Alexei Starovoitov says:
====================
pull-request: bpf 2021-03-20
The following pull-request contains BPF updates for your *net* tree.
We've added 5 non-merge commits during the last 3 day(s) which contain
a total of 8 files changed, 155 insertions(+), 12 deletions(-).
The main changes are:
1) Use correct nops in fexit trampoline, from Stanislav.
2) Fix BTF dump, from Jean-Philippe.
3) Fix umd memory leak, from Zqiang.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
__bpf_arch_text_poke does rewrite only for atomic nop5, emit_nops(xxx, 5)
emits non-atomic one which breaks fentry/fexit with k8 atomics:
P6_NOP5 == P6_NOP5_ATOMIC (0f1f440000 == 0f1f440000)
K8_NOP5 != K8_NOP5_ATOMIC (6666906690 != 6666666690)
Can be reproduced by doing "ideal_nops = k8_nops" in "arch_init_ideal_nops()
and running fexit_bpf2bpf selftest.
Fixes: e21aa34178 ("bpf: Fix fexit trampoline.")
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210320000001.915366-1-sdf@google.com
Jean-Philippe Brucker says:
====================
Fix an issue with the libbpf BTF dump, see patch 1 for details.
Since [v1] I added the selftest in patch 2, though I couldn't figure out
a way to make it independent from the order in which debug info is
issued by the compiler.
[v1]: https://lore.kernel.org/bpf/20210318122700.396574-1-jean-philippe@linaro.org/
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Bpftool used to issue forward declarations for a struct used as part of
a pointer to array, which is invalid. Add a test to check that the
struct is fully defined in this case:
@@ -134,9 +134,9 @@
};
};
-struct struct_in_array {};
+struct struct_in_array;
-struct struct_in_array_typed {};
+struct struct_in_array_typed;
typedef struct struct_in_array_typed struct_in_array_t[2];
@@ -189,3 +189,7 @@
struct struct_with_embedded_stuff _14;
};
+struct struct_in_array {};
+
+struct struct_in_array_typed {};
+
...
#13/1 btf_dump: syntax:FAIL
Suggested-by: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210319112554.794552-3-jean-philippe@linaro.org
The vmlinux.h generated from BTF is invalid when building
drivers/phy/ti/phy-gmii-sel.c with clang:
vmlinux.h:61702:27: error: array type has incomplete element type ‘struct reg_field’
61702 | const struct reg_field (*regfields)[3];
| ^~~~~~~~~
bpftool generates a forward declaration for this struct regfield, which
compilers aren't happy about. Here's a simplified reproducer:
struct inner {
int val;
};
struct outer {
struct inner (*ptr_to_array)[2];
} A;
After build with clang -> bpftool btf dump c -> clang/gcc:
./def-clang.h:11:23: error: array has incomplete element type 'struct inner'
struct inner (*ptr_to_array)[2];
Member ptr_to_array of struct outer is a pointer to an array of struct
inner. In the DWARF generated by clang, struct outer appears before
struct inner, so when converting BTF of struct outer into C, bpftool
issues a forward declaration to struct inner. With GCC the DWARF info is
reversed so struct inner gets fully defined.
That forward declaration is not sufficient when compilers handle an
array of the struct, even when it's only used through a pointer. Note
that we can trigger the same issue with an intermediate typedef:
struct inner {
int val;
};
typedef struct inner inner2_t[2];
struct outer {
inner2_t *ptr_to_array;
} A;
Becomes:
struct inner;
typedef struct inner inner2_t[2];
And causes:
./def-clang.h:10:30: error: array has incomplete element type 'struct inner'
typedef struct inner inner2_t[2];
To fix this, clear through_ptr whenever we encounter an intermediate
array, to make the inner struct part of a strong link and force full
declaration.
Fixes: 351131b51c ("libbpf: add btf_dump API for BTF-to-C conversion")
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210319112554.794552-2-jean-philippe@linaro.org
The ECN bit defines ECT(1) = 1, ECT(0) = 2. So inner 0x02 + outer 0x01
should be inner ECT(0) + outer ECT(1). Based on the description of
__INET_ECN_decapsulate, the final decapsulate value should be
ECT(1). So fix the test expect value to 0x01.
Before the fix:
TEST: VXLAN: ECN decap: 01/02->0x02 [FAIL]
Expected to capture 10 packets, got 0.
