lkml.org 
[lkml]   [2014]   [Feb]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Perf Oops on 3.14-rc2
On Wed, Feb 19, 2014 at 04:28:19PM +0000, Peter Zijlstra wrote:
> On Tue, Feb 18, 2014 at 10:18:31AM +0000, Will Deacon wrote:
> > On Mon, Feb 10, 2014 at 10:17:59PM +0000, Drew Richardson wrote:
> > > While adding CPU on/offlining support during perf captures I get an
> > > Oops both on ARM as well as my desktop x86_64. Below is a small
> > > program that duplicates the issue.
> >
> > [...]
> >
> > FWIW I can reproduce this easily with -rc3 on my x86 laptop running
> > hackbench in parallel with a tweaked version of your test (using
> > _SC_NPROCESSORS_ONLN instead of _SC_NPROCESSORS_CONF and hotplugging off
> > both CPU2 and CPU3).
> >
>
> OK; found it, or at least, I stopped being able to make my box explode.
>
> Drew, can you try the below and confirm? Its still got all the debug goo
> in too, but *shees* took me long enough.

This works great for me. No Oops or garbage data when a core is
offlined.

> + raw_spin_lock(&ctx->lock);
> + ctx_sched_out(ctx, cpuctx, EVENT_ALL);
> +
> + list_for_each_entry(event, &ctx->event_list, event_entry)
> + event->state = PERF_EVENT_STATE_off;
> +
> + raw_spin_unlock(&ctx->lock);

I changed PERF_EVENT_STATE_off to PERF_EVENT_STATE_OFF.

Drew


\
 
 \ /
  Last update: 2014-02-19 21:21    [W:0.067 / U:0.864 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site