1 files changed, 335 insertions, 78 deletions

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 44817c640e99..55aafcff5810 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -129,7 +129,11 @@ const_debug unsigned int sysctl_sched_features =
  * Number of tasks to iterate in a single balance run.
  * Limited because this is done with IRQs disabled.
  */
+#ifndef CONFIG_PREEMPT_RT_FULL
 const_debug unsigned int sysctl_sched_nr_migrate = 32;
+#else
+const_debug unsigned int sysctl_sched_nr_migrate = 8;
+#endif
 
 /*
  * period over which we average the RT time consumption, measured
@@ -345,6 +349,7 @@ static void init_rq_hrtick(struct rq *rq)
 
 	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
 	rq->hrtick_timer.function = hrtick;
+	rq->hrtick_timer.irqsafe = 1;
 }
 #else	/* CONFIG_SCHED_HRTICK */
 static inline void hrtick_clear(struct rq *rq)
@@ -449,7 +454,7 @@ void wake_q_add(struct wake_q_head *head, struct task_struct *task)
 	head->lastp = &node->next;
 }
 
-void wake_up_q(struct wake_q_head *head)
+void __wake_up_q(struct wake_q_head *head, bool sleeper)
 {
 	struct wake_q_node *node = head->first;
 
@@ -466,7 +471,10 @@ void wake_up_q(struct wake_q_head *head)
 		 * wake_up_process() implies a wmb() to pair with the queueing
 		 * in wake_q_add() so as not to miss wakeups.
 		 */
-		wake_up_process(task);
+		if (sleeper)
+			wake_up_lock_sleeper(task);
+		else
+			wake_up_process(task);
 		put_task_struct(task);
 	}
 }
@@ -502,6 +510,38 @@ void resched_curr(struct rq *rq)
 		trace_sched_wake_idle_without_ipi(cpu);
 }
 
+#ifdef CONFIG_PREEMPT_LAZY
+void resched_curr_lazy(struct rq *rq)
+{
+	struct task_struct *curr = rq->curr;
+	int cpu;
+
+	if (!sched_feat(PREEMPT_LAZY)) {
+		resched_curr(rq);
+		return;
+	}
+
+	lockdep_assert_held(&rq->lock);
+
+	if (test_tsk_need_resched(curr))
+		return;
+
+	if (test_tsk_need_resched_lazy(curr))
+		return;
+
+	set_tsk_need_resched_lazy(curr);
+
+	cpu = cpu_of(rq);
+	if (cpu == smp_processor_id())
+		return;
+
+	/* NEED_RESCHED_LAZY must be visible before we test polling */
+	smp_mb();
+	if (!tsk_is_polling(curr))
+		smp_send_reschedule(cpu);
+}
+#endif
+
 void resched_cpu(int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
@@ -525,11 +565,14 @@ void resched_cpu(int cpu)
  */
 int get_nohz_timer_target(void)
 {
-	int i, cpu = smp_processor_id();
+	int i, cpu;
 	struct sched_domain *sd;
 
+	preempt_disable_rt();
+	cpu = smp_processor_id();
+
 	if (!idle_cpu(cpu) && is_housekeeping_cpu(cpu))
-		return cpu;
+		goto preempt_en_rt;
 
 	rcu_read_lock();
 	for_each_domain(cpu, sd) {
@@ -548,6 +591,8 @@ int get_nohz_timer_target(void)
 		cpu = housekeeping_any_cpu();
 unlock:
 	rcu_read_unlock();
+preempt_en_rt:
+	preempt_enable_rt();
 	return cpu;
 }
 /*
@@ -1089,6 +1134,11 @@ void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
 
 	lockdep_assert_held(&p->pi_lock);
 
+	if (__migrate_disabled(p)) {
+		cpumask_copy(&p->cpus_allowed, new_mask);
+		return;
+	}
+
 	queued = task_on_rq_queued(p);
 	running = task_current(rq, p);
 
@@ -1111,6 +1161,84 @@ void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
 		enqueue_task(rq, p, ENQUEUE_RESTORE);
 }
 
