sched/deadline: Track the "total rq utilization" too

The total rq utilization is defined as the sum of the utilisations of
tasks that are "assigned" to a runqueue, independently from their state
(TASK_RUNNING or blocked)

Tested-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Signed-off-by: Luca Abeni <luca.abeni@santannapisa.it>
Signed-off-by: Claudio Scordino <claudio@evidence.eu.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tommaso Cucinotta <tommaso.cucinotta@sssup.it>
Link: http://lkml.kernel.org/r/1495138417-6203-8-git-send-email-luca.abeni@santannapisa.it
Signed-off-by: Ingo Molnar <mingo@kernel.org>

kernel/sched/deadline.c

@@ -51,6 +51,7 @@ void add_running_bw(u64 dl_bw, struct dl_rq *dl_rq)
 	lockdep_assert_held(&(rq_of_dl_rq(dl_rq))->lock);
 	dl_rq->running_bw += dl_bw;
 	SCHED_WARN_ON(dl_rq->running_bw < old); /* overflow */
+	SCHED_WARN_ON(dl_rq->running_bw > dl_rq->this_bw);
 }
 
 static inline
@@ -65,25 +66,52 @@ void sub_running_bw(u64 dl_bw, struct dl_rq *dl_rq)
 		dl_rq->running_bw = 0;
 }
 
+static inline
+void add_rq_bw(u64 dl_bw, struct dl_rq *dl_rq)
+{
+	u64 old = dl_rq->this_bw;
+
+	lockdep_assert_held(&(rq_of_dl_rq(dl_rq))->lock);
+	dl_rq->this_bw += dl_bw;
+	SCHED_WARN_ON(dl_rq->this_bw < old); /* overflow */
+}
+
+static inline
+void sub_rq_bw(u64 dl_bw, struct dl_rq *dl_rq)
+{
+	u64 old = dl_rq->this_bw;
+
+	lockdep_assert_held(&(rq_of_dl_rq(dl_rq))->lock);
+	dl_rq->this_bw -= dl_bw;
+	SCHED_WARN_ON(dl_rq->this_bw > old); /* underflow */
+	if (dl_rq->this_bw > old)
+		dl_rq->this_bw = 0;
+	SCHED_WARN_ON(dl_rq->running_bw > dl_rq->this_bw);
+}
+
 void dl_change_utilization(struct task_struct *p, u64 new_bw)
 {
+	struct rq *rq;
+
 	if (task_on_rq_queued(p))
 		return;
 
-	if (!p->dl.dl_non_contending)
+	rq = task_rq(p);
-		return;
+	if (p->dl.dl_non_contending) {
-
+		sub_running_bw(p->dl.dl_bw, &rq->dl);
-	sub_running_bw(p->dl.dl_bw, &task_rq(p)->dl);
+		p->dl.dl_non_contending = 0;
-	p->dl.dl_non_contending = 0;
+		/*
-	/*
+		 * If the timer handler is currently running and the
-	 * If the timer handler is currently running and the
+		 * timer cannot be cancelled, inactive_task_timer()
-	 * timer cannot be cancelled, inactive_task_timer()
+		 * will see that dl_not_contending is not set, and
-	 * will see that dl_not_contending is not set, and
+		 * will not touch the rq's active utilization,
-	 * will not touch the rq's active utilization,
+		 * so we are still safe.
-	 * so we are still safe.
+		 */
-	 */
+		if (hrtimer_try_to_cancel(&p->dl.inactive_timer) == 1)
-	if (hrtimer_try_to_cancel(&p->dl.inactive_timer) == 1)
+			put_task_struct(p);
-		put_task_struct(p);
+	}
+	sub_rq_bw(p->dl.dl_bw, &rq->dl);
+	add_rq_bw(new_bw, &rq->dl);
 }
 
 /*
@@ -178,6 +206,8 @@ static void task_non_contending(struct task_struct *p)
 		if (!dl_task(p) || p->state == TASK_DEAD) {
 			struct dl_bw *dl_b = dl_bw_of(task_cpu(p));
 
