linux/arch/x86/kernel/ftrace_64.S
Linus Torvalds 95f1fa9e34 New tracing features:
- PERAMAENT flag to ftrace_ops when attaching a callback to a function
    As /proc/sys/kernel/ftrace_enabled when set to zero will disable all
    attached callbacks in ftrace, this has a detrimental impact on live
    kernel tracing, as it disables all that it patched. If a ftrace_ops
    is registered to ftrace with the PERMANENT flag set, it will prevent
    ftrace_enabled from being disabled, and if ftrace_enabled is already
    disabled, it will prevent a ftrace_ops with PREMANENT flag set from
    being registered.
 
  - New register_ftrace_direct(). As eBPF would like to register its own
    trampolines to be called by the ftrace nop locations directly,
    without going through the ftrace trampoline, this function has been
    added. This allows for eBPF trampolines to live along side of
    ftrace, perf, kprobe and live patching. It also utilizes the ftrace
    enabled_functions file that keeps track of functions that have been
    modified in the kernel, to allow for security auditing.
 
  - Allow for kernel internal use of ftrace instances. Subsystems in
    the kernel can now create and destroy their own tracing instances
    which allows them to have their own tracing buffer, and be able
    to record events without worrying about other users from writing over
    their data.
 
  - New seq_buf_hex_dump() that lets users use the hex_dump() in their
    seq_buf usage.
 
  - Notifications now added to tracing_max_latency to allow user space
    to know when a new max latency is hit by one of the latency tracers.
 
  - Wider spread use of generic compare operations for use of bsearch and
    friends.
 
  - More synthetic event fields may be defined (32 up from 16)
 
  - Use of xarray for architectures with sparse system calls, for the
    system call trace events.
 
 This along with small clean ups and fixes.
 -----BEGIN PGP SIGNATURE-----
 
 iIoEABYIADIWIQRRSw7ePDh/lE+zeZMp5XQQmuv6qgUCXdwv4BQccm9zdGVkdEBn
 b29kbWlzLm9yZwAKCRAp5XQQmuv6qnB5AP91vsdHQjwE1+/UWG/cO+qFtKvn2QJK
 QmBRIJNH/s+1TAD/fAOhgw+ojSK3o/qc+NpvPTEW9AEwcJL1wacJUn+XbQc=
 =ztql
 -----END PGP SIGNATURE-----

Merge tag 'trace-v5.5' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace

Pull tracing updates from Steven Rostedt:
 "New tracing features:

   - New PERMANENT flag to ftrace_ops when attaching a callback to a
     function.

     As /proc/sys/kernel/ftrace_enabled when set to zero will disable
     all attached callbacks in ftrace, this has a detrimental impact on
     live kernel tracing, as it disables all that it patched. If a
     ftrace_ops is registered to ftrace with the PERMANENT flag set, it
     will prevent ftrace_enabled from being disabled, and if
     ftrace_enabled is already disabled, it will prevent a ftrace_ops
     with PREMANENT flag set from being registered.

   - New register_ftrace_direct().

     As eBPF would like to register its own trampolines to be called by
     the ftrace nop locations directly, without going through the ftrace
     trampoline, this function has been added. This allows for eBPF
     trampolines to live along side of ftrace, perf, kprobe and live
     patching. It also utilizes the ftrace enabled_functions file that
     keeps track of functions that have been modified in the kernel, to
     allow for security auditing.

   - Allow for kernel internal use of ftrace instances.

     Subsystems in the kernel can now create and destroy their own
     tracing instances which allows them to have their own tracing
     buffer, and be able to record events without worrying about other
     users from writing over their data.

   - New seq_buf_hex_dump() that lets users use the hex_dump() in their
     seq_buf usage.

   - Notifications now added to tracing_max_latency to allow user space
     to know when a new max latency is hit by one of the latency
     tracers.

   - Wider spread use of generic compare operations for use of bsearch
     and friends.

   - More synthetic event fields may be defined (32 up from 16)

   - Use of xarray for architectures with sparse system calls, for the
     system call trace events.

