[PATCH 2/5] sched-alter_uninterruptible_sleep_interactivity.patch

[Con Kolivas <kernel@kolivas.org>]

The interactivity estimator special cases tasks that are waking up from
uninterruptible sleep based on the fact that most uninterruptible sleep
represents a task waiting on disk I/O and is not truly interactive. The
current system uses a ceiling to the priority bonus said tasks can receive.
The problem with that system is that if there are enough interactive tasks
at high bonus levels it can lead to I/O starvation.

In order to remove the ceiling but still maintain some special case treatment
of uninterruptible sleep, we can make any sleep_avg incrementing to be purely
based on sleep time instead of being biased in the non-linear fashion that
interactive tasks are.

This will lead to a detriment in interactive behaviour under disk I/O however
the current system unfairly biases against them and leads to a loss of disk
throughput. This change should restore a better balance between disk
throughput and interactivity.

Signed-off-by: Con Kolivas <kernel@kolivas.org>

 kernel/sched.c |   36 ++++++++++++------------------------
 1 files changed, 12 insertions(+), 24 deletions(-)

Index: linux-2.6.15/kernel/sched.c
===================================================================
--- linux-2.6.15.orig/kernel/sched.c
+++ linux-2.6.15/kernel/sched.c
@@ -756,26 +756,17 @@ static int recalc_task_prio(task_t *p, u
 				p->sleep_avg = JIFFIES_TO_NS(MAX_SLEEP_AVG -
 						DEF_TIMESLICE);
 		} else {
-			/*
-			 * The lower the sleep avg a task has the more
-			 * rapidly it will rise with sleep time.
-			 */
-			sleep_time *= (MAX_BONUS - CURRENT_BONUS(p)) ? : 1;
 
 			/*
-			 * Tasks waking from uninterruptible sleep are
-			 * limited in their sleep_avg rise as they
-			 * are likely to be waiting on I/O
+			 * The lower the sleep avg a task has the more
+			 * rapidly it will rise with sleep time. This enables
+			 * tasks to rapidly recover to a low latency priority.
+			 * If a task was sleeping with the noninteractive
+			 * label do not apply this non-linear boost
 			 */
-			if (p->sleep_type == SLEEP_NONINTERACTIVE && p->mm) {
-				if (p->sleep_avg >= INTERACTIVE_SLEEP(p))
-					sleep_time = 0;
-				else if (p->sleep_avg + sleep_time >=
-						INTERACTIVE_SLEEP(p)) {
-					p->sleep_avg = INTERACTIVE_SLEEP(p);
-					sleep_time = 0;
-				}
-			}
+			if (p->sleep_type != SLEEP_NONINTERACTIVE || p->mm)
+				sleep_time *=
+					(MAX_BONUS - CURRENT_BONUS(p)) ? : 1;
 
 			/*
 			 * This code gives a bonus to interactive tasks.
@@ -818,11 +809,7 @@ static void activate_task(task_t *p, run
 	if (!rt_task(p))
 		p->prio = recalc_task_prio(p, now);
 
-	/*
-	 * This checks to make sure it's not an uninterruptible task
-	 * that is now waking up.
-	 */
-	if (p->sleep_type == SLEEP_NORMAL) {
+	if (p->sleep_type != SLEEP_NONINTERACTIVE) {
 		/*
 		 * Tasks which were woken up by interrupts (ie. hw events)
 		 * are most likely of interactive nature. So we give them
@@ -1356,8 +1343,9 @@ out_activate:
 	if (old_state == TASK_UNINTERRUPTIBLE) {
 		rq->nr_uninterruptible--;
 		/*
-		 * Tasks on involuntary sleep don't earn
-		 * sleep_avg beyond just interactive state.
+		 * Tasks waking from uninterruptible sleep are likely
+		 * to be sleeping involuntarily on I/O and are otherwise
+		 * cpu bound so label them as noninteractive.
 		 */
 		p->sleep_type = SLEEP_NONINTERACTIVE;
 	}

Attachment: pgp89CPxKIipY.pgp
Description: PGP signature