+			if (p->state == TASK_DEAD)
+				sub_rq_bw(p->dl.dl_bw, &rq->dl);
 			raw_spin_lock(&dl_b->lock);
 			__dl_clear(dl_b, p->dl.dl_bw);
 			__dl_clear_params(p);
@@ -192,7 +222,7 @@ static void task_non_contending(struct task_struct *p)
 	hrtimer_start(timer, ns_to_ktime(zerolag_time), HRTIMER_MODE_REL);
 }
 
-static void task_contending(struct sched_dl_entity *dl_se)
+static void task_contending(struct sched_dl_entity *dl_se, int flags)
 {
 	struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
 
@@ -203,6 +233,9 @@ static void task_contending(struct sched_dl_entity *dl_se)
 	if (dl_se->dl_runtime == 0)
 		return;
 
+	if (flags & ENQUEUE_MIGRATED)
+		add_rq_bw(dl_se->dl_bw, dl_rq);
+
 	if (dl_se->dl_non_contending) {
 		dl_se->dl_non_contending = 0;
 		/*
@@ -268,6 +301,7 @@ void init_dl_rq(struct dl_rq *dl_rq)
 #endif
 
 	dl_rq->running_bw = 0;
+	dl_rq->this_bw = 0;
 	init_dl_rq_bw_ratio(dl_rq);
 }
 
@@ -1042,6 +1076,7 @@ static enum hrtimer_restart inactive_task_timer(struct hrtimer *timer)
 
 		if (p->state == TASK_DEAD && dl_se->dl_non_contending) {
 			sub_running_bw(p->dl.dl_bw, dl_rq_of_se(&p->dl));
+			sub_rq_bw(p->dl.dl_bw, dl_rq_of_se(&p->dl));
 			dl_se->dl_non_contending = 0;
 		}
 
@@ -1207,7 +1242,7 @@ enqueue_dl_entity(struct sched_dl_entity *dl_se,
 	 * we want a replenishment of its runtime.
 	 */
 	if (flags & ENQUEUE_WAKEUP) {
-		task_contending(dl_se);
+		task_contending(dl_se, flags);
 		update_dl_entity(dl_se, pi_se);
 	} else if (flags & ENQUEUE_REPLENISH) {
 		replenish_dl_entity(dl_se, pi_se);
@@ -1260,8 +1295,10 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 	if (!p->dl.dl_throttled && dl_is_constrained(&p->dl))
 		dl_check_constrained_dl(&p->dl);
 
-	if (p->on_rq == TASK_ON_RQ_MIGRATING || flags & ENQUEUE_RESTORE)
+	if (p->on_rq == TASK_ON_RQ_MIGRATING || flags & ENQUEUE_RESTORE) {
+		add_rq_bw(p->dl.dl_bw, &rq->dl);
 		add_running_bw(p->dl.dl_bw, &rq->dl);
+	}
 
 	/*
 	 * If p is throttled, we do not enqueue it. In fact, if it exhausted
@@ -1277,7 +1314,7 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 	 */
 	if (p->dl.dl_throttled && !(flags & ENQUEUE_REPLENISH)) {
 		if (flags & ENQUEUE_WAKEUP)
-			task_contending(&p->dl);
+			task_contending(&p->dl, flags);
 
 		return;
 	}
@@ -1299,8 +1336,10 @@ static void dequeue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 	update_curr_dl(rq);
 	__dequeue_task_dl(rq, p, flags);
 
-	if (p->on_rq == TASK_ON_RQ_MIGRATING || flags & DEQUEUE_SAVE)
+	if (p->on_rq == TASK_ON_RQ_MIGRATING || flags & DEQUEUE_SAVE) {
 		sub_running_bw(p->dl.dl_bw, &rq->dl);
+		sub_rq_bw(p->dl.dl_bw, &rq->dl);
+	}
 
 	/*
 	 * This check allows to start the inactive timer (or to immediately
@@ -1394,7 +1433,7 @@ static void migrate_task_rq_dl(struct task_struct *p)
 {
 	struct rq *rq;
 