After the fix:
TEST: VXLAN: ECN decap: 01/02->0x01 [ OK ]
Fixes: a0b61f3d8e ("selftests: forwarding: vxlan_bridge_1d: Add an ECN decap test")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The mailing list for MPTCP maintenance has moved to the
kernel.org-supported mptcp@lists.linux.dev address.
Complete, combined archives for both lists are now hosted at
https://lore.kernel.org/mptcp
Cc: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Acked-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2021-03-19
This series contains updates to e1000e and igb drivers.
Tom Seewald fixes duplicate guard issues by including the driver name in
the guard for e1000e and igb.
Jesse adds checks that timestamping is on and valid to avoid possible
issues with a misinterpreted time stamp for igb.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
For AF_VSOCK, accept() currently returns sockets that are unlabelled.
Other socket families derive the child's SID from the SID of the parent
and the SID of the incoming packet. This is typically done as the
connected socket is placed in the queue that accept() removes from.
Reuse the existing 'security_sk_clone' hook to copy the SID from the
parent (server) socket to the child. There is no packet SID in this
case.
Fixes: d021c34405 ("VSOCK: Introduce VM Sockets")
Signed-off-by: David Brazdil <dbrazdil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
MTU cannot be changed on dwmac-sun8i. (ip link set eth0 mtu xxx returning EINVAL)
This is due to tx_fifo_size being 0, since this value is used to compute valid
MTU range.
Like dwmac-sunxi (with commit 806fd188ce ("net: stmmac: dwmac-sunxi: Provide TX and RX fifo sizes"))
dwmac-sun8i need to have tx and rx fifo sizes set.
I have used values from datasheets.
After this patch, setting a non-default MTU (like 1000) value works and network is still useable.
Tested-on: sun8i-h3-orangepi-pc
Tested-on: sun8i-r40-bananapi-m2-ultra
Tested-on: sun50i-a64-bananapi-m64
Tested-on: sun50i-h5-nanopi-neo-plus2
Tested-on: sun50i-h6-pine-h64
Fixes: 9f93ac8d40 ("net-next: stmmac: Add dwmac-sun8i")
Reported-by: Belisko Marek <marek.belisko@gmail.com>
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the USB host controller is EHCI, the throughput is reduced from
300Mb/s to 60Mb/s, when the rx buffer size is modified from 16K to
32K.
According to the EHCI spec, the maximum size of the qTD is 20K.
Therefore, when the driver uses more than 20K buffer, the latency
time of EHCI would be increased. And, it let the RTL8153A get worse
throughput.
However, the driver uses alloc_pages() for rx buffer, so I limit
the rx buffer to 16K rather than 20K.
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=205923
Fixes: ec5791c202 ("r8152: separate the rx buffer size")
Reported-by: Robert Davies <robdavies1977@gmail.com>
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The sk's sk_route_caps is set in sctp_packet_config, and later it
only needs to change when traversing the transport_list in a loop,
as the dst might be changed in the tx path.
So move sk_route_caps check and set into sctp_outq_flush_transports
from sctp_packet_transmit. This also fixes a dst leak reported by
Chen Yi:
https://bugzilla.kernel.org/show_bug.cgi?id=212227
As calling sk_setup_caps() in sctp_packet_transmit may also set the
sk_route_caps for the ctrl sock in a netns. When the netns is being
deleted, the ctrl sock's releasing is later than dst dev's deleting,
which will cause this dev's deleting to hang and dmesg error occurs:
unregister_netdevice: waiting for xxx to become free. Usage count = 1
Reported-by: Chen Yi <yiche@redhat.com>
Fixes: bcd623d8e9 ("sctp: call sk_setup_caps in sctp_packet_transmit instead")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a couple of checks to make sure timestamping is on and that the
timestamp value from DMA is valid. This avoids any functional issues
that could come from a misinterpreted time stamp.
One of the functions changed doesn't need a return value added because
there was no value in checking from the calling locations.
While here, fix a couple of reverse christmas tree issues next to
the code being changed.
Fixes: f56e7bba22 ("igb: Pull timestamp from fragment before adding it to skb")
Fixes: 9cbc948b5a ("igb: add XDP support")
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Dave Switzer <david.switzer@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
The include guard "_E1000_HW_H_" is used by two separate header files in
two different drivers (e1000/e1000_hw.h and igb/e1000_hw.h). Using the
same include guard macro in more than one header file may cause
unexpected behavior from the compiler. Fix this by renaming the
duplicate guard in the igb driver.
