(env, "type_id=%u", t->type);
}
static struct btf_kind_operations modifier_ops = {
.check_meta = btf_ref_type_check_meta,
.resolve = btf_modifier_resolve,
.check_member = btf_modifier_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_modifier_check_kflag_member,
.log_details = btf_ref_type_log,
.show = btf_modifier_show,
};
static struct btf_kind_operations ptr_ops = {
.check_meta = btf_ref_type_check_meta,
.resolve = btf_ptr_resolve,
.check_member = btf_ptr_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_generic_check_kflag_member,
.log_details = btf_ref_type_log,
.show = btf_ptr_show,
};
static s32 btf_fwd_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
if (btf_type_vlen(t)) {
btf_verifier_log_type(env, t, "vlen != 0");
return -EINVAL;
}
if (t->type) {
btf_verifier_log_type(env, t, "type != 0");
return -EINVAL;
}
/* fwd type must have a valid name */
if (!t->name_off ||
!btf_name_valid_identifier(env->btf, t->name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
return 0;
}
static void btf_fwd_type_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
btf_verifier_log(env, "%s", btf_type_kflag(t) ? "union" : "struct");
}
static struct btf_kind_operations fwd_ops = {
.check_meta = btf_fwd_check_meta,
.resolve = btf_df_resolve,
.check_member = btf_df_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_df_check_kflag_member,
.log_details = btf_fwd_type_log,
.show = btf_df_show,
};
static int btf_array_check_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
u32 struct_bits_off = member->offset;
u32 struct_size, bytes_offset;
u32 array_type_id, array_size;
struct btf *btf = env->btf;
if (BITS_PER_BYTE_MASKED(struct_bits_off)) {
btf_verifier_log_member(env, struct_type, member,
"Member is not byte aligned");
return -EINVAL;
}
array_type_id = member->type;
btf_type_id_size(btf, &array_type_id, &array_size);
struct_size = struct_type->size;
bytes_offset = BITS_ROUNDDOWN_BYTES(struct_bits_off);
if (struct_size - bytes_offset < array_size) {
btf_verifier_log_member(env, struct_type, member,
"Member exceeds struct_size");
return -EINVAL;
}
return 0;
}
static s32 btf_array_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
const struct btf_array *array = btf_type_array(t);
u32 meta_needed = sizeof(*array);
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
/* array type should not have a name */
if (t->name_off) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
if (btf_type_vlen(t)) {
btf_verifier_log_type(env, t, "vlen != 0");
return -EINVAL;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
if (t->size) {
btf_verifier_log_type(env, t, "size != 0");
return -EINVAL;
}
/* Array elem type and index type cannot be in type void,
* so !array->type and !array->index_type are not allowed.
*/
if (!array->type || !BTF_TYPE_ID_VALID(array->type)) {
btf_verifier_log_type(env, t, "Invalid elem");
return -EINVAL;
}
if (!array->index_type || !BTF_TYPE_ID_VALID(array->index_type)) {
btf_verifier_log_type(env, t, "Invalid index");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
return meta_needed;
}
static int btf_array_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
const struct btf_array *array = btf_type_array(v->t);
const struct btf_type *elem_type, *index_type;
u32 elem_type_id, index_type_id;
struct btf *btf = env->btf;
u32 elem_size;
/* Check array->index_type */
index_type_id = array->index_type;
index_type = btf_type_by_id(btf, index_type_id);
if (btf_type_nosize_or_null(index_type) ||
btf_type_is_resolve_source_only(index_type)) {
btf_verifier_log_type(env, v->t, "Invalid index");
return -EINVAL;
}
if (!env_type_is_resolve_sink(env, index_type) &&
!env_type_is_resolved(env, index_type_id))
return env_stack_push(env, index_type, index_type_id);
index_type = btf_type_id_size(btf, &index_type_id, NULL);
if (!index_type || !btf_type_is_int(index_type) ||
!btf_type_int_is_regular(index_type)) {
btf_verifier_log_type(env, v->t, "Invalid index");
return -EINVAL;
}
/* Check array->type */
elem_type_id = array->type;
elem_type = btf_type_by_id(btf, elem_type_id);
if (btf_type_nosize_or_null(elem_type) ||
btf_type_is_resolve_source_only(elem_type)) {
btf_verifier_log_type(env, v->t,
"Invalid elem");
return -EINVAL;
}
if (!env_type_is_resolve_sink(env, elem_type) &&
!env_type_is_resolved(env, elem_type_id))
return env_stack_push(env, elem_type, elem_type_id);
elem_type = btf_type_id_size(btf, &elem_type_id, &elem_size);
if (!elem_type) {
btf_verifier_log_type(env, v->t, "Invalid elem");
return -EINVAL;
}
if (btf_type_is_int(elem_type) && !btf_type_int_is_regular(elem_type)) {
btf_verifier_log_type(env, v->t, "Invalid array of int");
return -EINVAL;
}
if (array->nelems && elem_size > U32_MAX / array->nelems) {
btf_verifier_log_type(env, v->t,
"Array size overflows U32_MAX");
return -EINVAL;
}
env_stack_pop_resolved(env, elem_type_id, elem_size * array->nelems);
return 0;
}
static void btf_array_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
const struct btf_array *array = btf_type_array(t);
btf_verifier_log(env, "type_id=%u index_type_id=%u nr_elems=%u",
array->type, array->index_type, array->nelems);
}
static void __btf_array_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
{
const struct btf_array *array = btf_type_array(t);
const struct btf_kind_operations *elem_ops;
const struct btf_type *elem_type;
u32 i, elem_size = 0, elem_type_id;
u16 encoding = 0;
elem_type_id = array->type;
elem_type = btf_type_skip_modifiers(btf, elem_type_id, NULL);
if (elem_type && btf_type_has_size(elem_type))
elem_size = elem_type->size;
if (elem_type && btf_type_is_int(elem_type)) {
u32 int_type = btf_type_int(elem_type);
encoding = BTF_INT_ENCODING(int_type);
/*
* BTF_INT_CHAR encoding never seems to be set for
* char arrays, so if size is 1 and element is
* printable as a char, we'll do that.
*/
if (elem_size == 1)
encoding = BTF_INT_CHAR;
}
if (!btf_show_start_array_type(show, t, type_id, encoding, data))
return;
if (!elem_type)
goto out;
elem_ops = btf_type_ops(elem_type);
for (i = 0; i < array->nelems; i++) {
btf_show_start_array_member(show);
elem_ops->show(btf, elem_type, elem_type_id, data,
bits_offset, show);
data += elem_size;
btf_show_end_array_member(show);
if (show->state.array_terminated)
break;
}
out:
btf_show_end_array_type(show);
}
static void btf_array_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
{
const struct btf_member *m = show->state.member;
/*
* First check if any members would be shown (are non-zero).
* See comments above "struct btf_show" definition for more
* details on how this works at a high-level.
*/
if (show->state.depth > 0 && !(show->flags & BTF_SHOW_ZERO)) {
if (!show->state.depth_check) {
show->state.depth_check = show->state.depth + 1;
show->state.depth_to_show = 0;
}
__btf_array_show(btf, t, type_id, data, bits_offset, show);
show->state.member = m;
if (show->state.depth_check != show->state.depth + 1)
return;
show->state.depth_check = 0;
if (show->state.depth_to_show <= show->state.depth)
return;
/*
* Reaching here indicates we have recursed and found
* non-zero array member(s).
*/
}
__btf_array_show(btf, t, type_id, data, bits_offset, show);
}
static struct btf_kind_operations array_ops = {
.check_meta = btf_array_check_meta,
.resolve = btf_array_resolve,
.check_member = btf_array_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_generic_check_kflag_member,
.log_details = btf_array_log,
.show = btf_array_show,
};
static int btf_struct_check_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
u32 struct_bits_off = member->offset;
u32 struct_size, bytes_offset;
if (BITS_PER_BYTE_MASKED(struct_bits_off)) {
btf_verifier_log_member(env, struct_type, member,
"Member is not byte aligned");
return -EINVAL;
}
struct_size = struct_type->size;
bytes_offset = BITS_ROUNDDOWN_BYTES(struct_bits_off);
if (struct_size - bytes_offset < member_type->size) {
btf_verifier_log_member(env, struct_type, member,
"Member exceeds struct_size");
return -EINVAL;
}
return 0;
}
static s32 btf_struct_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
bool is_union = BTF_INFO_KIND(t->info) == BTF_KIND_UNION;
const struct btf_member *member;
u32 meta_needed, last_offset;
struct btf *btf = env->btf;
u32 struct_size = t->size;
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
u32 offset;
u16 i;
meta_needed = btf_type_vlen(t) * sizeof(*member);
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
/* struct type either no name or a valid one */
if (t->name_off &&
!btf_name_valid_identifier(env->btf, t->name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
last_offset = 0;
for_each_member(i, t, member) {
if (!btf_name_offset_valid(btf, member->name_off)) {
btf_verifier_log_member(env, t, member,
"Invalid member name_offset:%u",
member->name_off);
return -EINVAL;
}
/* struct member either no name or a valid one */
if (member->name_off &&
!btf_name_valid_identifier(btf, member->name_off)) {
btf_verifier_log_member(env, t, member, "Invalid name");
return -EINVAL;
}
/* A member cannot be in type void */
if (!member->type || !BTF_TYPE_ID_VALID(member->type)) {
btf_verifier_log_member(env, t, member,
"Invalid type_id");
return -EINVAL;
}
offset = __btf_member_bit_offset(t, member);
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (is_union && offset) {
btf_verifier_log_member(env, t, member,
"Invalid member bits_offset");
return -EINVAL;
}
/*
* ">" instead of ">=" because the last member could be
* "char a[0];"
*/
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (last_offset > offset) {
btf_verifier_log_member(env, t, member,
"Invalid member bits_offset");
return -EINVAL;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (BITS_ROUNDUP_BYTES(offset) > struct_size) {
btf_verifier_log_member(env, t, member,
"Member bits_offset exceeds its struct size");
return -EINVAL;
}
btf_verifier_log_member(env, t, member, NULL);
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
last_offset = offset;
}
return meta_needed;
}
static int btf_struct_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
const struct btf_member *member;
int err;
u16 i;
/* Before continue resolving the next_member,
* ensure the last member is indeed resolved to a
* type with size info.
*/
if (v->next_member) {
const struct btf_type *last_member_type;
const struct btf_member *last_member;
u32 last_member_type_id;
last_member = btf_type_member(v->t) + v->next_member - 1;
last_member_type_id = last_member->type;
if (WARN_ON_ONCE(!env_type_is_resolved(env,
last_member_type_id)))
return -EINVAL;
last_member_type = btf_type_by_id(env->btf,
last_member_type_id);
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (btf_type_kflag(v->t))
err = btf_type_ops(last_member_type)->check_kflag_member(env, v->t,
last_member,
last_member_type);
else
err = btf_type_ops(last_member_type)->check_member(env, v->t,
last_member,
last_member_type);
if (err)
return err;
}
for_each_member_from(i, v->next_member, v->t, member) {
u32 member_type_id = member->type;
const struct btf_type *member_type = btf_type_by_id(env->btf,
member_type_id);
if (btf_type_nosize_or_null(member_type) ||
btf_type_is_resolve_source_only(member_type)) {
btf_verifier_log_member(env, v->t, member,
"Invalid member");
return -EINVAL;
}
if (!env_type_is_resolve_sink(env, member_type) &&
!env_type_is_resolved(env, member_type_id)) {
env_stack_set_next_member(env, i + 1);
return env_stack_push(env, member_type, member_type_id);
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (btf_type_kflag(v->t))
err = btf_type_ops(member_type)->check_kflag_member(env, v->t,
member,
member_type);
else
err = btf_type_ops(member_type)->check_member(env, v->t,
member,
member_type);
if (err)
return err;
}
env_stack_pop_resolved(env, 0, 0);
return 0;
}
static void btf_struct_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
btf_verifier_log(env, "size=%u vlen=%u", t->size, btf_type_vlen(t));
}
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
enum {
BTF_FIELD_IGNORE = 0,
BTF_FIELD_FOUND = 1,
};
struct btf_field_info {
enum btf_field_type type;
u32 off;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
union {
struct {
u32 type_id;
} kptr;
struct {
const char *node_name;
u32 value_btf_id;
} graph_root;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
};
};
static int btf_find_struct(const struct btf *btf, const struct btf_type *t,
u32 off, int sz, enum btf_field_type field_type,
struct btf_field_info *info)
{
if (!__btf_type_is_struct(t))
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return BTF_FIELD_IGNORE;
if (t->size != sz)
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return BTF_FIELD_IGNORE;
info->type = field_type;
info->off = off;
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return BTF_FIELD_FOUND;
}
static int btf_find_kptr(const struct btf *btf, const struct btf_type *t,
u32 off, int sz, struct btf_field_info *info)
{
enum btf_field_type type;
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
u32 res_id;
/* Permit modifiers on the pointer itself */
if (btf_type_is_volatile(t))
t = btf_type_by_id(btf, t->type);
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
/* For PTR, sz is always == 8 */
if (!btf_type_is_ptr(t))
return BTF_FIELD_IGNORE;
t = btf_type_by_id(btf, t->type);
if (!btf_type_is_type_tag(t))
return BTF_FIELD_IGNORE;
/* Reject extra tags */
if (btf_type_is_type_tag(btf_type_by_id(btf, t->type)))
return -EINVAL;
if (!strcmp("kptr_untrusted", __btf_name_by_offset(btf, t->name_off)))
bpf: Allow storing referenced kptr in map Extending the code in previous commits, introduce referenced kptr support, which needs to be tagged using 'kptr_ref' tag instead. Unlike unreferenced kptr, referenced kptr have a lot more restrictions. In addition to the type matching, only a newly introduced bpf_kptr_xchg helper is allowed to modify the map value at that offset. This transfers the referenced pointer being stored into the map, releasing the references state for the program, and returning the old value and creating new reference state for the returned pointer. Similar to unreferenced pointer case, return value for this case will also be PTR_TO_BTF_ID_OR_NULL. The reference for the returned pointer must either be eventually released by calling the corresponding release function, otherwise it must be transferred into another map. It is also allowed to call bpf_kptr_xchg with a NULL pointer, to clear the value, and obtain the old value if any. BPF_LDX, BPF_STX, and BPF_ST cannot access referenced kptr. A future commit will permit using BPF_LDX for such pointers, but attempt at making it safe, since the lifetime of object won't be guaranteed. There are valid reasons to enforce the restriction of permitting only bpf_kptr_xchg to operate on referenced kptr. The pointer value must be consistent in face of concurrent modification, and any prior values contained in the map must also be released before a new one is moved into the map. To ensure proper transfer of this ownership, bpf_kptr_xchg returns the old value, which the verifier would require the user to either free or move into another map, and releases the reference held for the pointer being moved in. In the future, direct BPF_XCHG instruction may also be permitted to work like bpf_kptr_xchg helper. Note that process_kptr_func doesn't have to call check_helper_mem_access, since we already disallow rdonly/wronly flags for map, which is what check_map_access_type checks, and we already ensure the PTR_TO_MAP_VALUE refers to kptr by obtaining its off_desc, so check_map_access is also not required. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-4-memxor@gmail.com
2022-04-25 05:48:51 +08:00
type = BPF_KPTR_UNREF;
else if (!strcmp("kptr", __btf_name_by_offset(btf, t->name_off)))
bpf: Allow storing referenced kptr in map Extending the code in previous commits, introduce referenced kptr support, which needs to be tagged using 'kptr_ref' tag instead. Unlike unreferenced kptr, referenced kptr have a lot more restrictions. In addition to the type matching, only a newly introduced bpf_kptr_xchg helper is allowed to modify the map value at that offset. This transfers the referenced pointer being stored into the map, releasing the references state for the program, and returning the old value and creating new reference state for the returned pointer. Similar to unreferenced pointer case, return value for this case will also be PTR_TO_BTF_ID_OR_NULL. The reference for the returned pointer must either be eventually released by calling the corresponding release function, otherwise it must be transferred into another map. It is also allowed to call bpf_kptr_xchg with a NULL pointer, to clear the value, and obtain the old value if any. BPF_LDX, BPF_STX, and BPF_ST cannot access referenced kptr. A future commit will permit using BPF_LDX for such pointers, but attempt at making it safe, since the lifetime of object won't be guaranteed. There are valid reasons to enforce the restriction of permitting only bpf_kptr_xchg to operate on referenced kptr. The pointer value must be consistent in face of concurrent modification, and any prior values contained in the map must also be released before a new one is moved into the map. To ensure proper transfer of this ownership, bpf_kptr_xchg returns the old value, which the verifier would require the user to either free or move into another map, and releases the reference held for the pointer being moved in. In the future, direct BPF_XCHG instruction may also be permitted to work like bpf_kptr_xchg helper. Note that process_kptr_func doesn't have to call check_helper_mem_access, since we already disallow rdonly/wronly flags for map, which is what check_map_access_type checks, and we already ensure the PTR_TO_MAP_VALUE refers to kptr by obtaining its off_desc, so check_map_access is also not required. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-4-memxor@gmail.com
2022-04-25 05:48:51 +08:00
type = BPF_KPTR_REF;
else if (!strcmp("percpu_kptr", __btf_name_by_offset(btf, t->name_off)))
type = BPF_KPTR_PERCPU;
bpf: Allow storing referenced kptr in map Extending the code in previous commits, introduce referenced kptr support, which needs to be tagged using 'kptr_ref' tag instead. Unlike unreferenced kptr, referenced kptr have a lot more restrictions. In addition to the type matching, only a newly introduced bpf_kptr_xchg helper is allowed to modify the map value at that offset. This transfers the referenced pointer being stored into the map, releasing the references state for the program, and returning the old value and creating new reference state for the returned pointer. Similar to unreferenced pointer case, return value for this case will also be PTR_TO_BTF_ID_OR_NULL. The reference for the returned pointer must either be eventually released by calling the corresponding release function, otherwise it must be transferred into another map. It is also allowed to call bpf_kptr_xchg with a NULL pointer, to clear the value, and obtain the old value if any. BPF_LDX, BPF_STX, and BPF_ST cannot access referenced kptr. A future commit will permit using BPF_LDX for such pointers, but attempt at making it safe, since the lifetime of object won't be guaranteed. There are valid reasons to enforce the restriction of permitting only bpf_kptr_xchg to operate on referenced kptr. The pointer value must be consistent in face of concurrent modification, and any prior values contained in the map must also be released before a new one is moved into the map. To ensure proper transfer of this ownership, bpf_kptr_xchg returns the old value, which the verifier would require the user to either free or move into another map, and releases the reference held for the pointer being moved in. In the future, direct BPF_XCHG instruction may also be permitted to work like bpf_kptr_xchg helper. Note that process_kptr_func doesn't have to call check_helper_mem_access, since we already disallow rdonly/wronly flags for map, which is what check_map_access_type checks, and we already ensure the PTR_TO_MAP_VALUE refers to kptr by obtaining its off_desc, so check_map_access is also not required. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-4-memxor@gmail.com
2022-04-25 05:48:51 +08:00
else
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return -EINVAL;
/* Get the base type */
t = btf_type_skip_modifiers(btf, t->type, &res_id);
/* Only pointer to struct is allowed */
if (!__btf_type_is_struct(t))
return -EINVAL;
bpf: Allow storing referenced kptr in map Extending the code in previous commits, introduce referenced kptr support, which needs to be tagged using 'kptr_ref' tag instead. Unlike unreferenced kptr, referenced kptr have a lot more restrictions. In addition to the type matching, only a newly introduced bpf_kptr_xchg helper is allowed to modify the map value at that offset. This transfers the referenced pointer being stored into the map, releasing the references state for the program, and returning the old value and creating new reference state for the returned pointer. Similar to unreferenced pointer case, return value for this case will also be PTR_TO_BTF_ID_OR_NULL. The reference for the returned pointer must either be eventually released by calling the corresponding release function, otherwise it must be transferred into another map. It is also allowed to call bpf_kptr_xchg with a NULL pointer, to clear the value, and obtain the old value if any. BPF_LDX, BPF_STX, and BPF_ST cannot access referenced kptr. A future commit will permit using BPF_LDX for such pointers, but attempt at making it safe, since the lifetime of object won't be guaranteed. There are valid reasons to enforce the restriction of permitting only bpf_kptr_xchg to operate on referenced kptr. The pointer value must be consistent in face of concurrent modification, and any prior values contained in the map must also be released before a new one is moved into the map. To ensure proper transfer of this ownership, bpf_kptr_xchg returns the old value, which the verifier would require the user to either free or move into another map, and releases the reference held for the pointer being moved in. In the future, direct BPF_XCHG instruction may also be permitted to work like bpf_kptr_xchg helper. Note that process_kptr_func doesn't have to call check_helper_mem_access, since we already disallow rdonly/wronly flags for map, which is what check_map_access_type checks, and we already ensure the PTR_TO_MAP_VALUE refers to kptr by obtaining its off_desc, so check_map_access is also not required. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-4-memxor@gmail.com
2022-04-25 05:48:51 +08:00
info->type = type;
info->off = off;
info->kptr.type_id = res_id;
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return BTF_FIELD_FOUND;
}
int btf_find_next_decl_tag(const struct btf *btf, const struct btf_type *pt,
int comp_idx, const char *tag_key, int last_id)
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
{
int len = strlen(tag_key);
int i, n;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
for (i = last_id + 1, n = btf_nr_types(btf); i < n; i++) {
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
const struct btf_type *t = btf_type_by_id(btf, i);
if (!btf_type_is_decl_tag(t))
continue;
if (pt != btf_type_by_id(btf, t->type))
continue;
if (btf_type_decl_tag(t)->component_idx != comp_idx)
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
continue;
if (strncmp(__btf_name_by_offset(btf, t->name_off), tag_key, len))
continue;
return i;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
}
return -ENOENT;
}
const char *btf_find_decl_tag_value(const struct btf *btf, const struct btf_type *pt,
int comp_idx, const char *tag_key)
{
const char *value = NULL;
const struct btf_type *t;
int len, id;
id = btf_find_next_decl_tag(btf, pt, comp_idx, tag_key, 0);
if (id < 0)
return ERR_PTR(id);
t = btf_type_by_id(btf, id);
len = strlen(tag_key);
value = __btf_name_by_offset(btf, t->name_off) + len;
/* Prevent duplicate entries for same type */
id = btf_find_next_decl_tag(btf, pt, comp_idx, tag_key, id);
if (id >= 0)
return ERR_PTR(-EEXIST);
bpf: Add support for custom exception callbacks By default, the subprog generated by the verifier to handle a thrown exception hardcodes a return value of 0. To allow user-defined logic and modification of the return value when an exception is thrown, introduce the 'exception_callback:' declaration tag, which marks a callback as the default exception handler for the program. The format of the declaration tag is 'exception_callback:<value>', where <value> is the name of the exception callback. Each main program can be tagged using this BTF declaratiion tag to associate it with an exception callback. In case the tag is absent, the default callback is used. As such, the exception callback cannot be modified at runtime, only set during verification. Allowing modification of the callback for the current program execution at runtime leads to issues when the programs begin to nest, as any per-CPU state maintaing this information will have to be saved and restored. We don't want it to stay in bpf_prog_aux as this takes a global effect for all programs. An alternative solution is spilling the callback pointer at a known location on the program stack on entry, and then passing this location to bpf_throw as a parameter. However, since exceptions are geared more towards a use case where they are ideally never invoked, optimizing for this use case and adding to the complexity has diminishing returns. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20230912233214.1518551-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-09-13 07:32:03 +08:00
return value;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
}
static int
btf_find_graph_root(const struct btf *btf, const struct btf_type *pt,
const struct btf_type *t, int comp_idx, u32 off,
int sz, struct btf_field_info *info,
enum btf_field_type head_type)
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
{
const char *node_field_name;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
const char *value_type;
s32 id;
if (!__btf_type_is_struct(t))
return BTF_FIELD_IGNORE;
if (t->size != sz)
return BTF_FIELD_IGNORE;
value_type = btf_find_decl_tag_value(btf, pt, comp_idx, "contains:");
bpf: Add support for custom exception callbacks By default, the subprog generated by the verifier to handle a thrown exception hardcodes a return value of 0. To allow user-defined logic and modification of the return value when an exception is thrown, introduce the 'exception_callback:' declaration tag, which marks a callback as the default exception handler for the program. The format of the declaration tag is 'exception_callback:<value>', where <value> is the name of the exception callback. Each main program can be tagged using this BTF declaratiion tag to associate it with an exception callback. In case the tag is absent, the default callback is used. As such, the exception callback cannot be modified at runtime, only set during verification. Allowing modification of the callback for the current program execution at runtime leads to issues when the programs begin to nest, as any per-CPU state maintaing this information will have to be saved and restored. We don't want it to stay in bpf_prog_aux as this takes a global effect for all programs. An alternative solution is spilling the callback pointer at a known location on the program stack on entry, and then passing this location to bpf_throw as a parameter. However, since exceptions are geared more towards a use case where they are ideally never invoked, optimizing for this use case and adding to the complexity has diminishing returns. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20230912233214.1518551-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-09-13 07:32:03 +08:00
if (IS_ERR(value_type))
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
return -EINVAL;
node_field_name = strstr(value_type, ":");
if (!node_field_name)
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
return -EINVAL;
value_type = kstrndup(value_type, node_field_name - value_type, GFP_KERNEL | __GFP_NOWARN);
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
if (!value_type)
return -ENOMEM;
id = btf_find_by_name_kind(btf, value_type, BTF_KIND_STRUCT);
kfree(value_type);
if (id < 0)
return id;
node_field_name++;
if (str_is_empty(node_field_name))
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
return -EINVAL;
info->type = head_type;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
info->off = off;
info->graph_root.value_btf_id = id;
info->graph_root.node_name = node_field_name;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
return BTF_FIELD_FOUND;
}
#define field_mask_test_name(field_type, field_type_str) \
if (field_mask & field_type && !strcmp(name, field_type_str)) { \
type = field_type; \
goto end; \
}
static int btf_get_field_type(const struct btf *btf, const struct btf_type *var_type,
u32 field_mask, u32 *seen_mask,
int *align, int *sz)
{
int type = 0;
const char *name = __btf_name_by_offset(btf, var_type->name_off);
if (field_mask & BPF_SPIN_LOCK) {
if (!strcmp(name, "bpf_spin_lock")) {
if (*seen_mask & BPF_SPIN_LOCK)
return -E2BIG;
*seen_mask |= BPF_SPIN_LOCK;
type = BPF_SPIN_LOCK;
goto end;
}
}
if (field_mask & BPF_TIMER) {
if (!strcmp(name, "bpf_timer")) {
if (*seen_mask & BPF_TIMER)
return -E2BIG;
*seen_mask |= BPF_TIMER;
type = BPF_TIMER;
goto end;
}
}
if (field_mask & BPF_WORKQUEUE) {
if (!strcmp(name, "bpf_wq")) {
if (*seen_mask & BPF_WORKQUEUE)
return -E2BIG;
*seen_mask |= BPF_WORKQUEUE;
type = BPF_WORKQUEUE;
goto end;
}
}
field_mask_test_name(BPF_LIST_HEAD, "bpf_list_head");
field_mask_test_name(BPF_LIST_NODE, "bpf_list_node");
field_mask_test_name(BPF_RB_ROOT, "bpf_rb_root");
field_mask_test_name(BPF_RB_NODE, "bpf_rb_node");
field_mask_test_name(BPF_REFCOUNT, "bpf_refcount");
/* Only return BPF_KPTR when all other types with matchable names fail */
if (field_mask & BPF_KPTR && !__btf_type_is_struct(var_type)) {
type = BPF_KPTR_REF;
goto end;
}
return 0;
end:
*sz = btf_field_type_size(type);
*align = btf_field_type_align(type);
return type;
}
#undef field_mask_test_name
/* Repeat a number of fields for a specified number of times.
*
* Copy the fields starting from the first field and repeat them for
* repeat_cnt times. The fields are repeated by adding the offset of each
* field with
* (i + 1) * elem_size
* where i is the repeat index and elem_size is the size of an element.
*/
static int btf_repeat_fields(struct btf_field_info *info, int info_cnt,
u32 field_cnt, u32 repeat_cnt, u32 elem_size)
{
u32 i, j;
u32 cur;
/* Ensure not repeating fields that should not be repeated. */
for (i = 0; i < field_cnt; i++) {
switch (info[i].type) {
case BPF_KPTR_UNREF:
case BPF_KPTR_REF:
case BPF_KPTR_PERCPU:
case BPF_LIST_HEAD:
case BPF_RB_ROOT:
break;
default:
return -EINVAL;
}
}
/* The type of struct size or variable size is u32,
* so the multiplication will not overflow.
*/
if (field_cnt * (repeat_cnt + 1) > info_cnt)
return -E2BIG;
cur = field_cnt;
for (i = 0; i < repeat_cnt; i++) {
memcpy(&info[cur], &info[0], field_cnt * sizeof(info[0]));
for (j = 0; j < field_cnt; j++)
info[cur++].off += (i + 1) * elem_size;
}
return 0;
}
static int btf_find_struct_field(const struct btf *btf,
const struct btf_type *t, u32 field_mask,
struct btf_field_info *info, int info_cnt,
u32 level);
/* Find special fields in the struct type of a field.
*
* This function is used to find fields of special types that is not a
* global variable or a direct field of a struct type. It also handles the
* repetition if it is the element type of an array.
*/
static int btf_find_nested_struct(const struct btf *btf, const struct btf_type *t,
u32 off, u32 nelems,
u32 field_mask, struct btf_field_info *info,
int info_cnt, u32 level)
{
int ret, err, i;
level++;
if (level >= MAX_RESOLVE_DEPTH)
return -E2BIG;
ret = btf_find_struct_field(btf, t, field_mask, info, info_cnt, level);
if (ret <= 0)
return ret;
/* Shift the offsets of the nested struct fields to the offsets
* related to the container.
*/
for (i = 0; i < ret; i++)
info[i].off += off;
if (nelems > 1) {
err = btf_repeat_fields(info, info_cnt, ret, nelems - 1, t->size);
if (err == 0)
ret *= nelems;
else
ret = err;
}
return ret;
}
static int btf_find_field_one(const struct btf *btf,
const struct btf_type *var,
const struct btf_type *var_type,
int var_idx,
u32 off, u32 expected_size,
u32 field_mask, u32 *seen_mask,
struct btf_field_info *info, int info_cnt,
u32 level)
{
int ret, align, sz, field_type;
struct btf_field_info tmp;
const struct btf_array *array;
u32 i, nelems = 1;
/* Walk into array types to find the element type and the number of
* elements in the (flattened) array.
*/
for (i = 0; i < MAX_RESOLVE_DEPTH && btf_type_is_array(var_type); i++) {
array = btf_array(var_type);
nelems *= array->nelems;
var_type = btf_type_by_id(btf, array->type);
}
if (i == MAX_RESOLVE_DEPTH)
return -E2BIG;
if (nelems == 0)
return 0;
field_type = btf_get_field_type(btf, var_type,
field_mask, seen_mask, &align, &sz);
/* Look into variables of struct types */
if (!field_type && __btf_type_is_struct(var_type)) {
sz = var_type->size;
if (expected_size && expected_size != sz * nelems)
return 0;
ret = btf_find_nested_struct(btf, var_type, off, nelems, field_mask,
&info[0], info_cnt, level);
return ret;
}
if (field_type == 0)
return 0;
if (field_type < 0)
return field_type;
if (expected_size && expected_size != sz * nelems)
return 0;
if (off % align)
return 0;
switch (field_type) {
case BPF_SPIN_LOCK:
case BPF_TIMER:
case BPF_WORKQUEUE:
case BPF_LIST_NODE:
case BPF_RB_NODE:
case BPF_REFCOUNT:
ret = btf_find_struct(btf, var_type, off, sz, field_type,
info_cnt ? &info[0] : &tmp);
if (ret < 0)
return ret;
break;
case BPF_KPTR_UNREF:
case BPF_KPTR_REF:
case BPF_KPTR_PERCPU:
ret = btf_find_kptr(btf, var_type, off, sz,
info_cnt ? &info[0] : &tmp);
if (ret < 0)
return ret;
break;
case BPF_LIST_HEAD:
case BPF_RB_ROOT:
ret = btf_find_graph_root(btf, var, var_type,
var_idx, off, sz,
info_cnt ? &info[0] : &tmp,
field_type);
if (ret < 0)
return ret;
break;
default:
return -EFAULT;
}
if (ret == BTF_FIELD_IGNORE)
return 0;
if (!info_cnt)
return -E2BIG;
if (nelems > 1) {
ret = btf_repeat_fields(info, info_cnt, 1, nelems - 1, sz);
if (ret < 0)
return ret;
}
return nelems;
}
static int btf_find_struct_field(const struct btf *btf,
const struct btf_type *t, u32 field_mask,
struct btf_field_info *info, int info_cnt,
u32 level)
bpf: introduce bpf_spin_lock Introduce 'struct bpf_spin_lock' and bpf_spin_lock/unlock() helpers to let bpf program serialize access to other variables. Example: struct hash_elem { int cnt; struct bpf_spin_lock lock; }; struct hash_elem * val = bpf_map_lookup_elem(&hash_map, &key); if (val) { bpf_spin_lock(&val->lock); val->cnt++; bpf_spin_unlock(&val->lock); } Restrictions and safety checks: - bpf_spin_lock is only allowed inside HASH and ARRAY maps. - BTF description of the map is mandatory for safety analysis. - bpf program can take one bpf_spin_lock at a time, since two or more can cause dead locks. - only one 'struct bpf_spin_lock' is allowed per map element. It drastically simplifies implementation yet allows bpf program to use any number of bpf_spin_locks. - when bpf_spin_lock is taken the calls (either bpf2bpf or helpers) are not allowed. - bpf program must bpf_spin_unlock() before return. - bpf program can access 'struct bpf_spin_lock' only via bpf_spin_lock()/bpf_spin_unlock() helpers. - load/store into 'struct bpf_spin_lock lock;' field is not allowed. - to use bpf_spin_lock() helper the BTF description of map value must be a struct and have 'struct bpf_spin_lock anyname;' field at the top level. Nested lock inside another struct is not allowed. - syscall map_lookup doesn't copy bpf_spin_lock field to user space. - syscall map_update and program map_update do not update bpf_spin_lock field. - bpf_spin_lock cannot be on the stack or inside networking packet. bpf_spin_lock can only be inside HASH or ARRAY map value. - bpf_spin_lock is available to root only and to all program types. - bpf_spin_lock is not allowed in inner maps of map-in-map. - ld_abs is not allowed inside spin_lock-ed region. - tracing progs and socket filter progs cannot use bpf_spin_lock due to insufficient preemption checks Implementation details: - cgroup-bpf class of programs can nest with xdp/tc programs. Hence bpf_spin_lock is equivalent to spin_lock_irqsave. Other solutions to avoid nested bpf_spin_lock are possible. Like making sure that all networking progs run with softirq disabled. spin_lock_irqsave is the simplest and doesn't add overhead to the programs that don't use it. - arch_spinlock_t is used when its implemented as queued_spin_lock - archs can force their own arch_spinlock_t - on architectures where queued_spin_lock is not available and sizeof(arch_spinlock_t) != sizeof(__u32) trivial lock is used. - presence of bpf_spin_lock inside map value could have been indicated via extra flag during map_create, but specifying it via BTF is cleaner. It provides introspection for map key/value and reduces user mistakes. Next steps: - allow bpf_spin_lock in other map types (like cgroup local storage) - introduce BPF_F_LOCK flag for bpf_map_update() syscall and helper to request kernel to grab bpf_spin_lock before rewriting the value. That will serialize access to map elements. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-02-01 07:40:04 +08:00
{
int ret, idx = 0;
bpf: introduce bpf_spin_lock Introduce 'struct bpf_spin_lock' and bpf_spin_lock/unlock() helpers to let bpf program serialize access to other variables. Example: struct hash_elem { int cnt; struct bpf_spin_lock lock; }; struct hash_elem * val = bpf_map_lookup_elem(&hash_map, &key); if (val) { bpf_spin_lock(&val->lock); val->cnt++; bpf_spin_unlock(&val->lock); } Restrictions and safety checks: - bpf_spin_lock is only allowed inside HASH and ARRAY maps. - BTF description of the map is mandatory for safety analysis. - bpf program can take one bpf_spin_lock at a time, since two or more can cause dead locks. - only one 'struct bpf_spin_lock' is allowed per map element. It drastically simplifies implementation yet allows bpf program to use any number of bpf_spin_locks. - when bpf_spin_lock is taken the calls (either bpf2bpf or helpers) are not allowed. - bpf program must bpf_spin_unlock() before return. - bpf program can access 'struct bpf_spin_lock' only via bpf_spin_lock()/bpf_spin_unlock() helpers. - load/store into 'struct bpf_spin_lock lock;' field is not allowed. - to use bpf_spin_lock() helper the BTF description of map value must be a struct and have 'struct bpf_spin_lock anyname;' field at the top level. Nested lock inside another struct is not allowed. - syscall map_lookup doesn't copy bpf_spin_lock field to user space. - syscall map_update and program map_update do not update bpf_spin_lock field. - bpf_spin_lock cannot be on the stack or inside networking packet. bpf_spin_lock can only be inside HASH or ARRAY map value. - bpf_spin_lock is available to root only and to all program types. - bpf_spin_lock is not allowed in inner maps of map-in-map. - ld_abs is not allowed inside spin_lock-ed region. - tracing progs and socket filter progs cannot use bpf_spin_lock due to insufficient preemption checks Implementation details: - cgroup-bpf class of programs can nest with xdp/tc programs. Hence bpf_spin_lock is equivalent to spin_lock_irqsave. Other solutions to avoid nested bpf_spin_lock are possible. Like making sure that all networking progs run with softirq disabled. spin_lock_irqsave is the simplest and doesn't add overhead to the programs that don't use it. - arch_spinlock_t is used when its implemented as queued_spin_lock - archs can force their own arch_spinlock_t - on architectures where queued_spin_lock is not available and sizeof(arch_spinlock_t) != sizeof(__u32) trivial lock is used. - presence of bpf_spin_lock inside map value could have been indicated via extra flag during map_create, but specifying it via BTF is cleaner. It provides introspection for map key/value and reduces user mistakes. Next steps: - allow bpf_spin_lock in other map types (like cgroup local storage) - introduce BPF_F_LOCK flag for bpf_map_update() syscall and helper to request kernel to grab bpf_spin_lock before rewriting the value. That will serialize access to map elements. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-02-01 07:40:04 +08:00
const struct btf_member *member;
u32 i, off, seen_mask = 0;
bpf: introduce bpf_spin_lock Introduce 'struct bpf_spin_lock' and bpf_spin_lock/unlock() helpers to let bpf program serialize access to other variables. Example: struct hash_elem { int cnt; struct bpf_spin_lock lock; }; struct hash_elem * val = bpf_map_lookup_elem(&hash_map, &key); if (val) { bpf_spin_lock(&val->lock); val->cnt++; bpf_spin_unlock(&val->lock); } Restrictions and safety checks: - bpf_spin_lock is only allowed inside HASH and ARRAY maps. - BTF description of the map is mandatory for safety analysis. - bpf program can take one bpf_spin_lock at a time, since two or more can cause dead locks. - only one 'struct bpf_spin_lock' is allowed per map element. It drastically simplifies implementation yet allows bpf program to use any number of bpf_spin_locks. - when bpf_spin_lock is taken the calls (either bpf2bpf or helpers) are not allowed. - bpf program must bpf_spin_unlock() before return. - bpf program can access 'struct bpf_spin_lock' only via bpf_spin_lock()/bpf_spin_unlock() helpers. - load/store into 'struct bpf_spin_lock lock;' field is not allowed. - to use bpf_spin_lock() helper the BTF description of map value must be a struct and have 'struct bpf_spin_lock anyname;' field at the top level. Nested lock inside another struct is not allowed. - syscall map_lookup doesn't copy bpf_spin_lock field to user space. - syscall map_update and program map_update do not update bpf_spin_lock field. - bpf_spin_lock cannot be on the stack or inside networking packet. bpf_spin_lock can only be inside HASH or ARRAY map value. - bpf_spin_lock is available to root only and to all program types. - bpf_spin_lock is not allowed in inner maps of map-in-map. - ld_abs is not allowed inside spin_lock-ed region. - tracing progs and socket filter progs cannot use bpf_spin_lock due to insufficient preemption checks Implementation details: - cgroup-bpf class of programs can nest with xdp/tc programs. Hence bpf_spin_lock is equivalent to spin_lock_irqsave. Other solutions to avoid nested bpf_spin_lock are possible. Like making sure that all networking progs run with softirq disabled. spin_lock_irqsave is the simplest and doesn't add overhead to the programs that don't use it. - arch_spinlock_t is used when its implemented as queued_spin_lock - archs can force their own arch_spinlock_t - on architectures where queued_spin_lock is not available and sizeof(arch_spinlock_t) != sizeof(__u32) trivial lock is used. - presence of bpf_spin_lock inside map value could have been indicated via extra flag during map_create, but specifying it via BTF is cleaner. It provides introspection for map key/value and reduces user mistakes. Next steps: - allow bpf_spin_lock in other map types (like cgroup local storage) - introduce BPF_F_LOCK flag for bpf_map_update() syscall and helper to request kernel to grab bpf_spin_lock before rewriting the value. That will serialize access to map elements. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-02-01 07:40:04 +08:00
for_each_member(i, t, member) {
const struct btf_type *member_type = btf_type_by_id(btf,
member->type);
off = __btf_member_bit_offset(t, member);
bpf: introduce bpf_spin_lock Introduce 'struct bpf_spin_lock' and bpf_spin_lock/unlock() helpers to let bpf program serialize access to other variables. Example: struct hash_elem { int cnt; struct bpf_spin_lock lock; }; struct hash_elem * val = bpf_map_lookup_elem(&hash_map, &key); if (val) { bpf_spin_lock(&val->lock); val->cnt++; bpf_spin_unlock(&val->lock); } Restrictions and safety checks: - bpf_spin_lock is only allowed inside HASH and ARRAY maps. - BTF description of the map is mandatory for safety analysis. - bpf program can take one bpf_spin_lock at a time, since two or more can cause dead locks. - only one 'struct bpf_spin_lock' is allowed per map element. It drastically simplifies implementation yet allows bpf program to use any number of bpf_spin_locks. - when bpf_spin_lock is taken the calls (either bpf2bpf or helpers) are not allowed. - bpf program must bpf_spin_unlock() before return. - bpf program can access 'struct bpf_spin_lock' only via bpf_spin_lock()/bpf_spin_unlock() helpers. - load/store into 'struct bpf_spin_lock lock;' field is not allowed. - to use bpf_spin_lock() helper the BTF description of map value must be a struct and have 'struct bpf_spin_lock anyname;' field at the top level. Nested lock inside another struct is not allowed. - syscall map_lookup doesn't copy bpf_spin_lock field to user space. - syscall map_update and program map_update do not update bpf_spin_lock field. - bpf_spin_lock cannot be on the stack or inside networking packet. bpf_spin_lock can only be inside HASH or ARRAY map value. - bpf_spin_lock is available to root only and to all program types. - bpf_spin_lock is not allowed in inner maps of map-in-map. - ld_abs is not allowed inside spin_lock-ed region. - tracing progs and socket filter progs cannot use bpf_spin_lock due to insufficient preemption checks Implementation details: - cgroup-bpf class of programs can nest with xdp/tc programs. Hence bpf_spin_lock is equivalent to spin_lock_irqsave. Other solutions to avoid nested bpf_spin_lock are possible. Like making sure that all networking progs run with softirq disabled. spin_lock_irqsave is the simplest and doesn't add overhead to the programs that don't use it. - arch_spinlock_t is used when its implemented as queued_spin_lock - archs can force their own arch_spinlock_t - on architectures where queued_spin_lock is not available and sizeof(arch_spinlock_t) != sizeof(__u32) trivial lock is used. - presence of bpf_spin_lock inside map value could have been indicated via extra flag during map_create, but specifying it via BTF is cleaner. It provides introspection for map key/value and reduces user mistakes. Next steps: - allow bpf_spin_lock in other map types (like cgroup local storage) - introduce BPF_F_LOCK flag for bpf_map_update() syscall and helper to request kernel to grab bpf_spin_lock before rewriting the value. That will serialize access to map elements. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-02-01 07:40:04 +08:00
if (off % 8)
/* valid C code cannot generate such BTF */
return -EINVAL;
off /= 8;
ret = btf_find_field_one(btf, t, member_type, i,
off, 0,
field_mask, &seen_mask,
&info[idx], info_cnt - idx, level);
if (ret < 0)
return ret;
idx += ret;
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
}
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return idx;
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
}
static int btf_find_datasec_var(const struct btf *btf, const struct btf_type *t,
u32 field_mask, struct btf_field_info *info,
int info_cnt, u32 level)
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
{
int ret, idx = 0;
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
const struct btf_var_secinfo *vsi;
u32 i, off, seen_mask = 0;
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
for_each_vsi(i, t, vsi) {
const struct btf_type *var = btf_type_by_id(btf, vsi->type);
const struct btf_type *var_type = btf_type_by_id(btf, var->type);
off = vsi->offset;
ret = btf_find_field_one(btf, var, var_type, -1, off, vsi->size,
field_mask, &seen_mask,
&info[idx], info_cnt - idx,
level);
if (ret < 0)
return ret;
idx += ret;
bpf: introduce bpf_spin_lock Introduce 'struct bpf_spin_lock' and bpf_spin_lock/unlock() helpers to let bpf program serialize access to other variables. Example: struct hash_elem { int cnt; struct bpf_spin_lock lock; }; struct hash_elem * val = bpf_map_lookup_elem(&hash_map, &key); if (val) { bpf_spin_lock(&val->lock); val->cnt++; bpf_spin_unlock(&val->lock); } Restrictions and safety checks: - bpf_spin_lock is only allowed inside HASH and ARRAY maps. - BTF description of the map is mandatory for safety analysis. - bpf program can take one bpf_spin_lock at a time, since two or more can cause dead locks. - only one 'struct bpf_spin_lock' is allowed per map element. It drastically simplifies implementation yet allows bpf program to use any number of bpf_spin_locks. - when bpf_spin_lock is taken the calls (either bpf2bpf or helpers) are not allowed. - bpf program must bpf_spin_unlock() before return. - bpf program can access 'struct bpf_spin_lock' only via bpf_spin_lock()/bpf_spin_unlock() helpers. - load/store into 'struct bpf_spin_lock lock;' field is not allowed. - to use bpf_spin_lock() helper the BTF description of map value must be a struct and have 'struct bpf_spin_lock anyname;' field at the top level. Nested lock inside another struct is not allowed. - syscall map_lookup doesn't copy bpf_spin_lock field to user space. - syscall map_update and program map_update do not update bpf_spin_lock field. - bpf_spin_lock cannot be on the stack or inside networking packet. bpf_spin_lock can only be inside HASH or ARRAY map value. - bpf_spin_lock is available to root only and to all program types. - bpf_spin_lock is not allowed in inner maps of map-in-map. - ld_abs is not allowed inside spin_lock-ed region. - tracing progs and socket filter progs cannot use bpf_spin_lock due to insufficient preemption checks Implementation details: - cgroup-bpf class of programs can nest with xdp/tc programs. Hence bpf_spin_lock is equivalent to spin_lock_irqsave. Other solutions to avoid nested bpf_spin_lock are possible. Like making sure that all networking progs run with softirq disabled. spin_lock_irqsave is the simplest and doesn't add overhead to the programs that don't use it. - arch_spinlock_t is used when its implemented as queued_spin_lock - archs can force their own arch_spinlock_t - on architectures where queued_spin_lock is not available and sizeof(arch_spinlock_t) != sizeof(__u32) trivial lock is used. - presence of bpf_spin_lock inside map value could have been indicated via extra flag during map_create, but specifying it via BTF is cleaner. It provides introspection for map key/value and reduces user mistakes. Next steps: - allow bpf_spin_lock in other map types (like cgroup local storage) - introduce BPF_F_LOCK flag for bpf_map_update() syscall and helper to request kernel to grab bpf_spin_lock before rewriting the value. That will serialize access to map elements. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-02-01 07:40:04 +08:00
}
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return idx;
bpf: introduce bpf_spin_lock Introduce 'struct bpf_spin_lock' and bpf_spin_lock/unlock() helpers to let bpf program serialize access to other variables. Example: struct hash_elem { int cnt; struct bpf_spin_lock lock; }; struct hash_elem * val = bpf_map_lookup_elem(&hash_map, &key); if (val) { bpf_spin_lock(&val->lock); val->cnt++; bpf_spin_unlock(&val->lock); } Restrictions and safety checks: - bpf_spin_lock is only allowed inside HASH and ARRAY maps. - BTF description of the map is mandatory for safety analysis. - bpf program can take one bpf_spin_lock at a time, since two or more can cause dead locks. - only one 'struct bpf_spin_lock' is allowed per map element. It drastically simplifies implementation yet allows bpf program to use any number of bpf_spin_locks. - when bpf_spin_lock is taken the calls (either bpf2bpf or helpers) are not allowed. - bpf program must bpf_spin_unlock() before return. - bpf program can access 'struct bpf_spin_lock' only via bpf_spin_lock()/bpf_spin_unlock() helpers. - load/store into 'struct bpf_spin_lock lock;' field is not allowed. - to use bpf_spin_lock() helper the BTF description of map value must be a struct and have 'struct bpf_spin_lock anyname;' field at the top level. Nested lock inside another struct is not allowed. - syscall map_lookup doesn't copy bpf_spin_lock field to user space. - syscall map_update and program map_update do not update bpf_spin_lock field. - bpf_spin_lock cannot be on the stack or inside networking packet. bpf_spin_lock can only be inside HASH or ARRAY map value. - bpf_spin_lock is available to root only and to all program types. - bpf_spin_lock is not allowed in inner maps of map-in-map. - ld_abs is not allowed inside spin_lock-ed region. - tracing progs and socket filter progs cannot use bpf_spin_lock due to insufficient preemption checks Implementation details: - cgroup-bpf class of programs can nest with xdp/tc programs. Hence bpf_spin_lock is equivalent to spin_lock_irqsave. Other solutions to avoid nested bpf_spin_lock are possible. Like making sure that all networking progs run with softirq disabled. spin_lock_irqsave is the simplest and doesn't add overhead to the programs that don't use it. - arch_spinlock_t is used when its implemented as queued_spin_lock - archs can force their own arch_spinlock_t - on architectures where queued_spin_lock is not available and sizeof(arch_spinlock_t) != sizeof(__u32) trivial lock is used. - presence of bpf_spin_lock inside map value could have been indicated via extra flag during map_create, but specifying it via BTF is cleaner. It provides introspection for map key/value and reduces user mistakes. Next steps: - allow bpf_spin_lock in other map types (like cgroup local storage) - introduce BPF_F_LOCK flag for bpf_map_update() syscall and helper to request kernel to grab bpf_spin_lock before rewriting the value. That will serialize access to map elements. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2019-02-01 07:40:04 +08:00
}
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
static int btf_find_field(const struct btf *btf, const struct btf_type *t,
u32 field_mask, struct btf_field_info *info,
int info_cnt)
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
{
if (__btf_type_is_struct(t))
return btf_find_struct_field(btf, t, field_mask, info, info_cnt, 0);
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
else if (btf_type_is_datasec(t))
return btf_find_datasec_var(btf, t, field_mask, info, info_cnt, 0);
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
return -EINVAL;
}
/* Callers have to ensure the life cycle of btf if it is program BTF */
static int btf_parse_kptr(const struct btf *btf, struct btf_field *field,
struct btf_field_info *info)
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
{
struct module *mod = NULL;
const struct btf_type *t;
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
/* If a matching btf type is found in kernel or module BTFs, kptr_ref
* is that BTF, otherwise it's program BTF
*/
struct btf *kptr_btf;
int ret;
s32 id;
/* Find type in map BTF, and use it to look up the matching type
* in vmlinux or module BTFs, by name and kind.
