sched: introduce se->vruntime

introduce se->vruntime as a sum of weighted delta-exec's, and use that
as the key into the tree.

the idea to use absolute virtual time as the basic metric of scheduling
has been first raised by William Lee Irwin, advanced by Tong Li and first
prototyped by Roman Zippel in the "Really Fair Scheduler" (RFS) patchset.

also see:

   http://lkml.org/lkml/2007/9/2/76

for a simpler variant of this patch.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>

3 files changed, 50 insertions, 33 deletions

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 3c38a5040e8f..5e5c457fba86 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -902,6 +902,7 @@ struct sched_entity {
 
 	u64			exec_start;
 	u64			sum_exec_runtime;
+	u64			vruntime;
 	u64			prev_sum_exec_runtime;
 	u64			wait_start_fair;
 	u64			sleep_start_fair;
diff --git a/kernel/sched.c b/kernel/sched.c
index 992a1fae72a7..8f80ebafacc1 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -182,6 +182,7 @@ struct cfs_rq {
 
 	s64 fair_clock;
 	u64 exec_clock;
+	u64 min_vruntime;
 	s64 wait_runtime;
 	u64 sleeper_bonus;
 	unsigned long wait_runtime_overruns, wait_runtime_underruns;
diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index b46f8078e78f..a2af09cb6a70 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -92,14 +92,16 @@ unsigned int sysctl_sched_runtime_limit __read_mostly;
  */
 enum {
 	SCHED_FEAT_FAIR_SLEEPERS	= 1,
-	SCHED_FEAT_SLEEPER_AVG		= 2,
-	SCHED_FEAT_SLEEPER_LOAD_AVG	= 4,
-	SCHED_FEAT_START_DEBIT		= 8,
-	SCHED_FEAT_SKIP_INITIAL		= 16,
+	SCHED_FEAT_NEW_FAIR_SLEEPERS	= 2,
+	SCHED_FEAT_SLEEPER_AVG		= 4,
+	SCHED_FEAT_SLEEPER_LOAD_AVG	= 8,
+	SCHED_FEAT_START_DEBIT		= 16,
+	SCHED_FEAT_SKIP_INITIAL		= 32,
 };
 
 const_debug unsigned int sysctl_sched_features =
-		SCHED_FEAT_FAIR_SLEEPERS	*1 |
+		SCHED_FEAT_FAIR_SLEEPERS	*0 |
+		SCHED_FEAT_NEW_FAIR_SLEEPERS	*1 |
 		SCHED_FEAT_SLEEPER_AVG		*0 |
 		SCHED_FEAT_SLEEPER_LOAD_AVG	*1 |
 		SCHED_FEAT_START_DEBIT		*1 |
@@ -145,6 +147,19 @@ static inline struct task_struct *task_of(struct sched_entity *se)
  * Scheduling class tree data structure manipulation methods:
  */
 
+static inline void
+set_leftmost(struct cfs_rq *cfs_rq, struct rb_node *leftmost)
+{
+	struct sched_entity *se;
+
+	cfs_rq->rb_leftmost = leftmost;
+	if (leftmost) {
+		se = rb_entry(leftmost, struct sched_entity, run_node);
+		cfs_rq->min_vruntime = max(se->vruntime,
+						cfs_rq->min_vruntime);
+	}
+}
+
 /*
  * Enqueue an entity into the rb-tree:
  */
@@ -180,7 +195,7 @@ __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	 * used):
 	 */
 	if (leftmost)
-		cfs_rq->rb_leftmost = &se->run_node;
+		set_leftmost(cfs_rq, &se->run_node);
 
 	rb_link_node(&se->run_node, parent, link);
 	rb_insert_color(&se->run_node, &cfs_rq->tasks_timeline);
@@ -195,7 +210,8 @@ static void
 __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	if (cfs_rq->rb_leftmost == &se->run_node)
-		cfs_rq->rb_leftmost = rb_next(&se->run_node);
+		set_leftmost(cfs_rq, rb_next(&se->run_node));
+
 	rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
 	update_load_sub(&cfs_rq->load, se->load.weight);
 	cfs_rq->nr_running--;
@@ -336,7 +352,7 @@ static inline void
 __update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
 	      unsigned long delta_exec)
 {
-	unsigned long delta, delta_fair, delta_mine;
+	unsigned long delta, delta_fair, delta_mine, delta_exec_weighted;
 	struct load_weight *lw = &cfs_rq->load;
 	unsigned long load = lw->weight;
 
@@ -344,6 +360,12 @@ __update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
 
 	curr->sum_exec_runtime += delta_exec;
 	cfs_rq->exec_clock += delta_exec;
+	delta_exec_weighted = delta_exec;
+	if (unlikely(curr->load.weight != NICE_0_LOAD)) {
+		delta_exec_weighted = calc_delta_fair(delta_exec_weighted,
+							&curr->load);
+	}
+	curr->vruntime += delta_exec_weighted;
 
 	if (unlikely(!load))
 		return;
@@ -413,8 +435,6 @@ calc_weighted(unsigned long delta, struct sched_entity *se)
  */
 static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	s64 key;
-
 	/*
 	 * Are we enqueueing a waiting task? (for current tasks
 	 * a dequeue/enqueue event is a NOP)
@@ -424,28 +444,7 @@ static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	/*
 	 * Update the key:
 	 */
-	key = cfs_rq->fair_clock;
-
-	/*
-	 * Optimize the common nice 0 case:
-	 */
-	if (likely(se->load.weight == NICE_0_LOAD)) {
-		key -= se->wait_runtime;
-	} else {
-		u64 tmp;
-
-		if (se->wait_runtime < 0) {
-			tmp = -se->wait_runtime;
-			key += (tmp * se->load.inv_weight) >>
-					(WMULT_SHIFT - NICE_0_SHIFT);
-		} else {
-			tmp = se->wait_runtime;
-			key -= (tmp * se->load.inv_weight) >>
-					(WMULT_SHIFT - NICE_0_SHIFT);
-		}
-	}
-
-	se->fair_key = key;
+	se->fair_key = se->vruntime;
 }
 
 /*
@@ -615,8 +614,22 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
 	 */
 	update_curr(cfs_rq);
 
-	if (wakeup)
+	if (wakeup) {
+		u64 min_runtime, latency;
+
+		min_runtime = cfs_rq->min_vruntime;
+		min_runtime += sysctl_sched_latency/2;
+
+		if (sched_feat(NEW_FAIR_SLEEPERS)) {
+			latency = calc_weighted(sysctl_sched_latency, se);
+			if (min_runtime > latency)
+				min_runtime -= latency;
+		}
+
+		se->vruntime = max(se->vruntime, min_runtime);
+
 		enqueue_sleeper(cfs_rq, se);
+	}
 
 	update_stats_enqueue(cfs_rq, se);
 	__enqueue_entity(cfs_rq, se);
@@ -1155,6 +1168,8 @@ static void task_new_fair(struct rq *rq, struct task_struct *p)
 	if (sched_feat(START_DEBIT))
 		se->wait_runtime = -(sched_granularity(cfs_rq) / 2);
 
+	se->vruntime = cfs_rq->min_vruntime;
+	update_stats_enqueue(cfs_rq, se);
 	__enqueue_entity(cfs_rq, se);
 	resched_task(rq->curr);
 }