[PATCH 0/3] sched/fair: Init NUMA task_work in sched_fork()

[PATCH 0/3] sched/fair: Init NUMA task_work in sched_fork()

[Valentin Schneider]

A TODO has been sitting in task_tick_numa() regarding init'ing the
task_numa_work() task_work in sched_fork() rather than in task_tick_numa(),
so I figured I'd have a go at it.

Patches 1 & 2 do that, and patch 3 is a freebie cleanup.

Briefly tested on a 2 * (Xeon E5-2690) system, didn't see any obvious
breakage.

Valentin Schneider (3):
  sched/fair: Move init_numa_balancing() below task_numa_work()
  sched/fair: Move task_numa_work() init to init_numa_balancing()
  sched/fair: Change task_numa_work() storage to static

 kernel/sched/fair.c | 93 +++++++++++++++++++++++----------------------
 1 file changed, 47 insertions(+), 46 deletions(-)

--
2.22.0

[PATCH 1/3] sched/fair: Move init_numa_balancing() below task_numa_work()

[Valentin Schneider]

To reference task_numa_work() from within init_numa_balancing(), we
need the former to be declared before the latter. Do just that.

This is a pure code movement.

Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
---
 kernel/sched/fair.c | 82 ++++++++++++++++++++++-----------------------
 1 file changed, 41 insertions(+), 41 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 036be95a87e9..476b0201a8fb 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1169,47 +1169,6 @@ static unsigned int task_scan_max(struct task_struct *p)
 	return max(smin, smax);
 }
 
-void init_numa_balancing(unsigned long clone_flags, struct task_struct *p)
-{
-	int mm_users = 0;
-	struct mm_struct *mm = p->mm;
-
-	if (mm) {
-		mm_users = atomic_read(&mm->mm_users);
-		if (mm_users == 1) {
-			mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
-			mm->numa_scan_seq = 0;
-		}
-	}
-	p->node_stamp			= 0;
-	p->numa_scan_seq		= mm ? mm->numa_scan_seq : 0;
-	p->numa_scan_period		= sysctl_numa_balancing_scan_delay;
-	p->numa_work.next		= &p->numa_work;
-	p->numa_faults			= NULL;
-	p->numa_group			= NULL;
-	p->last_task_numa_placement	= 0;
-	p->last_sum_exec_runtime	= 0;
-
-	/* New address space, reset the preferred nid */
-	if (!(clone_flags & CLONE_VM)) {
-		p->numa_preferred_nid = NUMA_NO_NODE;
-		return;
-	}
-
-	/*
-	 * New thread, keep existing numa_preferred_nid which should be copied
-	 * already by arch_dup_task_struct but stagger when scans start.
-	 */
-	if (mm) {
-		unsigned int delay;
-
-		delay = min_t(unsigned int, task_scan_max(current),
-			current->numa_scan_period * mm_users * NSEC_PER_MSEC);
-		delay += 2 * TICK_NSEC;
-		p->node_stamp = delay;
-	}
-}
-
 static void account_numa_enqueue(struct rq *rq, struct task_struct *p)
 {
 	rq->nr_numa_running += (p->numa_preferred_nid != NUMA_NO_NODE);
@@ -2611,6 +2570,47 @@ void task_numa_work(struct callback_head *work)
 	}
 }
 
+void init_numa_balancing(unsigned long clone_flags, struct task_struct *p)
+{
+	int mm_users = 0;
+	struct mm_struct *mm = p->mm;
+
+	if (mm) {
+		mm_users = atomic_read(&mm->mm_users);
+		if (mm_users == 1) {
+			mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
+			mm->numa_scan_seq = 0;
+		}
+	}
+	p->node_stamp			= 0;
+	p->numa_scan_seq		= mm ? mm->numa_scan_seq : 0;
+	p->numa_scan_period		= sysctl_numa_balancing_scan_delay;
+	p->numa_work.next		= &p->numa_work;
+	p->numa_faults			= NULL;
+	p->numa_group			= NULL;
+	p->last_task_numa_placement	= 0;
+	p->last_sum_exec_runtime	= 0;
+
+	/* New address space, reset the preferred nid */
+	if (!(clone_flags & CLONE_VM)) {
+		p->numa_preferred_nid = NUMA_NO_NODE;
+		return;
+	}
+
+	/*
+	 * New thread, keep existing numa_preferred_nid which should be copied
+	 * already by arch_dup_task_struct but stagger when scans start.
+	 */
+	if (mm) {
+		unsigned int delay;
+
+		delay = min_t(unsigned int, task_scan_max(current),
+			current->numa_scan_period * mm_users * NSEC_PER_MSEC);
+		delay += 2 * TICK_NSEC;
+		p->node_stamp = delay;
+	}
+}
+
 /*
  * Drive the periodic memory faults..
  */
--
2.22.0

