sched/deadline: Add SCHED_DEADLINE SMP-related data structures & logic

Introduces data structures relevant for implementing dynamic
migration of -deadline tasks and the logic for checking if
runqueues are overloaded with -deadline tasks and for choosing
where a task should migrate, when it is the case.

Adds also dynamic migrations to SCHED_DEADLINE, so that tasks can
be moved among CPUs when necessary. It is also possible to bind a
task to a (set of) CPU(s), thus restricting its capability of
migrating, or forbidding migrations at all.

The very same approach used in sched_rt is utilised:
 - -deadline tasks are kept into CPU-specific runqueues,
 - -deadline tasks are migrated among runqueues to achieve the
   following:
    * on an M-CPU system the M earliest deadline ready tasks
      are always running;
    * affinity/cpusets settings of all the -deadline tasks is
      always respected.

Therefore, this very special form of "load balancing" is done with
an active method, i.e., the scheduler pushes or pulls tasks between
runqueues when they are woken up and/or (de)scheduled.
IOW, every time a preemption occurs, the descheduled task might be sent
to some other CPU (depending on its deadline) to continue executing
(push). On the other hand, every time a CPU becomes idle, it might pull
the second earliest deadline ready task from some other CPU.

To enforce this, a pull operation is always attempted before taking any
scheduling decision (pre_schedule()), as well as a push one after each
scheduling decision (post_schedule()). In addition, when a task arrives
or wakes up, the best CPU where to resume it is selected taking into
account its affinity mask, the system topology, but also its deadline.
E.g., from the scheduling point of view, the best CPU where to wake
up (and also where to push) a task is the one which is running the task
with the latest deadline among the M executing ones.

In order to facilitate these decisions, per-runqueue "caching" of the
deadlines of the currently running and of the first ready task is used.
Queued but not running tasks are also parked in another rb-tree to
speed-up pushes.
Signed-off-by: Juri Lelli <juri.lelli@gmail.com>
Signed-off-by: Dario Faggioli <raistlin@linux.it>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1383831828-15501-5-git-send-email-juri.lelli@gmail.comSigned-off-by: Ingo Molnar <mingo@kernel.org>

include/linux/sched.h

@@ -1201,6 +1201,7 @@ struct task_struct {
 	struct list_head tasks;
 #ifdef CONFIG_SMP
 	struct plist_node pushable_tasks;
+	struct rb_node pushable_dl_tasks;
 #endif
 
 	struct mm_struct *mm, *active_mm;

kernel/sched/core.c

@@ -1848,6 +1848,7 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 	init_task_preempt_count(p);
 #ifdef CONFIG_SMP
 	plist_node_init(&p->pushable_tasks, MAX_PRIO);
+	RB_CLEAR_NODE(&p->pushable_dl_tasks);
 #endif
 
 	put_cpu();
@@ -5040,6 +5041,7 @@ static void free_rootdomain(struct rcu_head *rcu)
 	struct root_domain *rd = container_of(rcu, struct root_domain, rcu);
 
 	cpupri_cleanup(&rd->cpupri);
+	free_cpumask_var(rd->dlo_mask);
 	free_cpumask_var(rd->rto_mask);
 	free_cpumask_var(rd->online);
 	free_cpumask_var(rd->span);
@@ -5091,8 +5093,10 @@ static int init_rootdomain(struct root_domain *rd)
 		goto out;
 	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
 		goto free_span;
-	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
+	if (!alloc_cpumask_var(&rd->dlo_mask, GFP_KERNEL))
 		goto free_online;
+	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
+		goto free_dlo_mask;
 
 	if (cpupri_init(&rd->cpupri) != 0)
 		goto free_rto_mask;
@@ -5100,6 +5104,8 @@ static int init_rootdomain(struct root_domain *rd)
 
 free_rto_mask:
 	free_cpumask_var(rd->rto_mask);
+free_dlo_mask:
+	free_cpumask_var(rd->dlo_mask);
 free_online:
 	free_cpumask_var(rd->online);
 free_span:
@@ -6451,6 +6457,7 @@ void __init sched_init_smp(void)
 	free_cpumask_var(non_isolated_cpus);
 
 	init_sched_rt_class();
+	init_sched_dl_class();
 }
 #else
 void __init sched_init_smp(void)

kernel/sched/deadline.c

@@ -10,6 +10,7 @@
  * miss some of their deadlines), and won't affect any other task.
  *
  * Copyright (C) 2012 Dario Faggioli <raistlin@linux.it>,
+ *                    Juri Lelli <juri.lelli@gmail.com>,
  *                    Michael Trimarchi <michael@amarulasolutions.com>,
  *                    Fabio Checconi <fchecconi@gmail.com>
  */
@@ -20,6 +21,15 @@ static inline int dl_time_before(u64 a, u64 b)
 	return (s64)(a - b) < 0;
 }
 
