Re: [PATCH 3/3] PM / Sleep: Make __pm_stay_awake() delete wakeup source timers

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Friday, February 17, 2012, Arve Hjønnevåg wrote:
> 2012/2/16 Rafael J. Wysocki <rjw@sisk.pl>:
> > On Thursday, February 16, 2012, Arve Hjønnevåg wrote:
> >> 2012/2/15 Rafael J. Wysocki <rjw@sisk.pl>:
> >> ...
> >> > --- linux.orig/drivers/base/power/wakeup.c
> >> > +++ linux/drivers/base/power/wakeup.c
> >> > @@ -365,9 +365,15 @@ void __pm_stay_awake(struct wakeup_sourc
> >> >                return;
> >> >
> >> >        spin_lock_irqsave(&ws->lock, flags);
> >> > +
> >> > +       del_timer(&ws->timer);
> >> > +       ws->timer_expires = 0;
> >>
> >> timer_expires gets overwritten in wakeup_source_activate,
> >
> > I actually don't remember why it does that and with the $subject changes
> > it's just pointless.
> >
> >> so __pm_relax followed by __pm_stay_awake is still not safe.
> >
> > Yes, I overlooked that timer_expires modification.  Updated patch follows.
> >
> >> ...
> >> > @@ -529,8 +535,12 @@ void __pm_wakeup_event(struct wakeup_sou
> >> >        spin_lock_irqsave(&ws->lock, flags);
> >> >
> >> >        ws->event_count++;
> >> > -       if (!ws->active)
> >> > +       if (ws->active) {
> >> > +               if (!ws->timer_expires)
> >> > +                       goto unlock;
> >> > +       } else {
> >> >                wakeup_source_activate(ws);
> >> > +       }
> >> >
> >> >        if (!msec) {
> >> >                wakeup_source_deactivate(ws);
> >> >
> >>
> >> I suggest dropping this and adding:
> >
> > Well, what exactly would you like to drop?  The above proposed changes I guess?
> 
> Yes.
> 
> >
> >> -       if (time_after(expires, ws->timer_expires)) {
> >> +       if (!ws->timer_expires || time_after(expires, ws->timer_expires)) {
> >
> > I've tried to follow your suggestion, so that __pm_wakeup_event() always
> > sets the timer, if msec is positive, or deactivates the wakeup source, if
> > msec is 0.  Please let me know if that's what you wanted. :-)
> >
> 
> Yes. I should be able to replace a wake_lock with a single wakeup_source now.

Good, thanks for the confirmation!

Rafael