When the device is in internal error state, command interface isn't
accessible and the driver decides which commands to fail and which
to ignore.
Move the MODIFY_RQT command to the ignore list in order to avoid
the following redundant warning messages in internal error state:
mlx5_core 0000:82:00.1: mlx5e_rss_disable:419:(pid 23754): Failed to redirect RQT 0x0 to drop RQ 0xc00848: err = -5
mlx5_core 0000:82:00.1: mlx5e_rx_res_channels_deactivate:598:(pid 23754): Failed to redirect direct RQT 0x1 to drop RQ 0xc00848 (channel 0): err = -5
mlx5_core 0000:82:00.1: mlx5e_rx_res_channels_deactivate:607:(pid 23754): Failed to redirect XSK RQT 0x19 to drop RQ 0xc00848 (channel 0): err = -5
Fixes: 43ec0f41fa ("net/mlx5e: Hide all implementation details of mlx5e_rx_res")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Transport Interface Receive (TIR) objects perform the packet processing and
reassembly and is also responsible for demultiplexing the packets into the
different RQs.
There are certain TIR context attributes that propagate to the pointed RQs
and applied to them (like packet_merge offloads (LRO/SHAMPO) and
tunneled_offload_en). When TIRs do not agree on attributes values, a "last
one wins" policy is applied. Hence, if not synced properly, a race between
TIR params update and a concurrent TIR create/modify operation might yield
to a mismatch between the shadow parameters in SW and the actual applied
state of the RQs in HW.
tunneled_offload_en is a fixed attribute per profile, while packet merge
offload state might be toggled and get out-of-sync. When this happens,
packet_merge offload might be working although not requested, or the
opposite.
All updates to packet_merge state and all create/modify operations of
regular redirection/steering TIRs are done under the same priv->state_lock,
so they do not run in parallel, and no race is possible.
However, there are other kind of TIRs (acceleration offloads TIRs, like TLS
TIRs) which are created on demand for each new connection without holding
the coarse priv->state_lock, hence might race.
Fix this by synchronizing all packet_merge state reads and writes against
all TIR create/modify operations. Include the modify operations of the
regular redirection steering TIRs under the new lock, for better code
layering and division of responsibilities.
Fixes: 1182f36593 ("net/mlx5e: kTLS, Add kTLS RX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The cited patch added the IPsec support to uplink representor, however
as uplink representors have his private statistics where IPsec stats
is not part of it, that effectively makes IPsec stats hidden when uplink
representor stats queried.
Resolve by adding IPsec stats to uplink representor private statistics.
Fixes: 5589b8f1a2 ("net/mlx5e: Add IPsec support to uplink representor")
Signed-off-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Alaa Hleihel <alaa@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Current code wrongly uses the skb->protocol field which reflects the
outer l3 protocol to set the inner l3 type in Software Parser (SWP)
fields settings in the ethernet segment (eseg) in flows where inner
l3 exists like in Vxlan over ESP flow, the above method wrongly use
the outer protocol type instead of the inner one. thus breaking cases
where inner and outer headers have different protocols.
Fix by setting the inner l3 type in SWP according to the inner l3 ip
header version.
Fixes: 2ac9cfe782 ("net/mlx5e: IPSec, Add Innova IPSec offload TX data path")
Signed-off-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Change sdhcN aliases to mmcN to make them actually work. Currently the
board uses non-standard aliases sdhcN, which do not work, resulting in
mmc0 and mmc1 hosts randomly changing indices between boots.
Fixes: c4da5a5616 ("arm64: dts: qcom: Add msm8916 sdhci configuration nodes")
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/20211201020559.1611890-1-dmitry.baryshkov@linaro.org
The IDP2 and CRD boards share the EC and H1 parts, so define
all related device nodes into a common file and include them
in the idp2 and crd dts files to avoid duplication.
Signed-off-by: Kshitiz Godara <kgodara@codeaurora.org>
Signed-off-by: Rajendra Nayak <quic_rjendra@quicinc.com>
Reviewed-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/1638185497-26477-4-git-send-email-quic_rjendra@quicinc.com
CRD (Compute Reference Design) is a sc7280 based board, largely
derived from the existing IDP board design with some key deltas
1. has EC and H1 over SPI similar to IDP2
2. touchscreen and trackpad support
3. eDP display
We just add the barebones dts file here, subsequent patches will
add support for EC/H1 and other components.