+/*
+ * Tells if entity @a should preempt entity @b.
+ */
+static inline
+int dl_entity_preempt(struct sched_dl_entity *a, struct sched_dl_entity *b)
+{
+	return dl_time_before(a->deadline, b->deadline);
+}
+
 static inline struct task_struct *dl_task_of(struct sched_dl_entity *dl_se)
 {
 	return container_of(dl_se, struct task_struct, dl);
@@ -53,8 +63,168 @@ static inline int is_leftmost(struct task_struct *p, struct dl_rq *dl_rq)
 void init_dl_rq(struct dl_rq *dl_rq, struct rq *rq)
 {
 	dl_rq->rb_root = RB_ROOT;
+
+#ifdef CONFIG_SMP
+	/* zero means no -deadline tasks */
+	dl_rq->earliest_dl.curr = dl_rq->earliest_dl.next = 0;
+
+	dl_rq->dl_nr_migratory = 0;
+	dl_rq->overloaded = 0;
+	dl_rq->pushable_dl_tasks_root = RB_ROOT;
+#endif
+}
+
+#ifdef CONFIG_SMP
+
+static inline int dl_overloaded(struct rq *rq)
+{
+	return atomic_read(&rq->rd->dlo_count);
+}
+
+static inline void dl_set_overload(struct rq *rq)
+{
+	if (!rq->online)
+		return;
+
+	cpumask_set_cpu(rq->cpu, rq->rd->dlo_mask);
+	/*
+	 * Must be visible before the overload count is
+	 * set (as in sched_rt.c).
+	 *
+	 * Matched by the barrier in pull_dl_task().
+	 */
+	smp_wmb();
+	atomic_inc(&rq->rd->dlo_count);
+}
+
+static inline void dl_clear_overload(struct rq *rq)
+{
+	if (!rq->online)
+		return;
+
+	atomic_dec(&rq->rd->dlo_count);
+	cpumask_clear_cpu(rq->cpu, rq->rd->dlo_mask);
+}
+
+static void update_dl_migration(struct dl_rq *dl_rq)
+{
+	if (dl_rq->dl_nr_migratory && dl_rq->dl_nr_total > 1) {
+		if (!dl_rq->overloaded) {
+			dl_set_overload(rq_of_dl_rq(dl_rq));
+			dl_rq->overloaded = 1;
+		}
+	} else if (dl_rq->overloaded) {
+		dl_clear_overload(rq_of_dl_rq(dl_rq));
+		dl_rq->overloaded = 0;
+	}
+}
+
+static void inc_dl_migration(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
+{
+	struct task_struct *p = dl_task_of(dl_se);
+	dl_rq = &rq_of_dl_rq(dl_rq)->dl;
+
+	dl_rq->dl_nr_total++;
+	if (p->nr_cpus_allowed > 1)
+		dl_rq->dl_nr_migratory++;
+
+	update_dl_migration(dl_rq);
+}
+
+static void dec_dl_migration(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
+{
+	struct task_struct *p = dl_task_of(dl_se);
+	dl_rq = &rq_of_dl_rq(dl_rq)->dl;
+
+	dl_rq->dl_nr_total--;
+	if (p->nr_cpus_allowed > 1)
+		dl_rq->dl_nr_migratory--;
+
+	update_dl_migration(dl_rq);
+}
+
+/*
+ * The list of pushable -deadline task is not a plist, like in
+ * sched_rt.c, it is an rb-tree with tasks ordered by deadline.
+ */
+static void enqueue_pushable_dl_task(struct rq *rq, struct task_struct *p)
+{
+	struct dl_rq *dl_rq = &rq->dl;
+	struct rb_node **link = &dl_rq->pushable_dl_tasks_root.rb_node;
+	struct rb_node *parent = NULL;
+	struct task_struct *entry;
+	int leftmost = 1;
+
+	BUG_ON(!RB_EMPTY_NODE(&p->pushable_dl_tasks));
+
+	while (*link) {
+		parent = *link;
+		entry = rb_entry(parent, struct task_struct,
+				 pushable_dl_tasks);
+		if (dl_entity_preempt(&p->dl, &entry->dl))
+			link = &parent->rb_left;
+		else {
+			link = &parent->rb_right;
+			leftmost = 0;
+		}
+	}
+
+	if (leftmost)
+		dl_rq->pushable_dl_tasks_leftmost = &p->pushable_dl_tasks;
+
+	rb_link_node(&p->pushable_dl_tasks, parent, link);
+	rb_insert_color(&p->pushable_dl_tasks, &dl_rq->pushable_dl_tasks_root);
 }
 
+static void dequeue_pushable_dl_task(struct rq *rq, struct task_struct *p)
+{
+	struct dl_rq *dl_rq = &rq->dl;
+
+	if (RB_EMPTY_NODE(&p->pushable_dl_tasks))
+		return;
+
+	if (dl_rq->pushable_dl_tasks_leftmost == &p->pushable_dl_tasks) {
+		struct rb_node *next_node;
+
+		next_node = rb_next(&p->pushable_dl_tasks);
+		dl_rq->pushable_dl_tasks_leftmost = next_node;
+	}
+
+	rb_erase(&p->pushable_dl_tasks, &dl_rq->pushable_dl_tasks_root);
+	RB_CLEAR_NODE(&p->pushable_dl_tasks);
+}
+
+static inline int has_pushable_dl_tasks(struct rq *rq)
+{
+	return !RB_EMPTY_ROOT(&rq->dl.pushable_dl_tasks_root);
+}
+
+static int push_dl_task(struct rq *rq);
+
+#else
+
+static inline
+void enqueue_pushable_dl_task(struct rq *rq, struct task_struct *p)
+{
+}
+
+static inline
+void dequeue_pushable_dl_task(struct rq *rq, struct task_struct *p)
+{
+}
+
+static inline
+void inc_dl_migration(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
+{
+}
+
+static inline
+void dec_dl_migration(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
+{
+}
+
+#endif /* CONFIG_SMP */
+
 static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags);
 static void __dequeue_task_dl(struct rq *rq, struct task_struct *p, int flags);
 static void check_preempt_curr_dl(struct rq *rq, struct task_struct *p,
@@ -309,6 +479,14 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer)
 			check_preempt_curr_dl(rq, p, 0);
 		else
 			resched_task(rq->curr);
+#ifdef CONFIG_SMP
+		/*
+		 * 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:
 	raw_spin_unlock(&rq->lock);
@@ -399,6 +577,100 @@ static void update_curr_dl(struct rq *rq)
 	}
 }
 
