Re: [PATCH v2] sched/fair: Reschedule the cfs_rq when current is ineligible

[Chen Yu <yu.c.chen@intel.com>]

On 2024-06-13 at 20:02:37 +0800, Chunxin Zang wrote:
> 
> 
> > On Jun 13, 2024, at 19:54, Chen Yu <yu.c.chen@intel.com> wrote:
> > 
> > On 2024-06-12 at 18:39:11 +0800, Chunxin Zang wrote:
> >> 
> >> 
> >>> On Jun 7, 2024, at 13:07, Chen Yu <yu.c.chen@intel.com> wrote:
> >>> 
> >>> On 2024-05-29 at 22:18:06 +0800, Chunxin Zang wrote:
> >> The purpose of the modification is to increase preemption opportunities without breaking the
> >> RUN_TO_PARITY rule. However, it clearly introduces some additional preemptions, or perhaps
> >> there should be a check for the eligibility of the se. Also, to avoid overwriting the scheduling
> >> strategy in entity_tick, would a modification like the following be more appropriate?
> >> 
> > 
> > I wonder if we can only take care of the NO_RUN_TO_PARITY case? Something like this,
> > 
> >> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> >> index 03be0d1330a6..5e49a15bbdd3 100644
> >> --- a/kernel/sched/fair.c
> >> +++ b/kernel/sched/fair.c
> >> @@ -745,6 +745,21 @@ int entity_eligible(struct cfs_rq *cfs_rq, struct sched_entity *se)
> >>        return vruntime_eligible(cfs_rq, se->vruntime);
> >> }
> >> 
> >> +static bool check_entity_need_preempt(struct cfs_rq *cfs_rq, struct sched_entity *se)
> >> +{
> > if (sched_feat(RUN_TO_PARITY) || cfs_rq->nr_running <= 1 ||
> >     !entity_eligible(cfs_rq, se))
> > return false;
> > 
> > return true;
> > 
> > Thoughts?
> > 
> 
> This does indeed look better. In that case, do I need to make the changes this way and send
> out a version 3?

If you mean the following changes, maybe we can continue the discussion here.
This is just my 2 cents, not sure what others think of it. Anyway, I can launch some tests.

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 8a5b1ae0aa55..c0fdb25f0695 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -744,6 +744,15 @@ int entity_eligible(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	return vruntime_eligible(cfs_rq, se->vruntime);
 }
 
+static bool check_curr_preempt(struct cfs_rq *cfs_rq, struct sched_entity *curr)
+{
+	if (sched_feat(RUN_TO_PARITY) || cfs_rq->nr_running <= 1 ||
+	    !entity_eligible(cfs_rq, curr))
+		return false;
+
+	return true;
+}
+
 static u64 __update_min_vruntime(struct cfs_rq *cfs_rq, u64 vruntime)
 {
 	u64 min_vruntime = cfs_rq->min_vruntime;
@@ -5536,6 +5545,9 @@ entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
 			hrtimer_active(&rq_of(cfs_rq)->hrtick_timer))
 		return;
 #endif
+
+	if (check_curr_preempt(cfs_rq, curr))
+		resched_curr(rq_of(cfs_rq));
 }
 
 
@@ -8415,7 +8427,7 @@ static void check_preempt_wakeup_fair(struct rq *rq, struct task_struct *p, int
 	/*
 	 * XXX pick_eevdf(cfs_rq) != se ?
 	 */
-	if (pick_eevdf(cfs_rq) == pse)
+	if (check_curr_preempt(cfs_rq, se) || pick_eevdf(cfs_rq) == pse)
 		goto preempt;
 
 	return;
-- 
2.25.1