perf/x86/intel/pt: Fix crash with stop filters in single-range mode
Subsystems (2)
PATCH 1/1 perf/x86/intel/pt: Fix crash with stop filters in single-range mode
Add a check for !buf->single before calling pt_buffer_region_size in a place where a missing check can cause a kernel crash.
Fixes a bug introduced by 670638477aede0d7a355ced04b569214aa3feacd, which added a support for PT single-range output mode. Since that commit if a PT stop filter range is hit while tracing, the kernel will crash because of a null pointer dereference in pt_handle_status due to calling pt_buffer_region_size without a ToPA configured.
The commit which introduced single-range mode guarded almost all uses of the ToPA buffer variables with checks of the buf->single variable, but missed the case where tracing was stopped by the PT hardware, which happens when execution hits a configured stop filter.
Tested that hitting a stop filter while PT recording successfully records a trace with this patch but crashes without this patch.
--- arch/x86/events/intel/pt.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
arch/x86/events/intel/pt.c +3 -2
@@ -897,8 +897,9 @@ static void pt_handle_status(struct pt *pt)* means we are already losing data; need to let the decoder * know. */ -if (!intel_pt_validate_hw_cap (PT_CAP_topa_multiple_entries) || - buf->output_off ==pt_buffer_region_size (buf)) { +if (!buf->single && + (!intel_pt_validate_hw_cap (PT_CAP_topa_multiple_entries) || + buf->output_off ==pt_buffer_region_size (buf))) {perf_aux_output_flag (&pt->handle,PERF_AUX_FLAG_TRUNCATED ); advance++;
Signature · 1 line
2.21.1 (Apple Git-122.3)
Quoted 9 lines · Tristan Hume
Add a check for !buf->single before calling
pt_buffer_region_sizein a place where a missing check can cause a kernel crash.Fixes a bug introduced by
670638477aede0d7a355ced04b569214aa3feacd, which added a support for PT single-range output mode. Since that commit if a PT stop filter range is hit while tracing, the kernel will crash because of a null pointer dereference inpt_handle_statusdue to callingpt_buffer_region_sizewithout a ToPA configured.
Quoted 4 lines
The commit which introduced single-range mode guarded almost all uses of the ToPA buffer variables with checks of the
buf->singlevariable, but missed the case where tracing was stopped by the PT hardware, which happens when execution hits a configured stop filter.
Please add to the commit message:
Example:
# perf record -S -e intel_pt// --filter tracestop __schedule sleep 1
[ 6148.371226] BUG: unable to handle page fault for address: fffffffffffff028 [ 6148.371236] #PF: supervisor read access in kernel mode [ 6148.371240] #PF: error_code(0x0000) - not-present page [ 6148.371244] PGD 6410067 P4D 6410067 PUD 6412067 PMD 0 [ 6148.371253] Oops: 0000 [#1] PREEMPT SMP PTI [ 6148.371258] CPU: 0 PID: 7279 Comm: perf Not tainted 5.16.0+ #24 [ 6148.371264] Hardware name: Intel(R) Client Systems NUC8i7BEH/NUC8BEB, BIOS BECFL357.86A.0089.2021.0621.1343 06/21/2021 [ 6148.371268] RIP: 0010:pt_handle_status+0x94/0x270 [ 6148.371278] Code: 26 01 00 00 8b 4b 28 4c 8b 43 20 48 8b 73 30 48 89 ca 49 8d b8 28 f0 ff ff 48 8d 0c cf 83 fa ff 75 08 49 63 50 20 48 8d 0c d7 <0f> b7 09 ba 01 00 00 00 66 c1 e9 06 83 e1 0f 83 c1 0c d3 e2 48 63 [ 6148.371283] RSP: 0018:ffffc9000308fad0 EFLAGS: 00010013 [ 6148.371288] RAX: 0000000000000002 RBX: ffff8881d62ed500 RCX: fffffffffffff028 [ 6148.371292] RDX: 0000000000000000 RSI: 0000000000000510 RDI: fffffffffffff028 [ 6148.371296] RBP: ffffc9000308fae8 R08: 0000000000000000 R09: 0000000000000000 [ 6148.371299] R10: 0000000000000000 R11: 0000000000000000 R12: 0000051000000002 [ 6148.371302] R13: ffff88846dc1e020 R14: 0000000000000004 R15: ffff88846dc36208 [ 6148.371306] FS: 00007f1f8630b3c0(0000) GS:ffff88846dc00000(0000) knlGS:0000000000000000 [ 6148.371311] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 6148.371315] CR2: fffffffffffff028 CR3: 0000000124ea6001 CR4: 00000000003706f0 [ 6148.371318] Call Trace: [ 6148.371321] <TASK> [ 6148.371325] pt_event_stop+0x8a/0xd0 [ 6148.371333] pt_event_del+0x13/0x20 [ 6148.371338] event_sched_out.isra.0+0x7a/0x1c0 [ 6148.371345] group_sched_out.part.0+0x43/0x90 [ 6148.371350] __perf_event_disable+0x105/0x1f0 [ 6148.371355] event_function+0xab/0xf0 [ 6148.371362] ? perf_duration_warn+0x30/0x30 [ 6148.371369] remote_function+0x51/0x60 [ 6148.371375] generic_exec_single+0x66/0xa0 [ 6148.371381] smp_call_function_single+0xb9/0x170 [ 6148.371386] ? perf_duration_warn+0x30/0x30 [ 6148.371393] ? preempt_count_add+0x74/0xc0 [ 6148.371399] event_function_call+0x114/0x120 [ 6148.371405] ? group_sched_out.part.0+0x90/0x90 [ 6148.371409] ? cpu_clock_event_read+0x30/0x30 [ 6148.371415] _perf_event_disable+0x48/0x60 [ 6148.371421] ? event_function_call+0x120/0x120 [ 6148.371427] perf_event_for_each_child+0x3f/0x90 [ 6148.371432] ? event_function_call+0x120/0x120 [ 6148.371438] _perf_ioctl+0x20c/0x910 [ 6148.371443] ? force_compatible_cpus_allowed_ptr+0x1d0/0x1d0 [ 6148.371451] ? affine_move_task+0x419/0x610 [ 6148.371459] ? __this_cpu_preempt_check+0x13/0x20 [ 6148.371468] ? xfd_validate_state+0x23/0x80 [ 6148.371475] perf_ioctl+0x48/0x70 [ 6148.371481] __x64_sys_ioctl+0x91/0xc0 [ 6148.371490] do_syscall_64+0x43/0xc0 [ 6148.371495] entry_SYSCALL_64_after_hwframe+0x44/0xae [ 6148.371504] RIP: 0033:0x7f1f86d8050b [ 6148.371509] Code: 0f 1e fa 48 8b 05 85 39 0d 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 55 39 0d 00 f7 d8 64 89 01 48 [ 6148.371513] RSP: 002b:00007fffd59a86c8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 [ 6148.371519] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f1f86d8050b [ 6148.371522] RDX: 0000000000000000 RSI: 0000000000002401 RDI: 0000000000000005 [ 6148.371525] RBP: 00007fffd59a8700 R08: 0000000000000000 R09: 00007f1f86e54cc0 [ 6148.371528] R10: 0000562bd7bad010 R11: 0000000000000246 R12: 0000000000000000 [ 6148.371531] R13: 0000562bd7bc8f20 R14: 0000562bd7bb1e10 R15: 0000562bd7bc8f20 [ 6148.371537] </TASK> [ 6148.371539] Modules linked in: [ 6148.371544] CR2: fffffffffffff028 [ 6148.371547] ---[ end trace 542919a16ee8853b ]---
Quoted 7 lines
Tested that hitting a stop filter while PT recording successfully records a trace with this patch but crashes without this patch.
Signed-off-by Tristan Hume <[email protected]>Cc Alexander Shishkin <[email protected]>Cc Peter Zijlstra (Intel) <[email protected]>
Quoted 21 lines
--- arch/x86/events/intel/pt.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)arch/x86/events/intel/pt.c +3
-2@@ -897,8 +897,9 @@ static void pt_handle_status(struct pt *pt)* means we are already losing data; need to let the decoder * know. */ -if (!intel_pt_validate_hw_cap (PT_CAP_topa_multiple_entries) || - buf->output_off ==pt_buffer_region_size (buf)) { +if (!buf->single && + (!intel_pt_validate_hw_cap (PT_CAP_topa_multiple_entries) || + buf->output_off ==pt_buffer_region_size (buf))) {perf_aux_output_flag (&pt->handle,PERF_AUX_FLAG_TRUNCATED ); advance++;
The following commit has been merged into the perf/urgent branch of tip:
The following commit has been merged into the perf/urgent branch of tip:
Commit-ID: 1d9093457b243061a9bba23543c38726e864a643 Gitweb: https://git.kernel.org/tip/1d9093457b243061a9bba23543c38726e864a643 Author: Tristan Hume <[email protected]> AuthorDate: Thu, 27 Jan 2022 17:08:06 -05:00 Committer: Peter Zijlstra <[email protected]> CommitterDate: Wed, 02 Feb 2022 13:11:40 +01:00
perf/x86/intel/pt: Fix crash with stop filters in single-range mode
Add a check for !buf->single before calling pt_buffer_region_size in a place where a missing check can cause a kernel crash.
Fixes a bug introduced by commit 670638477aed ("perf/x86/intel/pt:
Opportunistically use single range output mode"), which added a support for PT single-range output mode. Since that commit if a PT stop filter range is hit while tracing, the kernel will crash because of a null pointer dereference in pt_handle_status due to calling pt_buffer_region_size without a ToPA configured.
The commit which introduced single-range mode guarded almost all uses of the ToPA buffer variables with checks of the buf->single variable, but missed the case where tracing was stopped by the PT hardware, which happens when execution hits a configured stop filter.
Tested that hitting a stop filter while PT recording successfully records a trace with this patch but crashes without this patch.
--- arch/x86/events/intel/pt.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
arch/x86/events/intel/pt.c +3 -2
@@ -897,8 +897,9 @@ static void pt_handle_status(struct pt *pt)* means we are already losing data; need to let the decoder * know. */ -if (!intel_pt_validate_hw_cap (PT_CAP_topa_multiple_entries) || - buf->output_off ==pt_buffer_region_size (buf)) { +if (!buf->single && + (!intel_pt_validate_hw_cap (PT_CAP_topa_multiple_entries) || + buf->output_off ==pt_buffer_region_size (buf))) {perf_aux_output_flag (&pt->handle,PERF_AUX_FLAG_TRUNCATED ); advance++;