+#ifdef CONFIG_SMP
+
+static struct task_struct *pick_next_earliest_dl_task(struct rq *rq, int cpu);
+
+static inline u64 next_deadline(struct rq *rq)
+{
+	struct task_struct *next = pick_next_earliest_dl_task(rq, rq->cpu);
+
+	if (next && dl_prio(next->prio))
+		return next->dl.deadline;
+	else
+		return 0;
+}
+
+static void inc_dl_deadline(struct dl_rq *dl_rq, u64 deadline)
+{
+	struct rq *rq = rq_of_dl_rq(dl_rq);
+
+	if (dl_rq->earliest_dl.curr == 0 ||
+	    dl_time_before(deadline, dl_rq->earliest_dl.curr)) {
+		/*
+		 * If the dl_rq had no -deadline tasks, or if the new task
+		 * has shorter deadline than the current one on dl_rq, we
+		 * know that the previous earliest becomes our next earliest,
+		 * as the new task becomes the earliest itself.
+		 */
+		dl_rq->earliest_dl.next = dl_rq->earliest_dl.curr;
+		dl_rq->earliest_dl.curr = deadline;
+	} else if (dl_rq->earliest_dl.next == 0 ||
+		   dl_time_before(deadline, dl_rq->earliest_dl.next)) {
+		/*
+		 * On the other hand, if the new -deadline task has a
+		 * a later deadline than the earliest one on dl_rq, but
+		 * it is earlier than the next (if any), we must
+		 * recompute the next-earliest.
+		 */
+		dl_rq->earliest_dl.next = next_deadline(rq);
+	}
+}
+
+static void dec_dl_deadline(struct dl_rq *dl_rq, u64 deadline)
+{
+	struct rq *rq = rq_of_dl_rq(dl_rq);
+
+	/*
+	 * Since we may have removed our earliest (and/or next earliest)
+	 * task we must recompute them.
+	 */
+	if (!dl_rq->dl_nr_running) {
+		dl_rq->earliest_dl.curr = 0;
+		dl_rq->earliest_dl.next = 0;
+	} else {
+		struct rb_node *leftmost = dl_rq->rb_leftmost;
+		struct sched_dl_entity *entry;
+
+		entry = rb_entry(leftmost, struct sched_dl_entity, rb_node);
+		dl_rq->earliest_dl.curr = entry->deadline;
+		dl_rq->earliest_dl.next = next_deadline(rq);
+	}
+}
+
+#else
+
+static inline void inc_dl_deadline(struct dl_rq *dl_rq, u64 deadline) {}
+static inline void dec_dl_deadline(struct dl_rq *dl_rq, u64 deadline) {}
+
+#endif /* CONFIG_SMP */
+
+static inline
+void inc_dl_tasks(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
+{
+	int prio = dl_task_of(dl_se)->prio;
+	u64 deadline = dl_se->deadline;
+
+	WARN_ON(!dl_prio(prio));
+	dl_rq->dl_nr_running++;
+
+	inc_dl_deadline(dl_rq, deadline);
+	inc_dl_migration(dl_se, dl_rq);
+}
+
+static inline
+void dec_dl_tasks(struct sched_dl_entity *dl_se, struct dl_rq *dl_rq)
+{
+	int prio = dl_task_of(dl_se)->prio;
+
+	WARN_ON(!dl_prio(prio));
+	WARN_ON(!dl_rq->dl_nr_running);
+	dl_rq->dl_nr_running--;
+
+	dec_dl_deadline(dl_rq, dl_se->deadline);
+	dec_dl_migration(dl_se, dl_rq);
+}
+
 static void __enqueue_dl_entity(struct sched_dl_entity *dl_se)
 {
 	struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
@@ -426,7 +698,7 @@ static void __enqueue_dl_entity(struct sched_dl_entity *dl_se)
 	rb_link_node(&dl_se->rb_node, parent, link);
 	rb_insert_color(&dl_se->rb_node, &dl_rq->rb_root);
 
-	dl_rq->dl_nr_running++;
+	inc_dl_tasks(dl_se, dl_rq);
 }
 
 static void __dequeue_dl_entity(struct sched_dl_entity *dl_se)
@@ -446,7 +718,7 @@ static void __dequeue_dl_entity(struct sched_dl_entity *dl_se)
 	rb_erase(&dl_se->rb_node, &dl_rq->rb_root);
 	RB_CLEAR_NODE(&dl_se->rb_node);
 
