Merge branch 'sched-hrtimers-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull scheduler updates from Thomas Gleixner:
 "This series of scheduler updates depends on sched/core and timers/core
  branches, which are already in your tree:

   - Scheduler balancing overhaul to plug a hard to trigger race which
     causes an oops in the balancer (Peter Zijlstra)

   - Lockdep updates which are related to the balancing updates (Peter
     Zijlstra)"

* 'sched-hrtimers-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched,lockdep: Employ lock pinning
  lockdep: Implement lock pinning
  lockdep: Simplify lock_release()
  sched: Streamline the task migration locking a little
  sched: Move code around
  sched,dl: Fix sched class hopping CBS hole
  sched, dl: Convert switched_{from, to}_dl() / prio_changed_dl() to balance callbacks
  sched,dl: Remove return value from pull_dl_task()
  sched, rt: Convert switched_{from, to}_rt() / prio_changed_rt() to balance callbacks
  sched,rt: Remove return value from pull_rt_task()
  sched: Allow balance callbacks for check_class_changed()
  sched: Use replace normalize_task() with __sched_setscheduler()
  sched: Replace post_schedule with a balance callback list

include/linux/lockdep.h

@@ -255,6 +255,7 @@ struct held_lock {
 	unsigned int check:1;       /* see lock_acquire() comment */
 	unsigned int hardirqs_off:1;
 	unsigned int references:12;					/* 32 bits */
+	unsigned int pin_count;
 };
 
 /*
@@ -354,6 +355,9 @@ extern void lockdep_set_current_reclaim_state(gfp_t gfp_mask);
 extern void lockdep_clear_current_reclaim_state(void);
 extern void lockdep_trace_alloc(gfp_t mask);
 
+extern void lock_pin_lock(struct lockdep_map *lock);
+extern void lock_unpin_lock(struct lockdep_map *lock);
+
 # define INIT_LOCKDEP				.lockdep_recursion = 0, .lockdep_reclaim_gfp = 0,
 
 #define lockdep_depth(tsk)	(debug_locks ? (tsk)->lockdep_depth : 0)
@@ -368,6 +372,9 @@ extern void lockdep_trace_alloc(gfp_t mask);
 
 #define lockdep_recursing(tsk)	((tsk)->lockdep_recursion)
 
+#define lockdep_pin_lock(l)		lock_pin_lock(&(l)->dep_map)
+#define lockdep_unpin_lock(l)	lock_unpin_lock(&(l)->dep_map)
+
 #else /* !CONFIG_LOCKDEP */
 
 static inline void lockdep_off(void)
@@ -420,6 +427,9 @@ struct lock_class_key { };
 
 #define lockdep_recursing(tsk)			(0)
 
+#define lockdep_pin_lock(l)				do { (void)(l); } while (0)
+#define lockdep_unpin_lock(l)			do { (void)(l); } while (0)
+
 #endif /* !LOCKDEP */
 
 #ifdef CONFIG_LOCK_STAT

kernel/locking/lockdep.c

@@ -3157,6 +3157,7 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
 	hlock->waittime_stamp = 0;
 	hlock->holdtime_stamp = lockstat_clock();
 #endif
+	hlock->pin_count = 0;
 
 	if (check && !mark_irqflags(curr, hlock))
 		return 0;
@@ -3260,26 +3261,6 @@ print_unlock_imbalance_bug(struct task_struct *curr, struct lockdep_map *lock,
 	return 0;
 }
 
-/*
- * Common debugging checks for both nested and non-nested unlock:
- */
-static int check_unlock(struct task_struct *curr, struct lockdep_map *lock,
-			unsigned long ip)
-{
-	if (unlikely(!debug_locks))
-		return 0;
-	/*
-	 * Lockdep should run with IRQs disabled, recursion, head-ache, etc..
-	 */
-	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
-		return 0;
-
-	if (curr->lockdep_depth <= 0)
-		return print_unlock_imbalance_bug(curr, lock, ip);
-
-	return 1;
-}
-
 static int match_held_lock(struct held_lock *hlock, struct lockdep_map *lock)
 {
 	if (hlock->instance == lock)
@@ -3376,31 +3357,35 @@ __lock_set_class(struct lockdep_map *lock, const char *name,
 }
 
 /*
- * Remove the lock to the list of currently held locks in a
- * potentially non-nested (out of order) manner. This is a
- * relatively rare operation, as all the unlock APIs default
- * to nested mode (which uses lock_release()):
+ * Remove the lock to the list of currently held locks - this gets
+ * called on mutex_unlock()/spin_unlock*() (or on a failed
+ * mutex_lock_interruptible()).
+ *
+ * @nested is an hysterical artifact, needs a tree wide cleanup.
  */
 static int
-lock_release_non_nested(struct task_struct *curr,
-			struct lockdep_map *lock, unsigned long ip)
+__lock_release(struct lockdep_map *lock, int nested, unsigned long ip)
 {
+	struct task_struct *curr = current;
 	struct held_lock *hlock, *prev_hlock;
 	unsigned int depth;
 	int i;
 