Fixes: 9d5c824399 ("igb: PCI-Express 82575 Gigabit Ethernet driver")
Signed-off-by: Tom Seewald <tseewald@gmail.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
The include guard "_E1000_HW_H_" is used by header files in three
different drivers (e1000/e1000_hw.h, e1000e/hw.h, and igb/e1000_hw.h).
Using the same include guard macro in more than one header file may
cause unexpected behavior from the compiler. Fix the duplicate include
guard in the e1000e driver by renaming it.
Fixes: bc7f75fa97 ("[E1000E]: New pci-express e1000 driver (currently for ICH9 devices only)")
Signed-off-by: Tom Seewald <tseewald@gmail.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Acked-by: Sasha Neftin <sasha.neftin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Set the disconnected flag before releasing the data interface in case
netdev registration fails to avoid having the disconnect callback try to
deregister the never registered netdev (and trigger a WARN_ON()).
Fixes: 87cf65601e ("USB host CDC Phonet network interface driver")
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
1) Several patches to testore use of memory barriers instead of RCU to
ensure consistent access to ruleset, from Mark Tomlinson.
2) Fix dump of expectation via ctnetlink, from Florian Westphal.
3) GRE helper works for IPv6, from Ludovic Senecaux.
4) Set error on unsupported flowtable flags.
5) Use delayed instead of deferrable workqueue in the flowtable,
from Yinjun Zhang.
6) Fix spurious EEXIST in case of add-after-delete flowtable in
the same batch.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
__dev_alloc_name(), when supplied with a name containing '%d',
will search for the first available device number to generate a
unique device name.
Since commit ff92741270 ("net:
introduce name_node struct to be used in hashlist") network
devices may have alternate names. __dev_alloc_name() does take
these alternate names into account, possibly generating a name
that is already taken and failing with -ENFILE as a result.
This demonstrates the bug:
# rmmod dummy 2>/dev/null
# ip link property add dev lo altname dummy0
# modprobe dummy numdummies=1
modprobe: ERROR: could not insert 'dummy': Too many open files in system
Instead of creating a device named dummy1, modprobe fails.
Fix this by checking all the names in the d->name_node list, not just d->name.
Signed-off-by: Jiri Bohac <jbohac@suse.cz>
Fixes: ff92741270 ("net: introduce name_node struct to be used in hashlist")
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 92c8c16f34 ("powerpc/embedded6xx: Remove C2K board support")
removed last selector of CONFIG_MV64X60.
As it is not a user selectable config item, all references to it
are stale. Remove them.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hariprasad Kelam says:
====================
octeontx2: miscellaneous fixes
This series of patches fixes various issues related to NPC MCAM entry
management, debugfs, devlink, CGX LMAC mapping, RSS config etc
Change-log:
v2:
Fixed below review comments
- corrected Fixed tag syntax with 12 digits SHA1
and providing space between SHA1 and subject line
- remove code improvement patch
- make commit description more clear
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
unmapping npc counter works in a way by traversing all mcam
entries to find which mcam rule is associated with counter.
But loop cursor variable 'entry' is not incremented before
checking next mcam entry which resulting in infinite loop.
This in turn hogs the kworker thread forever and no other
mbox message is processed by AF driver after that.
Fix this by updating entry value before checking next
mcam entry.
Fixes: a958dd59f9 ("octeontx2-af: Map or unmap NPC MCAM entry and counter")
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RSS configuration can not be get/set when interface is in down state
as they required mbox communication. RSS enable flag status
is used for set/get configuration. Current code do not clear the
RSS enable flag on interface down which lead to mbox error while
trying to set/get RSS configuration.
Fixes: 85069e95e5 ("octeontx2-pf: Receive side scaling support")
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current devlink code try to free already freed irqs as the
irq_allocate flag is not cleared after free leading to kernel
crash while removing rvu driver. The patch fixes the irq free
sequence and clears the irq_allocate flag on free.
Fixes: 7304ac4567 ("octeontx2-af: Add mailbox IRQ and msg handlers")
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CGX receive buffer size is a constant value and
cannot be read from CGX0 block always since
CGX0 may not enabled everytime. Hence return CGX
receive buffer size from first enabled CGX block
instead of CGX0.
Fixes: 6e54e1c539 ("octeontx2-af: cn10K: MTU configuration")
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The MKEX profile describes what packet fields need to be extracted from
the input packet and how to place those packet fields in the output key
for MCAM matching. The MKEX profile can be in a way where higher layer
packet fields can overwrite lower layer packet fields in output MCAM
Key.