-	dl_rq->dl_nr_running--;
+	dec_dl_tasks(dl_se, dl_rq);
 }
 
 static void
@@ -484,12 +756,17 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 		return;
 
 	enqueue_dl_entity(&p->dl, flags);
+
+	if (!task_current(rq, p) && p->nr_cpus_allowed > 1)
+		enqueue_pushable_dl_task(rq, p);
+
 	inc_nr_running(rq);
 }
 
 static void __dequeue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 {
 	dequeue_dl_entity(&p->dl);
+	dequeue_pushable_dl_task(rq, p);
 }
 
 static void dequeue_task_dl(struct rq *rq, struct task_struct *p, int flags)
@@ -527,6 +804,74 @@ static void yield_task_dl(struct rq *rq)
 	update_curr_dl(rq);
 }
 
+#ifdef CONFIG_SMP
+
+static int find_later_rq(struct task_struct *task);
+static int latest_cpu_find(struct cpumask *span,
+			   struct task_struct *task,
+			   struct cpumask *later_mask);
+
+static int
+select_task_rq_dl(struct task_struct *p, int cpu, int sd_flag, int flags)
+{
+	struct task_struct *curr;
+	struct rq *rq;
+
+	if (sd_flag != SD_BALANCE_WAKE && sd_flag != SD_BALANCE_FORK)
+		goto out;
+
+	rq = cpu_rq(cpu);
+
+	rcu_read_lock();
+	curr = ACCESS_ONCE(rq->curr); /* unlocked access */
+
+	/*
+	 * If we are dealing with a -deadline task, we must
+	 * decide where to wake it up.
+	 * If it has a later deadline and the current task
+	 * on this rq can't move (provided the waking task
+	 * can!) we prefer to send it somewhere else. On the
+	 * other hand, if it has a shorter deadline, we
+	 * try to make it stay here, it might be important.
+	 */
+	if (unlikely(dl_task(curr)) &&
+	    (curr->nr_cpus_allowed < 2 ||
+	     !dl_entity_preempt(&p->dl, &curr->dl)) &&
+	    (p->nr_cpus_allowed > 1)) {
+		int target = find_later_rq(p);
+
+		if (target != -1)
+			cpu = target;
+	}
+	rcu_read_unlock();
+
+out:
+	return cpu;
+}
+
+static void check_preempt_equal_dl(struct rq *rq, struct task_struct *p)
+{
+	/*
+	 * Current can't be migrated, useless to reschedule,
+	 * let's hope p can move out.
+	 */
+	if (rq->curr->nr_cpus_allowed == 1 ||
+	    latest_cpu_find(rq->rd->span, rq->curr, NULL) == -1)
+		return;
+
+	/*
+	 * p is migratable, so let's not schedule it and
+	 * see if it is pushed or pulled somewhere else.
+	 */
+	if (p->nr_cpus_allowed != 1 &&
+	    latest_cpu_find(rq->rd->span, p, NULL) != -1)
+		return;
+
+	resched_task(rq->curr);
+}
+
+#endif /* CONFIG_SMP */
+
 /*
  * Only called when both the current and waking task are -deadline
  * tasks.
@@ -534,8 +879,20 @@ static void yield_task_dl(struct rq *rq)
 static void check_preempt_curr_dl(struct rq *rq, struct task_struct *p,
 				  int flags)
 {
-	if (dl_time_before(p->dl.deadline, rq->curr->dl.deadline))
+	if (dl_entity_preempt(&p->dl, &rq->curr->dl)) {
 		resched_task(rq->curr);
+		return;
+	}
+
+#ifdef CONFIG_SMP
+	/*
+	 * In the unlikely case current and p have the same deadline
+	 * let us try to decide what's the best thing to do...
+	 */
+	if ((s64)(p->dl.deadline - rq->curr->dl.deadline) == 0 &&
+	    !need_resched())
+		check_preempt_equal_dl(rq, p);
+#endif /* CONFIG_SMP */
 }
 
 #ifdef CONFIG_SCHED_HRTICK
@@ -575,16 +932,29 @@ struct task_struct *pick_next_task_dl(struct rq *rq)
 
 	p = dl_task_of(dl_se);
 	p->se.exec_start = rq_clock_task(rq);
+
+	/* Running task will never be pushed. */
+	if (p)
+		dequeue_pushable_dl_task(rq, p);
+
 #ifdef CONFIG_SCHED_HRTICK
 	if (hrtick_enabled(rq))
 		start_hrtick_dl(rq, p);
 #endif
+
+#ifdef CONFIG_SMP
+	rq->post_schedule = has_pushable_dl_tasks(rq);
+#endif /* CONFIG_SMP */
+
 	return p;
 }
 
 static void put_prev_task_dl(struct rq *rq, struct task_struct *p)
 {
 	update_curr_dl(rq);
+
+	if (on_dl_rq(&p->dl) && p->nr_cpus_allowed > 1)
+		enqueue_pushable_dl_task(rq, p);
 }
 
 static void task_tick_dl(struct rq *rq, struct task_struct *p, int queued)
@@ -618,16 +988,517 @@ static void set_curr_task_dl(struct rq *rq)
 	struct task_struct *p = rq->curr;
 
