Re: [PATCH 05/12] sched: Move sched_class::prio_changed() into the change pattern

[Juri Lelli]

On 15/01/26 14:13, Pierre Gondois wrote:
> Hello Peter,
> 
> On 1/15/26 10:05, Peter Zijlstra wrote:
> > On Thu, Jan 15, 2026 at 09:24:31AM +0100, Peter Zijlstra wrote:
> > > On Wed, Jan 14, 2026 at 03:20:48PM +0100, Juri Lelli wrote:
> > > 
> > > > > --- a/kernel/sched/syscalls.c
> > > > > +++ b/kernel/sched/syscalls.c
> > > > > @@ -639,7 +639,7 @@ int __sched_setscheduler(struct task_str
> > > > >   		 * itself.
> > > > >   		 */
> > > > >   		newprio = rt_effective_prio(p, newprio);
> > > > > -		if (newprio == oldprio)
> > > > > +		if (newprio == oldprio && !dl_prio(newprio))
> > > > >   			queue_flags &= ~DEQUEUE_MOVE;
> > > > >   	}
> > > > We have been using (improperly?) ENQUEUE_SAVE also to know when a new
> > > > entity gets setscheduled to DEADLINE (or its parameters are changed) and
> > > > it looks like this keeps that happening with DEQUEUE_MOVE. So, from a
> > > > quick first look, it does sound good to me.
> > > If this is strictly about tasks coming into SCHED_DEADLINE there are a
> > > number of alternative options:
> > > 
> > >   - there are the sched_class::switch{ing,ed}_to() callbacks;
> > >   - there is (the fairly recent) ENQUEUE_CLASS.
> > > 
> > > Anyway, let me break up this one patch into individual bits and write
> > > changelogs. I'll stick them in queue/sched/urgent for now; hopefully
> > > Pierre can given them a spin and report back if it all sorts his
> > > problem).
> > Now live at:
> > 
> > https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/log/?h=sched/urgent
> > 
> > Please test.
> I don't see the balance_callback or the double clock update warnings
> anymore.

FWIW (as I wasn't seeing the reported issue) I had a look as well and
tested locally. Patches look good and nothing to report on the test
side.

Thanks!
Juri