[PATCH] sched/fair: Sync load_sum with load_avg after dequeue

[PATCH] sched/fair: Sync load_sum with load_avg after dequeue

[Vincent Guittot]

commit 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
reported some inconsitencies between *_avg and *_sum.

commit 1c35b07e6d39 ("sched/fair: Ensure _sum and _avg values stay consistent")
fixed some but one remains when dequeuing load.

sync the cfs's load_sum with its load_avg after dequeuing the load of a
sched_entity.

Fixes: 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
---

I have been able to trigger a WARN on my system even with the patch
listed above. This patch fixes it.
Sachin could you test that it also fixes yours ?

kernel/sched/fair.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 11d22943753f..48fc7dfc2f66 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3037,8 +3037,9 @@ enqueue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
 static inline void
 dequeue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
+	u32 divider = get_pelt_divider(&se->avg);
 	sub_positive(&cfs_rq->avg.load_avg, se->avg.load_avg);
-	sub_positive(&cfs_rq->avg.load_sum, se_weight(se) * se->avg.load_sum);
+	cfs_rq->avg.load_sum = cfs_rq->avg.load_avg * divider;
 }
 #else
 static inline void
-- 
2.17.1

Re: [PATCH] sched/fair: Sync load_sum with load_avg after dequeue

[Sachin Sant]

> On 01-Jul-2021, at 10:48 PM, Vincent Guittot <vincent.guittot@linaro.org> wrote:
> 
> commit 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
> reported some inconsitencies between *_avg and *_sum.
> 
> commit 1c35b07e6d39 ("sched/fair: Ensure _sum and _avg values stay consistent")
> fixed some but one remains when dequeuing load.
> 
> sync the cfs's load_sum with its load_avg after dequeuing the load of a
> sched_entity.
> 
> Fixes: 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
> Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
> ---
> 
> I have been able to trigger a WARN on my system even with the patch
> listed above. This patch fixes it.
> Sachin could you test that it also fixes yours ?
> 

I ran various LTP stress tests, scheduler tests and kernel compile operation for about 5 hours.
Haven’t seen the warning during the testing.

Tested-by: Sachin Sant <sachinp@linux.vnet.ibm.com>

I have left the tests running, will let it run for few more hours.

Thanks
-Sachin

> kernel/sched/fair.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 11d22943753f..48fc7dfc2f66 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -3037,8 +3037,9 @@ enqueue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
> static inline void
> dequeue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
> {
> +	u32 divider = get_pelt_divider(&se->avg);
> 	sub_positive(&cfs_rq->avg.load_avg, se->avg.load_avg);
> -	sub_positive(&cfs_rq->avg.load_sum, se_weight(se) * se->avg.load_sum);
> +	cfs_rq->avg.load_sum = cfs_rq->avg.load_avg * divider;
> }
> #else
> static inline void
> -- 
> 2.17.1
> 

Re: [PATCH] sched/fair: Sync load_sum with load_avg after dequeue

[Vincent Guittot]

On Fri, 2 Jul 2021 at 08:16, Sachin Sant <sachinp@linux.vnet.ibm.com> wrote:
>
>
>
> > On 01-Jul-2021, at 10:48 PM, Vincent Guittot <vincent.guittot@linaro.org> wrote:
> >
> > commit 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
> > reported some inconsitencies between *_avg and *_sum.
> >
> > commit 1c35b07e6d39 ("sched/fair: Ensure _sum and _avg values stay consistent")
> > fixed some but one remains when dequeuing load.
> >
> > sync the cfs's load_sum with its load_avg after dequeuing the load of a
> > sched_entity.
> >
> > Fixes: 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
> > Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
> > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
> > ---
> >
> > I have been able to trigger a WARN on my system even with the patch
> > listed above. This patch fixes it.
> > Sachin could you test that it also fixes yours ?
> >
>
> I ran various LTP stress tests, scheduler tests and kernel compile operation for about 5 hours.
> Haven’t seen the warning during the testing.
>
> Tested-by: Sachin Sant <sachinp@linux.vnet.ibm.com>

Thanks

>
> I have left the tests running, will let it run for few more hours.
>
> Thanks
> -Sachin
>
> > kernel/sched/fair.c | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index 11d22943753f..48fc7dfc2f66 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -3037,8 +3037,9 @@ enqueue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
> > static inline void
> > dequeue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
> > {
> > +     u32 divider = get_pelt_divider(&se->avg);
> >       sub_positive(&cfs_rq->avg.load_avg, se->avg.load_avg);
> > -     sub_positive(&cfs_rq->avg.load_sum, se_weight(se) * se->avg.load_sum);
> > +     cfs_rq->avg.load_sum = cfs_rq->avg.load_avg * divider;
> > }
> > #else
> > static inline void
> > --
> > 2.17.1
> >
>

Re: [PATCH] sched/fair: Sync load_sum with load_avg after dequeue

[Odin Ugedal]

Thanks,

Reviewed-by: Odin Ugedal <odin@uged.al>

Odin

[tip: sched/urgent] sched/fair: Sync load_sum with load_avg after dequeue

[tip-bot2 for Vincent Guittot]

The following commit has been merged into the sched/urgent branch of tip:

Commit-ID:     ceb6ba45dc8074d2a1ec1117463dc94a20d4203d
Gitweb:        https://git.kernel.org/tip/ceb6ba45dc8074d2a1ec1117463dc94a20d4203d
Author:        Vincent Guittot <vincent.guittot@linaro.org>
AuthorDate:    Thu, 01 Jul 2021 19:18:37 +02:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Fri, 02 Jul 2021 15:58:23 +02:00

sched/fair: Sync load_sum with load_avg after dequeue

commit 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
reported some inconsitencies between *_avg and *_sum.

commit 1c35b07e6d39 ("sched/fair: Ensure _sum and _avg values stay consistent")
fixed some but one remains when dequeuing load.

sync the cfs's load_sum with its load_avg after dequeuing the load of a
sched_entity.

Fixes: 9e077b52d86a ("sched/pelt: Check that *_avg are null when *_sum are")
Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Odin Ugedal <odin@uged.al>
Tested-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
Link: https://lore.kernel.org/r/20210701171837.32156-1-vincent.guittot@linaro.org
---
 kernel/sched/fair.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 45edf61..1e263c9 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3037,8 +3037,9 @@ enqueue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
 static inline void
 dequeue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
+	u32 divider = get_pelt_divider(&se->avg);
 	sub_positive(&cfs_rq->avg.load_avg, se->avg.load_avg);
-	sub_positive(&cfs_rq->avg.load_sum, se_weight(se) * se->avg.load_sum);
+	cfs_rq->avg.load_sum = cfs_rq->avg.load_avg * divider;
 }
 #else
 static inline void