*/
t = btf_type_by_id(btf, info->kptr.type_id);
id = bpf_find_btf_id(__btf_name_by_offset(btf, t->name_off), BTF_INFO_KIND(t->info),
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
&kptr_btf);
if (id == -ENOENT) {
/* btf_parse_kptr should only be called w/ btf = program BTF */
WARN_ON_ONCE(btf_is_kernel(btf));
/* Type exists only in program BTF. Assume that it's a MEM_ALLOC
* kptr allocated via bpf_obj_new
*/
bpf: Disable migration when freeing stashed local kptr using obj drop When a local kptr is stashed in a map and freed when the map goes away, currently an error like the below appears: [ 39.195695] BUG: using smp_processor_id() in preemptible [00000000] code: kworker/u32:15/2875 [ 39.196549] caller is bpf_mem_free+0x56/0xc0 [ 39.196958] CPU: 15 PID: 2875 Comm: kworker/u32:15 Tainted: G O 6.2.0-13016-g22df776a9a86 #4477 [ 39.197897] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014 [ 39.198949] Workqueue: events_unbound bpf_map_free_deferred [ 39.199470] Call Trace: [ 39.199703] <TASK> [ 39.199911] dump_stack_lvl+0x60/0x70 [ 39.200267] check_preemption_disabled+0xbf/0xe0 [ 39.200704] bpf_mem_free+0x56/0xc0 [ 39.201032] ? bpf_obj_new_impl+0xa0/0xa0 [ 39.201430] bpf_obj_free_fields+0x1cd/0x200 [ 39.201838] array_map_free+0xad/0x220 [ 39.202193] ? finish_task_switch+0xe5/0x3c0 [ 39.202614] bpf_map_free_deferred+0xea/0x210 [ 39.203006] ? lockdep_hardirqs_on_prepare+0xe/0x220 [ 39.203460] process_one_work+0x64f/0xbe0 [ 39.203822] ? pwq_dec_nr_in_flight+0x110/0x110 [ 39.204264] ? do_raw_spin_lock+0x107/0x1c0 [ 39.204662] ? lockdep_hardirqs_on_prepare+0xe/0x220 [ 39.205107] worker_thread+0x74/0x7a0 [ 39.205451] ? process_one_work+0xbe0/0xbe0 [ 39.205818] kthread+0x171/0x1a0 [ 39.206111] ? kthread_complete_and_exit+0x20/0x20 [ 39.206552] ret_from_fork+0x1f/0x30 [ 39.206886] </TASK> This happens because the call to __bpf_obj_drop_impl I added in the patch adding support for stashing local kptrs doesn't disable migration. Prior to that patch, __bpf_obj_drop_impl logic only ran when called by a BPF progarm, whereas now it can be called from map free path, so it's necessary to explicitly disable migration. Also, refactor a bit to just call __bpf_obj_drop_impl directly instead of bothering w/ dtor union and setting pointer-to-obj_drop. Fixes: c8e187540914 ("bpf: Support __kptr to local kptrs") Reported-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230313214641.3731908-1-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-14 05:46:41 +08:00
field->kptr.dtor = NULL;
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
id = info->kptr.type_id;
kptr_btf = (struct btf *)btf;
goto found_dtor;
}
if (id < 0)
return id;
/* Find and stash the function pointer for the destruction function that
* needs to be eventually invoked from the map free path.
*/
if (info->type == BPF_KPTR_REF) {
const struct btf_type *dtor_func;
const char *dtor_func_name;
unsigned long addr;
s32 dtor_btf_id;
/* This call also serves as a whitelist of allowed objects that
* can be used as a referenced pointer and be stored in a map at
* the same time.
*/
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
dtor_btf_id = btf_find_dtor_kfunc(kptr_btf, id);
if (dtor_btf_id < 0) {
ret = dtor_btf_id;
goto end_btf;
}
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
dtor_func = btf_type_by_id(kptr_btf, dtor_btf_id);
if (!dtor_func) {
ret = -ENOENT;
goto end_btf;
}
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
if (btf_is_module(kptr_btf)) {
mod = btf_try_get_module(kptr_btf);
if (!mod) {
ret = -ENXIO;
goto end_btf;
}
}
/* We already verified dtor_func to be btf_type_is_func
* in register_btf_id_dtor_kfuncs.
*/
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
dtor_func_name = __btf_name_by_offset(kptr_btf, dtor_func->name_off);
addr = kallsyms_lookup_name(dtor_func_name);
if (!addr) {
ret = -EINVAL;
goto end_mod;
}
field->kptr.dtor = (void *)addr;
}
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
found_dtor:
field->kptr.btf_id = id;
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
field->kptr.btf = kptr_btf;
field->kptr.module = mod;
return 0;
end_mod:
module_put(mod);
end_btf:
bpf: Support __kptr to local kptrs If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module BTF - it must have been allocated by bpf_obj_new and therefore must be free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new. This patch adds support for treating __kptr-tagged pointers to "local kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr acquire / release semantics. Consider the following example: struct node_data { long key; long data; struct bpf_rb_node node; }; struct map_value { struct node_data __kptr *node; }; struct { __uint(type, BPF_MAP_TYPE_ARRAY); __type(key, int); __type(value, struct map_value); __uint(max_entries, 1); } some_nodes SEC(".maps"); If struct node_data had a matching definition in kernel BTF, the verifier would expect a destructor for the type to be registered. Since struct node_data does not match any type in kernel BTF, the verifier knows that there is no kfunc that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can only come from bpf_obj_new. So instead of searching for a registered dtor, a bpf_obj_drop dtor can be assumed. This allows the runtime to properly destruct such kptrs in bpf_obj_free_fields, which enables maps to clean up map_vals w/ such kptrs when going away. Implementation notes: * "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr. Before this patch, the variable would only ever point to vmlinux or module BTFs, but now it can point to some program BTF for local kptr type. It's later used to populate the (btf, btf_id) pair in kptr btf field. * It's necessary to btf_get the program BTF when populating btf_field for local kptr. btf_record_free later does a btf_put. * Behavior for non-local referenced kptrs is not modified, as bpf_find_btf_id helper only searches vmlinux and module BTFs for matching BTF type. If such a type is found, btf_field_kptr's btf will pass btf_is_kernel check, and the associated release function is some one-argument dtor. If btf_is_kernel check fails, associated release function is two-arg bpf_obj_drop_impl. Before this patch only btf_field_kptr's w/ kernel or module BTFs were created. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-11 07:07:41 +08:00
btf_put(kptr_btf);
return ret;
bpf: Add map side support for bpf timers. Restrict bpf timers to array, hash (both preallocated and kmalloced), and lru map types. The per-cpu maps with timers don't make sense, since 'struct bpf_timer' is a part of map value. bpf timers in per-cpu maps would mean that the number of timers depends on number of possible cpus and timers would not be accessible from all cpus. lpm map support can be added in the future. The timers in inner maps are supported. The bpf_map_update/delete_elem() helpers and sys_bpf commands cancel and free bpf_timer in a given map element. Similar to 'struct bpf_spin_lock' BTF is required and it is used to validate that map element indeed contains 'struct bpf_timer'. Make check_and_init_map_value() init both bpf_spin_lock and bpf_timer when map element data is reused in preallocated htab and lru maps. Teach copy_map_value() to support both bpf_spin_lock and bpf_timer in a single map element. There could be one of each, but not more than one. Due to 'one bpf_timer in one element' restriction do not support timers in global data, since global data is a map of single element, but from bpf program side it's seen as many global variables and restriction of single global timer would be odd. The sys_bpf map_freeze and sys_mmap syscalls are not allowed on maps with timers, since user space could have corrupted mmap element and crashed the kernel. The maps with timers cannot be readonly. Due to these restrictions search for bpf_timer in datasec BTF in case it was placed in the global data to report clear error. The previous patch allowed 'struct bpf_timer' as a first field in a map element only. Relax this restriction. Refactor lru map to s/bpf_lru_push_free/htab_lru_push_free/ to cancel and free the timer when lru map deletes an element as a part of it eviction algorithm. Make sure that bpf program cannot access 'struct bpf_timer' via direct load/store. The timer operation are done through helpers only. This is similar to 'struct bpf_spin_lock'. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20210715005417.78572-5-alexei.starovoitov@gmail.com
2021-07-15 08:54:10 +08:00
}
static int btf_parse_graph_root(const struct btf *btf,
struct btf_field *field,
struct btf_field_info *info,
const char *node_type_name,
size_t node_type_align)
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
{
const struct btf_type *t, *n = NULL;
const struct btf_member *member;
u32 offset;
int i;
t = btf_type_by_id(btf, info->graph_root.value_btf_id);
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
/* We've already checked that value_btf_id is a struct type. We
* just need to figure out the offset of the list_node, and
* verify its type.
*/
for_each_member(i, t, member) {
if (strcmp(info->graph_root.node_name,
__btf_name_by_offset(btf, member->name_off)))
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
continue;
/* Invalid BTF, two members with same name */
if (n)
return -EINVAL;
n = btf_type_by_id(btf, member->type);
if (!__btf_type_is_struct(n))
return -EINVAL;
if (strcmp(node_type_name, __btf_name_by_offset(btf, n->name_off)))
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
return -EINVAL;
offset = __btf_member_bit_offset(n, member);
if (offset % 8)
return -EINVAL;
offset /= 8;
if (offset % node_type_align)
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
return -EINVAL;
field->graph_root.btf = (struct btf *)btf;
field->graph_root.value_btf_id = info->graph_root.value_btf_id;
field->graph_root.node_offset = offset;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
}
if (!n)
return -ENOENT;
return 0;
}
static int btf_parse_list_head(const struct btf *btf, struct btf_field *field,
struct btf_field_info *info)
{
return btf_parse_graph_root(btf, field, info, "bpf_list_node",
__alignof__(struct bpf_list_node));
}
static int btf_parse_rb_root(const struct btf *btf, struct btf_field *field,
struct btf_field_info *info)
{
return btf_parse_graph_root(btf, field, info, "bpf_rb_node",
__alignof__(struct bpf_rb_node));
}
bpf: Remove btf_field_offs, use btf_record's fields instead The btf_field_offs struct contains (offset, size) for btf_record fields, sorted by offset. btf_field_offs is always used in conjunction with btf_record, which has btf_field 'fields' array with (offset, type), the latter of which btf_field_offs' size is derived from via btf_field_type_size. This patch adds a size field to struct btf_field and sorts btf_record's fields by offset, making it possible to get rid of btf_field_offs. Less data duplication and less code complexity results. Since btf_field_offs' lifetime closely followed the btf_record used to populate it, most complexity wins are from removal of initialization code like: if (btf_record_successfully_initialized) { foffs = btf_parse_field_offs(rec); if (IS_ERR_OR_NULL(foffs)) // free the btf_record and return err } Other changes in this patch are pretty mechanical: * foffs->field_off[i] -> rec->fields[i].offset * foffs->field_sz[i] -> rec->fields[i].size * Sort rec->fields in btf_parse_fields before returning * It's possible that this is necessary independently of other changes in this patch. btf_record_find in syscall.c expects btf_record's fields to be sorted by offset, yet there's no explicit sorting of them before this patch, record's fields are populated in the order they're read from BTF struct definition. BTF docs don't say anything about the sortedness of struct fields. * All functions taking struct btf_field_offs * input now instead take struct btf_record *. All callsites of these functions already have access to the correct btf_record. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230415201811.343116-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-16 04:18:03 +08:00
static int btf_field_cmp(const void *_a, const void *_b, const void *priv)
{
const struct btf_field *a = (const struct btf_field *)_a;
const struct btf_field *b = (const struct btf_field *)_b;
if (a->offset < b->offset)
return -1;
else if (a->offset > b->offset)
return 1;
return 0;
}
struct btf_record *btf_parse_fields(const struct btf *btf, const struct btf_type *t,
u32 field_mask, u32 value_size)
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
{
struct btf_field_info info_arr[BTF_FIELDS_MAX];
bpf: Remove btf_field_offs, use btf_record's fields instead The btf_field_offs struct contains (offset, size) for btf_record fields, sorted by offset. btf_field_offs is always used in conjunction with btf_record, which has btf_field 'fields' array with (offset, type), the latter of which btf_field_offs' size is derived from via btf_field_type_size. This patch adds a size field to struct btf_field and sorts btf_record's fields by offset, making it possible to get rid of btf_field_offs. Less data duplication and less code complexity results. Since btf_field_offs' lifetime closely followed the btf_record used to populate it, most complexity wins are from removal of initialization code like: if (btf_record_successfully_initialized) { foffs = btf_parse_field_offs(rec); if (IS_ERR_OR_NULL(foffs)) // free the btf_record and return err } Other changes in this patch are pretty mechanical: * foffs->field_off[i] -> rec->fields[i].offset * foffs->field_sz[i] -> rec->fields[i].size * Sort rec->fields in btf_parse_fields before returning * It's possible that this is necessary independently of other changes in this patch. btf_record_find in syscall.c expects btf_record's fields to be sorted by offset, yet there's no explicit sorting of them before this patch, record's fields are populated in the order they're read from BTF struct definition. BTF docs don't say anything about the sortedness of struct fields. * All functions taking struct btf_field_offs * input now instead take struct btf_record *. All callsites of these functions already have access to the correct btf_record. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230415201811.343116-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-16 04:18:03 +08:00
u32 next_off = 0, field_type_size;
struct btf_record *rec;
int ret, i, cnt;
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
ret = btf_find_field(btf, t, field_mask, info_arr, ARRAY_SIZE(info_arr));
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
if (ret < 0)
return ERR_PTR(ret);
if (!ret)
return NULL;
cnt = ret;
2022-11-18 09:56:08 +08:00
/* This needs to be kzalloc to zero out padding and unused fields, see
* comment in btf_record_equal.
*/
rec = kzalloc(offsetof(struct btf_record, fields[cnt]), GFP_KERNEL | __GFP_NOWARN);
if (!rec)
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return ERR_PTR(-ENOMEM);
rec->spin_lock_off = -EINVAL;
rec->timer_off = -EINVAL;
rec->wq_off = -EINVAL;
rec->refcount_off = -EINVAL;
for (i = 0; i < cnt; i++) {
bpf: Remove btf_field_offs, use btf_record's fields instead The btf_field_offs struct contains (offset, size) for btf_record fields, sorted by offset. btf_field_offs is always used in conjunction with btf_record, which has btf_field 'fields' array with (offset, type), the latter of which btf_field_offs' size is derived from via btf_field_type_size. This patch adds a size field to struct btf_field and sorts btf_record's fields by offset, making it possible to get rid of btf_field_offs. Less data duplication and less code complexity results. Since btf_field_offs' lifetime closely followed the btf_record used to populate it, most complexity wins are from removal of initialization code like: if (btf_record_successfully_initialized) { foffs = btf_parse_field_offs(rec); if (IS_ERR_OR_NULL(foffs)) // free the btf_record and return err } Other changes in this patch are pretty mechanical: * foffs->field_off[i] -> rec->fields[i].offset * foffs->field_sz[i] -> rec->fields[i].size * Sort rec->fields in btf_parse_fields before returning * It's possible that this is necessary independently of other changes in this patch. btf_record_find in syscall.c expects btf_record's fields to be sorted by offset, yet there's no explicit sorting of them before this patch, record's fields are populated in the order they're read from BTF struct definition. BTF docs don't say anything about the sortedness of struct fields. * All functions taking struct btf_field_offs * input now instead take struct btf_record *. All callsites of these functions already have access to the correct btf_record. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230415201811.343116-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-16 04:18:03 +08:00
field_type_size = btf_field_type_size(info_arr[i].type);
if (info_arr[i].off + field_type_size > value_size) {
WARN_ONCE(1, "verifier bug off %d size %d", info_arr[i].off, value_size);
ret = -EFAULT;
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
goto end;
}
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
if (info_arr[i].off < next_off) {
ret = -EEXIST;
goto end;
}
bpf: Remove btf_field_offs, use btf_record's fields instead The btf_field_offs struct contains (offset, size) for btf_record fields, sorted by offset. btf_field_offs is always used in conjunction with btf_record, which has btf_field 'fields' array with (offset, type), the latter of which btf_field_offs' size is derived from via btf_field_type_size. This patch adds a size field to struct btf_field and sorts btf_record's fields by offset, making it possible to get rid of btf_field_offs. Less data duplication and less code complexity results. Since btf_field_offs' lifetime closely followed the btf_record used to populate it, most complexity wins are from removal of initialization code like: if (btf_record_successfully_initialized) { foffs = btf_parse_field_offs(rec); if (IS_ERR_OR_NULL(foffs)) // free the btf_record and return err } Other changes in this patch are pretty mechanical: * foffs->field_off[i] -> rec->fields[i].offset * foffs->field_sz[i] -> rec->fields[i].size * Sort rec->fields in btf_parse_fields before returning * It's possible that this is necessary independently of other changes in this patch. btf_record_find in syscall.c expects btf_record's fields to be sorted by offset, yet there's no explicit sorting of them before this patch, record's fields are populated in the order they're read from BTF struct definition. BTF docs don't say anything about the sortedness of struct fields. * All functions taking struct btf_field_offs * input now instead take struct btf_record *. All callsites of these functions already have access to the correct btf_record. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230415201811.343116-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-16 04:18:03 +08:00
next_off = info_arr[i].off + field_type_size;
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
rec->field_mask |= info_arr[i].type;
rec->fields[i].offset = info_arr[i].off;
rec->fields[i].type = info_arr[i].type;
bpf: Remove btf_field_offs, use btf_record's fields instead The btf_field_offs struct contains (offset, size) for btf_record fields, sorted by offset. btf_field_offs is always used in conjunction with btf_record, which has btf_field 'fields' array with (offset, type), the latter of which btf_field_offs' size is derived from via btf_field_type_size. This patch adds a size field to struct btf_field and sorts btf_record's fields by offset, making it possible to get rid of btf_field_offs. Less data duplication and less code complexity results. Since btf_field_offs' lifetime closely followed the btf_record used to populate it, most complexity wins are from removal of initialization code like: if (btf_record_successfully_initialized) { foffs = btf_parse_field_offs(rec); if (IS_ERR_OR_NULL(foffs)) // free the btf_record and return err } Other changes in this patch are pretty mechanical: * foffs->field_off[i] -> rec->fields[i].offset * foffs->field_sz[i] -> rec->fields[i].size * Sort rec->fields in btf_parse_fields before returning * It's possible that this is necessary independently of other changes in this patch. btf_record_find in syscall.c expects btf_record's fields to be sorted by offset, yet there's no explicit sorting of them before this patch, record's fields are populated in the order they're read from BTF struct definition. BTF docs don't say anything about the sortedness of struct fields. * All functions taking struct btf_field_offs * input now instead take struct btf_record *. All callsites of these functions already have access to the correct btf_record. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230415201811.343116-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-16 04:18:03 +08:00
rec->fields[i].size = field_type_size;
switch (info_arr[i].type) {
case BPF_SPIN_LOCK:
WARN_ON_ONCE(rec->spin_lock_off >= 0);
/* Cache offset for faster lookup at runtime */
rec->spin_lock_off = rec->fields[i].offset;
break;
case BPF_TIMER:
WARN_ON_ONCE(rec->timer_off >= 0);
/* Cache offset for faster lookup at runtime */
rec->timer_off = rec->fields[i].offset;
break;
case BPF_WORKQUEUE:
WARN_ON_ONCE(rec->wq_off >= 0);
/* Cache offset for faster lookup at runtime */
rec->wq_off = rec->fields[i].offset;
break;
case BPF_REFCOUNT:
WARN_ON_ONCE(rec->refcount_off >= 0);
/* Cache offset for faster lookup at runtime */
rec->refcount_off = rec->fields[i].offset;
break;
case BPF_KPTR_UNREF:
case BPF_KPTR_REF:
case BPF_KPTR_PERCPU:
ret = btf_parse_kptr(btf, &rec->fields[i], &info_arr[i]);
if (ret < 0)
goto end;
break;
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
case BPF_LIST_HEAD:
ret = btf_parse_list_head(btf, &rec->fields[i], &info_arr[i]);
if (ret < 0)
goto end;
break;
case BPF_RB_ROOT:
ret = btf_parse_rb_root(btf, &rec->fields[i], &info_arr[i]);
if (ret < 0)
goto end;
break;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
case BPF_LIST_NODE:
case BPF_RB_NODE:
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
break;
default:
ret = -EFAULT;
goto end;
}
rec->cnt++;
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
}
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
/* bpf_{list_head, rb_node} require bpf_spin_lock */
if ((btf_record_has_field(rec, BPF_LIST_HEAD) ||
btf_record_has_field(rec, BPF_RB_ROOT)) && rec->spin_lock_off < 0) {
bpf: Support bpf_list_head in map values Add the support on the map side to parse, recognize, verify, and build metadata table for a new special field of the type struct bpf_list_head. To parameterize the bpf_list_head for a certain value type and the list_node member it will accept in that value type, we use BTF declaration tags. The definition of bpf_list_head in a map value will be done as follows: struct foo { struct bpf_list_node node; int data; }; struct map_value { struct bpf_list_head head __contains(foo, node); }; Then, the bpf_list_head only allows adding to the list 'head' using the bpf_list_node 'node' for the type struct foo. The 'contains' annotation is a BTF declaration tag composed of four parts, "contains:name:node" where the name is then used to look up the type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during the lookup. The node defines name of the member in this type that has the type struct bpf_list_node, which is actually used for linking into the linked list. For now, 'kind' part is hardcoded as struct. This allows building intrusive linked lists in BPF, using container_of to obtain pointer to entry, while being completely type safe from the perspective of the verifier. The verifier knows exactly the type of the nodes, and knows that list helpers return that type at some fixed offset where the bpf_list_node member used for this list exists. The verifier also uses this information to disallow adding types that are not accepted by a certain list. For now, no elements can be added to such lists. Support for that is coming in future patches, hence draining and freeing items is done with a TODO that will be resolved in a future patch. Note that the bpf_list_head_free function moves the list out to a local variable under the lock and releases it, doing the actual draining of the list items outside the lock. While this helps with not holding the lock for too long pessimizing other concurrent list operations, it is also necessary for deadlock prevention: unless every function called in the critical section would be notrace, a fentry/fexit program could attach and call bpf_map_update_elem again on the map, leading to the same lock being acquired if the key matches and lead to a deadlock. While this requires some special effort on part of the BPF programmer to trigger and is highly unlikely to occur in practice, it is always better if we can avoid such a condition. While notrace would prevent this, doing the draining outside the lock has advantages of its own, hence it is used to also fix the deadlock related problem. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-15 03:15:25 +08:00
ret = -EINVAL;
goto end;
}
bpf: Migrate bpf_rbtree_remove to possibly fail This patch modifies bpf_rbtree_remove to account for possible failure due to the input rb_node already not being in any collection. The function can now return NULL, and does when the aforementioned scenario occurs. As before, on successful removal an owning reference to the removed node is returned. Adding KF_RET_NULL to bpf_rbtree_remove's kfunc flags - now KF_RET_NULL | KF_ACQUIRE - provides the desired verifier semantics: * retval must be checked for NULL before use * if NULL, retval's ref_obj_id is released * retval is a "maybe acquired" owning ref, not a non-owning ref, so it will live past end of critical section (bpf_spin_unlock), and thus can be checked for NULL after the end of the CS BPF programs must add checks ============================ This does change bpf_rbtree_remove's verifier behavior. BPF program writers will need to add NULL checks to their programs, but the resulting UX looks natural: bpf_spin_lock(&glock); n = bpf_rbtree_first(&ghead); if (!n) { /* ... */} res = bpf_rbtree_remove(&ghead, &n->node); bpf_spin_unlock(&glock); if (!res) /* Newly-added check after this patch */ return 1; n = container_of(res, /* ... */); /* Do something else with n */ bpf_obj_drop(n); return 0; The "if (!res)" check above is the only addition necessary for the above program to pass verification after this patch. bpf_rbtree_remove no longer clobbers non-owning refs ==================================================== An issue arises when bpf_rbtree_remove fails, though. Consider this example: struct node_data { long key; struct bpf_list_node l; struct bpf_rb_node r; struct bpf_refcount ref; }; long failed_sum; void bpf_prog() { struct node_data *n = bpf_obj_new(/* ... */); struct bpf_rb_node *res; n->key = 10; bpf_spin_lock(&glock); bpf_list_push_back(&some_list, &n->l); /* n is now a non-owning ref */ res = bpf_rbtree_remove(&some_tree, &n->r, /* ... */); if (!res) failed_sum += n->key; /* not possible */ bpf_spin_unlock(&glock); /* if (res) { do something useful and drop } ... */ } The bpf_rbtree_remove in this example will always fail. Similarly to bpf_spin_unlock, bpf_rbtree_remove is a non-owning reference invalidation point. The verifier clobbers all non-owning refs after a bpf_rbtree_remove call, so the "failed_sum += n->key" line will fail verification, and in fact there's no good way to get information about the node which failed to add after the invalidation. This patch removes non-owning reference invalidation from bpf_rbtree_remove to allow the above usecase to pass verification. The logic for why this is now possible is as follows: Before this series, bpf_rbtree_add couldn't fail and thus assumed that its input, a non-owning reference, was in the tree. But it's easy to construct an example where two non-owning references pointing to the same underlying memory are acquired and passed to rbtree_remove one after another (see rbtree_api_release_aliasing in selftests/bpf/progs/rbtree_fail.c). So it was necessary to clobber non-owning refs to prevent this case and, more generally, to enforce "non-owning ref is definitely in some collection" invariant. This series removes that invariant and the failure / runtime checking added in this patch provide a clean way to deal with the aliasing issue - just fail to remove. Because the aliasing issue prevented by clobbering non-owning refs is no longer an issue, this patch removes the invalidate_non_owning_refs call from verifier handling of bpf_rbtree_remove. Note that bpf_spin_unlock - the other caller of invalidate_non_owning_refs - clobbers non-owning refs for a different reason, so its clobbering behavior remains unchanged. No BPF program changes are necessary for programs to remain valid as a result of this clobbering change. A valid program before this patch passed verification with its non-owning refs having shorter (or equal) lifetimes due to more aggressive clobbering. Also, update existing tests to check bpf_rbtree_remove retval for NULL where necessary, and move rbtree_api_release_aliasing from progs/rbtree_fail.c to progs/rbtree.c since it's now expected to pass verification. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230415201811.343116-8-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-16 04:18:09 +08:00
if (rec->refcount_off < 0 &&
btf_record_has_field(rec, BPF_LIST_NODE) &&
bpf: Special verifier handling for bpf_rbtree_{remove, first} Newly-added bpf_rbtree_{remove,first} kfuncs have some special properties that require handling in the verifier: * both bpf_rbtree_remove and bpf_rbtree_first return the type containing the bpf_rb_node field, with the offset set to that field's offset, instead of a struct bpf_rb_node * * mark_reg_graph_node helper added in previous patch generalizes this logic, use it * bpf_rbtree_remove's node input is a node that's been inserted in the tree - a non-owning reference. * bpf_rbtree_remove must invalidate non-owning references in order to avoid aliasing issue. Use previously-added invalidate_non_owning_refs helper to mark this function as a non-owning ref invalidation point. * Unlike other functions, which convert one of their input arg regs to non-owning reference, bpf_rbtree_first takes no arguments and just returns a non-owning reference (possibly null) * For now verifier logic for this is special-cased instead of adding new kfunc flag. This patch, along with the previous one, complete special verifier handling for all rbtree API functions added in this series. With functional verifier handling of rbtree_remove, under current non-owning reference scheme, a node type with both bpf_{list,rb}_node fields could cause the verifier to accept programs which remove such nodes from collections they haven't been added to. In order to prevent this, this patch adds a check to btf_parse_fields which rejects structs with both bpf_{list,rb}_node fields. This is a temporary measure that can be removed after "collection identity" followup. See comment added in btf_parse_fields. A linked_list BTF test exercising the new check is added in this patch as well. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230214004017.2534011-6-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-02-14 08:40:14 +08:00
btf_record_has_field(rec, BPF_RB_NODE)) {
ret = -EINVAL;
goto end;
}
bpf: Remove btf_field_offs, use btf_record's fields instead The btf_field_offs struct contains (offset, size) for btf_record fields, sorted by offset. btf_field_offs is always used in conjunction with btf_record, which has btf_field 'fields' array with (offset, type), the latter of which btf_field_offs' size is derived from via btf_field_type_size. This patch adds a size field to struct btf_field and sorts btf_record's fields by offset, making it possible to get rid of btf_field_offs. Less data duplication and less code complexity results. Since btf_field_offs' lifetime closely followed the btf_record used to populate it, most complexity wins are from removal of initialization code like: if (btf_record_successfully_initialized) { foffs = btf_parse_field_offs(rec); if (IS_ERR_OR_NULL(foffs)) // free the btf_record and return err } Other changes in this patch are pretty mechanical: * foffs->field_off[i] -> rec->fields[i].offset * foffs->field_sz[i] -> rec->fields[i].size * Sort rec->fields in btf_parse_fields before returning * It's possible that this is necessary independently of other changes in this patch. btf_record_find in syscall.c expects btf_record's fields to be sorted by offset, yet there's no explicit sorting of them before this patch, record's fields are populated in the order they're read from BTF struct definition. BTF docs don't say anything about the sortedness of struct fields. * All functions taking struct btf_field_offs * input now instead take struct btf_record *. All callsites of these functions already have access to the correct btf_record. Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Link: https://lore.kernel.org/r/20230415201811.343116-2-davemarchevsky@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-16 04:18:03 +08:00
sort_r(rec->fields, rec->cnt, sizeof(struct btf_field), btf_field_cmp,
NULL, rec);
return rec;
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
end:
btf_record_free(rec);
bpf: Allow storing unreferenced kptr in map This commit introduces a new pointer type 'kptr' which can be embedded in a map value to hold a PTR_TO_BTF_ID stored by a BPF program during its invocation. When storing such a kptr, BPF program's PTR_TO_BTF_ID register must have the same type as in the map value's BTF, and loading a kptr marks the destination register as PTR_TO_BTF_ID with the correct kernel BTF and BTF ID. Such kptr are unreferenced, i.e. by the time another invocation of the BPF program loads this pointer, the object which the pointer points to may not longer exist. Since PTR_TO_BTF_ID loads (using BPF_LDX) are patched to PROBE_MEM loads by the verifier, it would safe to allow user to still access such invalid pointer, but passing such pointers into BPF helpers and kfuncs should not be permitted. A future patch in this series will close this gap. The flexibility offered by allowing programs to dereference such invalid pointers while being safe at runtime frees the verifier from doing complex lifetime tracking. As long as the user may ensure that the object remains valid, it can ensure data read by it from the kernel object is valid. The user indicates that a certain pointer must be treated as kptr capable of accepting stores of PTR_TO_BTF_ID of a certain type, by using a BTF type tag 'kptr' on the pointed to type of the pointer. Then, this information is recorded in the object BTF which will be passed into the kernel by way of map's BTF information. The name and kind from the map value BTF is used to look up the in-kernel type, and the actual BTF and BTF ID is recorded in the map struct in a new kptr_off_tab member. For now, only storing pointers to structs is permitted. An example of this specification is shown below: #define __kptr __attribute__((btf_type_tag("kptr"))) struct map_value { ... struct task_struct __kptr *task; ... }; Then, in a BPF program, user may store PTR_TO_BTF_ID with the type task_struct into the map, and then load it later. Note that the destination register is marked PTR_TO_BTF_ID_OR_NULL, as the verifier cannot know whether the value is NULL or not statically, it must treat all potential loads at that map value offset as loading a possibly NULL pointer. Only BPF_LDX, BPF_STX, and BPF_ST (with insn->imm = 0 to denote NULL) are allowed instructions that can access such a pointer. On BPF_LDX, the destination register is updated to be a PTR_TO_BTF_ID, and on BPF_STX, it is checked whether the source register type is a PTR_TO_BTF_ID with same BTF type as specified in the map BTF. The access size must always be BPF_DW. For the map in map support, the kptr_off_tab for outer map is copied from the inner map's kptr_off_tab. It was chosen to do a deep copy instead of introducing a refcount to kptr_off_tab, because the copy only needs to be done when paramterizing using inner_map_fd in the map in map case, hence would be unnecessary for all other users. It is not permitted to use MAP_FREEZE command and mmap for BPF map having kptrs, similar to the bpf_timer case. A kptr also requires that BPF program has both read and write access to the map (hence both BPF_F_RDONLY_PROG and BPF_F_WRONLY_PROG are disallowed). Note that check_map_access must be called from both check_helper_mem_access and for the BPF instructions, hence the kptr check must distinguish between ACCESS_DIRECT and ACCESS_HELPER, and reject ACCESS_HELPER cases. We rename stack_access_src to bpf_access_src and reuse it for this purpose. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-2-memxor@gmail.com
2022-04-25 05:48:49 +08:00
return ERR_PTR(ret);
}
int btf_check_and_fixup_fields(const struct btf *btf, struct btf_record *rec)
{
int i;
/* There are three types that signify ownership of some other type:
* kptr_ref, bpf_list_head, bpf_rb_root.