-	if (!(p->state == TASK_WAKING) || !(p->dl.dl_non_contending))
+	if (p->state != TASK_WAKING)
 		return;
 
 	rq = task_rq(p);
@@ -1404,18 +1443,20 @@ static void migrate_task_rq_dl(struct task_struct *p)
 	 * rq->lock is not... So, lock it
 	 */
 	raw_spin_lock(&rq->lock);
-	sub_running_bw(p->dl.dl_bw, &rq->dl);
+	if (p->dl.dl_non_contending) {
-	p->dl.dl_non_contending = 0;
+		sub_running_bw(p->dl.dl_bw, &rq->dl);
-	/*
+		p->dl.dl_non_contending = 0;
-	 * If the timer handler is currently running and the
+		/*
-	 * timer cannot be cancelled, inactive_task_timer()
+		 * If the timer handler is currently running and the
-	 * will see that dl_not_contending is not set, and
+		 * timer cannot be cancelled, inactive_task_timer()
-	 * will not touch the rq's active utilization,
+		 * will see that dl_not_contending is not set, and
-	 * so we are still safe.
+		 * will not touch the rq's active utilization,
-	 */
+		 * so we are still safe.
-	if (hrtimer_try_to_cancel(&p->dl.inactive_timer) == 1)
+		 */
-		put_task_struct(p);
+		if (hrtimer_try_to_cancel(&p->dl.inactive_timer) == 1)
-
+			put_task_struct(p);
+	}
+	sub_rq_bw(p->dl.dl_bw, &rq->dl);
 	raw_spin_unlock(&rq->lock);
 }
 
@@ -1858,7 +1899,9 @@ retry:
 
 	deactivate_task(rq, next_task, 0);
 	sub_running_bw(next_task->dl.dl_bw, &rq->dl);
+	sub_rq_bw(next_task->dl.dl_bw, &rq->dl);
 	set_task_cpu(next_task, later_rq->cpu);
+	add_rq_bw(next_task->dl.dl_bw, &later_rq->dl);
 	add_running_bw(next_task->dl.dl_bw, &later_rq->dl);
 	activate_task(later_rq, next_task, 0);
 	ret = 1;
@@ -1948,7 +1991,9 @@ static void pull_dl_task(struct rq *this_rq)
 
 			deactivate_task(src_rq, p, 0);
 			sub_running_bw(p->dl.dl_bw, &src_rq->dl);
+			sub_rq_bw(p->dl.dl_bw, &src_rq->dl);
 			set_task_cpu(p, this_cpu);
+			add_rq_bw(p->dl.dl_bw, &this_rq->dl);
 			add_running_bw(p->dl.dl_bw, &this_rq->dl);
 			activate_task(this_rq, p, 0);
 			dmin = p->dl.deadline;
@@ -2057,6 +2102,9 @@ static void switched_from_dl(struct rq *rq, struct task_struct *p)
 	if (task_on_rq_queued(p) && p->dl.dl_runtime)
 		task_non_contending(p);
 
+	if (!task_on_rq_queued(p))
+		sub_rq_bw(p->dl.dl_bw, &rq->dl);
+
 	/*
 	 * We cannot use inactive_task_timer() to invoke sub_running_bw()
 	 * at the 0-lag time, because the task could have been migrated
@@ -2086,9 +2134,11 @@ static void switched_to_dl(struct rq *rq, struct task_struct *p)
 		put_task_struct(p);
 
 	/* If p is not queued we will update its parameters at next wakeup. */
-	if (!task_on_rq_queued(p))
+	if (!task_on_rq_queued(p)) {
-		return;
+		add_rq_bw(p->dl.dl_bw, &rq->dl);
 
+		return;
+	}
 	/*
 	 * If p is boosted we already updated its params in
 	 * rt_mutex_setprio()->enqueue_task(..., ENQUEUE_REPLENISH),

kernel/sched/sched.h

@@ -566,6 +566,17 @@ struct dl_rq {
 	 */
 	u64 running_bw;
 
+	/*
+	 * Utilization of the tasks "assigned" to this runqueue (including
+	 * the tasks that are in runqueue and the tasks that executed on this
+	 * CPU and blocked). Increased when a task moves to this runqueue, and
+	 * decreased when the task moves away (migrates, changes scheduling
+	 * policy, or terminates).
+	 * This is needed to compute the "inactive utilization" for the
+	 * runqueue (inactive utilization = this_bw - running_bw).
+	 */
+	u64 this_bw;
+
 	/*
 	 * Inverse of the fraction of CPU utilization that can be reclaimed
 	 * by the GRUB algorithm.