memcg: track all children over limit in the root
authorMichal Hocko <mhocko@suse.cz>
Thu, 12 Sep 2013 22:13:32 +0000 (15:13 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 12 Sep 2013 22:38:01 +0000 (15:38 -0700)
Children in soft limit excess are currently tracked up the hierarchy in
memcg->children_in_excess.  Nevertheless there still might exist tons of
groups that are not in hierarchy relation to the root cgroup (e.g.  all
first level groups if root_mem_cgroup->use_hierarchy == false).

As the whole tree walk has to be done when the iteration starts at
root_mem_cgroup the iterator should be able to skip the walk if there is
no child above the limit without iterating them.  This can be done
easily if the root tracks all children rather than only hierarchical
children.  This is done by this patch which updates root_mem_cgroup
children_in_excess if root_mem_cgroup->use_hierarchy == false so the
root knows about all children in excess.

Please note that this is not an issue for inner memcgs which have
use_hierarchy == false because then only the single group is visited so
no special optimization is necessary.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Ying Han <yinghan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memcontrol.c

index 848fc6c..46717d6 100644 (file)
@@ -864,9 +864,15 @@ static void mem_cgroup_update_soft_limit(struct mem_cgroup *memcg)
        /*
         * Necessary to update all ancestors when hierarchy is used
         * because their event counter is not touched.
+        * We track children even outside the hierarchy for the root
+        * cgroup because tree walk starting at root should visit
+        * all cgroups and we want to prevent from pointless tree
+        * walk if no children is below the limit.
         */
        while (delta && (parent = parent_mem_cgroup(parent)))
                atomic_add(delta, &parent->children_in_excess);
+       if (memcg != root_mem_cgroup && !root_mem_cgroup->use_hierarchy)
+               atomic_add(delta, &root_mem_cgroup->children_in_excess);
        spin_unlock(&memcg->soft_lock);
 }
 
@@ -6043,6 +6049,9 @@ static void mem_cgroup_css_offline(struct cgroup_subsys_state *css)
        if (memcg->soft_contributed) {
                while ((memcg = parent_mem_cgroup(memcg)))
                        atomic_dec(&memcg->children_in_excess);
+
+               if (memcg != root_mem_cgroup && !root_mem_cgroup->use_hierarchy)
+                       atomic_dec(&root_mem_cgroup->children_in_excess);
        }
        mem_cgroup_destroy_all_caches(memcg);
        vmpressure_cleanup(&memcg->vmpressure);