* kptr_ref only supports storing kernel types, which can't store
* references to program allocated local types.
*
* Hence we only need to ensure that bpf_{list_head,rb_root} ownership
* does not form cycles.
*/
if (IS_ERR_OR_NULL(rec) || !(rec->field_mask & BPF_GRAPH_ROOT))
return 0;
for (i = 0; i < rec->cnt; i++) {
struct btf_struct_meta *meta;
u32 btf_id;
if (!(rec->fields[i].type & BPF_GRAPH_ROOT))
continue;
btf_id = rec->fields[i].graph_root.value_btf_id;
meta = btf_find_struct_meta(btf, btf_id);
if (!meta)
return -EFAULT;
rec->fields[i].graph_root.value_rec = meta->record;
/* We need to set value_rec for all root types, but no need
* to check ownership cycle for a type unless it's also a
* node type.
*/
if (!(rec->field_mask & BPF_GRAPH_NODE))
continue;
/* We need to ensure ownership acyclicity among all types. The
* proper way to do it would be to topologically sort all BTF
* IDs based on the ownership edges, since there can be multiple
* bpf_{list_head,rb_node} in a type. Instead, we use the
* following resaoning:
*
* - A type can only be owned by another type in user BTF if it
* has a bpf_{list,rb}_node. Let's call these node types.
* - A type can only _own_ another type in user BTF if it has a
* bpf_{list_head,rb_root}. Let's call these root types.
*
* We ensure that if a type is both a root and node, its
* element types cannot be root types.
*
* To ensure acyclicity:
*
* When A is an root type but not a node, its ownership
* chain can be:
* A -> B -> C
* Where:
* - A is an root, e.g. has bpf_rb_root.
* - B is both a root and node, e.g. has bpf_rb_node and
* bpf_list_head.
* - C is only an root, e.g. has bpf_list_node
*
* When A is both a root and node, some other type already
* owns it in the BTF domain, hence it can not own
* another root type through any of the ownership edges.
* A -> B
* Where:
* - A is both an root and node.
* - B is only an node.
*/
if (meta->record->field_mask & BPF_GRAPH_ROOT)
return -ELOOP;
}
return 0;
}
static void __btf_struct_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
{
const struct btf_member *member;
void *safe_data;
u32 i;
safe_data = btf_show_start_struct_type(show, t, type_id, data);
if (!safe_data)
return;
for_each_member(i, t, member) {
const struct btf_type *member_type = btf_type_by_id(btf,
member->type);
const struct btf_kind_operations *ops;
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
u32 member_offset, bitfield_size;
u32 bytes_offset;
u8 bits8_offset;
btf_show_start_member(show, member);
member_offset = __btf_member_bit_offset(t, member);
bitfield_size = __btf_member_bitfield_size(t, member);
bytes_offset = BITS_ROUNDDOWN_BYTES(member_offset);
bits8_offset = BITS_PER_BYTE_MASKED(member_offset);
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (bitfield_size) {
safe_data = btf_show_start_type(show, member_type,
member->type,
data + bytes_offset);
if (safe_data)
btf_bitfield_show(safe_data,
bits8_offset,
bitfield_size, show);
btf_show_end_type(show);
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
} else {
ops = btf_type_ops(member_type);
ops->show(btf, member_type, member->type,
data + bytes_offset, bits8_offset, show);
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
}
btf_show_end_member(show);
}
btf_show_end_struct_type(show);
}
static void btf_struct_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
{
const struct btf_member *m = show->state.member;
/*
* First check if any members would be shown (are non-zero).
* See comments above "struct btf_show" definition for more
* details on how this works at a high-level.
*/
if (show->state.depth > 0 && !(show->flags & BTF_SHOW_ZERO)) {
if (!show->state.depth_check) {
show->state.depth_check = show->state.depth + 1;
show->state.depth_to_show = 0;
}
__btf_struct_show(btf, t, type_id, data, bits_offset, show);
/* Restore saved member data here */
show->state.member = m;
if (show->state.depth_check != show->state.depth + 1)
return;
show->state.depth_check = 0;
if (show->state.depth_to_show <= show->state.depth)
return;
/*
* Reaching here indicates we have recursed and found
* non-zero child values.
*/
}
__btf_struct_show(btf, t, type_id, data, bits_offset, show);
}
static struct btf_kind_operations struct_ops = {
.check_meta = btf_struct_check_meta,
.resolve = btf_struct_resolve,
.check_member = btf_struct_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_generic_check_kflag_member,
.log_details = btf_struct_log,
.show = btf_struct_show,
};
static int btf_enum_check_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
u32 struct_bits_off = member->offset;
u32 struct_size, bytes_offset;
if (BITS_PER_BYTE_MASKED(struct_bits_off)) {
btf_verifier_log_member(env, struct_type, member,
"Member is not byte aligned");
return -EINVAL;
}
struct_size = struct_type->size;
bytes_offset = BITS_ROUNDDOWN_BYTES(struct_bits_off);
if (struct_size - bytes_offset < member_type->size) {
btf_verifier_log_member(env, struct_type, member,
"Member exceeds struct_size");
return -EINVAL;
}
return 0;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
static int btf_enum_check_kflag_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
u32 struct_bits_off, nr_bits, bytes_end, struct_size;
u32 int_bitsize = sizeof(int) * BITS_PER_BYTE;
struct_bits_off = BTF_MEMBER_BIT_OFFSET(member->offset);
nr_bits = BTF_MEMBER_BITFIELD_SIZE(member->offset);
if (!nr_bits) {
if (BITS_PER_BYTE_MASKED(struct_bits_off)) {
btf_verifier_log_member(env, struct_type, member,
"Member is not byte aligned");
return -EINVAL;
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
}
nr_bits = int_bitsize;
} else if (nr_bits > int_bitsize) {
btf_verifier_log_member(env, struct_type, member,
"Invalid member bitfield_size");
return -EINVAL;
}
struct_size = struct_type->size;
bytes_end = BITS_ROUNDUP_BYTES(struct_bits_off + nr_bits);
if (struct_size < bytes_end) {
btf_verifier_log_member(env, struct_type, member,
"Member exceeds struct_size");
return -EINVAL;
}
return 0;
}
static s32 btf_enum_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
const struct btf_enum *enums = btf_type_enum(t);
struct btf *btf = env->btf;
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
const char *fmt_str;
u16 i, nr_enums;
u32 meta_needed;
nr_enums = btf_type_vlen(t);
meta_needed = nr_enums * sizeof(*enums);
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
if (t->size > 8 || !is_power_of_2(t->size)) {
btf_verifier_log_type(env, t, "Unexpected size");
return -EINVAL;
}
/* enum type either no name or a valid one */
if (t->name_off &&
!btf_name_valid_identifier(env->btf, t->name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
for (i = 0; i < nr_enums; i++) {
if (!btf_name_offset_valid(btf, enums[i].name_off)) {
btf_verifier_log(env, "\tInvalid name_offset:%u",
enums[i].name_off);
return -EINVAL;
}
/* enum member must have a valid name */
if (!enums[i].name_off ||
!btf_name_valid_identifier(btf, enums[i].name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
if (env->log.level == BPF_LOG_KERNEL)
continue;
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
fmt_str = btf_type_kflag(t) ? "\t%s val=%d\n" : "\t%s val=%u\n";
btf_verifier_log(env, fmt_str,
2018-12-14 02:41:46 +08:00
__btf_name_by_offset(btf, enums[i].name_off),
enums[i].val);
}
return meta_needed;
}
static void btf_enum_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
btf_verifier_log(env, "size=%u vlen=%u", t->size, btf_type_vlen(t));
}
static void btf_enum_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
{
const struct btf_enum *enums = btf_type_enum(t);
u32 i, nr_enums = btf_type_vlen(t);
void *safe_data;
int v;
safe_data = btf_show_start_type(show, t, type_id, data);
if (!safe_data)
return;
v = *(int *)safe_data;
for (i = 0; i < nr_enums; i++) {
if (v != enums[i].val)
continue;
btf_show_type_value(show, "%s",
__btf_name_by_offset(btf,
enums[i].name_off));
btf_show_end_type(show);
return;
}
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
if (btf_type_kflag(t))
btf_show_type_value(show, "%d", v);
else
btf_show_type_value(show, "%u", v);
btf_show_end_type(show);
}
static struct btf_kind_operations enum_ops = {
.check_meta = btf_enum_check_meta,
.resolve = btf_df_resolve,
.check_member = btf_enum_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_enum_check_kflag_member,
.log_details = btf_enum_log,
.show = btf_enum_show,
};
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
static s32 btf_enum64_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
const struct btf_enum64 *enums = btf_type_enum64(t);
struct btf *btf = env->btf;
const char *fmt_str;
u16 i, nr_enums;
u32 meta_needed;
nr_enums = btf_type_vlen(t);
meta_needed = nr_enums * sizeof(*enums);
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
if (t->size > 8 || !is_power_of_2(t->size)) {
btf_verifier_log_type(env, t, "Unexpected size");
return -EINVAL;
}
/* enum type either no name or a valid one */
if (t->name_off &&
!btf_name_valid_identifier(env->btf, t->name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
for (i = 0; i < nr_enums; i++) {
if (!btf_name_offset_valid(btf, enums[i].name_off)) {
btf_verifier_log(env, "\tInvalid name_offset:%u",
enums[i].name_off);
return -EINVAL;
}
/* enum member must have a valid name */
if (!enums[i].name_off ||
!btf_name_valid_identifier(btf, enums[i].name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
if (env->log.level == BPF_LOG_KERNEL)
continue;
fmt_str = btf_type_kflag(t) ? "\t%s val=%lld\n" : "\t%s val=%llu\n";
btf_verifier_log(env, fmt_str,
__btf_name_by_offset(btf, enums[i].name_off),
btf_enum64_value(enums + i));
}
return meta_needed;
}
static void btf_enum64_show(const struct btf *btf, const struct btf_type *t,
u32 type_id, void *data, u8 bits_offset,
struct btf_show *show)
{
const struct btf_enum64 *enums = btf_type_enum64(t);
u32 i, nr_enums = btf_type_vlen(t);
void *safe_data;
s64 v;
safe_data = btf_show_start_type(show, t, type_id, data);
if (!safe_data)
return;
v = *(u64 *)safe_data;
for (i = 0; i < nr_enums; i++) {
if (v != btf_enum64_value(enums + i))
continue;
btf_show_type_value(show, "%s",
__btf_name_by_offset(btf,
enums[i].name_off));
btf_show_end_type(show);
return;
}
if (btf_type_kflag(t))
btf_show_type_value(show, "%lld", v);
else
btf_show_type_value(show, "%llu", v);
btf_show_end_type(show);
}
static struct btf_kind_operations enum64_ops = {
.check_meta = btf_enum64_check_meta,
.resolve = btf_df_resolve,
.check_member = btf_enum_check_member,
.check_kflag_member = btf_enum_check_kflag_member,
.log_details = btf_enum_log,
.show = btf_enum64_show,
};
static s32 btf_func_proto_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
u32 meta_needed = btf_type_vlen(t) * sizeof(struct btf_param);
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
if (t->name_off) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
return meta_needed;
}
static void btf_func_proto_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
const struct btf_param *args = (const struct btf_param *)(t + 1);
u16 nr_args = btf_type_vlen(t), i;
btf_verifier_log(env, "return=%u args=(", t->type);
if (!nr_args) {
btf_verifier_log(env, "void");
goto done;
}
if (nr_args == 1 && !args[0].type) {
/* Only one vararg */
btf_verifier_log(env, "vararg");
goto done;
}
btf_verifier_log(env, "%u %s", args[0].type,
2018-12-14 02:41:46 +08:00
__btf_name_by_offset(env->btf,
args[0].name_off));
for (i = 1; i < nr_args - 1; i++)
btf_verifier_log(env, ", %u %s", args[i].type,
2018-12-14 02:41:46 +08:00
__btf_name_by_offset(env->btf,
args[i].name_off));
if (nr_args > 1) {
const struct btf_param *last_arg = &args[nr_args - 1];
if (last_arg->type)
btf_verifier_log(env, ", %u %s", last_arg->type,
2018-12-14 02:41:46 +08:00
__btf_name_by_offset(env->btf,
last_arg->name_off));
else
btf_verifier_log(env, ", vararg");
}
done:
btf_verifier_log(env, ")");
}
static struct btf_kind_operations func_proto_ops = {
.check_meta = btf_func_proto_check_meta,
.resolve = btf_df_resolve,
/*
* BTF_KIND_FUNC_PROTO cannot be directly referred by
* a struct's member.
*
* It should be a function pointer instead.
* (i.e. struct's member -> BTF_KIND_PTR -> BTF_KIND_FUNC_PROTO)
*
* Hence, there is no btf_func_check_member().
*/
.check_member = btf_df_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_df_check_kflag_member,
.log_details = btf_func_proto_log,
.show = btf_df_show,
};
static s32 btf_func_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
if (!t->name_off ||
!btf_name_valid_identifier(env->btf, t->name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
if (btf_type_vlen(t) > BTF_FUNC_GLOBAL) {
btf_verifier_log_type(env, t, "Invalid func linkage");
return -EINVAL;
}
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
return 0;
}
bpf: Fix a btf decl_tag bug when tagging a function syzbot reported a btf decl_tag bug with stack trace below: general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007] CPU: 0 PID: 3592 Comm: syz-executor914 Not tainted 5.16.0-syzkaller-11424-gb7892f7d5cb2 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:btf_type_vlen include/linux/btf.h:231 [inline] RIP: 0010:btf_decl_tag_resolve+0x83e/0xaa0 kernel/bpf/btf.c:3910 ... Call Trace: <TASK> btf_resolve+0x251/0x1020 kernel/bpf/btf.c:4198 btf_check_all_types kernel/bpf/btf.c:4239 [inline] btf_parse_type_sec kernel/bpf/btf.c:4280 [inline] btf_parse kernel/bpf/btf.c:4513 [inline] btf_new_fd+0x19fe/0x2370 kernel/bpf/btf.c:6047 bpf_btf_load kernel/bpf/syscall.c:4039 [inline] __sys_bpf+0x1cbb/0x5970 kernel/bpf/syscall.c:4679 __do_sys_bpf kernel/bpf/syscall.c:4738 [inline] __se_sys_bpf kernel/bpf/syscall.c:4736 [inline] __x64_sys_bpf+0x75/0xb0 kernel/bpf/syscall.c:4736 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae The kasan error is triggered with an illegal BTF like below: type 0: void type 1: int type 2: decl_tag to func type 3 type 3: func to func_proto type 8 The total number of types is 4 and the type 3 is illegal since its func_proto type is out of range. Currently, the target type of decl_tag can be struct/union, var or func. Both struct/union and var implemented their own 'resolve' callback functions and hence handled properly in kernel. But func type doesn't have 'resolve' callback function. When btf_decl_tag_resolve() tries to check func type, it tries to get vlen of its func_proto type, which triggered the above kasan error. To fix the issue, btf_decl_tag_resolve() needs to do btf_func_check() before trying to accessing func_proto type. In the current implementation, func type is checked with btf_func_check() in the main checking function btf_check_all_types(). To fix the above kasan issue, let us implement 'resolve' callback func type properly. The 'resolve' callback will be also called in btf_check_all_types() for func types. Fixes: b5ea834dde6b ("bpf: Support for new btf kind BTF_KIND_TAG") Reported-by: syzbot+53619be9444215e785ed@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20220203191727.741862-1-yhs@fb.com
2022-02-04 03:17:27 +08:00
static int btf_func_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
const struct btf_type *t = v->t;
u32 next_type_id = t->type;
int err;
err = btf_func_check(env, t);
if (err)
return err;
env_stack_pop_resolved(env, next_type_id, 0);
return 0;
}
static struct btf_kind_operations func_ops = {
.check_meta = btf_func_check_meta,
bpf: Fix a btf decl_tag bug when tagging a function syzbot reported a btf decl_tag bug with stack trace below: general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007] CPU: 0 PID: 3592 Comm: syz-executor914 Not tainted 5.16.0-syzkaller-11424-gb7892f7d5cb2 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:btf_type_vlen include/linux/btf.h:231 [inline] RIP: 0010:btf_decl_tag_resolve+0x83e/0xaa0 kernel/bpf/btf.c:3910 ... Call Trace: <TASK> btf_resolve+0x251/0x1020 kernel/bpf/btf.c:4198 btf_check_all_types kernel/bpf/btf.c:4239 [inline] btf_parse_type_sec kernel/bpf/btf.c:4280 [inline] btf_parse kernel/bpf/btf.c:4513 [inline] btf_new_fd+0x19fe/0x2370 kernel/bpf/btf.c:6047 bpf_btf_load kernel/bpf/syscall.c:4039 [inline] __sys_bpf+0x1cbb/0x5970 kernel/bpf/syscall.c:4679 __do_sys_bpf kernel/bpf/syscall.c:4738 [inline] __se_sys_bpf kernel/bpf/syscall.c:4736 [inline] __x64_sys_bpf+0x75/0xb0 kernel/bpf/syscall.c:4736 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae The kasan error is triggered with an illegal BTF like below: type 0: void type 1: int type 2: decl_tag to func type 3 type 3: func to func_proto type 8 The total number of types is 4 and the type 3 is illegal since its func_proto type is out of range. Currently, the target type of decl_tag can be struct/union, var or func. Both struct/union and var implemented their own 'resolve' callback functions and hence handled properly in kernel. But func type doesn't have 'resolve' callback function. When btf_decl_tag_resolve() tries to check func type, it tries to get vlen of its func_proto type, which triggered the above kasan error. To fix the issue, btf_decl_tag_resolve() needs to do btf_func_check() before trying to accessing func_proto type. In the current implementation, func type is checked with btf_func_check() in the main checking function btf_check_all_types(). To fix the above kasan issue, let us implement 'resolve' callback func type properly. The 'resolve' callback will be also called in btf_check_all_types() for func types. Fixes: b5ea834dde6b ("bpf: Support for new btf kind BTF_KIND_TAG") Reported-by: syzbot+53619be9444215e785ed@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20220203191727.741862-1-yhs@fb.com
2022-02-04 03:17:27 +08:00
.resolve = btf_func_resolve,
.check_member = btf_df_check_member,
bpf: btf: fix struct/union/fwd types with kind_flag This patch fixed two issues with BTF. One is related to struct/union bitfield encoding and the other is related to forward type. Issue #1 and solution: ====================== Current btf encoding of bitfield follows what pahole generates. For each bitfield, pahole will duplicate the type chain and put the bitfield size at the final int or enum type. Since the BTF enum type cannot encode bit size, pahole workarounds the issue by generating an int type whenever the enum bit size is not 32. For example, -bash-4.4$ cat t.c typedef int ___int; enum A { A1, A2, A3 }; struct t { int a[5]; ___int b:4; volatile enum A c:4; } g; -bash-4.4$ gcc -c -O2 -g t.c The current kernel supports the following BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t size=24 vlen=3 a type_id=5 bits_offset=0 b type_id=9 bits_offset=160 c type_id=11 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 [8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none) [9] TYPEDEF ___int type_id=8 [10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED [11] VOLATILE (anon) type_id=10 Two issues are in the above: . by changing enum type to int, we lost the original type information and this will not be ideal later when we try to convert BTF to a header file. . the type duplication for bitfields will cause BTF bloat. Duplicated types cannot be deduplicated later if the bitfield size is different. To fix this issue, this patch implemented a compatible change for BTF struct type encoding: . the bit 31 of struct_type->info, previously reserved, now is used to indicate whether bitfield_size is encoded in btf_member or not. . if bit 31 of struct_type->info is set, btf_member->offset will encode like: bit 0 - 23: bit offset bit 24 - 31: bitfield size if bit 31 is not set, the old behavior is preserved: bit 0 - 31: bit offset So if the struct contains a bit field, the maximum bit offset will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum bitfield size will be 256 which is enough for today as maximum bitfield in compiler can be 128 where int128 type is supported. This kernel patch intends to support the new BTF encoding: $ pahole -JV t.o [1] TYPEDEF ___int type_id=2 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [3] ENUM A size=4 vlen=3 A1 val=0 A2 val=1 A3 val=2 [4] STRUCT t kind_flag=1 size=24 vlen=3 a type_id=5 bitfield_size=0 bits_offset=0 b type_id=1 bitfield_size=4 bits_offset=160 c type_id=7 bitfield_size=4 bits_offset=164 [5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5 [6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none) [7] VOLATILE (anon) type_id=3 Issue #2 and solution: ====================== Current forward type in BTF does not specify whether the original type is struct or union. This will not work for type pretty print and BTF-to-header-file conversion as struct/union must be specified. $ cat tt.c struct t; union u; int foo(struct t *t, union u *u) { return 0; } $ gcc -c -g -O2 tt.c $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t type_id=0 [3] PTR (anon) type_id=2 [4] FWD u type_id=0 [5] PTR (anon) type_id=4 To fix this issue, similar to issue #1, type->info bit 31 is used. If the bit is set, it is union type. Otherwise, it is a struct type. $ pahole -JV tt.o [1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED [2] FWD t kind_flag=0 type_id=0 [3] PTR (anon) kind_flag=0 type_id=2 [4] FWD u kind_flag=1 type_id=0 [5] PTR (anon) kind_flag=0 type_id=4 Pahole/LLVM change: =================== The new kind_flag functionality has been implemented in pahole and llvm: https://github.com/yonghong-song/pahole/tree/bitfield https://github.com/yonghong-song/llvm/tree/bitfield Note that pahole hasn't implemented func/func_proto kind and .BTF.ext. So to print function signature with bpftool, the llvm compiler should be used. Fixes: 69b693f0aefa ("bpf: btf: Introduce BPF Type Format (BTF)") Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-12-16 14:13:51 +08:00
.check_kflag_member = btf_df_check_kflag_member,
.log_details = btf_ref_type_log,
.show = btf_df_show,
};
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
static s32 btf_var_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
const struct btf_var *var;
u32 meta_needed = sizeof(*var);
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
if (btf_type_vlen(t)) {
btf_verifier_log_type(env, t, "vlen != 0");
return -EINVAL;
}
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
if (!t->name_off ||
!btf_name_valid_identifier(env->btf, t->name_off)) {
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
/* A var cannot be in type void */
if (!t->type || !BTF_TYPE_ID_VALID(t->type)) {
btf_verifier_log_type(env, t, "Invalid type_id");
return -EINVAL;
}
var = btf_type_var(t);
if (var->linkage != BTF_VAR_STATIC &&
var->linkage != BTF_VAR_GLOBAL_ALLOCATED) {
btf_verifier_log_type(env, t, "Linkage not supported");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
return meta_needed;
}
static void btf_var_log(struct btf_verifier_env *env, const struct btf_type *t)
{
const struct btf_var *var = btf_type_var(t);
btf_verifier_log(env, "type_id=%u linkage=%u", t->type, var->linkage);
}
static const struct btf_kind_operations var_ops = {
.check_meta = btf_var_check_meta,
.resolve = btf_var_resolve,
.check_member = btf_df_check_member,
.check_kflag_member = btf_df_check_kflag_member,
.log_details = btf_var_log,
.show = btf_var_show,
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
};
static s32 btf_datasec_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
const struct btf_var_secinfo *vsi;
u64 last_vsi_end_off = 0, sum = 0;
u32 i, meta_needed;
meta_needed = btf_type_vlen(t) * sizeof(*vsi);
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
if (!t->size) {
btf_verifier_log_type(env, t, "size == 0");
return -EINVAL;
}
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
if (!t->name_off ||
!btf_name_valid_section(env->btf, t->name_off)) {
btf_verifier_log_type(env, t, "Invalid name");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
for_each_vsi(i, t, vsi) {
/* A var cannot be in type void */
if (!vsi->type || !BTF_TYPE_ID_VALID(vsi->type)) {
btf_verifier_log_vsi(env, t, vsi,
"Invalid type_id");
return -EINVAL;
}
if (vsi->offset < last_vsi_end_off || vsi->offset >= t->size) {
btf_verifier_log_vsi(env, t, vsi,
"Invalid offset");
return -EINVAL;
}
if (!vsi->size || vsi->size > t->size) {
btf_verifier_log_vsi(env, t, vsi,
"Invalid size");
return -EINVAL;
}
last_vsi_end_off = vsi->offset + vsi->size;
if (last_vsi_end_off > t->size) {
btf_verifier_log_vsi(env, t, vsi,
"Invalid offset+size");
return -EINVAL;
}
btf_verifier_log_vsi(env, t, vsi, NULL);
sum += vsi->size;
}
if (t->size < sum) {
btf_verifier_log_type(env, t, "Invalid btf_info size");
return -EINVAL;
}
return meta_needed;
}
static int btf_datasec_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
const struct btf_var_secinfo *vsi;
struct btf *btf = env->btf;
u16 i;
btf: fix resolving BTF_KIND_VAR after ARRAY, STRUCT, UNION, PTR btf_datasec_resolve contains a bug that causes the following BTF to fail loading: [1] DATASEC a size=2 vlen=2 type_id=4 offset=0 size=1 type_id=7 offset=1 size=1 [2] INT (anon) size=1 bits_offset=0 nr_bits=8 encoding=(none) [3] PTR (anon) type_id=2 [4] VAR a type_id=3 linkage=0 [5] INT (anon) size=1 bits_offset=0 nr_bits=8 encoding=(none) [6] TYPEDEF td type_id=5 [7] VAR b type_id=6 linkage=0 This error message is printed during btf_check_all_types: [1] DATASEC a size=2 vlen=2 type_id=7 offset=1 size=1 Invalid type By tracing btf_*_resolve we can pinpoint the problem: btf_datasec_resolve(depth: 1, type_id: 1, mode: RESOLVE_TBD) = 0 btf_var_resolve(depth: 2, type_id: 4, mode: RESOLVE_TBD) = 0 btf_ptr_resolve(depth: 3, type_id: 3, mode: RESOLVE_PTR) = 0 btf_var_resolve(depth: 2, type_id: 4, mode: RESOLVE_PTR) = 0 btf_datasec_resolve(depth: 1, type_id: 1, mode: RESOLVE_PTR) = -22 The last invocation of btf_datasec_resolve should invoke btf_var_resolve by means of env_stack_push, instead it returns EINVAL. The reason is that env_stack_push is never executed for the second VAR. if (!env_type_is_resolve_sink(env, var_type) && !env_type_is_resolved(env, var_type_id)) { env_stack_set_next_member(env, i + 1); return env_stack_push(env, var_type, var_type_id); } env_type_is_resolve_sink() changes its behaviour based on resolve_mode. For RESOLVE_PTR, we can simplify the if condition to the following: (btf_type_is_modifier() || btf_type_is_ptr) && !env_type_is_resolved() Since we're dealing with a VAR the clause evaluates to false. This is not sufficient to trigger the bug however. The log output and EINVAL are only generated if btf_type_id_size() fails. if (!btf_type_id_size(btf, &type_id, &type_size)) { btf_verifier_log_vsi(env, v->t, vsi, "Invalid type"); return -EINVAL; } Most types are sized, so for example a VAR referring to an INT is not a problem. The bug is only triggered if a VAR points at a modifier. Since we skipped btf_var_resolve that modifier was also never resolved, which means that btf_resolved_type_id returns 0 aka VOID for the modifier. This in turn causes btf_type_id_size to return NULL, triggering EINVAL. To summarise, the following conditions are necessary: - VAR pointing at PTR, STRUCT, UNION or ARRAY - Followed by a VAR pointing at TYPEDEF, VOLATILE, CONST, RESTRICT or TYPE_TAG The fix is to reset resolve_mode to RESOLVE_TBD before attempting to resolve a VAR from a DATASEC. Fixes: 1dc92851849c ("bpf: kernel side support for BTF Var and DataSec") Signed-off-by: Lorenz Bauer <lmb@isovalent.com> Link: https://lore.kernel.org/r/20230306112138.155352-2-lmb@isovalent.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2023-03-06 19:21:37 +08:00
env->resolve_mode = RESOLVE_TBD;
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
for_each_vsi_from(i, v->next_member, v->t, vsi) {
u32 var_type_id = vsi->type, type_id, type_size = 0;
const struct btf_type *var_type = btf_type_by_id(env->btf,
var_type_id);
if (!var_type || !btf_type_is_var(var_type)) {
btf_verifier_log_vsi(env, v->t, vsi,
"Not a VAR kind member");
return -EINVAL;
}
if (!env_type_is_resolve_sink(env, var_type) &&
!env_type_is_resolved(env, var_type_id)) {
env_stack_set_next_member(env, i + 1);
return env_stack_push(env, var_type, var_type_id);
}
type_id = var_type->type;
if (!btf_type_id_size(btf, &type_id, &type_size)) {
btf_verifier_log_vsi(env, v->t, vsi, "Invalid type");
return -EINVAL;
}
if (vsi->size < type_size) {
btf_verifier_log_vsi(env, v->t, vsi, "Invalid size");
return -EINVAL;
}
}
env_stack_pop_resolved(env, 0, 0);
return 0;
}
static void btf_datasec_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
btf_verifier_log(env, "size=%u vlen=%u", t->size, btf_type_vlen(t));
}
static void btf_datasec_show(const struct btf *btf,
const struct btf_type *t, u32 type_id,
void *data, u8 bits_offset,
struct btf_show *show)
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
{
const struct btf_var_secinfo *vsi;
const struct btf_type *var;
u32 i;
if (!btf_show_start_type(show, t, type_id, data))
return;
btf_show_type_value(show, "section (\"%s\") = {",
__btf_name_by_offset(btf, t->name_off));
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
for_each_vsi(i, t, vsi) {
var = btf_type_by_id(btf, vsi->type);
if (i)
btf_show(show, ",");
btf_type_ops(var)->show(btf, var, vsi->type,
data + vsi->offset, bits_offset, show);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
}
btf_show_end_type(show);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
}
static const struct btf_kind_operations datasec_ops = {
.check_meta = btf_datasec_check_meta,
.resolve = btf_datasec_resolve,
.check_member = btf_df_check_member,
.check_kflag_member = btf_df_check_kflag_member,
.log_details = btf_datasec_log,
.show = btf_datasec_show,
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
};
static s32 btf_float_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
if (btf_type_vlen(t)) {
btf_verifier_log_type(env, t, "vlen != 0");
return -EINVAL;
}
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
if (t->size != 2 && t->size != 4 && t->size != 8 && t->size != 12 &&
t->size != 16) {
btf_verifier_log_type(env, t, "Invalid type_size");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
return 0;
}
static int btf_float_check_member(struct btf_verifier_env *env,
const struct btf_type *struct_type,
const struct btf_member *member,
const struct btf_type *member_type)
{
u64 start_offset_bytes;
u64 end_offset_bytes;
u64 misalign_bits;
u64 align_bytes;
u64 align_bits;
/* Different architectures have different alignment requirements, so
* here we check only for the reasonable minimum. This way we ensure
* that types after CO-RE can pass the kernel BTF verifier.