 	p->se.exec_start = rq_clock_task(rq);
+
+	/* You can't push away the running task */
+	dequeue_pushable_dl_task(rq, p);
+}
+
+#ifdef CONFIG_SMP
+
+/* Only try algorithms three times */
+#define DL_MAX_TRIES 3
+
+static int pick_dl_task(struct rq *rq, struct task_struct *p, int cpu)
+{
+	if (!task_running(rq, p) &&
+	    (cpu < 0 || cpumask_test_cpu(cpu, &p->cpus_allowed)) &&
+	    (p->nr_cpus_allowed > 1))
+		return 1;
+
+	return 0;
+}
+
+/* Returns the second earliest -deadline task, NULL otherwise */
+static struct task_struct *pick_next_earliest_dl_task(struct rq *rq, int cpu)
+{
+	struct rb_node *next_node = rq->dl.rb_leftmost;
+	struct sched_dl_entity *dl_se;
+	struct task_struct *p = NULL;
+
+next_node:
+	next_node = rb_next(next_node);
+	if (next_node) {
+		dl_se = rb_entry(next_node, struct sched_dl_entity, rb_node);
+		p = dl_task_of(dl_se);
+
+		if (pick_dl_task(rq, p, cpu))
+			return p;
+
+		goto next_node;
+	}
+
+	return NULL;
+}
+
+static int latest_cpu_find(struct cpumask *span,
+			   struct task_struct *task,
+			   struct cpumask *later_mask)
+{
+	const struct sched_dl_entity *dl_se = &task->dl;
+	int cpu, found = -1, best = 0;
+	u64 max_dl = 0;
+
+	for_each_cpu(cpu, span) {
+		struct rq *rq = cpu_rq(cpu);
+		struct dl_rq *dl_rq = &rq->dl;
+
+		if (cpumask_test_cpu(cpu, &task->cpus_allowed) &&
+		    (!dl_rq->dl_nr_running || dl_time_before(dl_se->deadline,
+		     dl_rq->earliest_dl.curr))) {
+			if (later_mask)
+				cpumask_set_cpu(cpu, later_mask);
+			if (!best && !dl_rq->dl_nr_running) {
+				best = 1;
+				found = cpu;
+			} else if (!best &&
+				   dl_time_before(max_dl,
+						  dl_rq->earliest_dl.curr)) {
+				max_dl = dl_rq->earliest_dl.curr;
+				found = cpu;
+			}
+		} else if (later_mask)
+			cpumask_clear_cpu(cpu, later_mask);
+	}
+
+	return found;
+}
+
+static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask_dl);
+
+static int find_later_rq(struct task_struct *task)
+{
+	struct sched_domain *sd;
+	struct cpumask *later_mask = __get_cpu_var(local_cpu_mask_dl);
+	int this_cpu = smp_processor_id();
+	int best_cpu, cpu = task_cpu(task);
+
+	/* Make sure the mask is initialized first */
+	if (unlikely(!later_mask))
+		return -1;
+
+	if (task->nr_cpus_allowed == 1)
+		return -1;
+
+	best_cpu = latest_cpu_find(task_rq(task)->rd->span, task, later_mask);
+	if (best_cpu == -1)
+		return -1;
+
+	/*
+	 * If we are here, some target has been found,
+	 * the most suitable of which is cached in best_cpu.
+	 * This is, among the runqueues where the current tasks
+	 * have later deadlines than the task's one, the rq
+	 * with the latest possible one.
+	 *
+	 * Now we check how well this matches with task's
+	 * affinity and system topology.
+	 *
+	 * The last cpu where the task run is our first
+	 * guess, since it is most likely cache-hot there.
+	 */
+	if (cpumask_test_cpu(cpu, later_mask))
+		return cpu;
+	/*
+	 * Check if this_cpu is to be skipped (i.e., it is
+	 * not in the mask) or not.
+	 */
+	if (!cpumask_test_cpu(this_cpu, later_mask))
+		this_cpu = -1;
+
+	rcu_read_lock();
+	for_each_domain(cpu, sd) {
+		if (sd->flags & SD_WAKE_AFFINE) {
+
+			/*
+			 * If possible, preempting this_cpu is
+			 * cheaper than migrating.
+			 */
+			if (this_cpu != -1 &&
+			    cpumask_test_cpu(this_cpu, sched_domain_span(sd))) {
+				rcu_read_unlock();
+				return this_cpu;
+			}
+
+			/*
+			 * Last chance: if best_cpu is valid and is
+			 * in the mask, that becomes our choice.
+			 */
+			if (best_cpu < nr_cpu_ids &&
+			    cpumask_test_cpu(best_cpu, sched_domain_span(sd))) {
+				rcu_read_unlock();
+				return best_cpu;
+			}
+		}
+	}
+	rcu_read_unlock();
+
+	/*
+	 * At this point, all our guesses failed, we just return
