[PATCH v9 0/2] Make find_later_rq() choose a closer cpu in topology

[PATCH v9 0/2] Make find_later_rq() choose a closer cpu in topology

[Byungchul Park]

Change from v8
   -. add suggested-by Peterz
   -. add several comments

Change from v7
   -. fix a trivial typo
   -. modify commit messages to explain what it does more clearly
   -. simplify code with an existing macro

Change from v6
   -. add a comment about selection of fallback_cpu incase more than one exist
   -. modify a comment explaining what we do wrt PREFER_SIBLING

Change from v5
   -. exclude two patches already picked up by peterz
      (sched/deadline: Make find_later_rq() choose a closer cpu in topology)
      (sched/deadline: Change return value of cpudl_find())
   -. apply what peterz fixed for 'prefer sibling', into deadline and rt

Change from v4
   -. remove a patch that might cause huge lock contention
      (by spin lock(&cpudl.lock) in a hot path of scheduler)

Change from v3
   -. rename closest_cpu to best_cpu so that it align with rt
   -. protect referring cpudl.elements with cpudl.lock
   -. change return value of cpudl_find() to bool

Change from v2
   -. add support for SD_PREFER_SIBLING

Change from v1
   -. clean up the patch

Byungchul Park (2):
  sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()
  sched/rt: Add support for SD_PREFER_SIBLING on find_lowest_rq()

 kernel/sched/deadline.c | 80 ++++++++++++++++++++++++++++++++++++++++++++-----
 kernel/sched/rt.c       | 78 +++++++++++++++++++++++++++++++++++++++++++++--
 2 files changed, 148 insertions(+), 10 deletions(-)

-- 
1.9.1

[PATCH v9 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Byungchul Park]

It would be better to try to check other siblings first if
SD_PREFER_SIBLING is flaged when pushing tasks - migration.

Signed-off-by: Byungchul Park <byungchul.park@lge.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
---
 kernel/sched/deadline.c | 80 ++++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 73 insertions(+), 7 deletions(-)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 0223694..b6b3855 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1319,12 +1319,35 @@ static struct task_struct *pick_earliest_pushable_dl_task(struct rq *rq, int cpu
 
 static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask_dl);
 
+/*
+ * Find the first cpu in: mask & sd & ~prefer
+ */
+static int find_cpu(const struct cpumask *mask,
+		    const struct sched_domain *sd,
+		    const struct sched_domain *prefer)
+{
+	const struct cpumask *sds = sched_domain_span(sd);
+	const struct cpumask *ps  = prefer ? sched_domain_span(prefer) : NULL;
+	int cpu;
+
+	for_each_cpu(cpu, mask) {
+		if (!cpumask_test_cpu(cpu, sds))
+			continue;
+		if (ps && cpumask_test_cpu(cpu, ps))
+			continue;
+		break;
+	}
+
+	return cpu;
+}
+
 static int find_later_rq(struct task_struct *task)
 {
-	struct sched_domain *sd;
+	struct sched_domain *sd, *prefer = NULL;
 	struct cpumask *later_mask = this_cpu_cpumask_var_ptr(local_cpu_mask_dl);
 	int this_cpu = smp_processor_id();
 	int cpu = task_cpu(task);
+	int fallback_cpu = -1;
 
 	/* Make sure the mask is initialized first */
 	if (unlikely(!later_mask))
@@ -1376,15 +1399,35 @@ static int find_later_rq(struct task_struct *task)
 				return this_cpu;
 			}
 
-			best_cpu = cpumask_first_and(later_mask,
-							sched_domain_span(sd));
 			/*
-			 * Last chance: if a cpu being in both later_mask
-			 * and current sd span is valid, that becomes our
-			 * choice. Of course, the latest possible cpu is
-			 * already under consideration through later_mask.
+			 * If a cpu being in later_mask & current sd &
+			 * ~prefer sd is valid, that becomes our choice.
+			 * Of course, the latest possible cpu is already
+			 * under consideration through later_mask.
 			 */
+			best_cpu = find_cpu(later_mask, sd, prefer);
+
 			if (best_cpu < nr_cpu_ids) {
+				/*
+				 * If current domain is SD_PREFER_SIBLING
+				 * flaged, we have to try to check other
+				 * siblings first.
+				 */
+				if (sd->flags & SD_PREFER_SIBLING) {
+					prefer = sd;
+
+					/*
+					 * fallback_cpu should be one
+					 * in the closest domain among
+					 * SD_PREFER_SIBLING domains,
+					 * in case that more than one
+					 * SD_PREFER_SIBLING domains
+					 * exist in the hierachy.
+					 */
+					if (fallback_cpu == -1)
+						fallback_cpu = best_cpu;
+					continue;
+				}
 				rcu_read_unlock();
 				return best_cpu;
 			}
@@ -1393,6 +1436,29 @@ static int find_later_rq(struct task_struct *task)
 	rcu_read_unlock();
 
