Skip to contentSkip to conversation

sched/fair: Remove idle_balance() declaration in sched.h

PATCH In review Daniel Lezcano 3 participants 6 messages Feb 11, 2014 – Feb 21, 2014 Scheduler

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sched/fair: Remove idle_balance() declaration in sched.h

The idle_balance() function is called within a #ifdef CONFIG_SMP section.

Remove its declaration in sched.h for !CONFIG_SMP because it is pointless.

---
 kernel/sched/sched.h |    6 ------
 1 file changed, 6 deletions(-)
kernel/sched/sched.h +0 -6
index 1bf34c2..0eda78b 100644
@@ -1168,12 +1168,6 @@ extern int idle_balance(struct rq *this_rq);
 extern void idle_enter_fair(struct rq *this_rq);
 extern void idle_exit_fair(struct rq *this_rq);
 
-#else	/* CONFIG_SMP */
-
-static inline void idle_balance(int cpu, struct rq *rq)
-{
-}
-
 #endif
 
 extern void sysrq_sched_debug_show(void);
Signature · 1 line
1.7.9.5
maintainer
#1 lore raw
Quoted 5 lines · Daniel Lezcano

The idle_balance() function is called within a #ifdef CONFIG_SMP section.

Remove its declaration in sched.h for !CONFIG_SMP because it is pointless.

The below again makes a horrible mess of idle_balance() -- which you tried to clean up.. but it does rid us of some #ifdef goo.

Hmmm?

---
 kernel/sched/fair.c  | 47 +++++++++++++++++++++++++++++------------------
 kernel/sched/sched.h |  7 -------
 2 files changed, 29 insertions(+), 25 deletions(-)
kernel/sched/fair.c +29 -18
index 235cfa7ad8fc..d168c968195b 100644
@@ -2374,13 +2374,13 @@ static inline void __update_group_entity_contrib(struct sched_entity *se)
 		se->avg.load_avg_contrib >>= NICE_0_SHIFT;
 	}
 }
-#else
+#else /* CONFIG_FAIR_GROUP_SCHED */
 static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq,
 						 int force_update) {}
 static inline void __update_tg_runnable_avg(struct sched_avg *sa,
 						  struct cfs_rq *cfs_rq) {}
 static inline void __update_group_entity_contrib(struct sched_entity *se) {}
-#endif
+#endif /* CONFIG_FAIR_GROUP_SCHED */
 
 static inline void __update_task_entity_contrib(struct sched_entity *se)
 {
@@ -2571,6 +2571,8 @@ void idle_exit_fair(struct rq *this_rq)
 	update_rq_runnable_avg(this_rq, 0);
 }
 
+static int idle_balance(struct rq *this_rq);
+
 #else /* CONFIG_SMP */
 
 static inline void update_entity_load_avg(struct sched_entity *se,
@@ -2584,6 +2586,12 @@ static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
 					   int sleep) {}
 static inline void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq,
 					      int force_update) {}
+
+static inline int idle_balance(struct rq *rq)
+{
+	return 0;
+}
+
 #endif /* CONFIG_SMP */
 
 static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
@@ -4677,7 +4685,7 @@ pick_next_task_fair(struct rq *rq, struct task_struct *prev)
 	struct sched_entity *se;
 	struct task_struct *p;
 
-again: __maybe_unused
+again:
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	if (!cfs_rq->nr_running)
 		goto idle;
@@ -4775,18 +4783,8 @@ again: __maybe_unused
 	return p;
 
 idle:
-#ifdef CONFIG_SMP
-	idle_enter_fair(rq);
-	/*
-	 * We must set idle_stamp _before_ calling idle_balance(), such that we
-	 * measure the duration of idle_balance() as idle time.
-	 */
-	rq->idle_stamp = rq_clock(rq);
-	if (idle_balance(rq)) { /* drops rq->lock */
-		rq->idle_stamp = 0;
+	if (idle_balance(rq)) /* drops rq->lock */
 		goto again;
-	}
-#endif
 
 	return NULL;
 }
@@ -6634,7 +6632,7 @@ static int load_balance(int this_cpu, struct rq *this_rq,
  * idle_balance is called by schedule() if this_cpu is about to become
  * idle. Attempts to pull tasks from other CPUs.
  */
-int idle_balance(struct rq *this_rq)
+static int idle_balance(struct rq *this_rq)
 {
 	struct sched_domain *sd;
 	int pulled_task = 0;
@@ -6642,8 +6640,15 @@ int idle_balance(struct rq *this_rq)
 	u64 curr_cost = 0;
 	int this_cpu = this_rq->cpu;
 
