[v3] perf/x86/amd: check event before enable to avoid GPF

[PATCH] [PATCH v3] perf/x86/amd: check event before enable to avoid GPF

Posted by George Kennedy 1 year, 4 months ago

On AMD machines cpuc->events[idx] can become NULL in a subtle race
condition with NMI->throttle->x86_pmu_stop().

Check event for NULL in amd_pmu_enable_all() before enable to avoid a GPF.
This appears to be an AMD only issue.

Syzkaller reported a GPF in amd_pmu_enable_all.

INFO: NMI handler (perf_event_nmi_handler) took too long to run: 13.143
    msecs
Oops: general protection fault, probably for non-canonical address
    0xdffffc0000000034: 0000  PREEMPT SMP KASAN NOPTI
KASAN: null-ptr-deref in range [0x00000000000001a0-0x00000000000001a7]
CPU: 0 UID: 0 PID: 328415 Comm: repro_36674776 Not tainted 6.12.0-rc1-syzk
RIP: 0010:x86_pmu_enable_event (arch/x86/events/perf_event.h:1195
    arch/x86/events/core.c:1430)
RSP: 0018:ffff888118009d60 EFLAGS: 00010012
RAX: dffffc0000000000 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 0000000000000034 RSI: 0000000000000000 RDI: 00000000000001a0
RBP: 0000000000000001 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
R13: ffff88811802a440 R14: ffff88811802a240 R15: ffff8881132d8601
FS:  00007f097dfaa700(0000) GS:ffff888118000000(0000) GS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000200001c0 CR3: 0000000103d56000 CR4: 00000000000006f0
Call Trace:
 <IRQ>
amd_pmu_enable_all (arch/x86/events/amd/core.c:760 (discriminator 2))
x86_pmu_enable (arch/x86/events/core.c:1360)
event_sched_out (kernel/events/core.c:1191 kernel/events/core.c:1186
    kernel/events/core.c:2346)
__perf_remove_from_context (kernel/events/core.c:2435)
event_function (kernel/events/core.c:259)
remote_function (kernel/events/core.c:92 (discriminator 1)
    kernel/events/core.c:72 (discriminator 1))
__flush_smp_call_function_queue (./arch/x86/include/asm/jump_label.h:27
    ./include/linux/jump_label.h:207 ./include/trace/events/csd.h:64
    kernel/smp.c:135 kernel/smp.c:540)
__sysvec_call_function_single (./arch/x86/include/asm/jump_label.h:27
    ./include/linux/jump_label.h:207
    ./arch/x86/include/asm/trace/irq_vectors.h:99 arch/x86/kernel/smp.c:272)
sysvec_call_function_single (arch/x86/kernel/smp.c:266 (discriminator 47)
    arch/x86/kernel/smp.c:266 (discriminator 47))
 </IRQ>

Reported-by: syzkaller <syzkaller@googlegroups.com>
Signed-off-by: George Kennedy <george.kennedy@oracle.com>
---
 Suggested comment now with the code.

 arch/x86/events/amd/core.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index 920e3a640cad..6615ace15f5d 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -762,7 +762,12 @@ static void amd_pmu_enable_all(int added)
 		if (!test_bit(idx, cpuc->active_mask))
 			continue;
 
-		amd_pmu_enable_event(cpuc->events[idx]);
+		/*
+		 * FIXME: cpuc->events[idx] can become NULL in a subtle race
+		 * condition with NMI->throttle->x86_pmu_stop().
+		 */
+		if (cpuc->events[idx])
+			amd_pmu_enable_event(cpuc->events[idx]);
 	}
 }
 
-- 
2.39.3

Re: [PATCH] [PATCH v3] perf/x86/amd: check event before enable to avoid GPF

Posted by Peter Zijlstra 3 months, 3 weeks ago

On Tue, Oct 08, 2024 at 08:00:53AM -0500, George Kennedy wrote:
> On AMD machines cpuc->events[idx] can become NULL in a subtle race
> condition with NMI->throttle->x86_pmu_stop().
> 
> Check event for NULL in amd_pmu_enable_all() before enable to avoid a GPF.
> This appears to be an AMD only issue.
> 
> Syzkaller reported a GPF in amd_pmu_enable_all.
> 
> INFO: NMI handler (perf_event_nmi_handler) took too long to run: 13.143
>     msecs
> Oops: general protection fault, probably for non-canonical address
>     0xdffffc0000000034: 0000  PREEMPT SMP KASAN NOPTI
> KASAN: null-ptr-deref in range [0x00000000000001a0-0x00000000000001a7]
> CPU: 0 UID: 0 PID: 328415 Comm: repro_36674776 Not tainted 6.12.0-rc1-syzk
> RIP: 0010:x86_pmu_enable_event (arch/x86/events/perf_event.h:1195
>     arch/x86/events/core.c:1430)
> RSP: 0018:ffff888118009d60 EFLAGS: 00010012
> RAX: dffffc0000000000 RBX: 0000000000000000 RCX: 0000000000000000
> RDX: 0000000000000034 RSI: 0000000000000000 RDI: 00000000000001a0
> RBP: 0000000000000001 R08: 0000000000000000 R09: 0000000000000000
> R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
> R13: ffff88811802a440 R14: ffff88811802a240 R15: ffff8881132d8601
> FS:  00007f097dfaa700(0000) GS:ffff888118000000(0000) GS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00000000200001c0 CR3: 0000000103d56000 CR4: 00000000000006f0
> Call Trace:
>  <IRQ>
> amd_pmu_enable_all (arch/x86/events/amd/core.c:760 (discriminator 2))
> x86_pmu_enable (arch/x86/events/core.c:1360)
> event_sched_out (kernel/events/core.c:1191 kernel/events/core.c:1186
>     kernel/events/core.c:2346)
> __perf_remove_from_context (kernel/events/core.c:2435)
> event_function (kernel/events/core.c:259)
> remote_function (kernel/events/core.c:92 (discriminator 1)
>     kernel/events/core.c:72 (discriminator 1))
> __flush_smp_call_function_queue (./arch/x86/include/asm/jump_label.h:27
>     ./include/linux/jump_label.h:207 ./include/trace/events/csd.h:64
>     kernel/smp.c:135 kernel/smp.c:540)
> __sysvec_call_function_single (./arch/x86/include/asm/jump_label.h:27
>     ./include/linux/jump_label.h:207
>     ./arch/x86/include/asm/trace/irq_vectors.h:99 arch/x86/kernel/smp.c:272)
> sysvec_call_function_single (arch/x86/kernel/smp.c:266 (discriminator 47)
>     arch/x86/kernel/smp.c:266 (discriminator 47))
>  </IRQ>
> 
> Reported-by: syzkaller <syzkaller@googlegroups.com>
> Signed-off-by: George Kennedy <george.kennedy@oracle.com>
> ---