 	/*
+	 * If fallback_cpu is valid, all our guesses failed *except* for
+	 * SD_PREFER_SIBLING domain. Now, we can return the fallback cpu.
+	 *
+	 * XXX: Consider the following example, 4 cores SMT2 system:
+	 *
+	 *    LLC [0       -        7]
+	 *    SMT [0 1][2 3][4 5][6 7]
+	 *         o x  o x  x x  x x
+	 *
+	 *    where 'o': occupied and 'x': empty.
+	 *
+	 * A wakeup on cpu0 will exclude cpu1 and choose cpu3, since
+	 * cpu1 is in a SD_PREFER_SIBLING sd and cpu3 is not. However,
+	 * in this case, we have to choose cpu4 for better work, instead
+	 * of cpu3 that is fully loaded.
+	 *
+	 * We have to do the best if possible, but choose the second
+	 * best here since that is too expensive to adopt.
+	 */
+	if (fallback_cpu != -1)
+		return fallback_cpu;
+
+	/*
 	 * At this point, all our guesses failed, we just return
 	 * 'something', and let the caller sort the things out.
 	 */
-- 
1.9.1

Re: [PATCH v9 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Juri Lelli]

Hi,

On 22/08/17 17:30, Byungchul Park wrote:
> It would be better to try to check other siblings first if
> SD_PREFER_SIBLING is flaged when pushing tasks - migration.
> 
> Signed-off-by: Byungchul Park <byungchul.park@lge.com>
> Suggested-by: Peter Zijlstra <peterz@infradead.org>
       ^
This has to come before your SoB.

> ---
>  kernel/sched/deadline.c | 80 ++++++++++++++++++++++++++++++++++++++++++++-----
>  1 file changed, 73 insertions(+), 7 deletions(-)
> 
> diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
> index 0223694..b6b3855 100644
> --- a/kernel/sched/deadline.c
> +++ b/kernel/sched/deadline.c
> @@ -1319,12 +1319,35 @@ static struct task_struct *pick_earliest_pushable_dl_task(struct rq *rq, int cpu
>  
>  static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask_dl);
>  
> +/*
> + * Find the first cpu in: mask & sd & ~prefer
> + */
> +static int find_cpu(const struct cpumask *mask,
> +		    const struct sched_domain *sd,
> +		    const struct sched_domain *prefer)
> +{
> +	const struct cpumask *sds = sched_domain_span(sd);
> +	const struct cpumask *ps  = prefer ? sched_domain_span(prefer) : NULL;
> +	int cpu;
> +
> +	for_each_cpu(cpu, mask) {
> +		if (!cpumask_test_cpu(cpu, sds))
> +			continue;
> +		if (ps && cpumask_test_cpu(cpu, ps))
> +			continue;
> +		break;
> +	}
> +
> +	return cpu;
> +}
> +
>  static int find_later_rq(struct task_struct *task)
>  {
> -	struct sched_domain *sd;
> +	struct sched_domain *sd, *prefer = NULL;
>  	struct cpumask *later_mask = this_cpu_cpumask_var_ptr(local_cpu_mask_dl);
>  	int this_cpu = smp_processor_id();
>  	int cpu = task_cpu(task);
> +	int fallback_cpu = -1;
>  
>  	/* Make sure the mask is initialized first */
>  	if (unlikely(!later_mask))
> @@ -1376,15 +1399,35 @@ static int find_later_rq(struct task_struct *task)
>  				return this_cpu;
>  			}
>  
> -			best_cpu = cpumask_first_and(later_mask,
> -							sched_domain_span(sd));
>  			/*
> -			 * Last chance: if a cpu being in both later_mask
> -			 * and current sd span is valid, that becomes our
> -			 * choice. Of course, the latest possible cpu is
> -			 * already under consideration through later_mask.
> +			 * If a cpu being in later_mask & current sd &
> +			 * ~prefer sd is valid, that becomes our choice.
> +			 * Of course, the latest possible cpu is already
> +			 * under consideration through later_mask.
>  			 */
> +			best_cpu = find_cpu(later_mask, sd, prefer);
> +
>  			if (best_cpu < nr_cpu_ids) {
> +				/*
> +				 * If current domain is SD_PREFER_SIBLING
> +				 * flaged, we have to try to check other
> +				 * siblings first.
> +				 */
> +				if (sd->flags & SD_PREFER_SIBLING) {
> +					prefer = sd;
> +
> +					/*
> +					 * fallback_cpu should be one
> +					 * in the closest domain among
> +					 * SD_PREFER_SIBLING domains,
> +					 * in case that more than one
> +					 * SD_PREFER_SIBLING domains
> +					 * exist in the hierachy.
> +					 */
> +					if (fallback_cpu == -1)
> +						fallback_cpu = best_cpu;
> +					continue;
> +				}
>  				rcu_read_unlock();
>  				return best_cpu;
>  			}
> @@ -1393,6 +1436,29 @@ static int find_later_rq(struct task_struct *task)
>  	rcu_read_unlock();
>  
>  	/*
> +	 * If fallback_cpu is valid, all our guesses failed *except* for
> +	 * SD_PREFER_SIBLING domain. Now, we can return the fallback cpu.
> +	 *
> +	 * XXX: Consider the following example, 4 cores SMT2 system:
> +	 *
> +	 *    LLC [0       -        7]
> +	 *    SMT [0 1][2 3][4 5][6 7]
> +	 *         o x  o x  x x  x x
> +	 *
> +	 *    where 'o': occupied and 'x': empty.
> +	 *
> +	 * A wakeup on cpu0 will exclude cpu1 and choose cpu3, since
> +	 * cpu1 is in a SD_PREFER_SIBLING sd and cpu3 is not. However,
> +	 * in this case, we have to choose cpu4 for better work, instead

