Re: [PATCH v3] sched/rt: Do not account migrate-disabled tasks as pushable
From: Steven Rostedt
Date: Tue Nov 17 2026 - 10:26:05 EST
On Tue, 17 Nov 2026 16:20:26 +0100
Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> >
> > static void enqueue_pushable_task(struct rq *rq, struct task_struct *p)
> > {
> > + if (p->nr_cpus_allowed <= 1)
> > + return;
> > +#ifdef HAVE_RT_PUSH_IPI
> > + if (sched_feat(RT_PUSH_IPI) && is_migration_disabled(p))
> > + return;
> > +#endif
> > +
>
> if (p->nr_cpus_allowed <= 1 || is_migration_disabled(p))
> return;
>
> for both RT and DL. Sure, we might get an extra load, but is that really
> going to be noticable? Lots of scheduler code already looks at the
> migrate disable thing anyway.
>
> I don't see why it should depend on RT_PUSH_IPI, it never is actually
> migratable, and so it never makes sense to have on the pushable list.
When I took a look at this, I was thinking the same thing. That there's no
need to have the HAVE_RT_PUSH_IPI check here and always do the test.
>
> But perhaps I'm missing something; the above seems to suggest that is a
> theme today.
I'm still trying to dig myself out of the tracing patches backlog to spend
time on looking at this :-p
-- Steve