  This along with small clean ups and fixes"

* tag 'trace-v5.5' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace: (51 commits)
  tracing: Enable syscall optimization for MIPS
  tracing: Use xarray for syscall trace events
  tracing: Sample module to demonstrate kernel access to Ftrace instances.
  tracing: Adding new functions for kernel access to Ftrace instances
  tracing: Fix Kconfig indentation
  ring-buffer: Fix typos in function ring_buffer_producer
  ftrace: Use BIT() macro
  ftrace: Return ENOTSUPP when DYNAMIC_FTRACE_WITH_DIRECT_CALLS is not configured
  ftrace: Rename ftrace_graph_stub to ftrace_stub_graph
  ftrace: Add a helper function to modify_ftrace_direct() to allow arch optimization
  ftrace: Add helper find_direct_entry() to consolidate code
  ftrace: Add another check for match in register_ftrace_direct()
  ftrace: Fix accounting bug with direct->count in register_ftrace_direct()
  ftrace/selftests: Fix spelling mistake "wakeing" -> "waking"
  tracing: Increase SYNTH_FIELDS_MAX for synthetic_events
  ftrace/samples: Add a sample module that implements modify_ftrace_direct()
  ftrace: Add modify_ftrace_direct()
  tracing: Add missing "inline" in stub function of latency_fsnotify()
  tracing: Remove stray tab in TRACE_EVAL_MAP_FILE's help text
  tracing: Use seq_buf_hex_dump() to dump buffers
  ...
2019-11-27 11:42:01 -08:00

