Introduce a per-memory-cgroup oom_priority setting: an integer number
within the [-10000, 10000] range, which defines the order in which
the OOM killer selects victim memory cgroups.
OOM killer prefers memory cgroups with larger priority if they are
populated with eligible tasks.
The oom_priority value is compared within sibling cgroups.
The root cgroup has the oom_priority 0, which cannot be changed.
Signed-off-by: Roman Gushchin <g...@fb.com>
Cc: Michal Hocko <mho...@kernel.org>
Cc: Vladimir Davydov <vdavydov....@gmail.com>
Cc: Johannes Weiner <han...@cmpxchg.org>
Cc: David Rientjes <rient...@google.com>
Cc: Tejun Heo <t...@kernel.org>
Cc: Tetsuo Handa <penguin-ker...@i-love.sakura.ne.jp>
Cc: kernel-t...@fb.com
Cc: cgro...@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Cc: linux-ker...@vger.kernel.org
Cc: linux...@kvack.org
---
include/linux/memcontrol.h | 3 +++
mm/memcontrol.c | 52 ++++++++++++++++++++++++++++++++++++++++++++--
2 files changed, 53 insertions(+), 2 deletions(-)
diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index c57ee47c35bb..915f0c19a2b5 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -206,6 +206,9 @@ struct mem_cgroup {
/* cached OOM score */
long oom_score;
+ /* OOM killer priority */
+ short oom_priority;
+
/* handle for "memory.events" */
struct cgroup_file events_file;
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index a620aaae6201..a173e5b0d4d8 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -2749,6 +2749,7 @@ static void select_victim_memcg(struct mem_cgroup *root,
struct oom_control *oc)
for (;;) {
struct cgroup_subsys_state *css;
struct mem_cgroup *memcg = NULL;
+ short prio = SHRT_MIN;
long score = LONG_MIN;
css_for_each_child(css, &root->css) {
@@ -2760,7 +2761,12 @@ static void select_victim_memcg(struct mem_cgroup *root,
struct oom_control *oc)
if (iter->oom_score == 0)
continue;
- if (iter->oom_score > score) {
+ if (iter->oom_priority > prio) {
+ memcg = iter;
+ prio = iter->oom_priority;
+ score = iter->oom_score;
+ } else if (iter->oom_priority == prio &&
+ iter->oom_score > score) {
memcg = iter;
score = iter->oom_score;
}
@@ -2830,7 +2836,15 @@ bool mem_cgroup_select_oom_victim(struct oom_control *oc)
* For system-wide OOMs we should consider tasks in the root cgroup
* with oom_score larger than oc->chosen_points.
*/
- if (!oc->memcg) {
+ if (!oc->memcg && !(oc->chosen_memcg &&
+ oc->chosen_memcg->oom_priority > 0)) {
+ /*
+ * Root memcg has priority 0, so if chosen memcg has lower
+ * priority, any task in root cgroup is preferable.
+ */
+ if (oc->chosen_memcg && oc->chosen_memcg->oom_priority < 0)
+ oc->chosen_points = 0;
+
select_victim_root_cgroup_task(oc);
if (oc->chosen_task && oc->chosen_memcg) {
@@ -5426,6 +5440,34 @@ static ssize_t memory_oom_kill_all_write(struct
kernfs_open_file *of,
return nbytes;
}
+static int memory_oom_priority_show(struct seq_file *m, void *v)
+{
+ struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
+
+ seq_printf(m, "%d\n", memcg->oom_priority);
+
+ return 0;
+}
+
+static ssize_t memory_oom_priority_write(struct kernfs_open_file *of,
+ char *buf, size_t nbytes, loff_t off)
+{
+ struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
+ int oom_priority;
+ int err;
+
+ err = kstrtoint(strstrip(buf), 0, &oom_priority);
+ if (err)
+ return err;
+
+ if (oom_priority < -10000 || oom_priority > 10000)
+ return -EINVAL;
+
+ memcg->oom_priority = (short)oom_priority;
+
+ return nbytes;
+}
+
static int memory_events_show(struct seq_file *m, void *v)
{
struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
@@ -5552,6 +5594,12 @@ static struct cftype memory_files[] = {
.write = memory_oom_kill_all_write,
},
{
+ .name = "oom_priority",
+ .flags = CFTYPE_NOT_ON_ROOT,
+ .seq_show = memory_oom_priority_show,
+ .write = memory_oom_priority_write,
+ },
+ {
.name = "events",
.flags = CFTYPE_NOT_ON_ROOT,
.file_offset = offsetof(struct mem_cgroup, events_file),
--
2.13.5
--
To unsubscribe from this list: send the line "unsubscribe linux-doc" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
