memcg: oom wakeup filter

memcg's oom waitqueue is a system-wide wait_queue (for handling
hierarchy.) So, it's better to add custom wake function and do filtering
in wake up path.

This patch adds a filtering feature for waking up oom-waiters.  Hierarchy
is properly handled.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcontrol.c

@@ -1293,14 +1293,56 @@ static void mem_cgroup_oom_unlock(struct mem_cgroup *mem)
 static DEFINE_MUTEX(memcg_oom_mutex);
 static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);
 
+struct oom_wait_info {
+	struct mem_cgroup *mem;
+	wait_queue_t	wait;
+};
+
+static int memcg_oom_wake_function(wait_queue_t *wait,
+	unsigned mode, int sync, void *arg)
+{
+	struct mem_cgroup *wake_mem = (struct mem_cgroup *)arg;
+	struct oom_wait_info *oom_wait_info;
+
+	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
+
+	if (oom_wait_info->mem == wake_mem)
+		goto wakeup;
+	/* if no hierarchy, no match */
+	if (!oom_wait_info->mem->use_hierarchy || !wake_mem->use_hierarchy)
+		return 0;
+	/*
+	 * Both of oom_wait_info->mem and wake_mem are stable under us.
+	 * Then we can use css_is_ancestor without taking care of RCU.
+	 */
+	if (!css_is_ancestor(&oom_wait_info->mem->css, &wake_mem->css) &&
+	    !css_is_ancestor(&wake_mem->css, &oom_wait_info->mem->css))
+		return 0;
+
+wakeup:
+	return autoremove_wake_function(wait, mode, sync, arg);
+}
+
+static void memcg_wakeup_oom(struct mem_cgroup *mem)
+{
+	/* for filtering, pass "mem" as argument. */
+	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, mem);
+}
+
 /*
  * try to call OOM killer. returns false if we should exit memory-reclaim loop.
  */
 bool mem_cgroup_handle_oom(struct mem_cgroup *mem, gfp_t mask)
 {
-	DEFINE_WAIT(wait);
+	struct oom_wait_info owait;
 	bool locked;
 
+	owait.mem = mem;
+	owait.wait.flags = 0;
+	owait.wait.func = memcg_oom_wake_function;
+	owait.wait.private = current;
+	INIT_LIST_HEAD(&owait.wait.task_list);
+
 	/* At first, try to OOM lock hierarchy under mem.*/
 	mutex_lock(&memcg_oom_mutex);
 	locked = mem_cgroup_oom_lock(mem);
@@ -1310,31 +1352,18 @@ bool mem_cgroup_handle_oom(struct mem_cgroup *mem, gfp_t mask)
 	 * under OOM is always welcomed, use TASK_KILLABLE here.
 	 */
 	if (!locked)
-		prepare_to_wait(&memcg_oom_waitq, &wait, TASK_KILLABLE);
+		prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
 	mutex_unlock(&memcg_oom_mutex);
 
 	if (locked)
 		mem_cgroup_out_of_memory(mem, mask);
 	else {
 		schedule();
-		finish_wait(&memcg_oom_waitq, &wait);
+		finish_wait(&memcg_oom_waitq, &owait.wait);
 	}
 	mutex_lock(&memcg_oom_mutex);
 	mem_cgroup_oom_unlock(mem);
-	/*
-	 * Here, we use global waitq .....more fine grained waitq ?
-	 * Assume following hierarchy.
-	 * A/
-	 *   01
-	 *   02
-	 * assume OOM happens both in A and 01 at the same time. Tthey are
-	 * mutually exclusive by lock. (kill in 01 helps A.)
-	 * When we use per memcg waitq, we have to wake up waiters on A and 02
-	 * in addtion to waiters on 01. We use global waitq for avoiding mess.
-	 * It will not be a big problem.
-	 * (And a task may be moved to other groups while it's waiting for OOM.)
-	 */
-	wake_up_all(&memcg_oom_waitq);
+	memcg_wakeup_oom(mem);
 	mutex_unlock(&memcg_oom_mutex);
 
 	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))