346 lines
8.3 KiB
ArmAsm

/* SPDX-License-Identifier: GPL-2.0 */
/*
* Copyright (C) 2014 Steven Rostedt, Red Hat Inc
*/
#include <linux/linkage.h>
#include <asm/ptrace.h>
#include <asm/ftrace.h>
#include <asm/export.h>
#include <asm/nospec-branch.h>
#include <asm/unwind_hints.h>
#include <asm/frame.h>
.code64
.section .entry.text, "ax"
#ifdef CONFIG_FRAME_POINTER
/* Save parent and function stack frames (rip and rbp) */
# define MCOUNT_FRAME_SIZE (8+16*2)
#else
/* No need to save a stack frame */
# define MCOUNT_FRAME_SIZE 0
#endif /* CONFIG_FRAME_POINTER */
/* Size of stack used to save mcount regs in save_mcount_regs */
#define MCOUNT_REG_SIZE (SS+8 + MCOUNT_FRAME_SIZE)
/*
* gcc -pg option adds a call to 'mcount' in most functions.
* When -mfentry is used, the call is to 'fentry' and not 'mcount'
* and is done before the function's stack frame is set up.
* They both require a set of regs to be saved before calling
* any C code and restored before returning back to the function.
*
* On boot up, all these calls are converted into nops. When tracing
* is enabled, the call can jump to either ftrace_caller or
* ftrace_regs_caller. Callbacks (tracing functions) that require
* ftrace_regs_caller (like kprobes) need to have pt_regs passed to
* it. For this reason, the size of the pt_regs structure will be
* allocated on the stack and the required mcount registers will
* be saved in the locations that pt_regs has them in.
*/
/*
* @added: the amount of stack added before calling this
*
* After this is called, the following registers contain:
*
* %rdi - holds the address that called the trampoline
* %rsi - holds the parent function (traced function's return address)
* %rdx - holds the original %rbp
*/
.macro save_mcount_regs added=0
#ifdef CONFIG_FRAME_POINTER
/* Save the original rbp */
pushq %rbp
/*
* Stack traces will stop at the ftrace trampoline if the frame pointer
* is not set up properly. If fentry is used, we need to save a frame
* pointer for the parent as well as the function traced, because the
* fentry is called before the stack frame is set up, where as mcount
* is called afterward.
*/
/* Save the parent pointer (skip orig rbp and our return address) */
pushq \added+8*2(%rsp)
pushq %rbp
movq %rsp, %rbp
/* Save the return address (now skip orig rbp, rbp and parent) */
pushq \added+8*3(%rsp)
pushq %rbp
movq %rsp, %rbp
#endif /* CONFIG_FRAME_POINTER */
/*
* We add enough stack to save all regs.
*/
subq $(MCOUNT_REG_SIZE - MCOUNT_FRAME_SIZE), %rsp
movq %rax, RAX(%rsp)
movq %rcx, RCX(%rsp)
movq %rdx, RDX(%rsp)
movq %rsi, RSI(%rsp)
movq %rdi, RDI(%rsp)
movq %r8, R8(%rsp)
movq %r9, R9(%rsp)
movq $0, ORIG_RAX(%rsp)
/*
* Save the original RBP. Even though the mcount ABI does not
* require this, it helps out callers.
*/
#ifdef CONFIG_FRAME_POINTER
movq MCOUNT_REG_SIZE-8(%rsp), %rdx
#else
movq %rbp, %rdx
#endif
movq %rdx, RBP(%rsp)
/* Copy the parent address into %rsi (second parameter) */
movq MCOUNT_REG_SIZE+8+\added(%rsp), %rsi
/* Move RIP to its proper location */
movq MCOUNT_REG_SIZE+\added(%rsp), %rdi
movq %rdi, RIP(%rsp)
/*
* Now %rdi (the first parameter) has the return address of
* where ftrace_call returns. But the callbacks expect the
* address of the call itself.
*/
subq $MCOUNT_INSN_SIZE, %rdi
.endm
.macro restore_mcount_regs save=0
/* ftrace_regs_caller or frame pointers require this */
movq RBP(%rsp), %rbp
movq R9(%rsp), %r9
movq R8(%rsp), %r8
movq RDI(%rsp), %rdi
movq RSI(%rsp), %rsi
movq RDX(%rsp), %rdx
movq RCX(%rsp), %rcx
movq RAX(%rsp), %rax
addq $MCOUNT_REG_SIZE-\save, %rsp
.endm
#ifdef CONFIG_DYNAMIC_FTRACE
SYM_FUNC_START(__fentry__)
retq
SYM_FUNC_END(__fentry__)
EXPORT_SYMBOL(__fentry__)
SYM_FUNC_START(ftrace_caller)
/* save_mcount_regs fills in first two parameters */
save_mcount_regs
SYM_INNER_LABEL(ftrace_caller_op_ptr, SYM_L_GLOBAL)
/* Load the ftrace_ops into the 3rd parameter */
movq function_trace_op(%rip), %rdx
/* regs go into 4th parameter (but make it NULL) */
movq $0, %rcx
SYM_INNER_LABEL(ftrace_call, SYM_L_GLOBAL)
call ftrace_stub
restore_mcount_regs
/*
* The code up to this label is copied into trampolines so
