arm64 fixes for 6.13-rc3

Hi Linus,

Please pull the arm64 fixes below. Thanks.

The following changes since commit fac04efc5c793dccbd07e2d59af9f90b7fc0dca4:

Linux 6.13-rc2 (2024-12-08 14:03:39 -0800)

are available in the Git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux tags/arm64-fixes

for you to fetch changes up to ce03573a1917532da06057da9f8e74a2ee9e2ac9:

kselftest/arm64: abi: fix SVCR detection (2024-12-13 14:59:37 +0000)

----------------------------------------------------------------
arm64 fixes for 6.13-rc3:

- arm64 stacktrace: address some fallout from the recent changes to unwinding across exception boundaries

- Ensure the arm64 signal delivery failure is recoverable - only override the return registers after all the user accesses took place

- Fix the arm64 kselftest access to SVCR - only when SME is detected

----------------------------------------------------------------
Kevin Brodsky (1):
arm64: signal: Ensure signal delivery failure is recoverable

Mark Rutland (2):
arm64: stacktrace: Skip reporting LR at exception boundaries
arm64: stacktrace: Don't WARN when unwinding other tasks

Weizhao Ouyang (1):
kselftest/arm64: abi: fix SVCR detection

 arch/arm64/kernel/signal.c                         | 48 +++++++++++++++-------
 arch/arm64/kernel/stacktrace.c                     | 32 ++++-----------
 .../testing/selftests/arm64/abi/syscall-abi-asm.S  | 32 +++++++--------
 3 files changed, 55 insertions(+), 57 deletions(-)

Catalin

Hi Catalin, Mark,

It seems that kselftests can hit a warning that was addressed in the commit above:

<12>[   74.172204] kselftest: Running tests in ftrace
TAP version 13
1..1
# timeout set to 0
# selftests: ftrace: ftracetest-ktap
# unlink: cannot unlink '/opt/kselftests/default-in-kernel/ftrace/logs/latest': No such file or directory
# TAP version 13
# 1..145
# ok 1 Basic trace file check
<4>[  108.427656] hrtimer: interrupt took 13044036 ns
<4>[  161.249175] sched: DL replenish lagged too much
<4>[  163.845977] ------------[ cut here ]------------
<4>[  163.851787] WARNING: CPU: 1 PID: 31 at arch/arm64/kernel/stacktrace.c:141 arch_stack_walk+0x4a0/0x4b0
<4>[  163.861012] Modules linked in: crct10dif_ce sm3_ce sm3 sha3_ce sha512_ce sha512_arm64 fuse drm backlight ip_tables x_tables
<4>[  163.890268] CPU: 1 UID: 0 PID: 31 Comm: kworker/1:1 Not tainted 6.13.0-rc2 #1
<4>[  163.897386] Hardware name: linux,dummy-virt (DT)
<4>[  163.904217] Workqueue:  0x0 (mm_percpu_wq)
<4>[  163.915526] pstate: 624020c9 (nZCv daIF +PAN -UAO +TCO -DIT -SSBS BTYPE=--)
<4>[  163.924218] pc : arch_stack_walk+0x4a0/0x4b0
<4>[  163.930593] lr : arch_stack_walk+0x2bc/0x4b0
<4>[  163.937101] sp : ffff80008000b920
<4>[  163.942456] x29: ffff80008000b9d0 x28: fff00000ff6d9400 x27: fff00000ff6d9928
<4>[  163.954275] x26: 0000000000000001 x25: 00000000000000c0 x24: ffffae003d389d40
<4>[  163.967208] x23: 09cfae003ae216b4 x22: ffffae003ae2e4b0 x21: ffffae003ae35c70
<4>[  163.978823] x20: ffff80008000ba10 x19: ffffae003ae216f0 x18: 0000000000000014
<4>[  163.989751] x17: 0000000000000000 x16: ffff800080008000 x15: 00000000b5b23a12
<4>[  164.000780] x14: 000000005865dcaf x13: 00000000b77eca6d x12: 0000959eae5a34e8
<4>[  164.011592] x11: ffffae003d44a670 x10: ffff80008000b920 x9 : ffffae003ae2e4b0
<4>[  164.022498] x8 : ffff800080043e00 x7 : fff00000c682c000 x6 : 00000000ffffffff
<4>[  164.033396] x5 : fff00000c682c000 x4 : 0000000000000000 x3 : ffff80008000bda0
<4>[  164.045100] x2 : ffffae003ae2e4b0 x1 : ffffae003ae2e4b0 x0 : ffffae003ae2e4b0
<4>[  164.057084] Call trace:
<4>[  164.061489]  arch_stack_walk+0x4a0/0x4b0 (P)
<4>[  164.067225]  arch_stack_walk+0x2bc/0x4b0 (L)
<4>[  164.073009]  profile_pc+0x44/0x80
<4>[  164.078489]  profile_tick+0x50/0x80 (F)
<4>[  164.085151]  tick_nohz_handler+0xcc/0x160 (F)
<4>[  164.091427]  __hrtimer_run_queues+0x2c4/0x358 (F)
<4>[  164.098138]  hrtimer_interrupt+0xf4/0x268 (F)
<4>[  164.105401]  arch_timer_handler_phys+0x34/0x58 (F)
<4>[  164.114958]  handle_percpu_devid_irq+0x8c/0x218 (F)
<4>[  164.120914]  generic_handle_domain_irq+0x34/0x58 (F)
<4>[  164.127570]  gic_handle_irq+0x54/0x128 (F)
<4>[  164.135008]  do_interrupt_handler+0x58/0x98 (F)
<4>[  164.140769]  el1_interrupt+0x34/0x68 (F)
<4>[  164.146665]  el1h_64_irq_handler+0x18/0x28
<4>[  164.151813]  el1h_64_irq+0x6c/0x70
<4>[  164.160007]  _raw_spin_unlock_irq+0x28/0x58 (PF)
<4>[  164.176292] ------------[ cut here ]------------
<4>[  164.187816] WARNING: CPU: 1 PID: 31 at arch/arm64/kernel/stacktrace.c:141 dump_backtrace+0x620/0x660
<4>[  164.209007] Modules linked in: crct10dif_ce sm3_ce sm3 sha3_ce sha512_ce sha512_arm64 fuse drm backlight ip_tables x_tables
<4>[  164.252638] CPU: 1 UID: 0 PID: 31 Comm: kworker/1:1 Not tainted 6.13.0-rc2 #1
<4>[  164.260491] Hardware name: linux,dummy-virt (DT)
<4>[  164.266531] Workqueue:  0x0 (mm_percpu_wq)
<4>[  164.275972] pstate: 624023c9 (nZCv DAIF +PAN -UAO +TCO -DIT -SSBS BTYPE=--)
<4>[  164.282685] pc : dump_backtrace+0x620/0x660
<4>[  164.288093] lr : dump_backtrace+0x410/0x660
<4>[  164.293392] sp : ffff80008000b510