-	/*
-	 * Check whether the lock exists in the current stack
-	 * of held locks:
-	 */
+	if (unlikely(!debug_locks))
+		return 0;
+
 	depth = curr->lockdep_depth;
 	/*
 	 * So we're all set to release this lock.. wait what lock? We don't
 	 * own any locks, you've been drinking again?
 	 */
-	if (DEBUG_LOCKS_WARN_ON(!depth))
-		return 0;
+	if (DEBUG_LOCKS_WARN_ON(depth <= 0))
+		 return print_unlock_imbalance_bug(curr, lock, ip);
 
+	/*
+	 * Check whether the lock exists in the current stack
+	 * of held locks:
+	 */
 	prev_hlock = NULL;
 	for (i = depth-1; i >= 0; i--) {
 		hlock = curr->held_locks + i;
@@ -3419,6 +3404,8 @@ lock_release_non_nested(struct task_struct *curr,
 	if (hlock->instance == lock)
 		lock_release_holdtime(hlock);
 
+	WARN(hlock->pin_count, "releasing a pinned lock\n");
+
 	if (hlock->references) {
 		hlock->references--;
 		if (hlock->references) {
@@ -3456,91 +3443,66 @@ lock_release_non_nested(struct task_struct *curr,
 	 */
 	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth - 1))
 		return 0;
+
 	return 1;
 }
 
-/*
- * Remove the lock to the list of currently held locks - this gets
- * called on mutex_unlock()/spin_unlock*() (or on a failed
- * mutex_lock_interruptible()). This is done for unlocks that nest
- * perfectly. (i.e. the current top of the lock-stack is unlocked)
- */
-static int lock_release_nested(struct task_struct *curr,
-			       struct lockdep_map *lock, unsigned long ip)
+static int __lock_is_held(struct lockdep_map *lock)
 {
-	struct held_lock *hlock;
-	unsigned int depth;
-
-	/*
-	 * Pop off the top of the lock stack:
-	 */
-	depth = curr->lockdep_depth - 1;
-	hlock = curr->held_locks + depth;
-
-	/*
-	 * Is the unlock non-nested:
-	 */
-	if (hlock->instance != lock || hlock->references)
-		return lock_release_non_nested(curr, lock, ip);
-	curr->lockdep_depth--;
-
-	/*
-	 * No more locks, but somehow we've got hash left over, who left it?
-	 */
-	if (DEBUG_LOCKS_WARN_ON(!depth && (hlock->prev_chain_key != 0)))
-		return 0;
+	struct task_struct *curr = current;
+	int i;
 
-	curr->curr_chain_key = hlock->prev_chain_key;
+	for (i = 0; i < curr->lockdep_depth; i++) {
+		struct held_lock *hlock = curr->held_locks + i;
 
-	lock_release_holdtime(hlock);
+		if (match_held_lock(hlock, lock))
+			return 1;
+	}
 
-#ifdef CONFIG_DEBUG_LOCKDEP
-	hlock->prev_chain_key = 0;
-	hlock->class_idx = 0;
-	hlock->acquire_ip = 0;
-	hlock->irq_context = 0;
-#endif
-	return 1;
+	return 0;
 }
 
-/*
- * Remove the lock to the list of currently held locks - this gets
- * called on mutex_unlock()/spin_unlock*() (or on a failed
- * mutex_lock_interruptible()). This is done for unlocks that nest
- * perfectly. (i.e. the current top of the lock-stack is unlocked)
- */
-static void
-__lock_release(struct lockdep_map *lock, int nested, unsigned long ip)
+static void __lock_pin_lock(struct lockdep_map *lock)
 {
 	struct task_struct *curr = current;
+	int i;
 
-	if (!check_unlock(curr, lock, ip))
+	if (unlikely(!debug_locks))
 		return;
 
-	if (nested) {
-		if (!lock_release_nested(curr, lock, ip))
-			return;
-	} else {
-		if (!lock_release_non_nested(curr, lock, ip))
+	for (i = 0; i < curr->lockdep_depth; i++) {
+		struct held_lock *hlock = curr->held_locks + i;
+
+		if (match_held_lock(hlock, lock)) {
+			hlock->pin_count++;
 			return;
+		}
 	}
 
-	check_chain_key(curr);
+	WARN(1, "pinning an unheld lock\n");
 }
 
-static int __lock_is_held(struct lockdep_map *lock)
+static void __lock_unpin_lock(struct lockdep_map *lock)
 {
 	struct task_struct *curr = current;
 	int i;
 
+	if (unlikely(!debug_locks))
+		return;
+
 	for (i = 0; i < curr->lockdep_depth; i++) {
 		struct held_lock *hlock = curr->held_locks + i;
 
-		if (match_held_lock(hlock, lock))
-			return 1;
+		if (match_held_lock(hlock, lock)) {
+			if (WARN(!hlock->pin_count, "unpinning an unpinned lock\n"))
+				return;
+
+			hlock->pin_count--;
+			return;
+		}
 	}
 
