Re: [RFC v2] perf: Rewrite core context handling

[Peter Zijlstra]

On Mon, Aug 29, 2022 at 09:30:50AM +0530, Ravi Bangoria wrote:
> > With this, I can run 'perf test' and perf_event_tests without any error in
> > dmesg. I'll run perf fuzzer over night and see if it reports any issue.
> 
> I also ran fuzzer on Intel machine over the weekend. I see only one WARN_ON()
> hit. Otherwise system is running normal. FWIW, I was running fuzzer as normal
> user with perf_event_paranoid=0.
> 
>   WARNING: CPU: 3 PID: 2840537 at arch/x86/events/core.c:1606 x86_pmu_stop+0xd0/0x100

That's the WARN about PERF_HES_STOPPED already being set.

>   Call Trace:
>    <TASK>
>    x86_pmu_del+0x8e/0x2d0
>    ? debug_smp_processor_id+0x17/0x20
>    event_sched_out+0x10b/0x2b0
>    ? x86_pmu_del+0x5c/0x2d0
>    merge_sched_in+0x39f/0x410

And this callchain suggests this is the group_error path.

I can't immediately spot a fail there, but I'll try and stare at it
some.