sched: Drop the rq argument to sched_class::select_task_rq()

In preparation of calling select_task_rq() without rq->lock held, drop the dependency on the rq argument. Reviewed-by: Frank Rowand <frank.rowand@am.sony.com> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Link: http://lkml.kernel.org/r/20110405152729.031077745@chello.nl Signed-off-by: Ingo Molnar <mingo@elte.hu>

sched: Drop the rq argument to sched_class::select_task_rq()
In preparation of calling select_task_rq() without rq->lock held, drop the dependency on the rq argument. Reviewed-by: Frank Rowand <frank.rowand@am.sony.com> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Link: http://lkml.kernel.org/r/20110405152729.031077745@chello.nl Signed-off-by: Ingo Molnar <mingo@elte.hu>
Peter Zijlstra · Ingo Molnar
1 parent 013fdb8086
Showing 6 changed files with 41 additions and 27 deletions Side-by-side Diff
include/linux/sched.h
kernel/sched.c
kernel/sched_fair.c
kernel/sched_idletask.c
kernel/sched_rt.c
kernel/sched_stoptask.c
@@ -1067,8 +1067,7 @@
 	void (*put_prev_task) (struct rq *rq, struct task_struct *p);
  
 #ifdef CONFIG_SMP
-	int  (*select_task_rq)(struct rq *rq, struct task_struct *p,
-			       int sd_flag, int flags);
+	int  (*select_task_rq)(struct task_struct *p, int sd_flag, int flags);
  
 	void (*pre_schedule) (struct rq *this_rq, struct task_struct *task);
 	void (*post_schedule) (struct rq *this_rq);
@@ -2195,13 +2195,15 @@
  * The task's runqueue lock must be held.
  * Returns true if you have to wait for migration thread.
  */
-static bool migrate_task(struct task_struct *p, struct rq *rq)
+static bool need_migrate_task(struct task_struct *p)
 {
 	/*
 	 * If the task is not on a runqueue (and not running), then
 	 * the next wake-up will properly place the task.
 	 */
-	return p->on_rq || task_running(rq, p);
+	bool running = p->on_rq || p->on_cpu;
+	smp_rmb(); /* finish_lock_switch() */
+	return running;
 }
  
 /*
  
@@ -2376,9 +2378,9 @@
  * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
  */
 static inline
-int select_task_rq(struct rq *rq, struct task_struct *p, int sd_flags, int wake_flags)
+int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
 {
-	int cpu = p->sched_class->select_task_rq(rq, p, sd_flags, wake_flags);
+	int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
  
 	/*
 	 * In order not to call set_task_cpu() on a blocking task we need
@@ -2533,7 +2535,7 @@
 		en_flags |= ENQUEUE_WAKING;
 	}
  
-	cpu = select_task_rq(rq, p, SD_BALANCE_WAKE, wake_flags);
+	cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
 	if (cpu != orig_cpu)
 		set_task_cpu(p, cpu);
 	__task_rq_unlock(rq);
@@ -2744,7 +2746,7 @@
 	 * We set TASK_WAKING so that select_task_rq() can drop rq->lock
 	 * without people poking at ->cpus_allowed.
 	 */
-	cpu = select_task_rq(rq, p, SD_BALANCE_FORK, 0);
+	cpu = select_task_rq(p, SD_BALANCE_FORK, 0);
 	set_task_cpu(p, cpu);
  
 	p->state = TASK_RUNNING;
@@ -3474,7 +3476,7 @@
 	int dest_cpu;
  
 	rq = task_rq_lock(p, &flags);
-	dest_cpu = p->sched_class->select_task_rq(rq, p, SD_BALANCE_EXEC, 0);
+	dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
 	if (dest_cpu == smp_processor_id())
 		goto unlock;
  
@@ -3482,7 +3484,7 @@
 	 * select_task_rq() can race against ->cpus_allowed
 	 */
 	if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed) &&
-	    likely(cpu_active(dest_cpu)) && migrate_task(p, rq)) {
+	    likely(cpu_active(dest_cpu)) && need_migrate_task(p)) {
 		struct migration_arg arg = { p, dest_cpu };
  
 		task_rq_unlock(rq, &flags);
@@ -5911,7 +5913,7 @@
 		goto out;
  