* think twice before adding any new code or changing the
* layout here.
*/
SYM_INNER_LABEL(ftrace_epilogue, SYM_L_GLOBAL)
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
SYM_INNER_LABEL(ftrace_graph_call, SYM_L_GLOBAL)
jmp ftrace_stub
#endif
/*
* This is weak to keep gas from relaxing the jumps.
* It is also used to copy the retq for trampolines.
*/
SYM_INNER_LABEL_ALIGN(ftrace_stub, SYM_L_WEAK)
retq
SYM_FUNC_END(ftrace_caller)
SYM_FUNC_START(ftrace_regs_caller)
/* Save the current flags before any operations that can change them */
pushfq
UNWIND_HINT_SAVE
/* added 8 bytes to save flags */
save_mcount_regs 8
/* save_mcount_regs fills in first two parameters */
SYM_INNER_LABEL(ftrace_regs_caller_op_ptr, SYM_L_GLOBAL)
/* Load the ftrace_ops into the 3rd parameter */
movq function_trace_op(%rip), %rdx
/* Save the rest of pt_regs */
movq %r15, R15(%rsp)
movq %r14, R14(%rsp)
movq %r13, R13(%rsp)
movq %r12, R12(%rsp)
movq %r11, R11(%rsp)
movq %r10, R10(%rsp)
movq %rbx, RBX(%rsp)
/* Copy saved flags */
movq MCOUNT_REG_SIZE(%rsp), %rcx
movq %rcx, EFLAGS(%rsp)
/* Kernel segments */
movq $__KERNEL_DS, %rcx
movq %rcx, SS(%rsp)
movq $__KERNEL_CS, %rcx
movq %rcx, CS(%rsp)
/* Stack - skipping return address and flags */
leaq MCOUNT_REG_SIZE+8*2(%rsp), %rcx
movq %rcx, RSP(%rsp)
ENCODE_FRAME_POINTER
/* regs go into 4th parameter */
leaq (%rsp), %rcx
SYM_INNER_LABEL(ftrace_regs_call, SYM_L_GLOBAL)
call ftrace_stub
/* Copy flags back to SS, to restore them */
movq EFLAGS(%rsp), %rax
movq %rax, MCOUNT_REG_SIZE(%rsp)
/* Handlers can change the RIP */
movq RIP(%rsp), %rax
movq %rax, MCOUNT_REG_SIZE+8(%rsp)
/* restore the rest of pt_regs */
movq R15(%rsp), %r15
movq R14(%rsp), %r14
movq R13(%rsp), %r13
movq R12(%rsp), %r12
movq R10(%rsp), %r10
movq RBX(%rsp), %rbx
movq ORIG_RAX(%rsp), %rax
movq %rax, MCOUNT_REG_SIZE-8(%rsp)
/* If ORIG_RAX is anything but zero, make this a call to that */
movq ORIG_RAX(%rsp), %rax
cmpq $0, %rax
je 1f
/* Swap the flags with orig_rax */
movq MCOUNT_REG_SIZE(%rsp), %rdi
movq %rdi, MCOUNT_REG_SIZE-8(%rsp)
movq %rax, MCOUNT_REG_SIZE(%rsp)
restore_mcount_regs 8
jmp 2f
1: restore_mcount_regs
2:
/*
* The stack layout is nondetermistic here, depending on which path was
* taken. This confuses objtool and ORC, rightfully so. For now,
* pretend the stack always looks like the non-direct case.
*/
UNWIND_HINT_RESTORE
/* Restore flags */
popfq
/*
* As this jmp to ftrace_epilogue can be a short jump
* it must not be copied into the trampoline.
* The trampoline will add the code to jump
* to the return.
*/
SYM_INNER_LABEL(ftrace_regs_caller_end, SYM_L_GLOBAL)
jmp ftrace_epilogue
SYM_FUNC_END(ftrace_regs_caller)
#else /* ! CONFIG_DYNAMIC_FTRACE */
SYM_FUNC_START(__fentry__)
cmpq $ftrace_stub, ftrace_trace_function
jnz trace
fgraph_trace:
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
cmpq $ftrace_stub, ftrace_graph_return
jnz ftrace_graph_caller
cmpq $ftrace_graph_entry_stub, ftrace_graph_entry
jnz ftrace_graph_caller
#endif
SYM_INNER_LABEL(ftrace_stub, SYM_L_GLOBAL)
retq
trace:
/* save_mcount_regs fills in first two parameters */
save_mcount_regs
/*
* When DYNAMIC_FTRACE is not defined, ARCH_SUPPORTS_FTRACE_OPS is not
* set (see include/asm/ftrace.h and include/linux/ftrace.h). Only the
* ip and parent ip are used and the list function is called when
* function tracing is enabled.
*/
movq ftrace_trace_function, %r8
CALL_NOSPEC %r8
restore_mcount_regs
jmp fgraph_trace
SYM_FUNC_END(__fentry__)
EXPORT_SYMBOL(__fentry__)
#endif /* CONFIG_DYNAMIC_FTRACE */
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
SYM_FUNC_START(ftrace_graph_caller)
/* Saves rbp into %rdx and fills first parameter */
save_mcount_regs
leaq MCOUNT_REG_SIZE+8(%rsp), %rsi
movq $0, %rdx /* No framepointers needed */
call prepare_ftrace_return
restore_mcount_regs
retq
SYM_FUNC_END(ftrace_graph_caller)
SYM_CODE_START(return_to_handler)
UNWIND_HINT_EMPTY
subq $24, %rsp
/* Save the return values */
movq %rax, (%rsp)
movq %rdx, 8(%rsp)
movq %rbp, %rdi
call ftrace_return_to_handler
movq %rax, %rdi
movq 8(%rsp), %rdx
movq (%rsp), %rax
addq $24, %rsp
JMP_NOSPEC %rdi
SYM_CODE_END(return_to_handler)
#endif