Signed-off-by: Rajendra Nayak <quic_rjendra@quicinc.com>
Reviewed-by: Matthias Kaehlcke <mka@chromium.org>
Tested-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/1638185497-26477-3-git-send-email-quic_rjendra@quicinc.com
Document the qcom,sc7280-crd board based off sc7280 SoC,
The board is also known as hoglin in the Chrome OS builds,
so document the google,hoglin compatible as well.
Signed-off-by: Rajendra Nayak <quic_rjendra@quicinc.com>
Reviewed-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/1638185497-26477-2-git-send-email-quic_rjendra@quicinc.com
This property doesn't seem to exist in the documentation nor
in source code, but for some reason it is defined in a bunch
of device trees.
Signed-off-by: Dang Huynh <danct12@riseup.net>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/20211123162436.1507341-1-danct12@riseup.net
MSM8916 is similar to the other SoCs that had the RPM stats node added
in commit 290bc68465 ("arm64: dts: qcom: Enable RPM Sleep stats").
However, the dynamic offset readable at 0x14 seems only available on
some of the newer firmware versions. To be absolutely sure, make use
of the new qcom,msm8916-rpm-stats compatible that reads the sleep stats
from a fixed offset of 0xdba0.
Statistics are available for a "vmin" and "xosd" low power mode:
$ cat /sys/kernel/debug/qcom_stats/vmin
Count: 0
Last Entered At: 0
Last Exited At: 0
Accumulated Duration: 0
Client Votes: 0x0
$ cat /sys/kernel/debug/qcom_stats/xosd
Count: 0
Last Entered At: 0
Last Exited At: 0
Accumulated Duration: 0
Client Votes: 0x0
Cc: Maulik Shah <mkshah@codeaurora.org>
Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/20211119213953.31970-4-stephan@gerhold.net
Use eth_hw_addr_set function instead of writing the address directly to
net_device::dev_addr.
Fixes: adeef3e321 ("net: constify netdev->dev_addr")
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Link: https://lore.kernel.org/r/20211130143600.31970-1-jcmvbkbc@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix section mismatch warnings in xtsonic. The first one appears to be
bogus and after fixing the second one, the first one is gone.
WARNING: modpost: vmlinux.o(.text+0x529adc): Section mismatch in reference from the function sonic_get_stats() to the function .init.text:set_reset_devices()
The function sonic_get_stats() references
the function __init set_reset_devices().
This is often because sonic_get_stats lacks a __init
annotation or the annotation of set_reset_devices is wrong.
WARNING: modpost: vmlinux.o(.text+0x529b3b): Section mismatch in reference from the function xtsonic_probe() to the function .init.text:sonic_probe1()
The function xtsonic_probe() references
the function __init sonic_probe1().
This is often because xtsonic_probe lacks a __init
annotation or the annotation of sonic_probe1 is wrong.
Fixes: 74f2a5f0ef ("xtensa: Add support for the Sonic Ethernet device for the XT2000 board.")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: kernel test robot <lkp@intel.com>
Cc: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Cc: Finn Thain <fthain@telegraphics.com.au>
Cc: Chris Zankel <chris@zankel.net>
Cc: linux-xtensa@linux-xtensa.org
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Acked-by: Max Filippov <jcmvbkbc@gmail.com>
Link: https://lore.kernel.org/r/20211130063947.7529-1-rdunlap@infradead.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Rather than stealing bits from i915_sw_fence function pointer use
separate fields for function pointer and flags. If using two different
fields, the 4 byte alignment for the i915_sw_fence function pointer can
also be dropped.
v2:
(CI)
- Set new function field rather than flags in __i915_sw_fence_init
v3:
(Tvrtko)
- Remove BUG_ON(!fence->flags) in reinit as that will now blow up
- Only define fence->flags if CONFIG_DRM_I915_SW_FENCE_CHECK_DAG is
defined
v4:
- Rebase, resend for CI
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Alan Previn <alan.previn.teres.alexis@intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116194929.10211-1-matthew.brost@intel.com
The memory model has been updated to provide a stronger ordering
guarantee for unlock(A)+lock(B) on the same CPU/thread. Therefore add
two litmus tests describing this new guarantee, these tests are simple
yet can clearly show the usage of the new guarantee, also they can serve
as the self tests for the modification in the model.