 	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
-	if (migrate_task(p, rq)) {
+	if (need_migrate_task(p)) {
 		struct migration_arg arg = { p, dest_cpu };
 		/* Need help from migration thread: drop lock and wait. */
 		__task_rq_unlock(rq);
@@ -1657,7 +1657,7 @@
  * preempt must be disabled.
  */
 static int
-select_task_rq_fair(struct rq *rq, struct task_struct *p, int sd_flag, int wake_flags)
+select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
 {
 	struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL;
 	int cpu = smp_processor_id();
@@ -7,7 +7,7 @@
  
 #ifdef CONFIG_SMP
 static int
-select_task_rq_idle(struct rq *rq, struct task_struct *p, int sd_flag, int flags)
+select_task_rq_idle(struct task_struct *p, int sd_flag, int flags)
 {
 	return task_cpu(p); /* IDLE tasks as never migrated */
 }
@@ -977,13 +977,23 @@
 static int find_lowest_rq(struct task_struct *task);
  
 static int
-select_task_rq_rt(struct rq *rq, struct task_struct *p, int sd_flag, int flags)
+select_task_rq_rt(struct task_struct *p, int sd_flag, int flags)
 {
+	struct task_struct *curr;
+	struct rq *rq;
+	int cpu;
+
 	if (sd_flag != SD_BALANCE_WAKE)
 		return smp_processor_id();
  
+	cpu = task_cpu(p);
+	rq = cpu_rq(cpu);
+
+	rcu_read_lock();
+	curr = ACCESS_ONCE(rq->curr); /* unlocked access */
+
 	/*
-	 * If the current task is an RT task, then
+	 * If the current task on @p's runqueue is an RT task, then
 	 * try to see if we can wake this RT task up on another
 	 * runqueue. Otherwise simply start this RT task
 	 * on its current runqueue.
  
  
  
  
  
@@ -997,21 +1007,25 @@
 	 * lock?
 	 *
 	 * For equal prio tasks, we just let the scheduler sort it out.
+	 *
+	 * Otherwise, just let it ride on the affined RQ and the
+	 * post-schedule router will push the preempted task away
+	 *
+	 * This test is optimistic, if we get it wrong the load-balancer
+	 * will have to sort it out.
 	 */
-	if (unlikely(rt_task(rq->curr)) &&
-	    (rq->curr->rt.nr_cpus_allowed < 2 ||
-	     rq->curr->prio < p->prio) &&
+	if (curr && unlikely(rt_task(curr)) &&
+	    (curr->rt.nr_cpus_allowed < 2 ||
+	     curr->prio < p->prio) &&
 	    (p->rt.nr_cpus_allowed > 1)) {
-		int cpu = find_lowest_rq(p);
+		int target = find_lowest_rq(p);
  
-		return (cpu == -1) ? task_cpu(p) : cpu;
+		if (target != -1)
+			cpu = target;
 	}
+	rcu_read_unlock();
  
-	/*
-	 * Otherwise, just let it ride on the affined RQ and the
-	 * post-schedule router will push the preempted task away
-	 */
-	return task_cpu(p);
+	return cpu;
 }
  
