On 9 October 2014 16:16, Peter Zijlstra [off-list ref] wrote:

On Tue, Oct 07, 2014 at 02:13:36PM +0200, Vincent Guittot wrote:

quoted

@@ -6214,17 +6178,21 @@ static inline void update_sd_lb_stats(struct lb_env *env, struct sd_lb_stats *sd

              /*
               * In case the child domain prefers tasks go to siblings
+              * first, lower the sg capacity to one so that we'll try
               * and move all the excess tasks away. We lower the capacity
               * of a group only if the local group has the capacity to fit
+              * these excess tasks, i.e. group_capacity > 0. The
               * extra check prevents the case where you always pull from the
               * heaviest group when it is already under-utilized (possible
               * with a large weight task outweighs the tasks on the system).
               */
              if (prefer_sibling && sds->local &&
+                 group_has_capacity(env, &sds->local_stat)) {
+                     if (sgs->sum_nr_running > 1)
+                             sgs->group_no_capacity = 1;
+                     sgs->group_capacity = min(sgs->group_capacity,
+                                             SCHED_CAPACITY_SCALE);
+             }

              if (update_sd_pick_busiest(env, sds, sg, sgs)) {
                      sds->busiest = sg;

quoted

@@ -6490,8 +6460,8 @@ static struct sched_group *find_busiest_group(struct lb_env *env)
              goto force_balance;

      /* SD_BALANCE_NEWIDLE trumps SMP nice when underutilized */
-     if (env->idle == CPU_NEWLY_IDLE && local->group_has_free_capacity &&
-         !busiest->group_has_free_capacity)
+     if (env->idle == CPU_NEWLY_IDLE && group_has_capacity(env, local) &&
+         busiest->group_no_capacity)
              goto force_balance;

      /*

This is two calls to group_has_capacity() on the local group. Why not
compute once in update_sd_lb_stats()?

mainly because of the place in the code, so it is not always used
during load balance unlike group_no_capacity

Now that i have said that, i just noticed that it's better to move the
call to the last tested condition

 +     if (env->idle == CPU_NEWLY_IDLE && busiest->group_no_capacity &&
 +         group_has_capacity(env, local))