Re: [RFC][PATCH 2/3] PM / sleep: Mechanism to avoid resuming runtime-suspended devices unnecessarily

["Rafael J. Wysocki" <rjw@rjwysocki.net>]

On Tuesday, May 13, 2014 11:12:28 AM Alan Stern wrote:
> On Tue, 13 May 2014, Rafael J. Wysocki wrote:
> 
> > > > +	dev->power.direct_complete = ret > 0 && state.event == PM_EVENT_SUSPEND
> > > > +					&& pm_runtime_suspended(dev);
> > > 
> > > Shouldn't the flag be set under the spinlock?
> > 
> > I guess you're worried about runtime PM flags being modified in parallel to
> > this?  But we've just done the barrier a while ago, so is that still a concern
> > here?
> 
> A wakeup request from the hardware could cause a runtime resume to 
> occur at this time.  The barrier wouldn't prevent that.
> 
> It's unlikely, I agree, but not impossible.

Yeah, I didn't think about that.

But that also can occur in __device_suspend(), after we've checked the flag
and decided not to invoke the ->suspend() callback, right?  So moving the
check in there doesn't help much I'd say.  It closes the race window, but
that's it.

That means that the whole approach based on ->prepare() is problematic
unless we somehow mix it with disabling runtime PM.

Rafael