+static DEFINE_PER_CPU(struct cpumask, sched_cpumasks);
+static DEFINE_MUTEX(sched_down_mutex);
+static cpumask_t sched_down_cpumask;
+
+void tell_sched_cpu_down_begin(int cpu)
+{
+	mutex_lock(&sched_down_mutex);
+	cpumask_set_cpu(cpu, &sched_down_cpumask);
+	mutex_unlock(&sched_down_mutex);
+}
+
+void tell_sched_cpu_down_done(int cpu)
+{
+	mutex_lock(&sched_down_mutex);
+	cpumask_clear_cpu(cpu, &sched_down_cpumask);
+	mutex_unlock(&sched_down_mutex);
+}
+
+/**
+ * migrate_me - try to move the current task off this cpu
+ *
+ * Used by the pin_current_cpu() code to try to get tasks
+ * to move off the current CPU as it is going down.
+ * It will only move the task if the task isn't pinned to
+ * the CPU (with migrate_disable, affinity or NO_SETAFFINITY)
+ * and the task has to be in a RUNNING state. Otherwise the
+ * movement of the task will wake it up (change its state
+ * to running) when the task did not expect it.
+ *
+ * Returns 1 if it succeeded in moving the current task
+ *         0 otherwise.
+ */
+int migrate_me(void)
+{
+	struct task_struct *p = current;
+	struct migration_arg arg;
+	struct cpumask *cpumask;
+	struct cpumask *mask;
+	unsigned int dest_cpu;
+	struct rq_flags rf;
+	struct rq *rq;
+
+	/*
+	 * We can not migrate tasks bounded to a CPU or tasks not
+	 * running. The movement of the task will wake it up.
+	 */
+	if (p->flags & PF_NO_SETAFFINITY || p->state)
+		return 0;
+
+	mutex_lock(&sched_down_mutex);
+	rq = task_rq_lock(p, &rf);
+
+	cpumask = this_cpu_ptr(&sched_cpumasks);
+	mask = &p->cpus_allowed;
+
+	cpumask_andnot(cpumask, mask, &sched_down_cpumask);
+
+	if (!cpumask_weight(cpumask)) {
+		/* It's only on this CPU? */
+		task_rq_unlock(rq, p, &rf);
+		mutex_unlock(&sched_down_mutex);
+		return 0;
+	}
+
+	dest_cpu = cpumask_any_and(cpu_active_mask, cpumask);
+
+	arg.task = p;
+	arg.dest_cpu = dest_cpu;
+
+	task_rq_unlock(rq, p, &rf);
+
+	stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
+	tlb_migrate_finish(p->mm);
+	mutex_unlock(&sched_down_mutex);
+
+	return 1;
+}
+
 /*
  * Change a given task's CPU affinity. Migrate the thread to a
  * proper CPU and schedule it away if the CPU it's executing on
@@ -1168,7 +1296,7 @@ static int __set_cpus_allowed_ptr(struct task_struct *p,
 	}
 
 	/* Can the task run on the task's current CPU? If so, we're done */
-	if (cpumask_test_cpu(task_cpu(p), new_mask))
+	if (cpumask_test_cpu(task_cpu(p), new_mask) || __migrate_disabled(p))
 		goto out;
 
 	dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);
@@ -1355,6 +1483,18 @@ out:
 	return ret;
 }
 
+static bool check_task_state(struct task_struct *p, long match_state)
+{
+	bool match = false;
+
+	raw_spin_lock_irq(&p->pi_lock);
+	if (p->state == match_state || p->saved_state == match_state)
+		match = true;
+	raw_spin_unlock_irq(&p->pi_lock);
+
+	return match;
+}
+
 /*
  * wait_task_inactive - wait for a thread to unschedule.
  *
@@ -1399,7 +1539,7 @@ unsigned long wait_task_inactive(struct task_struct *p, long match_state)
 		 * is actually now running somewhere else!
 		 */
 		while (task_running(rq, p)) {
-			if (match_state && unlikely(p->state != match_state))
+			if (match_state && !check_task_state(p, match_state))
 				return 0;
 			cpu_relax();
 		}
@@ -1414,7 +1554,8 @@ unsigned long wait_task_inactive(struct task_struct *p, long match_state)
 		running = task_running(rq, p);
 		queued = task_on_rq_queued(p);
 		ncsw = 0;
-		if (!match_state || p->state == match_state)
+		if (!match_state || p->state == match_state ||
+		    p->saved_state == match_state)
 			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
 		task_rq_unlock(rq, p, &rf);
 
@@ -1670,10 +1811,6 @@ static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_fl
 {
 	activate_task(rq, p, en_flags);
 	p->on_rq = TASK_ON_RQ_QUEUED;
-
-	/* if a worker is waking up, notify workqueue */
-	if (p->flags & PF_WQ_WORKER)
-		wq_worker_waking_up(p, cpu_of(rq));
 }
 