Hence MKEX profile is always ensured that there are no overlaps between
any of the layers. But the commit 42006910b5
("octeontx2-af: cleanup KPU config data") introduced TX TOS field which
overlaps with DMAC in MCAM key.
This led to AF driver returning error when TX rule is installed with
DMAC as match criteria since DMAC gets overwritten and cannot be
supported. This patch fixes the issue by removing TOS field from MKEX TX
profile.
Fixes: 42006910b5 ("octeontx2-af: cleanup KPU config data")
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the ETHTOOL_GRXCLSRLALL ioctl ethtool uses
below structure to read number of rules from the driver.
struct ethtool_rxnfc {
__u32 cmd;
__u32 flow_type;
__u64 data;
struct ethtool_rx_flow_spec fs;
union {
__u32 rule_cnt;
__u32 rss_context;
};
__u32 rule_locs[0];
};
Driver must not modify rule_cnt member. But currently driver
modifies it by modifying rss_context. Hence fix it by using a
local variable.
Fixes: 81a4362016 ("octeontx2-pf: Add RSS multi group support")
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 8e850f25b5 ("net: socionext: Stop PHY before resetting
netsec") netsec_netdev_init() power downs phy before resetting the
controller. However, the state is not restored once the reset is
complete. As a result it is not possible to bring up network on a
platform with Broadcom BCM5482 phy.
Fix the issue by restoring phy power state after controller reset is
complete.
Fixes: 8e850f25b5 ("net: socionext: Stop PHY before resetting netsec")
Cc: stable@vger.kernel.org
Signed-off-by: Mian Yousaf Kaukab <ykaukab@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 6af1799aaf.
Commit 6af1799aaf ("ipv6: drop incoming packets having a v4mapped
source address") introduced an input check against v4mapped addresses.
Use of such addresses on the wire is indeed questionable and not
allowed on public Internet. As the commit pointed out
https://tools.ietf.org/html/draft-itojun-v6ops-v4mapped-harmful-02
lists potential issues.
Unfortunately there are applications which use v4mapped addresses,
and breaking them is a clear regression. For example v4mapped
addresses (or any semi-valid addresses, really) may be used
for uni-direction event streams or packet export.
Since the issue which sparked the addition of the check was with
TCP and request_socks in particular push the check down to TCPv6
and DCCP. This restores the ability to receive UDPv6 packets with
v4mapped address as the source.
Keep using the IPSTATS_MIB_INHDRERRORS statistic to minimize the
user-visible changes.
Fixes: 6af1799aaf ("ipv6: drop incoming packets having a v4mapped source address")
Reported-by: Sunyi Shao <sunyishao@fb.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a selftest for commit e21aa34178 ("bpf: Fix fexit trampoline.")
to make sure that attaching fexit prog to a sleeping kernel function
will trigger appropriate trampoline and program destruction.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20210318004523.55908-1-alexei.starovoitov@gmail.com
Daniel Borkmann says:
====================
pull-request: bpf 2021-03-18
The following pull-request contains BPF updates for your *net* tree.
We've added 10 non-merge commits during the last 4 day(s) which contain
a total of 14 files changed, 336 insertions(+), 94 deletions(-).
The main changes are:
1) Fix fexit/fmod_ret trampoline for sleepable programs, and also fix a ftrace
splat in modify_ftrace_direct() on address change, from Alexei Starovoitov.
2) Fix two oob speculation possibilities that allows unprivileged to leak mem
via side-channel, from Piotr Krysiuk and Daniel Borkmann.
3) Fix libbpf's netlink handling wrt SOCK_CLOEXEC, from Kumar Kartikeya Dwivedi.
4) Fix libbpf's error handling on failure in getting section names, from Namhyung Kim.
5) Fix tunnel collect_md BPF selftest wrt Geneve option handling, from Hangbin Liu.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
If the flowtable has been previously removed in this batch, skip the
hook overlap checks. This fixes spurious EEXIST errors when removing and
adding the flowtable in the same batch.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Otherwise, there exists a small window between the opening and closing
of the socket fd where it may leak into processes launched by some other
thread.
Fixes: 949abbe884 ("libbpf: add function to setup XDP")
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/bpf/20210317115857.6536-1-memxor@gmail.com
When it failed to get section names, it should call into
bpf_object__elf_finish() like others.