 static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
@@ -9,8 +9,7 @@
  
 #ifdef CONFIG_SMP
 static int
-select_task_rq_stop(struct rq *rq, struct task_struct *p,
-		    int sd_flag, int flags)
+select_task_rq_stop(struct task_struct *p, int sd_flag, int flags)
 {
 	return task_cpu(p); /* stop tasks as never migrate */
 }
...	...	@@ -1067,8 +1067,7 @@
1067	1067	void (put_prev_task) (struct rq rq, struct task_struct *p);
1068	1068
1069	1069	#ifdef CONFIG_SMP
1070		- int (select_task_rq)(struct rq rq, struct task_struct *p,
1071		- int sd_flag, int flags);
	1070	+ int (select_task_rq)(struct task_struct p, int sd_flag, int flags);
1072	1071
1073	1072	void (pre_schedule) (struct rq this_rq, struct task_struct *task);
1074	1073	void (post_schedule) (struct rq this_rq);
...	...	@@ -2195,13 +2195,15 @@
2195	2195	* The task's runqueue lock must be held.
2196	2196	* Returns true if you have to wait for migration thread.
2197	2197	*/
2198		-static bool migrate_task(struct task_struct p, struct rq rq)
	2198	+static bool need_migrate_task(struct task_struct *p)
2199	2199	{
2200	2200	/*
2201	2201	* If the task is not on a runqueue (and not running), then
2202	2202	* the next wake-up will properly place the task.
2203	2203	*/
2204		- return p->on_rq \|\| task_running(rq, p);
	2204	+ bool running = p->on_rq \|\| p->on_cpu;
	2205	+ smp_rmb(); /* finish_lock_switch() */
	2206	+ return running;
2205	2207	}
2206	2208
2207	2209	/*
2208	2210
...	...	@@ -2376,9 +2378,9 @@
2376	2378	* The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
2377	2379	*/
2378	2380	static inline
2379		-int select_task_rq(struct rq rq, struct task_struct p, int sd_flags, int wake_flags)
	2381	+int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
2380	2382	{
2381		- int cpu = p->sched_class->select_task_rq(rq, p, sd_flags, wake_flags);
	2383	+ int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
2382	2384
2383	2385	/*
2384	2386	* In order not to call set_task_cpu() on a blocking task we need
...	...	@@ -2533,7 +2535,7 @@
2533	2535	en_flags \|= ENQUEUE_WAKING;
2534	2536	}
2535	2537
2536		- cpu = select_task_rq(rq, p, SD_BALANCE_WAKE, wake_flags);
	2538	+ cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
2537	2539	if (cpu != orig_cpu)
2538	2540	set_task_cpu(p, cpu);
2539	2541	__task_rq_unlock(rq);
...	...	@@ -2744,7 +2746,7 @@
2744	2746	* We set TASK_WAKING so that select_task_rq() can drop rq->lock
2745	2747	* without people poking at ->cpus_allowed.
2746	2748	*/
2747		- cpu = select_task_rq(rq, p, SD_BALANCE_FORK, 0);
	2749	+ cpu = select_task_rq(p, SD_BALANCE_FORK, 0);
2748	2750	set_task_cpu(p, cpu);
2749	2751
2750	2752	p->state = TASK_RUNNING;
...	...	@@ -3474,7 +3476,7 @@
3474	3476	int dest_cpu;
3475	3477
3476	3478	rq = task_rq_lock(p, &flags);
3477		- dest_cpu = p->sched_class->select_task_rq(rq, p, SD_BALANCE_EXEC, 0);
	3479	+ dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
3478	3480	if (dest_cpu == smp_processor_id())
3479	3481	goto unlock;
3480	3482
...	...	@@ -3482,7 +3484,7 @@
3482	3484	* select_task_rq() can race against ->cpus_allowed
3483	3485	*/
3484	3486	if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed) &&
3485		- likely(cpu_active(dest_cpu)) && migrate_task(p, rq)) {
	3487	+ likely(cpu_active(dest_cpu)) && need_migrate_task(p)) {
3486	3488	struct migration_arg arg = { p, dest_cpu };
3487	3489
3488	3490	task_rq_unlock(rq, &flags);
...	...	@@ -5911,7 +5913,7 @@