 /*
@@ -2008,8 +2145,27 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 	 */
 	smp_mb__before_spinlock();
 	raw_spin_lock_irqsave(&p->pi_lock, flags);
-	if (!(p->state & state))
+	if (!(p->state & state)) {
+		/*
+		 * The task might be running due to a spinlock sleeper
+		 * wakeup. Check the saved state and set it to running
+		 * if the wakeup condition is true.
+		 */
+		if (!(wake_flags & WF_LOCK_SLEEPER)) {
+			if (p->saved_state & state) {
+				p->saved_state = TASK_RUNNING;
+				success = 1;
+			}
+		}
 		goto out;
+	}
+
+	/*
+	 * If this is a regular wakeup, then we can unconditionally
+	 * clear the saved state of a "lock sleeper".
+	 */
+	if (!(wake_flags & WF_LOCK_SLEEPER))
+		p->saved_state = TASK_RUNNING;
 
 	trace_sched_waking(p);
 
@@ -2093,53 +2249,6 @@ out:
 }
 
 /**
- * try_to_wake_up_local - try to wake up a local task with rq lock held
- * @p: the thread to be awakened
- *
- * Put @p on the run-queue if it's not already there. The caller must
- * ensure that this_rq() is locked, @p is bound to this_rq() and not
- * the current task.
- */
-static void try_to_wake_up_local(struct task_struct *p, struct pin_cookie cookie)
-{
-	struct rq *rq = task_rq(p);
-
-	if (WARN_ON_ONCE(rq != this_rq()) ||
-	    WARN_ON_ONCE(p == current))
-		return;
-
-	lockdep_assert_held(&rq->lock);
-
-	if (!raw_spin_trylock(&p->pi_lock)) {
-		/*
-		 * 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've
-		 * not yet picked a replacement task.
-		 */
-		lockdep_unpin_lock(&rq->lock, cookie);
-		raw_spin_unlock(&rq->lock);
-		raw_spin_lock(&p->pi_lock);
-		raw_spin_lock(&rq->lock);
-		lockdep_repin_lock(&rq->lock, cookie);
-	}
-
-	if (!(p->state & TASK_NORMAL))
-		goto out;
-
-	trace_sched_waking(p);
-
-	if (!task_on_rq_queued(p))
-		ttwu_activate(rq, p, ENQUEUE_WAKEUP);
-
-	ttwu_do_wakeup(rq, p, 0, cookie);
-	if (schedstat_enabled())
-		ttwu_stat(p, smp_processor_id(), 0);
-out:
-	raw_spin_unlock(&p->pi_lock);
-}
-
-/**
  * wake_up_process - Wake up a specific process
  * @p: The process to be woken up.
  *
@@ -2157,6 +2266,18 @@ int wake_up_process(struct task_struct *p)
 }
 EXPORT_SYMBOL(wake_up_process);
 
+/**
+ * wake_up_lock_sleeper - Wake up a specific process blocked on a "sleeping lock"
+ * @p: The process to be woken up.
+ *
+ * Same as wake_up_process() above, but wake_flags=WF_LOCK_SLEEPER to indicate
+ * the nature of the wakeup.
+ */
+int wake_up_lock_sleeper(struct task_struct *p)
+{
+	return try_to_wake_up(p, TASK_ALL, WF_LOCK_SLEEPER);
+}
+
 int wake_up_state(struct task_struct *p, unsigned int state)
 {
 	return try_to_wake_up(p, state, 0);
@@ -2433,6 +2554,9 @@ int sched_fork(unsigned long clone_flags, struct task_struct *p)
 	p->on_cpu = 0;
 #endif
 	init_task_preempt_count(p);
+#ifdef CONFIG_HAVE_PREEMPT_LAZY
+	task_thread_info(p)->preempt_lazy_count = 0;
+#endif
 #ifdef CONFIG_SMP
 	plist_node_init(&p->pushable_tasks, MAX_PRIO);
 	RB_CLEAR_NODE(&p->pushable_dl_tasks);
@@ -2761,8 +2885,12 @@ static struct rq *finish_task_switch(struct task_struct *prev)
 	finish_arch_post_lock_switch();
 
 	fire_sched_in_preempt_notifiers(current);
+	/*
+	 * We use mmdrop_delayed() here so we don't have to do the
+	 * full __mmdrop() when we are the last user.
+	 */
 	if (mm)
-		mmdrop(mm);
+		mmdrop_delayed(mm);
 	if (unlikely(prev_state == TASK_DEAD)) {
 		if (prev->sched_class->task_dead)
 			prev->sched_class->task_dead(prev);
@@ -3237,6 +3365,77 @@ static inline void schedule_debug(struct task_struct *prev)
 	schedstat_inc(this_rq(), sched_count);
 }
 
