[PATCH v2 10/10] cpuset: fix to migrate mm correctly in a corner case

Li Zefan lizefan at huawei.com
Wed Jun 5 09:17:30 UTC 2013


Before moving tasks out of empty cpusets, update_tasks_nodemask()
is called, which calls do_migrate_pages(xx, from, to). Then those
tasks are moved to an ancestor, and do_migrate_pages() is called
again.

The first time: from = node_to_be_offlined, to = empty.
The second time: from = empty, to = ancestor's nodemask.

so looks like no pages will be migrated.

Fix this by:

- Don't call update_tasks_nodemask() on empty cpusets.
- Pass cs->old_mems_allowed to do_migrate_pages().

Signed-off-by: Li Zefan <lizefan at huawei.com>
---
 kernel/cpuset.c | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/kernel/cpuset.c b/kernel/cpuset.c
index 9bb6a47..de7f6c1 100644
--- a/kernel/cpuset.c
+++ b/kernel/cpuset.c
@@ -1574,9 +1574,16 @@ static void cpuset_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 		struct cpuset *mems_oldcs = effective_nodemask_cpuset(oldcs);
 
 		mpol_rebind_mm(mm, &cpuset_attach_nodemask_to);
-		if (is_memory_migrate(cs))
-			cpuset_migrate_mm(mm, &mems_oldcs->mems_allowed,
+		if (is_memory_migrate(cs)) {
+			/*
+			 * old_mems_allowed is the same with mems_allowed,
+			 * except if this task is being moved automatically
+			 * due to hotplug, and in this case mems_allowed is
+			 * empty and old_mems_allowed is the offflined node.
+			 */
+			cpuset_migrate_mm(mm, &mems_oldcs->old_mems_allowed,
 					  &cpuset_attach_nodemask_to);
+		}
 		mmput(mm);
 	}
 
@@ -2168,7 +2175,7 @@ static void cpuset_propagate_hotplug_workfn(struct work_struct *work)
 	 * for empty cpuset to take on ancestor's cpumask
 	 */
 	if ((sane && cpumask_empty(cs->cpus_allowed)) ||
-	    !cpumask_empty(&off_cpus))
+	    (!cpumask_empty(&off_cpus) && !cpumask_empty(cs->cpus_allowed)))
 		update_tasks_cpumask(cs, NULL);
 
 	mutex_lock(&callback_mutex);
@@ -2180,7 +2187,7 @@ static void cpuset_propagate_hotplug_workfn(struct work_struct *work)
 	 * for empty cpuset to take on ancestor's nodemask
 	 */
 	if ((sane && nodes_empty(cs->mems_allowed)) ||
-	    !nodes_empty(off_mems))
+	    (!nodes_empty(off_mems) && !nodes_empty(cs->mems_allowed)))
 		update_tasks_nodemask(cs, NULL);
 
 	is_empty = cpumask_empty(cs->cpus_allowed) ||
-- 
1.8.0.2


More information about the Containers mailing list