Re: [PATCH 1/1] sched/fair:Reduce unnecessary check preempt in the sched tick

[Peter Zijlstra]

On Tue, Apr 13, 2021 at 09:18:42PM +0800, qianjun.kernel@xxxxxxxxx wrote:
> From: jun qian <qianjun.kernel@xxxxxxxxx>
> 
> If it has been determined that the current cpu need resched in the
> early stage of for_each_sched_entity, then there is no need to check
> preempt in the subsequent se->parent entity_tick.

Right, but does it actually do anything, except increase linecount?

> Signed-off-by: jun qian <qianjun.kernel@xxxxxxxxx>
> ---
>  kernel/sched/fair.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 1a68a0536add..c0d135100d54 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -4352,8 +4352,13 @@ check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
>  {
>  	unsigned long ideal_runtime, delta_exec;
>  	struct sched_entity *se;
> +	struct rq *rq = rq_of(cfs_rq);
>  	s64 delta;
>  
> +	/* If the TIF_NEED_RESCHED has been set, it is no need to check again */
> +	if (test_tsk_need_resched(rq->curr))
> +		return;
> +
>  	ideal_runtime = sched_slice(cfs_rq, curr);
>  	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
>  	if (delta_exec > ideal_runtime) {

Also, I think that's placed wrong; this way we can mis clear_buddies().