On Wed, Oct 26, 2016 at 05:12:49PM +0200, Peter Zijlstra wrote:
> On Wed, Oct 26, 2016 at 11:48:24AM +0200, Jiri Olsa wrote:
>
> > diff --git a/kernel/events/core.c b/kernel/events/core.c
> > index c6e47e97b33f..04477983945e 100644
> > --- a/kernel/events/core.c
> > +++ b/kernel/events/core.c
> > @@ -1960,6 +1960,13 @@ void perf_event_disable(struct perf_event *event)
> > }
> > EXPORT_SYMBOL_GPL(perf_event_disable);
> >
> > +void perf_event_disable_inatomic(struct perf_event *event, int kill)
> > +{
> > + event->pending_kill = kill;
> > + event->pending_disable = 1;
> > + irq_work_queue(&event->pending);
> > +}
> > +
> > static void perf_set_shadow_time(struct perf_event *event,
> > struct perf_event_context *ctx,
> > u64 tstamp)
> > @@ -7074,9 +7081,7 @@ static int __perf_event_overflow(struct perf_event
> > *event,
> > event->pending_kill = POLL_IN;
> > if (events && atomic_dec_and_test(&event->event_limit)) {
> > ret = 1;
> > - event->pending_kill = POLL_HUP;
> > - event->pending_disable = 1;
> > - irq_work_queue(&event->pending);
> > + perf_event_disable_inatomic(event, POLL_HUP);
> > }
>
> So the pending_kill stuff is independent of the disable here. No need to
> combine the two. I've change the patch as per the below.
>
> That is, pending_kill is part of pending_wakeup, not of pending_disable.
> Here we simply use both, its just that on disable we need a different
> kind of wakeup (HANGUP instead of IN).
>
> See how after ->overflow_handler() we send a wakeup if there's a
> registered signal.
ok, seems good
thanks,
jirka
