[PATCH, 2.6.9] improved load_balance() tolerance for pinned tasks

[John Hawkes <hawkes@oss.sgi.com>]

A large number of processes that are pinned to a single CPU results in
every other CPU's load_balance() seeing this overloaded CPU as "busiest",
yet move_tasks() never finds a task to pull-migrate.  This condition
occurs during module unload, but can also occur as a denial-of-service
using sys_sched_setaffinity().  Several hundred CPUs performing this
fruitless load_balance() will livelock on the busiest CPU's runqueue
lock.  A smaller number of CPUs will livelock if the pinned task count
gets high.  This simple patch remedies the more common first problem:
after a move_tasks() failure to migrate anything, the balance_interval
increments.  Using a simple increment, vs.  the more dramatic doubling of
the balance_interval, is conservative and yet also effective.

John Hawkes


Signed-off-by: John Hawkes <hawkes@sgi.com>




Index: linux/kernel/sched.c
===================================================================
--- linux.orig/kernel/sched.c	2004-10-19 15:04:11.000000000 -0700
+++ linux/kernel/sched.c	2004-10-19 15:09:50.000000000 -0700
@@ -2123,11 +2123,19 @@
 			 */
 			sd->nr_balance_failed = sd->cache_nice_tries;
 		}
-	} else
-		sd->nr_balance_failed = 0;
 
-	/* We were unbalanced, so reset the balancing interval */
-	sd->balance_interval = sd->min_interval;
+		/*
+		 * We were unbalanced, but unsuccessful in move_tasks(),
+		 * so bump the balance_interval to lessen the lock contention.
+		 */
+		if (sd->balance_interval < sd->max_interval)
+			sd->balance_interval++;
+	} else {
+                sd->nr_balance_failed = 0;
+
+		/* We were unbalanced, so reset the balancing interval */
+		sd->balance_interval = sd->min_interval;
+	}
 
 	return nr_moved;