memcg: fix hierarchical reclaim
authorDaisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Thu, 15 Jan 2009 21:51:13 +0000 (13:51 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 16 Jan 2009 00:39:39 +0000 (16:39 -0800)
If root_mem has no children, last_scaned_child is set to root_mem itself.
But after some children added to root_mem, mem_cgroup_get_next_node can
mem_cgroup_put the root_mem although root_mem has not been mem_cgroup_get.

This patch fixes this behavior by:

- Set last_scanned_child to NULL if root_mem has no children or DFS
  search has returned to root_mem itself(root_mem is not a "child" of
  root_mem).  Make mem_cgroup_get_first_node return root_mem in this case.
   There are no mem_cgroup_get/put for root_mem.

- Rename mem_cgroup_get_next_node to __mem_cgroup_get_next_node, and
  mem_cgroup_get_first_node to mem_cgroup_get_next_node.  Make
  mem_cgroup_hierarchical_reclaim call only new mem_cgroup_get_next_node.

Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memcontrol.c

index 7be9b35d7ffb293ac6296354ea1e3493d5ef09d6..322625f551c27504c89d3f0f743bd58da32f909f 100644 (file)
@@ -633,7 +633,7 @@ unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
  * called with hierarchy_mutex held
  */
 static struct mem_cgroup *
-mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem)
+__mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem)
 {
        struct cgroup *cgroup, *curr_cgroup, *root_cgroup;
 
@@ -644,19 +644,16 @@ mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem)
                /*
                 * Walk down to children
                 */
-               mem_cgroup_put(curr);
                cgroup = list_entry(curr_cgroup->children.next,
                                                struct cgroup, sibling);
                curr = mem_cgroup_from_cont(cgroup);
-               mem_cgroup_get(curr);
                goto done;
        }
 
 visit_parent:
        if (curr_cgroup == root_cgroup) {
-               mem_cgroup_put(curr);
-               curr = root_mem;
-               mem_cgroup_get(curr);
+               /* caller handles NULL case */
+               curr = NULL;
                goto done;
        }
 
@@ -664,11 +661,9 @@ visit_parent:
         * Goto next sibling
         */
        if (curr_cgroup->sibling.next != &curr_cgroup->parent->children) {
-               mem_cgroup_put(curr);
                cgroup = list_entry(curr_cgroup->sibling.next, struct cgroup,
                                                sibling);
                curr = mem_cgroup_from_cont(cgroup);
-               mem_cgroup_get(curr);
                goto done;
        }
 
@@ -679,7 +674,6 @@ visit_parent:
        goto visit_parent;
 
 done:
-       root_mem->last_scanned_child = curr;
        return curr;
 }
 
@@ -689,40 +683,46 @@ done:
  * that to reclaim free pages from.
  */
 static struct mem_cgroup *
-mem_cgroup_get_first_node(struct mem_cgroup *root_mem)
+mem_cgroup_get_next_node(struct mem_cgroup *root_mem)
 {
        struct cgroup *cgroup;
-       struct mem_cgroup *ret;
+       struct mem_cgroup *orig, *next;
        bool obsolete;
 
-       obsolete = mem_cgroup_is_obsolete(root_mem->last_scanned_child);
-
        /*
         * Scan all children under the mem_cgroup mem
         */
        mutex_lock(&mem_cgroup_subsys.hierarchy_mutex);
+
+       orig = root_mem->last_scanned_child;
+       obsolete = mem_cgroup_is_obsolete(orig);
+
        if (list_empty(&root_mem->css.cgroup->children)) {
-               ret = root_mem;
+               /*
+                * root_mem might have children before and last_scanned_child
+                * may point to one of them. We put it later.
+                */
+               if (orig)
+                       VM_BUG_ON(!obsolete);
+               next = NULL;
                goto done;
        }
 
-       if (!root_mem->last_scanned_child || obsolete) {
-
-               if (obsolete && root_mem->last_scanned_child)
-                       mem_cgroup_put(root_mem->last_scanned_child);
-
+       if (!orig || obsolete) {
                cgroup = list_first_entry(&root_mem->css.cgroup->children,
                                struct cgroup, sibling);
-               ret = mem_cgroup_from_cont(cgroup);
-               mem_cgroup_get(ret);
+               next = mem_cgroup_from_cont(cgroup);
        } else
-               ret = mem_cgroup_get_next_node(root_mem->last_scanned_child,
-                                               root_mem);
+               next = __mem_cgroup_get_next_node(orig, root_mem);
 
 done:
-       root_mem->last_scanned_child = ret;
+       if (next)
+               mem_cgroup_get(next);
+       root_mem->last_scanned_child = next;
+       if (orig)
+               mem_cgroup_put(orig);
        mutex_unlock(&mem_cgroup_subsys.hierarchy_mutex);
-       return ret;
+       return (next) ? next : root_mem;
 }
 
 static bool mem_cgroup_check_under_limit(struct mem_cgroup *mem)
@@ -780,21 +780,18 @@ static int mem_cgroup_hierarchical_reclaim(struct mem_cgroup *root_mem,
        if (!root_mem->use_hierarchy)
                return ret;
 
-       next_mem = mem_cgroup_get_first_node(root_mem);
+       next_mem = mem_cgroup_get_next_node(root_mem);
 
        while (next_mem != root_mem) {
                if (mem_cgroup_is_obsolete(next_mem)) {
-                       mem_cgroup_put(next_mem);
-                       next_mem = mem_cgroup_get_first_node(root_mem);
+                       next_mem = mem_cgroup_get_next_node(root_mem);
                        continue;
                }
                ret = try_to_free_mem_cgroup_pages(next_mem, gfp_mask, noswap,
                                                   get_swappiness(next_mem));
                if (mem_cgroup_check_under_limit(root_mem))
                        return 0;
-               mutex_lock(&mem_cgroup_subsys.hierarchy_mutex);
-               next_mem = mem_cgroup_get_next_node(next_mem, root_mem);
-               mutex_unlock(&mem_cgroup_subsys.hierarchy_mutex);
+               next_mem = mem_cgroup_get_next_node(root_mem);
        }
        return ret;
 }
@@ -2254,7 +2251,14 @@ static void mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
 static void mem_cgroup_destroy(struct cgroup_subsys *ss,
                                struct cgroup *cont)
 {
-       mem_cgroup_put(mem_cgroup_from_cont(cont));
+       struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
+       struct mem_cgroup *last_scanned_child = mem->last_scanned_child;
+
+       if (last_scanned_child) {
+               VM_BUG_ON(!mem_cgroup_is_obsolete(last_scanned_child));
+               mem_cgroup_put(last_scanned_child);
+       }
+       mem_cgroup_put(mem);
 }
 
 static int mem_cgroup_populate(struct cgroup_subsys *ss,