[PATCH] sched/core: remove get_cpu() from sched_fork()

[PATCH] sched/core: remove get_cpu() from sched_fork()

[Sebastian Andrzej Siewior]

get_cpu() disables preemption for the entire sched_fork() function.
This get_cpu() was introduced in commit dd41f596cda0 ("sched: cfs core
code") which also invoked sched_balance_self() and this function
required preemption do be off.
Today, sched_balance_self() seems to be moved to ->task_fork callback
which is invoked while the ->pi_lock is held.
set_load_weight() could invoke reweight_task() which then via $callchain
might end up in smp_processor_id() but since `update_load' is false
this won't happen.
I didn't find any this_cpu*() or similar usage during the initialisation
of the task_struct.
The `cpu' value (from get_cpu()) is only used later in __set_task_cpu()
while the ->pi_lock lock is held.

Based on this it is possible to remove get_cpu() and use
smp_processor_id() for the `cpu' variable without breaking anything.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 kernel/sched/core.c | 13 ++++---------
 1 file changed, 4 insertions(+), 9 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 78d8facba456..b6ec6ebe620d 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2318,7 +2318,6 @@ static inline void init_schedstats(void) {}
 int sched_fork(unsigned long clone_flags, struct task_struct *p)
 {
 	unsigned long flags;
-	int cpu = get_cpu();
 
 	__sched_fork(clone_flags, p);
 	/*
@@ -2354,14 +2353,12 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 		p->sched_reset_on_fork = 0;
 	}
 
-	if (dl_prio(p->prio)) {
-		put_cpu();
+	if (dl_prio(p->prio))
 		return -EAGAIN;
-	} else if (rt_prio(p->prio)) {
+	else if (rt_prio(p->prio))
 		p->sched_class = &rt_sched_class;
-	} else {
+	else
 		p->sched_class = &fair_sched_class;
-	}
 
 	init_entity_runnable_average(&p->se);
 
@@ -2377,7 +2374,7 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 	 * We're setting the CPU for the first time, we don't migrate,
 	 * so use __set_task_cpu().
 	 */
-	__set_task_cpu(p, cpu);
+	__set_task_cpu(p, smp_processor_id());
 	if (p->sched_class->task_fork)
 		p->sched_class->task_fork(p);
 	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
@@ -2394,8 +2391,6 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 	plist_node_init(&p->pushable_tasks, MAX_PRIO);
 	RB_CLEAR_NODE(&p->pushable_dl_tasks);
 #endif
-
-	put_cpu();
 	return 0;
 }
 
-- 
2.18.0

Re: [PATCH] sched/core: remove get_cpu() from sched_fork()

[Peter Zijlstra]

On Fri, Jul 06, 2018 at 03:06:15PM +0200, Sebastian Andrzej Siewior wrote:
> Based on this it is possible to remove get_cpu() and use
> smp_processor_id() for the `cpu' variable without breaking anything.

Almost.. I think, see init_task_preempt_count().

Re: [PATCH] sched/core: remove get_cpu() from sched_fork()

[Sebastian Andrzej Siewior]

On 2018-07-06 15:18:08 [+0200], Peter Zijlstra wrote:
> On Fri, Jul 06, 2018 at 03:06:15PM +0200, Sebastian Andrzej Siewior wrote:
> > Based on this it is possible to remove get_cpu() and use
> > smp_processor_id() for the `cpu' variable without breaking anything.
> 
> Almost.. I think, see init_task_preempt_count().

 #define init_task_preempt_count(p) do { \
         task_thread_info(p)->preempt_count = FORK_PREEMPT_COUNT; \
 } while (0)

and task_thread_info() references p only. It looks good, what did I
miss?