... in this case cpu4 would have been a better choice, since cpu3 is a
(SMT) thread of an already loaded core.

> +	 * of cpu3 that is fully loaded.
> +	 *
> +	 * We have to do the best if possible, but choose the second
> +	 * best here since that is too expensive to adopt.
> +	 */

I'd also modify this last paragraph with something like:

"Doing it 'right' is difficult and expensive. The current solution is an
acceptable approximation."

Apart from these minor points, patch looks ok to me.

Acked-by: Juri Lelli <juri.lelli@arm.com>

Best,

- Juri

Re: [PATCH v9 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Byungchul Park]

On Tue, Aug 22, 2017 at 10:25:42AM +0100, Juri Lelli wrote:
> Hi,
> 
> On 22/08/17 17:30, Byungchul Park wrote:
> > It would be better to try to check other siblings first if
> > SD_PREFER_SIBLING is flaged when pushing tasks - migration.
> > 
> > Signed-off-by: Byungchul Park <byungchul.park@lge.com>
> > Suggested-by: Peter Zijlstra <peterz@infradead.org>
>        ^
> This has to come before your SoB.

Thank you, I will.

> > ---
> >  kernel/sched/deadline.c | 80 ++++++++++++++++++++++++++++++++++++++++++++-----
> >  1 file changed, 73 insertions(+), 7 deletions(-)
> > 
> > diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
> > index 0223694..b6b3855 100644
> > --- a/kernel/sched/deadline.c
> > +++ b/kernel/sched/deadline.c
> > @@ -1319,12 +1319,35 @@ static struct task_struct *pick_earliest_pushable_dl_task(struct rq *rq, int cpu
> >  
> >  static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask_dl);
> >  
> > +/*
> > + * Find the first cpu in: mask & sd & ~prefer
> > + */
> > +static int find_cpu(const struct cpumask *mask,
> > +		    const struct sched_domain *sd,
> > +		    const struct sched_domain *prefer)
> > +{
> > +	const struct cpumask *sds = sched_domain_span(sd);
> > +	const struct cpumask *ps  = prefer ? sched_domain_span(prefer) : NULL;
> > +	int cpu;
> > +
> > +	for_each_cpu(cpu, mask) {
> > +		if (!cpumask_test_cpu(cpu, sds))
> > +			continue;
> > +		if (ps && cpumask_test_cpu(cpu, ps))
> > +			continue;
> > +		break;
> > +	}
> > +
> > +	return cpu;
> > +}
> > +
> >  static int find_later_rq(struct task_struct *task)
> >  {
> > -	struct sched_domain *sd;
> > +	struct sched_domain *sd, *prefer = NULL;
> >  	struct cpumask *later_mask = this_cpu_cpumask_var_ptr(local_cpu_mask_dl);
> >  	int this_cpu = smp_processor_id();
> >  	int cpu = task_cpu(task);
> > +	int fallback_cpu = -1;
> >  
> >  	/* Make sure the mask is initialized first */
> >  	if (unlikely(!later_mask))
> > @@ -1376,15 +1399,35 @@ static int find_later_rq(struct task_struct *task)
> >  				return this_cpu;
> >  			}
> >  
> > -			best_cpu = cpumask_first_and(later_mask,
> > -							sched_domain_span(sd));