Co-developed-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
It's better that we have some "standard" about which test should be put
in the litmus-tests directory because it helps future contributors
understand whether they should work on litmus-tests in kernel or Paul's
GitHub repo. Therefore explain a little bit on what a "representative"
litmus test is.
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
A recent discussion[1] shows that we are in favor of strengthening the
ordering of unlock + lock on the same CPU: a unlock and a po-after lock
should provide the so-called RCtso ordering, that is a memory access S
po-before the unlock should be ordered against a memory access R
po-after the lock, unless S is a store and R is a load.
The strengthening meets programmers' expection that "sequence of two
locked regions to be ordered wrt each other" (from Linus), and can
reduce the mental burden when using locks. Therefore add it in LKMM.
[1]: https://lore.kernel.org/lkml/20210909185937.GA12379@rowland.harvard.edu/
Co-developed-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Reviewed-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc)
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com> (RISC-V)
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The echo commands following initialization of the "oldrun" variable need
to be "tee"d to $oldrun/remote-log. This commit fixes several stragglers.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The "exit 4" in kvm-remote.sh is pointlessly redirected, so this commit
removes the redirection.
Fixes: 0092eae4cb ("torture: Add kvm-remote.sh script for distributed rcutorture test runs")
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit adjusts the TRACE02 scenario to use a pair of callback-flood
kthreads. This in turn forces lock contention on the single RCU Tasks
Trace callback queue, which forces use of all CPUs' queues, thus testing
this transition. (No, there is not yet any way to transition back.
Cc: Neeraj Upadhyay <neeraj.iitr10@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit enables two callback-flood kthreads for the TREE02 scenario
and 28 for the TREE10 scenario.
Cc: Neeraj Upadhyay <neeraj.iitr10@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Currently, a transient network error can kill a run if it happens while
downloading the tarball to one of the target systems. This commit
therefore does a 60-second wait and then a retry. If further experience
indicates, a more elaborate mechanism might be used later.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit makes kvm-find-errors.sh check for and report undefined
symbols that are detected at link time.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit brings the kvm.sh script's help text up to date with recently
(and some not-so-recently) added parameters.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
0Day/LKP observed that the refscale results fail to complete when larger
values of nrun (such as 300) are specified. The problem is that printk()
can accept at most a 1024-byte buffer. This commit therefore prints
the buffer whenever its length exceeds 800 bytes.
CC: Philip Li <philip.li@intel.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
There is only the one OOM error case in main_func(), so this commit
eliminates the errexit local variable in favor of a branch to cleanup
code.
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Because Tiny srcu_read_unlock() directly calls swake_up_one(), lockdep
complains when a pi lock is held across that srcu_read_unlock().
Although this is a lockdep false positive (there is no other CPU to
complete the deadlock cycle), lockdep is what it is at the moment.
This commit therefore prevents rcutorture from holding pi lock across
a Tiny srcu_read_unlock().
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Currently, nested readers occur only when a timer handler interrupts a
reader. This is rare, and is thus insufficient testing of the transition
between nesting levels. This commit therefore causes rcutorture nested
readers to be the rule rather than the exception.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
RCUTORTURE_RDR_MASK is currently not the bit indicated by
RCUTORTURE_RDR_SHIFT, but is instead all the bits less significant than
that one. This is an accident waiting to happen, so this commit makes
RCUTORTURE_RDR_MASK be that one bit and adjusts uses accordingly.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Currently, the check_all_holdout_tasks_trace() function removes all tasks
marked with ->trc_reader_checked from the holdout list, including those
with IPIs pending. This means that the IPI handler might arrive at
a task that has already been removed from the list, which is at best
an accident waiting to happen.
