On Thu, 28 Nov 2024 at 10:49, Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Thu, Nov 28, 2024 at 10:27:48AM +0100, Vincent Guittot wrote:
> > Migrating a delayed dequeued task doesn't help in balancing the number
> > of runnable tasks in the system.
>
> But it can help balance the weight; furthermore, by moving them to a
> lighter queue, they'll get picked sooner and disappear sooner.
When groups are not overloaded, we don't compare load but only running
tasks t balance them across cpus
It's only when both src and dst groups are overloaded that we look at
the load and the weight
>
> Perhaps make it: p->se.sched_delayed && !env->sd->nr_balance_failed ?
So we could take into account which type of migration with
env->migration_type == migrate_load
In this case, migrating a delayed dequeue task would help
>
> > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
> > ---
> > 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 1b4f1b610543..9d80f3a61082 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -9405,11 +9405,15 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
> >
> > /*
> > * We do not migrate tasks that are:
> > - * 1) throttled_lb_pair, or
> > - * 2) cannot be migrated to this CPU due to cpus_ptr, or
> > - * 3) running (obviously), or
> > - * 4) are cache-hot on their current CPU.
> > + * 1) delayed dequeued, or
> > + * 2) throttled_lb_pair, or
> > + * 3) cannot be migrated to this CPU due to cpus_ptr, or
> > + * 4) running (obviously), or
> > + * 5) are cache-hot on their current CPU.
> > */
> > + if (p->se.sched_delayed)
> > + return 0;
> > +
> > if (throttled_lb_pair(task_group(p), env->src_cpu, env->dst_cpu))
> > return 0;
> >
> > --
> > 2.43.0
> >