lkml.org 
[lkml]   [2015]   [Feb]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [RFC][PATCH 2/3] perf: Add a bit of paranoia
On Mon, Feb 02, 2015 at 06:32:32PM +0100, Peter Zijlstra wrote:
> On Mon, Feb 02, 2015 at 04:42:40PM +0100, Peter Zijlstra wrote:
> > On Mon, Feb 02, 2015 at 01:33:14AM -0500, Vince Weaver wrote:
> > > [407484.309136] ------------[ cut here ]------------
>
> > > [407484.588602] <<EOE>> <IRQ> [<ffffffff8115c28c>] perf_prepare_sample+0x2ec/0x3c0
> > > [407484.597358] [<ffffffff8115c46e>] __perf_event_overflow+0x10e/0x270
> > > [407484.604708] [<ffffffff8115c439>] ? __perf_event_overflow+0xd9/0x270
> > > [407484.612215] [<ffffffff8115c924>] ? perf_tp_event+0xc4/0x210
> > > [407484.619000] [<ffffffff8115cfe2>] ? __perf_sw_event+0x72/0x1f0
> > > [407484.625937] [<ffffffff8115c799>] ? perf_swevent_overflow+0xa9/0xe0
> > > [407484.633287] [<ffffffff8115c799>] perf_swevent_overflow+0xa9/0xe0
> > > [407484.640467] [<ffffffff8115c837>] perf_swevent_event+0x67/0x90
> > > [407484.647343] [<ffffffff8115c924>] perf_tp_event+0xc4/0x210
> > > [407484.653923] [<ffffffff810b6fa9>] ? lock_acquire+0x119/0x130
> > > [407484.660606] [<ffffffff810b3cf6>] ? perf_trace_lock_acquire+0x146/0x180
> > > [407484.668332] [<ffffffff810b594f>] ? __lock_acquire.isra.31+0x3af/0xfe0
> > > [407484.675962] [<ffffffff810b3cf6>] perf_trace_lock_acquire+0x146/0x180
> > > [407484.683490] [<ffffffff810b6fa9>] ? lock_acquire+0x119/0x130
> > > [407484.690211] [<ffffffff810b6fa9>] lock_acquire+0x119/0x130
> > > [407484.696750] [<ffffffff8115b7f5>] ? perf_event_wakeup+0x5/0xf0
> > > [407484.703640] [<ffffffff811f50ef>] ? kill_fasync+0xf/0xf0
> > > [407484.710008] [<ffffffff8115b828>] perf_event_wakeup+0x38/0xf0
> > > [407484.716798] [<ffffffff8115b7f5>] ? perf_event_wakeup+0x5/0xf0
> > > [407484.723696] [<ffffffff8115b913>] perf_pending_event+0x33/0x60
> > > [407484.730570] [<ffffffff8114cc7c>] irq_work_run_list+0x4c/0x80
> > > [407484.737392] [<ffffffff8114ccc8>] irq_work_run+0x18/0x40
> > > [407484.743765] [<ffffffff8101955f>] smp_trace_irq_work_interrupt+0x3f/0xc0
> > > [407484.751579] [<ffffffff816c01fd>] trace_irq_work_interrupt+0x6d/0x80
>
> > > [407484.799195] ---[ end trace 55752a03ec8ab979 ]---
> >
> > That looks like tail recursive fun! An irq work that raises and irq work
> > ad infinitum. Lemme see if I can squash that.. didn't we have something
> > like this before... /me goes look.
>
>
> Does this make it go away?
>
> --- a/kernel/events/core.c
> +++ b/kernel/events/core.c
> @@ -4413,6 +4413,8 @@ static void perf_pending_event(struct ir
> struct perf_event *event = container_of(entry,
> struct perf_event, pending);
>
> + int rctx = perf_swevent_get_recursion_context();
> +

hum, you should check the rctx

if (rctx == -1)
return;

also this recursion is bound to swevent_htable, should we rather add
separate ctx data for irq_work to limit the clashing with SW events?

jirka

> if (event->pending_disable) {
> event->pending_disable = 0;
> __perf_event_disable(event);
> @@ -4422,6 +4424,8 @@ static void perf_pending_event(struct ir
> event->pending_wakeup = 0;
> perf_event_wakeup(event);
> }
> +
> + perf_swevent_put_recursion_context(rctx);
> }
>
> /*


\
 
 \ /
  Last update: 2015-02-04 16:01    [W:0.085 / U:1.588 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site