-	return 0;
+	WARN(1, "unpinning an unheld lock\n");
 }
 
 /*
@@ -3639,7 +3601,8 @@ void lock_release(struct lockdep_map *lock, int nested,
 	check_flags(flags);
 	current->lockdep_recursion = 1;
 	trace_lock_release(lock, ip);
-	__lock_release(lock, nested, ip);
+	if (__lock_release(lock, nested, ip))
+		check_chain_key(current);
 	current->lockdep_recursion = 0;
 	raw_local_irq_restore(flags);
 }
@@ -3665,6 +3628,40 @@ int lock_is_held(struct lockdep_map *lock)
 }
 EXPORT_SYMBOL_GPL(lock_is_held);
 
+void lock_pin_lock(struct lockdep_map *lock)
+{
+	unsigned long flags;
+
+	if (unlikely(current->lockdep_recursion))
+		return;
+
+	raw_local_irq_save(flags);
+	check_flags(flags);
+
+	current->lockdep_recursion = 1;
+	__lock_pin_lock(lock);
+	current->lockdep_recursion = 0;
+	raw_local_irq_restore(flags);
+}
+EXPORT_SYMBOL_GPL(lock_pin_lock);
+
+void lock_unpin_lock(struct lockdep_map *lock)
+{
+	unsigned long flags;
+
+	if (unlikely(current->lockdep_recursion))
+		return;
+
+	raw_local_irq_save(flags);
+	check_flags(flags);
+
+	current->lockdep_recursion = 1;
+	__lock_unpin_lock(lock);
+	current->lockdep_recursion = 0;
+	raw_local_irq_restore(flags);
+}
+EXPORT_SYMBOL_GPL(lock_unpin_lock);
+
 void lockdep_set_current_reclaim_state(gfp_t gfp_mask)
 {
 	current->lockdep_reclaim_gfp = gfp_mask;

kernel/sched/deadline.c

@@ -213,14 +213,28 @@ static inline bool need_pull_dl_task(struct rq *rq, struct task_struct *prev)
 	return dl_task(prev);
 }
 
-static inline void set_post_schedule(struct rq *rq)
+static DEFINE_PER_CPU(struct callback_head, dl_push_head);
+static DEFINE_PER_CPU(struct callback_head, dl_pull_head);
+
+static void push_dl_tasks(struct rq *);
+static void pull_dl_task(struct rq *);
+
+static inline void queue_push_tasks(struct rq *rq)
 {
-	rq->post_schedule = has_pushable_dl_tasks(rq);
+	if (!has_pushable_dl_tasks(rq))
+		return;
+
+	queue_balance_callback(rq, &per_cpu(dl_push_head, rq->cpu), push_dl_tasks);
+}
+
+static inline void queue_pull_task(struct rq *rq)
+{
+	queue_balance_callback(rq, &per_cpu(dl_pull_head, rq->cpu), pull_dl_task);
 }
 
 static struct rq *find_lock_later_rq(struct task_struct *task, struct rq *rq);
 
-static void dl_task_offline_migration(struct rq *rq, struct task_struct *p)
+static struct rq *dl_task_offline_migration(struct rq *rq, struct task_struct *p)
 {
 	struct rq *later_rq = NULL;
 	bool fallback = false;
@@ -254,14 +268,19 @@ static void dl_task_offline_migration(struct rq *rq, struct task_struct *p)
 		double_lock_balance(rq, later_rq);
 	}
 
+	/*
+	 * By now the task is replenished and enqueued; migrate it.
+	 */
 	deactivate_task(rq, p, 0);
 	set_task_cpu(p, later_rq->cpu);
-	activate_task(later_rq, p, ENQUEUE_REPLENISH);
+	activate_task(later_rq, p, 0);
 
 	if (!fallback)
 		resched_curr(later_rq);
 
-	double_unlock_balance(rq, later_rq);
+	double_unlock_balance(later_rq, rq);
+
+	return later_rq;
 }
 
 #else
@@ -291,12 +310,15 @@ static inline bool need_pull_dl_task(struct rq *rq, struct task_struct *prev)
 	return false;
 }
 
-static inline int pull_dl_task(struct rq *rq)
+static inline void pull_dl_task(struct rq *rq)
+{
+}
+
+static inline void queue_push_tasks(struct rq *rq)
 {
-	return 0;
 }
 
-static inline void set_post_schedule(struct rq *rq)
+static inline void queue_pull_task(struct rq *rq)
 {
 }
 #endif /* CONFIG_SMP */
@@ -498,22 +520,23 @@ static void update_dl_entity(struct sched_dl_entity *dl_se,
  * actually started or not (i.e., the replenishment instant is in
  * the future or in the past).
  */
-static int start_dl_timer(struct sched_dl_entity *dl_se, bool boosted)
+static int start_dl_timer(struct task_struct *p)
 {
-	struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
-	struct rq *rq = rq_of_dl_rq(dl_rq);
+	struct sched_dl_entity *dl_se = &p->dl;
+	struct hrtimer *timer = &dl_se->dl_timer;
+	struct rq *rq = task_rq(p);
 	ktime_t now, act;
 	s64 delta;
 