*/
align_bytes = min_t(u64, sizeof(void *), member_type->size);
align_bits = align_bytes * BITS_PER_BYTE;
div64_u64_rem(member->offset, align_bits, &misalign_bits);
if (misalign_bits) {
btf_verifier_log_member(env, struct_type, member,
"Member is not properly aligned");
return -EINVAL;
}
start_offset_bytes = member->offset / BITS_PER_BYTE;
end_offset_bytes = start_offset_bytes + member_type->size;
if (end_offset_bytes > struct_type->size) {
btf_verifier_log_member(env, struct_type, member,
"Member exceeds struct_size");
return -EINVAL;
}
return 0;
}
static void btf_float_log(struct btf_verifier_env *env,
const struct btf_type *t)
{
btf_verifier_log(env, "size=%u", t->size);
}
static const struct btf_kind_operations float_ops = {
.check_meta = btf_float_check_meta,
.resolve = btf_df_resolve,
.check_member = btf_float_check_member,
.check_kflag_member = btf_generic_check_kflag_member,
.log_details = btf_float_log,
.show = btf_df_show,
};
static s32 btf_decl_tag_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
const struct btf_decl_tag *tag;
u32 meta_needed = sizeof(*tag);
s32 component_idx;
const char *value;
if (meta_left < meta_needed) {
btf_verifier_log_basic(env, t,
"meta_left:%u meta_needed:%u",
meta_left, meta_needed);
return -EINVAL;
}
value = btf_name_by_offset(env->btf, t->name_off);
if (!value || !value[0]) {
btf_verifier_log_type(env, t, "Invalid value");
return -EINVAL;
}
if (btf_type_vlen(t)) {
btf_verifier_log_type(env, t, "vlen != 0");
return -EINVAL;
}
if (btf_type_kflag(t)) {
btf_verifier_log_type(env, t, "Invalid btf_info kind_flag");
return -EINVAL;
}
component_idx = btf_type_decl_tag(t)->component_idx;
if (component_idx < -1) {
btf_verifier_log_type(env, t, "Invalid component_idx");
return -EINVAL;
}
btf_verifier_log_type(env, t, NULL);
return meta_needed;
}
static int btf_decl_tag_resolve(struct btf_verifier_env *env,
const struct resolve_vertex *v)
{
const struct btf_type *next_type;
const struct btf_type *t = v->t;
u32 next_type_id = t->type;
struct btf *btf = env->btf;
s32 component_idx;
u32 vlen;
next_type = btf_type_by_id(btf, next_type_id);
if (!next_type || !btf_type_is_decl_tag_target(next_type)) {
btf_verifier_log_type(env, v->t, "Invalid type_id");
return -EINVAL;
}
if (!env_type_is_resolve_sink(env, next_type) &&
!env_type_is_resolved(env, next_type_id))
return env_stack_push(env, next_type, next_type_id);
component_idx = btf_type_decl_tag(t)->component_idx;
if (component_idx != -1) {
if (btf_type_is_var(next_type) || btf_type_is_typedef(next_type)) {
btf_verifier_log_type(env, v->t, "Invalid component_idx");
return -EINVAL;
}
if (btf_type_is_struct(next_type)) {
vlen = btf_type_vlen(next_type);
} else {
/* next_type should be a function */
next_type = btf_type_by_id(btf, next_type->type);
vlen = btf_type_vlen(next_type);
}
if ((u32)component_idx >= vlen) {
btf_verifier_log_type(env, v->t, "Invalid component_idx");
return -EINVAL;
}
}
env_stack_pop_resolved(env, next_type_id, 0);
return 0;
}
static void btf_decl_tag_log(struct btf_verifier_env *env, const struct btf_type *t)
{
btf_verifier_log(env, "type=%u component_idx=%d", t->type,
btf_type_decl_tag(t)->component_idx);
}
static const struct btf_kind_operations decl_tag_ops = {
.check_meta = btf_decl_tag_check_meta,
.resolve = btf_decl_tag_resolve,
.check_member = btf_df_check_member,
.check_kflag_member = btf_df_check_kflag_member,
.log_details = btf_decl_tag_log,
.show = btf_df_show,
};
static int btf_func_proto_check(struct btf_verifier_env *env,
const struct btf_type *t)
{
const struct btf_type *ret_type;
const struct btf_param *args;
const struct btf *btf;
u16 nr_args, i;
int err;
btf = env->btf;
args = (const struct btf_param *)(t + 1);
nr_args = btf_type_vlen(t);
/* Check func return type which could be "void" (t->type == 0) */
if (t->type) {
u32 ret_type_id = t->type;
ret_type = btf_type_by_id(btf, ret_type_id);
if (!ret_type) {
btf_verifier_log_type(env, t, "Invalid return type");
return -EINVAL;
}
bpf: prevent decl_tag from being referenced in func_proto Syzkaller was able to hit the following issue: ------------[ cut here ]------------ WARNING: CPU: 0 PID: 3609 at kernel/bpf/btf.c:1946 btf_type_id_size+0x2d5/0x9d0 kernel/bpf/btf.c:1946 Modules linked in: CPU: 0 PID: 3609 Comm: syz-executor361 Not tainted 6.0.0-syzkaller-02734-g0326074ff465 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 09/22/2022 RIP: 0010:btf_type_id_size+0x2d5/0x9d0 kernel/bpf/btf.c:1946 Code: ef e8 7f 8e e4 ff 41 83 ff 0b 77 28 f6 44 24 10 18 75 3f e8 6d 91 e4 ff 44 89 fe bf 0e 00 00 00 e8 20 8e e4 ff e8 5b 91 e4 ff <0f> 0b 45 31 f6 e9 98 02 00 00 41 83 ff 12 74 18 e8 46 91 e4 ff 44 RSP: 0018:ffffc90003cefb40 EFLAGS: 00010293 RAX: 0000000000000000 RBX: 0000000000000002 RCX: 0000000000000000 RDX: ffff8880259c0000 RSI: ffffffff81968415 RDI: 0000000000000005 RBP: ffff88801270ca00 R08: 0000000000000005 R09: 000000000000000e R10: 0000000000000011 R11: 0000000000000000 R12: 0000000000000000 R13: 0000000000000011 R14: ffff888026ee6424 R15: 0000000000000011 FS: 000055555641b300(0000) GS:ffff8880b9a00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000f2e258 CR3: 000000007110e000 CR4: 00000000003506f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: <TASK> btf_func_proto_check kernel/bpf/btf.c:4447 [inline] btf_check_all_types kernel/bpf/btf.c:4723 [inline] btf_parse_type_sec kernel/bpf/btf.c:4752 [inline] btf_parse kernel/bpf/btf.c:5026 [inline] btf_new_fd+0x1926/0x1e70 kernel/bpf/btf.c:6892 bpf_btf_load kernel/bpf/syscall.c:4324 [inline] __sys_bpf+0xb7d/0x4cf0 kernel/bpf/syscall.c:5010 __do_sys_bpf kernel/bpf/syscall.c:5069 [inline] __se_sys_bpf kernel/bpf/syscall.c:5067 [inline] __x64_sys_bpf+0x75/0xb0 kernel/bpf/syscall.c:5067 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd RIP: 0033:0x7f0fbae41c69 Code: 28 c3 e8 2a 14 00 00 66 2e 0f 1f 84 00 00 00 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 c0 ff ff ff f7 d8 64 89 01 48 RSP: 002b:00007ffc8aeb6228 EFLAGS: 00000246 ORIG_RAX: 0000000000000141 RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f0fbae41c69 RDX: 0000000000000020 RSI: 0000000020000140 RDI: 0000000000000012 RBP: 00007f0fbae05e10 R08: 0000000000000000 R09: 0000000000000000 R10: 00000000ffffffff R11: 0000000000000246 R12: 00007f0fbae05ea0 R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000 </TASK> Looks like it tries to create a func_proto which return type is decl_tag. For the details, see Martin's spot on analysis in [0]. 0: https://lore.kernel.org/bpf/CAKH8qBuQDLva_hHxxBuZzyAcYNO4ejhovz6TQeVSk8HY-2SO6g@mail.gmail.com/T/#mea6524b3fcd6298347432226e81b1e6155efc62c Cc: Yonghong Song <yhs@fb.com> Cc: Martin KaFai Lau <martin.lau@kernel.org> Fixes: bd16dee66ae4 ("bpf: Add BTF_KIND_DECL_TAG typedef support") Reported-by: syzbot+d8bd751aef7c6b39a344@syzkaller.appspotmail.com Signed-off-by: Stanislav Fomichev <sdf@google.com> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20221015002444.2680969-2-sdf@google.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2022-10-15 08:24:44 +08:00
if (btf_type_is_resolve_source_only(ret_type)) {
btf_verifier_log_type(env, t, "Invalid return type");
return -EINVAL;
}
if (btf_type_needs_resolve(ret_type) &&
!env_type_is_resolved(env, ret_type_id)) {
err = btf_resolve(env, ret_type, ret_type_id);
if (err)
return err;
}
/* Ensure the return type is a type that has a size */
if (!btf_type_id_size(btf, &ret_type_id, NULL)) {
btf_verifier_log_type(env, t, "Invalid return type");
return -EINVAL;
}
}
if (!nr_args)
return 0;
/* Last func arg type_id could be 0 if it is a vararg */
if (!args[nr_args - 1].type) {
if (args[nr_args - 1].name_off) {
btf_verifier_log_type(env, t, "Invalid arg#%u",
nr_args);
return -EINVAL;
}
nr_args--;
}
for (i = 0; i < nr_args; i++) {
const struct btf_type *arg_type;
u32 arg_type_id;
arg_type_id = args[i].type;
arg_type = btf_type_by_id(btf, arg_type_id);
if (!arg_type) {
btf_verifier_log_type(env, t, "Invalid arg#%u", i + 1);
return -EINVAL;
}
if (btf_type_is_resolve_source_only(arg_type)) {
btf_verifier_log_type(env, t, "Invalid arg#%u", i + 1);
return -EINVAL;
}
if (args[i].name_off &&
(!btf_name_offset_valid(btf, args[i].name_off) ||
!btf_name_valid_identifier(btf, args[i].name_off))) {
btf_verifier_log_type(env, t,
"Invalid arg#%u", i + 1);
return -EINVAL;
}
if (btf_type_needs_resolve(arg_type) &&
!env_type_is_resolved(env, arg_type_id)) {
err = btf_resolve(env, arg_type, arg_type_id);
if (err)
return err;
}
if (!btf_type_id_size(btf, &arg_type_id, NULL)) {
btf_verifier_log_type(env, t, "Invalid arg#%u", i + 1);
return -EINVAL;
}
}
return 0;
}
static int btf_func_check(struct btf_verifier_env *env,
const struct btf_type *t)
{
const struct btf_type *proto_type;
const struct btf_param *args;
const struct btf *btf;
u16 nr_args, i;
btf = env->btf;
proto_type = btf_type_by_id(btf, t->type);
if (!proto_type || !btf_type_is_func_proto(proto_type)) {
btf_verifier_log_type(env, t, "Invalid type_id");
return -EINVAL;
}
args = (const struct btf_param *)(proto_type + 1);
nr_args = btf_type_vlen(proto_type);
for (i = 0; i < nr_args; i++) {
if (!args[i].name_off && args[i].type) {
btf_verifier_log_type(env, t, "Invalid arg#%u", i + 1);
return -EINVAL;
}
}
return 0;
}
static const struct btf_kind_operations * const kind_ops[NR_BTF_KINDS] = {
[BTF_KIND_INT] = &int_ops,
[BTF_KIND_PTR] = &ptr_ops,
[BTF_KIND_ARRAY] = &array_ops,
[BTF_KIND_STRUCT] = &struct_ops,
[BTF_KIND_UNION] = &struct_ops,
[BTF_KIND_ENUM] = &enum_ops,
[BTF_KIND_FWD] = &fwd_ops,
[BTF_KIND_TYPEDEF] = &modifier_ops,
[BTF_KIND_VOLATILE] = &modifier_ops,
[BTF_KIND_CONST] = &modifier_ops,
[BTF_KIND_RESTRICT] = &modifier_ops,
[BTF_KIND_FUNC] = &func_ops,
[BTF_KIND_FUNC_PROTO] = &func_proto_ops,
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
[BTF_KIND_VAR] = &var_ops,
[BTF_KIND_DATASEC] = &datasec_ops,
[BTF_KIND_FLOAT] = &float_ops,
[BTF_KIND_DECL_TAG] = &decl_tag_ops,
[BTF_KIND_TYPE_TAG] = &modifier_ops,
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
[BTF_KIND_ENUM64] = &enum64_ops,
};
static s32 btf_check_meta(struct btf_verifier_env *env,
const struct btf_type *t,
u32 meta_left)
{
u32 saved_meta_left = meta_left;
s32 var_meta_size;
if (meta_left < sizeof(*t)) {
btf_verifier_log(env, "[%u] meta_left:%u meta_needed:%zu",
env->log_type_id, meta_left, sizeof(*t));
return -EINVAL;
}
meta_left -= sizeof(*t);
if (t->info & ~BTF_INFO_MASK) {
btf_verifier_log(env, "[%u] Invalid btf_info:%x",
env->log_type_id, t->info);
return -EINVAL;
}
if (BTF_INFO_KIND(t->info) > BTF_KIND_MAX ||
BTF_INFO_KIND(t->info) == BTF_KIND_UNKN) {
btf_verifier_log(env, "[%u] Invalid kind:%u",
env->log_type_id, BTF_INFO_KIND(t->info));
return -EINVAL;
}
if (!btf_name_offset_valid(env->btf, t->name_off)) {
btf_verifier_log(env, "[%u] Invalid name_offset:%u",
env->log_type_id, t->name_off);
return -EINVAL;
}
var_meta_size = btf_type_ops(t)->check_meta(env, t, meta_left);
if (var_meta_size < 0)
return var_meta_size;
meta_left -= var_meta_size;
return saved_meta_left - meta_left;
}
static int btf_check_all_metas(struct btf_verifier_env *env)
{
struct btf *btf = env->btf;
struct btf_header *hdr;
void *cur, *end;
hdr = &btf->hdr;
cur = btf->nohdr_data + hdr->type_off;
end = cur + hdr->type_len;
env->log_type_id = btf->base_btf ? btf->start_id : 1;
while (cur < end) {
struct btf_type *t = cur;
s32 meta_size;
meta_size = btf_check_meta(env, t, end - cur);
if (meta_size < 0)
return meta_size;
btf_add_type(env, t);
cur += meta_size;
env->log_type_id++;
}
return 0;
}
static bool btf_resolve_valid(struct btf_verifier_env *env,
const struct btf_type *t,
u32 type_id)
{
struct btf *btf = env->btf;
if (!env_type_is_resolved(env, type_id))
return false;
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
if (btf_type_is_struct(t) || btf_type_is_datasec(t))
return !btf_resolved_type_id(btf, type_id) &&
!btf_resolved_type_size(btf, type_id);
bpf: Fix a btf decl_tag bug when tagging a function syzbot reported a btf decl_tag bug with stack trace below: general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007] CPU: 0 PID: 3592 Comm: syz-executor914 Not tainted 5.16.0-syzkaller-11424-gb7892f7d5cb2 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:btf_type_vlen include/linux/btf.h:231 [inline] RIP: 0010:btf_decl_tag_resolve+0x83e/0xaa0 kernel/bpf/btf.c:3910 ... Call Trace: <TASK> btf_resolve+0x251/0x1020 kernel/bpf/btf.c:4198 btf_check_all_types kernel/bpf/btf.c:4239 [inline] btf_parse_type_sec kernel/bpf/btf.c:4280 [inline] btf_parse kernel/bpf/btf.c:4513 [inline] btf_new_fd+0x19fe/0x2370 kernel/bpf/btf.c:6047 bpf_btf_load kernel/bpf/syscall.c:4039 [inline] __sys_bpf+0x1cbb/0x5970 kernel/bpf/syscall.c:4679 __do_sys_bpf kernel/bpf/syscall.c:4738 [inline] __se_sys_bpf kernel/bpf/syscall.c:4736 [inline] __x64_sys_bpf+0x75/0xb0 kernel/bpf/syscall.c:4736 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae The kasan error is triggered with an illegal BTF like below: type 0: void type 1: int type 2: decl_tag to func type 3 type 3: func to func_proto type 8 The total number of types is 4 and the type 3 is illegal since its func_proto type is out of range. Currently, the target type of decl_tag can be struct/union, var or func. Both struct/union and var implemented their own 'resolve' callback functions and hence handled properly in kernel. But func type doesn't have 'resolve' callback function. When btf_decl_tag_resolve() tries to check func type, it tries to get vlen of its func_proto type, which triggered the above kasan error. To fix the issue, btf_decl_tag_resolve() needs to do btf_func_check() before trying to accessing func_proto type. In the current implementation, func type is checked with btf_func_check() in the main checking function btf_check_all_types(). To fix the above kasan issue, let us implement 'resolve' callback func type properly. The 'resolve' callback will be also called in btf_check_all_types() for func types. Fixes: b5ea834dde6b ("bpf: Support for new btf kind BTF_KIND_TAG") Reported-by: syzbot+53619be9444215e785ed@syzkaller.appspotmail.com Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20220203191727.741862-1-yhs@fb.com
2022-02-04 03:17:27 +08:00
if (btf_type_is_decl_tag(t) || btf_type_is_func(t))
return btf_resolved_type_id(btf, type_id) &&
!btf_resolved_type_size(btf, type_id);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
if (btf_type_is_modifier(t) || btf_type_is_ptr(t) ||
btf_type_is_var(t)) {
t = btf_type_id_resolve(btf, &type_id);
bpf: kernel side support for BTF Var and DataSec This work adds kernel-side verification, logging and seq_show dumping of BTF Var and DataSec kinds which are emitted with latest LLVM. The following constraints apply: BTF Var must have: - Its kind_flag is 0 - Its vlen is 0 - Must point to a valid type - Type must not resolve to a forward type - Size of underlying type must be > 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. in case of static variables inside functions) - Cannot be a member of a struct/union - Linkage so far can either only be static or global/allocated BTF DataSec must have: - Its kind_flag is 0 - Its vlen cannot be 0 - Its size cannot be 0 - Must have a valid name - Can only be a source type, not sink or intermediate one - Name may include dots (e.g. to represent .bss, .data, .rodata etc) - Cannot be a member of a struct/union - Inner btf_var_secinfo array with {type,offset,size} triple must be sorted by offset in ascending order - Type must always point to BTF Var - BTF resolved size of Var must be <= size provided by triple - DataSec size must be >= sum of triple sizes (thus holes are allowed) btf_var_resolve(), btf_ptr_resolve() and btf_modifier_resolve() are on a high level quite similar but each come with slight, subtle differences. They could potentially be a bit refactored in future which hasn't been done here to ease review. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-10 05:20:09 +08:00
return t &&
!btf_type_is_modifier(t) &&
!btf_type_is_var(t) &&
!btf_type_is_datasec(t);
}
if (btf_type_is_array(t)) {
const struct btf_array *array = btf_type_array(t);
const struct btf_type *elem_type;
u32 elem_type_id = array->type;
u32 elem_size;
elem_type = btf_type_id_size(btf, &elem_type_id, &elem_size);
return elem_type && !btf_type_is_modifier(elem_type) &&
(array->nelems * elem_size ==
btf_resolved_type_size(btf, type_id));
}
return false;
}
static int btf_resolve(struct btf_verifier_env *env,
const struct btf_type *t, u32 type_id)
{
u32 save_log_type_id = env->log_type_id;
const struct resolve_vertex *v;
int err = 0;
env->resolve_mode = RESOLVE_TBD;
env_stack_push(env, t, type_id);
while (!err && (v = env_stack_peak(env))) {
env->log_type_id = v->type_id;
err = btf_type_ops(v->t)->resolve(env, v);
}
env->log_type_id = type_id;
if (err == -E2BIG) {
btf_verifier_log_type(env, t,
"Exceeded max resolving depth:%u",
MAX_RESOLVE_DEPTH);
} else if (err == -EEXIST) {
btf_verifier_log_type(env, t, "Loop detected");
}
/* Final sanity check */
if (!err && !btf_resolve_valid(env, t, type_id)) {
btf_verifier_log_type(env, t, "Invalid resolve state");
err = -EINVAL;
}
env->log_type_id = save_log_type_id;
return err;
}
static int btf_check_all_types(struct btf_verifier_env *env)
{
struct btf *btf = env->btf;
const struct btf_type *t;
u32 type_id, i;
int err;
err = env_resolve_init(env);
if (err)
return err;
env->phase++;
for (i = btf->base_btf ? 0 : 1; i < btf->nr_types; i++) {
type_id = btf->start_id + i;
t = btf_type_by_id(btf, type_id);
env->log_type_id = type_id;
if (btf_type_needs_resolve(t) &&
!env_type_is_resolved(env, type_id)) {
err = btf_resolve(env, t, type_id);
if (err)
return err;
}
if (btf_type_is_func_proto(t)) {
err = btf_func_proto_check(env, t);
if (err)
return err;
}
}
return 0;
}
static int btf_parse_type_sec(struct btf_verifier_env *env)
{
const struct btf_header *hdr = &env->btf->hdr;
int err;
/* Type section must align to 4 bytes */
if (hdr->type_off & (sizeof(u32) - 1)) {
btf_verifier_log(env, "Unaligned type_off");
return -EINVAL;
}
if (!env->btf->base_btf && !hdr->type_len) {
btf_verifier_log(env, "No type found");
return -EINVAL;
}
err = btf_check_all_metas(env);
if (err)
return err;
return btf_check_all_types(env);
}
static int btf_parse_str_sec(struct btf_verifier_env *env)
{
const struct btf_header *hdr;
struct btf *btf = env->btf;
const char *start, *end;
hdr = &btf->hdr;
start = btf->nohdr_data + hdr->str_off;
end = start + hdr->str_len;
if (end != btf->data + btf->data_size) {
btf_verifier_log(env, "String section is not at the end");
return -EINVAL;
}
btf->strings = start;
if (btf->base_btf && !hdr->str_len)
return 0;
if (!hdr->str_len || hdr->str_len - 1 > BTF_MAX_NAME_OFFSET || end[-1]) {
btf_verifier_log(env, "Invalid string section");
return -EINVAL;
}
if (!btf->base_btf && start[0]) {
btf_verifier_log(env, "Invalid string section");
return -EINVAL;
}
return 0;
}
static const size_t btf_sec_info_offset[] = {
offsetof(struct btf_header, type_off),
offsetof(struct btf_header, str_off),
};
static int btf_sec_info_cmp(const void *a, const void *b)
{
const struct btf_sec_info *x = a;
const struct btf_sec_info *y = b;
return (int)(x->off - y->off) ? : (int)(x->len - y->len);
}
static int btf_check_sec_info(struct btf_verifier_env *env,
u32 btf_data_size)
{
struct btf_sec_info secs[ARRAY_SIZE(btf_sec_info_offset)];
u32 total, expected_total, i;
const struct btf_header *hdr;
const struct btf *btf;
btf = env->btf;
hdr = &btf->hdr;
/* Populate the secs from hdr */
for (i = 0; i < ARRAY_SIZE(btf_sec_info_offset); i++)
secs[i] = *(struct btf_sec_info *)((void *)hdr +
btf_sec_info_offset[i]);
sort(secs, ARRAY_SIZE(btf_sec_info_offset),
sizeof(struct btf_sec_info), btf_sec_info_cmp, NULL);
/* Check for gaps and overlap among sections */
total = 0;
expected_total = btf_data_size - hdr->hdr_len;
for (i = 0; i < ARRAY_SIZE(btf_sec_info_offset); i++) {
if (expected_total < secs[i].off) {
btf_verifier_log(env, "Invalid section offset");
return -EINVAL;
}
if (total < secs[i].off) {
/* gap */
btf_verifier_log(env, "Unsupported section found");
return -EINVAL;
}
if (total > secs[i].off) {
btf_verifier_log(env, "Section overlap found");
return -EINVAL;
}
if (expected_total - total < secs[i].len) {
btf_verifier_log(env,
"Total section length too long");
return -EINVAL;
}
total += secs[i].len;
}
/* There is data other than hdr and known sections */
if (expected_total != total) {
btf_verifier_log(env, "Unsupported section found");
return -EINVAL;
}
return 0;
}
static int btf_parse_hdr(struct btf_verifier_env *env)
{
u32 hdr_len, hdr_copy, btf_data_size;
const struct btf_header *hdr;
struct btf *btf;
btf = env->btf;
btf_data_size = btf->data_size;
if (btf_data_size < offsetofend(struct btf_header, hdr_len)) {
btf_verifier_log(env, "hdr_len not found");
return -EINVAL;
}
hdr = btf->data;
hdr_len = hdr->hdr_len;
if (btf_data_size < hdr_len) {
btf_verifier_log(env, "btf_header not found");
return -EINVAL;
}
/* Ensure the unsupported header fields are zero */
if (hdr_len > sizeof(btf->hdr)) {
u8 *expected_zero = btf->data + sizeof(btf->hdr);
u8 *end = btf->data + hdr_len;
for (; expected_zero < end; expected_zero++) {
if (*expected_zero) {
btf_verifier_log(env, "Unsupported btf_header");
return -E2BIG;
}
}
}
hdr_copy = min_t(u32, hdr_len, sizeof(btf->hdr));
memcpy(&btf->hdr, btf->data, hdr_copy);
hdr = &btf->hdr;
btf_verifier_log_hdr(env, btf_data_size);
if (hdr->magic != BTF_MAGIC) {
btf_verifier_log(env, "Invalid magic");
return -EINVAL;
}
if (hdr->version != BTF_VERSION) {
btf_verifier_log(env, "Unsupported version");
return -ENOTSUPP;
}
if (hdr->flags) {
btf_verifier_log(env, "Unsupported flags");
return -ENOTSUPP;
}
if (!btf->base_btf && btf_data_size == hdr->hdr_len) {
btf_verifier_log(env, "No data");
return -EINVAL;
}
return btf_check_sec_info(env, btf_data_size);
}
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
static const char *alloc_obj_fields[] = {
"bpf_spin_lock",
"bpf_list_head",
"bpf_list_node",
"bpf_rb_root",
"bpf_rb_node",
"bpf_refcount",
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
};
static struct btf_struct_metas *
btf_parse_struct_metas(struct bpf_verifier_log *log, struct btf *btf)
{
struct btf_struct_metas *tab = NULL;
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
struct btf_id_set *aof;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
int i, n, id, ret;
BUILD_BUG_ON(offsetof(struct btf_id_set, cnt) != 0);
BUILD_BUG_ON(sizeof(struct btf_id_set) != sizeof(u32));
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
aof = kmalloc(sizeof(*aof), GFP_KERNEL | __GFP_NOWARN);
if (!aof)
return ERR_PTR(-ENOMEM);
aof->cnt = 0;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
for (i = 0; i < ARRAY_SIZE(alloc_obj_fields); i++) {
/* Try to find whether this special type exists in user BTF, and
* if so remember its ID so we can easily find it among members
* of structs that we iterate in the next loop.
*/
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
struct btf_id_set *new_aof;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
id = btf_find_by_name_kind(btf, alloc_obj_fields[i], BTF_KIND_STRUCT);
if (id < 0)
continue;
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
new_aof = krealloc(aof, offsetof(struct btf_id_set, ids[aof->cnt + 1]),
GFP_KERNEL | __GFP_NOWARN);
if (!new_aof) {
ret = -ENOMEM;
goto free_aof;
}
aof = new_aof;
aof->ids[aof->cnt++] = id;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
}
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
n = btf_nr_types(btf);
for (i = 1; i < n; i++) {
/* Try to find if there are kptrs in user BTF and remember their ID */
struct btf_id_set *new_aof;
struct btf_field_info tmp;
const struct btf_type *t;
t = btf_type_by_id(btf, i);
if (!t) {
ret = -EINVAL;
goto free_aof;
}
ret = btf_find_kptr(btf, t, 0, 0, &tmp);
if (ret != BTF_FIELD_FOUND)
continue;
new_aof = krealloc(aof, offsetof(struct btf_id_set, ids[aof->cnt + 1]),
GFP_KERNEL | __GFP_NOWARN);
if (!new_aof) {
ret = -ENOMEM;
goto free_aof;
}
aof = new_aof;
aof->ids[aof->cnt++] = i;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
}
if (!aof->cnt) {
kfree(aof);
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
return NULL;
}
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
sort(&aof->ids, aof->cnt, sizeof(aof->ids[0]), btf_id_cmp_func, NULL);
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
for (i = 1; i < n; i++) {
struct btf_struct_metas *new_tab;
const struct btf_member *member;
struct btf_struct_meta *type;
struct btf_record *record;
const struct btf_type *t;
int j, tab_cnt;
t = btf_type_by_id(btf, i);
if (!__btf_type_is_struct(t))
continue;
cond_resched();
for_each_member(j, t, member) {
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
if (btf_id_set_contains(aof, member->type))
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
goto parse;
}
continue;
parse:
tab_cnt = tab ? tab->cnt : 0;
new_tab = krealloc(tab, offsetof(struct btf_struct_metas, types[tab_cnt + 1]),
GFP_KERNEL | __GFP_NOWARN);
if (!new_tab) {
ret = -ENOMEM;
goto free;
}
if (!tab)
new_tab->cnt = 0;
tab = new_tab;
type = &tab->types[tab->cnt];
type->btf_id = i;
record = btf_parse_fields(btf, t, BPF_SPIN_LOCK | BPF_LIST_HEAD | BPF_LIST_NODE |
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
BPF_RB_ROOT | BPF_RB_NODE | BPF_REFCOUNT |
BPF_KPTR, t->size);
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
/* The record cannot be unset, treat it as an error if so */
if (IS_ERR_OR_NULL(record)) {
ret = PTR_ERR_OR_ZERO(record) ?: -EFAULT;
goto free;
}
type->record = record;
tab->cnt++;
}
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
kfree(aof);
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
return tab;
free:
btf_struct_metas_free(tab);
bpf: Search for kptrs in prog BTF structs Currently btf_parse_fields is used in two places to create struct btf_record's for structs: when looking at mapval type, and when looking at any struct in program BTF. The former looks for kptr fields while the latter does not. This patch modifies the btf_parse_fields call made when looking at prog BTF struct types to search for kptrs as well. Before this series there was no reason to search for kptrs in non-mapval types: a referenced kptr needs some owner to guarantee resource cleanup, and map values were the only owner that supported this. If a struct with a kptr field were to have some non-kptr-aware owner, the kptr field might not be properly cleaned up and result in resources leaking. Only searching for kptr fields in mapval was a simple way to avoid this problem. In practice, though, searching for BPF_KPTR when populating struct_meta_tab does not expose us to this risk, as struct_meta_tab is only accessed through btf_find_struct_meta helper, and that helper is only called in contexts where recognizing the kptr field is safe: * PTR_TO_BTF_ID reg w/ MEM_ALLOC flag * Such a reg is a local kptr and must be free'd via bpf_obj_drop, which will correctly handle kptr field * When handling specific kfuncs which either expect MEM_ALLOC input or return MEM_ALLOC output (obj_{new,drop}, percpu_obj_{new,drop}, list+rbtree funcs, refcount_acquire) * Will correctly handle kptr field for same reasons as above * When looking at kptr pointee type * Called by functions which implement "correct kptr resource handling" * In btf_check_and_fixup_fields * Helper that ensures no ownership loops for lists and rbtrees, doesn't care about kptr field existence So we should be able to find BPF_KPTR fields in all prog BTF structs without leaking resources. Further patches in the series will build on this change to support kptr_xchg into non-mapval local kptr. Without this change there would be no kptr field found in such a type. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com> Signed-off-by: Amery Hung <amery.hung@bytedance.com> Link: https://lore.kernel.org/r/20240813212424.2871455-3-amery.hung@bytedance.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-08-14 05:24:21 +08:00
free_aof:
kfree(aof);
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
return ERR_PTR(ret);
}
struct btf_struct_meta *btf_find_struct_meta(const struct btf *btf, u32 btf_id)
{
struct btf_struct_metas *tab;
BUILD_BUG_ON(offsetof(struct btf_struct_meta, btf_id) != 0);
tab = btf->struct_meta_tab;
if (!tab)
return NULL;
return bsearch(&btf_id, tab->types, tab->cnt, sizeof(tab->types[0]), btf_id_cmp_func);
}
static int btf_check_type_tags(struct btf_verifier_env *env,
struct btf *btf, int start_id)
{
int i, n, good_id = start_id - 1;
bool in_tags;
n = btf_nr_types(btf);
for (i = start_id; i < n; i++) {
const struct btf_type *t;
int chain_limit = 32;
u32 cur_id = i;
t = btf_type_by_id(btf, i);
if (!t)
return -EINVAL;
if (!btf_type_is_modifier(t))
continue;
cond_resched();
in_tags = btf_type_is_type_tag(t);
while (btf_type_is_modifier(t)) {
if (!chain_limit--) {
btf_verifier_log(env, "Max chain length or cycle detected");
return -ELOOP;
}
if (btf_type_is_type_tag(t)) {
if (!in_tags) {
btf_verifier_log(env, "Type tags don't precede modifiers");
return -EINVAL;
}
} else if (in_tags) {
in_tags = false;
}
if (cur_id <= good_id)
break;
/* Move to next type */
cur_id = t->type;
t = btf_type_by_id(btf, cur_id);
if (!t)
return -EINVAL;
}
good_id = i;
}
return 0;
}
static int finalize_log(struct bpf_verifier_log *log, bpfptr_t uattr, u32 uattr_size)
{
u32 log_true_size;
int err;
err = bpf_vlog_finalize(log, &log_true_size);
if (uattr_size >= offsetofend(union bpf_attr, btf_log_true_size) &&
copy_to_bpfptr_offset(uattr, offsetof(union bpf_attr, btf_log_true_size),
&log_true_size, sizeof(log_true_size)))
err = -EFAULT;
return err;
}
static struct btf *btf_parse(const union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size)
{
bpfptr_t btf_data = make_bpfptr(attr->btf, uattr.is_kernel);
char __user *log_ubuf = u64_to_user_ptr(attr->btf_log_buf);
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
struct btf_struct_metas *struct_meta_tab;
struct btf_verifier_env *env = NULL;
struct btf *btf = NULL;
u8 *data;
int err, ret;
if (attr->btf_size > BTF_MAX_SIZE)
return ERR_PTR(-E2BIG);
env = kzalloc(sizeof(*env), GFP_KERNEL | __GFP_NOWARN);
if (!env)
return ERR_PTR(-ENOMEM);
/* user could have requested verbose verifier output
* and supplied buffer to store the verification trace
*/
err = bpf_vlog_init(&env->log, attr->btf_log_level,
log_ubuf, attr->btf_log_size);
if (err)
goto errout_free;
btf = kzalloc(sizeof(*btf), GFP_KERNEL | __GFP_NOWARN);
if (!btf) {
err = -ENOMEM;
goto errout;
}
env->btf = btf;
data = kvmalloc(attr->btf_size, GFP_KERNEL | __GFP_NOWARN);
if (!data) {
err = -ENOMEM;
goto errout;
}
btf->data = data;
btf->data_size = attr->btf_size;
if (copy_from_bpfptr(data, btf_data, attr->btf_size)) {
err = -EFAULT;
goto errout;
}
err = btf_parse_hdr(env);
if (err)
goto errout;
btf->nohdr_data = btf->data + btf->hdr.hdr_len;
err = btf_parse_str_sec(env);
if (err)
goto errout;
err = btf_parse_type_sec(env);
if (err)
goto errout;
err = btf_check_type_tags(env, btf, 1);
if (err)
goto errout;
struct_meta_tab = btf_parse_struct_metas(&env->log, btf);
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
if (IS_ERR(struct_meta_tab)) {
err = PTR_ERR(struct_meta_tab);
goto errout;
}
btf->struct_meta_tab = struct_meta_tab;
if (struct_meta_tab) {
int i;
for (i = 0; i < struct_meta_tab->cnt; i++) {
err = btf_check_and_fixup_fields(btf, struct_meta_tab->types[i].record);
if (err < 0)
goto errout_meta;
}
}
err = finalize_log(&env->log, uattr, uattr_size);
if (err)
goto errout_free;
btf_verifier_env_free(env);
refcount_set(&btf->refcnt, 1);
return btf;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
errout_meta:
btf_free_struct_meta_tab(btf);
errout:
/* overwrite err with -ENOSPC or -EFAULT */
ret = finalize_log(&env->log, uattr, uattr_size);
if (ret)
err = ret;
errout_free:
btf_verifier_env_free(env);
if (btf)
btf_free(btf);
return ERR_PTR(err);
}
extern char __start_BTF[];
extern char __stop_BTF[];
extern struct btf *btf_vmlinux;
#define BPF_MAP_TYPE(_id, _ops)
#define BPF_LINK_TYPE(_id, _name)
static union {
struct bpf_ctx_convert {
#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
prog_ctx_type _id##_prog; \
kern_ctx_type _id##_kern;
#include <linux/bpf_types.h>
#undef BPF_PROG_TYPE
} *__t;
/* 't' is written once under lock. Read many times. */
const struct btf_type *t;
} bpf_ctx_convert;
enum {
#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
__ctx_convert##_id,
#include <linux/bpf_types.h>
#undef BPF_PROG_TYPE
__ctx_convert_unused, /* to avoid empty enum in extreme .config */
};
static u8 bpf_ctx_convert_map[] = {
#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
[_id] = __ctx_convert##_id,
#include <linux/bpf_types.h>
#undef BPF_PROG_TYPE
0, /* avoid empty array */
};
#undef BPF_MAP_TYPE
#undef BPF_LINK_TYPE
static const struct btf_type *find_canonical_prog_ctx_type(enum bpf_prog_type prog_type)
{
const struct btf_type *conv_struct;
const struct btf_member *ctx_type;
conv_struct = bpf_ctx_convert.t;
if (!conv_struct)
return NULL;
/* prog_type is valid bpf program type. No need for bounds check. */
ctx_type = btf_type_member(conv_struct) + bpf_ctx_convert_map[prog_type] * 2;
/* ctx_type is a pointer to prog_ctx_type in vmlinux.
* Like 'struct __sk_buff'
*/
return btf_type_by_id(btf_vmlinux, ctx_type->type);
}
static int find_kern_ctx_type_id(enum bpf_prog_type prog_type)
{
const struct btf_type *conv_struct;
const struct btf_member *ctx_type;
conv_struct = bpf_ctx_convert.t;
if (!conv_struct)
return -EFAULT;
/* prog_type is valid bpf program type. No need for bounds check. */
ctx_type = btf_type_member(conv_struct) + bpf_ctx_convert_map[prog_type] * 2 + 1;
/* ctx_type is a pointer to prog_ctx_type in vmlinux.
* Like 'struct sk_buff'
*/
return ctx_type->type;
}
bool btf_is_projection_of(const char *pname, const char *tname)
{
if (strcmp(pname, "__sk_buff") == 0 && strcmp(tname, "sk_buff") == 0)
return true;
if (strcmp(pname, "xdp_md") == 0 && strcmp(tname, "xdp_buff") == 0)
return true;
return false;
}
bool btf_is_prog_ctx_type(struct bpf_verifier_log *log, const struct btf *btf,
const struct btf_type *t, enum bpf_prog_type prog_type,
int arg)
{
const struct btf_type *ctx_type;
const char *tname, *ctx_tname;
t = btf_type_by_id(btf, t->type);
/* KPROBE programs allow bpf_user_pt_regs_t typedef, which we need to
* check before we skip all the typedef below.
*/
if (prog_type == BPF_PROG_TYPE_KPROBE) {
while (btf_type_is_modifier(t) && !btf_type_is_typedef(t))
t = btf_type_by_id(btf, t->type);
if (btf_type_is_typedef(t)) {
tname = btf_name_by_offset(btf, t->name_off);
if (tname && strcmp(tname, "bpf_user_pt_regs_t") == 0)
return true;
}
}
while (btf_type_is_modifier(t))
t = btf_type_by_id(btf, t->type);
if (!btf_type_is_struct(t)) {
/* Only pointer to struct is supported for now.
* That means that BPF_PROG_TYPE_TRACEPOINT with BTF
* is not supported yet.
* BPF_PROG_TYPE_RAW_TRACEPOINT is fine.
*/
return false;
}
tname = btf_name_by_offset(btf, t->name_off);
if (!tname) {
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
bpf_log(log, "arg#%d struct doesn't have a name\n", arg);
return false;
}
ctx_type = find_canonical_prog_ctx_type(prog_type);
if (!ctx_type) {
bpf_log(log, "btf_vmlinux is malformed\n");
/* should not happen */
return false;
}
again:
ctx_tname = btf_name_by_offset(btf_vmlinux, ctx_type->name_off);
if (!ctx_tname) {
/* should not happen */
bpf_log(log, "Please fix kernel include/linux/bpf_types.h\n");
return false;
}
/* program types without named context types work only with arg:ctx tag */
if (ctx_tname[0] == '\0')
return false;
/* only compare that prog's ctx type name is the same as
* kernel expects. No need to compare field by field.
* It's ok for bpf prog to do:
* struct __sk_buff {};
* int socket_filter_bpf_prog(struct __sk_buff *skb)
* { // no fields of skb are ever used }
*/
if (btf_is_projection_of(ctx_tname, tname))
return true;
if (strcmp(ctx_tname, tname)) {
/* bpf_user_pt_regs_t is a typedef, so resolve it to
* underlying struct and check name again
*/
if (!btf_type_is_modifier(ctx_type))
return false;
while (btf_type_is_modifier(ctx_type))
ctx_type = btf_type_by_id(btf_vmlinux, ctx_type->type);
goto again;
}
return true;
}
bpf: enforce types for __arg_ctx-tagged arguments in global subprogs Add enforcement of expected types for context arguments tagged with arg:ctx (__arg_ctx) tag. First, any program type will accept generic `void *` context type when combined with __arg_ctx tag. Besides accepting "canonical" struct names and `void *`, for a bunch of program types for which program context is actually a named struct, we allows a bunch of pragmatic exceptions to match real-world and expected usage: - for both kprobes and perf_event we allow `bpf_user_pt_regs_t *` as canonical context argument type, where `bpf_user_pt_regs_t` is a *typedef*, not a struct; - for kprobes, we also always accept `struct pt_regs *`, as that's what actually is passed as a context to any kprobe program; - for perf_event, we resolve typedefs (unless it's `bpf_user_pt_regs_t`) down to actual struct type and accept `struct pt_regs *`, or `struct user_pt_regs *`, or `struct user_regs_struct *`, depending on the actual struct type kernel architecture points `bpf_user_pt_regs_t` typedef to; otherwise, canonical `struct bpf_perf_event_data *` is expected; - for raw_tp/raw_tp.w programs, `u64/long *` are accepted, as that's what's expected with BPF_PROG() usage; otherwise, canonical `struct bpf_raw_tracepoint_args *` is expected; - tp_btf supports both `struct bpf_raw_tracepoint_args *` and `u64 *` formats, both are coded as expections as tp_btf is actually a TRACING program type, which has no canonical context type; - iterator programs accept `struct bpf_iter__xxx *` structs, currently with no further iterator-type specific enforcement; - fentry/fexit/fmod_ret/lsm/struct_ops all accept `u64 *`; - classic tracepoint programs, as well as syscall and freplace programs allow any user-provided type. In all other cases kernel will enforce exact match of struct name to expected canonical type. And if user-provided type doesn't match that expectation, verifier will emit helpful message with expected type name. Note a bit unnatural way the check is done after processing all the arguments. This is done to avoid conflict between bpf and bpf-next trees. Once trees converge, a small follow up patch will place a simple btf_validate_prog_ctx_type() check into a proper ARG_PTR_TO_CTX branch (which bpf-next tree patch refactored already), removing duplicated arg:ctx detection logic. Suggested-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20240118033143.3384355-4-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2024-01-18 11:31:41 +08:00
/* forward declarations for arch-specific underlying types of
* bpf_user_pt_regs_t; this avoids the need for arch-specific #ifdef
* compilation guards below for BPF_PROG_TYPE_PERF_EVENT checks, but still
* works correctly with __builtin_types_compatible_p() on respective
* architectures
*/
struct user_regs_struct;
struct user_pt_regs;
static int btf_validate_prog_ctx_type(struct bpf_verifier_log *log, const struct btf *btf,
const struct btf_type *t, int arg,
enum bpf_prog_type prog_type,
enum bpf_attach_type attach_type)
{
const struct btf_type *ctx_type;
const char *tname, *ctx_tname;
if (!btf_is_ptr(t)) {
bpf_log(log, "arg#%d type isn't a pointer\n", arg);
return -EINVAL;
}
t = btf_type_by_id(btf, t->type);
/* KPROBE and PERF_EVENT programs allow bpf_user_pt_regs_t typedef */
if (prog_type == BPF_PROG_TYPE_KPROBE || prog_type == BPF_PROG_TYPE_PERF_EVENT) {
while (btf_type_is_modifier(t) && !btf_type_is_typedef(t))
t = btf_type_by_id(btf, t->type);
if (btf_type_is_typedef(t)) {
tname = btf_name_by_offset(btf, t->name_off);
if (tname && strcmp(tname, "bpf_user_pt_regs_t") == 0)
return 0;
}
}
/* all other program types don't use typedefs for context type */
while (btf_type_is_modifier(t))
t = btf_type_by_id(btf, t->type);
/* `void *ctx __arg_ctx` is always valid */
if (btf_type_is_void(t))
return 0;
tname = btf_name_by_offset(btf, t->name_off);
if (str_is_empty(tname)) {
bpf_log(log, "arg#%d type doesn't have a name\n", arg);
return -EINVAL;
}
/* special cases */
switch (prog_type) {
case BPF_PROG_TYPE_KPROBE:
if (__btf_type_is_struct(t) && strcmp(tname, "pt_regs") == 0)
return 0;
break;
case BPF_PROG_TYPE_PERF_EVENT:
if (__builtin_types_compatible_p(bpf_user_pt_regs_t, struct pt_regs) &&
__btf_type_is_struct(t) && strcmp(tname, "pt_regs") == 0)
return 0;
if (__builtin_types_compatible_p(bpf_user_pt_regs_t, struct user_pt_regs) &&
__btf_type_is_struct(t) && strcmp(tname, "user_pt_regs") == 0)
return 0;
if (__builtin_types_compatible_p(bpf_user_pt_regs_t, struct user_regs_struct) &&
__btf_type_is_struct(t) && strcmp(tname, "user_regs_struct") == 0)
return 0;
break;
case BPF_PROG_TYPE_RAW_TRACEPOINT:
case BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE:
/* allow u64* as ctx */
if (btf_is_int(t) && t->size == 8)
return 0;
break;
case BPF_PROG_TYPE_TRACING:
switch (attach_type) {
case BPF_TRACE_RAW_TP:
/* tp_btf program is TRACING, so need special case here */
if (__btf_type_is_struct(t) &&
strcmp(tname, "bpf_raw_tracepoint_args") == 0)
return 0;
/* allow u64* as ctx */
if (btf_is_int(t) && t->size == 8)
return 0;
break;
case BPF_TRACE_ITER:
/* allow struct bpf_iter__xxx types only */
if (__btf_type_is_struct(t) &&
strncmp(tname, "bpf_iter__", sizeof("bpf_iter__") - 1) == 0)
return 0;
break;
case BPF_TRACE_FENTRY:
case BPF_TRACE_FEXIT:
case BPF_MODIFY_RETURN:
/* allow u64* as ctx */
if (btf_is_int(t) && t->size == 8)
return 0;
break;
default:
break;
}
break;
case BPF_PROG_TYPE_LSM:
case BPF_PROG_TYPE_STRUCT_OPS:
/* allow u64* as ctx */
if (btf_is_int(t) && t->size == 8)
return 0;
break;
case BPF_PROG_TYPE_TRACEPOINT:
case BPF_PROG_TYPE_SYSCALL:
case BPF_PROG_TYPE_EXT:
return 0; /* anything goes */
default:
break;
}
ctx_type = find_canonical_prog_ctx_type(prog_type);
if (!ctx_type) {
/* should not happen */
bpf_log(log, "btf_vmlinux is malformed\n");
return -EINVAL;
}
/* resolve typedefs and check that underlying structs are matching as well */
while (btf_type_is_modifier(ctx_type))
ctx_type = btf_type_by_id(btf_vmlinux, ctx_type->type);
/* if program type doesn't have distinctly named struct type for
* context, then __arg_ctx argument can only be `void *`, which we
* already checked above
*/
if (!__btf_type_is_struct(ctx_type)) {
bpf_log(log, "arg#%d should be void pointer\n", arg);
return -EINVAL;
}
ctx_tname = btf_name_by_offset(btf_vmlinux, ctx_type->name_off);
if (!__btf_type_is_struct(t) || strcmp(ctx_tname, tname) != 0) {
bpf_log(log, "arg#%d should be `struct %s *`\n", arg, ctx_tname);
return -EINVAL;
}
return 0;
}
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
static int btf_translate_to_vmlinux(struct bpf_verifier_log *log,
struct btf *btf,
const struct btf_type *t,
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
enum bpf_prog_type prog_type,
int arg)
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
{
if (!btf_is_prog_ctx_type(log, btf, t, prog_type, arg))
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
return -ENOENT;
return find_kern_ctx_type_id(prog_type);
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
}
int get_kern_ctx_btf_id(struct bpf_verifier_log *log, enum bpf_prog_type prog_type)
{
const struct btf_member *kctx_member;
const struct btf_type *conv_struct;
const struct btf_type *kctx_type;
u32 kctx_type_id;
conv_struct = bpf_ctx_convert.t;
/* get member for kernel ctx type */
kctx_member = btf_type_member(conv_struct) + bpf_ctx_convert_map[prog_type] * 2 + 1;
kctx_type_id = kctx_member->type;
kctx_type = btf_type_by_id(btf_vmlinux, kctx_type_id);
if (!btf_type_is_struct(kctx_type)) {
bpf_log(log, "kern ctx type id %u is not a struct\n", kctx_type_id);
return -EINVAL;
}
return kctx_type_id;
}
BTF_ID_LIST(bpf_ctx_convert_btf_id)
BTF_ID(struct, bpf_ctx_convert)
static struct btf *btf_parse_base(struct btf_verifier_env *env, const char *name,
void *data, unsigned int data_size)
{
struct btf *btf = NULL;
int err;
if (!IS_ENABLED(CONFIG_DEBUG_INFO_BTF))
return ERR_PTR(-ENOENT);
btf = kzalloc(sizeof(*btf), GFP_KERNEL | __GFP_NOWARN);
if (!btf) {
err = -ENOMEM;
goto errout;
}
env->btf = btf;
btf->data = data;
btf->data_size = data_size;
btf->kernel_btf = true;
snprintf(btf->name, sizeof(btf->name), "%s", name);
err = btf_parse_hdr(env);
if (err)
goto errout;
btf->nohdr_data = btf->data + btf->hdr.hdr_len;
err = btf_parse_str_sec(env);
if (err)
goto errout;
err = btf_check_all_metas(env);
if (err)
goto errout;
err = btf_check_type_tags(env, btf, 1);
if (err)
goto errout;
refcount_set(&btf->refcnt, 1);
return btf;
errout:
if (btf) {
kvfree(btf->types);
kfree(btf);
}
return ERR_PTR(err);
}
struct btf *btf_parse_vmlinux(void)
{
struct btf_verifier_env *env = NULL;
struct bpf_verifier_log *log;
struct btf *btf;
int err;
env = kzalloc(sizeof(*env), GFP_KERNEL | __GFP_NOWARN);
if (!env)
return ERR_PTR(-ENOMEM);
log = &env->log;
log->level = BPF_LOG_KERNEL;
btf = btf_parse_base(env, "vmlinux", __start_BTF, __stop_BTF - __start_BTF);
if (IS_ERR(btf))
goto err_out;
/* btf_parse_vmlinux() runs under bpf_verifier_lock */
bpf_ctx_convert.t = btf_type_by_id(btf, bpf_ctx_convert_btf_id[0]);
err = btf_alloc_id(btf);
if (err) {
btf_free(btf);
btf = ERR_PTR(err);
}
err_out:
btf_verifier_env_free(env);
return btf;
}
/* If .BTF_ids section was created with distilled base BTF, both base and
* split BTF ids will need to be mapped to actual base/split ids for
* BTF now that it has been relocated.