+	 * 'something', and let the caller sort the things out.
+	 */
+	if (this_cpu != -1)
+		return this_cpu;
+
+	cpu = cpumask_any(later_mask);
+	if (cpu < nr_cpu_ids)
+		return cpu;
+
+	return -1;
+}
+
+/* Locks the rq it finds */
+static struct rq *find_lock_later_rq(struct task_struct *task, struct rq *rq)
+{
+	struct rq *later_rq = NULL;
+	int tries;
+	int cpu;
+
+	for (tries = 0; tries < DL_MAX_TRIES; tries++) {
+		cpu = find_later_rq(task);
+
+		if ((cpu == -1) || (cpu == rq->cpu))
+			break;
+
+		later_rq = cpu_rq(cpu);
+
+		/* Retry if something changed. */
+		if (double_lock_balance(rq, later_rq)) {
+			if (unlikely(task_rq(task) != rq ||
+				     !cpumask_test_cpu(later_rq->cpu,
+				                       &task->cpus_allowed) ||
+				     task_running(rq, task) || !task->on_rq)) {
+				double_unlock_balance(rq, later_rq);
+				later_rq = NULL;
+				break;
+			}
+		}
+
+		/*
+		 * If the rq we found has no -deadline task, or
+		 * its earliest one has a later deadline than our
+		 * task, the rq is a good one.
+		 */
+		if (!later_rq->dl.dl_nr_running ||
+		    dl_time_before(task->dl.deadline,
+				   later_rq->dl.earliest_dl.curr))
+			break;
+
+		/* Otherwise we try again. */
+		double_unlock_balance(rq, later_rq);
+		later_rq = NULL;
+	}
+
+	return later_rq;
+}
+
+static struct task_struct *pick_next_pushable_dl_task(struct rq *rq)
+{
+	struct task_struct *p;
+
+	if (!has_pushable_dl_tasks(rq))
+		return NULL;
+
+	p = rb_entry(rq->dl.pushable_dl_tasks_leftmost,
+		     struct task_struct, pushable_dl_tasks);
+
+	BUG_ON(rq->cpu != task_cpu(p));
+	BUG_ON(task_current(rq, p));
+	BUG_ON(p->nr_cpus_allowed <= 1);
+
+	BUG_ON(!p->se.on_rq);
+	BUG_ON(!dl_task(p));
+
+	return p;
+}
+
+/*
+ * See if the non running -deadline tasks on this rq
+ * can be sent to some other CPU where they can preempt
+ * and start executing.
+ */
+static int push_dl_task(struct rq *rq)
+{
+	struct task_struct *next_task;
+	struct rq *later_rq;
+
+	if (!rq->dl.overloaded)
+		return 0;
+
+	next_task = pick_next_pushable_dl_task(rq);
+	if (!next_task)
+		return 0;
+
+retry:
+	if (unlikely(next_task == rq->curr)) {
+		WARN_ON(1);
+		return 0;
+	}
+
+	/*
+	 * If next_task preempts rq->curr, and rq->curr
+	 * can move away, it makes sense to just reschedule
+	 * without going further in pushing next_task.
+	 */
+	if (dl_task(rq->curr) &&
+	    dl_time_before(next_task->dl.deadline, rq->curr->dl.deadline) &&
+	    rq->curr->nr_cpus_allowed > 1) {
+		resched_task(rq->curr);
+		return 0;
+	}
+
+	/* We might release rq lock */
+	get_task_struct(next_task);
+
+	/* Will lock the rq it'll find */
+	later_rq = find_lock_later_rq(next_task, rq);
+	if (!later_rq) {
+		struct task_struct *task;
+
+		/*
+		 * We must check all this again, since
+		 * find_lock_later_rq releases rq->lock and it is
+		 * then possible that next_task has migrated.
+		 */
+		task = pick_next_pushable_dl_task(rq);
+		if (task_cpu(next_task) == rq->cpu && task == next_task) {
+			/*
+			 * The task is still there. We don't try
+			 * again, some other cpu will pull it when ready.
+			 */
+			dequeue_pushable_dl_task(rq, next_task);
+			goto out;
+		}
+
+		if (!task)
+			/* No more tasks */
+			goto out;
+
+		put_task_struct(next_task);
+		next_task = task;
+		goto retry;
+	}
+
+	deactivate_task(rq, next_task, 0);
+	set_task_cpu(next_task, later_rq->cpu);
+	activate_task(later_rq, next_task, 0);
+
+	resched_task(later_rq->curr);
+
+	double_unlock_balance(rq, later_rq);
+
+out:
+	put_task_struct(next_task);
+
+	return 1;
+}
+
+static void push_dl_tasks(struct rq *rq)
+{
+	/* Terminates as it moves a -deadline task */
+	while (push_dl_task(rq))
+		;
 }
 
