Re: Regression seen for patch "sched:dont decrease idle sleep avg"

[Mike Galbraith <efault@gmx.de>]

On Thu, 2006-05-18 at 09:04 +0200, Mike Galbraith wrote:

> OK, after some brief testing, I think this is a step in the right
> direct, but there is another problem.  In the case where the queue isn't
> empty, the stated intent is utterly defeated by the on runqueue bonus.

The overly verbose one liner below could serve as a minimal ~fix.

Prevent the on-runqueue bonus logic from defeating the idle sleep logic.

Signed-off-by: Mike Galbraith <efault@gmx.de>

--- linux-2.6.17-rc4-mm1/kernel/sched.c.org	2006-05-18 08:38:13.000000000 +0200
+++ linux-2.6.17-rc4-mm1/kernel/sched.c	2006-05-18 14:47:09.000000000 +0200
@@ -917,6 +917,16 @@ static int recalc_task_prio(task_t *p, u
 			 * with one single large enough sleep.
 			 */
 			p->sleep_avg = ceiling;
+			/*
+			 * Using INTERACTIVE_SLEEP() as a ceiling places a
+			 * nice(0) task 1ms sleep away from promotion, and
+			 * gives it 700ms to round-robin with no chance of
+			 * being demoted.  This is more than generous, so
+			 * mark this sleep as non-interactive to prevent the
+			 * on-runqueue bonus logic from intervening should
+			 * this task not receive cpu immediately.
+			 */
+			p->sleep_type = SLEEP_NONINTERACTIVE;
 		} else {
 			/*
 			 * Tasks waking from uninterruptible sleep are