*/
static __u32 btf_relocate_id(const struct btf *btf, __u32 id)
{
if (!btf->base_btf || !btf->base_id_map)
return id;
return btf->base_id_map[id];
}
#ifdef CONFIG_DEBUG_INFO_BTF_MODULES
static struct btf *btf_parse_module(const char *module_name, const void *data,
unsigned int data_size, void *base_data,
unsigned int base_data_size)
{
struct btf *btf = NULL, *vmlinux_btf, *base_btf = NULL;
struct btf_verifier_env *env = NULL;
struct bpf_verifier_log *log;
int err = 0;
vmlinux_btf = bpf_get_btf_vmlinux();
if (IS_ERR(vmlinux_btf))
return vmlinux_btf;
if (!vmlinux_btf)
return ERR_PTR(-EINVAL);
env = kzalloc(sizeof(*env), GFP_KERNEL | __GFP_NOWARN);
if (!env)
return ERR_PTR(-ENOMEM);
log = &env->log;
log->level = BPF_LOG_KERNEL;
if (base_data) {
base_btf = btf_parse_base(env, ".BTF.base", base_data, base_data_size);
if (IS_ERR(base_btf)) {
err = PTR_ERR(base_btf);
goto errout;
}
} else {
base_btf = vmlinux_btf;
}
btf = kzalloc(sizeof(*btf), GFP_KERNEL | __GFP_NOWARN);
if (!btf) {
err = -ENOMEM;
goto errout;
}
env->btf = btf;
btf->base_btf = base_btf;
btf->start_id = base_btf->nr_types;
btf->start_str_off = base_btf->hdr.str_len;
btf->kernel_btf = true;
snprintf(btf->name, sizeof(btf->name), "%s", module_name);
btf->data = kvmemdup(data, data_size, GFP_KERNEL | __GFP_NOWARN);
if (!btf->data) {
err = -ENOMEM;
goto errout;
}
btf->data_size = data_size;
err = btf_parse_hdr(env);
if (err)
goto errout;
btf->nohdr_data = btf->data + btf->hdr.hdr_len;
err = btf_parse_str_sec(env);
if (err)
goto errout;
err = btf_check_all_metas(env);
if (err)
goto errout;
err = btf_check_type_tags(env, btf, btf_nr_types(base_btf));
if (err)
goto errout;
if (base_btf != vmlinux_btf) {
err = btf_relocate(btf, vmlinux_btf, &btf->base_id_map);
if (err)
goto errout;
btf_free(base_btf);
base_btf = vmlinux_btf;
}
btf_verifier_env_free(env);
refcount_set(&btf->refcnt, 1);
return btf;
errout:
btf_verifier_env_free(env);
if (!IS_ERR(base_btf) && base_btf != vmlinux_btf)
btf_free(base_btf);
if (btf) {
kvfree(btf->data);
kvfree(btf->types);
kfree(btf);
}
return ERR_PTR(err);
}
#endif /* CONFIG_DEBUG_INFO_BTF_MODULES */
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
struct btf *bpf_prog_get_target_btf(const struct bpf_prog *prog)
{
struct bpf_prog *tgt_prog = prog->aux->dst_prog;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
if (tgt_prog)
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
return tgt_prog->aux->btf;
else
return prog->aux->attach_btf;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
}
static bool is_int_ptr(struct btf *btf, const struct btf_type *t)
{
/* skip modifiers */
t = btf_type_skip_modifiers(btf, t->type, NULL);
return btf_type_is_int(t);
}
static u32 get_ctx_arg_idx(struct btf *btf, const struct btf_type *func_proto,
int off)
{
const struct btf_param *args;
const struct btf_type *t;
u32 offset = 0, nr_args;
int i;
if (!func_proto)
return off / 8;
nr_args = btf_type_vlen(func_proto);
args = (const struct btf_param *)(func_proto + 1);
for (i = 0; i < nr_args; i++) {
t = btf_type_skip_modifiers(btf, args[i].type, NULL);
offset += btf_type_is_ptr(t) ? 8 : roundup(t->size, 8);
if (off < offset)
return i;
}
t = btf_type_skip_modifiers(btf, func_proto->type, NULL);
offset += btf_type_is_ptr(t) ? 8 : roundup(t->size, 8);
if (off < offset)
return nr_args;
return nr_args + 1;
}
static bool prog_args_trusted(const struct bpf_prog *prog)
bpf: Allow trusted pointers to be passed to KF_TRUSTED_ARGS kfuncs Kfuncs currently support specifying the KF_TRUSTED_ARGS flag to signal to the verifier that it should enforce that a BPF program passes it a "safe", trusted pointer. Currently, "safe" means that the pointer is either PTR_TO_CTX, or is refcounted. There may be cases, however, where the kernel passes a BPF program a safe / trusted pointer to an object that the BPF program wishes to use as a kptr, but because the object does not yet have a ref_obj_id from the perspective of the verifier, the program would be unable to pass it to a KF_ACQUIRE | KF_TRUSTED_ARGS kfunc. The solution is to expand the set of pointers that are considered trusted according to KF_TRUSTED_ARGS, so that programs can invoke kfuncs with these pointers without getting rejected by the verifier. There is already a PTR_UNTRUSTED flag that is set in some scenarios, such as when a BPF program reads a kptr directly from a map without performing a bpf_kptr_xchg() call. These pointers of course can and should be rejected by the verifier. Unfortunately, however, PTR_UNTRUSTED does not cover all the cases for safety that need to be addressed to adequately protect kfuncs. Specifically, pointers obtained by a BPF program "walking" a struct are _not_ considered PTR_UNTRUSTED according to BPF. For example, say that we were to add a kfunc called bpf_task_acquire(), with KF_ACQUIRE | KF_TRUSTED_ARGS, to acquire a struct task_struct *. If we only used PTR_UNTRUSTED to signal that a task was unsafe to pass to a kfunc, the verifier would mistakenly allow the following unsafe BPF program to be loaded: SEC("tp_btf/task_newtask") int BPF_PROG(unsafe_acquire_task, struct task_struct *task, u64 clone_flags) { struct task_struct *acquired, *nested; nested = task->last_wakee; /* Would not be rejected by the verifier. */ acquired = bpf_task_acquire(nested); if (!acquired) return 0; bpf_task_release(acquired); return 0; } To address this, this patch defines a new type flag called PTR_TRUSTED which tracks whether a PTR_TO_BTF_ID pointer is safe to pass to a KF_TRUSTED_ARGS kfunc or a BPF helper function. PTR_TRUSTED pointers are passed directly from the kernel as a tracepoint or struct_ops callback argument. Any nested pointer that is obtained from walking a PTR_TRUSTED pointer is no longer PTR_TRUSTED. From the example above, the struct task_struct *task argument is PTR_TRUSTED, but the 'nested' pointer obtained from 'task->last_wakee' is not PTR_TRUSTED. A subsequent patch will add kfuncs for storing a task kfunc as a kptr, and then another patch will add selftests to validate. Signed-off-by: David Vernet <void@manifault.com> Link: https://lore.kernel.org/r/20221120051004.3605026-3-void@manifault.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-20 13:10:02 +08:00
{
enum bpf_attach_type atype = prog->expected_attach_type;
switch (prog->type) {
case BPF_PROG_TYPE_TRACING:
return atype == BPF_TRACE_RAW_TP || atype == BPF_TRACE_ITER;
case BPF_PROG_TYPE_LSM:
return bpf_lsm_is_trusted(prog);
case BPF_PROG_TYPE_STRUCT_OPS:
return true;
default:
return false;
}
bpf: Allow trusted pointers to be passed to KF_TRUSTED_ARGS kfuncs Kfuncs currently support specifying the KF_TRUSTED_ARGS flag to signal to the verifier that it should enforce that a BPF program passes it a "safe", trusted pointer. Currently, "safe" means that the pointer is either PTR_TO_CTX, or is refcounted. There may be cases, however, where the kernel passes a BPF program a safe / trusted pointer to an object that the BPF program wishes to use as a kptr, but because the object does not yet have a ref_obj_id from the perspective of the verifier, the program would be unable to pass it to a KF_ACQUIRE | KF_TRUSTED_ARGS kfunc. The solution is to expand the set of pointers that are considered trusted according to KF_TRUSTED_ARGS, so that programs can invoke kfuncs with these pointers without getting rejected by the verifier. There is already a PTR_UNTRUSTED flag that is set in some scenarios, such as when a BPF program reads a kptr directly from a map without performing a bpf_kptr_xchg() call. These pointers of course can and should be rejected by the verifier. Unfortunately, however, PTR_UNTRUSTED does not cover all the cases for safety that need to be addressed to adequately protect kfuncs. Specifically, pointers obtained by a BPF program "walking" a struct are _not_ considered PTR_UNTRUSTED according to BPF. For example, say that we were to add a kfunc called bpf_task_acquire(), with KF_ACQUIRE | KF_TRUSTED_ARGS, to acquire a struct task_struct *. If we only used PTR_UNTRUSTED to signal that a task was unsafe to pass to a kfunc, the verifier would mistakenly allow the following unsafe BPF program to be loaded: SEC("tp_btf/task_newtask") int BPF_PROG(unsafe_acquire_task, struct task_struct *task, u64 clone_flags) { struct task_struct *acquired, *nested; nested = task->last_wakee; /* Would not be rejected by the verifier. */ acquired = bpf_task_acquire(nested); if (!acquired) return 0; bpf_task_release(acquired); return 0; } To address this, this patch defines a new type flag called PTR_TRUSTED which tracks whether a PTR_TO_BTF_ID pointer is safe to pass to a KF_TRUSTED_ARGS kfunc or a BPF helper function. PTR_TRUSTED pointers are passed directly from the kernel as a tracepoint or struct_ops callback argument. Any nested pointer that is obtained from walking a PTR_TRUSTED pointer is no longer PTR_TRUSTED. From the example above, the struct task_struct *task argument is PTR_TRUSTED, but the 'nested' pointer obtained from 'task->last_wakee' is not PTR_TRUSTED. A subsequent patch will add kfuncs for storing a task kfunc as a kptr, and then another patch will add selftests to validate. Signed-off-by: David Vernet <void@manifault.com> Link: https://lore.kernel.org/r/20221120051004.3605026-3-void@manifault.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-20 13:10:02 +08:00
}
bpf: Create argument information for nullable arguments. Collect argument information from the type information of stub functions to mark arguments of BPF struct_ops programs with PTR_MAYBE_NULL if they are nullable. A nullable argument is annotated by suffixing "__nullable" at the argument name of stub function. For nullable arguments, this patch sets a struct bpf_ctx_arg_aux to label their reg_type with PTR_TO_BTF_ID | PTR_TRUSTED | PTR_MAYBE_NULL. This makes the verifier to check programs and ensure that they properly check the pointer. The programs should check if the pointer is null before accessing the pointed memory. The implementer of a struct_ops type should annotate the arguments that can be null. The implementer should define a stub function (empty) as a placeholder for each defined operator. The name of a stub function should be in the pattern "<st_op_type>__<operator name>". For example, for test_maybe_null of struct bpf_testmod_ops, it's stub function name should be "bpf_testmod_ops__test_maybe_null". You mark an argument nullable by suffixing the argument name with "__nullable" at the stub function. Since we already has stub functions for kCFI, we just reuse these stub functions with the naming convention mentioned earlier. These stub functions with the naming convention is only required if there are nullable arguments to annotate. For functions having not nullable arguments, stub functions are not necessary for the purpose of this patch. This patch will prepare a list of struct bpf_ctx_arg_aux, aka arg_info, for each member field of a struct_ops type. "arg_info" will be assigned to "prog->aux->ctx_arg_info" of BPF struct_ops programs in check_struct_ops_btf_id() so that it can be used by btf_ctx_access() later to set reg_type properly for the verifier. Signed-off-by: Kui-Feng Lee <thinker.li@gmail.com> Link: https://lore.kernel.org/r/20240209023750.1153905-4-thinker.li@gmail.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2024-02-09 10:37:49 +08:00
int btf_ctx_arg_offset(const struct btf *btf, const struct btf_type *func_proto,
u32 arg_no)
{
const struct btf_param *args;
const struct btf_type *t;
int off = 0, i;
u32 sz;
args = btf_params(func_proto);
for (i = 0; i < arg_no; i++) {
t = btf_type_by_id(btf, args[i].type);
t = btf_resolve_size(btf, t, &sz);
if (IS_ERR(t))
return PTR_ERR(t);
off += roundup(sz, 8);
}
return off;
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
bool btf_ctx_access(int off, int size, enum bpf_access_type type,
const struct bpf_prog *prog,
struct bpf_insn_access_aux *info)
{
const struct btf_type *t = prog->aux->attach_func_proto;
struct bpf_prog *tgt_prog = prog->aux->dst_prog;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
struct btf *btf = bpf_prog_get_target_btf(prog);
const char *tname = prog->aux->attach_func_name;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
struct bpf_verifier_log *log = info->log;
const struct btf_param *args;
const char *tag_value;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
u32 nr_args, arg;
int i, ret;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
if (off % 8) {
bpf_log(log, "func '%s' offset %d is not multiple of 8\n",
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
tname, off);
return false;
}
arg = get_ctx_arg_idx(btf, t, off);
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
args = (const struct btf_param *)(t + 1);
/* if (t == NULL) Fall back to default BPF prog with
* MAX_BPF_FUNC_REG_ARGS u64 arguments.
*/
nr_args = t ? btf_type_vlen(t) : MAX_BPF_FUNC_REG_ARGS;
if (prog->aux->attach_btf_trace) {
/* skip first 'void *__data' argument in btf_trace_##name typedef */
args++;
nr_args--;
}
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
if (arg > nr_args) {
bpf_log(log, "func '%s' doesn't have %d-th argument\n",
tname, arg + 1);
return false;
}
if (arg == nr_args) {
switch (prog->expected_attach_type) {
case BPF_LSM_MAC:
/* mark we are accessing the return value */
info->is_retval = true;
fallthrough;
case BPF_LSM_CGROUP:
case BPF_TRACE_FEXIT:
/* When LSM programs are attached to void LSM hooks
* they use FEXIT trampolines and when attached to
* int LSM hooks, they use MODIFY_RETURN trampolines.
*
* While the LSM programs are BPF_MODIFY_RETURN-like
* the check:
*
* if (ret_type != 'int')
* return -EINVAL;
*
* is _not_ done here. This is still safe as LSM hooks
* have only void and int return types.
*/
if (!t)
return true;
t = btf_type_by_id(btf, t->type);
break;
case BPF_MODIFY_RETURN:
/* For now the BPF_MODIFY_RETURN can only be attached to
* functions that return an int.
*/
if (!t)
return false;
t = btf_type_skip_modifiers(btf, t->type, NULL);
if (!btf_type_is_small_int(t)) {
bpf_log(log,
"ret type %s not allowed for fmod_ret\n",
btf_type_str(t));
return false;
}
break;
default:
bpf_log(log, "func '%s' doesn't have %d-th argument\n",
tname, arg + 1);
return false;
}
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
} else {
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
if (!t)
/* Default prog with MAX_BPF_FUNC_REG_ARGS args */
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
return true;
t = btf_type_by_id(btf, args[arg].type);
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* skip modifiers */
while (btf_type_is_modifier(t))
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
t = btf_type_by_id(btf, t->type);
if (btf_type_is_small_int(t) || btf_is_any_enum(t) || __btf_type_is_struct(t))
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* accessing a scalar */
return true;
if (!btf_type_is_ptr(t)) {
bpf_log(log,
"func '%s' arg%d '%s' has type %s. Only pointer access is allowed\n",
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
tname, arg,
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
__btf_name_by_offset(btf, t->name_off),
btf_type_str(t));
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
return false;
}
/* check for PTR_TO_RDONLY_BUF_OR_NULL or PTR_TO_RDWR_BUF_OR_NULL */
for (i = 0; i < prog->aux->ctx_arg_info_size; i++) {
const struct bpf_ctx_arg_aux *ctx_arg_info = &prog->aux->ctx_arg_info[i];
u32 type, flag;
type = base_type(ctx_arg_info->reg_type);
flag = type_flag(ctx_arg_info->reg_type);
if (ctx_arg_info->offset == off && type == PTR_TO_BUF &&
(flag & PTR_MAYBE_NULL)) {
info->reg_type = ctx_arg_info->reg_type;
return true;
}
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
if (t->type == 0)
/* This is a pointer to void.
* It is the same as scalar from the verifier safety pov.
* No further pointer walking is allowed.
*/
return true;
if (is_int_ptr(btf, t))
return true;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* this is a pointer to another type */
for (i = 0; i < prog->aux->ctx_arg_info_size; i++) {
const struct bpf_ctx_arg_aux *ctx_arg_info = &prog->aux->ctx_arg_info[i];
if (ctx_arg_info->offset == off) {
bpf: Emit better log message if bpf_iter ctx arg btf_id == 0 To avoid kernel build failure due to some missing .BTF-ids referenced functions/types, the patch ([1]) tries to fill btf_id 0 for these types. In bpf verifier, for percpu variable and helper returning btf_id cases, verifier already emitted proper warning with something like verbose(env, "Helper has invalid btf_id in R%d\n", regno); verbose(env, "invalid return type %d of func %s#%d\n", fn->ret_type, func_id_name(func_id), func_id); But this is not the case for bpf_iter context arguments. I hacked resolve_btfids to encode btf_id 0 for struct task_struct. With `./test_progs -n 7/5`, I got, 0: (79) r2 = *(u64 *)(r1 +0) func 'bpf_iter_task' arg0 has btf_id 29739 type STRUCT 'bpf_iter_meta' ; struct seq_file *seq = ctx->meta->seq; 1: (79) r6 = *(u64 *)(r2 +0) ; struct task_struct *task = ctx->task; 2: (79) r7 = *(u64 *)(r1 +8) ; if (task == (void *)0) { 3: (55) if r7 != 0x0 goto pc+11 ... ; BPF_SEQ_PRINTF(seq, "%8d %8d\n", task->tgid, task->pid); 26: (61) r1 = *(u32 *)(r7 +1372) Type '(anon)' is not a struct Basically, verifier will return btf_id 0 for task_struct. Later on, when the code tries to access task->tgid, the verifier correctly complains the type is '(anon)' and it is not a struct. Users still need to backtrace to find out what is going on. Let us catch the invalid btf_id 0 earlier and provide better message indicating btf_id is wrong. The new error message looks like below: R1 type=ctx expected=fp ; struct seq_file *seq = ctx->meta->seq; 0: (79) r2 = *(u64 *)(r1 +0) func 'bpf_iter_task' arg0 has btf_id 29739 type STRUCT 'bpf_iter_meta' ; struct seq_file *seq = ctx->meta->seq; 1: (79) r6 = *(u64 *)(r2 +0) ; struct task_struct *task = ctx->task; 2: (79) r7 = *(u64 *)(r1 +8) invalid btf_id for context argument offset 8 invalid bpf_context access off=8 size=8 [1] https://lore.kernel.org/bpf/20210727132532.2473636-1-hengqi.chen@gmail.com/ Signed-off-by: Yonghong Song <yhs@fb.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/bpf/20210728183025.1461750-1-yhs@fb.com
2021-07-29 02:30:25 +08:00
if (!ctx_arg_info->btf_id) {
bpf_log(log,"invalid btf_id for context argument offset %u\n", off);
return false;
}
info->reg_type = ctx_arg_info->reg_type;
info->btf = ctx_arg_info->btf ? : btf_vmlinux;
info->btf_id = ctx_arg_info->btf_id;
return true;
}
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
info->reg_type = PTR_TO_BTF_ID;
if (prog_args_trusted(prog))
bpf: Allow trusted pointers to be passed to KF_TRUSTED_ARGS kfuncs Kfuncs currently support specifying the KF_TRUSTED_ARGS flag to signal to the verifier that it should enforce that a BPF program passes it a "safe", trusted pointer. Currently, "safe" means that the pointer is either PTR_TO_CTX, or is refcounted. There may be cases, however, where the kernel passes a BPF program a safe / trusted pointer to an object that the BPF program wishes to use as a kptr, but because the object does not yet have a ref_obj_id from the perspective of the verifier, the program would be unable to pass it to a KF_ACQUIRE | KF_TRUSTED_ARGS kfunc. The solution is to expand the set of pointers that are considered trusted according to KF_TRUSTED_ARGS, so that programs can invoke kfuncs with these pointers without getting rejected by the verifier. There is already a PTR_UNTRUSTED flag that is set in some scenarios, such as when a BPF program reads a kptr directly from a map without performing a bpf_kptr_xchg() call. These pointers of course can and should be rejected by the verifier. Unfortunately, however, PTR_UNTRUSTED does not cover all the cases for safety that need to be addressed to adequately protect kfuncs. Specifically, pointers obtained by a BPF program "walking" a struct are _not_ considered PTR_UNTRUSTED according to BPF. For example, say that we were to add a kfunc called bpf_task_acquire(), with KF_ACQUIRE | KF_TRUSTED_ARGS, to acquire a struct task_struct *. If we only used PTR_UNTRUSTED to signal that a task was unsafe to pass to a kfunc, the verifier would mistakenly allow the following unsafe BPF program to be loaded: SEC("tp_btf/task_newtask") int BPF_PROG(unsafe_acquire_task, struct task_struct *task, u64 clone_flags) { struct task_struct *acquired, *nested; nested = task->last_wakee; /* Would not be rejected by the verifier. */ acquired = bpf_task_acquire(nested); if (!acquired) return 0; bpf_task_release(acquired); return 0; } To address this, this patch defines a new type flag called PTR_TRUSTED which tracks whether a PTR_TO_BTF_ID pointer is safe to pass to a KF_TRUSTED_ARGS kfunc or a BPF helper function. PTR_TRUSTED pointers are passed directly from the kernel as a tracepoint or struct_ops callback argument. Any nested pointer that is obtained from walking a PTR_TRUSTED pointer is no longer PTR_TRUSTED. From the example above, the struct task_struct *task argument is PTR_TRUSTED, but the 'nested' pointer obtained from 'task->last_wakee' is not PTR_TRUSTED. A subsequent patch will add kfuncs for storing a task kfunc as a kptr, and then another patch will add selftests to validate. Signed-off-by: David Vernet <void@manifault.com> Link: https://lore.kernel.org/r/20221120051004.3605026-3-void@manifault.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-20 13:10:02 +08:00
info->reg_type |= PTR_TRUSTED;
if (btf_param_match_suffix(btf, &args[arg], "__nullable"))
info->reg_type |= PTR_MAYBE_NULL;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
if (tgt_prog) {
bpf: Fix context type resolving for extension programs Eelco reported we can't properly access arguments if the tracing program is attached to extension program. Having following program: SEC("classifier/test_pkt_md_access") int test_pkt_md_access(struct __sk_buff *skb) with its extension: SEC("freplace/test_pkt_md_access") int test_pkt_md_access_new(struct __sk_buff *skb) and tracing that extension with: SEC("fentry/test_pkt_md_access_new") int BPF_PROG(fentry, struct sk_buff *skb) It's not possible to access skb argument in the fentry program, with following error from verifier: ; int BPF_PROG(fentry, struct sk_buff *skb) 0: (79) r1 = *(u64 *)(r1 +0) invalid bpf_context access off=0 size=8 The problem is that btf_ctx_access gets the context type for the traced program, which is in this case the extension. But when we trace extension program, we want to get the context type of the program that the extension is attached to, so we can access the argument properly in the trace program. This version of the patch is tweaked slightly from Jiri's original one, since the refactoring in the previous patches means we have to get the target prog type from the new variable in prog->aux instead of directly from the target prog. Reported-by: Eelco Chaudron <echaudro@redhat.com> Suggested-by: Jiri Olsa <jolsa@kernel.org> Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Link: https://lore.kernel.org/bpf/160138355278.48470.17057040257274725638.stgit@toke.dk
2020-09-29 20:45:52 +08:00
enum bpf_prog_type tgt_type;
if (tgt_prog->type == BPF_PROG_TYPE_EXT)
tgt_type = tgt_prog->aux->saved_dst_prog_type;
else
tgt_type = tgt_prog->type;
ret = btf_translate_to_vmlinux(log, btf, t, tgt_type, arg);
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
if (ret > 0) {
info->btf = btf_vmlinux;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
info->btf_id = ret;
return true;
} else {
return false;
}
}
info->btf = btf;
info->btf_id = t->type;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
t = btf_type_by_id(btf, t->type);
if (btf_type_is_type_tag(t)) {
tag_value = __btf_name_by_offset(btf, t->name_off);
if (strcmp(tag_value, "user") == 0)
info->reg_type |= MEM_USER;
if (strcmp(tag_value, "percpu") == 0)
info->reg_type |= MEM_PERCPU;
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* skip modifiers */
while (btf_type_is_modifier(t)) {
info->btf_id = t->type;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
t = btf_type_by_id(btf, t->type);
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
if (!btf_type_is_struct(t)) {
bpf_log(log,
"func '%s' arg%d type %s is not a struct\n",
tname, arg, btf_type_str(t));
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
return false;
}
bpf_log(log, "func '%s' arg%d has btf_id %d type %s '%s'\n",
tname, arg, info->btf_id, btf_type_str(t),
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
__btf_name_by_offset(btf, t->name_off));
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
return true;
}
EXPORT_SYMBOL_GPL(btf_ctx_access);
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
enum bpf_struct_walk_result {
/* < 0 error */
WALK_SCALAR = 0,
WALK_PTR,
WALK_STRUCT,
};
static int btf_struct_walk(struct bpf_verifier_log *log, const struct btf *btf,
const struct btf_type *t, int off, int size,
u32 *next_btf_id, enum bpf_type_flag *flag,
const char **field_name)
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
{
u32 i, moff, mtrue_end, msize = 0, total_nelems = 0;
const struct btf_type *mtype, *elem_type = NULL;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
const struct btf_member *member;
const char *tname, *mname, *tag_value;
u32 vlen, elem_id, mid;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
again:
if (btf_type_is_modifier(t))
t = btf_type_skip_modifiers(btf, t->type, NULL);
tname = __btf_name_by_offset(btf, t->name_off);
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
if (!btf_type_is_struct(t)) {
bpf_log(log, "Type '%s' is not a struct\n", tname);
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
return -EINVAL;
}
vlen = btf_type_vlen(t);
if (BTF_INFO_KIND(t->info) == BTF_KIND_UNION && vlen != 1 && !(*flag & PTR_UNTRUSTED))
/*
* walking unions yields untrusted pointers
* with exception of __bpf_md_ptr and other
* unions with a single member
*/
*flag |= PTR_UNTRUSTED;
if (off + size > t->size) {
/* If the last element is a variable size array, we may
* need to relax the rule.
*/
struct btf_array *array_elem;
if (vlen == 0)
goto error;
member = btf_type_member(t) + vlen - 1;
mtype = btf_type_skip_modifiers(btf, member->type,
NULL);
if (!btf_type_is_array(mtype))
goto error;
array_elem = (struct btf_array *)(mtype + 1);
if (array_elem->nelems != 0)
goto error;
moff = __btf_member_bit_offset(t, member) / 8;
if (off < moff)
goto error;
/* allow structure and integer */
t = btf_type_skip_modifiers(btf, array_elem->type,
NULL);
if (btf_type_is_int(t))
return WALK_SCALAR;
if (!btf_type_is_struct(t))
goto error;
off = (off - moff) % t->size;
goto again;
error:
bpf_log(log, "access beyond struct %s at off %u size %u\n",
tname, off, size);
return -EACCES;
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
for_each_member(i, t, member) {
/* offset of the field in bytes */
moff = __btf_member_bit_offset(t, member) / 8;
if (off + size <= moff)
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* won't find anything, field is already too far */
break;
if (__btf_member_bitfield_size(t, member)) {
u32 end_bit = __btf_member_bit_offset(t, member) +
__btf_member_bitfield_size(t, member);
/* off <= moff instead of off == moff because clang
* does not generate a BTF member for anonymous
* bitfield like the ":16" here:
* struct {
* int :16;
* int x:8;
* };
*/
if (off <= moff &&
BITS_ROUNDUP_BYTES(end_bit) <= off + size)
return WALK_SCALAR;
/* off may be accessing a following member
*
* or
*
* Doing partial access at either end of this
* bitfield. Continue on this case also to
* treat it as not accessing this bitfield
* and eventually error out as field not
* found to keep it simple.
* It could be relaxed if there was a legit
* partial access case later.
*/
continue;
}
/* In case of "off" is pointing to holes of a struct */
if (off < moff)
break;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* type of the field */
mid = member->type;
mtype = btf_type_by_id(btf, member->type);
mname = __btf_name_by_offset(btf, member->name_off);
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
mtype = __btf_resolve_size(btf, mtype, &msize,
&elem_type, &elem_id, &total_nelems,
&mid);
if (IS_ERR(mtype)) {
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
bpf_log(log, "field %s doesn't have size\n", mname);
return -EFAULT;
}
mtrue_end = moff + msize;
if (off >= mtrue_end)
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* no overlap with member, keep iterating */
continue;
if (btf_type_is_array(mtype)) {
u32 elem_idx;
/* __btf_resolve_size() above helps to
* linearize a multi-dimensional array.
*
* The logic here is treating an array
* in a struct as the following way:
*
* struct outer {
* struct inner array[2][2];
* };
*
* looks like:
*
* struct outer {
* struct inner array_elem0;
* struct inner array_elem1;
* struct inner array_elem2;
* struct inner array_elem3;
* };
*
* When accessing outer->array[1][0], it moves
* moff to "array_elem2", set mtype to
* "struct inner", and msize also becomes
* sizeof(struct inner). Then most of the
* remaining logic will fall through without
* caring the current member is an array or
* not.
*
* Unlike mtype/msize/moff, mtrue_end does not
* change. The naming difference ("_true") tells
* that it is not always corresponding to
* the current mtype/msize/moff.
* It is the true end of the current
* member (i.e. array in this case). That
* will allow an int array to be accessed like
* a scratch space,
* i.e. allow access beyond the size of
* the array's element as long as it is
* within the mtrue_end boundary.
*/
/* skip empty array */
if (moff == mtrue_end)
continue;
msize /= total_nelems;
elem_idx = (off - moff) / msize;
moff += elem_idx * msize;
mtype = elem_type;
mid = elem_id;
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* the 'off' we're looking for is either equal to start
* of this field or inside of this struct
*/
if (btf_type_is_struct(mtype)) {
/* our field must be inside that union or struct */
t = mtype;
/* return if the offset matches the member offset */
if (off == moff) {
*next_btf_id = mid;
return WALK_STRUCT;
}
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
/* adjust offset we're looking for */
off -= moff;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
goto again;
}
if (btf_type_is_ptr(mtype)) {
const struct btf_type *stype, *t;
enum bpf_type_flag tmp_flag = 0;
u32 id;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
if (msize != size || off != moff) {
bpf_log(log,
"cannot access ptr member %s with moff %u in struct %s with off %u size %u\n",
mname, moff, tname, off, size);
return -EACCES;
}
/* check type tag */
t = btf_type_by_id(btf, mtype->type);
if (btf_type_is_type_tag(t)) {
tag_value = __btf_name_by_offset(btf, t->name_off);
/* check __user tag */
if (strcmp(tag_value, "user") == 0)
tmp_flag = MEM_USER;
/* check __percpu tag */
if (strcmp(tag_value, "percpu") == 0)
tmp_flag = MEM_PERCPU;
bpf: Add kfunc bpf_rcu_read_lock/unlock() Add two kfunc's bpf_rcu_read_lock() and bpf_rcu_read_unlock(). These two kfunc's can be used for all program types. The following is an example about how rcu pointer are used w.r.t. bpf_rcu_read_lock()/bpf_rcu_read_unlock(). struct task_struct { ... struct task_struct *last_wakee; struct task_struct __rcu *real_parent; ... }; Let us say prog does 'task = bpf_get_current_task_btf()' to get a 'task' pointer. The basic rules are: - 'real_parent = task->real_parent' should be inside bpf_rcu_read_lock region. This is to simulate rcu_dereference() operation. The 'real_parent' is marked as MEM_RCU only if (1). task->real_parent is inside bpf_rcu_read_lock region, and (2). task is a trusted ptr. So MEM_RCU marked ptr can be 'trusted' inside the bpf_rcu_read_lock region. - 'last_wakee = real_parent->last_wakee' should be inside bpf_rcu_read_lock region since it tries to access rcu protected memory. - the ptr 'last_wakee' will be marked as PTR_UNTRUSTED since in general it is not clear whether the object pointed by 'last_wakee' is valid or not even inside bpf_rcu_read_lock region. The verifier will reset all rcu pointer register states to untrusted at bpf_rcu_read_unlock() kfunc call site, so any such rcu pointer won't be trusted any more outside the bpf_rcu_read_lock() region. The current implementation does not support nested rcu read lock region in the prog. Acked-by: Martin KaFai Lau <martin.lau@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20221124053217.2373910-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-24 13:32:17 +08:00
/* check __rcu tag */
if (strcmp(tag_value, "rcu") == 0)
tmp_flag = MEM_RCU;
}
stype = btf_type_skip_modifiers(btf, mtype->type, &id);
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
if (btf_type_is_struct(stype)) {
*next_btf_id = id;
bpf: Refactor RCU enforcement in the verifier. bpf_rcu_read_lock/unlock() are only available in clang compiled kernels. Lack of such key mechanism makes it impossible for sleepable bpf programs to use RCU pointers. Allow bpf_rcu_read_lock/unlock() in GCC compiled kernels (though GCC doesn't support btf_type_tag yet) and allowlist certain field dereferences in important data structures like tast_struct, cgroup, socket that are used by sleepable programs either as RCU pointer or full trusted pointer (which is valid outside of RCU CS). Use BTF_TYPE_SAFE_RCU and BTF_TYPE_SAFE_TRUSTED macros for such tagging. They will be removed once GCC supports btf_type_tag. With that refactor check_ptr_to_btf_access(). Make it strict in enforcing PTR_TRUSTED and PTR_UNTRUSTED while deprecating old PTR_TO_BTF_ID without modifier flags. There is a chance that this strict enforcement might break existing programs (especially on GCC compiled kernels), but this cleanup has to start sooner than later. Note PTR_TO_CTX access still yields old deprecated PTR_TO_BTF_ID. Once it's converted to strict PTR_TRUSTED or PTR_UNTRUSTED the kfuncs and helpers will be able to default to KF_TRUSTED_ARGS. KF_RCU will remain as a weaker version of KF_TRUSTED_ARGS where obj refcnt could be 0. Adjust rcu_read_lock selftest to run on gcc and clang compiled kernels. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: David Vernet <void@manifault.com> Link: https://lore.kernel.org/bpf/20230303041446.3630-7-alexei.starovoitov@gmail.com
2023-03-03 12:14:46 +08:00
*flag |= tmp_flag;
if (field_name)
*field_name = mname;
return WALK_PTR;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
}
}
/* Allow more flexible access within an int as long as
* it is within mtrue_end.
* Since mtrue_end could be the end of an array,
* that also allows using an array of int as a scratch
* space. e.g. skb->cb[].
*/
bpf: Fix an error in verifying a field in a union We are utilizing BPF LSM to monitor BPF operations within our container environment. When we add support for raw_tracepoint, it hits below error. ; (const void *)attr->raw_tracepoint.name); 27: (79) r3 = *(u64 *)(r2 +0) access beyond the end of member map_type (mend:4) in struct (anon) with off 0 size 8 It can be reproduced with below BPF prog. SEC("lsm/bpf") int BPF_PROG(bpf_audit, int cmd, union bpf_attr *attr, unsigned int size) { switch (cmd) { case BPF_RAW_TRACEPOINT_OPEN: bpf_printk("raw_tracepoint is %s", attr->raw_tracepoint.name); break; default: break; } return 0; } The reason is that when accessing a field in a union, such as bpf_attr, if the field is located within a nested struct that is not the first member of the union, it can result in incorrect field verification. union bpf_attr { struct { __u32 map_type; <<<< Actually it will find that field. __u32 key_size; __u32 value_size; ... }; ... struct { __u64 name; <<<< We want to verify this field. __u32 prog_fd; } raw_tracepoint; }; Considering the potential deep nesting levels, finding a perfect solution to address this issue has proven challenging. Therefore, I propose a solution where we simply skip the verification process if the field in question is located within a union. Fixes: 7e3617a72df3 ("bpf: Add array support to btf_struct_access") Signed-off-by: Yafang Shao <laoar.shao@gmail.com> Link: https://lore.kernel.org/r/20230713025642.27477-4-laoar.shao@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-07-13 10:56:41 +08:00
if (off + size > mtrue_end && !(*flag & PTR_UNTRUSTED)) {
bpf_log(log,
"access beyond the end of member %s (mend:%u) in struct %s with off %u size %u\n",
mname, mtrue_end, tname, off, size);
return -EACCES;
}
return WALK_SCALAR;
bpf: Implement accurate raw_tp context access via BTF libbpf analyzes bpf C program, searches in-kernel BTF for given type name and stores it into expected_attach_type. The kernel verifier expects this btf_id to point to something like: typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc); which represents signature of raw_tracepoint "kfree_skb". Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb' and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint. In first case it passes btf_id of 'struct sk_buff *' back to the verifier core and 'void *' in second case. Then the verifier tracks PTR_TO_BTF_ID as any other pointer type. Like PTR_TO_SOCKET points to 'struct bpf_sock', PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on. PTR_TO_BTF_ID points to in-kernel structs. If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF then PTR_TO_BTF_ID#1234 points to one of in kernel skbs. When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32) the btf_struct_access() checks which field of 'struct sk_buff' is at offset 32. Checks that size of access matches type definition of the field and continues to track the dereferenced type. If that field was a pointer to 'struct net_device' the r2's type will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device' in vmlinux's BTF. Such verifier analysis prevents "cheating" in BPF C program. The program cannot cast arbitrary pointer to 'struct sk_buff *' and access it. C compiler would allow type cast, of course, but the verifier will notice type mismatch based on BPF assembly and in-kernel BTF. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-7-ast@kernel.org
2019-10-16 11:25:00 +08:00
}
bpf_log(log, "struct %s doesn't have field at offset %d\n", tname, off);
return -EINVAL;
}
int btf_struct_access(struct bpf_verifier_log *log,
const struct bpf_reg_state *reg,
int off, int size, enum bpf_access_type atype __maybe_unused,
u32 *next_btf_id, enum bpf_type_flag *flag,
const char **field_name)
{
const struct btf *btf = reg->btf;
enum bpf_type_flag tmp_flag = 0;
const struct btf_type *t;
u32 id = reg->btf_id;
int err;
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
while (type_is_alloc(reg->type)) {
struct btf_struct_meta *meta;
struct btf_record *rec;
int i;
meta = btf_find_struct_meta(btf, id);
if (!meta)
break;
rec = meta->record;
for (i = 0; i < rec->cnt; i++) {
struct btf_field *field = &rec->fields[i];
u32 offset = field->offset;
if (off < offset + field->size && offset < off + size) {
bpf: Recognize lock and list fields in allocated objects Allow specifying bpf_spin_lock, bpf_list_head, bpf_list_node fields in a allocated object. Also update btf_struct_access to reject direct access to these special fields. A bpf_list_head allows implementing map-in-map style use cases, where an allocated object with bpf_list_head is linked into a list in a map value. This would require embedding a bpf_list_node, support for which is also included. The bpf_spin_lock is used to protect the bpf_list_head and other data. While we strictly don't require to hold a bpf_spin_lock while touching the bpf_list_head in such objects, as when have access to it, we have complete ownership of the object, the locking constraint is still kept and may be conditionally lifted in the future. Note that the specification of such types can be done just like map values, e.g.: struct bar { struct bpf_list_node node; }; struct foo { struct bpf_spin_lock lock; struct bpf_list_head head __contains(bar, node); struct bpf_list_node node; }; struct map_value { struct bpf_spin_lock lock; struct bpf_list_head head __contains(foo, node); }; To recognize such types in user BTF, we build a btf_struct_metas array of metadata items corresponding to each BTF ID. This is done once during the btf_parse stage to avoid having to do it each time during the verification process's requirement to inspect the metadata. Moreover, the computed metadata needs to be passed to some helpers in future patches which requires allocating them and storing them in the BTF that is pinned by the program itself, so that valid access can be assumed to such data during program runtime. A key thing to note is that once a btf_struct_meta is available for a type, both the btf_record and btf_field_offs should be available. It is critical that btf_field_offs is available in case special fields are present, as we extensively rely on special fields being zeroed out in map values and allocated objects in later patches. The code ensures that by bailing out in case of errors and ensuring both are available together. If the record is not available, the special fields won't be recognized, so not having both is also fine (in terms of being a verification error and not a runtime bug). Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20221118015614.2013203-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 09:55:56 +08:00
bpf_log(log,
"direct access to %s is disallowed\n",
btf_field_type_name(field->type));
return -EACCES;
}
}
break;
}
t = btf_type_by_id(btf, id);
do {
err = btf_struct_walk(log, btf, t, off, size, &id, &tmp_flag, field_name);
switch (err) {
case WALK_PTR:
/* For local types, the destination register cannot
* become a pointer again.
*/
if (type_is_alloc(reg->type))
return SCALAR_VALUE;
/* If we found the pointer or scalar on t+off,
* we're done.
*/
*next_btf_id = id;
*flag = tmp_flag;
return PTR_TO_BTF_ID;
case WALK_SCALAR:
return SCALAR_VALUE;
case WALK_STRUCT:
/* We found nested struct, so continue the search
* by diving in it. At this point the offset is
* aligned with the new type, so set it to 0.
*/
t = btf_type_by_id(btf, id);
off = 0;
break;
default:
/* It's either error or unknown return value..
* scream and leave.
*/
if (WARN_ONCE(err > 0, "unknown btf_struct_walk return value"))
return -EINVAL;
return err;
}
} while (t);
return -EINVAL;
}
/* Check that two BTF types, each specified as an BTF object + id, are exactly
* the same. Trivial ID check is not enough due to module BTFs, because we can
* end up with two different module BTFs, but IDs point to the common type in
* vmlinux BTF.
*/
bool btf_types_are_same(const struct btf *btf1, u32 id1,
const struct btf *btf2, u32 id2)
{
if (id1 != id2)
return false;
if (btf1 == btf2)
return true;
return btf_type_by_id(btf1, id1) == btf_type_by_id(btf2, id2);
}
bool btf_struct_ids_match(struct bpf_verifier_log *log,
const struct btf *btf, u32 id, int off,
const struct btf *need_btf, u32 need_type_id,
bool strict)
{
const struct btf_type *type;
enum bpf_type_flag flag = 0;
int err;
/* Are we already done? */
if (off == 0 && btf_types_are_same(btf, id, need_btf, need_type_id))
return true;
/* In case of strict type match, we do not walk struct, the top level
* type match must succeed. When strict is true, off should have already
* been 0.