> >  			/*
> > -			 * Last chance: if a cpu being in both later_mask
> > -			 * and current sd span is valid, that becomes our
> > -			 * choice. Of course, the latest possible cpu is
> > -			 * already under consideration through later_mask.
> > +			 * If a cpu being in later_mask & current sd &
> > +			 * ~prefer sd is valid, that becomes our choice.
> > +			 * Of course, the latest possible cpu is already
> > +			 * under consideration through later_mask.
> >  			 */
> > +			best_cpu = find_cpu(later_mask, sd, prefer);
> > +
> >  			if (best_cpu < nr_cpu_ids) {
> > +				/*
> > +				 * If current domain is SD_PREFER_SIBLING
> > +				 * flaged, we have to try to check other
> > +				 * siblings first.
> > +				 */
> > +				if (sd->flags & SD_PREFER_SIBLING) {
> > +					prefer = sd;
> > +
> > +					/*
> > +					 * fallback_cpu should be one
> > +					 * in the closest domain among
> > +					 * SD_PREFER_SIBLING domains,
> > +					 * in case that more than one
> > +					 * SD_PREFER_SIBLING domains
> > +					 * exist in the hierachy.
> > +					 */
> > +					if (fallback_cpu == -1)
> > +						fallback_cpu = best_cpu;
> > +					continue;
> > +				}
> >  				rcu_read_unlock();
> >  				return best_cpu;
> >  			}
> > @@ -1393,6 +1436,29 @@ static int find_later_rq(struct task_struct *task)
> >  	rcu_read_unlock();
> >  
> >  	/*
> > +	 * If fallback_cpu is valid, all our guesses failed *except* for
> > +	 * SD_PREFER_SIBLING domain. Now, we can return the fallback cpu.
> > +	 *
> > +	 * XXX: Consider the following example, 4 cores SMT2 system:
> > +	 *
> > +	 *    LLC [0       -        7]
> > +	 *    SMT [0 1][2 3][4 5][6 7]
> > +	 *         o x  o x  x x  x x
> > +	 *
> > +	 *    where 'o': occupied and 'x': empty.
> > +	 *
> > +	 * A wakeup on cpu0 will exclude cpu1 and choose cpu3, since
> > +	 * cpu1 is in a SD_PREFER_SIBLING sd and cpu3 is not. However,
> > +	 * in this case, we have to choose cpu4 for better work, instead
> 
> ... in this case cpu4 would have been a better choice, since cpu3 is a
> (SMT) thread of an already loaded core.

Thank you, I will.

> > +	 * of cpu3 that is fully loaded.
> > +	 *
> > +	 * We have to do the best if possible, but choose the second
> > +	 * best here since that is too expensive to adopt.
> > +	 */
> 
> I'd also modify this last paragraph with something like:
> 
> "Doing it 'right' is difficult and expensive. The current solution is an
> acceptable approximation."

Thank you, I will.

> Apart from these minor points, patch looks ok to me.
> 
> Acked-by: Juri Lelli <juri.lelli@arm.com>
> 
> Best,
> 
> - Juri

[PATCH v9 2/2] sched/rt: Add support for SD_PREFER_SIBLING on find_lowest_rq()

[Byungchul Park]

Hello Steven,

I added several comments so I'm not sure if I could add your reviewed-by.