+	idle_enter_fair(this_rq);
+	/*
+	 * We must set idle_stamp _before_ calling idle_balance(), such that we
+	 * measure the duration of idle_balance() as idle time.
+	 */
+	this_rq->idle_stamp = rq_clock(this_rq);
+
 	if (this_rq->avg_idle < sysctl_sched_migration_cost)
-		return 0;
+		goto out;
 
 	/*
 	 * Drop the rq->lock, but keep IRQ/preempt disabled.
@@ -6692,8 +6697,10 @@ int idle_balance(struct rq *this_rq)
 	 * While browsing the domains, we released the rq lock.
 	 * A task could have be enqueued in the meantime
 	 */
-	if (this_rq->nr_running && !pulled_task)
-		return 1;
+	if (this_rq->nr_running && !pulled_task) {
+		pulled_task = 1;
+		goto out;
+	}
 
 	if (pulled_task || time_after(jiffies, this_rq->next_balance)) {
 		/*
@@ -6706,6 +6713,10 @@ int idle_balance(struct rq *this_rq)
 	if (curr_cost > this_rq->max_idle_balance_cost)
 		this_rq->max_idle_balance_cost = curr_cost;
 
+out:
+	if (pulled_task)
+		this_rq->idle_stamp = 0;
+
 	return pulled_task;
 }
 
kernel/sched/sched.h +0 -7
index 1bf34c257d3b..92018f9821e8 100644
@@ -1163,17 +1163,10 @@ extern const struct sched_class idle_sched_class;
 extern void update_group_power(struct sched_domain *sd, int cpu);
 
 extern void trigger_load_balance(struct rq *rq);
-extern int idle_balance(struct rq *this_rq);
 
 extern void idle_enter_fair(struct rq *this_rq);
 extern void idle_exit_fair(struct rq *this_rq);
 
-#else	/* CONFIG_SMP */
-
-static inline void idle_balance(int cpu, struct rq *rq)
-{
-}
-
 #endif
 
 extern void sysrq_sched_debug_show(void);
#2 lore raw
Quoted 11 lines · Peter Zijlstra
Quoted 5 lines · Daniel Lezcano

The idle_balance() function is called within a #ifdef CONFIG_SMP section.

Remove its declaration in sched.h for !CONFIG_SMP because it is pointless.

The below again makes a horrible mess of idle_balance() -- which you tried to clean up.. but it does rid us of some #ifdef goo.

Hmmm?

Yes, it sounds ok.

Why is idle_enter_fair() called unconditionally in idle_balance() ?
Isn't the call in pick_next_task_idle enough ? Shouldn't be called when we will go to idle effectively ?

If I am not wrong idle_enter_fair() is called from idle_balance() but a task may be pulled, so we the next task won't be the idle task and idle_exit_fair won't be called at put_prev_task.

May be I missed this change which was done at purpose in the previous patchset you sent...

Quoted 149 lines
---
  kernel/sched/fair.c  | 47 +++++++++++++++++++++++++++++------------------
  kernel/sched/sched.h |  7 -------
  2 files changed, 29 insertions(+), 25 deletions(-)
kernel/sched/fair.c +29 -18
index 235cfa7ad8fc..d168c968195b 100644
@@ -2374,13 +2374,13 @@ static inline void __update_group_entity_contrib(struct sched_entity *se)
  		se->avg.load_avg_contrib >>= NICE_0_SHIFT;
  	}
  }
-#else
+#else /* CONFIG_FAIR_GROUP_SCHED */
  static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq,
  						 int force_update) {}
  static inline void __update_tg_runnable_avg(struct sched_avg *sa,
  						  struct cfs_rq *cfs_rq) {}
  static inline void __update_group_entity_contrib(struct sched_entity *se) {}
-#endif
+#endif /* CONFIG_FAIR_GROUP_SCHED */
 
  static inline void __update_task_entity_contrib(struct sched_entity *se)
  {
@@ -2571,6 +2571,8 @@ void idle_exit_fair(struct rq *this_rq)
  	update_rq_runnable_avg(this_rq, 0);
  }
 