This commit therefore avoids removing tasks with IPIs pending from
the holdout list. This in turn means that the "if" condition in the
for_each_online_cpu() loop in rcu_tasks_trace_postgp() should always
evaluate to false, so a WARN_ON_ONCE() is added to check that.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Tiny SRCU readers can appear at task level, but also in interrupt and
softirq handlers. Because Tiny SRCU is selected only in kernels built
with CONFIG_SMP=n and CONFIG_PREEMPTION=n, it is not possible for a grace
period to start while there is a non-task-level SRCU reader executing.
This means that it does not make sense for __srcu_read_unlock() to awaken
the Tiny SRCU grace period, because that can only happen when the grace
period is waiting for one value of ->srcu_idx and __srcu_read_unlock()
is ending the last reader for some other value of ->srcu_idx. After all,
any such wakeup will be redundant.
Worse yet, in some cases, such wakeups generate lockdep splats:
======================================================
WARNING: possible circular locking dependency detected
5.15.0-rc1+ #3758 Not tainted
------------------------------------------------------
rcu_torture_rea/53 is trying to acquire lock:
ffffffff9514e6a8 (srcu_ctl.srcu_wq.lock){..-.}-{2:2}, at:
xa/0x30
but task is already holding lock:
ffff95c642479d80 (&p->pi_lock){-.-.}-{2:2}, at:
_extend+0x370/0x400
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (&p->pi_lock){-.-.}-{2:2}:
_raw_spin_lock_irqsave+0x2f/0x50
try_to_wake_up+0x50/0x580
swake_up_locked.part.7+0xe/0x30
swake_up_one+0x22/0x30
rcutorture_one_extend+0x1b6/0x400
rcu_torture_one_read+0x290/0x5d0
rcu_torture_timer+0x1a/0x70
call_timer_fn+0xa6/0x230
run_timer_softirq+0x493/0x4c0
__do_softirq+0xc0/0x371
irq_exit+0x73/0x90
sysvec_apic_timer_interrupt+0x63/0x80
asm_sysvec_apic_timer_interrupt+0x12/0x20
default_idle+0xb/0x10
default_idle_call+0x5e/0x170
do_idle+0x18a/0x1f0
cpu_startup_entry+0xa/0x10
start_kernel+0x678/0x69f
secondary_startup_64_no_verify+0xc2/0xcb
-> #0 (srcu_ctl.srcu_wq.lock){..-.}-{2:2}:
__lock_acquire+0x130c/0x2440
lock_acquire+0xc2/0x270
_raw_spin_lock_irqsave+0x2f/0x50
swake_up_one+0xa/0x30
rcutorture_one_extend+0x387/0x400
rcu_torture_one_read+0x290/0x5d0
rcu_torture_reader+0xac/0x200
kthread+0x12d/0x150
ret_from_fork+0x22/0x30
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&p->pi_lock);
lock(srcu_ctl.srcu_wq.lock);
lock(&p->pi_lock);
lock(srcu_ctl.srcu_wq.lock);
*** DEADLOCK ***
1 lock held by rcu_torture_rea/53:
#0: ffff95c642479d80 (&p->pi_lock){-.-.}-{2:2}, at:
_extend+0x370/0x400
stack backtrace:
CPU: 0 PID: 53 Comm: rcu_torture_rea Not tainted 5.15.0-rc1+
Hardware name: Red Hat KVM/RHEL-AV, BIOS
e_el8.5.0+746+bbd5d70c 04/01/2014
Call Trace:
check_noncircular+0xfe/0x110
? find_held_lock+0x2d/0x90
__lock_acquire+0x130c/0x2440
lock_acquire+0xc2/0x270
? swake_up_one+0xa/0x30
? find_held_lock+0x72/0x90
_raw_spin_lock_irqsave+0x2f/0x50
? swake_up_one+0xa/0x30
swake_up_one+0xa/0x30
rcutorture_one_extend+0x387/0x400
rcu_torture_one_read+0x290/0x5d0
rcu_torture_reader+0xac/0x200
? rcutorture_oom_notify+0xf0/0xf0
? __kthread_parkme+0x61/0x90
? rcu_torture_one_read+0x5d0/0x5d0
kthread+0x12d/0x150
? set_kthread_struct+0x40/0x40
ret_from_fork+0x22/0x30
This is a false positive because there is only one CPU, and both locks
are raw (non-preemptible) spinlocks. However, it is worthwhile getting
rid of the redundant wakeup, which has the side effect of breaking
the theoretical deadlock cycle. This commit therefore eliminates the
redundant wakeups.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Allow test programs to determine their thread ID.
