Re: [GIT PULL] perf crash fix

[Frederic Weisbecker <fweisbec@gmail.com>]

On Thu, Jun 03, 2010 at 10:02:39AM +0200, Peter Zijlstra wrote:
> On Thu, 2010-06-03 at 05:13 +0200, Frederic Weisbecker wrote:
> 
> > diff --git a/kernel/perf_event.c b/kernel/perf_event.c
> > index 858f56f..b666d7d 100644
> > --- a/kernel/perf_event.c
> > +++ b/kernel/perf_event.c
> > @@ -1510,20 +1510,16 @@ do {					\
> >  	return div64_u64(dividend, divisor);
> >  }
> >  
> > -static void perf_event_stop(struct perf_event *event)
> > +static void perf_event_stop_hwevent(struct perf_event *event)
> >  {
> > -	if (!event->pmu->stop)
> > -		return event->pmu->disable(event);
> > -
> > -	return event->pmu->stop(event);
> > +	if (event->pmu->stop && !is_software_event(event))
> > +		return event->pmu->stop(event);
> >  }
> >  
> > -static int perf_event_start(struct perf_event *event)
> > +static int perf_event_start_hwevent(struct perf_event *event)
> >  {
> > -	if (!event->pmu->start)
> > -		return event->pmu->enable(event);
> > -
> > -	return event->pmu->start(event);
> > +	if (event->pmu->start && !is_software_event(event))
> > +		return event->pmu->start(event);
> >  }
> >  
> >  static void perf_adjust_period(struct perf_event *event, u64 nsec, u64 count)
> > @@ -1546,9 +1542,9 @@ static void perf_adjust_period(struct perf_event *event, u64 nsec, u64 count)
> >  
> >  	if (atomic64_read(&hwc->period_left) > 8*sample_period) {
> >  		perf_disable();
> > -		perf_event_stop(event);
> > +		perf_event_stop_hwevent(event);
> >  		atomic64_set(&hwc->period_left, 0);
> > -		perf_event_start(event);
> > +		perf_event_start_hwevent(event);
> >  		perf_enable();
> >  	}
> >  }
> 
> Urhm,. isn't is much easier to simply give the software events a NOP
> stop/start callback?


I wanted to, but I thought we could avoid two indirect calls on each
ticks and I was also afraid of breaking start/stop original semantics,
more especially the role of perf_event_stop/start


But that's about quite small details. I'm ok with your patch (the version
that also handles trace events ;)

Thanks.