-	if (boosted)
-		return 0;
+	lockdep_assert_held(&rq->lock);
+
 	/*
 	 * We want the timer to fire at the deadline, but considering
 	 * that it is actually coming from rq->clock and not from
 	 * hrtimer's time base reading.
 	 */
 	act = ns_to_ktime(dl_se->deadline);
-	now = hrtimer_cb_get_time(&dl_se->dl_timer);
+	now = hrtimer_cb_get_time(timer);
 	delta = ktime_to_ns(now) - rq_clock(rq);
 	act = ktime_add_ns(act, delta);
 
@@ -525,7 +548,19 @@ static int start_dl_timer(struct sched_dl_entity *dl_se, bool boosted)
 	if (ktime_us_delta(act, now) < 0)
 		return 0;
 
-	hrtimer_start(&dl_se->dl_timer, act, HRTIMER_MODE_ABS);
+	/*
+	 * !enqueued will guarantee another callback; even if one is already in
+	 * progress. This ensures a balanced {get,put}_task_struct().
+	 *
+	 * The race against __run_timer() clearing the enqueued state is
+	 * harmless because we're holding task_rq()->lock, therefore the timer
+	 * expiring after we've done the check will wait on its task_rq_lock()
+	 * and observe our state.
+	 */
+	if (!hrtimer_is_queued(timer)) {
+		get_task_struct(p);
+		hrtimer_start(timer, act, HRTIMER_MODE_ABS);
+	}
 
 	return 1;
 }
@@ -555,35 +590,40 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer)
 	rq = task_rq_lock(p, &flags);
 
 	/*
-	 * We need to take care of several possible races here:
-	 *
-	 *   - the task might have changed its scheduling policy
-	 *     to something different than SCHED_DEADLINE
-	 *   - the task might have changed its reservation parameters
-	 *     (through sched_setattr())
-	 *   - the task might have been boosted by someone else and
-	 *     might be in the boosting/deboosting path
+	 * The task might have changed its scheduling policy to something
+	 * different than SCHED_DEADLINE (through switched_fromd_dl()).
+	 */
+	if (!dl_task(p)) {
+		__dl_clear_params(p);
+		goto unlock;
+	}
+
+	/*
+	 * This is possible if switched_from_dl() raced against a running
+	 * callback that took the above !dl_task() path and we've since then
+	 * switched back into SCHED_DEADLINE.
 	 *
-	 * In all this cases we bail out, as the task is already
-	 * in the runqueue or is going to be enqueued back anyway.
+	 * There's nothing to do except drop our task reference.
 	 */
-	if (!dl_task(p) || dl_se->dl_new ||
-	    dl_se->dl_boosted || !dl_se->dl_throttled)
+	if (dl_se->dl_new)
 		goto unlock;
 
-	sched_clock_tick();
-	update_rq_clock(rq);
+	/*
+	 * The task might have been boosted by someone else and might be in the
+	 * boosting/deboosting path, its not throttled.
+	 */
+	if (dl_se->dl_boosted)
+		goto unlock;
 
-#ifdef CONFIG_SMP
 	/*
-	 * If we find that the rq the task was on is no longer
-	 * available, we need to select a new rq.
+	 * Spurious timer due to start_dl_timer() race; or we already received
+	 * a replenishment from rt_mutex_setprio().
 	 */
-	if (unlikely(!rq->online)) {
-		dl_task_offline_migration(rq, p);
+	if (!dl_se->dl_throttled)
 		goto unlock;
-	}
-#endif
+
+	sched_clock_tick();
+	update_rq_clock(rq);
 
 	/*
 	 * If the throttle happened during sched-out; like:
@@ -609,17 +649,38 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer)
 		check_preempt_curr_dl(rq, p, 0);
 	else
 		resched_curr(rq);
+
 #ifdef CONFIG_SMP
 	/*
-	 * Queueing this task back might have overloaded rq,
-	 * check if we need to kick someone away.
+	 * Perform balancing operations here; after the replenishments.  We
+	 * cannot drop rq->lock before this, otherwise the assertion in
+	 * start_dl_timer() about not missing updates is not true.
+	 *
+	 * If we find that the rq the task was on is no longer available, we
+	 * need to select a new rq.
+	 *
+	 * XXX figure out if select_task_rq_dl() deals with offline cpus.
+	 */
+	if (unlikely(!rq->online))
+		rq = dl_task_offline_migration(rq, p);
+
+	/*
+	 * Queueing this task back might have overloaded rq, check if we need
+	 * to kick someone away.
 	 */
 	if (has_pushable_dl_tasks(rq))
 		push_dl_task(rq);
 #endif
+
 unlock:
 	task_rq_unlock(rq, p, &flags);
 
+	/*
+	 * This can free the task_struct, including this hrtimer, do not touch
+	 * anything related to that after this.
+	 */
+	put_task_struct(p);
+
 	return HRTIMER_NORESTART;
 }
 
