Re: [PATCH v3 2/2] sched: update the rq->avg_idle when a task is moved to an idle CPU

[K Prateek Nayak <kprateek.nayak@amd.com>]

Hello Huang Shijie,

On 11/27/2025 2:44 PM, Huang Shijie wrote:
>  void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
>  {
> +	int delayed = p->se.sched_delayed;
> +
>  	if (!(flags & ENQUEUE_NOCLOCK))
>  		update_rq_clock(rq);
>  
> @@ -2100,6 +2117,13 @@ void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
>  
>  	if (sched_core_enabled(rq))
>  		sched_core_enqueue(rq, p);
> +
> +	if (delayed) {
> +		if (entity_eligible(cfs_rq_of(&p->se), &p->se))
> +			update_rq_avg_idle(rq);

Question: Why do we want to treat the delayed case like this?

If entity is not eligible, we want to consider that it hasn't
even gone through a wakeup? Wouldn't this lead to the next
wakeup seeing rq->idle_stamp to be non-zero and inaccurately
account more idle time?

Also if we've done newidle balance and the rq->idle_stamp is
set, we cannot have delayed tasks since pick_next_task() would
have dequeued all delayed tasks before reaching newidle
balance.

Just doing a update_rq_avg_idle() unconditionally should be
fine.

> +	} else {
> +		update_rq_avg_idle(rq);
> +	}
>  }
>  
>  /*
-- 
Thanks and Regards,
Prateek