memcg: new force_empty to free pages under group

By memcg-move-all-accounts-to-parent-at-rmdir.patch, there is no leak of memory usage and force_empty is removed. This patch adds "force_empty" again, in reasonable manner. memory.force_empty file works when #echo 0 (or some) > memory.force_empty and have following function. 1. only works when there are no task in this cgroup. 2. free all page under this cgroup as much as possible. 3. page which cannot be freed will be moved up to parent. 4. Then, memcg will be empty after above echo returns. This is much better behavior than old "force_empty" which just forget all accounts. This patch also check signal_pending() and above "echo" can be stopped by "Ctrl-C". [akpm@linux-foundation.org: cleanup] Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Li Zefan <lizf@cn.fujitsu.com> Cc: Balbir Singh <balbir@in.ibm.com> Cc: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: new force_empty to free pages under group
By memcg-move-all-accounts-to-parent-at-rmdir.patch, there is no leak of memory usage and force_empty is removed. This patch adds "force_empty" again, in reasonable manner. memory.force_empty file works when #echo 0 (or some) > memory.force_empty and have following function. 1. only works when there are no task in this cgroup. 2. free all page under this cgroup as much as possible. 3. page which cannot be freed will be moved up to parent. 4. Then, memcg will be empty after above echo returns. This is much better behavior than old "force_empty" which just forget all accounts. This patch also check signal_pending() and above "echo" can be stopped by "Ctrl-C". [akpm@linux-foundation.org: cleanup] Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Li Zefan <lizf@cn.fujitsu.com> Cc: Balbir Singh <balbir@in.ibm.com> Cc: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
KAMEZAWA Hiroyuki · Linus Torvalds
1 parent c8dad2bb63
Showing 2 changed files with 57 additions and 11 deletions Side-by-side Diff
Documentation/controllers/memory.txt
mm/memcontrol.c
@@ -237,11 +237,30 @@
 A cgroup can be removed by rmdir, but as discussed in sections 4.1 and 4.2, a
 cgroup might have some charge associated with it, even though all
 tasks have migrated away from it.
-Such charges are moved to its parent as much as possible and freed if parent
-is full. Both of RSS and CACHES are moved to parent.
-If both of them are busy, rmdir() returns -EBUSY.
+Such charges are freed(at default) or moved to its parent. When moved,
+both of RSS and CACHES are moved to parent.
+If both of them are busy, rmdir() returns -EBUSY. See 5.1 Also.
  
-5. TODO
+5. Misc. interfaces.
+
+5.1 force_empty
+  memory.force_empty interface is provided to make cgroup's memory usage empty.
+  You can use this interface only when the cgroup has no tasks.
+  When writing anything to this
+
+  # echo 0 > memory.force_empty
+
+  Almost all pages tracked by this memcg will be unmapped and freed. Some of
+  pages cannot be freed because it's locked or in-use. Such pages are moved
+  to parent and this cgroup will be empty. But this may return -EBUSY in
+  some too busy case.
+
+  Typical use case of this interface is that calling this before rmdir().
+  Because rmdir() moves all pages to parent, some out-of-use page caches can be
+  moved to the parent. If you want to avoid that, force_empty will be useful.
+
+
+6. TODO
  
 1. Add support for accounting huge pages (as a separate controller)
 2. Make per-cgroup scanner reclaim not-shared pages first
@@ -1062,21 +1062,27 @@
  * make mem_cgroup's charge to be 0 if there is no task.
  * This enables deleting this mem_cgroup.
  */
-static int mem_cgroup_force_empty(struct mem_cgroup *mem)
+static int mem_cgroup_force_empty(struct mem_cgroup *mem, bool free_all)
 {
 	int ret;
 	int node, zid, shrink;
 	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
+	struct cgroup *cgrp = mem->css.cgroup;
  
 	css_get(&mem->css);
  
 	shrink = 0;
+	/* should free all ? */
+	if (free_all)
+		goto try_to_free;
 move_account:
 	while (mem->res.usage > 0) {
 		ret = -EBUSY;
-		if (atomic_read(&mem->css.cgroup->count) > 0)
+		if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
 			goto out;
-
+		ret = -EINTR;
+		if (signal_pending(current))
+			goto out;
 		/* This is for making all *used* pages to be on LRU. */
 		lru_add_drain_all();
 		ret = 0;
  
