[PATCH v4] sched/eevdf: Fix vprot across reweighting

[PATCH v4] sched/eevdf: Fix vprot across reweighting

[Kayra Cizmeci]

Currently in reweighting, while a tasks vruntime changes dead protection
could go alive, or alive protection could decrease, increase and go dead.
Fix it in reweight_eevdf(), and just cancel it in the sched_change_end()
path.

Fixes: 80390ead2080 ("sched/fair: Separate se->vlag from se->vprot")
Fixes: bcd74b2ffdd0 ("sched/fair: Only set slice protection at pick time")
Suggested-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Kayra Cizmeci <kayracizmeci@gmail.com>
---

Changes v3->v4:

* Change enqueue_task_fair() block so that instead of extending the protection
in sched_change_end() path just cancel the protection if it is active. (Vincent)

v3: https://lore.kernel.org/lkml/20261016084604.17999-1-kayracizmeci@gmail.com/

Compile tested only.

 kernel/sched/fair.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 0feec172fd37..61266c3a7f2f 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4892,12 +4892,15 @@ static void reweight_eevdf(struct cfs_rq *cfs_rq, struct sched_entity *se,
 	update_load_set(&se->h_load, weight);
 
 	if (on_rq) {
-		if (rel_vprot)
-			se->vprot += avruntime;
 		se->deadline += avruntime;
 		se->rel_deadline = 0;
 		se->vruntime = avruntime - se->vlag;
 
+		if (rel_vprot)
+			se->vprot += avruntime;
+		else if (curr)
+			se->vprot = se->vruntime;
+
 		if (!curr)
 			__enqueue_entity(cfs_rq, se);
 		cfs_rq->h_nr_queued++;
@@ -8202,6 +8205,13 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
 	if (!curr) {
 		reweight_eevdf(cfs_rq, se, weight, false);
 		place_entity(cfs_rq, se, flags | ENQUEUE_QUEUED);
+		/*
+		 * This happens in sched_change_end(), while cfs_rq->curr == NULL
+		 * and rq->donor == p. In this case, cancel the protection rather
+		 * than extending it.
+		 */
+		if (task_current_donor(rq, p))
+			cancel_protect_slice(se);
 		__enqueue_entity(cfs_rq, se);
 	}
 
-- 
2.53.0

Re: [PATCH v4] sched/eevdf: Fix vprot across reweighting

[Vincent Guittot]

On Fri, 16 Oct 2026 at 17:04, Kayra Cizmeci <kayracizmeci@gmail.com> wrote:
>
> Currently in reweighting, while a tasks vruntime changes dead protection
> could go alive, or alive protection could decrease, increase and go dead.
> Fix it in reweight_eevdf(), and just cancel it in the sched_change_end()
> path.
>
> Fixes: 80390ead2080 ("sched/fair: Separate se->vlag from se->vprot")
> Fixes: bcd74b2ffdd0 ("sched/fair: Only set slice protection at pick time")
> Suggested-by: Vincent Guittot <vincent.guittot@linaro.org>
> Signed-off-by: Kayra Cizmeci <kayracizmeci@gmail.com>

Besides the comment below, this looks good to me.

Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>

> ---
>
> Changes v3->v4:
>
> * Change enqueue_task_fair() block so that instead of extending the protection
> in sched_change_end() path just cancel the protection if it is active. (Vincent)
>
> v3: https://lore.kernel.org/lkml/20261016084604.17999-1-kayracizmeci@gmail.com/
>
> Compile tested only.
>
>  kernel/sched/fair.c | 14 ++++++++++++--
>  1 file changed, 12 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 0feec172fd37..61266c3a7f2f 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -4892,12 +4892,15 @@ static void reweight_eevdf(struct cfs_rq *cfs_rq, struct sched_entity *se,
>         update_load_set(&se->h_load, weight);
>
>         if (on_rq) {
> -               if (rel_vprot)
> -                       se->vprot += avruntime;
>                 se->deadline += avruntime;
>                 se->rel_deadline = 0;
>                 se->vruntime = avruntime - se->vlag;
>
> +               if (rel_vprot)
> +                       se->vprot += avruntime;
> +               else if (curr)
> +                       se->vprot = se->vruntime;
> +
>                 if (!curr)
>                         __enqueue_entity(cfs_rq, se);
>                 cfs_rq->h_nr_queued++;
> @@ -8202,6 +8205,13 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
>         if (!curr) {
>                 reweight_eevdf(cfs_rq, se, weight, false);
>                 place_entity(cfs_rq, se, flags | ENQUEUE_QUEUED);
> +               /*
> +                * This happens in sched_change_end(), while cfs_rq->curr == NULL
> +                * and rq->donor == p. In this case, cancel the protection rather
> +                * than extending it.
> +                */

what about :

/*
* During a sched_change pattern, current is dequeued and
* cfs_rq->curr becomes NULL while rq->donor == p. As all
* properties can change, cancel slice protection which is
* not reliable anymore.
*/

> +               if (task_current_donor(rq, p))
> +                       cancel_protect_slice(se);
>                 __enqueue_entity(cfs_rq, se);
>         }
>
> --
> 2.53.0
>