[PATCH-next v6 7/6] cgroup/cpuset: Set old_mems_allowed from guarantee_online_mems() consistently

[Waiman Long <longman@redhat.com>]

An earlier patch has added an optimization in guarantee_online_mems()
to just return effective_mems for v2. However there is a short window
during memory hotunplug operation that it can return a nodemask with
no online node leading to possible memory OOM. To avoid this scenario,
though highly unlikely, the optimization is dropped.

Also set old_mems_allowed of the cpuset structure consistently with
the output of guarantee_online_mems() whenever an attach or a related
operation is in progress.

Signed-off-by: Waiman Long <longman@redhat.com>
---
 kernel/cgroup/cpuset.c | 28 ++++++++++++++--------------
 1 file changed, 14 insertions(+), 14 deletions(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index d624cd0a1e04..5dabe9d040e9 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -502,10 +502,6 @@ static void guarantee_active_cpus(struct task_struct *tsk,
  */
 static void guarantee_online_mems(struct cpuset *cs, nodemask_t *pmask)
 {
-	if (cpuset_v2()) {
-		*pmask = cs->effective_mems;
-		return;
-	}
 	while (!nodes_and(*pmask, cs->effective_mems, node_states[N_MEMORY]))
 		cs = parent_cs(cs);
 }
@@ -3350,22 +3346,24 @@ static void cpuset_attach(struct cgroup_taskset *tset)
 	}
 
 	cs = css_cs(css);
+	guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
+
 	/*
 	 * In the default hierarchy, enabling cpuset in the child cgroups
 	 * will trigger a cpuset_attach() call with no change in effective cpus
 	 * and mems. In that case, we can optimize out by skipping the task
 	 * iteration and update, but the destination cpuset list is iterated to
-	 * set old_mems_sllowed.
+	 * set old_mems_allowed.
 	 */
-	if (cpuset_v2()) {
-		cpuset_attach_nodemask_to = cs->effective_mems;
-		if (!attach_cpus_updated && !attach_mems_updated) {
-			llist_for_each_entry(cs, dst_cs_head.first, attach_node)
-				cs->old_mems_allowed = cs->effective_mems;
-			goto out;
+	if (cpuset_v2() && !attach_cpus_updated && !attach_mems_updated) {
+		struct cpuset *tcs;
+
+		llist_for_each_entry(tcs, dst_cs_head.first, attach_node) {
+			if (tcs == cs)
+				continue;
+			guarantee_online_mems(tcs, &tcs->old_mems_allowed);
 		}
-	} else {
-		guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
+		goto out;
 	}
 
 	cgroup_taskset_for_each(task, css, tset) {
@@ -3381,8 +3379,8 @@ static void cpuset_attach(struct cgroup_taskset *tset)
 
 	if (queue_task_work)
 		schedule_flush_migrate_mm();
-	cs->old_mems_allowed = cpuset_attach_nodemask_to;
 out:
+	cs->old_mems_allowed = cpuset_attach_nodemask_to;
 	reset_attach_in_progress();
 	clear_attach_data(false);
 	mutex_unlock(&cpuset_mutex);
@@ -3824,6 +3822,8 @@ static void cpuset_fork(struct task_struct *task)
 	/* CLONE_INTO_CGROUP */
 	mutex_lock(&cpuset_mutex);
 	guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
+	cs->old_mems_allowed = cpuset_attach_nodemask_to;
+
 	/*
 	 * Assume CPUs and memory nodes are updated
 	 * A CLONE_INTO_CGROUP operation should have taken the cgroup mutex
-- 
2.54.0