Signed-off-by: Mark Brown <broonie@kernel.org>
Cc: Willy Tarreau <w@1wt.eu>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Note that mov to 32-bit register will zero extend to 64-bit register.
Thus `mov $60,%eax` has the same effect with `mov $60,%rax`. Use the
shorter opcode to achieve the same thing.
```
b8 3c 00 00 00 mov $60,%eax (5 bytes) [1]
48 c7 c0 3c 00 00 00 mov $60,%rax (7 bytes) [2]
```
Currently, we use [2]. Change it to [1] for shorter code.
Signed-off-by: Ammar Faizi <ammar.faizi@students.amikom.ac.id>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Linux x86-64 syscall only clobbers rax, rcx and r11 (and "memory").
- rax for the return value.
- rcx to save the return address.
- r11 to save the rflags.
Other registers are preserved.
Having r8, r9 and r10 in the syscall clobber list is harmless, but this
results in a missed-optimization.
As the syscall doesn't clobber r8-r10, GCC should be allowed to reuse
their value after the syscall returns to userspace. But since they are
in the clobber list, GCC will always miss this opportunity.
Remove them from the x86-64 syscall clobber list to help GCC generate
better code and fix the comment.
See also the x86-64 ABI, section A.2 AMD64 Linux Kernel Conventions,
A.2.1 Calling Conventions [1].
Extra note:
Some people may think it does not really give a benefit to remove r8,
r9 and r10 from the syscall clobber list because the impression of
syscall is a C function call, and function call always clobbers those 3.
However, that is not the case for nolibc.h, because we have a potential
to inline the "syscall" instruction (which its opcode is "0f 05") to the
user functions.
All syscalls in the nolibc.h are written as a static function with inline
ASM and are likely always inline if we use optimization flag, so this is
a profit not to have r8, r9 and r10 in the clobber list.
Here is the example where this matters.
Consider the following C code:
```
#include "tools/include/nolibc/nolibc.h"
#define read_abc(a, b, c) __asm__ volatile("nop"::"r"(a),"r"(b),"r"(c))
int main(void)
{
int a = 0xaa;
int b = 0xbb;
int c = 0xcc;
read_abc(a, b, c);
write(1, "test\n", 5);
read_abc(a, b, c);
return 0;
}
```
Compile with:
gcc -Os test.c -o test -nostdlib
With r8, r9, r10 in the clobber list, GCC generates this:
0000000000001000 <main>:
1000: f3 0f 1e fa endbr64
1004: 41 54 push %r12
1006: 41 bc cc 00 00 00 mov $0xcc,%r12d
100c: 55 push %rbp
100d: bd bb 00 00 00 mov $0xbb,%ebp
1012: 53 push %rbx
1013: bb aa 00 00 00 mov $0xaa,%ebx
1018: 90 nop
1019: b8 01 00 00 00 mov $0x1,%eax
101e: bf 01 00 00 00 mov $0x1,%edi
1023: ba 05 00 00 00 mov $0x5,%edx
1028: 48 8d 35 d1 0f 00 00 lea 0xfd1(%rip),%rsi
102f: 0f 05 syscall
1031: 90 nop
1032: 31 c0 xor %eax,%eax
1034: 5b pop %rbx
1035: 5d pop %rbp
1036: 41 5c pop %r12
1038: c3 ret
GCC thinks that syscall will clobber r8, r9, r10. So it spills 0xaa,
0xbb and 0xcc to callee saved registers (r12, rbp and rbx). This is
clearly extra memory access and extra stack size for preserving them.
But syscall does not actually clobber them, so this is a missed
optimization.
