[PATCH 3/4] sched/fair: Do not replace recent_used_cpu with the new target

[Mel Gorman]

On Fri, Dec 11, 2020 at 05:34:43PM +0800, Hillf Danton wrote:
> On Fri, 11 Dec 2020 09:02:28 +0000 Mel Gorman wrote:
> >On Fri, Dec 11, 2020 at 02:25:42PM +0800, Hillf Danton wrote:
> >> On Tue,  8 Dec 2020 15:35:00 +0000 Mel Gorman wrote:
> >> > @@ -6277,17 +6277,13 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)
> >> >  
> >> >  	/* Check a recently used CPU as a potential idle candidate: */
> >> >  	recent_used_cpu = p->recent_used_cpu;
> >> > +	p->recent_used_cpu = prev;
> >> >  	if (recent_used_cpu != prev &&
> >> >  	    recent_used_cpu != target &&
> >> >  	    cpus_share_cache(recent_used_cpu, target) &&
> >> >  	    (available_idle_cpu(recent_used_cpu) || sched_idle_cpu(recent_used_cpu)) &&
> >> >  	    cpumask_test_cpu(p->recent_used_cpu, p->cpus_ptr) &&
> >> 
> >> Typo? Fix it in spin if so.
> >> 
> >
> >What typo?
> 
> After your change it is prev that we check against p->cpus_ptr instead of
> the recent CPU. Wonder the point to do such a check for returning the
> recent one.

Ah... yes, this is indeed wrong. It wouldn't affect Vincent's case
that showed a problem with a hackbench configuration (which I'm still
disappointed about as it's a trade-off depending on machine and workload)
but it allows a task to run on the wrong cpu if sched_setscheduler()
was called between wakeup events.

-- 
Mel Gorman
SUSE Labs