@@ -679,7 +740,7 @@ static void update_curr_dl(struct rq *rq)
 	if (dl_runtime_exceeded(dl_se)) {
 		dl_se->dl_throttled = 1;
 		__dequeue_task_dl(rq, curr, 0);
-		if (unlikely(!start_dl_timer(dl_se, curr->dl.dl_boosted)))
+		if (unlikely(dl_se->dl_boosted || !start_dl_timer(curr)))
 			enqueue_task_dl(rq, curr, ENQUEUE_REPLENISH);
 
 		if (!is_leftmost(curr, &rq->dl))
@@ -1036,8 +1097,6 @@ static void check_preempt_equal_dl(struct rq *rq, struct task_struct *p)
 	resched_curr(rq);
 }
 
-static int pull_dl_task(struct rq *this_rq);
-
 #endif /* CONFIG_SMP */
 
 /*
@@ -1094,7 +1153,15 @@ struct task_struct *pick_next_task_dl(struct rq *rq, struct task_struct *prev)
 	dl_rq = &rq->dl;
 
 	if (need_pull_dl_task(rq, prev)) {
+		/*
+		 * This is OK, because current is on_cpu, which avoids it being
+		 * picked for load-balance and preemption/IRQs are still
+		 * disabled avoiding further scheduler activity on it and we're
+		 * being very careful to re-start the picking loop.
+		 */
+		lockdep_unpin_lock(&rq->lock);
 		pull_dl_task(rq);
+		lockdep_pin_lock(&rq->lock);
 		/*
 		 * pull_rt_task() can drop (and re-acquire) rq->lock; this
 		 * means a stop task can slip in, in which case we need to
@@ -1128,7 +1195,7 @@ struct task_struct *pick_next_task_dl(struct rq *rq, struct task_struct *prev)
 	if (hrtick_enabled(rq))
 		start_hrtick_dl(rq, p);
 
-	set_post_schedule(rq);
+	queue_push_tasks(rq);
 
 	return p;
 }
@@ -1165,7 +1232,6 @@ static void task_fork_dl(struct task_struct *p)
 
 static void task_dead_dl(struct task_struct *p)
 {
-	struct hrtimer *timer = &p->dl.dl_timer;
 	struct dl_bw *dl_b = dl_bw_of(task_cpu(p));
 
 	/*
@@ -1175,8 +1241,6 @@ static void task_dead_dl(struct task_struct *p)
 	/* XXX we should retain the bw until 0-lag */
 	dl_b->total_bw -= p->dl.dl_bw;
 	raw_spin_unlock_irq(&dl_b->lock);
-
-	hrtimer_cancel(timer);
 }
 
 static void set_curr_task_dl(struct rq *rq)
@@ -1504,15 +1568,16 @@ static void push_dl_tasks(struct rq *rq)
 		;
 }
 
-static int pull_dl_task(struct rq *this_rq)
+static void pull_dl_task(struct rq *this_rq)
 {
-	int this_cpu = this_rq->cpu, ret = 0, cpu;
+	int this_cpu = this_rq->cpu, cpu;
 	struct task_struct *p;
+	bool resched = false;
 	struct rq *src_rq;
 	u64 dmin = LONG_MAX;
 
 	if (likely(!dl_overloaded(this_rq)))
-		return 0;
+		return;
 
 	/*
 	 * Match the barrier from dl_set_overloaded; this guarantees that if we
@@ -1567,7 +1632,7 @@ static int pull_dl_task(struct rq *this_rq)
 					   src_rq->curr->dl.deadline))
 				goto skip;
 
-			ret = 1;
+			resched = true;
 
 			deactivate_task(src_rq, p, 0);
 			set_task_cpu(p, this_cpu);
@@ -1580,12 +1645,8 @@ static int pull_dl_task(struct rq *this_rq)
 		double_unlock_balance(this_rq, src_rq);
 	}
 
-	return ret;
-}
-
-static void post_schedule_dl(struct rq *rq)
-{
-	push_dl_tasks(rq);
+	if (resched)
+		resched_curr(this_rq);
 }
 
 /*
@@ -1701,37 +1762,16 @@ void __init init_sched_dl_class(void)
 