[PATCH 2/3] sched/fair: Move task_numa_work() init to init_numa_balancing()

[Valentin Schneider]

We only need to set the callback_head worker function once, do it
during sched_fork().

While at it, move the comment regarding double task_work addition to
init_numa_balancing(), since the double add sentinel is first set there.

Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
---
 kernel/sched/fair.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 476b0201a8fb..74faa55bc52a 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2441,7 +2441,7 @@ void task_numa_work(struct callback_head *work)
 
 	SCHED_WARN_ON(p != container_of(work, struct task_struct, numa_work));
 
-	work->next = work; /* protect against double add */
+	work->next = work;
 	/*
 	 * Who cares about NUMA placement when they're dying.
 	 *
@@ -2585,12 +2585,15 @@ void init_numa_balancing(unsigned long clone_flags, struct task_struct *p)
 	p->node_stamp			= 0;
 	p->numa_scan_seq		= mm ? mm->numa_scan_seq : 0;
 	p->numa_scan_period		= sysctl_numa_balancing_scan_delay;
+	/* Protect against double add, see task_tick_numa and task_numa_work */
 	p->numa_work.next		= &p->numa_work;
 	p->numa_faults			= NULL;
 	p->numa_group			= NULL;
 	p->last_task_numa_placement	= 0;
 	p->last_sum_exec_runtime	= 0;
 
+	init_task_work(&p->numa_work, task_numa_work);
+
 	/* New address space, reset the preferred nid */
 	if (!(clone_flags & CLONE_VM)) {
 		p->numa_preferred_nid = NUMA_NO_NODE;
@@ -2639,10 +2642,8 @@ static void task_tick_numa(struct rq *rq, struct task_struct *curr)
 			curr->numa_scan_period = task_scan_start(curr);
 		curr->node_stamp += period;
 
-		if (!time_before(jiffies, curr->mm->numa_next_scan)) {
-			init_task_work(work, task_numa_work); /* TODO: move this into sched_fork() */
+		if (!time_before(jiffies, curr->mm->numa_next_scan))
 			task_work_add(curr, work, true);
-		}
 	}
 }
 
--
2.22.0

[PATCH 3/3] sched/fair: Change task_numa_work() storage to static

[Valentin Schneider]

There are no callers outside of fair.c.

Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
---
 kernel/sched/fair.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 74faa55bc52a..c747ce05e726 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2428,7 +2428,7 @@ static void reset_ptenuma_scan(struct task_struct *p)
  * The expensive part of numa migration is done from task_work context.
  * Triggered from task_tick_numa().
  */
-void task_numa_work(struct callback_head *work)
+static void task_numa_work(struct callback_head *work)
 {
 	unsigned long migrate, next_scan, now = jiffies;
 	struct task_struct *p = current;
-- 
2.22.0

Re: [PATCH 0/3] sched/fair: Init NUMA task_work in sched_fork()

[Peter Zijlstra]

On Mon, Jul 15, 2019 at 11:25:05AM +0100, Valentin Schneider wrote:
> A TODO has been sitting in task_tick_numa() regarding init'ing the
> task_numa_work() task_work in sched_fork() rather than in task_tick_numa(),
> so I figured I'd have a go at it.
> 
> Patches 1 & 2 do that, and patch 3 is a freebie cleanup.
> 
> Briefly tested on a 2 * (Xeon E5-2690) system, didn't see any obvious
> breakage.
> 
> Valentin Schneider (3):
>   sched/fair: Move init_numa_balancing() below task_numa_work()
>   sched/fair: Move task_numa_work() init to init_numa_balancing()
>   sched/fair: Change task_numa_work() storage to static
> 
>  kernel/sched/fair.c | 93 +++++++++++++++++++++++----------------------
>  1 file changed, 47 insertions(+), 46 deletions(-)

Thanks!