+static int idle_balance(struct rq *this_rq);
+
  #else /* CONFIG_SMP */
 
  static inline void update_entity_load_avg(struct sched_entity *se,
@@ -2584,6 +2586,12 @@ static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
  					   int sleep) {}
  static inline void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq,
  					      int force_update) {}
+
+static inline int idle_balance(struct rq *rq)
+{
+	return 0;
+}
+
  #endif /* CONFIG_SMP */
 
  static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
@@ -4677,7 +4685,7 @@ pick_next_task_fair(struct rq *rq, struct task_struct *prev)
  	struct sched_entity *se;
  	struct task_struct *p;
 
-again: __maybe_unused
+again:
  #ifdef CONFIG_FAIR_GROUP_SCHED
  	if (!cfs_rq->nr_running)
  		goto idle;
@@ -4775,18 +4783,8 @@ again: __maybe_unused
  	return p;
 
  idle:
-#ifdef CONFIG_SMP
-	idle_enter_fair(rq);
-	/*
-	 * We must set idle_stamp _before_ calling idle_balance(), such that we
-	 * measure the duration of idle_balance() as idle time.
-	 */
-	rq->idle_stamp = rq_clock(rq);
-	if (idle_balance(rq)) { /* drops rq->lock */
-		rq->idle_stamp = 0;
+	if (idle_balance(rq)) /* drops rq->lock */
  		goto again;
-	}
-#endif
 
  	return NULL;
  }
@@ -6634,7 +6632,7 @@ static int load_balance(int this_cpu, struct rq *this_rq,
   * idle_balance is called by schedule() if this_cpu is about to become
   * idle. Attempts to pull tasks from other CPUs.
   */
-int idle_balance(struct rq *this_rq)
+static int idle_balance(struct rq *this_rq)
  {
  	struct sched_domain *sd;
  	int pulled_task = 0;
@@ -6642,8 +6640,15 @@ int idle_balance(struct rq *this_rq)
  	u64 curr_cost = 0;
  	int this_cpu = this_rq->cpu;
 
+	idle_enter_fair(this_rq);
+	/*
+	 * We must set idle_stamp _before_ calling idle_balance(), such that we
+	 * measure the duration of idle_balance() as idle time.
+	 */
+	this_rq->idle_stamp = rq_clock(this_rq);
+
  	if (this_rq->avg_idle < sysctl_sched_migration_cost)
-		return 0;
+		goto out;
 
  	/*
  	 * Drop the rq->lock, but keep IRQ/preempt disabled.
@@ -6692,8 +6697,10 @@ int idle_balance(struct rq *this_rq)
  	 * While browsing the domains, we released the rq lock.
  	 * A task could have be enqueued in the meantime
  	 */
-	if (this_rq->nr_running && !pulled_task)
-		return 1;
+	if (this_rq->nr_running && !pulled_task) {
+		pulled_task = 1;
+		goto out;
+	}
 
  	if (pulled_task || time_after(jiffies, this_rq->next_balance)) {
  		/*
@@ -6706,6 +6713,10 @@ int idle_balance(struct rq *this_rq)
  	if (curr_cost > this_rq->max_idle_balance_cost)
  		this_rq->max_idle_balance_cost = curr_cost;
 
+out:
+	if (pulled_task)
+		this_rq->idle_stamp = 0;
+
  	return pulled_task;
  }
 
kernel/sched/sched.h +0 -7
index 1bf34c257d3b..92018f9821e8 100644
@@ -1163,17 +1163,10 @@ extern const struct sched_class idle_sched_class;
  extern void update_group_power(struct sched_domain *sd, int cpu);
 
  extern void trigger_load_balance(struct rq *rq);
-extern int idle_balance(struct rq *this_rq);
 
  extern void idle_enter_fair(struct rq *this_rq);
  extern void idle_exit_fair(struct rq *this_rq);
 
-#else	/* CONFIG_SMP */
-
-static inline void idle_balance(int cpu, struct rq *rq)
-{
-}
-
  #endif
 
  extern void sysrq_sched_debug_show(void);
Signature · 5 lines
 <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog
maintainer
#3 lore raw
Quoted 19 lines · Daniel Lezcano
Quoted 11 lines · Peter Zijlstra
Quoted 5 lines · Daniel Lezcano

The idle_balance() function is called within a #ifdef CONFIG_SMP section.

