commit 95938a35c5562afa7af7252821e44132391a3db8
author Mike Galbraith <efault@gmx.de> Mon Oct 15 17:00:14 2007 +0200
committer Ingo Molnar <mingo@elte.hu> Mon Oct 15 17:00:14 2007 +0200
tree 548cda134d007105f3a819a438ab384ab9f04e27
parent ce6c131131df442f0d49d064129ecc52d9fe8ca9

sched: prevent wakeup over-scheduling

Prevent wakeup over-scheduling.  Once a task has been preempted by a
task of the same or lower priority, it becomes ineligible for repeated
preemption by same until it has been ticked, or slept.  Instead, the
task is marked for preemption at the next tick.  Tasks of higher
priority still preempt immediately.

Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

include/linux/sched.h

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 04233c8..8be5b57 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -912,6 +912,7 @@
 	struct load_weight	load;		/* for load-balancing */
 	struct rb_node		run_node;
 	unsigned int		on_rq;
+	int			peer_preempt;
 
 	u64			exec_start;
 	u64			sum_exec_runtime;

kernel/sched.c

diff --git a/kernel/sched.c b/kernel/sched.c
index 0bd8f2c..e8051bd 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -445,6 +445,7 @@
 	SCHED_FEAT_TREE_AVG             = 4,
 	SCHED_FEAT_APPROX_AVG           = 8,
 	SCHED_FEAT_WAKEUP_PREEMPT	= 16,
+	SCHED_FEAT_PREEMPT_RESTRICT	= 32,
 };
 
 const_debug unsigned int sysctl_sched_features =
@@ -452,7 +453,8 @@
 		SCHED_FEAT_START_DEBIT		*1 |
 		SCHED_FEAT_TREE_AVG		*0 |
 		SCHED_FEAT_APPROX_AVG		*0 |
-		SCHED_FEAT_WAKEUP_PREEMPT	*1;
+		SCHED_FEAT_WAKEUP_PREEMPT	*1 |
+		SCHED_FEAT_PREEMPT_RESTRICT	*1;
 
 #define sched_feat(x) (sysctl_sched_features & SCHED_FEAT_##x)
 

kernel/sched_fair.c

diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index 3843ec7..f819f94 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -526,6 +526,7 @@
 
 	update_stats_dequeue(cfs_rq, se);
 	if (sleep) {
+		se->peer_preempt = 0;
 #ifdef CONFIG_SCHEDSTATS
 		if (entity_is_task(se)) {
 			struct task_struct *tsk = task_of(se);
@@ -553,8 +554,10 @@
 
 	ideal_runtime = sched_slice(cfs_rq, curr);
 	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
-	if (delta_exec > ideal_runtime)
+	if (delta_exec > ideal_runtime ||
+			(sched_feat(PREEMPT_RESTRICT) && curr->peer_preempt))
 		resched_task(rq_of(cfs_rq)->curr);
+	curr->peer_preempt = 0;
 }
 
 static void
@@ -839,8 +842,12 @@
 		if (unlikely(se->load.weight != NICE_0_LOAD))
 			gran = calc_delta_fair(gran, &se->load);
 
-		if (delta > gran)
-			resched_task(curr);
+		if (delta > gran) {
+			int now = !sched_feat(PREEMPT_RESTRICT);
+
+			if (now || p->prio < curr->prio || !se->peer_preempt++)
+				resched_task(curr);
+		}
 	}
 }
 
@@ -1034,6 +1041,7 @@
 	check_spread(cfs_rq, curr);
 	__enqueue_entity(cfs_rq, se);
 	account_entity_enqueue(cfs_rq, se);
+	se->peer_preempt = 0;
 	resched_task(rq->curr);
 }