Now without r8, r9, r10 in the clobber list, GCC generates better code:
0000000000001000 <main>:
1000: f3 0f 1e fa endbr64
1004: 41 b8 aa 00 00 00 mov $0xaa,%r8d
100a: 41 b9 bb 00 00 00 mov $0xbb,%r9d
1010: 41 ba cc 00 00 00 mov $0xcc,%r10d
1016: 90 nop
1017: b8 01 00 00 00 mov $0x1,%eax
101c: bf 01 00 00 00 mov $0x1,%edi
1021: ba 05 00 00 00 mov $0x5,%edx
1026: 48 8d 35 d3 0f 00 00 lea 0xfd3(%rip),%rsi
102d: 0f 05 syscall
102f: 90 nop
1030: 31 c0 xor %eax,%eax
1032: c3 ret
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: x86@kernel.org
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: David Laight <David.Laight@ACULAB.COM>
Acked-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Ammar Faizi <ammar.faizi@students.amikom.ac.id>
Link: https://gitlab.com/x86-psABIs/x86-64-ABI/-/wikis/x86-64-psABI [1]
Link: https://lore.kernel.org/lkml/20211011040344.437264-1-ammar.faizi@students.amikom.ac.id/
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Ammar Faizi reported that our exit code handling is wrong. We truncate
it to the lowest 8 bits but the syscall itself is expected to take a
regular 32-bit signed integer, not an unsigned char. It's the kernel
that later truncates it to the lowest 8 bits. The difference is visible
in strace, where the program below used to show exit(255) instead of
exit(-1):
int main(void)
{
return -1;
}
This patch applies the fix to all archs. x86_64, i386, arm64, armv7 and
mips were all tested and confirmed to work fine now. Risc-v was not
tested but the change is trivial and exactly the same as for other archs.
Reported-by: Ammar Faizi <ammar.faizi@students.amikom.ac.id>
Cc: stable@vger.kernel.org
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
After re-checking in the spec and comparing stack offsets with glibc,
The last pushed argument must be 16-byte aligned (i.e. aligned before the
call) so that in the callee esp+4 is multiple of 16, so the principle is
the 32-bit equivalent to what Ammar fixed for x86_64. It's possible that
32-bit code using SSE2 or MMX could have been affected. In addition the
frame pointer ought to be zero at the deepest level.
Link: https://gitlab.com/x86-psABIs/i386-ABI/-/wikis/Intel386-psABI
Cc: Ammar Faizi <ammar.faizi@students.amikom.ac.id>
Cc: stable@vger.kernel.org
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Before this patch, the `_start` function looks like this:
```
0000000000001170 <_start>:
1170: pop %rdi
1171: mov %rsp,%rsi
1174: lea 0x8(%rsi,%rdi,8),%rdx
1179: and $0xfffffffffffffff0,%rsp
117d: sub $0x8,%rsp
1181: call 1000 <main>
1186: movzbq %al,%rdi
118a: mov $0x3c,%rax
1191: syscall
1193: hlt
1194: data16 cs nopw 0x0(%rax,%rax,1)
119f: nop
```
Note the "and" to %rsp with $-16, it makes the %rsp be 16-byte aligned,
but then there is a "sub" with $0x8 which makes the %rsp no longer
16-byte aligned, then it calls main. That's the bug!
What actually the x86-64 System V ABI mandates is that right before the
"call", the %rsp must be 16-byte aligned, not after the "call". So the
"sub" with $0x8 here breaks the alignment. Remove it.
An example where this rule matters is when the callee needs to align
its stack at 16-byte for aligned move instruction, like `movdqa` and
`movaps`. If the callee can't align its stack properly, it will result
in segmentation fault.