+#if defined(CONFIG_PREEMPT_RT_FULL) && defined(CONFIG_SMP)
+
+void migrate_disable(void)
+{
+	struct task_struct *p = current;
+
+	if (in_atomic() || irqs_disabled()) {
+#ifdef CONFIG_SCHED_DEBUG
+		p->migrate_disable_atomic++;
+#endif
+		return;
+	}
+
+#ifdef CONFIG_SCHED_DEBUG
+	if (unlikely(p->migrate_disable_atomic)) {
+		tracing_off();
+		WARN_ON_ONCE(1);
+	}
+#endif
+
+	if (p->migrate_disable) {
+		p->migrate_disable++;
+		return;
+	}
+
+	preempt_disable();
+	preempt_lazy_disable();
+	pin_current_cpu();
+	p->migrate_disable = 1;
+	preempt_enable();
+}
+EXPORT_SYMBOL(migrate_disable);
+
+void migrate_enable(void)
+{
+	struct task_struct *p = current;
+
+	if (in_atomic() || irqs_disabled()) {
+#ifdef CONFIG_SCHED_DEBUG
+		p->migrate_disable_atomic--;
+#endif
+		return;
+	}
+
+#ifdef CONFIG_SCHED_DEBUG
+	if (unlikely(p->migrate_disable_atomic)) {
+		tracing_off();
+		WARN_ON_ONCE(1);
+	}
+#endif
+	WARN_ON_ONCE(p->migrate_disable <= 0);
+
+	if (p->migrate_disable > 1) {
+		p->migrate_disable--;
+		return;
+	}
+
+	preempt_disable();
+	/*
+	 * Clearing migrate_disable causes tsk_cpus_allowed to
+	 * show the tasks original cpu affinity.
+	 */
+	p->migrate_disable = 0;
+
+	unpin_current_cpu();
+	preempt_enable();
+	preempt_lazy_enable();
+}
+EXPORT_SYMBOL(migrate_enable);
+#endif
+
 /*
  * Pick up the highest-prio task:
  */
@@ -3364,19 +3563,6 @@ static void __sched notrace __schedule(bool preempt)
 		} else {
 			deactivate_task(rq, prev, DEQUEUE_SLEEP);
 			prev->on_rq = 0;
-
-			/*
-			 * If a worker went to sleep, notify and ask workqueue
-			 * whether it wants to wake up a task to maintain
-			 * concurrency.
-			 */
-			if (prev->flags & PF_WQ_WORKER) {
-				struct task_struct *to_wakeup;
-
-				to_wakeup = wq_worker_sleeping(prev);
-				if (to_wakeup)
-					try_to_wake_up_local(to_wakeup, cookie);
-			}
 		}
 		switch_count = &prev->nvcsw;
 	}
@@ -3386,6 +3572,7 @@ static void __sched notrace __schedule(bool preempt)
 
 	next = pick_next_task(rq, prev, cookie);
 	clear_tsk_need_resched(prev);
+	clear_tsk_need_resched_lazy(prev);
 	clear_preempt_need_resched();
 	rq->clock_skip_update = 0;
 
@@ -3407,9 +3594,20 @@ STACK_FRAME_NON_STANDARD(__schedule); /* switch_to() */
 
 static inline void sched_submit_work(struct task_struct *tsk)
 {
-	if (!tsk->state || tsk_is_pi_blocked(tsk))
+	if (!tsk->state)
 		return;
 	/*
+	 * If a worker went to sleep, notify and ask workqueue whether
+	 * it wants to wake up a task to maintain concurrency.
+	 */
+	if (tsk->flags & PF_WQ_WORKER)
+		wq_worker_sleeping(tsk);
+
+
+	if (tsk_is_pi_blocked(tsk))
+		return;
+
+	/*
 	 * If we are going to sleep and we have plugged IO queued,
 	 * make sure to submit it to avoid deadlocks.
 	 */
@@ -3417,6 +3615,12 @@ static inline void sched_submit_work(struct task_struct *tsk)
 		blk_schedule_flush_plug(tsk);
 }
 
+static void sched_update_worker(struct task_struct *tsk)
+{
+	if (tsk->flags & PF_WQ_WORKER)
+		wq_worker_running(tsk);
+}
+
 asmlinkage __visible void __sched schedule(void)
 {
 	struct task_struct *tsk = current;
@@ -3427,6 +3631,7 @@ asmlinkage __visible void __sched schedule(void)
 		__schedule(false);
 		sched_preempt_enable_no_resched();
 	} while (need_resched());
+	sched_update_worker(tsk);
 }
 EXPORT_SYMBOL(schedule);
 
@@ -3490,6 +3695,30 @@ static void __sched notrace preempt_schedule_common(void)
 	} while (need_resched());
 }
 