*/
if (strict)
return false;
again:
type = btf_type_by_id(btf, id);
if (!type)
return false;
err = btf_struct_walk(log, btf, type, off, 1, &id, &flag, NULL);
if (err != WALK_STRUCT)
return false;
/* We found nested struct object. If it matches
* the requested ID, we're done. Otherwise let's
* continue the search with offset 0 in the new
* type.
*/
if (!btf_types_are_same(btf, id, need_btf, need_type_id)) {
off = 0;
goto again;
}
return true;
}
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
static int __get_type_size(struct btf *btf, u32 btf_id,
const struct btf_type **ret_type)
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
{
const struct btf_type *t;
*ret_type = btf_type_by_id(btf, 0);
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
if (!btf_id)
/* void */
return 0;
t = btf_type_by_id(btf, btf_id);
while (t && btf_type_is_modifier(t))
t = btf_type_by_id(btf, t->type);
if (!t)
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
return -EINVAL;
*ret_type = t;
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
if (btf_type_is_ptr(t))
/* kernel size of pointer. Not BPF's size of pointer*/
return sizeof(void *);
if (btf_type_is_int(t) || btf_is_any_enum(t) || __btf_type_is_struct(t))
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
return t->size;
return -EINVAL;
}
static u8 __get_type_fmodel_flags(const struct btf_type *t)
{
u8 flags = 0;
if (__btf_type_is_struct(t))
flags |= BTF_FMODEL_STRUCT_ARG;
if (btf_type_is_signed_int(t))
flags |= BTF_FMODEL_SIGNED_ARG;
return flags;
}
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
int btf_distill_func_proto(struct bpf_verifier_log *log,
struct btf *btf,
const struct btf_type *func,
const char *tname,
struct btf_func_model *m)
{
const struct btf_param *args;
const struct btf_type *t;
u32 i, nargs;
int ret;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
if (!func) {
/* BTF function prototype doesn't match the verifier types.
* Fall back to MAX_BPF_FUNC_REG_ARGS u64 args.
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
*/
for (i = 0; i < MAX_BPF_FUNC_REG_ARGS; i++) {
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
m->arg_size[i] = 8;
m->arg_flags[i] = 0;
}
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
m->ret_size = 8;
m->ret_flags = 0;
m->nr_args = MAX_BPF_FUNC_REG_ARGS;
bpf: Support attaching tracing BPF program to other BPF programs Allow FENTRY/FEXIT BPF programs to attach to other BPF programs of any type including their subprograms. This feature allows snooping on input and output packets in XDP, TC programs including their return values. In order to do that the verifier needs to track types not only of vmlinux, but types of other BPF programs as well. The verifier also needs to translate uapi/linux/bpf.h types used by networking programs into kernel internal BTF types used by FENTRY/FEXIT BPF programs. In some cases LLVM optimizations can remove arguments from BPF subprograms without adjusting BTF info that LLVM backend knows. When BTF info disagrees with actual types that the verifiers sees the BPF trampoline has to fallback to conservative and treat all arguments as u64. The FENTRY/FEXIT program can still attach to such subprograms, but it won't be able to recognize pointer types like 'struct sk_buff *' and it won't be able to pass them to bpf_skb_output() for dumping packets to user space. The FENTRY/FEXIT program would need to use bpf_probe_read_kernel() instead. The BPF_PROG_LOAD command is extended with attach_prog_fd field. When it's set to zero the attach_btf_id is one vmlinux BTF type ids. When attach_prog_fd points to previously loaded BPF program the attach_btf_id is BTF type id of main function or one of its subprograms. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-18-ast@kernel.org
2019-11-15 02:57:17 +08:00
return 0;
}
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
args = (const struct btf_param *)(func + 1);
nargs = btf_type_vlen(func);
if (nargs > MAX_BPF_FUNC_ARGS) {
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
bpf_log(log,
"The function %s has %d arguments. Too many.\n",
tname, nargs);
return -EINVAL;
}
ret = __get_type_size(btf, func->type, &t);
if (ret < 0 || __btf_type_is_struct(t)) {
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
bpf_log(log,
"The function %s return type %s is unsupported.\n",
tname, btf_type_str(t));
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
return -EINVAL;
}
m->ret_size = ret;
m->ret_flags = __get_type_fmodel_flags(t);
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
for (i = 0; i < nargs; i++) {
if (i == nargs - 1 && args[i].type == 0) {
bpf_log(log,
"The function %s with variable args is unsupported.\n",
tname);
return -EINVAL;
}
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
ret = __get_type_size(btf, args[i].type, &t);
/* No support of struct argument size greater than 16 bytes */
if (ret < 0 || ret > 16) {
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
bpf_log(log,
"The function %s arg%d type %s is unsupported.\n",
tname, i, btf_type_str(t));
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
return -EINVAL;
}
if (ret == 0) {
bpf_log(log,
"The function %s has malformed void argument.\n",
tname);
return -EINVAL;
}
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
m->arg_size[i] = ret;
m->arg_flags[i] = __get_type_fmodel_flags(t);
bpf: Introduce BPF trampoline Introduce BPF trampoline concept to allow kernel code to call into BPF programs with practically zero overhead. The trampoline generation logic is architecture dependent. It's converting native calling convention into BPF calling convention. BPF ISA is 64-bit (even on 32-bit architectures). The registers R1 to R5 are used to pass arguments into BPF functions. The main BPF program accepts only single argument "ctx" in R1. Whereas CPU native calling convention is different. x86-64 is passing first 6 arguments in registers and the rest on the stack. x86-32 is passing first 3 arguments in registers. sparc64 is passing first 6 in registers. And so on. The trampolines between BPF and kernel already exist. BPF_CALL_x macros in include/linux/filter.h statically compile trampolines from BPF into kernel helpers. They convert up to five u64 arguments into kernel C pointers and integers. On 64-bit architectures this BPF_to_kernel trampolines are nops. On 32-bit architecture they're meaningful. The opposite job kernel_to_BPF trampolines is done by CAST_TO_U64 macros and __bpf_trace_##call() shim functions in include/trace/bpf_probe.h. They convert kernel function arguments into array of u64s that BPF program consumes via R1=ctx pointer. This patch set is doing the same job as __bpf_trace_##call() static trampolines, but dynamically for any kernel function. There are ~22k global kernel functions that are attachable via nop at function entry. The function arguments and types are described in BTF. The job of btf_distill_func_proto() function is to extract useful information from BTF into "function model" that architecture dependent trampoline generators will use to generate assembly code to cast kernel function arguments into array of u64s. For example the kernel function eth_type_trans has two pointers. They will be casted to u64 and stored into stack of generated trampoline. The pointer to that stack space will be passed into BPF program in R1. On x86-64 such generated trampoline will consume 16 bytes of stack and two stores of %rdi and %rsi into stack. The verifier will make sure that only two u64 are accessed read-only by BPF program. The verifier will also recognize the precise type of the pointers being accessed and will not allow typecasting of the pointer to a different type within BPF program. The tracing use case in the datacenter demonstrated that certain key kernel functions have (like tcp_retransmit_skb) have 2 or more kprobes that are always active. Other functions have both kprobe and kretprobe. So it is essential to keep both kernel code and BPF programs executing at maximum speed. Hence generated BPF trampoline is re-generated every time new program is attached or detached to maintain maximum performance. To avoid the high cost of retpoline the attached BPF programs are called directly. __bpf_prog_enter/exit() are used to support per-program execution stats. In the future this logic will be optimized further by adding support for bpf_stats_enabled_key inside generated assembly code. Introduction of preemptible and sleepable BPF programs will completely remove the need to call to __bpf_prog_enter/exit(). Detach of a BPF program from the trampoline should not fail. To avoid memory allocation in detach path the half of the page is used as a reserve and flipped after each attach/detach. 2k bytes is enough to call 40+ BPF programs directly which is enough for BPF tracing use cases. This limit can be increased in the future. BPF_TRACE_FENTRY programs have access to raw kernel function arguments while BPF_TRACE_FEXIT programs have access to kernel return value as well. Often kprobe BPF program remembers function arguments in a map while kretprobe fetches arguments from a map and analyzes them together with return value. BPF_TRACE_FEXIT accelerates this typical use case. Recursion prevention for kprobe BPF programs is done via per-cpu bpf_prog_active counter. In practice that turned out to be a mistake. It caused programs to randomly skip execution. The tracing tools missed results they were looking for. Hence BPF trampoline doesn't provide builtin recursion prevention. It's a job of BPF program itself and will be addressed in the follow up patches. BPF trampoline is intended to be used beyond tracing and fentry/fexit use cases in the future. For example to remove retpoline cost from XDP programs. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20191114185720.1641606-5-ast@kernel.org
2019-11-15 02:57:04 +08:00
}
m->nr_args = nargs;
return 0;
}
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
/* Compare BTFs of two functions assuming only scalars and pointers to context.
* t1 points to BTF_KIND_FUNC in btf1
* t2 points to BTF_KIND_FUNC in btf2
* Returns:
* EINVAL - function prototype mismatch
* EFAULT - verifier bug
* 0 - 99% match. The last 1% is validated by the verifier.
*/
static int btf_check_func_type_match(struct bpf_verifier_log *log,
struct btf *btf1, const struct btf_type *t1,
struct btf *btf2, const struct btf_type *t2)
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
{
const struct btf_param *args1, *args2;
const char *fn1, *fn2, *s1, *s2;
u32 nargs1, nargs2, i;
fn1 = btf_name_by_offset(btf1, t1->name_off);
fn2 = btf_name_by_offset(btf2, t2->name_off);
if (btf_func_linkage(t1) != BTF_FUNC_GLOBAL) {
bpf_log(log, "%s() is not a global function\n", fn1);
return -EINVAL;
}
if (btf_func_linkage(t2) != BTF_FUNC_GLOBAL) {
bpf_log(log, "%s() is not a global function\n", fn2);
return -EINVAL;
}
t1 = btf_type_by_id(btf1, t1->type);
if (!t1 || !btf_type_is_func_proto(t1))
return -EFAULT;
t2 = btf_type_by_id(btf2, t2->type);
if (!t2 || !btf_type_is_func_proto(t2))
return -EFAULT;
args1 = (const struct btf_param *)(t1 + 1);
nargs1 = btf_type_vlen(t1);
args2 = (const struct btf_param *)(t2 + 1);
nargs2 = btf_type_vlen(t2);
if (nargs1 != nargs2) {
bpf_log(log, "%s() has %d args while %s() has %d args\n",
fn1, nargs1, fn2, nargs2);
return -EINVAL;
}
t1 = btf_type_skip_modifiers(btf1, t1->type, NULL);
t2 = btf_type_skip_modifiers(btf2, t2->type, NULL);
if (t1->info != t2->info) {
bpf_log(log,
"Return type %s of %s() doesn't match type %s of %s()\n",
btf_type_str(t1), fn1,
btf_type_str(t2), fn2);
return -EINVAL;
}
for (i = 0; i < nargs1; i++) {
t1 = btf_type_skip_modifiers(btf1, args1[i].type, NULL);
t2 = btf_type_skip_modifiers(btf2, args2[i].type, NULL);
if (t1->info != t2->info) {
bpf_log(log, "arg%d in %s() is %s while %s() has %s\n",
i, fn1, btf_type_str(t1),
fn2, btf_type_str(t2));
return -EINVAL;
}
if (btf_type_has_size(t1) && t1->size != t2->size) {
bpf_log(log,
"arg%d in %s() has size %d while %s() has %d\n",
i, fn1, t1->size,
fn2, t2->size);
return -EINVAL;
}
/* global functions are validated with scalars and pointers
* to context only. And only global functions can be replaced.
* Hence type check only those types.
*/
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
if (btf_type_is_int(t1) || btf_is_any_enum(t1))
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
continue;
if (!btf_type_is_ptr(t1)) {
bpf_log(log,
"arg%d in %s() has unrecognized type\n",
i, fn1);
return -EINVAL;
}
t1 = btf_type_skip_modifiers(btf1, t1->type, NULL);
t2 = btf_type_skip_modifiers(btf2, t2->type, NULL);
if (!btf_type_is_struct(t1)) {
bpf_log(log,
"arg%d in %s() is not a pointer to context\n",
i, fn1);
return -EINVAL;
}
if (!btf_type_is_struct(t2)) {
bpf_log(log,
"arg%d in %s() is not a pointer to context\n",
i, fn2);
return -EINVAL;
}
/* This is an optional check to make program writing easier.
* Compare names of structs and report an error to the user.
* btf_prepare_func_args() already checked that t2 struct
* is a context type. btf_prepare_func_args() will check
* later that t1 struct is a context type as well.
*/
s1 = btf_name_by_offset(btf1, t1->name_off);
s2 = btf_name_by_offset(btf2, t2->name_off);
if (strcmp(s1, s2)) {
bpf_log(log,
"arg%d %s(struct %s *) doesn't match %s(struct %s *)\n",
i, fn1, s1, fn2, s2);
return -EINVAL;
}
}
return 0;
}
/* Compare BTFs of given program with BTF of target program */
int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
struct btf *btf2, const struct btf_type *t2)
{
struct btf *btf1 = prog->aux->btf;
const struct btf_type *t1;
u32 btf_id = 0;
if (!prog->aux->func_info) {
bpf_log(log, "Program extension requires BTF\n");
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
return -EINVAL;
}
btf_id = prog->aux->func_info[0].type_id;
if (!btf_id)
return -EFAULT;
t1 = btf_type_by_id(btf1, btf_id);
if (!t1 || !btf_type_is_func(t1))
return -EFAULT;
return btf_check_func_type_match(log, btf1, t1, btf2, t2);
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
}
2023-12-15 09:13:31 +08:00
static bool btf_is_dynptr_ptr(const struct btf *btf, const struct btf_type *t)
{
const char *name;
t = btf_type_by_id(btf, t->type); /* skip PTR */
while (btf_type_is_modifier(t))
t = btf_type_by_id(btf, t->type);
/* allow either struct or struct forward declaration */
if (btf_type_is_struct(t) ||
(btf_type_is_fwd(t) && btf_type_kflag(t) == 0)) {
name = btf_str_by_offset(btf, t->name_off);
return name && strcmp(name, "bpf_dynptr") == 0;
}
return false;
}
struct bpf_cand_cache {
const char *name;
u32 name_len;
u16 kind;
u16 cnt;
struct {
const struct btf *btf;
u32 id;
} cands[];
};
static DEFINE_MUTEX(cand_cache_mutex);
static struct bpf_cand_cache *
bpf_core_find_cands(struct bpf_core_ctx *ctx, u32 local_type_id);
static int btf_get_ptr_to_btf_id(struct bpf_verifier_log *log, int arg_idx,
const struct btf *btf, const struct btf_type *t)
{
struct bpf_cand_cache *cc;
struct bpf_core_ctx ctx = {
.btf = btf,
.log = log,
};
u32 kern_type_id, type_id;
int err = 0;
/* skip PTR and modifiers */
type_id = t->type;
t = btf_type_by_id(btf, t->type);
while (btf_type_is_modifier(t)) {
type_id = t->type;
t = btf_type_by_id(btf, t->type);
}
mutex_lock(&cand_cache_mutex);
cc = bpf_core_find_cands(&ctx, type_id);
if (IS_ERR(cc)) {
err = PTR_ERR(cc);
bpf_log(log, "arg#%d reference type('%s %s') candidate matching error: %d\n",
arg_idx, btf_type_str(t), __btf_name_by_offset(btf, t->name_off),
err);
goto cand_cache_unlock;
}
if (cc->cnt != 1) {
bpf_log(log, "arg#%d reference type('%s %s') %s\n",
arg_idx, btf_type_str(t), __btf_name_by_offset(btf, t->name_off),
cc->cnt == 0 ? "has no matches" : "is ambiguous");
err = cc->cnt == 0 ? -ENOENT : -ESRCH;
goto cand_cache_unlock;
}
if (btf_is_module(cc->cands[0].btf)) {
bpf_log(log, "arg#%d reference type('%s %s') points to kernel module type (unsupported)\n",
arg_idx, btf_type_str(t), __btf_name_by_offset(btf, t->name_off));
err = -EOPNOTSUPP;
goto cand_cache_unlock;
}
kern_type_id = cc->cands[0].id;
cand_cache_unlock:
mutex_unlock(&cand_cache_mutex);
if (err)
return err;
return kern_type_id;
}
enum btf_arg_tag {
ARG_TAG_CTX = BIT_ULL(0),
ARG_TAG_NONNULL = BIT_ULL(1),
ARG_TAG_TRUSTED = BIT_ULL(2),
ARG_TAG_NULLABLE = BIT_ULL(3),
ARG_TAG_ARENA = BIT_ULL(4),
};
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
/* Process BTF of a function to produce high-level expectation of function
* arguments (like ARG_PTR_TO_CTX, or ARG_PTR_TO_MEM, etc). This information
* is cached in subprog info for reuse.
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
* Returns:
* EFAULT - there is a verifier bug. Abort verification.
* EINVAL - cannot convert BTF.
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
* 0 - Successfully processed BTF and constructed argument expectations.
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
*/
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog)
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
{
bpf: prepare btf_prepare_func_args() for handling static subprogs Generalize btf_prepare_func_args() to support both global and static subprogs. We are going to utilize this property in the next patch, reusing btf_prepare_func_args() for subprog call logic instead of reparsing BTF information in a completely separate implementation. btf_prepare_func_args() now detects whether subprog is global or static makes slight logic adjustments for static func cases, like not failing fatally (-EFAULT) for conditions that are allowable for static subprogs. Somewhat subtle (but major!) difference is the handling of pointer arguments. Both global and static functions need to handle special context arguments (which are pointers to predefined type names), but static subprogs give up on any other pointers, falling back to marking subprog as "unreliable", disabling the use of BTF type information altogether. For global functions, though, we are assuming that such pointers to unrecognized types are just pointers to fixed-sized memory region (or error out if size cannot be established, like for `void *` pointers). This patch accommodates these small differences and sets up a stage for refactoring in the next patch, eliminating a separate BTF-based parsing logic in btf_check_func_arg_match(). Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-4-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:27 +08:00
bool is_global = subprog_aux(env, subprog)->linkage == BTF_FUNC_GLOBAL;
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
struct bpf_subprog_info *sub = subprog_info(env, subprog);
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
struct bpf_verifier_log *log = &env->log;
struct bpf_prog *prog = env->prog;
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
enum bpf_prog_type prog_type = prog->type;
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
struct btf *btf = prog->aux->btf;
const struct btf_param *args;
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
const struct btf_type *t, *ref_t, *fn_t;
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
u32 i, nargs, btf_id;
const char *tname;
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
if (sub->args_cached)
return 0;
bpf: prepare btf_prepare_func_args() for handling static subprogs Generalize btf_prepare_func_args() to support both global and static subprogs. We are going to utilize this property in the next patch, reusing btf_prepare_func_args() for subprog call logic instead of reparsing BTF information in a completely separate implementation. btf_prepare_func_args() now detects whether subprog is global or static makes slight logic adjustments for static func cases, like not failing fatally (-EFAULT) for conditions that are allowable for static subprogs. Somewhat subtle (but major!) difference is the handling of pointer arguments. Both global and static functions need to handle special context arguments (which are pointers to predefined type names), but static subprogs give up on any other pointers, falling back to marking subprog as "unreliable", disabling the use of BTF type information altogether. For global functions, though, we are assuming that such pointers to unrecognized types are just pointers to fixed-sized memory region (or error out if size cannot be established, like for `void *` pointers). This patch accommodates these small differences and sets up a stage for refactoring in the next patch, eliminating a separate BTF-based parsing logic in btf_check_func_arg_match(). Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-4-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:27 +08:00
if (!prog->aux->func_info) {
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
bpf_log(log, "Verifier bug\n");
return -EFAULT;
}
btf_id = prog->aux->func_info[subprog].type_id;
if (!btf_id) {
bpf: prepare btf_prepare_func_args() for handling static subprogs Generalize btf_prepare_func_args() to support both global and static subprogs. We are going to utilize this property in the next patch, reusing btf_prepare_func_args() for subprog call logic instead of reparsing BTF information in a completely separate implementation. btf_prepare_func_args() now detects whether subprog is global or static makes slight logic adjustments for static func cases, like not failing fatally (-EFAULT) for conditions that are allowable for static subprogs. Somewhat subtle (but major!) difference is the handling of pointer arguments. Both global and static functions need to handle special context arguments (which are pointers to predefined type names), but static subprogs give up on any other pointers, falling back to marking subprog as "unreliable", disabling the use of BTF type information altogether. For global functions, though, we are assuming that such pointers to unrecognized types are just pointers to fixed-sized memory region (or error out if size cannot be established, like for `void *` pointers). This patch accommodates these small differences and sets up a stage for refactoring in the next patch, eliminating a separate BTF-based parsing logic in btf_check_func_arg_match(). Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-4-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:27 +08:00
if (!is_global) /* not fatal for static funcs */
return -EINVAL;
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
bpf_log(log, "Global functions need valid BTF\n");
return -EFAULT;
}
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
fn_t = btf_type_by_id(btf, btf_id);
if (!fn_t || !btf_type_is_func(fn_t)) {
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
/* These checks were already done by the verifier while loading
* struct bpf_func_info
*/
bpf_log(log, "BTF of func#%d doesn't point to KIND_FUNC\n",
subprog);
return -EFAULT;
}
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
tname = btf_name_by_offset(btf, fn_t->name_off);
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
if (prog->aux->func_info_aux[subprog].unreliable) {
bpf_log(log, "Verifier bug in function %s()\n", tname);
return -EFAULT;
}
bpf: Introduce dynamic program extensions Introduce dynamic program extensions. The users can load additional BPF functions and replace global functions in previously loaded BPF programs while these programs are executing. Global functions are verified individually by the verifier based on their types only. Hence the global function in the new program which types match older function can safely replace that corresponding function. This new function/program is called 'an extension' of old program. At load time the verifier uses (attach_prog_fd, attach_btf_id) pair to identify the function to be replaced. The BPF program type is derived from the target program into extension program. Technically bpf_verifier_ops is copied from target program. The BPF_PROG_TYPE_EXT program type is a placeholder. It has empty verifier_ops. The extension program can call the same bpf helper functions as target program. Single BPF_PROG_TYPE_EXT type is used to extend XDP, SKB and all other program types. The verifier allows only one level of replacement. Meaning that the extension program cannot recursively extend an extension. That also means that the maximum stack size is increasing from 512 to 1024 bytes and maximum function nesting level from 8 to 16. The programs don't always consume that much. The stack usage is determined by the number of on-stack variables used by the program. The verifier could have enforced 512 limit for combined original plus extension program, but it makes for difficult user experience. The main use case for extensions is to provide generic mechanism to plug external programs into policy program or function call chaining. BPF trampoline is used to track both fentry/fexit and program extensions because both are using the same nop slot at the beginning of every BPF function. Attaching fentry/fexit to a function that was replaced is not allowed. The opposite is true as well. Replacing a function that currently being analyzed with fentry/fexit is not allowed. The executable page allocated by BPF trampoline is not used by program extensions. This inefficiency will be optimized in future patches. Function by function verification of global function supports scalars and pointer to context only. Hence program extensions are supported for such class of global functions only. In the future the verifier will be extended with support to pointers to structures, arrays with sizes, etc. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/bpf/20200121005348.2769920-2-ast@kernel.org
2020-01-21 08:53:46 +08:00
if (prog_type == BPF_PROG_TYPE_EXT)
prog_type = prog->aux->dst_prog->type;
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
t = btf_type_by_id(btf, fn_t->type);
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
if (!t || !btf_type_is_func_proto(t)) {
bpf_log(log, "Invalid type of function %s()\n", tname);
return -EFAULT;
}
args = (const struct btf_param *)(t + 1);
nargs = btf_type_vlen(t);
if (nargs > MAX_BPF_FUNC_REG_ARGS) {
if (!is_global)
return -EINVAL;
bpf_log(log, "Global function %s() with %d > %d args. Buggy compiler.\n",
tname, nargs, MAX_BPF_FUNC_REG_ARGS);
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
return -EINVAL;
}
bpf: Add support for custom exception callbacks By default, the subprog generated by the verifier to handle a thrown exception hardcodes a return value of 0. To allow user-defined logic and modification of the return value when an exception is thrown, introduce the 'exception_callback:' declaration tag, which marks a callback as the default exception handler for the program. The format of the declaration tag is 'exception_callback:<value>', where <value> is the name of the exception callback. Each main program can be tagged using this BTF declaratiion tag to associate it with an exception callback. In case the tag is absent, the default callback is used. As such, the exception callback cannot be modified at runtime, only set during verification. Allowing modification of the callback for the current program execution at runtime leads to issues when the programs begin to nest, as any per-CPU state maintaing this information will have to be saved and restored. We don't want it to stay in bpf_prog_aux as this takes a global effect for all programs. An alternative solution is spilling the callback pointer at a known location on the program stack on entry, and then passing this location to bpf_throw as a parameter. However, since exceptions are geared more towards a use case where they are ideally never invoked, optimizing for this use case and adding to the complexity has diminishing returns. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20230912233214.1518551-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-09-13 07:32:03 +08:00
/* check that function returns int, exception cb also requires this */
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
t = btf_type_by_id(btf, t->type);
while (btf_type_is_modifier(t))
t = btf_type_by_id(btf, t->type);
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
if (!btf_type_is_int(t) && !btf_is_any_enum(t)) {
if (!is_global)
return -EINVAL;
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
bpf_log(log,
"Global function %s() doesn't return scalar. Only those are supported.\n",
tname);
return -EINVAL;
}
/* Convert BTF function arguments into verifier types.
* Only PTR_TO_CTX and SCALAR are supported atm.
*/
for (i = 0; i < nargs; i++) {
u32 tags = 0;
int id = 0;
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
/* 'arg:<tag>' decl_tag takes precedence over derivation of
* register type from BTF type itself
*/
while ((id = btf_find_next_decl_tag(btf, fn_t, i, "arg:", id)) > 0) {
const struct btf_type *tag_t = btf_type_by_id(btf, id);
const char *tag = __btf_name_by_offset(btf, tag_t->name_off) + 4;
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
/* disallow arg tags in static subprogs */
if (!is_global) {
bpf_log(log, "arg#%d type tag is not supported in static functions\n", i);
return -EOPNOTSUPP;
}
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
if (strcmp(tag, "ctx") == 0) {
tags |= ARG_TAG_CTX;
} else if (strcmp(tag, "trusted") == 0) {
tags |= ARG_TAG_TRUSTED;
} else if (strcmp(tag, "nonnull") == 0) {
tags |= ARG_TAG_NONNULL;
} else if (strcmp(tag, "nullable") == 0) {
tags |= ARG_TAG_NULLABLE;
} else if (strcmp(tag, "arena") == 0) {
tags |= ARG_TAG_ARENA;
} else {
bpf_log(log, "arg#%d has unsupported set of tags\n", i);
return -EOPNOTSUPP;
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
}
}
if (id != -ENOENT) {
bpf_log(log, "arg#%d type tag fetching failure: %d\n", i, id);
return id;
}
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
t = btf_type_by_id(btf, args[i].type);
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
while (btf_type_is_modifier(t))
t = btf_type_by_id(btf, t->type);
if (!btf_type_is_ptr(t))
goto skip_pointer;
if ((tags & ARG_TAG_CTX) || btf_is_prog_ctx_type(log, btf, t, prog_type, i)) {
if (tags & ~ARG_TAG_CTX) {
bpf_log(log, "arg#%d has invalid combination of tags\n", i);
return -EINVAL;
}
if ((tags & ARG_TAG_CTX) &&
btf_validate_prog_ctx_type(log, btf, t, i, prog_type,
prog->expected_attach_type))
return -EINVAL;
bpf: prepare btf_prepare_func_args() for handling static subprogs Generalize btf_prepare_func_args() to support both global and static subprogs. We are going to utilize this property in the next patch, reusing btf_prepare_func_args() for subprog call logic instead of reparsing BTF information in a completely separate implementation. btf_prepare_func_args() now detects whether subprog is global or static makes slight logic adjustments for static func cases, like not failing fatally (-EFAULT) for conditions that are allowable for static subprogs. Somewhat subtle (but major!) difference is the handling of pointer arguments. Both global and static functions need to handle special context arguments (which are pointers to predefined type names), but static subprogs give up on any other pointers, falling back to marking subprog as "unreliable", disabling the use of BTF type information altogether. For global functions, though, we are assuming that such pointers to unrecognized types are just pointers to fixed-sized memory region (or error out if size cannot be established, like for `void *` pointers). This patch accommodates these small differences and sets up a stage for refactoring in the next patch, eliminating a separate BTF-based parsing logic in btf_check_func_arg_match(). Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-4-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:27 +08:00
sub->args[i].arg_type = ARG_PTR_TO_CTX;
continue;
}
if (btf_is_dynptr_ptr(btf, t)) {
if (tags) {
bpf_log(log, "arg#%d has invalid combination of tags\n", i);
return -EINVAL;
}
2023-12-15 09:13:31 +08:00
sub->args[i].arg_type = ARG_PTR_TO_DYNPTR | MEM_RDONLY;
continue;
}
if (tags & ARG_TAG_TRUSTED) {
int kern_type_id;
if (tags & ARG_TAG_NONNULL) {
bpf_log(log, "arg#%d has invalid combination of tags\n", i);
return -EINVAL;
}
kern_type_id = btf_get_ptr_to_btf_id(log, i, btf, t);
if (kern_type_id < 0)
return kern_type_id;
sub->args[i].arg_type = ARG_PTR_TO_BTF_ID | PTR_TRUSTED;
if (tags & ARG_TAG_NULLABLE)
sub->args[i].arg_type |= PTR_MAYBE_NULL;
sub->args[i].btf_id = kern_type_id;
continue;
}
if (tags & ARG_TAG_ARENA) {
if (tags & ~ARG_TAG_ARENA) {
bpf_log(log, "arg#%d arena cannot be combined with any other tags\n", i);
return -EINVAL;
}
sub->args[i].arg_type = ARG_PTR_TO_ARENA;
continue;
}
if (is_global) { /* generic user data pointer */
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
u32 mem_size;
if (tags & ARG_TAG_NULLABLE) {
bpf_log(log, "arg#%d has invalid combination of tags\n", i);
return -EINVAL;
}
bpf: Support pointers in global func args Add an ability to pass a pointer to a type with known size in arguments of a global function. Such pointers may be used to overcome the limit on the maximum number of arguments, avoid expensive and tricky workarounds and to have multiple output arguments. A referenced type may contain pointers but indirect access through them isn't supported. The implementation consists of two parts. If a global function has an argument that is a pointer to a type with known size then: 1) In btf_check_func_arg_match(): check that the corresponding register points to NULL or to a valid memory region that is large enough to contain the expected argument's type. 2) In btf_prepare_func_args(): set the corresponding register type to PTR_TO_MEM_OR_NULL and its size to the size of the expected type. Only global functions are supported because allowance of pointers for static functions might break validation. Consider the following scenario. A static function has a pointer argument. A caller passes pointer to its stack memory. Because the callee can change referenced memory verifier cannot longer assume any particular slot type of the caller's stack memory hence the slot type is changed to SLOT_MISC. If there is an operation that relies on slot type other than SLOT_MISC then verifier won't be able to infer safety of the operation. When verifier sees a static function that has a pointer argument different from PTR_TO_CTX then it skips arguments check and continues with "inline" validation with more information available. The operation that relies on the particular slot type now succeeds. Because global functions were not allowed to have pointer arguments different from PTR_TO_CTX it's not possible to break existing and valid code. Signed-off-by: Dmitrii Banshchikov <me@ubique.spb.ru> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/bpf/20210212205642.620788-4-me@ubique.spb.ru
2021-02-13 04:56:41 +08:00
t = btf_type_skip_modifiers(btf, t->type, NULL);
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
ref_t = btf_resolve_size(btf, t, &mem_size);
bpf: Support pointers in global func args Add an ability to pass a pointer to a type with known size in arguments of a global function. Such pointers may be used to overcome the limit on the maximum number of arguments, avoid expensive and tricky workarounds and to have multiple output arguments. A referenced type may contain pointers but indirect access through them isn't supported. The implementation consists of two parts. If a global function has an argument that is a pointer to a type with known size then: 1) In btf_check_func_arg_match(): check that the corresponding register points to NULL or to a valid memory region that is large enough to contain the expected argument's type. 2) In btf_prepare_func_args(): set the corresponding register type to PTR_TO_MEM_OR_NULL and its size to the size of the expected type. Only global functions are supported because allowance of pointers for static functions might break validation. Consider the following scenario. A static function has a pointer argument. A caller passes pointer to its stack memory. Because the callee can change referenced memory verifier cannot longer assume any particular slot type of the caller's stack memory hence the slot type is changed to SLOT_MISC. If there is an operation that relies on slot type other than SLOT_MISC then verifier won't be able to infer safety of the operation. When verifier sees a static function that has a pointer argument different from PTR_TO_CTX then it skips arguments check and continues with "inline" validation with more information available. The operation that relies on the particular slot type now succeeds. Because global functions were not allowed to have pointer arguments different from PTR_TO_CTX it's not possible to break existing and valid code. Signed-off-by: Dmitrii Banshchikov <me@ubique.spb.ru> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/bpf/20210212205642.620788-4-me@ubique.spb.ru
2021-02-13 04:56:41 +08:00
if (IS_ERR(ref_t)) {
bpf_log(log, "arg#%d reference type('%s %s') size cannot be determined: %ld\n",
i, btf_type_str(t), btf_name_by_offset(btf, t->name_off),
bpf: Support pointers in global func args Add an ability to pass a pointer to a type with known size in arguments of a global function. Such pointers may be used to overcome the limit on the maximum number of arguments, avoid expensive and tricky workarounds and to have multiple output arguments. A referenced type may contain pointers but indirect access through them isn't supported. The implementation consists of two parts. If a global function has an argument that is a pointer to a type with known size then: 1) In btf_check_func_arg_match(): check that the corresponding register points to NULL or to a valid memory region that is large enough to contain the expected argument's type. 2) In btf_prepare_func_args(): set the corresponding register type to PTR_TO_MEM_OR_NULL and its size to the size of the expected type. Only global functions are supported because allowance of pointers for static functions might break validation. Consider the following scenario. A static function has a pointer argument. A caller passes pointer to its stack memory. Because the callee can change referenced memory verifier cannot longer assume any particular slot type of the caller's stack memory hence the slot type is changed to SLOT_MISC. If there is an operation that relies on slot type other than SLOT_MISC then verifier won't be able to infer safety of the operation. When verifier sees a static function that has a pointer argument different from PTR_TO_CTX then it skips arguments check and continues with "inline" validation with more information available. The operation that relies on the particular slot type now succeeds. Because global functions were not allowed to have pointer arguments different from PTR_TO_CTX it's not possible to break existing and valid code. Signed-off-by: Dmitrii Banshchikov <me@ubique.spb.ru> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/bpf/20210212205642.620788-4-me@ubique.spb.ru
2021-02-13 04:56:41 +08:00
PTR_ERR(ref_t));
return -EINVAL;
}
sub->args[i].arg_type = ARG_PTR_TO_MEM | PTR_MAYBE_NULL;
if (tags & ARG_TAG_NONNULL)
sub->args[i].arg_type &= ~PTR_MAYBE_NULL;
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
sub->args[i].mem_size = mem_size;
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
continue;
}
skip_pointer:
if (tags) {
bpf_log(log, "arg#%d has pointer tag, but is not a pointer type\n", i);
bpf: support 'arg:xxx' btf_decl_tag-based hints for global subprog args Add support for annotating global BPF subprog arguments to provide more information about expected semantics of the argument. Currently, verifier relies purely on argument's BTF type information, and supports three general use cases: scalar, pointer-to-context, and pointer-to-fixed-size-memory. Scalar and pointer-to-fixed-mem work well in practice and are quite natural to use. But pointer-to-context is a bit problematic, as typical BPF users don't realize that they need to use a special type name to signal to verifier that argument is not just some pointer, but actually a PTR_TO_CTX. Further, even if users do know which type to use, it is limiting in situations where the same BPF program logic is used across few different program types. Common case is kprobes, tracepoints, and perf_event programs having a helper to send some data over BPF perf buffer. bpf_perf_event_output() requires `ctx` argument, and so it's quite cumbersome to share such global subprog across few BPF programs of different types, necessitating extra static subprog that is context type-agnostic. Long story short, there is a need to go beyond types and allow users to add hints to global subprog arguments to define expectations. This patch adds such support for two initial special tags: - pointer to context; - non-null qualifier for generic pointer arguments. All of the above came up in practice already and seem generally useful additions. Non-null qualifier is an often requested feature, which currently has to be worked around by having unnecessary NULL checks inside subprogs even if we know that arguments are never NULL. Pointer to context was discussed earlier. As for implementation, we utilize btf_decl_tag attribute and set up an "arg:xxx" convention to specify argument hint. As such: - btf_decl_tag("arg:ctx") is a PTR_TO_CTX hint; - btf_decl_tag("arg:nonnull") marks pointer argument as not allowed to be NULL, making NULL check inside global subprog unnecessary. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-7-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:30 +08:00
return -EINVAL;
}
if (btf_type_is_int(t) || btf_is_any_enum(t)) {
sub->args[i].arg_type = ARG_ANYTHING;
continue;
}
if (!is_global)
return -EINVAL;
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
bpf_log(log, "Arg#%d type %s in %s() is not supported yet.\n",
i, btf_type_str(t), tname);
bpf: Introduce function-by-function verification New llvm and old llvm with libbpf help produce BTF that distinguish global and static functions. Unlike arguments of static function the arguments of global functions cannot be removed or optimized away by llvm. The compiler has to use exactly the arguments specified in a function prototype. The argument type information allows the verifier validate each global function independently. For now only supported argument types are pointer to context and scalars. In the future pointers to structures, sizes, pointer to packet data can be supported as well. Consider the following example: static int f1(int ...) { ... } int f3(int b); int f2(int a) { f1(a) + f3(a); } int f3(int b) { ... } int main(...) { f1(...) + f2(...) + f3(...); } The verifier will start its safety checks from the first global function f2(). It will recursively descend into f1() because it's static. Then it will check that arguments match for the f3() invocation inside f2(). It will not descend into f3(). It will finish f2() that has to be successfully verified for all possible values of 'a'. Then it will proceed with f3(). That function also has to be safe for all possible values of 'b'. Then it will start subprog 0 (which is main() function). It will recursively descend into f1() and will skip full check of f2() and f3(), since they are global. The order of processing global functions doesn't affect safety, since all global functions must be proven safe based on their arguments only. Such function by function verification can drastically improve speed of the verification and reduce complexity. Note that the stack limit of 512 still applies to the call chain regardless whether functions were static or global. The nested level of 8 also still applies. The same recursion prevention checks are in place as well. The type information and static/global kind is preserved after the verification hence in the above example global function f2() and f3() can be replaced later by equivalent functions with the same types that are loaded and verified later without affecting safety of this main() program. Such replacement (re-linking) of global functions is a subject of future patches. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200110064124.1760511-3-ast@kernel.org
2020-01-10 14:41:20 +08:00
return -EINVAL;
}
bpf: abstract away global subprog arg preparation logic from reg state setup btf_prepare_func_args() is used to understand expectations and restrictions on global subprog arguments. But current implementation is hard to extend, as it intermixes BTF-based func prototype parsing and interpretation logic with setting up register state at subprog entry. Worse still, those registers are not completely set up inside btf_prepare_func_args(), requiring some more logic later in do_check_common(). Like calling mark_reg_unknown() and similar initialization operations. This intermixing of BTF interpretation and register state setup is problematic. First, it causes duplication of BTF parsing logic for global subprog verification (to set up initial state of global subprog) and global subprog call sites analysis (when we need to check that whatever is being passed into global subprog matches expectations), performed in btf_check_subprog_call(). Given we want to extend global func argument with tags later, this duplication is problematic. So refactor btf_prepare_func_args() to do only BTF-based func proto and args parsing, returning high-level argument "expectations" only, with no regard to specifics of register state. I.e., if it's a context argument, instead of setting register state to PTR_TO_CTX, we return ARG_PTR_TO_CTX enum for that argument as "an argument specification" for further processing inside do_check_common(). Similarly for SCALAR arguments, PTR_TO_MEM, etc. This allows to reuse btf_prepare_func_args() in following patches at global subprog call site analysis time. It also keeps register setup code consistently in one place, do_check_common(). Besides all this, we cache this argument specs information inside env->subprog_info, eliminating the need to redo these potentially expensive BTF traversals, especially if BPF program's BTF is big and/or there are lots of global subprog calls. Acked-by: Eduard Zingerman <eddyz87@gmail.com> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20231215011334.2307144-2-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-12-15 09:13:25 +08:00
sub->arg_cnt = nargs;
sub->args_cached = true;
return 0;
}
static void btf_type_show(const struct btf *btf, u32 type_id, void *obj,
struct btf_show *show)
{
const struct btf_type *t = btf_type_by_id(btf, type_id);
show->btf = btf;
memset(&show->state, 0, sizeof(show->state));
memset(&show->obj, 0, sizeof(show->obj));
btf_type_ops(t)->show(btf, t, type_id, obj, 0, show);
}
__printf(2, 0) static void btf_seq_show(struct btf_show *show, const char *fmt,
va_list args)
{
seq_vprintf((struct seq_file *)show->target, fmt, args);
}
int btf_type_seq_show_flags(const struct btf *btf, u32 type_id,
void *obj, struct seq_file *m, u64 flags)
{
struct btf_show sseq;
sseq.target = m;
sseq.showfn = btf_seq_show;
sseq.flags = flags;
btf_type_show(btf, type_id, obj, &sseq);
return sseq.state.status;
}
void btf_type_seq_show(const struct btf *btf, u32 type_id, void *obj,
struct seq_file *m)
{
(void) btf_type_seq_show_flags(btf, type_id, obj, m,
BTF_SHOW_NONAME | BTF_SHOW_COMPACT |
BTF_SHOW_ZERO | BTF_SHOW_UNSAFE);
}
struct btf_show_snprintf {
struct btf_show show;
int len_left; /* space left in string */
int len; /* length we would have written */
};
__printf(2, 0) static void btf_snprintf_show(struct btf_show *show, const char *fmt,
va_list args)
{
struct btf_show_snprintf *ssnprintf = (struct btf_show_snprintf *)show;
int len;
len = vsnprintf(show->target, ssnprintf->len_left, fmt, args);
if (len < 0) {
ssnprintf->len_left = 0;
ssnprintf->len = len;
} else if (len >= ssnprintf->len_left) {
/* no space, drive on to get length we would have written */
ssnprintf->len_left = 0;
ssnprintf->len += len;
} else {
ssnprintf->len_left -= len;
ssnprintf->len += len;
show->target += len;
}
}
int btf_type_snprintf_show(const struct btf *btf, u32 type_id, void *obj,
char *buf, int len, u64 flags)
{
struct btf_show_snprintf ssnprintf;
ssnprintf.show.target = buf;
ssnprintf.show.flags = flags;
ssnprintf.show.showfn = btf_snprintf_show;
ssnprintf.len_left = len;
ssnprintf.len = 0;
btf_type_show(btf, type_id, obj, (struct btf_show *)&ssnprintf);
/* If we encountered an error, return it. */
if (ssnprintf.show.state.status)
return ssnprintf.show.state.status;
/* Otherwise return length we would have written */
return ssnprintf.len;
}
#ifdef CONFIG_PROC_FS
static void bpf_btf_show_fdinfo(struct seq_file *m, struct file *filp)
{
const struct btf *btf = filp->private_data;
seq_printf(m, "btf_id:\t%u\n", btf->id);
}
#endif
static int btf_release(struct inode *inode, struct file *filp)
{
btf_put(filp->private_data);
return 0;
}
const struct file_operations btf_fops = {
#ifdef CONFIG_PROC_FS
.show_fdinfo = bpf_btf_show_fdinfo,
#endif
.release = btf_release,
};
static int __btf_new_fd(struct btf *btf)
{
return anon_inode_getfd("btf", &btf_fops, btf, O_RDONLY | O_CLOEXEC);
}
int btf_new_fd(const union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size)
{
struct btf *btf;
int ret;
btf = btf_parse(attr, uattr, uattr_size);
if (IS_ERR(btf))
return PTR_ERR(btf);
ret = btf_alloc_id(btf);
if (ret) {
btf_free(btf);
return ret;
}
/*
* The BTF ID is published to the userspace.