+static int pull_dl_task(struct rq *this_rq)
+{
+	int this_cpu = this_rq->cpu, ret = 0, cpu;
+	struct task_struct *p;
+	struct rq *src_rq;
+	u64 dmin = LONG_MAX;
+
+	if (likely(!dl_overloaded(this_rq)))
+		return 0;
+
+	/*
+	 * Match the barrier from dl_set_overloaded; this guarantees that if we
+	 * see overloaded we must also see the dlo_mask bit.
+	 */
+	smp_rmb();
+
+	for_each_cpu(cpu, this_rq->rd->dlo_mask) {
+		if (this_cpu == cpu)
+			continue;
+
+		src_rq = cpu_rq(cpu);
+
+		/*
+		 * It looks racy, abd it is! However, as in sched_rt.c,
+		 * we are fine with this.
+		 */
+		if (this_rq->dl.dl_nr_running &&
+		    dl_time_before(this_rq->dl.earliest_dl.curr,
+				   src_rq->dl.earliest_dl.next))
+			continue;
+
+		/* Might drop this_rq->lock */
+		double_lock_balance(this_rq, src_rq);
+
+		/*
+		 * If there are no more pullable tasks on the
+		 * rq, we're done with it.
+		 */
+		if (src_rq->dl.dl_nr_running <= 1)
+			goto skip;
+
+		p = pick_next_earliest_dl_task(src_rq, this_cpu);
+
+		/*
+		 * We found a task to be pulled if:
+		 *  - it preempts our current (if there's one),
+		 *  - it will preempt the last one we pulled (if any).
+		 */
+		if (p && dl_time_before(p->dl.deadline, dmin) &&
+		    (!this_rq->dl.dl_nr_running ||
+		     dl_time_before(p->dl.deadline,
+				    this_rq->dl.earliest_dl.curr))) {
+			WARN_ON(p == src_rq->curr);
+			WARN_ON(!p->se.on_rq);
+
+			/*
+			 * Then we pull iff p has actually an earlier
+			 * deadline than the current task of its runqueue.
+			 */
+			if (dl_time_before(p->dl.deadline,
+					   src_rq->curr->dl.deadline))
+				goto skip;
+
+			ret = 1;
+
+			deactivate_task(src_rq, p, 0);
+			set_task_cpu(p, this_cpu);
+			activate_task(this_rq, p, 0);
+			dmin = p->dl.deadline;
+
+			/* Is there any other task even earlier? */
+		}
+skip:
+		double_unlock_balance(this_rq, src_rq);
+	}
+
+	return ret;
+}
+
+static void pre_schedule_dl(struct rq *rq, struct task_struct *prev)
+{
+	/* Try to pull other tasks here */
+	if (dl_task(prev))
+		pull_dl_task(rq);
+}
+
+static void post_schedule_dl(struct rq *rq)
+{
+	push_dl_tasks(rq);
+}
+
+/*
+ * Since the task is not running and a reschedule is not going to happen
+ * anytime soon on its runqueue, we try pushing it away now.
+ */
+static void task_woken_dl(struct rq *rq, struct task_struct *p)
+{
+	if (!task_running(rq, p) &&
+	    !test_tsk_need_resched(rq->curr) &&
+	    has_pushable_dl_tasks(rq) &&
+	    p->nr_cpus_allowed > 1 &&
+	    dl_task(rq->curr) &&
+	    (rq->curr->nr_cpus_allowed < 2 ||
+	     dl_entity_preempt(&rq->curr->dl, &p->dl))) {
+		push_dl_tasks(rq);
+	}
+}
+
+static void set_cpus_allowed_dl(struct task_struct *p,
+				const struct cpumask *new_mask)
+{
+	struct rq *rq;
+	int weight;
+
+	BUG_ON(!dl_task(p));
+
+	/*
+	 * Update only if the task is actually running (i.e.,
+	 * it is on the rq AND it is not throttled).
+	 */
+	if (!on_dl_rq(&p->dl))
+		return;
+
+	weight = cpumask_weight(new_mask);
+
+	/*
+	 * Only update if the process changes its state from whether it
+	 * can migrate or not.
+	 */
+	if ((p->nr_cpus_allowed > 1) == (weight > 1))
+		return;
+
+	rq = task_rq(p);
+
+	/*
+	 * The process used to be able to migrate OR it can now migrate
+	 */
+	if (weight <= 1) {
+		if (!task_current(rq, p))
+			dequeue_pushable_dl_task(rq, p);
+		BUG_ON(!rq->dl.dl_nr_migratory);
+		rq->dl.dl_nr_migratory--;
+	} else {
+		if (!task_current(rq, p))
+			enqueue_pushable_dl_task(rq, p);
+		rq->dl.dl_nr_migratory++;
+	}
+
+	update_dl_migration(&rq->dl);
+}
+
+/* Assumes rq->lock is held */
+static void rq_online_dl(struct rq *rq)
+{
+	if (rq->dl.overloaded)
+		dl_set_overload(rq);
+}
+
+/* Assumes rq->lock is held */
+static void rq_offline_dl(struct rq *rq)
+{
+	if (rq->dl.overloaded)
+		dl_clear_overload(rq);
+}
+
+void init_sched_dl_class(void)
+{
+	unsigned int i;
+
+	for_each_possible_cpu(i)
+		zalloc_cpumask_var_node(&per_cpu(local_cpu_mask_dl, i),
+					GFP_KERNEL, cpu_to_node(i));
+}
+
+#endif /* CONFIG_SMP */
+
 static void switched_from_dl(struct rq *rq, struct task_struct *p)
 {
-	if (hrtimer_active(&p->dl.dl_timer))
+	if (hrtimer_active(&p->dl.dl_timer) && !dl_policy(p->policy))
 		hrtimer_try_to_cancel(&p->dl.dl_timer);
+
+#ifdef CONFIG_SMP
+	/*
+	 * Since this might be the only -deadline task on the rq,
+	 * this is the right place to try to pull some other one
+	 * from an overloaded cpu, if any.
+	 */
+	if (!rq->dl.dl_nr_running)
+		pull_dl_task(rq);
+#endif
 }
 
