[PATCH] sched: allow the load to grow upto its cpu_power
authorSiddha, Suresh B <suresh.b.siddha@intel.com>
Sat, 10 Sep 2005 07:26:21 +0000 (00:26 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Sat, 10 Sep 2005 17:06:24 +0000 (10:06 -0700)
Don't pull tasks from a group if that would cause the group's total load to
drop below its total cpu_power (ie.  cause the group to start going idle).

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
kernel/sched.c

index 1dc29dec38a9dd8c4baa95c081ac24db65a23e9d..dbd4490afec14b1f2e71d11099feb04331836509 100644 (file)
@@ -1910,6 +1910,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
 {
        struct sched_group *busiest = NULL, *this = NULL, *group = sd->groups;
        unsigned long max_load, avg_load, total_load, this_load, total_pwr;
+       unsigned long max_pull;
        int load_idx;
 
        max_load = this_load = total_load = total_pwr = 0;
@@ -1959,7 +1960,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                group = group->next;
        } while (group != sd->groups);
 
-       if (!busiest || this_load >= max_load)
+       if (!busiest || this_load >= max_load || max_load <= SCHED_LOAD_SCALE)
                goto out_balanced;
 
        avg_load = (SCHED_LOAD_SCALE * total_load) / total_pwr;
@@ -1979,8 +1980,12 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
         * by pulling tasks to us.  Be careful of negative numbers as they'll
         * appear as very large values with unsigned longs.
         */
+
+       /* Don't want to pull so many tasks that a group would go idle */
+       max_pull = min(max_load - avg_load, max_load - SCHED_LOAD_SCALE);
+
        /* How much load to actually move to equalise the imbalance */
-       *imbalance = min((max_load - avg_load) * busiest->cpu_power,
+       *imbalance = min(max_pull * busiest->cpu_power,
                                (avg_load - this_load) * this->cpu_power)
                        / SCHED_LOAD_SCALE;