Commit c268e994 authored by Daisuke Nishimura's avatar Daisuke Nishimura Committed by Linus Torvalds

memcg: fix hierarchical reclaim

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: default avatarDaisuke 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: default avatarAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
parent 40d58138
...@@ -633,7 +633,7 @@ unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan, ...@@ -633,7 +633,7 @@ unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
* called with hierarchy_mutex held * called with hierarchy_mutex held
*/ */
static struct mem_cgroup * 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; 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) ...@@ -644,19 +644,16 @@ mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem)
/* /*
* Walk down to children * Walk down to children
*/ */
mem_cgroup_put(curr);
cgroup = list_entry(curr_cgroup->children.next, cgroup = list_entry(curr_cgroup->children.next,
struct cgroup, sibling); struct cgroup, sibling);
curr = mem_cgroup_from_cont(cgroup); curr = mem_cgroup_from_cont(cgroup);
mem_cgroup_get(curr);
goto done; goto done;
} }
visit_parent: visit_parent:
if (curr_cgroup == root_cgroup) { if (curr_cgroup == root_cgroup) {
mem_cgroup_put(curr); /* caller handles NULL case */
curr = root_mem; curr = NULL;
mem_cgroup_get(curr);
goto done; goto done;
} }
...@@ -664,11 +661,9 @@ mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem) ...@@ -664,11 +661,9 @@ mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem)
* Goto next sibling * Goto next sibling
*/ */
if (curr_cgroup->sibling.next != &curr_cgroup->parent->children) { if (curr_cgroup->sibling.next != &curr_cgroup->parent->children) {
mem_cgroup_put(curr);
cgroup = list_entry(curr_cgroup->sibling.next, struct cgroup, cgroup = list_entry(curr_cgroup->sibling.next, struct cgroup,
sibling); sibling);
curr = mem_cgroup_from_cont(cgroup); curr = mem_cgroup_from_cont(cgroup);
mem_cgroup_get(curr);
goto done; goto done;
} }
...@@ -679,7 +674,6 @@ mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem) ...@@ -679,7 +674,6 @@ mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem)
goto visit_parent; goto visit_parent;
done: done:
root_mem->last_scanned_child = curr;
return curr; return curr;
} }
...@@ -689,40 +683,46 @@ mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem) ...@@ -689,40 +683,46 @@ mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem)
* that to reclaim free pages from. * that to reclaim free pages from.
*/ */
static struct mem_cgroup * 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 cgroup *cgroup;
struct mem_cgroup *ret; struct mem_cgroup *orig, *next;
bool obsolete; bool obsolete;
obsolete = mem_cgroup_is_obsolete(root_mem->last_scanned_child);
/* /*
* Scan all children under the mem_cgroup mem * Scan all children under the mem_cgroup mem
*/ */
mutex_lock(&mem_cgroup_subsys.hierarchy_mutex); 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)) { 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; goto done;
} }
if (!root_mem->last_scanned_child || obsolete) { if (!orig || obsolete) {
if (obsolete && root_mem->last_scanned_child)
mem_cgroup_put(root_mem->last_scanned_child);
cgroup = list_first_entry(&root_mem->css.cgroup->children, cgroup = list_first_entry(&root_mem->css.cgroup->children,
struct cgroup, sibling); struct cgroup, sibling);
ret = mem_cgroup_from_cont(cgroup); next = mem_cgroup_from_cont(cgroup);
mem_cgroup_get(ret);
} else } else
ret = mem_cgroup_get_next_node(root_mem->last_scanned_child, next = __mem_cgroup_get_next_node(orig, root_mem);
root_mem);
done: 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); 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) 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, ...@@ -780,21 +780,18 @@ static int mem_cgroup_hierarchical_reclaim(struct mem_cgroup *root_mem,
if (!root_mem->use_hierarchy) if (!root_mem->use_hierarchy)
return ret; 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) { while (next_mem != root_mem) {
if (mem_cgroup_is_obsolete(next_mem)) { if (mem_cgroup_is_obsolete(next_mem)) {
mem_cgroup_put(next_mem); next_mem = mem_cgroup_get_next_node(root_mem);
next_mem = mem_cgroup_get_first_node(root_mem);
continue; continue;
} }
ret = try_to_free_mem_cgroup_pages(next_mem, gfp_mask, noswap, ret = try_to_free_mem_cgroup_pages(next_mem, gfp_mask, noswap,
get_swappiness(next_mem)); get_swappiness(next_mem));
if (mem_cgroup_check_under_limit(root_mem)) if (mem_cgroup_check_under_limit(root_mem))
return 0; return 0;
mutex_lock(&mem_cgroup_subsys.hierarchy_mutex); next_mem = mem_cgroup_get_next_node(root_mem);
next_mem = mem_cgroup_get_next_node(next_mem, root_mem);
mutex_unlock(&mem_cgroup_subsys.hierarchy_mutex);
} }
return ret; return ret;
} }
...@@ -2254,7 +2251,14 @@ static void mem_cgroup_pre_destroy(struct cgroup_subsys *ss, ...@@ -2254,7 +2251,14 @@ static void mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
static void mem_cgroup_destroy(struct cgroup_subsys *ss, static void mem_cgroup_destroy(struct cgroup_subsys *ss,
struct cgroup *cont) 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, static int mem_cgroup_populate(struct cgroup_subsys *ss,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment