sched/fair: fixes in presence of lot of sched_idle tasks

[PATCH 2/4] sched/fair: cleanup loop_max and loop_break

Posted by Vincent Guittot 3 years, 7 months ago

sched_nr_migrate_break is set to a fix value and never changes so we can
replace it by a define SCHED_NR_MIGRATE_BREAK.

Also, we adjust SCHED_NR_MIGRATE_BREAK to be aligned with the init value
of sysctl_sched_nr_migrate which can be init to different values.

Then, use SCHED_NR_MIGRATE_BREAK to init sysctl_sched_nr_migrate.

The behavior stays unchanged unless you modify sysctl_sched_nr_migrate
trough debugfs.

Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
---
 kernel/sched/core.c  |  6 +-----
 kernel/sched/fair.c  | 11 ++++-------
 kernel/sched/sched.h |  6 ++++++
 3 files changed, 11 insertions(+), 12 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 64c08993221b..a21e817bdd1c 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -142,11 +142,7 @@ __read_mostly int sysctl_resched_latency_warn_once = 1;
  * Number of tasks to iterate in a single balance run.
  * Limited because this is done with IRQs disabled.
  */
-#ifdef CONFIG_PREEMPT_RT
-const_debug unsigned int sysctl_sched_nr_migrate = 8;
-#else
-const_debug unsigned int sysctl_sched_nr_migrate = 32;
-#endif
+const_debug unsigned int sysctl_sched_nr_migrate = SCHED_NR_MIGRATE_BREAK;
 
 __read_mostly int scheduler_running;
 
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 02b7b808e186..6972a1a29a48 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -8012,8 +8012,6 @@ static struct task_struct *detach_one_task(struct lb_env *env)
 	return NULL;
 }
 
-static const unsigned int sched_nr_migrate_break = 32;
-
 /*
  * detach_tasks() -- tries to detach up to imbalance load/util/tasks from
  * busiest_rq, as part of a balancing operation within domain "sd".
@@ -8062,7 +8060,7 @@ static int detach_tasks(struct lb_env *env)
 
 		/* take a breather every nr_migrate tasks */
 		if (env->loop > env->loop_break) {
-			env->loop_break += sched_nr_migrate_break;
+			env->loop_break += SCHED_NR_MIGRATE_BREAK;
 			env->flags |= LBF_NEED_BREAK;
 			break;
 		}
@@ -10103,14 +10101,13 @@ static int load_balance(int this_cpu, struct rq *this_rq,
 	struct rq *busiest;
 	struct rq_flags rf;
 	struct cpumask *cpus = this_cpu_cpumask_var_ptr(load_balance_mask);
-
 	struct lb_env env = {
 		.sd		= sd,
 		.dst_cpu	= this_cpu,
 		.dst_rq		= this_rq,
 		.dst_grpmask    = sched_group_span(sd->groups),
 		.idle		= idle,
-		.loop_break	= sched_nr_migrate_break,
+		.loop_break	= SCHED_NR_MIGRATE_BREAK,
 		.cpus		= cpus,
 		.fbq_type	= all,
 		.tasks		= LIST_HEAD_INIT(env.tasks),
@@ -10219,7 +10216,7 @@ static int load_balance(int this_cpu, struct rq *this_rq,
 			env.dst_cpu	 = env.new_dst_cpu;
 			env.flags	&= ~LBF_DST_PINNED;
 			env.loop	 = 0;
-			env.loop_break	 = sched_nr_migrate_break;
+			env.loop_break	 = SCHED_NR_MIGRATE_BREAK;
 
 			/*
 			 * Go back to "more_balance" rather than "redo" since we
@@ -10251,7 +10248,7 @@ static int load_balance(int this_cpu, struct rq *this_rq,
 			 */
 			if (!cpumask_subset(cpus, env.dst_grpmask)) {
 				env.loop = 0;
-				env.loop_break = sched_nr_migrate_break;
+				env.loop_break = SCHED_NR_MIGRATE_BREAK;
 				goto redo;
 			}
 			goto out_all_pinned;
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 3ccd35c22f0f..d5cfd1b5bfe9 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2423,6 +2423,12 @@ extern void deactivate_task(struct rq *rq, struct task_struct *p, int flags);
 
 extern void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags);
 
+#ifdef CONFIG_PREEMPT_RT
+#define SCHED_NR_MIGRATE_BREAK 8
+#else
+#define SCHED_NR_MIGRATE_BREAK 32
+#endif
+
 extern const_debug unsigned int sysctl_sched_nr_migrate;
 extern const_debug unsigned int sysctl_sched_migration_cost;
 
-- 
2.17.1

Re: [PATCH 2/4] sched/fair: cleanup loop_max and loop_break

Posted by Dietmar Eggemann 3 years, 6 months ago

On 25/08/2022 14:27, Vincent Guittot wrote:
> sched_nr_migrate_break is set to a fix value and never changes so we can
> replace it by a define SCHED_NR_MIGRATE_BREAK.
> 
> Also, we adjust SCHED_NR_MIGRATE_BREAK to be aligned with the init value
> of sysctl_sched_nr_migrate which can be init to different values.
> 
> Then, use SCHED_NR_MIGRATE_BREAK to init sysctl_sched_nr_migrate.
> 
> The behavior stays unchanged unless you modify sysctl_sched_nr_migrate
> trough debugfs.

I don't quite get this sentence. The behavior would potentially change
if you change sysctl_sched_nr_migrate before this patch too?

> 
> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
> ---
>  kernel/sched/core.c  |  6 +-----
>  kernel/sched/fair.c  | 11 ++++-------
>  kernel/sched/sched.h |  6 ++++++
>  3 files changed, 11 insertions(+), 12 deletions(-)
> 
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 64c08993221b..a21e817bdd1c 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -142,11 +142,7 @@ __read_mostly int sysctl_resched_latency_warn_once = 1;
>   * Number of tasks to iterate in a single balance run.
>   * Limited because this is done with IRQs disabled.
>   */

    ^^^
Shouldn't this comment be removed as well?

> -#ifdef CONFIG_PREEMPT_RT
> -const_debug unsigned int sysctl_sched_nr_migrate = 8;
> -#else
> -const_debug unsigned int sysctl_sched_nr_migrate = 32;
> -#endif
> +const_debug unsigned int sysctl_sched_nr_migrate = SCHED_NR_MIGRATE_BREAK;
>  
>  __read_mostly int scheduler_running;

[...]

Re: [PATCH 2/4] sched/fair: cleanup loop_max and loop_break

Posted by Vincent Guittot 3 years, 6 months ago

On Mon, 12 Sept 2022 at 10:45, Dietmar Eggemann
<dietmar.eggemann@arm.com> wrote:
>
> On 25/08/2022 14:27, Vincent Guittot wrote:
> > sched_nr_migrate_break is set to a fix value and never changes so we can
> > replace it by a define SCHED_NR_MIGRATE_BREAK.
> >
> > Also, we adjust SCHED_NR_MIGRATE_BREAK to be aligned with the init value
> > of sysctl_sched_nr_migrate which can be init to different values.
> >
> > Then, use SCHED_NR_MIGRATE_BREAK to init sysctl_sched_nr_migrate.
> >
> > The behavior stays unchanged unless you modify sysctl_sched_nr_migrate
> > trough debugfs.
>
> I don't quite get this sentence. The behavior would potentially change
> if you change sysctl_sched_nr_migrate before this patch too?

the behavior is different if you change the sysctl_sched_nr_migrate.

With this patch, loop_break is now aligned with
sysctl_sched_nr_migrate value which was not the case for
CONFIG_PREEMPT_RT. For the latter, the behavior can change if you
increase sysctl_sched_nr_migrate at runtime because there is now at
least one break whereas it was not the case before as long as
sysctl_sched_nr_migrate stayed below 32

>
> >
> > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
> > ---
> >  kernel/sched/core.c  |  6 +-----
> >  kernel/sched/fair.c  | 11 ++++-------
> >  kernel/sched/sched.h |  6 ++++++
> >  3 files changed, 11 insertions(+), 12 deletions(-)
> >
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index 64c08993221b..a21e817bdd1c 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -142,11 +142,7 @@ __read_mostly int sysctl_resched_latency_warn_once = 1;
> >   * Number of tasks to iterate in a single balance run.
> >   * Limited because this is done with IRQs disabled.
> >   */
>
>     ^^^
> Shouldn't this comment be removed as well?
>
> > -#ifdef CONFIG_PREEMPT_RT
> > -const_debug unsigned int sysctl_sched_nr_migrate = 8;
> > -#else
> > -const_debug unsigned int sysctl_sched_nr_migrate = 32;
> > -#endif
> > +const_debug unsigned int sysctl_sched_nr_migrate = SCHED_NR_MIGRATE_BREAK;
> >
> >  __read_mostly int scheduler_running;
>
> [...]

[tip: sched/core] sched/fair: Cleanup loop_max and loop_break

Posted by tip-bot2 for Vincent Guittot 3 years, 6 months ago

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

Commit-ID:     c59862f8265f8060b6650ee1dc12159fe5c89779
Gitweb:        https://git.kernel.org/tip/c59862f8265f8060b6650ee1dc12159fe5c89779
Author:        Vincent Guittot <vincent.guittot@linaro.org>
AuthorDate:    Thu, 25 Aug 2022 14:27:24 +02:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Thu, 15 Sep 2022 16:13:51 +02:00

sched/fair: Cleanup loop_max and loop_break

sched_nr_migrate_break is set to a fix value and never changes so we can
replace it by a define SCHED_NR_MIGRATE_BREAK.