  
  
  
@@ -1106,19 +1112,29 @@
 	return ret;
  
 try_to_free:
-	/* returns EBUSY if we come here twice. */
-	if (shrink)  {
+	/* returns EBUSY if there is a task or if we come here twice. */
+	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
 		ret = -EBUSY;
 		goto out;
 	}
+	/* we call try-to-free pages for make this cgroup empty */
+	lru_add_drain_all();
 	/* try to free all pages in this cgroup */
 	shrink = 1;
 	while (nr_retries && mem->res.usage > 0) {
 		int progress;
+
+		if (signal_pending(current)) {
+			ret = -EINTR;
+			goto out;
+		}
 		progress = try_to_free_mem_cgroup_pages(mem,
 						  GFP_HIGHUSER_MOVABLE);
-		if (!progress)
+		if (!progress) {
 			nr_retries--;
+			/* maybe some writeback is necessary */
+			congestion_wait(WRITE, HZ/10);
+		}
  
 	}
 	/* try move_account...there may be some *locked* pages. */
@@ -1128,6 +1144,12 @@
 	goto out;
 }
  
+int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
+{
+	return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
+}
+
+
 static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
 {
 	return res_counter_read_u64(&mem_cgroup_from_cont(cont)->res,
@@ -1225,6 +1247,7 @@
 	return 0;
 }
  
+
 static struct cftype mem_cgroup_files[] = {
 	{
 		.name = "usage_in_bytes",
@@ -1253,6 +1276,10 @@
 		.name = "stat",
 		.read_map = mem_control_stat_show,
 	},
+	{
+		.name = "force_empty",
+		.trigger = mem_cgroup_force_empty_write,
+	},
 };
  
 static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
@@ -1350,7 +1377,7 @@
 					struct cgroup *cont)
 {
 	struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
-	mem_cgroup_force_empty(mem);
+	mem_cgroup_force_empty(mem, false);
 }
  