* All BTF free must go through call_rcu() from
* now on (i.e. free by calling btf_put()).
*/
ret = __btf_new_fd(btf);
if (ret < 0)
btf_put(btf);
return ret;
}
struct btf *btf_get_by_fd(int fd)
{
struct btf *btf;
CLASS(fd, f)(fd);
if (fd_empty(f))
return ERR_PTR(-EBADF);
if (fd_file(f)->f_op != &btf_fops)
return ERR_PTR(-EINVAL);
btf = fd_file(f)->private_data;
refcount_inc(&btf->refcnt);
return btf;
}
int btf_get_info_by_fd(const struct btf *btf,
const union bpf_attr *attr,
union bpf_attr __user *uattr)
{
struct bpf_btf_info __user *uinfo;
struct bpf_btf_info info;
u32 info_copy, btf_copy;
void __user *ubtf;
char __user *uname;
u32 uinfo_len, uname_len, name_len;
int ret = 0;
uinfo = u64_to_user_ptr(attr->info.info);
uinfo_len = attr->info.info_len;
info_copy = min_t(u32, uinfo_len, sizeof(info));
memset(&info, 0, sizeof(info));
if (copy_from_user(&info, uinfo, info_copy))
return -EFAULT;
info.id = btf->id;
ubtf = u64_to_user_ptr(info.btf);
btf_copy = min_t(u32, btf->data_size, info.btf_size);
if (copy_to_user(ubtf, btf->data, btf_copy))
return -EFAULT;
info.btf_size = btf->data_size;
info.kernel_btf = btf->kernel_btf;
uname = u64_to_user_ptr(info.name);
uname_len = info.name_len;
if (!uname ^ !uname_len)
return -EINVAL;
name_len = strlen(btf->name);
info.name_len = name_len;
if (uname) {
if (uname_len >= name_len + 1) {
if (copy_to_user(uname, btf->name, name_len + 1))
return -EFAULT;
} else {
char zero = '\0';
if (copy_to_user(uname, btf->name, uname_len - 1))
return -EFAULT;
if (put_user(zero, uname + uname_len - 1))
return -EFAULT;
/* let user-space know about too short buffer */
ret = -ENOSPC;
}
}
if (copy_to_user(uinfo, &info, info_copy) ||
put_user(info_copy, &uattr->info.info_len))
return -EFAULT;
return ret;
}
int btf_get_fd_by_id(u32 id)
{
struct btf *btf;
int fd;
rcu_read_lock();
btf = idr_find(&btf_idr, id);
if (!btf || !refcount_inc_not_zero(&btf->refcnt))
btf = ERR_PTR(-ENOENT);
rcu_read_unlock();
if (IS_ERR(btf))
return PTR_ERR(btf);
fd = __btf_new_fd(btf);
if (fd < 0)
btf_put(btf);
return fd;
}
u32 btf_obj_id(const struct btf *btf)
{
return btf->id;
}
bool btf_is_kernel(const struct btf *btf)
{
return btf->kernel_btf;
}
bool btf_is_module(const struct btf *btf)
{
return btf->kernel_btf && strcmp(btf->name, "vmlinux") != 0;
}
bpf: Fix UAF due to race between btf_try_get_module and load_module While working on code to populate kfunc BTF ID sets for module BTF from its initcall, I noticed that by the time the initcall is invoked, the module BTF can already be seen by userspace (and the BPF verifier). The existing btf_try_get_module calls try_module_get which only fails if mod->state == MODULE_STATE_GOING, i.e. it can increment module reference when module initcall is happening in parallel. Currently, BTF parsing happens from MODULE_STATE_COMING notifier callback. At this point, the module initcalls have not been invoked. The notifier callback parses and prepares the module BTF, allocates an ID, which publishes it to userspace, and then adds it to the btf_modules list allowing the kernel to invoke btf_try_get_module for the BTF. However, at this point, the module has not been fully initialized (i.e. its initcalls have not finished). The code in module.c can still fail and free the module, without caring for other users. However, nothing stops btf_try_get_module from succeeding between the state transition from MODULE_STATE_COMING to MODULE_STATE_LIVE. This leads to a use-after-free issue when BPF program loads successfully in the state transition, load_module's do_init_module call fails and frees the module, and BPF program fd on close calls module_put for the freed module. Future patch has test case to verify we don't regress in this area in future. There are multiple points after prepare_coming_module (in load_module) where failure can occur and module loading can return error. We illustrate and test for the race using the last point where it can practically occur (in module __init function). An illustration of the race: CPU 0 CPU 1 load_module notifier_call(MODULE_STATE_COMING) btf_parse_module btf_alloc_id // Published to userspace list_add(&btf_mod->list, btf_modules) mod->init(...) ... ^ bpf_check | check_pseudo_btf_id | btf_try_get_module | returns true | ... ... | module __init in progress return prog_fd | ... ... V if (ret < 0) free_module(mod) ... close(prog_fd) ... bpf_prog_free_deferred module_put(used_btf.mod) // use-after-free We fix this issue by setting a flag BTF_MODULE_F_LIVE, from the notifier callback when MODULE_STATE_LIVE state is reached for the module, so that we return NULL from btf_try_get_module for modules that are not fully formed. Since try_module_get already checks that module is not in MODULE_STATE_GOING state, and that is the only transition a live module can make before being removed from btf_modules list, this is enough to close the race and prevent the bug. A later selftest patch crafts the race condition artifically to verify that it has been fixed, and that verifier fails to load program (with ENXIO). Lastly, a couple of comments: 1. Even if this race didn't exist, it seems more appropriate to only access resources (ksyms and kfuncs) of a fully formed module which has been initialized completely. 2. This patch was born out of need for synchronization against module initcall for the next patch, so it is needed for correctness even without the aforementioned race condition. The BTF resources initialized by module initcall are set up once and then only looked up, so just waiting until the initcall has finished ensures correct behavior. Fixes: 541c3bad8dc5 ("bpf: Support BPF ksym variables in kernel modules") Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-2-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:44 +08:00
enum {
BTF_MODULE_F_LIVE = (1 << 0),
};
#ifdef CONFIG_DEBUG_INFO_BTF_MODULES
struct btf_module {
struct list_head list;
struct module *module;
struct btf *btf;
struct bin_attribute *sysfs_attr;
bpf: Fix UAF due to race between btf_try_get_module and load_module While working on code to populate kfunc BTF ID sets for module BTF from its initcall, I noticed that by the time the initcall is invoked, the module BTF can already be seen by userspace (and the BPF verifier). The existing btf_try_get_module calls try_module_get which only fails if mod->state == MODULE_STATE_GOING, i.e. it can increment module reference when module initcall is happening in parallel. Currently, BTF parsing happens from MODULE_STATE_COMING notifier callback. At this point, the module initcalls have not been invoked. The notifier callback parses and prepares the module BTF, allocates an ID, which publishes it to userspace, and then adds it to the btf_modules list allowing the kernel to invoke btf_try_get_module for the BTF. However, at this point, the module has not been fully initialized (i.e. its initcalls have not finished). The code in module.c can still fail and free the module, without caring for other users. However, nothing stops btf_try_get_module from succeeding between the state transition from MODULE_STATE_COMING to MODULE_STATE_LIVE. This leads to a use-after-free issue when BPF program loads successfully in the state transition, load_module's do_init_module call fails and frees the module, and BPF program fd on close calls module_put for the freed module. Future patch has test case to verify we don't regress in this area in future. There are multiple points after prepare_coming_module (in load_module) where failure can occur and module loading can return error. We illustrate and test for the race using the last point where it can practically occur (in module __init function). An illustration of the race: CPU 0 CPU 1 load_module notifier_call(MODULE_STATE_COMING) btf_parse_module btf_alloc_id // Published to userspace list_add(&btf_mod->list, btf_modules) mod->init(...) ... ^ bpf_check | check_pseudo_btf_id | btf_try_get_module | returns true | ... ... | module __init in progress return prog_fd | ... ... V if (ret < 0) free_module(mod) ... close(prog_fd) ... bpf_prog_free_deferred module_put(used_btf.mod) // use-after-free We fix this issue by setting a flag BTF_MODULE_F_LIVE, from the notifier callback when MODULE_STATE_LIVE state is reached for the module, so that we return NULL from btf_try_get_module for modules that are not fully formed. Since try_module_get already checks that module is not in MODULE_STATE_GOING state, and that is the only transition a live module can make before being removed from btf_modules list, this is enough to close the race and prevent the bug. A later selftest patch crafts the race condition artifically to verify that it has been fixed, and that verifier fails to load program (with ENXIO). Lastly, a couple of comments: 1. Even if this race didn't exist, it seems more appropriate to only access resources (ksyms and kfuncs) of a fully formed module which has been initialized completely. 2. This patch was born out of need for synchronization against module initcall for the next patch, so it is needed for correctness even without the aforementioned race condition. The BTF resources initialized by module initcall are set up once and then only looked up, so just waiting until the initcall has finished ensures correct behavior. Fixes: 541c3bad8dc5 ("bpf: Support BPF ksym variables in kernel modules") Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-2-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:44 +08:00
int flags;
};
static LIST_HEAD(btf_modules);
static DEFINE_MUTEX(btf_module_mutex);
static ssize_t
btf_module_read(struct file *file, struct kobject *kobj,
struct bin_attribute *bin_attr,
char *buf, loff_t off, size_t len)
{
const struct btf *btf = bin_attr->private;
memcpy(buf, btf->data + off, len);
return len;
}
static void purge_cand_cache(struct btf *btf);
static int btf_module_notify(struct notifier_block *nb, unsigned long op,
void *module)
{
struct btf_module *btf_mod, *tmp;
struct module *mod = module;
struct btf *btf;
int err = 0;
if (mod->btf_data_size == 0 ||
bpf: Fix UAF due to race between btf_try_get_module and load_module While working on code to populate kfunc BTF ID sets for module BTF from its initcall, I noticed that by the time the initcall is invoked, the module BTF can already be seen by userspace (and the BPF verifier). The existing btf_try_get_module calls try_module_get which only fails if mod->state == MODULE_STATE_GOING, i.e. it can increment module reference when module initcall is happening in parallel. Currently, BTF parsing happens from MODULE_STATE_COMING notifier callback. At this point, the module initcalls have not been invoked. The notifier callback parses and prepares the module BTF, allocates an ID, which publishes it to userspace, and then adds it to the btf_modules list allowing the kernel to invoke btf_try_get_module for the BTF. However, at this point, the module has not been fully initialized (i.e. its initcalls have not finished). The code in module.c can still fail and free the module, without caring for other users. However, nothing stops btf_try_get_module from succeeding between the state transition from MODULE_STATE_COMING to MODULE_STATE_LIVE. This leads to a use-after-free issue when BPF program loads successfully in the state transition, load_module's do_init_module call fails and frees the module, and BPF program fd on close calls module_put for the freed module. Future patch has test case to verify we don't regress in this area in future. There are multiple points after prepare_coming_module (in load_module) where failure can occur and module loading can return error. We illustrate and test for the race using the last point where it can practically occur (in module __init function). An illustration of the race: CPU 0 CPU 1 load_module notifier_call(MODULE_STATE_COMING) btf_parse_module btf_alloc_id // Published to userspace list_add(&btf_mod->list, btf_modules) mod->init(...) ... ^ bpf_check | check_pseudo_btf_id | btf_try_get_module | returns true | ... ... | module __init in progress return prog_fd | ... ... V if (ret < 0) free_module(mod) ... close(prog_fd) ... bpf_prog_free_deferred module_put(used_btf.mod) // use-after-free We fix this issue by setting a flag BTF_MODULE_F_LIVE, from the notifier callback when MODULE_STATE_LIVE state is reached for the module, so that we return NULL from btf_try_get_module for modules that are not fully formed. Since try_module_get already checks that module is not in MODULE_STATE_GOING state, and that is the only transition a live module can make before being removed from btf_modules list, this is enough to close the race and prevent the bug. A later selftest patch crafts the race condition artifically to verify that it has been fixed, and that verifier fails to load program (with ENXIO). Lastly, a couple of comments: 1. Even if this race didn't exist, it seems more appropriate to only access resources (ksyms and kfuncs) of a fully formed module which has been initialized completely. 2. This patch was born out of need for synchronization against module initcall for the next patch, so it is needed for correctness even without the aforementioned race condition. The BTF resources initialized by module initcall are set up once and then only looked up, so just waiting until the initcall has finished ensures correct behavior. Fixes: 541c3bad8dc5 ("bpf: Support BPF ksym variables in kernel modules") Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-2-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:44 +08:00
(op != MODULE_STATE_COMING && op != MODULE_STATE_LIVE &&
op != MODULE_STATE_GOING))
goto out;
switch (op) {
case MODULE_STATE_COMING:
btf_mod = kzalloc(sizeof(*btf_mod), GFP_KERNEL);
if (!btf_mod) {
err = -ENOMEM;
goto out;
}
btf = btf_parse_module(mod->name, mod->btf_data, mod->btf_data_size,
mod->btf_base_data, mod->btf_base_data_size);
if (IS_ERR(btf)) {
kfree(btf_mod);
if (!IS_ENABLED(CONFIG_MODULE_ALLOW_BTF_MISMATCH)) {
pr_warn("failed to validate module [%s] BTF: %ld\n",
mod->name, PTR_ERR(btf));
err = PTR_ERR(btf);
} else {
pr_warn_once("Kernel module BTF mismatch detected, BTF debug info may be unavailable for some modules\n");
}
goto out;
}
err = btf_alloc_id(btf);
if (err) {
btf_free(btf);
kfree(btf_mod);
goto out;
}
purge_cand_cache(NULL);
mutex_lock(&btf_module_mutex);
btf_mod->module = module;
btf_mod->btf = btf;
list_add(&btf_mod->list, &btf_modules);
mutex_unlock(&btf_module_mutex);
if (IS_ENABLED(CONFIG_SYSFS)) {
struct bin_attribute *attr;
attr = kzalloc(sizeof(*attr), GFP_KERNEL);
if (!attr)
goto out;
sysfs_bin_attr_init(attr);
attr->attr.name = btf->name;
attr->attr.mode = 0444;
attr->size = btf->data_size;
attr->private = btf;
attr->read = btf_module_read;
err = sysfs_create_bin_file(btf_kobj, attr);
if (err) {
pr_warn("failed to register module [%s] BTF in sysfs: %d\n",
mod->name, err);
kfree(attr);
err = 0;
goto out;
}
btf_mod->sysfs_attr = attr;
}
bpf: Fix UAF due to race between btf_try_get_module and load_module While working on code to populate kfunc BTF ID sets for module BTF from its initcall, I noticed that by the time the initcall is invoked, the module BTF can already be seen by userspace (and the BPF verifier). The existing btf_try_get_module calls try_module_get which only fails if mod->state == MODULE_STATE_GOING, i.e. it can increment module reference when module initcall is happening in parallel. Currently, BTF parsing happens from MODULE_STATE_COMING notifier callback. At this point, the module initcalls have not been invoked. The notifier callback parses and prepares the module BTF, allocates an ID, which publishes it to userspace, and then adds it to the btf_modules list allowing the kernel to invoke btf_try_get_module for the BTF. However, at this point, the module has not been fully initialized (i.e. its initcalls have not finished). The code in module.c can still fail and free the module, without caring for other users. However, nothing stops btf_try_get_module from succeeding between the state transition from MODULE_STATE_COMING to MODULE_STATE_LIVE. This leads to a use-after-free issue when BPF program loads successfully in the state transition, load_module's do_init_module call fails and frees the module, and BPF program fd on close calls module_put for the freed module. Future patch has test case to verify we don't regress in this area in future. There are multiple points after prepare_coming_module (in load_module) where failure can occur and module loading can return error. We illustrate and test for the race using the last point where it can practically occur (in module __init function). An illustration of the race: CPU 0 CPU 1 load_module notifier_call(MODULE_STATE_COMING) btf_parse_module btf_alloc_id // Published to userspace list_add(&btf_mod->list, btf_modules) mod->init(...) ... ^ bpf_check | check_pseudo_btf_id | btf_try_get_module | returns true | ... ... | module __init in progress return prog_fd | ... ... V if (ret < 0) free_module(mod) ... close(prog_fd) ... bpf_prog_free_deferred module_put(used_btf.mod) // use-after-free We fix this issue by setting a flag BTF_MODULE_F_LIVE, from the notifier callback when MODULE_STATE_LIVE state is reached for the module, so that we return NULL from btf_try_get_module for modules that are not fully formed. Since try_module_get already checks that module is not in MODULE_STATE_GOING state, and that is the only transition a live module can make before being removed from btf_modules list, this is enough to close the race and prevent the bug. A later selftest patch crafts the race condition artifically to verify that it has been fixed, and that verifier fails to load program (with ENXIO). Lastly, a couple of comments: 1. Even if this race didn't exist, it seems more appropriate to only access resources (ksyms and kfuncs) of a fully formed module which has been initialized completely. 2. This patch was born out of need for synchronization against module initcall for the next patch, so it is needed for correctness even without the aforementioned race condition. The BTF resources initialized by module initcall are set up once and then only looked up, so just waiting until the initcall has finished ensures correct behavior. Fixes: 541c3bad8dc5 ("bpf: Support BPF ksym variables in kernel modules") Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-2-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:44 +08:00
break;
case MODULE_STATE_LIVE:
mutex_lock(&btf_module_mutex);
list_for_each_entry_safe(btf_mod, tmp, &btf_modules, list) {
if (btf_mod->module != module)
continue;
btf_mod->flags |= BTF_MODULE_F_LIVE;
break;
}
mutex_unlock(&btf_module_mutex);
break;
case MODULE_STATE_GOING:
mutex_lock(&btf_module_mutex);
list_for_each_entry_safe(btf_mod, tmp, &btf_modules, list) {
if (btf_mod->module != module)
continue;
list_del(&btf_mod->list);
if (btf_mod->sysfs_attr)
sysfs_remove_bin_file(btf_kobj, btf_mod->sysfs_attr);
purge_cand_cache(btf_mod->btf);
btf_put(btf_mod->btf);
kfree(btf_mod->sysfs_attr);
kfree(btf_mod);
break;
}
mutex_unlock(&btf_module_mutex);
break;
}
out:
return notifier_from_errno(err);
}
static struct notifier_block btf_module_nb = {
.notifier_call = btf_module_notify,
};
static int __init btf_module_init(void)
{
register_module_notifier(&btf_module_nb);
return 0;
}
fs_initcall(btf_module_init);
#endif /* CONFIG_DEBUG_INFO_BTF_MODULES */
struct module *btf_try_get_module(const struct btf *btf)
{
struct module *res = NULL;
#ifdef CONFIG_DEBUG_INFO_BTF_MODULES
struct btf_module *btf_mod, *tmp;
mutex_lock(&btf_module_mutex);
list_for_each_entry_safe(btf_mod, tmp, &btf_modules, list) {
if (btf_mod->btf != btf)
continue;
bpf: Fix UAF due to race between btf_try_get_module and load_module While working on code to populate kfunc BTF ID sets for module BTF from its initcall, I noticed that by the time the initcall is invoked, the module BTF can already be seen by userspace (and the BPF verifier). The existing btf_try_get_module calls try_module_get which only fails if mod->state == MODULE_STATE_GOING, i.e. it can increment module reference when module initcall is happening in parallel. Currently, BTF parsing happens from MODULE_STATE_COMING notifier callback. At this point, the module initcalls have not been invoked. The notifier callback parses and prepares the module BTF, allocates an ID, which publishes it to userspace, and then adds it to the btf_modules list allowing the kernel to invoke btf_try_get_module for the BTF. However, at this point, the module has not been fully initialized (i.e. its initcalls have not finished). The code in module.c can still fail and free the module, without caring for other users. However, nothing stops btf_try_get_module from succeeding between the state transition from MODULE_STATE_COMING to MODULE_STATE_LIVE. This leads to a use-after-free issue when BPF program loads successfully in the state transition, load_module's do_init_module call fails and frees the module, and BPF program fd on close calls module_put for the freed module. Future patch has test case to verify we don't regress in this area in future. There are multiple points after prepare_coming_module (in load_module) where failure can occur and module loading can return error. We illustrate and test for the race using the last point where it can practically occur (in module __init function). An illustration of the race: CPU 0 CPU 1 load_module notifier_call(MODULE_STATE_COMING) btf_parse_module btf_alloc_id // Published to userspace list_add(&btf_mod->list, btf_modules) mod->init(...) ... ^ bpf_check | check_pseudo_btf_id | btf_try_get_module | returns true | ... ... | module __init in progress return prog_fd | ... ... V if (ret < 0) free_module(mod) ... close(prog_fd) ... bpf_prog_free_deferred module_put(used_btf.mod) // use-after-free We fix this issue by setting a flag BTF_MODULE_F_LIVE, from the notifier callback when MODULE_STATE_LIVE state is reached for the module, so that we return NULL from btf_try_get_module for modules that are not fully formed. Since try_module_get already checks that module is not in MODULE_STATE_GOING state, and that is the only transition a live module can make before being removed from btf_modules list, this is enough to close the race and prevent the bug. A later selftest patch crafts the race condition artifically to verify that it has been fixed, and that verifier fails to load program (with ENXIO). Lastly, a couple of comments: 1. Even if this race didn't exist, it seems more appropriate to only access resources (ksyms and kfuncs) of a fully formed module which has been initialized completely. 2. This patch was born out of need for synchronization against module initcall for the next patch, so it is needed for correctness even without the aforementioned race condition. The BTF resources initialized by module initcall are set up once and then only looked up, so just waiting until the initcall has finished ensures correct behavior. Fixes: 541c3bad8dc5 ("bpf: Support BPF ksym variables in kernel modules") Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-2-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:44 +08:00
/* We must only consider module whose __init routine has
* finished, hence we must check for BTF_MODULE_F_LIVE flag,
* which is set from the notifier callback for
* MODULE_STATE_LIVE.
*/
if ((btf_mod->flags & BTF_MODULE_F_LIVE) && try_module_get(btf_mod->module))
res = btf_mod->module;
break;
}
mutex_unlock(&btf_module_mutex);
#endif
return res;
}
/* Returns struct btf corresponding to the struct module.
* This function can return NULL or ERR_PTR.
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
*/
static struct btf *btf_get_module_btf(const struct module *module)
{
#ifdef CONFIG_DEBUG_INFO_BTF_MODULES
struct btf_module *btf_mod, *tmp;
#endif
struct btf *btf = NULL;
if (!module) {
btf = bpf_get_btf_vmlinux();
if (!IS_ERR_OR_NULL(btf))
btf_get(btf);
return btf;
}
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
#ifdef CONFIG_DEBUG_INFO_BTF_MODULES
mutex_lock(&btf_module_mutex);
list_for_each_entry_safe(btf_mod, tmp, &btf_modules, list) {
if (btf_mod->module != module)
continue;
btf_get(btf_mod->btf);
btf = btf_mod->btf;
break;
}
mutex_unlock(&btf_module_mutex);
#endif
return btf;
}
static int check_btf_kconfigs(const struct module *module, const char *feature)
{
if (!module && IS_ENABLED(CONFIG_DEBUG_INFO_BTF)) {
pr_err("missing vmlinux BTF, cannot register %s\n", feature);
return -ENOENT;
}
if (module && IS_ENABLED(CONFIG_DEBUG_INFO_BTF_MODULES))
pr_warn("missing module BTF, cannot register %s\n", feature);
return 0;
}
BPF_CALL_4(bpf_btf_find_by_name_kind, char *, name, int, name_sz, u32, kind, int, flags)
{
struct btf *btf = NULL;
int btf_obj_fd = 0;
long ret;
if (flags)
return -EINVAL;
if (name_sz <= 1 || name[name_sz - 1])
return -EINVAL;
ret = bpf_find_btf_id(name, kind, &btf);
if (ret > 0 && btf_is_module(btf)) {
btf_obj_fd = __btf_new_fd(btf);
if (btf_obj_fd < 0) {
btf_put(btf);
return btf_obj_fd;
}
return ret | (((u64)btf_obj_fd) << 32);
}
if (ret > 0)
btf_put(btf);
return ret;
}
const struct bpf_func_proto bpf_btf_find_by_name_kind_proto = {
.func = bpf_btf_find_by_name_kind,
.gpl_only = false,
.ret_type = RET_INTEGER,
.arg1_type = ARG_PTR_TO_MEM | MEM_RDONLY,
.arg2_type = ARG_CONST_SIZE,
.arg3_type = ARG_ANYTHING,
.arg4_type = ARG_ANYTHING,
};
BTF_ID_LIST_GLOBAL(btf_tracing_ids, MAX_BTF_TRACING_TYPE)
#define BTF_TRACING_TYPE(name, type) BTF_ID(struct, type)
BTF_TRACING_TYPE_xxx
#undef BTF_TRACING_TYPE
/* Validate well-formedness of iter argument type.
* On success, return positive BTF ID of iter state's STRUCT type.
* On error, negative error is returned.
*/
int btf_check_iter_arg(struct btf *btf, const struct btf_type *func, int arg_idx)
{
const struct btf_param *arg;
const struct btf_type *t;
const char *name;
int btf_id;
if (btf_type_vlen(func) <= arg_idx)
return -EINVAL;
arg = &btf_params(func)[arg_idx];
t = btf_type_skip_modifiers(btf, arg->type, NULL);
if (!t || !btf_type_is_ptr(t))
return -EINVAL;
t = btf_type_skip_modifiers(btf, t->type, &btf_id);
if (!t || !__btf_type_is_struct(t))
return -EINVAL;
name = btf_name_by_offset(btf, t->name_off);
if (!name || strncmp(name, ITER_PREFIX, sizeof(ITER_PREFIX) - 1))
return -EINVAL;
return btf_id;
}
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
static int btf_check_iter_kfuncs(struct btf *btf, const char *func_name,
const struct btf_type *func, u32 func_flags)
{
u32 flags = func_flags & (KF_ITER_NEW | KF_ITER_NEXT | KF_ITER_DESTROY);
const char *sfx, *iter_name;
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
const struct btf_type *t;
char exp_name[128];
u32 nr_args;
int btf_id;
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
/* exactly one of KF_ITER_{NEW,NEXT,DESTROY} can be set */
if (!flags || (flags & (flags - 1)))
return -EINVAL;
/* any BPF iter kfunc should have `struct bpf_iter_<type> *` first arg */
nr_args = btf_type_vlen(func);
if (nr_args < 1)
return -EINVAL;
btf_id = btf_check_iter_arg(btf, func, 0);
if (btf_id < 0)
return btf_id;
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
/* sizeof(struct bpf_iter_<type>) should be a multiple of 8 to
* fit nicely in stack slots
*/
t = btf_type_by_id(btf, btf_id);
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
if (t->size == 0 || (t->size % 8))
return -EINVAL;
/* validate bpf_iter_<type>_{new,next,destroy}(struct bpf_iter_<type> *)
* naming pattern
*/
iter_name = btf_name_by_offset(btf, t->name_off) + sizeof(ITER_PREFIX) - 1;
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
if (flags & KF_ITER_NEW)
sfx = "new";
else if (flags & KF_ITER_NEXT)
sfx = "next";
else /* (flags & KF_ITER_DESTROY) */
sfx = "destroy";
snprintf(exp_name, sizeof(exp_name), "bpf_iter_%s_%s", iter_name, sfx);
if (strcmp(func_name, exp_name))
return -EINVAL;
/* only iter constructor should have extra arguments */
if (!(flags & KF_ITER_NEW) && nr_args != 1)
return -EINVAL;
if (flags & KF_ITER_NEXT) {
/* bpf_iter_<type>_next() should return pointer */
t = btf_type_skip_modifiers(btf, func->type, NULL);
if (!t || !btf_type_is_ptr(t))
return -EINVAL;
}
if (flags & KF_ITER_DESTROY) {
/* bpf_iter_<type>_destroy() should return void */
t = btf_type_by_id(btf, func->type);
if (!t || !btf_type_is_void(t))
return -EINVAL;
}
return 0;
}
static int btf_check_kfunc_protos(struct btf *btf, u32 func_id, u32 func_flags)
{
const struct btf_type *func;
const char *func_name;
int err;
/* any kfunc should be FUNC -> FUNC_PROTO */
func = btf_type_by_id(btf, func_id);
if (!func || !btf_type_is_func(func))
return -EINVAL;
/* sanity check kfunc name */
func_name = btf_name_by_offset(btf, func->name_off);
if (!func_name || !func_name[0])
return -EINVAL;
func = btf_type_by_id(btf, func->type);
if (!func || !btf_type_is_func_proto(func))
return -EINVAL;
if (func_flags & (KF_ITER_NEW | KF_ITER_NEXT | KF_ITER_DESTROY)) {
err = btf_check_iter_kfuncs(btf, func_name, func, func_flags);
if (err)
return err;
}
return 0;
}
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
/* Kernel Function (kfunc) BTF ID set registration API */
static int btf_populate_kfunc_set(struct btf *btf, enum btf_kfunc_hook hook,
const struct btf_kfunc_id_set *kset)
{
struct btf_kfunc_hook_filter *hook_filter;
struct btf_id_set8 *add_set = kset->set;
bool vmlinux_set = !btf_is_module(btf);
bool add_filter = !!kset->filter;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
struct btf_kfunc_set_tab *tab;
struct btf_id_set8 *set;
u32 set_cnt, i;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
int ret;
if (hook >= BTF_KFUNC_HOOK_MAX) {
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
ret = -EINVAL;
goto end;
}
if (!add_set->cnt)
return 0;
tab = btf->kfunc_set_tab;
if (tab && add_filter) {
u32 i;
hook_filter = &tab->hook_filters[hook];
for (i = 0; i < hook_filter->nr_filters; i++) {
if (hook_filter->filters[i] == kset->filter) {
add_filter = false;
break;
}
}
if (add_filter && hook_filter->nr_filters == BTF_KFUNC_FILTER_MAX_CNT) {
ret = -E2BIG;
goto end;
}
}
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
if (!tab) {
tab = kzalloc(sizeof(*tab), GFP_KERNEL | __GFP_NOWARN);
if (!tab)
return -ENOMEM;
btf->kfunc_set_tab = tab;
}
set = tab->sets[hook];
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
/* Warn when register_btf_kfunc_id_set is called twice for the same hook
* for module sets.
*/
if (WARN_ON_ONCE(set && !vmlinux_set)) {
ret = -EINVAL;
goto end;
}
/* In case of vmlinux sets, there may be more than one set being
* registered per hook. To create a unified set, we allocate a new set
* and concatenate all individual sets being registered. While each set
* is individually sorted, they may become unsorted when concatenated,
* hence re-sorting the final set again is required to make binary
* searching the set using btf_id_set8_contains function work.
*
* For module sets, we need to allocate as we may need to relocate
* BTF ids.
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
*/
set_cnt = set ? set->cnt : 0;
if (set_cnt > U32_MAX - add_set->cnt) {
ret = -EOVERFLOW;
goto end;
}
if (set_cnt + add_set->cnt > BTF_KFUNC_SET_MAX_CNT) {
ret = -E2BIG;
goto end;
}
/* Grow set */
set = krealloc(tab->sets[hook],
offsetof(struct btf_id_set8, pairs[set_cnt + add_set->cnt]),
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
GFP_KERNEL | __GFP_NOWARN);
if (!set) {
ret = -ENOMEM;
goto end;
}
/* For newly allocated set, initialize set->cnt to 0 */
if (!tab->sets[hook])
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
set->cnt = 0;
tab->sets[hook] = set;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
/* Concatenate the two sets */
memcpy(set->pairs + set->cnt, add_set->pairs, add_set->cnt * sizeof(set->pairs[0]));
/* Now that the set is copied, update with relocated BTF ids */
for (i = set->cnt; i < set->cnt + add_set->cnt; i++)
set->pairs[i].id = btf_relocate_id(btf, set->pairs[i].id);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
set->cnt += add_set->cnt;
sort(set->pairs, set->cnt, sizeof(set->pairs[0]), btf_id_cmp_func, NULL);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
if (add_filter) {
hook_filter = &tab->hook_filters[hook];
hook_filter->filters[hook_filter->nr_filters++] = kset->filter;
}
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
return 0;
end:
btf_free_kfunc_set_tab(btf);
return ret;
}
static u32 *__btf_kfunc_id_set_contains(const struct btf *btf,
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
enum btf_kfunc_hook hook,
u32 kfunc_btf_id,
const struct bpf_prog *prog)
{
struct btf_kfunc_hook_filter *hook_filter;
struct btf_id_set8 *set;
u32 *id, i;
if (hook >= BTF_KFUNC_HOOK_MAX)
return NULL;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
if (!btf->kfunc_set_tab)
return NULL;
hook_filter = &btf->kfunc_set_tab->hook_filters[hook];
for (i = 0; i < hook_filter->nr_filters; i++) {
if (hook_filter->filters[i](prog, kfunc_btf_id))
return NULL;
}
set = btf->kfunc_set_tab->sets[hook];
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
if (!set)
return NULL;
id = btf_id_set8_contains(set, kfunc_btf_id);
if (!id)
return NULL;
/* The flags for BTF ID are located next to it */
return id + 1;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
}
static int bpf_prog_type_to_kfunc_hook(enum bpf_prog_type prog_type)
{
switch (prog_type) {
case BPF_PROG_TYPE_UNSPEC:
return BTF_KFUNC_HOOK_COMMON;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
case BPF_PROG_TYPE_XDP:
return BTF_KFUNC_HOOK_XDP;
case BPF_PROG_TYPE_SCHED_CLS:
return BTF_KFUNC_HOOK_TC;
case BPF_PROG_TYPE_STRUCT_OPS:
return BTF_KFUNC_HOOK_STRUCT_OPS;
case BPF_PROG_TYPE_TRACING:
case BPF_PROG_TYPE_TRACEPOINT:
case BPF_PROG_TYPE_PERF_EVENT:
case BPF_PROG_TYPE_LSM:
return BTF_KFUNC_HOOK_TRACING;
case BPF_PROG_TYPE_SYSCALL:
return BTF_KFUNC_HOOK_SYSCALL;
case BPF_PROG_TYPE_CGROUP_SKB:
case BPF_PROG_TYPE_CGROUP_SOCK:
case BPF_PROG_TYPE_CGROUP_DEVICE:
case BPF_PROG_TYPE_CGROUP_SOCK_ADDR:
case BPF_PROG_TYPE_CGROUP_SOCKOPT:
case BPF_PROG_TYPE_CGROUP_SYSCTL:
return BTF_KFUNC_HOOK_CGROUP;
case BPF_PROG_TYPE_SCHED_ACT:
return BTF_KFUNC_HOOK_SCHED_ACT;
case BPF_PROG_TYPE_SK_SKB:
return BTF_KFUNC_HOOK_SK_SKB;
case BPF_PROG_TYPE_SOCKET_FILTER:
return BTF_KFUNC_HOOK_SOCKET_FILTER;
case BPF_PROG_TYPE_LWT_OUT:
case BPF_PROG_TYPE_LWT_IN:
case BPF_PROG_TYPE_LWT_XMIT:
case BPF_PROG_TYPE_LWT_SEG6LOCAL:
return BTF_KFUNC_HOOK_LWT;
case BPF_PROG_TYPE_NETFILTER:
return BTF_KFUNC_HOOK_NETFILTER;
case BPF_PROG_TYPE_KPROBE:
return BTF_KFUNC_HOOK_KPROBE;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
default:
return BTF_KFUNC_HOOK_MAX;
}
}
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
/* Caution:
* Reference to the module (obtained using btf_try_get_module) corresponding to
* the struct btf *MUST* be held when calling this function from verifier
* context. This is usually true as we stash references in prog's kfunc_btf_tab;
* keeping the reference for the duration of the call provides the necessary
* protection for looking up a well-formed btf->kfunc_set_tab.
*/
u32 *btf_kfunc_id_set_contains(const struct btf *btf,
u32 kfunc_btf_id,
const struct bpf_prog *prog)
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
{
enum bpf_prog_type prog_type = resolve_prog_type(prog);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
enum btf_kfunc_hook hook;
u32 *kfunc_flags;
kfunc_flags = __btf_kfunc_id_set_contains(btf, BTF_KFUNC_HOOK_COMMON, kfunc_btf_id, prog);
if (kfunc_flags)
return kfunc_flags;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
hook = bpf_prog_type_to_kfunc_hook(prog_type);
return __btf_kfunc_id_set_contains(btf, hook, kfunc_btf_id, prog);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
}
u32 *btf_kfunc_is_modify_return(const struct btf *btf, u32 kfunc_btf_id,
const struct bpf_prog *prog)
{
return __btf_kfunc_id_set_contains(btf, BTF_KFUNC_HOOK_FMODRET, kfunc_btf_id, prog);
}
static int __register_btf_kfunc_id_set(enum btf_kfunc_hook hook,
const struct btf_kfunc_id_set *kset)
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
{
struct btf *btf;
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
int ret, i;
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
btf = btf_get_module_btf(kset->owner);
if (!btf)
return check_btf_kconfigs(kset->owner, "kfunc");
if (IS_ERR(btf))
return PTR_ERR(btf);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
for (i = 0; i < kset->set->cnt; i++) {
ret = btf_check_kfunc_protos(btf, btf_relocate_id(btf, kset->set->pairs[i].id),
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
kset->set->pairs[i].flags);
if (ret)
goto err_out;
}
ret = btf_populate_kfunc_set(btf, hook, kset);
bpf: add iterator kfuncs registration and validation logic Add ability to register kfuncs that implement BPF open-coded iterator contract and enforce naming and function proto convention. Enforcement happens at the time of kfunc registration and significantly simplifies the rest of iterators logic in the verifier. More details follow in subsequent patches, but we enforce the following conditions. All kfuncs (constructor, next, destructor) have to be named consistenly as bpf_iter_<type>_{new,next,destroy}(), respectively. <type> represents iterator type, and iterator state should be represented as a matching `struct bpf_iter_<type>` state type. Also, all iter kfuncs should have a pointer to this `struct bpf_iter_<type>` as the very first argument. Additionally: - Constructor, i.e., bpf_iter_<type>_new(), can have arbitrary extra number of arguments. Return type is not enforced either. - Next method, i.e., bpf_iter_<type>_next(), has to return a pointer type and should have exactly one argument: `struct bpf_iter_<type> *` (const/volatile/restrict and typedefs are ignored). - Destructor, i.e., bpf_iter_<type>_destroy(), should return void and should have exactly one argument, similar to the next method. - struct bpf_iter_<type> size is enforced to be positive and a multiple of 8 bytes (to fit stack slots correctly). Such strictness and consistency allows to build generic helpers abstracting important, but boilerplate, details to be able to use open-coded iterators effectively and ergonomically (see bpf_for_each() in subsequent patches). It also simplifies the verifier logic in some places. At the same time, this doesn't hurt generality of possible iterator implementations. Win-win. Constructor kfunc is marked with a new KF_ITER_NEW flags, next method is marked with KF_ITER_NEXT (and should also have KF_RET_NULL, of course), while destructor kfunc is marked as KF_ITER_DESTROY. Additionally, we add a trivial kfunc name validation: it should be a valid non-NULL and non-empty string. Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20230308184121.1165081-3-andrii@kernel.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-03-09 02:41:15 +08:00
err_out:
btf_put(btf);
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
return ret;
}
/* This function must be invoked only from initcalls/module init functions */
int register_btf_kfunc_id_set(enum bpf_prog_type prog_type,
const struct btf_kfunc_id_set *kset)
{
enum btf_kfunc_hook hook;
/* All kfuncs need to be tagged as such in BTF.
* WARN() for initcall registrations that do not check errors.