Also, we adjust SCHED_NR_MIGRATE_BREAK to be aligned with the init value
of sysctl_sched_nr_migrate which can be init to different values.

Then, use SCHED_NR_MIGRATE_BREAK to init sysctl_sched_nr_migrate.

The behavior stays unchanged unless you modify sysctl_sched_nr_migrate
trough debugfs.

Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220825122726.20819-3-vincent.guittot@linaro.org
---
 kernel/sched/core.c  |  6 +-----
 kernel/sched/fair.c  | 11 ++++-------
 kernel/sched/sched.h |  6 ++++++
 3 files changed, 11 insertions(+), 12 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 2b85d1b..4fa4a3d 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -142,11 +142,7 @@ __read_mostly int sysctl_resched_latency_warn_once = 1;
  * Number of tasks to iterate in a single balance run.
  * Limited because this is done with IRQs disabled.
  */
-#ifdef CONFIG_PREEMPT_RT
-const_debug unsigned int sysctl_sched_nr_migrate = 8;
-#else
-const_debug unsigned int sysctl_sched_nr_migrate = 32;
-#endif
+const_debug unsigned int sysctl_sched_nr_migrate = SCHED_NR_MIGRATE_BREAK;
 
 __read_mostly int scheduler_running;
 
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index dae3bfa..7b3a58f 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -8009,8 +8009,6 @@ static struct task_struct *detach_one_task(struct lb_env *env)
 	return NULL;
 }
 
-static const unsigned int sched_nr_migrate_break = 32;
-
 /*
  * detach_tasks() -- tries to detach up to imbalance load/util/tasks from
  * busiest_rq, as part of a balancing operation within domain "sd".
@@ -8059,7 +8057,7 @@ static int detach_tasks(struct lb_env *env)
 
 		/* take a breather every nr_migrate tasks */
 		if (env->loop > env->loop_break) {
-			env->loop_break += sched_nr_migrate_break;
+			env->loop_break += SCHED_NR_MIGRATE_BREAK;
 			env->flags |= LBF_NEED_BREAK;
 			break;
 		}
@@ -10100,14 +10098,13 @@ static int load_balance(int this_cpu, struct rq *this_rq,
 	struct rq *busiest;
 	struct rq_flags rf;
 	struct cpumask *cpus = this_cpu_cpumask_var_ptr(load_balance_mask);
-
 	struct lb_env env = {
 		.sd		= sd,
 		.dst_cpu	= this_cpu,
 		.dst_rq		= this_rq,
 		.dst_grpmask    = sched_group_span(sd->groups),
 		.idle		= idle,
-		.loop_break	= sched_nr_migrate_break,
+		.loop_break	= SCHED_NR_MIGRATE_BREAK,
 		.cpus		= cpus,
 		.fbq_type	= all,
 		.tasks		= LIST_HEAD_INIT(env.tasks),
@@ -10216,7 +10213,7 @@ more_balance:
 			env.dst_cpu	 = env.new_dst_cpu;
 			env.flags	&= ~LBF_DST_PINNED;
 			env.loop	 = 0;
-			env.loop_break	 = sched_nr_migrate_break;
+			env.loop_break	 = SCHED_NR_MIGRATE_BREAK;
 
 			/*
 			 * Go back to "more_balance" rather than "redo" since we
@@ -10248,7 +10245,7 @@ more_balance:
 			 */
 			if (!cpumask_subset(cpus, env.dst_grpmask)) {
 				env.loop = 0;
-				env.loop_break = sched_nr_migrate_break;
+				env.loop_break = SCHED_NR_MIGRATE_BREAK;
 				goto redo;
 			}
 			goto out_all_pinned;
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 91b2c7e..1fc198b 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2423,6 +2423,12 @@ extern void deactivate_task(struct rq *rq, struct task_struct *p, int flags);
 
 extern void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags);
 
+#ifdef CONFIG_PREEMPT_RT
+#define SCHED_NR_MIGRATE_BREAK 8
+#else
+#define SCHED_NR_MIGRATE_BREAK 32
+#endif
+
 extern const_debug unsigned int sysctl_sched_nr_migrate;
 extern const_debug unsigned int sysctl_sched_migration_cost;

[PATCH 1/4] sched/fair: make sure to try to detach at least one movable task
[PATCH 2/4] sched/fair: cleanup loop_max and loop_break
[PATCH 3/4] sched/fair: move call to list_last_entry() in detach_tasks
[PATCH 4/4] sched/fair: limit sched slice duration