 #endif /* CONFIG_SMP */
 
-/*
- *  Ensure p's dl_timer is cancelled. May drop rq->lock for a while.
- */
-static void cancel_dl_timer(struct rq *rq, struct task_struct *p)
-{
-	struct hrtimer *dl_timer = &p->dl.dl_timer;
-
-	/* Nobody will change task's class if pi_lock is held */
-	lockdep_assert_held(&p->pi_lock);
-
-	if (hrtimer_active(dl_timer)) {
-		int ret = hrtimer_try_to_cancel(dl_timer);
-
-		if (unlikely(ret == -1)) {
-			/*
-			 * Note, p may migrate OR new deadline tasks
-			 * may appear in rq when we are unlocking it.
-			 * A caller of us must be fine with that.
-			 */
-			raw_spin_unlock(&rq->lock);
-			hrtimer_cancel(dl_timer);
-			raw_spin_lock(&rq->lock);
-		}
-	}
-}
-
 static void switched_from_dl(struct rq *rq, struct task_struct *p)
 {
-	/* XXX we should retain the bw until 0-lag */
-	cancel_dl_timer(rq, p);
-	__dl_clear_params(p);
+	/*
+	 * Start the deadline timer; if we switch back to dl before this we'll
+	 * continue consuming our current CBS slice. If we stay outside of
+	 * SCHED_DEADLINE until the deadline passes, the timer will reset the
+	 * task.
+	 */
+	if (!start_dl_timer(p))
+		__dl_clear_params(p);
 
 	/*
 	 * Since this might be the only -deadline task on the rq,
@@ -1741,8 +1781,7 @@ static void switched_from_dl(struct rq *rq, struct task_struct *p)
 	if (!task_on_rq_queued(p) || rq->dl.dl_nr_running)
 		return;
 
-	if (pull_dl_task(rq))
-		resched_curr(rq);
+	queue_pull_task(rq);
 }
 
 /*
@@ -1751,21 +1790,16 @@ static void switched_from_dl(struct rq *rq, struct task_struct *p)
  */
 static void switched_to_dl(struct rq *rq, struct task_struct *p)
 {
-	int check_resched = 1;
-
 	if (task_on_rq_queued(p) && rq->curr != p) {
 #ifdef CONFIG_SMP
-		if (p->nr_cpus_allowed > 1 && rq->dl.overloaded &&
-			push_dl_task(rq) && rq != task_rq(p))
-			/* Only reschedule if pushing failed */
-			check_resched = 0;
-#endif /* CONFIG_SMP */
-		if (check_resched) {
-			if (dl_task(rq->curr))
-				check_preempt_curr_dl(rq, p, 0);
-			else
-				resched_curr(rq);
-		}
+		if (p->nr_cpus_allowed > 1 && rq->dl.overloaded)
+			queue_push_tasks(rq);
+#else
+		if (dl_task(rq->curr))
+			check_preempt_curr_dl(rq, p, 0);
+		else
+			resched_curr(rq);
+#endif
 	}
 }
 
@@ -1785,15 +1819,14 @@ static void prio_changed_dl(struct rq *rq, struct task_struct *p,
 		 * or lowering its prio, so...
 		 */
 		if (!rq->dl.overloaded)
-			pull_dl_task(rq);
+			queue_pull_task(rq);
 
 		/*
 		 * If we now have a earlier deadline task than p,
 		 * then reschedule, provided p is still on this
 		 * runqueue.
 		 */
-		if (dl_time_before(rq->dl.earliest_dl.curr, p->dl.deadline) &&
-		    rq->curr == p)
+		if (dl_time_before(rq->dl.earliest_dl.curr, p->dl.deadline))
 			resched_curr(rq);
 #else
 		/*
@@ -1823,7 +1856,6 @@ const struct sched_class dl_sched_class = {
 	.set_cpus_allowed       = set_cpus_allowed_dl,
 	.rq_online              = rq_online_dl,
 	.rq_offline             = rq_offline_dl,
-	.post_schedule		= post_schedule_dl,
 	.task_woken		= task_woken_dl,
 #endif
 

kernel/sched/fair.c

@@ -5392,7 +5392,15 @@ pick_next_task_fair(struct rq *rq, struct task_struct *prev)
 	return p;
 
 idle:
+	/*
+	 * This is OK, because current is on_cpu, which avoids it being picked
+	 * for load-balance and preemption/IRQs are still disabled avoiding
+	 * further scheduler activity on it and we're being very careful to
+	 * re-start the picking loop.
+	 */
+	lockdep_unpin_lock(&rq->lock);
 	new_tasks = idle_balance(rq);
+	lockdep_pin_lock(&rq->lock);
 	/*
 	 * Because idle_balance() releases (and re-acquires) rq->lock, it is
 	 * possible for any higher priority task to appear. In that case we
@@ -7426,9 +7434,6 @@ static int idle_balance(struct rq *this_rq)
 		goto out;
 	}
 
-	/*
-	 * Drop the rq->lock, but keep IRQ/preempt disabled.
-	 */
 	raw_spin_unlock(&this_rq->lock);
 
 	update_blocked_averages(this_cpu);

kernel/sched/rt.c

@@ -260,7 +260,7 @@ int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
 
 #ifdef CONFIG_SMP
 
-static int pull_rt_task(struct rq *this_rq);
+static void pull_rt_task(struct rq *this_rq);
 
 static inline bool need_pull_rt_task(struct rq *rq, struct task_struct *prev)
 {
@@ -354,13 +354,23 @@ static inline int has_pushable_tasks(struct rq *rq)
 	return !plist_head_empty(&rq->rt.pushable_tasks);
 }
 
