Re: [RFC PATCH 2/5] clockevents: Add generic timer broadcast receiver

[Mark Rutland <mark.rutland@arm.com>]

On Tue, Dec 18, 2012 at 10:17:11PM +0000, Stephen Boyd wrote:
> On 12/18/12 04:06, Mark Rutland wrote:
> > diff --git a/kernel/time/tick-broadcast.c b/kernel/time/tick-broadcast.c
> > index f113755..c2dd022 100644
> > --- a/kernel/time/tick-broadcast.c
> > +++ b/kernel/time/tick-broadcast.c
> > @@ -125,6 +125,21 @@ int tick_device_uses_broadcast(struct clock_event_device *dev, int cpu)
> >  	return ret;
> >  }
> >  
> > +int tick_receive_broadcast(void)
> > +{
> > +	struct tick_device *td = this_cpu_ptr(&tick_cpu_device);
> > +	struct clock_event_device *evt = td->evtdev;
> > +
> > +	if (!evt)
> > +		return -ENODEV;
> > +
> > +	if (!evt->event_handler)
> > +		return -EINVAL;
> > +
> 
> Does this actually happen? It's not obvious to me how it does.

Taking a look at it, I don't think it does. Any tick_device's
clock_event_device should have an event_handler. I'll drop this check.

A tick device might not have a clock_event_device registered in some cases --
the x86 lapic timer has to deal with interrupts pending across a kexec, where
interrupts are enabled before local timers are registered. I'm not sure how
common this problem might be though, and whether it should be dealt with here
or elsewhere.

> > +	evt->event_handler(evt);
> > +	return 0;
> > +}
> > +
> >  /*
> >   * Broadcast the event to the cpus, which are set in the mask (mangled).
> >   */
> 
> 
> -- 
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
> hosted by The Linux Foundation
> 

Thanks,
Mark.