Re: [PATCH 4/6] posix-cpu-timers: Force next_expiration recalc after timer reset

[Peter Zijlstra]

On Wed, Jun 16, 2021 at 01:21:11PM +0200, Frederic Weisbecker wrote:
> On Wed, Jun 16, 2021 at 11:23:33AM +0200, Peter Zijlstra wrote:
> > On Fri, Jun 04, 2021 at 01:31:57PM +0200, Frederic Weisbecker wrote:
> > 
> > > @@ -647,8 +651,6 @@ static int posix_cpu_timer_set(struct k_itimer *timer, int timer_flags,
> > >  	if (unlikely(timer->it.cpu.firing)) {
> > >  		timer->it.cpu.firing = -1;
> > >  		ret = TIMER_RETRY;
> > > -	} else {
> > > -		cpu_timer_dequeue(ctmr);
> > >  	}
> > >  
> > >  	/*
> > > @@ -713,9 +715,13 @@ static int posix_cpu_timer_set(struct k_itimer *timer, int timer_flags,
> > >  	 * For a timer with no notification action, we don't actually
> > >  	 * arm the timer (we'll just fake it for timer_gettime).
> > >  	 */
> > > -	cpu_timer_setexpires(ctmr, new_expires);
> > > -	if (new_expires != 0 && val < new_expires) {
> > > -		arm_timer(timer, p);
> > > +	if (new_expires != 0) {
> > > +		cpu_timer_dequeue(ctmr);
> > > +		cpu_timer_setexpires(ctmr, new_expires);
> > > +		if (val < new_expires)
> > > +			arm_timer(timer, p);
> > > +	} else {
> > > +		disarm_timer(timer, p);
> > >  	}
> > >  
> > >  	unlock_task_sighand(p, &flags);
> > 
> > AFAICT there's an error path in between where you've removed
> > cpu_timer_dequeue() and added it back. This error path will now leave
> > the timer enqueued.
> 
> Ah that's the case where the timer is firing. In this case it can't be queued
> anyway. Also it's a retry path so we'll eventually dequeue it in any case
> (should it be concurrently requeued after firing).

Urgh, I see.. this code is a maze :-(