Re: [PATCH] sched/fair: sanitize vruntime of entity being migrated

[Vincent Guittot]

On Sat, 4 Mar 2023 at 09:29, Zhang Qiao <zhangqiao22@xxxxxxxxxx> wrote:
>
> Commit 829c1651e9c4 ("sched/fair: sanitize vruntime of
> entity being placed") fix an overflowing bug, but ignore
> a case that se->exec_start is reset after a migration.
>
> For fix this case, we reset the vruntime of a loog sleep
> task in migrate_task_rq_fair().

some typo:
"For fixing this case, we reset the vruntime of a long sleeping task
in migrate_task_rq_fair()."

>
> Fixes: 829c1651e9c4 ("sched/fair: sanitize vruntime of entity being placed")
> Suggested-by: Vincent Guittot <vincent.guittot@xxxxxxxxxx>
> Signed-off-by: Zhang Qiao <zhangqiao22@xxxxxxxxxx>

Your patch doesn't apply. It seems to be malformed

> ---
>  kernel/sched/fair.c | 73 ++++++++++++++++++++++++++++++++-------------
>  1 file changed, 53 insertions(+), 20 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index ff4dbbae3b10..6697462baf0f 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -4648,6 +4648,41 @@ static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se)
>  #endif
>  }
>  +static inline bool entity_is_long_sleep(struct sched_entity *se)

This extra space at the beg of the line above looks strange

> +{
> +       struct cfs_rq *cfs_rq;
> +       u64 sleep_time;
> +
> +       if (se->exec_start == 0)
> +               return false;
> +
> +       cfs_rq = cfs_rq_of(se);
> +       sleep_time = rq_clock_task(rq_of(cfs_rq)) - se->exec_start;
> +       if ((s64)sleep_time > 60LL * NSEC_PER_SEC)
> +               return true;
> +
> +       return false;
> +}
> +
> +static inline u64 sched_sleeper_credit(struct sched_entity *se)
> +{
> +       unsigned long thresh;
> +
> +       if (se_is_idle(se))
> +               thresh = sysctl_sched_min_granularity;
> +       else
> +               thresh = sysctl_sched_latency;
> +
> +       /*
> +        * Halve their sleep time's effect, to allow
> +        * for a gentler effect of sleepers:
> +        */
> +       if (sched_feat(GENTLE_FAIR_SLEEPERS))
> +               thresh >>= 1;
> +
> +       return thresh;
> +}
> +
>  static void
>  place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
>  {
> @@ -4664,23 +4699,8 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
>                 vruntime += sched_vslice(cfs_rq, se);
>         /* sleeps up to a single latency don't count. */
> -       if (!initial) {
> -               unsigned long thresh;
> -
> -               if (se_is_idle(se))
> -                       thresh = sysctl_sched_min_granularity;
> -               else
> -                       thresh = sysctl_sched_latency;
> -
> -               /*
> -                * Halve their sleep time's effect, to allow
> -                * for a gentler effect of sleepers:
> -                */
> -               if (sched_feat(GENTLE_FAIR_SLEEPERS))
> -                       thresh >>= 1;
> -
> -               vruntime -= thresh;
> -       }
> +       if (!initial)
> +               vruntime -= sched_sleeper_credit(se);
>         /*
>          * Pull vruntime of the entity being placed to the base level of
> @@ -4689,8 +4709,7 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
>          * the base as it may be too far off and the comparison may get
>          * inversed due to s64 overflow.
>          */
> -       sleep_time = rq_clock_task(rq_of(cfs_rq)) - se->exec_start;
> -       if ((s64)sleep_time > 60LL * NSEC_PER_SEC)
> +       if (entity_is_long_sleep(se))
>                 se->vruntime = vruntime;
>         else
>                 se->vruntime = max_vruntime(se->vruntime, vruntime);
> @@ -7635,7 +7654,21 @@ static void migrate_task_rq_fair(struct task_struct *p, int new_cpu)
>         if (READ_ONCE(p->__state) == TASK_WAKING) {
>                 struct cfs_rq *cfs_rq = cfs_rq_of(se);
>  -              se->vruntime -= u64_u32_load(cfs_rq->min_vruntime);
> +               /*
> +                * We determine a task whether sleep for long by checking se->exec_start,
> +                * and if it is, sanitize its vruntime at place_entity(). However, after
> +                * a migration, this detection method fails due to se->exec_start is reset.

some typo

"We determine whether a task sleeps for long by checking
se->exec_start, and if it is, we sanitize its vruntime at
place_entity(). However, after a migration, this detection method
fails due to se->exec_start being reset."

> +                *
> +                * For fix this case, we add the same check at here. For a task which has
> +                * slept for long, its vruntime should be reset cfs_rq->min_vruntime and get
> +                * a sleep credit. Because waking task's vruntime will be added cfs_rq->min_vruntime

For fixing this case, we add the same check here. For a task which has
slept for a long time, its vruntime should be reset to
cfs_rq->min_vruntime with a sleep credit.

Beside the typo and malformed patch, the proposed fix looks good to me

> +                * when enqueue, so we only need resetting the se->vruntime of waking task
> +                * to a credit at here.
> +                */
> +               if (entity_is_long_sleep(se))
> +                       se->vruntime = -sched_sleeper_credit(se);
> +               else
> +                       se->vruntime -= u64_u32_load(cfs_rq->min_vruntime);
>         }
>         if (!task_on_rq_migrating(p)) {
> --
> 2.17.1
>
> .