----->8-----
>From f0710d99759ed28c1409a527166780899f00d236 Mon Sep 17 00:00:00 2001
From: Byungchul Park <byungchul.park@lge.com>
Date: Tue, 22 Aug 2017 17:19:21 +0900
Subject: [PATCH v9 2/2] sched/rt: Add support for SD_PREFER_SIBLING on
 find_lowest_rq()

It would be better to try to check other siblings first if
SD_PREFER_SIBLING is flaged when pushing tasks - migration.

Signed-off-by: Byungchul Park <byungchul.park@lge.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
---
 kernel/sched/rt.c | 78 ++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 75 insertions(+), 3 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 979b734..196ffc7 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1618,12 +1618,35 @@ static struct task_struct *pick_highest_pushable_task(struct rq *rq, int cpu)
 
 static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask);
 
+/*
+ * Find the first cpu in: mask & sd & ~prefer
+ */
+static int find_cpu(const struct cpumask *mask,
+		    const struct sched_domain *sd,
+		    const struct sched_domain *prefer)
+{
+	const struct cpumask *sds = sched_domain_span(sd);
+	const struct cpumask *ps  = prefer ? sched_domain_span(prefer) : NULL;
+	int cpu;
+
+	for_each_cpu(cpu, mask) {
+		if (!cpumask_test_cpu(cpu, sds))
+			continue;
+		if (ps && cpumask_test_cpu(cpu, ps))
+			continue;
+		break;
+	}
+
+	return cpu;
+}
+
 static int find_lowest_rq(struct task_struct *task)
 {
-	struct sched_domain *sd;
+	struct sched_domain *sd, *prefer = NULL;
 	struct cpumask *lowest_mask = this_cpu_cpumask_var_ptr(local_cpu_mask);
 	int this_cpu = smp_processor_id();
 	int cpu      = task_cpu(task);
+	int fallback_cpu = -1;
 
 	/* Make sure the mask is initialized first */
 	if (unlikely(!lowest_mask))
@@ -1668,9 +1691,35 @@ static int find_lowest_rq(struct task_struct *task)
 				return this_cpu;
 			}
 
-			best_cpu = cpumask_first_and(lowest_mask,
-						     sched_domain_span(sd));
+			/*
+			 * If a cpu being in lowest_mask & current sd &
+			 * ~prefer sd is valid, that becomes our choice.
+			 * Of course, the lowest possible cpu is already
+			 * under consideration through lowest_mask.
+			 */
+			best_cpu = find_cpu(lowest_mask, sd, prefer);
+
 			if (best_cpu < nr_cpu_ids) {
+				/*
+				 * If current domain is SD_PREFER_SIBLING
+				 * flaged, we have to try to check other
+				 * siblings first.
+				 */
+				if (sd->flags & SD_PREFER_SIBLING) {
+					prefer = sd;
+
+					/*
+					 * fallback_cpu should be one
+					 * in the closest domain among
+					 * SD_PREFER_SIBLING domains,
+					 * in case that more than one
+					 * SD_PREFER_SIBLING domains
+					 * exist in the hierachy.
+					 */
+					if (fallback_cpu == -1)
+						fallback_cpu = best_cpu;
+					continue;
+				}
 				rcu_read_unlock();
 				return best_cpu;
 			}
@@ -1679,6 +1728,29 @@ static int find_lowest_rq(struct task_struct *task)
 	rcu_read_unlock();
 
 	/*
+	 * If fallback_cpu is valid, all our guesses failed *except* for
+	 * SD_PREFER_SIBLING domain. Now, we can return the fallback cpu.
+	 *
+	 * XXX: Consider the following example, 4 cores SMT2 system:
+	 *
+	 *    LLC [0       -        7]
+	 *    SMT [0 1][2 3][4 5][6 7]
+	 *         o x  o x  x x  x x
+	 *
+	 *    where 'o': occupied and 'x': empty.
+	 *
+	 * A wakeup on cpu0 will exclude cpu1 and choose cpu3, since
+	 * cpu1 is in a SD_PREFER_SIBLING sd and cpu3 is not. However,
+	 * in this case, we have to choose cpu4 for better work, instead
+	 * of cpu3 that is fully loaded.
+	 *
+	 * We have to do the best if possible, but choose the second
+	 * best here since that is too expensive to adopt.
+	 */
+	if (fallback_cpu != -1)
+		return fallback_cpu;
+
+	/*
 	 * And finally, if there were no matches within the domains
 	 * just give the caller *something* to work with from the compatible
 	 * locations.
-- 
1.9.1