Remove its declaration in sched.h for !CONFIG_SMP because it is pointless.

The below again makes a horrible mess of idle_balance() -- which you tried to clean up.. but it does rid us of some #ifdef goo.

Hmmm?

Yes, it sounds ok.

Why is idle_enter_fair() called unconditionally in idle_balance() ? Isn't the call in pick_next_task_idle enough ? Shouldn't be called when we will go to idle effectively ?

Quoted 6 lines

If I am not wrong idle_enter_fair() is called from idle_balance() but a task may be pulled, so we the next task won't be the idle task and idle_exit_fair won't be called at put_prev_task.

May be I missed this change which was done at purpose in the previous patchset you sent...

lkml.kernel.org/r/CAKfTPtAMmhq0tBpu+pk_uzwrfUM-6Fz4Do1yrzc4NRQS4cZ_-A@mail.gmail.com

I seem to be terminally confused on the subject, but Vincent explains it there I think.

#4 lore raw
Quoted 24 lines · Peter Zijlstra
Quoted 23 lines · Daniel Lezcano
Quoted 11 lines · Peter Zijlstra
Quoted 5 lines · Daniel Lezcano

The idle_balance() function is called within a #ifdef CONFIG_SMP section.

Remove its declaration in sched.h for !CONFIG_SMP because it is pointless.

The below again makes a horrible mess of idle_balance() -- which you tried to clean up.. but it does rid us of some #ifdef goo.

Hmmm?

Yes, it sounds ok.

Why is idle_enter_fair() called unconditionally in idle_balance() ? Isn't the call in pick_next_task_idle enough ? Shouldn't be called when we will go to idle effectively ?

If I am not wrong idle_enter_fair() is called from idle_balance() but a task may be pulled, so we the next task won't be the idle task and idle_exit_fair won't be called at put_prev_task.

Quoted 7 lines
Quoted 2 lines

May be I missed this change which was done at purpose in the previous patchset you sent...

lkml.kernel.org/r/CAKfTPtAMmhq0tBpu+pk_uzwrfUM-6Fz4Do1yrzc4NRQS4cZ_-A@mail.gmail.com

I seem to be terminally confused on the subject, but Vincent explains it there I think.

Ok, thanks. I will read again the thread carefully.

-- Daniel

Signature · 5 lines
 <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog
#5 lore raw
In reply to Peter Zijlstra
sched/fair: Remove idle_balance() declaration in sched.h
Commit-ID:  6e83125c6b151afa139c8852c099d6d92954fe3b
Gitweb:     http://git.kernel.org/tip/6e83125c6b151afa139c8852c099d6d92954fe3b
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Tue, 11 Feb 2014 16:11:48 +0100
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Fri, 21 Feb 2014 21:43:17 +0100

sched/fair: Remove idle_balance() declaration in sched.h

Remove idle_balance() from the public life; also reduce some #ifdef clutter by folding the pick_next_task_fair() idle path into idle_balance().

---
 kernel/sched/fair.c  | 47 +++++++++++++++++++++++++++++------------------
 kernel/sched/sched.h |  7 -------
 2 files changed, 29 insertions(+), 25 deletions(-)
kernel/sched/fair.c +29 -18
index 280da89..40c758b 100644
@@ -2374,13 +2374,13 @@ static inline void __update_group_entity_contrib(struct sched_entity *se)
 		se->avg.load_avg_contrib >>= NICE_0_SHIFT;
 	}
 }
-#else
+#else /* CONFIG_FAIR_GROUP_SCHED */
 static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq,
 						 int force_update) {}
 static inline void __update_tg_runnable_avg(struct sched_avg *sa,
 						  struct cfs_rq *cfs_rq) {}
 static inline void __update_group_entity_contrib(struct sched_entity *se) {}
-#endif
+#endif /* CONFIG_FAIR_GROUP_SCHED */
 
 static inline void __update_task_entity_contrib(struct sched_entity *se)
 {
@@ -2571,6 +2571,8 @@ void idle_exit_fair(struct rq *this_rq)
 	update_rq_runnable_avg(this_rq, 0);
 }
 