*/
if (!(kset->set->flags & BTF_SET8_KFUNCS)) {
WARN_ON(!kset->owner);
return -EINVAL;
}
hook = bpf_prog_type_to_kfunc_hook(prog_type);
return __register_btf_kfunc_id_set(hook, kset);
}
bpf: Populate kfunc BTF ID sets in struct btf This patch prepares the kernel to support putting all kinds of kfunc BTF ID sets in the struct btf itself. The various kernel subsystems will make register_btf_kfunc_id_set call in the initcalls (for built-in code and modules). The 'hook' is one of the many program types, e.g. XDP and TC/SCHED_CLS, STRUCT_OPS, and 'types' are check (allowed or not), acquire, release, and ret_null (with PTR_TO_BTF_ID_OR_NULL return type). A maximum of BTF_KFUNC_SET_MAX_CNT (32) kfunc BTF IDs are permitted in a set of certain hook and type for vmlinux sets, since they are allocated on demand, and otherwise set as NULL. Module sets can only be registered once per hook and type, hence they are directly assigned. A new btf_kfunc_id_set_contains function is exposed for use in verifier, this new method is faster than the existing list searching method, and is also automatic. It also lets other code not care whether the set is unallocated or not. Note that module code can only do single register_btf_kfunc_id_set call per hook. This is why sorting is only done for in-kernel vmlinux sets, because there might be multiple sets for the same hook and type that must be concatenated, hence sorting them is required to ensure bsearch in btf_id_set_contains continues to work correctly. Next commit will update the kernel users to make use of this infrastructure. Finally, add __maybe_unused annotation for BTF ID macros for the !CONFIG_DEBUG_INFO_BTF case, so that they don't produce warnings during build time. The previous patch is also needed to provide synchronization against initialization for module BTF's kfunc_set_tab introduced here, as described below: The kfunc_set_tab pointer in struct btf is write-once (if we consider the registration phase (comprised of multiple register_btf_kfunc_id_set calls) as a single operation). In this sense, once it has been fully prepared, it isn't modified, only used for lookup (from the verifier context). For btf_vmlinux, it is initialized fully during the do_initcalls phase, which happens fairly early in the boot process, before any processes are present. This also eliminates the possibility of bpf_check being called at that point, thus relieving us of ensuring any synchronization between the registration and lookup function (btf_kfunc_id_set_contains). However, the case for module BTF is a bit tricky. The BTF is parsed, prepared, and published from the MODULE_STATE_COMING notifier callback. After this, the module initcalls are invoked, where our registration function will be called to populate the kfunc_set_tab for module BTF. At this point, BTF may be available to userspace while its corresponding module is still intializing. A BTF fd can then be passed to verifier using bpf syscall (e.g. for kfunc call insn). Hence, there is a race window where verifier may concurrently try to lookup the kfunc_set_tab. To prevent this race, we must ensure the operations are serialized, or waiting for the __init functions to complete. In the earlier registration API, this race was alleviated as verifier bpf_check_mod_kfunc_call didn't find the kfunc BTF ID until it was added by the registration function (called usually at the end of module __init function after all module resources have been initialized). If the verifier made the check_kfunc_call before kfunc BTF ID was added to the list, it would fail verification (saying call isn't allowed). The access to list was protected using a mutex. Now, it would still fail verification, but for a different reason (returning ENXIO due to the failed btf_try_get_module call in add_kfunc_call), because if the __init call is in progress the module will be in the middle of MODULE_STATE_COMING -> MODULE_STATE_LIVE transition, and the BTF_MODULE_LIVE flag for btf_module instance will not be set, so the btf_try_get_module call will fail. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Link: https://lore.kernel.org/r/20220114163953.1455836-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-01-15 00:39:45 +08:00
EXPORT_SYMBOL_GPL(register_btf_kfunc_id_set);
Merge https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next Andrii Nakryiko says: ==================== bpf-next 2021-12-10 v2 We've added 115 non-merge commits during the last 26 day(s) which contain a total of 182 files changed, 5747 insertions(+), 2564 deletions(-). The main changes are: 1) Various samples fixes, from Alexander Lobakin. 2) BPF CO-RE support in kernel and light skeleton, from Alexei Starovoitov. 3) A batch of new unified APIs for libbpf, logging improvements, version querying, etc. Also a batch of old deprecations for old APIs and various bug fixes, in preparation for libbpf 1.0, from Andrii Nakryiko. 4) BPF documentation reorganization and improvements, from Christoph Hellwig and Dave Tucker. 5) Support for declarative initialization of BPF_MAP_TYPE_PROG_ARRAY in libbpf, from Hengqi Chen. 6) Verifier log fixes, from Hou Tao. 7) Runtime-bounded loops support with bpf_loop() helper, from Joanne Koong. 8) Extend branch record capturing to all platforms that support it, from Kajol Jain. 9) Light skeleton codegen improvements, from Kumar Kartikeya Dwivedi. 10) bpftool doc-generating script improvements, from Quentin Monnet. 11) Two libbpf v0.6 bug fixes, from Shuyi Cheng and Vincent Minet. 12) Deprecation warning fix for perf/bpf_counter, from Song Liu. 13) MAX_TAIL_CALL_CNT unification and MIPS build fix for libbpf, from Tiezhu Yang. 14) BTF_KING_TYPE_TAG follow-up fixes, from Yonghong Song. 15) Selftests fixes and improvements, from Ilya Leoshkevich, Jean-Philippe Brucker, Jiri Olsa, Maxim Mikityanskiy, Tirthendu Sarkar, Yucong Sun, and others. * https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (115 commits) libbpf: Add "bool skipped" to struct bpf_map libbpf: Fix typo in btf__dedup@LIBBPF_0.0.2 definition bpftool: Switch bpf_object__load_xattr() to bpf_object__load() selftests/bpf: Remove the only use of deprecated bpf_object__load_xattr() selftests/bpf: Add test for libbpf's custom log_buf behavior selftests/bpf: Replace all uses of bpf_load_btf() with bpf_btf_load() libbpf: Deprecate bpf_object__load_xattr() libbpf: Add per-program log buffer setter and getter libbpf: Preserve kernel error code and remove kprobe prog type guessing libbpf: Improve logging around BPF program loading libbpf: Allow passing user log setting through bpf_object_open_opts libbpf: Allow passing preallocated log_buf when loading BTF into kernel libbpf: Add OPTS-based bpf_btf_load() API libbpf: Fix bpf_prog_load() log_buf logic for log_level 0 samples/bpf: Remove unneeded variable bpf: Remove redundant assignment to pointer t selftests/bpf: Fix a compilation warning perf/bpf_counter: Use bpf_map_create instead of bpf_create_map samples: bpf: Fix 'unknown warning group' build warning on Clang samples: bpf: Fix xdp_sample_user.o linking with Clang ... ==================== Link: https://lore.kernel.org/r/20211210234746.2100561-1-andrii@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2021-12-11 07:56:10 +08:00
/* This function must be invoked only from initcalls/module init functions */
int register_btf_fmodret_id_set(const struct btf_kfunc_id_set *kset)
{
return __register_btf_kfunc_id_set(BTF_KFUNC_HOOK_FMODRET, kset);
}
EXPORT_SYMBOL_GPL(register_btf_fmodret_id_set);
s32 btf_find_dtor_kfunc(struct btf *btf, u32 btf_id)
{
struct btf_id_dtor_kfunc_tab *tab = btf->dtor_kfunc_tab;
struct btf_id_dtor_kfunc *dtor;
if (!tab)
return -ENOENT;
/* Even though the size of tab->dtors[0] is > sizeof(u32), we only need
* to compare the first u32 with btf_id, so we can reuse btf_id_cmp_func.
*/
BUILD_BUG_ON(offsetof(struct btf_id_dtor_kfunc, btf_id) != 0);
dtor = bsearch(&btf_id, tab->dtors, tab->cnt, sizeof(tab->dtors[0]), btf_id_cmp_func);
if (!dtor)
return -ENOENT;
return dtor->kfunc_btf_id;
}
bpf: Wire up freeing of referenced kptr A destructor kfunc can be defined as void func(type *), where type may be void or any other pointer type as per convenience. In this patch, we ensure that the type is sane and capture the function pointer into off_desc of ptr_off_tab for the specific pointer offset, with the invariant that the dtor pointer is always set when 'kptr_ref' tag is applied to the pointer's pointee type, which is indicated by the flag BPF_MAP_VALUE_OFF_F_REF. Note that only BTF IDs whose destructor kfunc is registered, thus become the allowed BTF IDs for embedding as referenced kptr. Hence it serves the purpose of finding dtor kfunc BTF ID, as well acting as a check against the whitelist of allowed BTF IDs for this purpose. Finally, wire up the actual freeing of the referenced pointer if any at all available offsets, so that no references are leaked after the BPF map goes away and the BPF program previously moved the ownership a referenced pointer into it. The behavior is similar to BPF timers, where bpf_map_{update,delete}_elem will free any existing referenced kptr. The same case is with LRU map's bpf_lru_push_free/htab_lru_push_free functions, which are extended to reset unreferenced and free referenced kptr. Note that unlike BPF timers, kptr is not reset or freed when map uref drops to zero. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-8-memxor@gmail.com
2022-04-25 05:48:55 +08:00
static int btf_check_dtor_kfuncs(struct btf *btf, const struct btf_id_dtor_kfunc *dtors, u32 cnt)
{
const struct btf_type *dtor_func, *dtor_func_proto, *t;
const struct btf_param *args;
s32 dtor_btf_id;
u32 nr_args, i;
for (i = 0; i < cnt; i++) {
dtor_btf_id = btf_relocate_id(btf, dtors[i].kfunc_btf_id);
bpf: Wire up freeing of referenced kptr A destructor kfunc can be defined as void func(type *), where type may be void or any other pointer type as per convenience. In this patch, we ensure that the type is sane and capture the function pointer into off_desc of ptr_off_tab for the specific pointer offset, with the invariant that the dtor pointer is always set when 'kptr_ref' tag is applied to the pointer's pointee type, which is indicated by the flag BPF_MAP_VALUE_OFF_F_REF. Note that only BTF IDs whose destructor kfunc is registered, thus become the allowed BTF IDs for embedding as referenced kptr. Hence it serves the purpose of finding dtor kfunc BTF ID, as well acting as a check against the whitelist of allowed BTF IDs for this purpose. Finally, wire up the actual freeing of the referenced pointer if any at all available offsets, so that no references are leaked after the BPF map goes away and the BPF program previously moved the ownership a referenced pointer into it. The behavior is similar to BPF timers, where bpf_map_{update,delete}_elem will free any existing referenced kptr. The same case is with LRU map's bpf_lru_push_free/htab_lru_push_free functions, which are extended to reset unreferenced and free referenced kptr. Note that unlike BPF timers, kptr is not reset or freed when map uref drops to zero. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-8-memxor@gmail.com
2022-04-25 05:48:55 +08:00
dtor_func = btf_type_by_id(btf, dtor_btf_id);
if (!dtor_func || !btf_type_is_func(dtor_func))
return -EINVAL;
dtor_func_proto = btf_type_by_id(btf, dtor_func->type);
if (!dtor_func_proto || !btf_type_is_func_proto(dtor_func_proto))
return -EINVAL;
/* Make sure the prototype of the destructor kfunc is 'void func(type *)' */
t = btf_type_by_id(btf, dtor_func_proto->type);
if (!t || !btf_type_is_void(t))
return -EINVAL;
nr_args = btf_type_vlen(dtor_func_proto);
if (nr_args != 1)
return -EINVAL;
args = btf_params(dtor_func_proto);
t = btf_type_by_id(btf, args[0].type);
/* Allow any pointer type, as width on targets Linux supports
* will be same for all pointer types (i.e. sizeof(void *))
*/
if (!t || !btf_type_is_ptr(t))
return -EINVAL;
}
return 0;
}
/* This function must be invoked only from initcalls/module init functions */
int register_btf_id_dtor_kfuncs(const struct btf_id_dtor_kfunc *dtors, u32 add_cnt,
struct module *owner)
{
struct btf_id_dtor_kfunc_tab *tab;
struct btf *btf;
u32 tab_cnt, i;
int ret;
btf = btf_get_module_btf(owner);
if (!btf)
return check_btf_kconfigs(owner, "dtor kfuncs");
if (IS_ERR(btf))
return PTR_ERR(btf);
if (add_cnt >= BTF_DTOR_KFUNC_MAX_CNT) {
pr_err("cannot register more than %d kfunc destructors\n", BTF_DTOR_KFUNC_MAX_CNT);
ret = -E2BIG;
goto end;
}
bpf: Wire up freeing of referenced kptr A destructor kfunc can be defined as void func(type *), where type may be void or any other pointer type as per convenience. In this patch, we ensure that the type is sane and capture the function pointer into off_desc of ptr_off_tab for the specific pointer offset, with the invariant that the dtor pointer is always set when 'kptr_ref' tag is applied to the pointer's pointee type, which is indicated by the flag BPF_MAP_VALUE_OFF_F_REF. Note that only BTF IDs whose destructor kfunc is registered, thus become the allowed BTF IDs for embedding as referenced kptr. Hence it serves the purpose of finding dtor kfunc BTF ID, as well acting as a check against the whitelist of allowed BTF IDs for this purpose. Finally, wire up the actual freeing of the referenced pointer if any at all available offsets, so that no references are leaked after the BPF map goes away and the BPF program previously moved the ownership a referenced pointer into it. The behavior is similar to BPF timers, where bpf_map_{update,delete}_elem will free any existing referenced kptr. The same case is with LRU map's bpf_lru_push_free/htab_lru_push_free functions, which are extended to reset unreferenced and free referenced kptr. Note that unlike BPF timers, kptr is not reset or freed when map uref drops to zero. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20220424214901.2743946-8-memxor@gmail.com
2022-04-25 05:48:55 +08:00
/* Ensure that the prototype of dtor kfuncs being registered is sane */
ret = btf_check_dtor_kfuncs(btf, dtors, add_cnt);
if (ret < 0)
goto end;
tab = btf->dtor_kfunc_tab;
/* Only one call allowed for modules */
if (WARN_ON_ONCE(tab && btf_is_module(btf))) {
ret = -EINVAL;
goto end;
}
tab_cnt = tab ? tab->cnt : 0;
if (tab_cnt > U32_MAX - add_cnt) {
ret = -EOVERFLOW;
goto end;
}
if (tab_cnt + add_cnt >= BTF_DTOR_KFUNC_MAX_CNT) {
pr_err("cannot register more than %d kfunc destructors\n", BTF_DTOR_KFUNC_MAX_CNT);
ret = -E2BIG;
goto end;
}
tab = krealloc(btf->dtor_kfunc_tab,
offsetof(struct btf_id_dtor_kfunc_tab, dtors[tab_cnt + add_cnt]),
GFP_KERNEL | __GFP_NOWARN);
if (!tab) {
ret = -ENOMEM;
goto end;
}
if (!btf->dtor_kfunc_tab)
tab->cnt = 0;
btf->dtor_kfunc_tab = tab;
memcpy(tab->dtors + tab->cnt, dtors, add_cnt * sizeof(tab->dtors[0]));
/* remap BTF ids based on BTF relocation (if any) */
for (i = tab_cnt; i < tab_cnt + add_cnt; i++) {
tab->dtors[i].btf_id = btf_relocate_id(btf, tab->dtors[i].btf_id);
tab->dtors[i].kfunc_btf_id = btf_relocate_id(btf, tab->dtors[i].kfunc_btf_id);
}
tab->cnt += add_cnt;
sort(tab->dtors, tab->cnt, sizeof(tab->dtors[0]), btf_id_cmp_func, NULL);
end:
if (ret)
btf_free_dtor_kfunc_tab(btf);
btf_put(btf);
return ret;
}
EXPORT_SYMBOL_GPL(register_btf_id_dtor_kfuncs);
#define MAX_TYPES_ARE_COMPAT_DEPTH 2
/* Check local and target types for compatibility. This check is used for
* type-based CO-RE relocations and follow slightly different rules than
* field-based relocations. This function assumes that root types were already
* checked for name match. Beyond that initial root-level name check, names
* are completely ignored. Compatibility rules are as follows:
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
* - any two STRUCTs/UNIONs/FWDs/ENUMs/INTs/ENUM64s are considered compatible, but
* kind should match for local and target types (i.e., STRUCT is not
* compatible with UNION);
bpf: Add btf enum64 support Currently, BTF only supports upto 32bit enum value with BTF_KIND_ENUM. But in kernel, some enum indeed has 64bit values, e.g., in uapi bpf.h, we have enum { BPF_F_INDEX_MASK = 0xffffffffULL, BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, BPF_F_CTXLEN_MASK = (0xfffffULL << 32), }; In this case, BTF_KIND_ENUM will encode the value of BPF_F_CTXLEN_MASK as 0, which certainly is incorrect. This patch added a new btf kind, BTF_KIND_ENUM64, which permits 64bit value to cover the above use case. The BTF_KIND_ENUM64 has the following three fields followed by the common type: struct bpf_enum64 { __u32 nume_off; __u32 val_lo32; __u32 val_hi32; }; Currently, btf type section has an alignment of 4 as all element types are u32. Representing the value with __u64 will introduce a pad for bpf_enum64 and may also introduce misalignment for the 64bit value. Hence, two members of val_hi32 and val_lo32 are chosen to avoid these issues. The kflag is also introduced for BTF_KIND_ENUM and BTF_KIND_ENUM64 to indicate whether the value is signed or unsigned. The kflag intends to provide consistent output of BTF C fortmat with the original source code. For example, the original BTF_KIND_ENUM bit value is 0xffffffff. The format C has two choices, printing out 0xffffffff or -1 and current libbpf prints out as unsigned value. But if the signedness is preserved in btf, the value can be printed the same as the original source code. The kflag value 0 means unsigned values, which is consistent to the default by libbpf and should also cover most cases as well. The new BTF_KIND_ENUM64 is intended to support the enum value represented as 64bit value. But it can represent all BTF_KIND_ENUM values as well. The compiler ([1]) and pahole will generate BTF_KIND_ENUM64 only if the value has to be represented with 64 bits. In addition, a static inline function btf_kind_core_compat() is introduced which will be used later when libbpf relo_core.c changed. Here the kernel shares the same relo_core.c with libbpf. [1] https://reviews.llvm.org/D124641 Acked-by: Andrii Nakryiko <andrii@kernel.org> Signed-off-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20220607062600.3716578-1-yhs@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-06-07 14:26:00 +08:00
* - for ENUMs/ENUM64s, the size is ignored;
* - for INT, size and signedness are ignored;
* - for ARRAY, dimensionality is ignored, element types are checked for
* compatibility recursively;
* - CONST/VOLATILE/RESTRICT modifiers are ignored;
* - TYPEDEFs/PTRs are compatible if types they pointing to are compatible;
* - FUNC_PROTOs are compatible if they have compatible signature: same
* number of input args and compatible return and argument types.
* These rules are not set in stone and probably will be adjusted as we get
* more experience with using BPF CO-RE relocations.
*/
int bpf_core_types_are_compat(const struct btf *local_btf, __u32 local_id,
const struct btf *targ_btf, __u32 targ_id)
{
return __bpf_core_types_are_compat(local_btf, local_id, targ_btf, targ_id,
MAX_TYPES_ARE_COMPAT_DEPTH);
}
#define MAX_TYPES_MATCH_DEPTH 2
int bpf_core_types_match(const struct btf *local_btf, u32 local_id,
const struct btf *targ_btf, u32 targ_id)
{
return __bpf_core_types_match(local_btf, local_id, targ_btf, targ_id, false,
MAX_TYPES_MATCH_DEPTH);
}
static bool bpf_core_is_flavor_sep(const char *s)
{
/* check X___Y name pattern, where X and Y are not underscores */
return s[0] != '_' && /* X */
s[1] == '_' && s[2] == '_' && s[3] == '_' && /* ___ */
s[4] != '_'; /* Y */
}
size_t bpf_core_essential_name_len(const char *name)
{
size_t n = strlen(name);
int i;
for (i = n - 5; i >= 0; i--) {
if (bpf_core_is_flavor_sep(name + i))
return i + 1;
}
return n;
}
static void bpf_free_cands(struct bpf_cand_cache *cands)
{
if (!cands->cnt)
/* empty candidate array was allocated on stack */
return;
kfree(cands);
}
static void bpf_free_cands_from_cache(struct bpf_cand_cache *cands)
{
kfree(cands->name);
kfree(cands);
}
#define VMLINUX_CAND_CACHE_SIZE 31
static struct bpf_cand_cache *vmlinux_cand_cache[VMLINUX_CAND_CACHE_SIZE];
#define MODULE_CAND_CACHE_SIZE 31
static struct bpf_cand_cache *module_cand_cache[MODULE_CAND_CACHE_SIZE];
static void __print_cand_cache(struct bpf_verifier_log *log,
struct bpf_cand_cache **cache,
int cache_size)
{
struct bpf_cand_cache *cc;
int i, j;
for (i = 0; i < cache_size; i++) {
cc = cache[i];
if (!cc)
continue;
bpf_log(log, "[%d]%s(", i, cc->name);
for (j = 0; j < cc->cnt; j++) {
bpf_log(log, "%d", cc->cands[j].id);
if (j < cc->cnt - 1)
bpf_log(log, " ");
}
bpf_log(log, "), ");
}
}
static void print_cand_cache(struct bpf_verifier_log *log)
{
mutex_lock(&cand_cache_mutex);
bpf_log(log, "vmlinux_cand_cache:");
__print_cand_cache(log, vmlinux_cand_cache, VMLINUX_CAND_CACHE_SIZE);
bpf_log(log, "\nmodule_cand_cache:");
__print_cand_cache(log, module_cand_cache, MODULE_CAND_CACHE_SIZE);
bpf_log(log, "\n");
mutex_unlock(&cand_cache_mutex);
}
static u32 hash_cands(struct bpf_cand_cache *cands)
{
return jhash(cands->name, cands->name_len, 0);
}
static struct bpf_cand_cache *check_cand_cache(struct bpf_cand_cache *cands,
struct bpf_cand_cache **cache,
int cache_size)
{
struct bpf_cand_cache *cc = cache[hash_cands(cands) % cache_size];
if (cc && cc->name_len == cands->name_len &&
!strncmp(cc->name, cands->name, cands->name_len))
return cc;
return NULL;
}
static size_t sizeof_cands(int cnt)
{
return offsetof(struct bpf_cand_cache, cands[cnt]);
}
static struct bpf_cand_cache *populate_cand_cache(struct bpf_cand_cache *cands,
struct bpf_cand_cache **cache,
int cache_size)
{
struct bpf_cand_cache **cc = &cache[hash_cands(cands) % cache_size], *new_cands;
if (*cc) {
bpf_free_cands_from_cache(*cc);
*cc = NULL;
}
new_cands = kmemdup(cands, sizeof_cands(cands->cnt), GFP_KERNEL);
if (!new_cands) {
bpf_free_cands(cands);
return ERR_PTR(-ENOMEM);
}
/* strdup the name, since it will stay in cache.
* the cands->name points to strings in prog's BTF and the prog can be unloaded.
*/
new_cands->name = kmemdup_nul(cands->name, cands->name_len, GFP_KERNEL);
bpf_free_cands(cands);
if (!new_cands->name) {
kfree(new_cands);
return ERR_PTR(-ENOMEM);
}
*cc = new_cands;
return new_cands;
}
#ifdef CONFIG_DEBUG_INFO_BTF_MODULES
static void __purge_cand_cache(struct btf *btf, struct bpf_cand_cache **cache,
int cache_size)
{
struct bpf_cand_cache *cc;
int i, j;
for (i = 0; i < cache_size; i++) {
cc = cache[i];
if (!cc)
continue;
if (!btf) {
/* when new module is loaded purge all of module_cand_cache,
* since new module might have candidates with the name
* that matches cached cands.
*/
bpf_free_cands_from_cache(cc);
cache[i] = NULL;
continue;
}
/* when module is unloaded purge cache entries
* that match module's btf
*/
for (j = 0; j < cc->cnt; j++)
if (cc->cands[j].btf == btf) {
bpf_free_cands_from_cache(cc);
cache[i] = NULL;
break;
}
}
}
static void purge_cand_cache(struct btf *btf)
{
mutex_lock(&cand_cache_mutex);
__purge_cand_cache(btf, module_cand_cache, MODULE_CAND_CACHE_SIZE);
mutex_unlock(&cand_cache_mutex);
}
#endif
static struct bpf_cand_cache *
bpf_core_add_cands(struct bpf_cand_cache *cands, const struct btf *targ_btf,
int targ_start_id)
{
struct bpf_cand_cache *new_cands;
const struct btf_type *t;
const char *targ_name;
size_t targ_essent_len;
int n, i;
n = btf_nr_types(targ_btf);
for (i = targ_start_id; i < n; i++) {
t = btf_type_by_id(targ_btf, i);
if (btf_kind(t) != cands->kind)
continue;
targ_name = btf_name_by_offset(targ_btf, t->name_off);
if (!targ_name)
continue;
/* the resched point is before strncmp to make sure that search
* for non-existing name will have a chance to schedule().
*/
cond_resched();
if (strncmp(cands->name, targ_name, cands->name_len) != 0)
continue;
targ_essent_len = bpf_core_essential_name_len(targ_name);
if (targ_essent_len != cands->name_len)
continue;
/* most of the time there is only one candidate for a given kind+name pair */
new_cands = kmalloc(sizeof_cands(cands->cnt + 1), GFP_KERNEL);
if (!new_cands) {
bpf_free_cands(cands);
return ERR_PTR(-ENOMEM);
}
memcpy(new_cands, cands, sizeof_cands(cands->cnt));
bpf_free_cands(cands);
cands = new_cands;
cands->cands[cands->cnt].btf = targ_btf;
cands->cands[cands->cnt].id = i;
cands->cnt++;
}
return cands;
}
static struct bpf_cand_cache *
bpf_core_find_cands(struct bpf_core_ctx *ctx, u32 local_type_id)
{
struct bpf_cand_cache *cands, *cc, local_cand = {};
const struct btf *local_btf = ctx->btf;
const struct btf_type *local_type;
const struct btf *main_btf;
size_t local_essent_len;
struct btf *mod_btf;
const char *name;
int id;
main_btf = bpf_get_btf_vmlinux();
if (IS_ERR(main_btf))
return ERR_CAST(main_btf);
if (!main_btf)
return ERR_PTR(-EINVAL);
local_type = btf_type_by_id(local_btf, local_type_id);
if (!local_type)
return ERR_PTR(-EINVAL);
name = btf_name_by_offset(local_btf, local_type->name_off);
if (str_is_empty(name))
return ERR_PTR(-EINVAL);
local_essent_len = bpf_core_essential_name_len(name);
cands = &local_cand;
cands->name = name;
cands->kind = btf_kind(local_type);
cands->name_len = local_essent_len;
cc = check_cand_cache(cands, vmlinux_cand_cache, VMLINUX_CAND_CACHE_SIZE);
/* cands is a pointer to stack here */
if (cc) {
if (cc->cnt)
return cc;
goto check_modules;
}
/* Attempt to find target candidates in vmlinux BTF first */
cands = bpf_core_add_cands(cands, main_btf, 1);
if (IS_ERR(cands))
return ERR_CAST(cands);
/* cands is a pointer to kmalloced memory here if cands->cnt > 0 */
/* populate cache even when cands->cnt == 0 */
cc = populate_cand_cache(cands, vmlinux_cand_cache, VMLINUX_CAND_CACHE_SIZE);
if (IS_ERR(cc))
return ERR_CAST(cc);
/* if vmlinux BTF has any candidate, don't go for module BTFs */
if (cc->cnt)
return cc;
check_modules:
/* cands is a pointer to stack here and cands->cnt == 0 */
cc = check_cand_cache(cands, module_cand_cache, MODULE_CAND_CACHE_SIZE);
if (cc)
/* if cache has it return it even if cc->cnt == 0 */
return cc;
/* If candidate is not found in vmlinux's BTF then search in module's BTFs */
spin_lock_bh(&btf_idr_lock);
idr_for_each_entry(&btf_idr, mod_btf, id) {
if (!btf_is_module(mod_btf))
continue;
/* linear search could be slow hence unlock/lock
* the IDR to avoiding holding it for too long
*/
btf_get(mod_btf);
spin_unlock_bh(&btf_idr_lock);
cands = bpf_core_add_cands(cands, mod_btf, btf_nr_types(main_btf));
btf_put(mod_btf);
if (IS_ERR(cands))
return ERR_CAST(cands);
spin_lock_bh(&btf_idr_lock);
}
spin_unlock_bh(&btf_idr_lock);
/* cands is a pointer to kmalloced memory here if cands->cnt > 0
* or pointer to stack if cands->cnd == 0.
* Copy it into the cache even when cands->cnt == 0 and
* return the result.
*/
return populate_cand_cache(cands, module_cand_cache, MODULE_CAND_CACHE_SIZE);
}
int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo,
int relo_idx, void *insn)
{
bool need_cands = relo->kind != BPF_CORE_TYPE_ID_LOCAL;
struct bpf_core_cand_list cands = {};
struct bpf_core_relo_res targ_res;
struct bpf_core_spec *specs;
const struct btf_type *type;
int err;
/* ~4k of temp memory necessary to convert LLVM spec like "0:1:0:5"
* into arrays of btf_ids of struct fields and array indices.
*/
specs = kcalloc(3, sizeof(*specs), GFP_KERNEL);
if (!specs)
return -ENOMEM;
type = btf_type_by_id(ctx->btf, relo->type_id);
if (!type) {
bpf_log(ctx->log, "relo #%u: bad type id %u\n",
relo_idx, relo->type_id);
kfree(specs);
return -EINVAL;
}
if (need_cands) {
struct bpf_cand_cache *cc;
int i;
mutex_lock(&cand_cache_mutex);
cc = bpf_core_find_cands(ctx, relo->type_id);
if (IS_ERR(cc)) {
bpf_log(ctx->log, "target candidate search failed for %d\n",
relo->type_id);
err = PTR_ERR(cc);
goto out;
}
if (cc->cnt) {
cands.cands = kcalloc(cc->cnt, sizeof(*cands.cands), GFP_KERNEL);
if (!cands.cands) {
err = -ENOMEM;
goto out;
}
}
for (i = 0; i < cc->cnt; i++) {
bpf_log(ctx->log,
"CO-RE relocating %s %s: found target candidate [%d]\n",
btf_kind_str[cc->kind], cc->name, cc->cands[i].id);
cands.cands[i].btf = cc->cands[i].btf;
cands.cands[i].id = cc->cands[i].id;
}
cands.len = cc->cnt;
/* cand_cache_mutex needs to span the cache lookup and
* copy of btf pointer into bpf_core_cand_list,
* since module can be unloaded while bpf_core_calc_relo_insn
* is working with module's btf.
*/
}
err = bpf_core_calc_relo_insn((void *)ctx->log, relo, relo_idx, ctx->btf, &cands, specs,
&targ_res);
if (err)
goto out;
err = bpf_core_patch_insn((void *)ctx->log, insn, relo->insn_off / 8, relo, relo_idx,
&targ_res);
out:
kfree(specs);
if (need_cands) {
kfree(cands.cands);
mutex_unlock(&cand_cache_mutex);
if (ctx->log->level & BPF_LOG_LEVEL2)
print_cand_cache(ctx->log);
}
return err;
}
bool btf_nested_type_is_trusted(struct bpf_verifier_log *log,
const struct bpf_reg_state *reg,
const char *field_name, u32 btf_id, const char *suffix)
{
struct btf *btf = reg->btf;
const struct btf_type *walk_type, *safe_type;
const char *tname;
char safe_tname[64];
long ret, safe_id;
const struct btf_member *member;
u32 i;
walk_type = btf_type_by_id(btf, reg->btf_id);
if (!walk_type)
return false;
tname = btf_name_by_offset(btf, walk_type->name_off);
bpf: Refactor RCU enforcement in the verifier. bpf_rcu_read_lock/unlock() are only available in clang compiled kernels. Lack of such key mechanism makes it impossible for sleepable bpf programs to use RCU pointers. Allow bpf_rcu_read_lock/unlock() in GCC compiled kernels (though GCC doesn't support btf_type_tag yet) and allowlist certain field dereferences in important data structures like tast_struct, cgroup, socket that are used by sleepable programs either as RCU pointer or full trusted pointer (which is valid outside of RCU CS). Use BTF_TYPE_SAFE_RCU and BTF_TYPE_SAFE_TRUSTED macros for such tagging. They will be removed once GCC supports btf_type_tag. With that refactor check_ptr_to_btf_access(). Make it strict in enforcing PTR_TRUSTED and PTR_UNTRUSTED while deprecating old PTR_TO_BTF_ID without modifier flags. There is a chance that this strict enforcement might break existing programs (especially on GCC compiled kernels), but this cleanup has to start sooner than later. Note PTR_TO_CTX access still yields old deprecated PTR_TO_BTF_ID. Once it's converted to strict PTR_TRUSTED or PTR_UNTRUSTED the kfuncs and helpers will be able to default to KF_TRUSTED_ARGS. KF_RCU will remain as a weaker version of KF_TRUSTED_ARGS where obj refcnt could be 0. Adjust rcu_read_lock selftest to run on gcc and clang compiled kernels. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: David Vernet <void@manifault.com> Link: https://lore.kernel.org/bpf/20230303041446.3630-7-alexei.starovoitov@gmail.com
2023-03-03 12:14:46 +08:00
ret = snprintf(safe_tname, sizeof(safe_tname), "%s%s", tname, suffix);
if (ret >= sizeof(safe_tname))
return false;
safe_id = btf_find_by_name_kind(btf, safe_tname, BTF_INFO_KIND(walk_type->info));
if (safe_id < 0)
return false;
safe_type = btf_type_by_id(btf, safe_id);
if (!safe_type)
return false;
for_each_member(i, safe_type, member) {
const char *m_name = __btf_name_by_offset(btf, member->name_off);
const struct btf_type *mtype = btf_type_by_id(btf, member->type);
u32 id;
if (!btf_type_is_ptr(mtype))
continue;
btf_type_skip_modifiers(btf, mtype->type, &id);
/* If we match on both type and name, the field is considered trusted. */
if (btf_id == id && !strcmp(field_name, m_name))
return true;
}
return false;
}
bpf: Allow trusted args to walk struct when checking BTF IDs When validating BTF types for KF_TRUSTED_ARGS kfuncs, the verifier currently enforces that the top-level type must match when calling the kfunc. In other words, the verifier does not allow the BPF program to pass a bitwise equivalent struct, despite it being allowed according to the C standard. For example, if you have the following type: struct nf_conn___init { struct nf_conn ct; }; The C standard stipulates that it would be safe to pass a struct nf_conn___init to a kfunc expecting a struct nf_conn. The verifier currently disallows this, however, as semantically kfuncs may want to enforce that structs that have equivalent types according to the C standard, but have different BTF IDs, are not able to be passed to kfuncs expecting one or the other. For example, struct nf_conn___init may not be queried / looked up, as it is allocated but may not yet be fully initialized. On the other hand, being able to pass types that are equivalent according to the C standard will be useful for other types of kfunc / kptrs enabled by BPF. For example, in a follow-on patch, a series of kfuncs will be added which allow programs to do bitwise queries on cpumasks that are either allocated by the program (in which case they'll be a 'struct bpf_cpumask' type that wraps a cpumask_t as its first element), or a cpumask that was allocated by the main kernel (in which case it will just be a straight cpumask_t, as in task->cpus_ptr). Having the two types of cpumasks allows us to distinguish between the two for when a cpumask is read-only vs. mutatable. A struct bpf_cpumask can be mutated by e.g. bpf_cpumask_clear(), whereas a regular cpumask_t cannot be. On the other hand, a struct bpf_cpumask can of course be queried in the exact same manner as a cpumask_t, with e.g. bpf_cpumask_test_cpu(). If we were to enforce that top level types match, then a user that's passing a struct bpf_cpumask to a read-only cpumask_t argument would have to cast with something like bpf_cast_to_kern_ctx() (which itself would need to be updated to expect the alias, and currently it only accommodates a single alias per prog type). Additionally, not specifying KF_TRUSTED_ARGS is not an option, as some kfuncs take one argument as a struct bpf_cpumask *, and another as a struct cpumask * (i.e. cpumask_t). In order to enable this, this patch relaxes the constraint that a KF_TRUSTED_ARGS kfunc must have strict type matching, and instead only enforces strict type matching if a type is observed to be a "no-cast alias" (i.e., that the type names are equivalent, but one is suffixed with ___init). Additionally, in order to try and be conservative and match existing behavior / expectations, this patch also enforces strict type checking for acquire kfuncs. We were already enforcing it for release kfuncs, so this should also improve the consistency of the semantics for kfuncs. Signed-off-by: David Vernet <void@manifault.com> Link: https://lore.kernel.org/r/20230120192523.3650503-3-void@manifault.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-01-21 03:25:16 +08:00
bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log,
const struct btf *reg_btf, u32 reg_id,
const struct btf *arg_btf, u32 arg_id)
{
const char *reg_name, *arg_name, *search_needle;
const struct btf_type *reg_type, *arg_type;
int reg_len, arg_len, cmp_len;
size_t pattern_len = sizeof(NOCAST_ALIAS_SUFFIX) - sizeof(char);
reg_type = btf_type_by_id(reg_btf, reg_id);
if (!reg_type)
return false;
arg_type = btf_type_by_id(arg_btf, arg_id);
if (!arg_type)
return false;
reg_name = btf_name_by_offset(reg_btf, reg_type->name_off);
arg_name = btf_name_by_offset(arg_btf, arg_type->name_off);
reg_len = strlen(reg_name);
arg_len = strlen(arg_name);
/* Exactly one of the two type names may be suffixed with ___init, so
* if the strings are the same size, they can't possibly be no-cast
* aliases of one another. If you have two of the same type names, e.g.
* they're both nf_conn___init, it would be improper to return true
* because they are _not_ no-cast aliases, they are the same type.
*/
if (reg_len == arg_len)
return false;
/* Either of the two names must be the other name, suffixed with ___init. */
if ((reg_len != arg_len + pattern_len) &&
(arg_len != reg_len + pattern_len))
return false;
if (reg_len < arg_len) {
search_needle = strstr(arg_name, NOCAST_ALIAS_SUFFIX);
cmp_len = reg_len;
} else {
search_needle = strstr(reg_name, NOCAST_ALIAS_SUFFIX);
cmp_len = arg_len;
}
if (!search_needle)
return false;
/* ___init suffix must come at the end of the name */
if (*(search_needle + pattern_len) != '\0')
return false;
return !strncmp(reg_name, arg_name, cmp_len);
}
#ifdef CONFIG_BPF_JIT
static int
btf_add_struct_ops(struct btf *btf, struct bpf_struct_ops *st_ops,
struct bpf_verifier_log *log)
{
struct btf_struct_ops_tab *tab, *new_tab;
int i, err;
tab = btf->struct_ops_tab;
if (!tab) {
tab = kzalloc(offsetof(struct btf_struct_ops_tab, ops[4]),
GFP_KERNEL);
if (!tab)
return -ENOMEM;
tab->capacity = 4;
btf->struct_ops_tab = tab;
}
for (i = 0; i < tab->cnt; i++)
if (tab->ops[i].st_ops == st_ops)
return -EEXIST;
if (tab->cnt == tab->capacity) {
new_tab = krealloc(tab,
offsetof(struct btf_struct_ops_tab,
ops[tab->capacity * 2]),
GFP_KERNEL);
if (!new_tab)
return -ENOMEM;
tab = new_tab;
tab->capacity *= 2;
btf->struct_ops_tab = tab;
}
tab->ops[btf->struct_ops_tab->cnt].st_ops = st_ops;
err = bpf_struct_ops_desc_init(&tab->ops[btf->struct_ops_tab->cnt], btf, log);
if (err)
return err;
btf->struct_ops_tab->cnt++;
return 0;
}
const struct bpf_struct_ops_desc *
bpf_struct_ops_find_value(struct btf *btf, u32 value_id)
{
const struct bpf_struct_ops_desc *st_ops_list;
unsigned int i;
u32 cnt;
if (!value_id)
return NULL;
if (!btf->struct_ops_tab)
return NULL;
cnt = btf->struct_ops_tab->cnt;
st_ops_list = btf->struct_ops_tab->ops;
for (i = 0; i < cnt; i++) {
if (st_ops_list[i].value_id == value_id)
return &st_ops_list[i];
}
return NULL;
}
const struct bpf_struct_ops_desc *
bpf_struct_ops_find(struct btf *btf, u32 type_id)
{
const struct bpf_struct_ops_desc *st_ops_list;
unsigned int i;
u32 cnt;
if (!type_id)
return NULL;
if (!btf->struct_ops_tab)
return NULL;
cnt = btf->struct_ops_tab->cnt;
st_ops_list = btf->struct_ops_tab->ops;
for (i = 0; i < cnt; i++) {
if (st_ops_list[i].type_id == type_id)
return &st_ops_list[i];
}
return NULL;
}
int __register_bpf_struct_ops(struct bpf_struct_ops *st_ops)
{
struct bpf_verifier_log *log;
struct btf *btf;
int err = 0;
btf = btf_get_module_btf(st_ops->owner);
if (!btf)
return check_btf_kconfigs(st_ops->owner, "struct_ops");
if (IS_ERR(btf))
return PTR_ERR(btf);
log = kzalloc(sizeof(*log), GFP_KERNEL | __GFP_NOWARN);
if (!log) {
err = -ENOMEM;
goto errout;
}
log->level = BPF_LOG_KERNEL;
err = btf_add_struct_ops(btf, st_ops, log);
errout:
kfree(log);
btf_put(btf);
return err;
}
EXPORT_SYMBOL_GPL(__register_bpf_struct_ops);
#endif
bool btf_param_match_suffix(const struct btf *btf,
const struct btf_param *arg,
const char *suffix)
{
int suffix_len = strlen(suffix), len;
const char *param_name;
/* In the future, this can be ported to use BTF tagging */
param_name = btf_name_by_offset(btf, arg->name_off);
if (str_is_empty(param_name))
return false;
len = strlen(param_name);
if (len <= suffix_len)
return false;
param_name += len - suffix_len;
return !strncmp(param_name, suffix, suffix_len);
}