[PATCH] nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt

[PATCH] nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt

[Nicolas Saenz Julienne]

dequeue_task_rt() only decrements 'rt_rq->rt_nr_running' after having
called sched_update_tick_dependency() preventing it from re-enabling the
tick on systems that no longer have pending SCHED_RT tasks but have
multiple runnable SCHED_OTHER tasks:

  dequeue_task_rt()
    dequeue_rt_entity()
      dequeue_rt_stack()
        dequeue_top_rt_rq()
	  sub_nr_running()	// decrements rq->nr_running
	    sched_update_tick_dependency()
	      sched_can_stop_tick()	// checks rq->rt.rt_nr_running,
	      ...
        __dequeue_rt_entity()
          dec_rt_tasks()	// decrements rq->rt.rt_nr_running
	  ...

Every other scheduler class performs the operation in the opposite
order, and sched_update_tick_dependency() expects the values to be
updated as such. So avoid the misbehaviour by inverting the order in
which the above operations are performed in the RT scheduler.

Fixes: 76d92ac305f2 ("sched: Migrate sched to use new tick dependency mask model")
Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
---
 kernel/sched/rt.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 8c9ed96648409..55f39c8f42032 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -480,7 +480,7 @@ static inline void rt_queue_push_tasks(struct rq *rq)
 #endif /* CONFIG_SMP */
 
 static void enqueue_top_rt_rq(struct rt_rq *rt_rq);
-static void dequeue_top_rt_rq(struct rt_rq *rt_rq);
+static void dequeue_top_rt_rq(struct rt_rq *rt_rq, unsigned int count);
 
 static inline int on_rt_rq(struct sched_rt_entity *rt_se)
 {
@@ -601,7 +601,7 @@ static void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
 	rt_se = rt_rq->tg->rt_se[cpu];
 
 	if (!rt_se) {
-		dequeue_top_rt_rq(rt_rq);
+		dequeue_top_rt_rq(rt_rq, rt_rq->rt_nr_running);
 		/* Kick cpufreq (see the comment in kernel/sched/sched.h). */
 		cpufreq_update_util(rq_of_rt_rq(rt_rq), 0);
 	}
@@ -687,7 +687,7 @@ static inline void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
 
 static inline void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
 {
-	dequeue_top_rt_rq(rt_rq);
+	dequeue_top_rt_rq(rt_rq, rt_rq->rt_nr_running);
 }
 
 static inline int rt_rq_throttled(struct rt_rq *rt_rq)
@@ -1089,7 +1089,7 @@ static void update_curr_rt(struct rq *rq)
 }
 
 static void
-dequeue_top_rt_rq(struct rt_rq *rt_rq)
+dequeue_top_rt_rq(struct rt_rq *rt_rq, unsigned int count)
 {
 	struct rq *rq = rq_of_rt_rq(rt_rq);
 
@@ -1100,7 +1100,7 @@ dequeue_top_rt_rq(struct rt_rq *rt_rq)
 
 	BUG_ON(!rq->nr_running);
 
-	sub_nr_running(rq, rt_rq->rt_nr_running);
+	sub_nr_running(rq, count);
 	rt_rq->rt_queued = 0;
 
 }
@@ -1486,18 +1486,21 @@ static void __dequeue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flag
 static void dequeue_rt_stack(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct sched_rt_entity *back = NULL;
+	unsigned int rt_nr_running;
 
 	for_each_sched_rt_entity(rt_se) {
 		rt_se->back = back;
 		back = rt_se;
 	}
 
-	dequeue_top_rt_rq(rt_rq_of_se(back));
+	rt_nr_running = rt_rq_of_se(back)->rt_nr_running;
 
 	for (rt_se = back; rt_se; rt_se = rt_se->back) {
 		if (on_rt_rq(rt_se))
 			__dequeue_rt_entity(rt_se, flags);
 	}
+
+	dequeue_top_rt_rq(rt_rq_of_se(back), rt_nr_running);
 }
 
 static void enqueue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
-- 
2.36.1

Re: [PATCH] nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt

[Valentin Schneider]