+/*
+ * When switching to -deadline, we may overload the rq, then
+ * we try to push someone off, if possible.
+ */
 static void switched_to_dl(struct rq *rq, struct task_struct *p)
 {
+	int check_resched = 1;
+
 	/*
 	 * If p is throttled, don't consider the possibility
 	 * of preempting rq->curr, the check will be done right
@@ -637,26 +1508,53 @@ static void switched_to_dl(struct rq *rq, struct task_struct *p)
 		return;
 
 	if (p->on_rq || rq->curr != p) {
-		if (task_has_dl_policy(rq->curr))
+#ifdef CONFIG_SMP
+		if (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 && task_has_dl_policy(rq->curr))
 			check_preempt_curr_dl(rq, p, 0);
-		else
-			resched_task(rq->curr);
 	}
 }
 
+/*
+ * If the scheduling parameters of a -deadline task changed,
+ * a push or pull operation might be needed.
+ */
 static void prio_changed_dl(struct rq *rq, struct task_struct *p,
 			    int oldprio)
 {
-	switched_to_dl(rq, p);
-}
-
+	if (p->on_rq || rq->curr == p) {
 #ifdef CONFIG_SMP
-static int
-select_task_rq_dl(struct task_struct *p, int prev_cpu, int sd_flag, int flags)
-{
-	return task_cpu(p);
+		/*
+		 * This might be too much, but unfortunately
+		 * we don't have the old deadline value, and
+		 * we can't argue if the task is increasing
+		 * or lowering its prio, so...
+		 */
+		if (!rq->dl.overloaded)
+			pull_dl_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)
+			resched_task(p);
+#else
+		/*
+		 * Again, we don't know if p has a earlier
+		 * or later deadline, so let's blindly set a
+		 * (maybe not needed) rescheduling point.
+		 */
+		resched_task(p);
+#endif /* CONFIG_SMP */
+	} else
+		switched_to_dl(rq, p);
 }
-#endif
 
 const struct sched_class dl_sched_class = {
 	.next			= &rt_sched_class,
@@ -671,6 +1569,12 @@ const struct sched_class dl_sched_class = {
 
 #ifdef CONFIG_SMP
 	.select_task_rq		= select_task_rq_dl,
+	.set_cpus_allowed       = set_cpus_allowed_dl,
+	.rq_online              = rq_online_dl,
+	.rq_offline             = rq_offline_dl,
+	.pre_schedule		= pre_schedule_dl,
+	.post_schedule		= post_schedule_dl,
+	.task_woken		= task_woken_dl,
 #endif
 
 	.set_curr_task		= set_curr_task_dl,

kernel/sched/rt.c

@@ -1738,7 +1738,7 @@ static void task_woken_rt(struct rq *rq, struct task_struct *p)
 	    !test_tsk_need_resched(rq->curr) &&
 	    has_pushable_tasks(rq) &&
 	    p->nr_cpus_allowed > 1 &&
-	    rt_task(rq->curr) &&
+	    (dl_task(rq->curr) || rt_task(rq->curr)) &&
 	    (rq->curr->nr_cpus_allowed < 2 ||
 	     rq->curr->prio <= p->prio))
 		push_rt_tasks(rq);

kernel/sched/sched.h

@@ -385,6 +385,31 @@ struct dl_rq {
 	struct rb_node *rb_leftmost;
 
 	unsigned long dl_nr_running;
+
+#ifdef CONFIG_SMP
+	/*
+	 * Deadline values of the currently executing and the
+	 * earliest ready task on this rq. Caching these facilitates
+	 * the decision wether or not a ready but not running task
+	 * should migrate somewhere else.
+	 */
+	struct {
+		u64 curr;
+		u64 next;
+	} earliest_dl;
+
+	unsigned long dl_nr_migratory;
+	unsigned long dl_nr_total;
+	int overloaded;
+
+	/*
+	 * Tasks on this rq that can be pushed away. They are kept in
+	 * an rb-tree, ordered by tasks' deadlines, with caching
+	 * of the leftmost (earliest deadline) element.
+	 */
+	struct rb_root pushable_dl_tasks_root;
+	struct rb_node *pushable_dl_tasks_leftmost;
+#endif
 };
 
 #ifdef CONFIG_SMP
@@ -404,6 +429,13 @@ struct root_domain {
 	cpumask_var_t span;
 	cpumask_var_t online;
 
+	/*
+	 * The bit corresponding to a CPU gets set here if such CPU has more
+	 * than one runnable -deadline task (as it is below for RT tasks).
+	 */
+	cpumask_var_t dlo_mask;
+	atomic_t dlo_count;
+
 	/*
 	 * The "RT overload" flag: it gets set if a CPU has more than
 	 * one runnable RT task.
@@ -1095,6 +1127,8 @@ static inline void idle_balance(int cpu, struct rq *rq)
 extern void sysrq_sched_debug_show(void);
 extern void sched_init_granularity(void);
 extern void update_max_interval(void);
+
+extern void init_sched_dl_class(void);
 extern void init_sched_rt_class(void);
 extern void init_sched_fair_class(void);