x86-64 System V ABI also mandates the deepest stack frame should be
zero. Just to be safe, let's zero the %rbp on startup as the content
of %rbp may be unspecified when the program starts. Now it looks like
this:
```
0000000000001170 <_start>:
1170: pop %rdi
1171: mov %rsp,%rsi
1174: lea 0x8(%rsi,%rdi,8),%rdx
1179: xor %ebp,%ebp # zero the %rbp
117b: and $0xfffffffffffffff0,%rsp # align the %rsp
117f: call 1000 <main>
1184: movzbq %al,%rdi
1188: mov $0x3c,%rax
118f: syscall
1191: hlt
1192: data16 cs nopw 0x0(%rax,%rax,1)
119d: nopl (%rax)
```
Cc: Bedirhan KURT <windowz414@gnuweeb.org>
Cc: Louvian Lyndal <louvianlyndal@gmail.com>
Reported-by: Peter Cordes <peter@cordes.ca>
Signed-off-by: Ammar Faizi <ammar.faizi@students.amikom.ac.id>
[wt: I did this on purpose due to a misunderstanding of the spec, other
archs will thus have to be rechecked, particularly i386]
Cc: stable@vger.kernel.org
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The default kasan_record_aux_stack() calls stack_depot_save() with GFP_NOWAIT,
which in turn can then call alloc_pages(GFP_NOWAIT, ...). In general, however,
it is not even possible to use either GFP_ATOMIC nor GFP_NOWAIT in certain
non-preemptive contexts/RT kernel including raw_spin_locks (see gfp.h and ab00db216c).
Fix it by instructing stackdepot to not expand stack storage via alloc_pages()
in case it runs out by using kasan_record_aux_stack_noalloc().
Jianwei Hu reported:
BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:969
in_atomic(): 0, irqs_disabled(): 1, non_block: 0, pid: 15319, name: python3
INFO: lockdep is turned off.
irq event stamp: 0
hardirqs last enabled at (0): [<0000000000000000>] 0x0
hardirqs last disabled at (0): [<ffffffff856c8b13>] copy_process+0xaf3/0x2590
softirqs last enabled at (0): [<ffffffff856c8b13>] copy_process+0xaf3/0x2590
softirqs last disabled at (0): [<0000000000000000>] 0x0
CPU: 6 PID: 15319 Comm: python3 Tainted: G W O 5.15-rc7-preempt-rt #1
Hardware name: Supermicro SYS-E300-9A-8C/A2SDi-8C-HLN4F, BIOS 1.1b 12/17/2018
Call Trace:
show_stack+0x52/0x58
dump_stack+0xa1/0xd6
___might_sleep.cold+0x11c/0x12d
rt_spin_lock+0x3f/0xc0
rmqueue+0x100/0x1460
rmqueue+0x100/0x1460
mark_usage+0x1a0/0x1a0
ftrace_graph_ret_addr+0x2a/0xb0
rmqueue_pcplist.constprop.0+0x6a0/0x6a0
__kasan_check_read+0x11/0x20
__zone_watermark_ok+0x114/0x270
get_page_from_freelist+0x148/0x630
is_module_text_address+0x32/0xa0
__alloc_pages_nodemask+0x2f6/0x790
__alloc_pages_slowpath.constprop.0+0x12d0/0x12d0
create_prof_cpu_mask+0x30/0x30
alloc_pages_current+0xb1/0x150
stack_depot_save+0x39f/0x490
kasan_save_stack+0x42/0x50
kasan_save_stack+0x23/0x50
kasan_record_aux_stack+0xa9/0xc0
__call_rcu+0xff/0x9c0
call_rcu+0xe/0x10
put_object+0x53/0x70
__delete_object+0x7b/0x90
kmemleak_free+0x46/0x70
slab_free_freelist_hook+0xb4/0x160
kfree+0xe5/0x420
kfree_const+0x17/0x30
kobject_cleanup+0xaa/0x230
kobject_put+0x76/0x90
netdev_queue_update_kobjects+0x17d/0x1f0
... ...
ksys_write+0xd9/0x180
__x64_sys_write+0x42/0x50
do_syscall_64+0x38/0x50
entry_SYSCALL_64_after_hwframe+0x44/0xa9
Links: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/include/linux/kasan.h?id=7cb3007ce2da27ec02a1a3211941e7fe6875b642
Fixes: 84109ab585 ("rcu: Record kvfree_call_rcu() call stack for KASAN")
Fixes: 26e760c9a7 ("rcu: kasan: record and print call_rcu() call stack")
Reported-by: Jianwei Hu <jianwei.hu@windriver.com>
Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Acked-by: Marco Elver <elver@google.com>
Tested-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Jun Miao <jun.miao@intel.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