On 28/06/22 11:22, Nicolas Saenz Julienne wrote:
> dequeue_task_rt() only decrements 'rt_rq->rt_nr_running' after having
> called sched_update_tick_dependency() preventing it from re-enabling the
> tick on systems that no longer have pending SCHED_RT tasks but have
> multiple runnable SCHED_OTHER tasks:
>
>   dequeue_task_rt()
>     dequeue_rt_entity()
>       dequeue_rt_stack()
>         dequeue_top_rt_rq()
>         sub_nr_running()	// decrements rq->nr_running
>           sched_update_tick_dependency()
>             sched_can_stop_tick()	// checks rq->rt.rt_nr_running,
>             ...
>         __dequeue_rt_entity()
>           dec_rt_tasks()	// decrements rq->rt.rt_nr_running
>         ...
>
> Every other scheduler class performs the operation in the opposite
> order, and sched_update_tick_dependency() expects the values to be
> updated as such. So avoid the misbehaviour by inverting the order in
> which the above operations are performed in the RT scheduler.
>

I can't see anything wrong with your approach, though I did have to spend
some time re-learning RT_GROUP_SCHED. The designated Fixes: commit looks
about right too.

> Fixes: 76d92ac305f2 ("sched: Migrate sched to use new tick dependency mask model")
> Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>

Reviewed-by: Valentin Schneider <vschneid@redhat.com>

Re: [PATCH] nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt

[Nicolas Saenz Julienne]

On Fri, 2022-07-01 at 12:25 +0100, Valentin Schneider wrote:
> On 28/06/22 11:22, Nicolas Saenz Julienne wrote:
> > dequeue_task_rt() only decrements 'rt_rq->rt_nr_running' after having
> > called sched_update_tick_dependency() preventing it from re-enabling the
> > tick on systems that no longer have pending SCHED_RT tasks but have
> > multiple runnable SCHED_OTHER tasks:
> > 
> >   dequeue_task_rt()
> >     dequeue_rt_entity()
> >       dequeue_rt_stack()
> >         dequeue_top_rt_rq()
> >         sub_nr_running()	// decrements rq->nr_running
> >           sched_update_tick_dependency()
> >             sched_can_stop_tick()	// checks rq->rt.rt_nr_running,
> >             ...
> >         __dequeue_rt_entity()
> >           dec_rt_tasks()	// decrements rq->rt.rt_nr_running
> >         ...
> > 
> > Every other scheduler class performs the operation in the opposite
> > order, and sched_update_tick_dependency() expects the values to be
> > updated as such. So avoid the misbehaviour by inverting the order in
> > which the above operations are performed in the RT scheduler.
> > 
> 
> I can't see anything wrong with your approach, though I did have to spend
> some time re-learning RT_GROUP_SCHED. The designated Fixes: commit looks
> about right too.
> 
> > Fixes: 76d92ac305f2 ("sched: Migrate sched to use new tick dependency mask model")
> > Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
> 
> Reviewed-by: Valentin Schneider <vschneid@redhat.com>

Thanks!

-- 
Nicolás Sáenz

Re: [PATCH] nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt

[Nicolas Saenz Julienne]

On Tue, 2022-06-28 at 11:22 +0200, Nicolas Saenz Julienne wrote:
> dequeue_task_rt() only decrements 'rt_rq->rt_nr_running' after having
> called sched_update_tick_dependency() preventing it from re-enabling the
> tick on systems that no longer have pending SCHED_RT tasks but have
> multiple runnable SCHED_OTHER tasks:
> 
>   dequeue_task_rt()
>     dequeue_rt_entity()
>       dequeue_rt_stack()
>         dequeue_top_rt_rq()
> 	  sub_nr_running()	// decrements rq->nr_running
> 	    sched_update_tick_dependency()
> 	      sched_can_stop_tick()	// checks rq->rt.rt_nr_running,
> 	      ...
>         __dequeue_rt_entity()
>           dec_rt_tasks()	// decrements rq->rt.rt_nr_running
> 	  ...
> 
> Every other scheduler class performs the operation in the opposite
> order, and sched_update_tick_dependency() expects the values to be
> updated as such. So avoid the misbehaviour by inverting the order in
> which the above operations are performed in the RT scheduler.
> 
> Fixes: 76d92ac305f2 ("sched: Migrate sched to use new tick dependency mask model")
> Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
> ---

Ping :)

-- 
Nicolás Sáenz

Re: [PATCH] nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt

[Phil Auld]

On Tue, Jun 28, 2022 at 11:22:59AM +0200 Nicolas Saenz Julienne wrote:
> dequeue_task_rt() only decrements 'rt_rq->rt_nr_running' after having
> called sched_update_tick_dependency() preventing it from re-enabling the
> tick on systems that no longer have pending SCHED_RT tasks but have
> multiple runnable SCHED_OTHER tasks:
> 
>   dequeue_task_rt()
>     dequeue_rt_entity()
>       dequeue_rt_stack()
>         dequeue_top_rt_rq()
> 	  sub_nr_running()	// decrements rq->nr_running
> 	    sched_update_tick_dependency()
> 	      sched_can_stop_tick()	// checks rq->rt.rt_nr_running,
> 	      ...
>         __dequeue_rt_entity()
>           dec_rt_tasks()	// decrements rq->rt.rt_nr_running
> 	  ...
> 
> Every other scheduler class performs the operation in the opposite
> order, and sched_update_tick_dependency() expects the values to be
> updated as such. So avoid the misbehaviour by inverting the order in
> which the above operations are performed in the RT scheduler.
> 
> Fixes: 76d92ac305f2 ("sched: Migrate sched to use new tick dependency mask model")
> Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>

That's been around for a while. This change looks right and makes sense
to me.

Thanks,
Phil

Reviewed-by: Phil Auld <pauld@redhat.com>

-- 

Re: [PATCH] nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt

[Peter Zijlstra]

On Wed, Jul 13, 2022 at 07:18:42PM +0200, Nicolas Saenz Julienne wrote:
> On Tue, 2022-06-28 at 11:22 +0200, Nicolas Saenz Julienne wrote:
> > dequeue_task_rt() only decrements 'rt_rq->rt_nr_running' after having
> > called sched_update_tick_dependency() preventing it from re-enabling the
> > tick on systems that no longer have pending SCHED_RT tasks but have
> > multiple runnable SCHED_OTHER tasks:
> > 
> >   dequeue_task_rt()
> >     dequeue_rt_entity()
> >       dequeue_rt_stack()
> >         dequeue_top_rt_rq()
> > 	  sub_nr_running()	// decrements rq->nr_running
> > 	    sched_update_tick_dependency()
> > 	      sched_can_stop_tick()	// checks rq->rt.rt_nr_running,
> > 	      ...
> >         __dequeue_rt_entity()
> >           dec_rt_tasks()	// decrements rq->rt.rt_nr_running
> > 	  ...
> > 
> > Every other scheduler class performs the operation in the opposite
> > order, and sched_update_tick_dependency() expects the values to be
> > updated as such. So avoid the misbehaviour by inverting the order in
> > which the above operations are performed in the RT scheduler.
> > 
> > Fixes: 76d92ac305f2 ("sched: Migrate sched to use new tick dependency mask model")
> > Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
> > ---
> 
> Ping :)

Thanks! got stuck in the retbleed backlog :/

[tip: sched/core] nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt()

[tip-bot2 for Nicolas Saenz Julienne]

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

Commit-ID:     5c66d1b9b30f737fcef85a0b75bfe0590e16b62a
Gitweb:        https://git.kernel.org/tip/5c66d1b9b30f737fcef85a0b75bfe0590e16b62a
Author:        Nicolas Saenz Julienne <nsaenzju@redhat.com>
AuthorDate:    Tue, 28 Jun 2022 11:22:59 +02:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Thu, 21 Jul 2022 10:39:38 +02:00

nohz/full, sched/rt: Fix missed tick-reenabling bug in dequeue_task_rt()

dequeue_task_rt() only decrements 'rt_rq->rt_nr_running' after having
called sched_update_tick_dependency() preventing it from re-enabling the
tick on systems that no longer have pending SCHED_RT tasks but have
multiple runnable SCHED_OTHER tasks:

  dequeue_task_rt()
    dequeue_rt_entity()
      dequeue_rt_stack()
        dequeue_top_rt_rq()
	  sub_nr_running()	// decrements rq->nr_running
	    sched_update_tick_dependency()
	      sched_can_stop_tick()	// checks rq->rt.rt_nr_running,
	      ...
        __dequeue_rt_entity()
          dec_rt_tasks()	// decrements rq->rt.rt_nr_running
	  ...