<4>[  164.297983] x29: ffff80008000b5c0 x28: fff00000c08c91c0 x27: fff00000ff6d9928
<4>[  164.309075] x26: ffff80008000bda0 x25: ffffae003cd0c000 x24: ffffae003cedfaa0
<4>[  164.320534] x23: ffffae003cce90b8 x22: ffffae003ae35c70 x21: ffffae003ccd80a8
<4>[  164.331352] x20: ffffae003cd1fd38 x19: fff00000c08c91c0 x18: 0000000000000006
<4>[  164.343050] x17: 0000000000000000 x16: ffff800080008000 x15: ffff80008000af40
<4>[  164.354100] x14: ffff80010000b0fa x13: ffff80008000b102 x12: ffffae003d410098
<4>[  164.366177] x11: ffffae003d44a670 x10: ffff80008000b350 x9 : ffffae003ae2e4b0
<4>[  164.378146] x8 : ffff800080043e00 x7 : fff00000c682c000 x6 : 00000000ffffffff
<4>[  164.389041] x5 : fff00000c682c000 x4 : 0000000000000000 x3 : ffff80008000bda0
<4>[  164.400262] x2 : ffffae003ae2e4b0 x1 : ffffae003ae2e4b0 x0 : ffffae003ae2e4b0
<4>[  164.411526] Call trace:
<4>[  164.415800]  dump_backtrace+0x620/0x660 (P)
<4>[  164.421415]  dump_backtrace+0x410/0x660 (L)
<4>[  164.427393]  show_regs+0x34/0x50 (F)
<4>[  164.434342]  __warn+0x98/0x198 (F)
<4>[  164.439610]  report_bug+0x1dc/0x200 (F)
<4>[  164.444981]  bug_handler+0x2c/0x90
<4>[  164.450151]  call_break_hook+0x6c/0x88 (F)
<4>[  164.455802]  brk_handler+0x24/0x70 (F)
<4>[  164.460935]  do_debug_exception+0x74/0x118 (F)
<4>[  164.466742]  el1_dbg+0x70/0x90 (F)
<4>[  164.471595]  el1h_64_sync_handler+0xd4/0x120
<4>[  164.477225]  el1h_64_sync+0x6c/0x70
<4>[  164.482730]  arch_stack_walk+0x4a0/0x4b0 (P)
<4>[  164.488313]  arch_stack_walk+0x2bc/0x4b0 (L)
<4>[  164.494329]  profile_pc+0x44/0x80
<4>[  164.499430]  profile_tick+0x50/0x80 (F)
<4>[  164.504776]  tick_nohz_handler+0xcc/0x160 (F)
<4>[  164.511508]  __hrtimer_run_queues+0x2c4/0x358 (F)
<4>[  164.519213]  hrtimer_interrupt+0xf4/0x268 (F)
<4>[  164.525536]  arch_timer_handler_phys+0x34/0x58 (F)
<4>[  164.531384]  handle_percpu_devid_irq+0x8c/0x218 (F)
<4>[  164.537153]  generic_handle_domain_irq+0x34/0x58 (F)
<4>[  164.543448]  gic_handle_irq+0x54/0x128 (F)
<4>[  164.548744]  do_interrupt_handler+0x58/0x98 (F)
<4>[  164.555359]  el1_interrupt+0x34/0x68 (F)
<4>[  164.561123]  el1h_64_irq_handler+0x18/0x28
<4>[  164.567261]  el1h_64_irq+0x6c/0x70
<4>[  164.573197]  _raw_spin_unlock_irq+0x28/0x58 (PF)
<4>[  164.580018] ---[ end trace 0000000000000000 ]---
<4>[  164.585812] ---[ end trace 0000000000000000 ]---

Thanks,
Sasha

I assume you're referring to commit:

32ed1205682ec42a ("arm64: stacktrace: Skip reporting LR at exception boundaries")

... ?

That was intended to fix this specific issue, as explained in the commit message (with a very similar splat from running the ftrace kselftests).

Am I missing something? i.e. are you saying it *doesn't* fix that?

Mark.

I was actually thinking 65ac33bed8b9 ("arm64: stacktrace: Don't WARN when unwinding other tasks"), sorry for the confusion

Right - LKFT is able to trigger the warning I copied on Linus's tree.

Thanks,
Sasha

That commit should have no impact the case you report, which is seen when unwinding the current task, not another task.

The commit immediately before that (32ed1205682ec42a) *should* fix the issue triggered by the ftrace kselftests.

Is that "Right" as in "Right, that commit fixes it", or "Right, that commit does not fix it"?

The warning you reported says:

	Not tainted 6.13.0-rc2 #1

... which is *before* either of the two fixes above, since the arm64 for-next/fixes branch is based on v6.13-rc2.

Are you seeing the warning with those applied? e.g. do you see this in v6.13-rc3? I cannot reproduce the issue on v6.13-rc3 so far.

Mark.