Sebastian

Re: [PATCH] sched/core: remove get_cpu() from sched_fork()

[Peter Zijlstra]

On Fri, Jul 06, 2018 at 03:32:00PM +0200, Sebastian Andrzej Siewior wrote:
> On 2018-07-06 15:18:08 [+0200], Peter Zijlstra wrote:
> > On Fri, Jul 06, 2018 at 03:06:15PM +0200, Sebastian Andrzej Siewior wrote:
> > > Based on this it is possible to remove get_cpu() and use
> > > smp_processor_id() for the `cpu' variable without breaking anything.
> > 
> > Almost.. I think, see init_task_preempt_count().
> 
>  #define init_task_preempt_count(p) do { \
>          task_thread_info(p)->preempt_count = FORK_PREEMPT_COUNT; \
>  } while (0)
> 
> and task_thread_info() references p only. It looks good, what did I
> miss?

Argh, dammit, I read that with p == current :/

Then yes, looks ok.

[tip:sched/core] sched/core: Remove get_cpu() from sched_fork()

[tip-bot for Sebastian Andrzej Siewior]

Commit-ID:  af0fffd9300b97d8875aa745bc78e2f6fdb3c1f0
Gitweb:     https://git.kernel.org/tip/af0fffd9300b97d8875aa745bc78e2f6fdb3c1f0
Author:     Sebastian Andrzej Siewior <bigeasy@linutronix.de>
AuthorDate: Fri, 6 Jul 2018 15:06:15 +0200
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Mon, 16 Jul 2018 00:16:29 +0200

sched/core: Remove get_cpu() from sched_fork()

get_cpu() disables preemption for the entire sched_fork() function.
This get_cpu() was introduced in commit:

  dd41f596cda0 ("sched: cfs core code")

... which also invoked sched_balance_self() and this function
required preemption do be off.

Today, sched_balance_self() seems to be moved to ->task_fork callback
which is invoked while the ->pi_lock is held.

set_load_weight() could invoke reweight_task() which then via $callchain
might end up in smp_processor_id() but since `update_load' is false
this won't happen.

I didn't find any this_cpu*() or similar usage during the initialisation
of the task_struct.

The `cpu' value (from get_cpu()) is only used later in __set_task_cpu()
while the ->pi_lock lock is held.

Based on this it is possible to remove get_cpu() and use
smp_processor_id() for the `cpu' variable without breaking anything.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
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>
Link: http://lkml.kernel.org/r/20180706130615.g2ex2kmfu5kcvlq6@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 kernel/sched/core.c | 13 ++++---------
 1 file changed, 4 insertions(+), 9 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index ba6bb805693a..c3cf7d992159 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2294,7 +2294,6 @@ static inline void init_schedstats(void) {}
 int sched_fork(unsigned long clone_flags, struct task_struct *p)
 {
 	unsigned long flags;
-	int cpu = get_cpu();
 
 	__sched_fork(clone_flags, p);
 	/*
@@ -2330,14 +2329,12 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 		p->sched_reset_on_fork = 0;
 	}
 
-	if (dl_prio(p->prio)) {
-		put_cpu();
+	if (dl_prio(p->prio))
 		return -EAGAIN;
-	} else if (rt_prio(p->prio)) {
+	else if (rt_prio(p->prio))
 		p->sched_class = &rt_sched_class;
-	} else {
+	else
 		p->sched_class = &fair_sched_class;
-	}
 
 	init_entity_runnable_average(&p->se);
 
@@ -2353,7 +2350,7 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 	 * We're setting the CPU for the first time, we don't migrate,
 	 * so use __set_task_cpu().
 	 */
-	__set_task_cpu(p, cpu);
+	__set_task_cpu(p, smp_processor_id());
 	if (p->sched_class->task_fork)
 		p->sched_class->task_fork(p);
 	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
@@ -2370,8 +2367,6 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 	plist_node_init(&p->pushable_tasks, MAX_PRIO);
 	RB_CLEAR_NODE(&p->pushable_dl_tasks);
 #endif
-
-	put_cpu();
 	return 0;
 }