+static int idle_balance(struct rq *this_rq);
+
 #else /* CONFIG_SMP */
 
 static inline void update_entity_load_avg(struct sched_entity *se,
@@ -2584,6 +2586,12 @@ static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
 					   int sleep) {}
 static inline void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq,
 					      int force_update) {}
+
+static inline int idle_balance(struct rq *rq)
+{
+	return 0;
+}
+
 #endif /* CONFIG_SMP */
 
 static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
@@ -4677,7 +4685,7 @@ pick_next_task_fair(struct rq *rq, struct task_struct *prev)
 	struct sched_entity *se;
 	struct task_struct *p;
 
-again: __maybe_unused
+again:
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	if (!cfs_rq->nr_running)
 		goto idle;
@@ -4775,18 +4783,8 @@ simple:
 	return p;
 
 idle:
-#ifdef CONFIG_SMP
-	idle_enter_fair(rq);
-	/*
-	 * We must set idle_stamp _before_ calling idle_balance(), such that we
-	 * measure the duration of idle_balance() as idle time.
-	 */
-	rq->idle_stamp = rq_clock(rq);
-	if (idle_balance(rq)) { /* drops rq->lock */
-		rq->idle_stamp = 0;
+	if (idle_balance(rq)) /* drops rq->lock */
 		goto again;
-	}
-#endif
 
 	return NULL;
 }
@@ -6634,7 +6632,7 @@ out:
  * idle_balance is called by schedule() if this_cpu is about to become
  * idle. Attempts to pull tasks from other CPUs.
  */
-int idle_balance(struct rq *this_rq)
+static int idle_balance(struct rq *this_rq)
 {
 	struct sched_domain *sd;
 	int pulled_task = 0;
@@ -6642,8 +6640,15 @@ int idle_balance(struct rq *this_rq)
 	u64 curr_cost = 0;
 	int this_cpu = this_rq->cpu;
 
+	idle_enter_fair(this_rq);
+	/*
+	 * We must set idle_stamp _before_ calling idle_balance(), such that we
+	 * measure the duration of idle_balance() as idle time.
+	 */
+	this_rq->idle_stamp = rq_clock(this_rq);
+
 	if (this_rq->avg_idle < sysctl_sched_migration_cost)
-		return 0;
+		goto out;
 
 	/*
 	 * Drop the rq->lock, but keep IRQ/preempt disabled.
@@ -6692,8 +6697,10 @@ int idle_balance(struct rq *this_rq)
 	 * While browsing the domains, we released the rq lock.
 	 * A task could have be enqueued in the meantime
 	 */
-	if (this_rq->nr_running && !pulled_task)
-		return 1;
+	if (this_rq->nr_running && !pulled_task) {
+		pulled_task = 1;
+		goto out;
+	}
 
 	if (pulled_task || time_after(jiffies, this_rq->next_balance)) {
 		/*
@@ -6706,6 +6713,10 @@ int idle_balance(struct rq *this_rq)
 	if (curr_cost > this_rq->max_idle_balance_cost)
 		this_rq->max_idle_balance_cost = curr_cost;
 
+out:
+	if (pulled_task)
+		this_rq->idle_stamp = 0;
+
 	return pulled_task;
 }
 
kernel/sched/sched.h +0 -7
index 1bf34c2..92018f9 100644
@@ -1163,17 +1163,10 @@ extern const struct sched_class idle_sched_class;
 extern void update_group_power(struct sched_domain *sd, int cpu);
 
 extern void trigger_load_balance(struct rq *rq);
-extern int idle_balance(struct rq *this_rq);
 
 extern void idle_enter_fair(struct rq *this_rq);
 extern void idle_exit_fair(struct rq *this_rq);
 
-#else	/* CONFIG_SMP */
-
-static inline void idle_balance(int cpu, struct rq *rq)
-{
-}
-
 #endif
 
 extern void sysrq_sched_debug_show(void);