Fixes: 88a8212028 ("libbpf: Factor out common ELF operations and improve logging")
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20210317145414.884817-1-namhyung@kernel.org
Currently flowtable's GC work is initialized as deferrable, which
means GC cannot work on time when system is idle. So the hardware
offloaded flow may be deleted for timeout, since its used time is
not timely updated.
Resolve it by initializing the GC work as delayed work instead of
deferrable.
Fixes: c29f74e0df ("netfilter: nf_flow_table: hardware offload support")
Signed-off-by: Yinjun Zhang <yinjun.zhang@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Honor flowtable flags from the control update path. Disallow disabling
to toggle hardware offload support though.
Fixes: 8bb69f3b29 ("netfilter: nf_tables: add flowtable offload control plane")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Error was not set accordingly.
Fixes: 8bb69f3b29 ("netfilter: nf_tables: add flowtable offload control plane")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
This fix permits gre connections to be tracked within ip6tables rules
Signed-off-by: Ludovic Senecaux <linuxludo@free.fr>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The fexit/fmod_ret programs can be attached to kernel functions that can sleep.
The synchronize_rcu_tasks() will not wait for such tasks to complete.
In such case the trampoline image will be freed and when the task
wakes up the return IP will point to freed memory causing the crash.
Solve this by adding percpu_ref_get/put for the duration of trampoline
and separate trampoline vs its image life times.
The "half page" optimization has to be removed, since
first_half->second_half->first_half transition cannot be guaranteed to
complete in deterministic time. Every trampoline update becomes a new image.
The image with fmod_ret or fexit progs will be freed via percpu_ref_kill and
call_rcu_tasks. Together they will wait for the original function and
trampoline asm to complete. The trampoline is patched from nop to jmp to skip
fexit progs. They are freed independently from the trampoline. The image with
fentry progs only will be freed via call_rcu_tasks_trace+call_rcu_tasks which
will wait for both sleepable and non-sleepable progs to complete.
Fixes: fec56f5890 ("bpf: Introduce BPF trampoline")
Reported-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Paul E. McKenney <paulmck@kernel.org> # for RCU
Link: https://lore.kernel.org/bpf/20210316210007.38949-1-alexei.starovoitov@gmail.com
Currently, napi_thread_wait() checks for NAPI_STATE_SCHED bit to
determine if the kthread owns this napi and could call napi->poll() on
it. However, if socket busy poll is enabled, it is possible that the
busy poll thread grabs this SCHED bit (after the previous napi->poll()
invokes napi_complete_done() and clears SCHED bit) and tries to poll
on the same napi. napi_disable() could grab the SCHED bit as well.
This patch tries to fix this race by adding a new bit
NAPI_STATE_SCHED_THREADED in napi->state. This bit gets set in
____napi_schedule() if the threaded mode is enabled, and gets cleared
in napi_complete_done(), and we only poll the napi in kthread if this
bit is set. This helps distinguish the ownership of the napi between
kthread and other scenarios and fixes the race issue.
Fixes: 29863d41bb ("net: implement threaded-able napi poll loop support")
Reported-by: Martin Zaharinov <micron10@gmail.com>
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Wei Wang <weiwan@google.com>
Cc: Alexander Duyck <alexanderduyck@fb.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix up test_verifier error messages for the case where the original error
message changed, or for the case where pointer alu errors differ between
privileged and unprivileged tests. Also, add alternative tests for keeping
coverage of the original verifier rejection error message (fp alu), and
newly reject map_ptr += rX where rX == 0 given we now forbid alu on these
types for unprivileged. All test_verifier cases pass after the change. The
test case fixups were kept separate to ease backporting of core changes.
Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Given we know the max possible value of ptr_limit at the time of retrieving
the latter, add basic assertions, so that the verifier can bail out if
anything looks odd and reject the program. Nothing triggered this so far,
but it also does not hurt to have these.
Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
* HE (802.11ax) elements can be extended, handle that
* fix locking in network namespace changes that was
broken due to the RTNL-redux work
* various other small fixes
-----BEGIN PGP SIGNATURE-----
iQIzBAABCgAdFiEEH1e1rEeCd0AIMq6MB8qZga/fl8QFAmBR1WoACgkQB8qZga/f
l8QvKQ/+I4a4gmno6A05asm6PY4sf8lQ6j90SGS0u82nJlFCMGlPNpDcTYDwozy1
bjwTBGn5Jc808UUriPKt0aY8sukgFuJmBHZsxDFOfx48t/hMDnbkf4u3pIvtqhsf
OfcqsmS4Zp3r1r7xfKhResAg5t4phaRz7WGW+RYyAF8vVwSR6hDR/TmUI7sft3l4
0FlQEpdW/dm2cvy2tbUJtSyHrKEbrlVn55DDE3fWpfwaOQzHSFZPKbishmk4mhIR
zJgdehaogkUxa5KbuImWIRHFudJ+Seths6+c8FPERQGjK7QGGyB/qrC1+mvw3wX+
pAf9+R54aBqcA8mEEryqprmEwo0i+YPo1kuXig6fE4D62d2dD9nZSghmgcqMM1Vp
6FwpU4g7Fn5mpyw3yAxdWbYzP/Ai4fb9QVaFsQvIv4PnpRB0wt+/1Aftd68HDNYk
kWLRRsoCVrddr3yImbYTA0ytIwgi6wAUqoH9+JQSGkJbOh18G1gAwxsdz0BWZoYK
9GI8LST0Pr+mnwjo0E9ILHbIVKyGCcx7Porxbr2yEeIB3JBpBtZA1qTIX6MHcote
BIx8LyEWrnEVCyWq1ZWcEdLdzzTRVgzYmFprdY+Zq/Lk1L9Gr7M8m4ri1VzMX2rF
9IwfKRjZPLD5v3A5ls9PrkUrns+6IjIpIYhwoqFuVyv24QHyyoo=
=ePzY
-----END PGP SIGNATURE-----
Merge tag 'mac80211-for-net-2021-03-17' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211
Johannes Berg says:
====================
First round of fixes for 5.12-rc:
* HE (802.11ax) elements can be extended, handle that
* fix locking in network namespace changes that was
broken due to the RTNL-redux work
* various other small fixes
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The ct_state validate should not only check the mask bit and also
check mask_bit & key_bit..
For the +new+est case example, The 'new' and 'est' bits should be
set in both state_mask and state flags. Or the -new-est case also
will be reject by kernel.
When Openvswitch with two flows
ct_state=+trk+new,action=commit,forward
ct_state=+trk+est,action=forward
A packet go through the kernel and the contrack state is invalid,
The ct_state will be +trk-inv. Upcall to the ovs-vswitchd, the
finally dp action will be drop with -new-est+trk.
Fixes: 1bcc51ac07 ("net/sched: cls_flower: Reject invalid ct_state flags rules")
Fixes: 3aed8b6333 ("net/sched: cls_flower: validate ct_state for invalid and reply flags")
Signed-off-by: wenxu <wenxu@ucloud.cn>
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We were linearizing non-TSO skbs that had too many frags, but
we weren't checking number of frags on TSO skbs. This could
lead to a bad page reference when we received a TSO skb with
more frags than the Tx descriptor could support.
v2: use gso_segs rather than yet another division
don't rework the check on the nr_frags
Fixes: 0f3154e6bc ("ionic: Add Tx and Rx handling")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of having the mov32 with aux->alu_limit - 1 immediate, move this
operation to retrieve_ptr_limit() instead to simplify the logic and to
allow for subsequent sanity boundary checks inside retrieve_ptr_limit().
This avoids in future that at the time of the verifier masking rewrite
we'd run into an underflow which would not sign extend due to the nature
of mov32 instruction.
Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
retrieve_ptr_limit() computes the ptr_limit for registers with stack and
map_value type. ptr_limit is the size of the memory area that is still
valid / in-bounds from the point of the current position and direction
of the operation (add / sub). This size will later be used for masking
the operation such that attempting out-of-bounds access in the speculative
domain is redirected to remain within the bounds of the current map value.
When masking to the right the size is correct, however, when masking to
the left, the size is off-by-one which would lead to an incorrect mask
and thus incorrect arithmetic operation in the non-speculative domain.
Piotr found that if the resulting alu_limit value is zero, then the
BPF_MOV32_IMM() from the fixup_bpf_calls() rewrite will end up loading
0xffffffff into AX instead of sign-extending to the full 64 bit range,
and as a result, this allows abuse for executing speculatively out-of-
bounds loads against 4GB window of address space and thus extracting the
contents of kernel memory via side-channel.
Fixes: 979d63d50c ("bpf: prevent out of bounds speculation on pointer arithmetic")
Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