Every other scheduler class performs the operation in the opposite
order, and sched_update_tick_dependency() expects the values to be
updated as such. So avoid the misbehaviour by inverting the order in
which the above operations are performed in the RT scheduler.

Fixes: 76d92ac305f2 ("sched: Migrate sched to use new tick dependency mask model")
Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Reviewed-by: Phil Auld <pauld@redhat.com>
Link: https://lore.kernel.org/r/20220628092259.330171-1-nsaenzju@redhat.com
---
 kernel/sched/rt.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 8c9ed96..55f39c8 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -480,7 +480,7 @@ static inline void rt_queue_push_tasks(struct rq *rq)
 #endif /* CONFIG_SMP */
 
 static void enqueue_top_rt_rq(struct rt_rq *rt_rq);
-static void dequeue_top_rt_rq(struct rt_rq *rt_rq);
+static void dequeue_top_rt_rq(struct rt_rq *rt_rq, unsigned int count);
 
 static inline int on_rt_rq(struct sched_rt_entity *rt_se)
 {
@@ -601,7 +601,7 @@ static void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
 	rt_se = rt_rq->tg->rt_se[cpu];
 
 	if (!rt_se) {
-		dequeue_top_rt_rq(rt_rq);
+		dequeue_top_rt_rq(rt_rq, rt_rq->rt_nr_running);
 		/* Kick cpufreq (see the comment in kernel/sched/sched.h). */
 		cpufreq_update_util(rq_of_rt_rq(rt_rq), 0);
 	}
@@ -687,7 +687,7 @@ static inline void sched_rt_rq_enqueue(struct rt_rq *rt_rq)
 
 static inline void sched_rt_rq_dequeue(struct rt_rq *rt_rq)
 {
-	dequeue_top_rt_rq(rt_rq);
+	dequeue_top_rt_rq(rt_rq, rt_rq->rt_nr_running);
 }
 
 static inline int rt_rq_throttled(struct rt_rq *rt_rq)
@@ -1089,7 +1089,7 @@ static void update_curr_rt(struct rq *rq)
 }
 
 static void
-dequeue_top_rt_rq(struct rt_rq *rt_rq)
+dequeue_top_rt_rq(struct rt_rq *rt_rq, unsigned int count)
 {
 	struct rq *rq = rq_of_rt_rq(rt_rq);
 
@@ -1100,7 +1100,7 @@ dequeue_top_rt_rq(struct rt_rq *rt_rq)
 
 	BUG_ON(!rq->nr_running);
 
-	sub_nr_running(rq, rt_rq->rt_nr_running);
+	sub_nr_running(rq, count);
 	rt_rq->rt_queued = 0;
 
 }
@@ -1486,18 +1486,21 @@ static void __dequeue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flag
 static void dequeue_rt_stack(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct sched_rt_entity *back = NULL;
+	unsigned int rt_nr_running;
 
 	for_each_sched_rt_entity(rt_se) {
 		rt_se->back = back;
 		back = rt_se;
 	}
 
-	dequeue_top_rt_rq(rt_rq_of_se(back));
+	rt_nr_running = rt_rq_of_se(back)->rt_nr_running;
 
 	for (rt_se = back; rt_se; rt_se = rt_se->back) {
 		if (on_rt_rq(rt_se))
 			__dequeue_rt_entity(rt_se, flags);
 	}
+
+	dequeue_top_rt_rq(rt_rq_of_se(back), rt_nr_running);
 }
 
 static void enqueue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)