Without a Fixes tag it goes into perf/core.

[tip: perf/core] perf/x86/amd: Check event before enable to avoid GPF

Posted by tip-bot2 for George Kennedy 3 months, 3 weeks ago

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     866cf36bfee4fba6a492d2dcc5133f857e3446b0
Gitweb:        https://git.kernel.org/tip/866cf36bfee4fba6a492d2dcc5133f857e3446b0
Author:        George Kennedy <george.kennedy@oracle.com>
AuthorDate:    Tue, 08 Oct 2024 08:00:53 -05:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Thu, 16 Oct 2025 11:13:47 +02:00

perf/x86/amd: Check event before enable to avoid GPF

On AMD machines cpuc->events[idx] can become NULL in a subtle race
condition with NMI->throttle->x86_pmu_stop().

Check event for NULL in amd_pmu_enable_all() before enable to avoid a GPF.
This appears to be an AMD only issue.

Syzkaller reported a GPF in amd_pmu_enable_all.

INFO: NMI handler (perf_event_nmi_handler) took too long to run: 13.143
    msecs
Oops: general protection fault, probably for non-canonical address
    0xdffffc0000000034: 0000  PREEMPT SMP KASAN NOPTI
KASAN: null-ptr-deref in range [0x00000000000001a0-0x00000000000001a7]
CPU: 0 UID: 0 PID: 328415 Comm: repro_36674776 Not tainted 6.12.0-rc1-syzk
RIP: 0010:x86_pmu_enable_event (arch/x86/events/perf_event.h:1195
    arch/x86/events/core.c:1430)
RSP: 0018:ffff888118009d60 EFLAGS: 00010012
RAX: dffffc0000000000 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 0000000000000034 RSI: 0000000000000000 RDI: 00000000000001a0
RBP: 0000000000000001 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
R13: ffff88811802a440 R14: ffff88811802a240 R15: ffff8881132d8601
FS:  00007f097dfaa700(0000) GS:ffff888118000000(0000) GS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000200001c0 CR3: 0000000103d56000 CR4: 00000000000006f0
Call Trace:
 <IRQ>
amd_pmu_enable_all (arch/x86/events/amd/core.c:760 (discriminator 2))
x86_pmu_enable (arch/x86/events/core.c:1360)
event_sched_out (kernel/events/core.c:1191 kernel/events/core.c:1186
    kernel/events/core.c:2346)
__perf_remove_from_context (kernel/events/core.c:2435)
event_function (kernel/events/core.c:259)
remote_function (kernel/events/core.c:92 (discriminator 1)
    kernel/events/core.c:72 (discriminator 1))
__flush_smp_call_function_queue (./arch/x86/include/asm/jump_label.h:27
    ./include/linux/jump_label.h:207 ./include/trace/events/csd.h:64
    kernel/smp.c:135 kernel/smp.c:540)
__sysvec_call_function_single (./arch/x86/include/asm/jump_label.h:27
    ./include/linux/jump_label.h:207
    ./arch/x86/include/asm/trace/irq_vectors.h:99 arch/x86/kernel/smp.c:272)
sysvec_call_function_single (arch/x86/kernel/smp.c:266 (discriminator 47)
    arch/x86/kernel/smp.c:266 (discriminator 47))
 </IRQ>

Reported-by: syzkaller <syzkaller@googlegroups.com>
Signed-off-by: George Kennedy <george.kennedy@oracle.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/x86/events/amd/core.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index b20661b..8868f5f 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -763,7 +763,12 @@ static void amd_pmu_enable_all(int added)
 		if (!test_bit(idx, cpuc->active_mask))
 			continue;
 
-		amd_pmu_enable_event(cpuc->events[idx]);
+		/*
+		 * FIXME: cpuc->events[idx] can become NULL in a subtle race
+		 * condition with NMI->throttle->x86_pmu_stop().
+		 */
+		if (cpuc->events[idx])
+			amd_pmu_enable_event(cpuc->events[idx]);
 	}
 }