-static inline void set_post_schedule(struct rq *rq)
+static DEFINE_PER_CPU(struct callback_head, rt_push_head);
+static DEFINE_PER_CPU(struct callback_head, rt_pull_head);
+
+static void push_rt_tasks(struct rq *);
+static void pull_rt_task(struct rq *);
+
+static inline void queue_push_tasks(struct rq *rq)
 {
-	/*
-	 * We detect this state here so that we can avoid taking the RQ
-	 * lock again later if there is no need to push
-	 */
-	rq->post_schedule = has_pushable_tasks(rq);
+	if (!has_pushable_tasks(rq))
+		return;
+
+	queue_balance_callback(rq, &per_cpu(rt_push_head, rq->cpu), push_rt_tasks);
+}
+
+static inline void queue_pull_task(struct rq *rq)
+{
+	queue_balance_callback(rq, &per_cpu(rt_pull_head, rq->cpu), pull_rt_task);
 }
 
 static void enqueue_pushable_task(struct rq *rq, struct task_struct *p)
@@ -412,12 +422,11 @@ static inline bool need_pull_rt_task(struct rq *rq, struct task_struct *prev)
 	return false;
 }
 
-static inline int pull_rt_task(struct rq *this_rq)
+static inline void pull_rt_task(struct rq *this_rq)
 {
-	return 0;
 }
 
-static inline void set_post_schedule(struct rq *rq)
+static inline void queue_push_tasks(struct rq *rq)
 {
 }
 #endif /* CONFIG_SMP */
@@ -1469,7 +1478,15 @@ pick_next_task_rt(struct rq *rq, struct task_struct *prev)
 	struct rt_rq *rt_rq = &rq->rt;
 
 	if (need_pull_rt_task(rq, prev)) {
+		/*
+		 * This is OK, because current is on_cpu, which avoids it being
+		 * picked for load-balance and preemption/IRQs are still
+		 * disabled avoiding further scheduler activity on it and we're
+		 * being very careful to re-start the picking loop.
+		 */
+		lockdep_unpin_lock(&rq->lock);
 		pull_rt_task(rq);
+		lockdep_pin_lock(&rq->lock);
 		/*
 		 * pull_rt_task() can drop (and re-acquire) rq->lock; this
 		 * means a dl or stop task can slip in, in which case we need
@@ -1497,7 +1514,7 @@ pick_next_task_rt(struct rq *rq, struct task_struct *prev)
 	/* The running task is never eligible for pushing */
 	dequeue_pushable_task(rq, p);
 
-	set_post_schedule(rq);
+	queue_push_tasks(rq);
 
 	return p;
 }
@@ -1952,14 +1969,15 @@ static void push_irq_work_func(struct irq_work *work)
 }
 #endif /* HAVE_RT_PUSH_IPI */
 
-static int pull_rt_task(struct rq *this_rq)
+static void pull_rt_task(struct rq *this_rq)
 {
-	int this_cpu = this_rq->cpu, ret = 0, cpu;
+	int this_cpu = this_rq->cpu, cpu;
+	bool resched = false;
 	struct task_struct *p;
 	struct rq *src_rq;
 
 	if (likely(!rt_overloaded(this_rq)))
-		return 0;
+		return;
 
 	/*
 	 * Match the barrier from rt_set_overloaded; this guarantees that if we
@@ -1970,7 +1988,7 @@ static int pull_rt_task(struct rq *this_rq)
 #ifdef HAVE_RT_PUSH_IPI
 	if (sched_feat(RT_PUSH_IPI)) {
 		tell_cpu_to_push(this_rq);
-		return 0;
+		return;
 	}
 #endif
 
@@ -2023,7 +2041,7 @@ static int pull_rt_task(struct rq *this_rq)
 			if (p->prio < src_rq->curr->prio)
 				goto skip;
 
-			ret = 1;
+			resched = true;
 
 			deactivate_task(src_rq, p, 0);
 			set_task_cpu(p, this_cpu);
@@ -2039,12 +2057,8 @@ static int pull_rt_task(struct rq *this_rq)
 		double_unlock_balance(this_rq, src_rq);
 	}
 
-	return ret;
-}
-
-static void post_schedule_rt(struct rq *rq)
-{
-	push_rt_tasks(rq);
+	if (resched)
+		resched_curr(this_rq);
 }
 
 /*
@@ -2140,8 +2154,7 @@ static void switched_from_rt(struct rq *rq, struct task_struct *p)
 	if (!task_on_rq_queued(p) || rq->rt.rt_nr_running)
 		return;
 