5911	5913	goto out;
5912	5914
5913	5915	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
5914		- if (migrate_task(p, rq)) {
	5916	+ if (need_migrate_task(p)) {
5915	5917	struct migration_arg arg = { p, dest_cpu };
5916	5918	/* Need help from migration thread: drop lock and wait. */
5917	5919	__task_rq_unlock(rq);
...	...	@@ -1657,7 +1657,7 @@
1657	1657	* preempt must be disabled.
1658	1658	*/
1659	1659	static int
1660		-select_task_rq_fair(struct rq rq, struct task_struct p, int sd_flag, int wake_flags)
	1660	+select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
1661	1661	{
1662	1662	struct sched_domain tmp, affine_sd = NULL, *sd = NULL;
1663	1663	int cpu = smp_processor_id();
...	...	@@ -7,7 +7,7 @@
7	7
8	8	#ifdef CONFIG_SMP
9	9	static int
10		-select_task_rq_idle(struct rq rq, struct task_struct p, int sd_flag, int flags)
	10	+select_task_rq_idle(struct task_struct *p, int sd_flag, int flags)
11	11	{
12	12	return task_cpu(p); /* IDLE tasks as never migrated */
13	13	}
...	...	@@ -977,13 +977,23 @@
977	977	static int find_lowest_rq(struct task_struct *task);
978	978
979	979	static int
980		-select_task_rq_rt(struct rq rq, struct task_struct p, int sd_flag, int flags)
	980	+select_task_rq_rt(struct task_struct *p, int sd_flag, int flags)
981	981	{
	982	+ struct task_struct *curr;
	983	+ struct rq *rq;
	984	+ int cpu;
	985	+
982	986	if (sd_flag != SD_BALANCE_WAKE)
983	987	return smp_processor_id();
984	988
	989	+ cpu = task_cpu(p);
	990	+ rq = cpu_rq(cpu);
	991	+
	992	+ rcu_read_lock();
	993	+ curr = ACCESS_ONCE(rq->curr); /* unlocked access */
	994	+
985	995	/*
986		- * If the current task is an RT task, then
	996	+ * If the current task on @p's runqueue is an RT task, then
987	997	* try to see if we can wake this RT task up on another
988	998	* runqueue. Otherwise simply start this RT task
989	999	* on its current runqueue.
990	1000
991	1001
992	1002
993	1003
994	1004
...	...	@@ -997,21 +1007,25 @@
997	1007	* lock?
998	1008	*
999	1009	* For equal prio tasks, we just let the scheduler sort it out.
	1010	+ *
	1011	+ * Otherwise, just let it ride on the affined RQ and the
	1012	+ * post-schedule router will push the preempted task away
	1013	+ *
	1014	+ * This test is optimistic, if we get it wrong the load-balancer
	1015	+ * will have to sort it out.
1000	1016	*/
1001		- if (unlikely(rt_task(rq->curr)) &&
1002		- (rq->curr->rt.nr_cpus_allowed < 2 \|\|
1003		- rq->curr->prio < p->prio) &&
	1017	+ if (curr && unlikely(rt_task(curr)) &&
	1018	+ (curr->rt.nr_cpus_allowed < 2 \|\|
	1019	+ curr->prio < p->prio) &&
1004	1020	(p->rt.nr_cpus_allowed > 1)) {
1005		- int cpu = find_lowest_rq(p);
	1021	+ int target = find_lowest_rq(p);
1006	1022
1007		- return (cpu == -1) ? task_cpu(p) : cpu;
	1023	+ if (target != -1)
	1024	+ cpu = target;
1008	1025	}
	1026	+ rcu_read_unlock();
1009	1027
1010		- /*
1011		- * Otherwise, just let it ride on the affined RQ and the
1012		- * post-schedule router will push the preempted task away
1013		- */
1014		- return task_cpu(p);
	1028	+ return cpu;
1015	1029	}
1016	1030
1017	1031	static void check_preempt_equal_prio(struct rq rq, struct task_struct p)
...	...	@@ -9,8 +9,7 @@
9	9
10	10	#ifdef CONFIG_SMP
11	11	static int
12		-select_task_rq_stop(struct rq rq, struct task_struct p,
13		- int sd_flag, int flags)
	12	+select_task_rq_stop(struct task_struct *p, int sd_flag, int flags)
14	13	{
15	14	return task_cpu(p); /* stop tasks as never migrate */
16	15	}