 static void mem_cgroup_destroy(struct cgroup_subsys *ss,
...	...	@@ -237,11 +237,30 @@
237	237	A cgroup can be removed by rmdir, but as discussed in sections 4.1 and 4.2, a
238	238	cgroup might have some charge associated with it, even though all
239	239	tasks have migrated away from it.
240		-Such charges are moved to its parent as much as possible and freed if parent
241		-is full. Both of RSS and CACHES are moved to parent.
242		-If both of them are busy, rmdir() returns -EBUSY.
	240	+Such charges are freed(at default) or moved to its parent. When moved,
	241	+both of RSS and CACHES are moved to parent.
	242	+If both of them are busy, rmdir() returns -EBUSY. See 5.1 Also.
243	243
244		-5. TODO
	244	+5. Misc. interfaces.
	245	+
	246	+5.1 force_empty
	247	+ memory.force_empty interface is provided to make cgroup's memory usage empty.
	248	+ You can use this interface only when the cgroup has no tasks.
	249	+ When writing anything to this
	250	+
	251	+ # echo 0 > memory.force_empty
	252	+
	253	+ Almost all pages tracked by this memcg will be unmapped and freed. Some of
	254	+ pages cannot be freed because it's locked or in-use. Such pages are moved
	255	+ to parent and this cgroup will be empty. But this may return -EBUSY in
	256	+ some too busy case.
	257	+
	258	+ Typical use case of this interface is that calling this before rmdir().
	259	+ Because rmdir() moves all pages to parent, some out-of-use page caches can be
	260	+ moved to the parent. If you want to avoid that, force_empty will be useful.
	261	+
	262	+
	263	+6. TODO
245	264
246	265	1. Add support for accounting huge pages (as a separate controller)
247	266	2. Make per-cgroup scanner reclaim not-shared pages first
...	...	@@ -1062,21 +1062,27 @@
1062	1062	* make mem_cgroup's charge to be 0 if there is no task.
1063	1063	* This enables deleting this mem_cgroup.
1064	1064	*/
1065		-static int mem_cgroup_force_empty(struct mem_cgroup *mem)
	1065	+static int mem_cgroup_force_empty(struct mem_cgroup *mem, bool free_all)
1066	1066	{
1067	1067	int ret;
1068	1068	int node, zid, shrink;
1069	1069	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
	1070	+ struct cgroup *cgrp = mem->css.cgroup;
1070	1071
1071	1072	css_get(&mem->css);
1072	1073
1073	1074	shrink = 0;
	1075	+ /* should free all ? */
	1076	+ if (free_all)
	1077	+ goto try_to_free;
1074	1078	move_account:
1075	1079	while (mem->res.usage > 0) {
1076	1080	ret = -EBUSY;
1077		- if (atomic_read(&mem->css.cgroup->count) > 0)
	1081	+ if (cgroup_task_count(cgrp) \|\| !list_empty(&cgrp->children))
1078	1082	goto out;
1079		-
	1083	+ ret = -EINTR;
	1084	+ if (signal_pending(current))
	1085	+ goto out;
1080	1086	/* This is for making all used pages to be on LRU. */
1081	1087	lru_add_drain_all();
1082	1088	ret = 0;
1083	1089
1084	1090
1085	1091
1086	1092
...	...	@@ -1106,19 +1112,29 @@
1106	1112	return ret;
1107	1113
1108	1114	try_to_free:
1109		- /* returns EBUSY if we come here twice. */
1110		- if (shrink) {
	1115	+ /* returns EBUSY if there is a task or if we come here twice. */
	1116	+ if (cgroup_task_count(cgrp) \|\| !list_empty(&cgrp->children) \|\| shrink) {
1111	1117	ret = -EBUSY;
1112	1118	goto out;
1113	1119	}
	1120	+ /* we call try-to-free pages for make this cgroup empty */
	1121	+ lru_add_drain_all();
1114	1122	/* try to free all pages in this cgroup */
1115	1123	shrink = 1;
1116	1124	while (nr_retries && mem->res.usage > 0) {
1117	1125	int progress;
	1126	+
	1127	+ if (signal_pending(current)) {
	1128	+ ret = -EINTR;
	1129	+ goto out;
	1130	+ }
1118	1131	progress = try_to_free_mem_cgroup_pages(mem,
1119	1132	GFP_HIGHUSER_MOVABLE);
1120		- if (!progress)
	1133	+ if (!progress) {
1121	1134	nr_retries--;
	1135	+ /* maybe some writeback is necessary */
	1136	+ congestion_wait(WRITE, HZ/10);
	1137	+ }
1122	1138
1123	1139	}
1124	1140	/* try move_account...there may be some locked pages. */
...	...	@@ -1128,6 +1144,12 @@
1128	1144	goto out;
1129	1145	}
1130	1146
	1147	+int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
	1148	+{
	1149	+ return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
	1150	+}
	1151	+
	1152	+
1131	1153	static u64 mem_cgroup_read(struct cgroup cont, struct cftype cft)
1132	1154	{
1133	1155	return res_counter_read_u64(&mem_cgroup_from_cont(cont)->res,
...	...	@@ -1225,6 +1247,7 @@
1225	1247	return 0;
1226	1248	}
1227	1249
	1250	+
1228	1251	static struct cftype mem_cgroup_files[] = {
1229	1252	{
1230	1253	.name = "usage_in_bytes",
...	...	@@ -1253,6 +1276,10 @@
1253	1276	.name = "stat",
1254	1277	.read_map = mem_control_stat_show,
1255	1278	},
	1279	+ {
	1280	+ .name = "force_empty",
	1281	+ .trigger = mem_cgroup_force_empty_write,
	1282	+ },
1256	1283	};
1257	1284
1258	1285	static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
...	...	@@ -1350,7 +1377,7 @@
1350	1377	struct cgroup *cont)
1351	1378	{
1352	1379	struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
1353		- mem_cgroup_force_empty(mem);
	1380	+ mem_cgroup_force_empty(mem, false);
1354	1381	}
1355	1382
1356	1383	static void mem_cgroup_destroy(struct cgroup_subsys *ss,