-	if (pull_rt_task(rq))
-		resched_curr(rq);
+	queue_pull_task(rq);
 }
 
 void __init init_sched_rt_class(void)
@@ -2162,8 +2175,6 @@ void __init init_sched_rt_class(void)
  */
 static void switched_to_rt(struct rq *rq, struct task_struct *p)
 {
-	int check_resched = 1;
-
 	/*
 	 * If we are already running, then there's nothing
 	 * that needs to be done. But if we are not running
@@ -2173,13 +2184,12 @@ static void switched_to_rt(struct rq *rq, struct task_struct *p)
 	 */
 	if (task_on_rq_queued(p) && rq->curr != p) {
 #ifdef CONFIG_SMP
-		if (p->nr_cpus_allowed > 1 && rq->rt.overloaded &&
-		    /* Don't resched if we changed runqueues */
-		    push_rt_task(rq) && rq != task_rq(p))
-			check_resched = 0;
-#endif /* CONFIG_SMP */
-		if (check_resched && p->prio < rq->curr->prio)
+		if (p->nr_cpus_allowed > 1 && rq->rt.overloaded)
+			queue_push_tasks(rq);
+#else
+		if (p->prio < rq->curr->prio)
 			resched_curr(rq);
+#endif /* CONFIG_SMP */
 	}
 }
 
@@ -2200,14 +2210,13 @@ prio_changed_rt(struct rq *rq, struct task_struct *p, int oldprio)
 		 * may need to pull tasks to this runqueue.
 		 */
 		if (oldprio < p->prio)
-			pull_rt_task(rq);
+			queue_pull_task(rq);
+
 		/*
 		 * If there's a higher priority task waiting to run
-		 * then reschedule. Note, the above pull_rt_task
-		 * can release the rq lock and p could migrate.
-		 * Only reschedule if p is still on the same runqueue.
+		 * then reschedule.
 		 */
-		if (p->prio > rq->rt.highest_prio.curr && rq->curr == p)
+		if (p->prio > rq->rt.highest_prio.curr)
 			resched_curr(rq);
 #else
 		/* For UP simply resched on drop of prio */
@@ -2318,7 +2327,6 @@ const struct sched_class rt_sched_class = {
 	.set_cpus_allowed       = set_cpus_allowed_rt,
 	.rq_online              = rq_online_rt,
 	.rq_offline             = rq_offline_rt,
-	.post_schedule		= post_schedule_rt,
 	.task_woken		= task_woken_rt,
 	.switched_from		= switched_from_rt,
 #endif

kernel/sched/sched.h

@@ -624,9 +624,10 @@ struct rq {
 	unsigned long cpu_capacity;
 	unsigned long cpu_capacity_orig;
 
+	struct callback_head *balance_callback;
+
 	unsigned char idle_balance;
 	/* For active balancing */
-	int post_schedule;
 	int active_balance;
 	int push_cpu;
 	struct cpu_stop_work active_balance_work;
@@ -767,6 +768,21 @@ extern int migrate_swap(struct task_struct *, struct task_struct *);
 
 #ifdef CONFIG_SMP
 
+static inline void
+queue_balance_callback(struct rq *rq,
+		       struct callback_head *head,
+		       void (*func)(struct rq *rq))
+{
+	lockdep_assert_held(&rq->lock);
+
+	if (unlikely(head->next))
+		return;
+
+	head->func = (void (*)(struct callback_head *))func;
+	head->next = rq->balance_callback;
+	rq->balance_callback = head;
+}
+
 extern void sched_ttwu_pending(void);
 
 #define rcu_dereference_check_sched_domain(p) \
@@ -1192,7 +1208,6 @@ struct sched_class {
 	int  (*select_task_rq)(struct task_struct *p, int task_cpu, int sd_flag, int flags);
 	void (*migrate_task_rq)(struct task_struct *p, int next_cpu);
 
-	void (*post_schedule) (struct rq *this_rq);
 	void (*task_waking) (struct task_struct *task);
 	void (*task_woken) (struct rq *this_rq, struct task_struct *task);
 
@@ -1423,8 +1438,10 @@ static inline struct rq *__task_rq_lock(struct task_struct *p)
 	for (;;) {
 		rq = task_rq(p);
 		raw_spin_lock(&rq->lock);
-		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p)))
+		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
+			lockdep_pin_lock(&rq->lock);
 			return rq;
+		}
 		raw_spin_unlock(&rq->lock);
 
 		while (unlikely(task_on_rq_migrating(p)))
@@ -1461,8 +1478,10 @@ static inline struct rq *task_rq_lock(struct task_struct *p, unsigned long *flag
 		 * If we observe the new cpu in task_rq_lock, the acquire will
 		 * pair with the WMB to ensure we must then also see migrating.
 		 */
-		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p)))
+		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
+			lockdep_pin_lock(&rq->lock);
 			return rq;
+		}
 		raw_spin_unlock(&rq->lock);
 		raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
 
@@ -1474,6 +1493,7 @@ static inline struct rq *task_rq_lock(struct task_struct *p, unsigned long *flag
 static inline void __task_rq_unlock(struct rq *rq)
 	__releases(rq->lock)
 {
+	lockdep_unpin_lock(&rq->lock);
 	raw_spin_unlock(&rq->lock);
 }
 
@@ -1482,6 +1502,7 @@ task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags)
 	__releases(rq->lock)
 	__releases(p->pi_lock)
 {
+	lockdep_unpin_lock(&rq->lock);
 	raw_spin_unlock(&rq->lock);
 	raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
 }