[tip:sched/core] sched/fair: Move init_numa_balancing() below task_numa_work()

[tip-bot for Valentin Schneider]

Commit-ID:  d35927a144641700c8328d707d1c89d305b4ecb8
Gitweb:     https://git.kernel.org/tip/d35927a144641700c8328d707d1c89d305b4ecb8
Author:     Valentin Schneider <valentin.schneider@arm.com>
AuthorDate: Mon, 15 Jul 2019 11:25:06 +0100
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Thu, 25 Jul 2019 15:51:51 +0200

sched/fair: Move init_numa_balancing() below task_numa_work()

To reference task_numa_work() from within init_numa_balancing(), we
need the former to be declared before the latter. Do just that.

This is a pure code movement.

Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: mgorman@suse.de
Cc: riel@surriel.com
Link: https://lkml.kernel.org/r/20190715102508.32434-2-valentin.schneider@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 kernel/sched/fair.c | 82 ++++++++++++++++++++++++++---------------------------
 1 file changed, 41 insertions(+), 41 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index bc9cfeaac8bd..f0c488015649 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1188,47 +1188,6 @@ static unsigned int task_scan_max(struct task_struct *p)
 	return max(smin, smax);
 }
 
-void init_numa_balancing(unsigned long clone_flags, struct task_struct *p)
-{
-	int mm_users = 0;
-	struct mm_struct *mm = p->mm;
-
-	if (mm) {
-		mm_users = atomic_read(&mm->mm_users);
-		if (mm_users == 1) {
-			mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
-			mm->numa_scan_seq = 0;
-		}
-	}
-	p->node_stamp			= 0;
-	p->numa_scan_seq		= mm ? mm->numa_scan_seq : 0;
-	p->numa_scan_period		= sysctl_numa_balancing_scan_delay;
-	p->numa_work.next		= &p->numa_work;
-	p->numa_faults			= NULL;
-	RCU_INIT_POINTER(p->numa_group, NULL);
-	p->last_task_numa_placement	= 0;
-	p->last_sum_exec_runtime	= 0;
-
-	/* New address space, reset the preferred nid */
-	if (!(clone_flags & CLONE_VM)) {
-		p->numa_preferred_nid = NUMA_NO_NODE;
-		return;
-	}
-
-	/*
-	 * New thread, keep existing numa_preferred_nid which should be copied
-	 * already by arch_dup_task_struct but stagger when scans start.
-	 */
-	if (mm) {
-		unsigned int delay;
-
-		delay = min_t(unsigned int, task_scan_max(current),
-			current->numa_scan_period * mm_users * NSEC_PER_MSEC);
-		delay += 2 * TICK_NSEC;
-		p->node_stamp = delay;
-	}
-}
-
 static void account_numa_enqueue(struct rq *rq, struct task_struct *p)
 {
 	rq->nr_numa_running += (p->numa_preferred_nid != NUMA_NO_NODE);
@@ -2665,6 +2624,47 @@ out:
 	}
 }
 
+void init_numa_balancing(unsigned long clone_flags, struct task_struct *p)
+{
+	int mm_users = 0;
+	struct mm_struct *mm = p->mm;
+
+	if (mm) {
+		mm_users = atomic_read(&mm->mm_users);
+		if (mm_users == 1) {
+			mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
+			mm->numa_scan_seq = 0;
+		}
+	}
+	p->node_stamp			= 0;
+	p->numa_scan_seq		= mm ? mm->numa_scan_seq : 0;
+	p->numa_scan_period		= sysctl_numa_balancing_scan_delay;
+	p->numa_work.next		= &p->numa_work;
+	p->numa_faults			= NULL;
+	RCU_INIT_POINTER(p->numa_group, NULL);
+	p->last_task_numa_placement	= 0;
+	p->last_sum_exec_runtime	= 0;
+
+	/* New address space, reset the preferred nid */
+	if (!(clone_flags & CLONE_VM)) {
+		p->numa_preferred_nid = NUMA_NO_NODE;
+		return;
+	}
+
+	/*
+	 * New thread, keep existing numa_preferred_nid which should be copied
+	 * already by arch_dup_task_struct but stagger when scans start.
+	 */
+	if (mm) {
+		unsigned int delay;
+
+		delay = min_t(unsigned int, task_scan_max(current),
+			current->numa_scan_period * mm_users * NSEC_PER_MSEC);
+		delay += 2 * TICK_NSEC;
+		p->node_stamp = delay;
+	}
+}
+
 /*
  * Drive the periodic memory faults..
  */

[tip:sched/core] sched/fair: Move task_numa_work() init to init_numa_balancing()

[tip-bot for Valentin Schneider]

Commit-ID:  b34920d4ce6e6fc9424c20a4be98676eb543122f
Gitweb:     https://git.kernel.org/tip/b34920d4ce6e6fc9424c20a4be98676eb543122f
Author:     Valentin Schneider <valentin.schneider@arm.com>
AuthorDate: Mon, 15 Jul 2019 11:25:07 +0100
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Thu, 25 Jul 2019 15:51:51 +0200

sched/fair: Move task_numa_work() init to init_numa_balancing()

We only need to set the callback_head worker function once, do it
during sched_fork().

While at it, move the comment regarding double task_work addition to
init_numa_balancing(), since the double add sentinel is first set there.

Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: mgorman@suse.de
Cc: riel@surriel.com
Link: https://lkml.kernel.org/r/20190715102508.32434-3-valentin.schneider@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 kernel/sched/fair.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index f0c488015649..fd391fc00ed8 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2495,7 +2495,7 @@ void task_numa_work(struct callback_head *work)
 
 	SCHED_WARN_ON(p != container_of(work, struct task_struct, numa_work));
 
-	work->next = work; /* protect against double add */
+	work->next = work;
 	/*
 	 * Who cares about NUMA placement when they're dying.
 	 *
@@ -2639,12 +2639,15 @@ void init_numa_balancing(unsigned long clone_flags, struct task_struct *p)
 	p->node_stamp			= 0;
 	p->numa_scan_seq		= mm ? mm->numa_scan_seq : 0;
 	p->numa_scan_period		= sysctl_numa_balancing_scan_delay;
+	/* Protect against double add, see task_tick_numa and task_numa_work */
 	p->numa_work.next		= &p->numa_work;
 	p->numa_faults			= NULL;
 	RCU_INIT_POINTER(p->numa_group, NULL);
 	p->last_task_numa_placement	= 0;
 	p->last_sum_exec_runtime	= 0;
 
+	init_task_work(&p->numa_work, task_numa_work);
+
 	/* New address space, reset the preferred nid */
 	if (!(clone_flags & CLONE_VM)) {
 		p->numa_preferred_nid = NUMA_NO_NODE;
@@ -2693,10 +2696,8 @@ static void task_tick_numa(struct rq *rq, struct task_struct *curr)
 			curr->numa_scan_period = task_scan_start(curr);
 		curr->node_stamp += period;
 
-		if (!time_before(jiffies, curr->mm->numa_next_scan)) {
-			init_task_work(work, task_numa_work); /* TODO: move this into sched_fork() */
+		if (!time_before(jiffies, curr->mm->numa_next_scan))
 			task_work_add(curr, work, true);
-		}
 	}
 }
 

[tip:sched/core] sched/fair: Change task_numa_work() storage to static

[tip-bot for Valentin Schneider]

Commit-ID:  9434f9f5d117302cc7ddf038e7879f6871dc7a81
Gitweb:     https://git.kernel.org/tip/9434f9f5d117302cc7ddf038e7879f6871dc7a81
Author:     Valentin Schneider <valentin.schneider@arm.com>
AuthorDate: Mon, 15 Jul 2019 11:25:08 +0100
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Thu, 25 Jul 2019 15:51:52 +0200

sched/fair: Change task_numa_work() storage to static

There are no callers outside of fair.c.

Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: mgorman@suse.de
Cc: riel@surriel.com
Link: https://lkml.kernel.org/r/20190715102508.32434-4-valentin.schneider@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 kernel/sched/fair.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index fd391fc00ed8..b5546a15206c 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2482,7 +2482,7 @@ static void reset_ptenuma_scan(struct task_struct *p)
  * The expensive part of numa migration is done from task_work context.
  * Triggered from task_tick_numa().
  */
-void task_numa_work(struct callback_head *work)
+static void task_numa_work(struct callback_head *work)
 {
 	unsigned long migrate, next_scan, now = jiffies;
 	struct task_struct *p = current;