+#ifdef CONFIG_PREEMPT_LAZY
+/*
+ * If TIF_NEED_RESCHED is then we allow to be scheduled away since this is
+ * set by a RT task. Oterwise we try to avoid beeing scheduled out as long as
+ * preempt_lazy_count counter >0.
+ */
+static __always_inline int preemptible_lazy(void)
+{
+	if (test_thread_flag(TIF_NEED_RESCHED))
+		return 1;
+	if (current_thread_info()->preempt_lazy_count)
+		return 0;
+	return 1;
+}
+
+#else
+
+static inline int preemptible_lazy(void)
+{
+	return 1;
+}
+
+#endif
+
 #ifdef CONFIG_PREEMPT
 /*
  * this is the entry point to schedule() from in-kernel preemption
@@ -3504,7 +3733,8 @@ asmlinkage __visible void __sched notrace preempt_schedule(void)
 	 */
 	if (likely(!preemptible()))
 		return;
-
+	if (!preemptible_lazy())
+		return;
 	preempt_schedule_common();
 }
 NOKPROBE_SYMBOL(preempt_schedule);
@@ -3531,6 +3761,9 @@ asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
 	if (likely(!preemptible()))
 		return;
 
+	if (!preemptible_lazy())
+		return;
+
 	do {
 		/*
 		 * Because the function tracer can trace preempt_count_sub()
@@ -3553,7 +3786,16 @@ asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
 		 * an infinite recursion.
 		 */
 		prev_ctx = exception_enter();
+		/*
+		 * The add/subtract must not be traced by the function
+		 * tracer. But we still want to account for the
+		 * preempt off latency tracer. Since the _notrace versions
+		 * of add/subtract skip the accounting for latency tracer
+		 * we must force it manually.
+		 */
+		start_critical_timings();
 		__schedule(true);
+		stop_critical_timings();
 		exception_exit(prev_ctx);
 
 		preempt_latency_stop(1);
@@ -4901,6 +5143,7 @@ int __cond_resched_lock(spinlock_t *lock)
 }
 EXPORT_SYMBOL(__cond_resched_lock);
 
+#ifndef CONFIG_PREEMPT_RT_FULL
 int __sched __cond_resched_softirq(void)
 {
 	BUG_ON(!in_softirq());
@@ -4914,6 +5157,7 @@ int __sched __cond_resched_softirq(void)
 	return 0;
 }
 EXPORT_SYMBOL(__cond_resched_softirq);
+#endif
 
 /**
  * yield - yield the current processor to other threads.
@@ -5283,7 +5527,9 @@ void init_idle(struct task_struct *idle, int cpu)
 
 	/* Set the preempt count _outside_ the spinlocks! */
 	init_idle_preempt_count(idle, cpu);
-
+#ifdef CONFIG_HAVE_PREEMPT_LAZY
+	task_thread_info(idle)->preempt_lazy_count = 0;
+#endif
 	/*
 	 * The idle tasks have their own, simple scheduling class:
 	 */
@@ -5426,6 +5672,8 @@ void sched_setnuma(struct task_struct *p, int nid)
 #endif /* CONFIG_NUMA_BALANCING */
 
 #ifdef CONFIG_HOTPLUG_CPU
+static DEFINE_PER_CPU(struct mm_struct *, idle_last_mm);
+
 /*
  * Ensures that the idle task is using init_mm right before its cpu goes
  * offline.
@@ -5440,7 +5688,12 @@ void idle_task_exit(void)
 		switch_mm_irqs_off(mm, &init_mm, current);
 		finish_arch_post_lock_switch();
 	}
-	mmdrop(mm);
+	/*
+	 * Defer the cleanup to an alive cpu. On RT we can neither
+	 * call mmdrop() nor mmdrop_delayed() from here.
+	 */
+	per_cpu(idle_last_mm, smp_processor_id()) = mm;
+
 }
 
 /*
@@ -7315,6 +7568,10 @@ int sched_cpu_dying(unsigned int cpu)
 	update_max_interval();
 	nohz_balance_exit_idle(cpu);
 	hrtick_clear(rq);
+	if (per_cpu(idle_last_mm, cpu)) {
+		mmdrop_delayed(per_cpu(idle_last_mm, cpu));
+		per_cpu(idle_last_mm, cpu) = NULL;
+	}
 	return 0;
 }
 #endif
@@ -7566,7 +7823,7 @@ void __init sched_init(void)
 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
 static inline int preempt_count_equals(int preempt_offset)
 {
-	int nested = preempt_count() + rcu_preempt_depth();
+	int nested = preempt_count() + sched_rcu_preempt_depth();
 
 	return (nested == preempt_offset);
 }