Re: [tip:perf/urgent] perf_event, x86: Fix 'perf sched record' crashing the machine

[Chris Malley <mail@chrismalley.co.uk>]

2009/9/23 tip-bot for Peter Zijlstra <peterz@infradead.org>:
> Commit-ID:  7d42896628202a551ad1107697cd215dc5fca099
> Gitweb:     http://git.kernel.org/tip/7d42896628202a551ad1107697cd215dc5fca099
> Author:     Peter Zijlstra <peterz@infradead.org>
> AuthorDate: Wed, 23 Sep 2009 11:03:37 +0200
> Committer:  Ingo Molnar <mingo@elte.hu>
> CommitDate: Wed, 23 Sep 2009 11:25:56 +0200
>
> perf_event, x86: Fix 'perf sched record' crashing the machine
>
> Chris Malley reported that 'perf sched record' sometimes
> crashes his box with:
>
> [  389.272175] BUG: unable to handle kernel paging request at ffffb300
> [  389.272294] IP: [<c011b0bd>] default_send_IPI_self+0x1d/0x50
> [  389.272366] *pde = 0073f067 *pte = 00000000
> [  389.274708] Call Trace:
> [  389.274752]  [<c010e3b4>] ?  set_perf_event_pending+0x14/0x20
> [  389.274801]  [<c01b9751>] ?  perf_output_unlock+0x121/0x1a0
> [  389.274848]  [<c01b981a>] ? perf_output_end+0x4a/0x70
> [  389.274893]  [<c01ba690>] ?  __perf_event_overflow+0x240/0x2f0
> [  389.274942]  [<c030963e>] ? atomic64_cmpxchg+0x1e/0x30
> [  389.274988]  [<c01ba8f4>] ?  perf_swevent_ctx_event+0x1b4/0x1c0
> [  389.275035]  [<c01ba773>] ?  perf_swevent_ctx_event+0x33/0x1c0
> [  389.275081]  [<c01ba9a7>] ? do_perf_sw_event+0xa7/0x160
> [  389.275127]  [<c01baae2>] ? perf_tp_event+0x82/0xa0
> [  389.275174]  [<c012e9c6>] ?  ftrace_profile_sched_stat_runtime+0xe6/0x120
> [  389.275224]  [<c012e8e0>] ?  ftrace_profile_sched_stat_runtime+0x0/0x120
> [  389.275273]  [<c013c85a>] ? update_curr+0x18a/0x230
> [  389.275318]  [<c013cdc5>] ?  put_prev_task_fair+0x155/0x160
> [  389.275366]  [<c01618b5>] ? sched_clock_cpu+0xd5/0x110
> [  389.275413]  [<c04e7525>] ? _spin_lock_irq+0x45/0x50
> [  389.275458]  [<c04e424e>] ? schedule+0x20e/0xb10
>
> The problem is that the box has no lapic enabled:
>
>  [    0.042445] Local APIC not detected. Using dummy APIC emulation.
>
> The below seems like the best fix. We disabled all lapic bits, except
> the self-IPI-resend logic.
>
> Reported-by: Chris Malley <mail@chrismalley.co.uk>
> Signed-off-by: Peter Zijlstra <peterz@infradead.org>
> Cc: Cyrill Gorcunov <gorcunov@gmail.com>
> Cc: Frederic Weisbecker <fweisbec@gmail.com>
> LKML-Reference: <7863dc4c0909221409v7893bfd3o4b590d5951a233ba@mail.gmail.com>
> Signed-off-by: Ingo Molnar <mingo@elte.hu>
>
>
> ---
>  arch/x86/kernel/cpu/perf_event.c |    3 +++
>  1 files changed, 3 insertions(+), 0 deletions(-)
>
> diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
> index a3c7adb..b5801c3 100644
> --- a/arch/x86/kernel/cpu/perf_event.c
> +++ b/arch/x86/kernel/cpu/perf_event.c
> @@ -1790,6 +1790,9 @@ void smp_perf_pending_interrupt(struct pt_regs *regs)
>  void set_perf_event_pending(void)
>  {
>  #ifdef CONFIG_X86_LOCAL_APIC
> +       if (!x86_pmu.apic || !x86_pmu_initialized())
> +               return;
> +
>        apic->send_IPI_self(LOCAL_PENDING_VECTOR);
>  #endif
>  }
>

Confirmed, this fixes the problem, so FWIW:

Tested-by: Chris Malley <mail@chrismalley.co.uk>