Re: [PATCH 18/18 v2] sched/fair: Take into account slice in EAS

[Vincent Guittot]

On Fri, 9 Oct 2026 at 00:53, Tim Chen <tim.c.chen@xxxxxxxxxxxxxxx> wrote:
>
> On Fri, 2026-10-02 at 17:44 +0200, Vincent Guittot wrote:
> > When the cost is the same, take into account the slice of a task to try to
> > select a CPU where is will run first.
> >
> > Signed-off-by: Vincent Guittot <vincent.guittot@xxxxxxxxxx>
> > ---
> >  kernel/sched/fair.c | 13 +++++++++++--
> >  1 file changed, 11 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index 94554f165f42..d60bb6db4ce9 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -9598,8 +9598,17 @@ static int check_cpu_with_task(struct task_struct *p, int cpu)
> >   */
> >  static bool update_best_cpu(struct energy_cpu_stat *target,
> >                           struct energy_cpu_stat *min,
> > -                         int prev, struct sched_domain *sd)
> > +                         int prev, struct sched_domain *sd,
> > +                         struct task_struct *p)
> >  {
> > +     unsigned long task_slice = p->se.slice;
> > +
> > +     /*  Select the one where you can run first */
> > +     if (task_slice < get_rq_min_slice(cpu_rq(target->cpu)) &&
> > +         task_slice >= get_rq_min_slice(cpu_rq(min->cpu)))
> > +             return true;
> > +
> > +     /* Favor previous CPU */
> >       if (target->cpu == prev)
> >               return true;
>
> Hi Vincent,
>
> The slice check above comes before "Favor previous CPU", so I read the intent
> as: the CPU where the task can run first wins, even
> against prev_cpu. But the slice check can fail for prev CPU
> and we can skip to "Favor previous CPU" and pick prev CPU, even
> though the task can run on min CPU first but not
> necessarily on prev CPU first.

yes, I wil fix it

>
> Take an idle CPU X and a busy prev_cpu. Any task can
> run first on X, because an empty rq has min_slice == ULONG_MAX:
>
> - prev_cpu scanned first: when X is the target, the slice check
>   selects X.
> - X scanned first: when prev_cpu is the target, the slice check fails,
>   and "Favor previous CPU" then selects prev_cpu.
>
> In the second case the task is stacked on the busy prev_cpu even though
> it may not run first there.
>
> Perhaps something like the following is better.
>
> ---
>  kernel/sched/fair.c | 12 ++++++++----
>  1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index a47c6521a411..3fe305d690eb 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -9830,11 +9830,15 @@ static bool update_best_cpu(struct energy_cpu_stat *target,
>                             struct task_struct *p)
>  {
>         unsigned long task_slice = p->se.slice;
> +       bool target_first = task_slice < get_rq_min_slice(cpu_rq(target->cpu));
> +       bool min_first = task_slice < get_rq_min_slice(cpu_rq(min->cpu));
>
> -       /*  Select the one where you can run first */
> -       if (task_slice < get_rq_min_slice(cpu_rq(target->cpu)) &&
> -           task_slice >= get_rq_min_slice(cpu_rq(min->cpu)))
> -               return true;
> +       /*
> +        * Select the one where you can run first. Check both ways, or the
> +        * result depends on the order of the CPUs in the PD.
> +        */
> +       if (target_first != min_first)
> +               return target_first;
>
>         /* Favor previous CPU */
>         if (target->cpu == prev)
>
>
> Tim
>
> >       if (min->cpu == prev)
> > @@ -9762,7 +9771,7 @@ static int find_energy_efficient_cpu(struct task_struct *p, int prev_cpu)
> >                        */
> >                       if (target_perf < min_stat.min_perf)
> >                               find_pd_cost(pd->em_pd, target_perf, &target_stat);
> > -                     else if (!update_best_cpu(&target_stat, &min_stat, prev_cpu, sd))
> > +                     else if (!update_best_cpu(&target_stat, &min_stat, prev_cpu, sd, p))
> >                               continue;
> >
> >                       /* Save the new most efficient CPU of the PD */