cgroup: make sure a parent css isn't freed before its children
authorTejun Heo <tj@kernel.org>
Thu, 21 Jan 2016 20:32:15 +0000 (15:32 -0500)
committerSasha Levin <sasha.levin@oracle.com>
Mon, 11 Jul 2016 03:07:11 +0000 (23:07 -0400)
[ Upstream commit 8bb5ef79bc0f4016ecf79e8dce6096a3c63603e4 ]

There are three subsystem callbacks in css shutdown path -
css_offline(), css_released() and css_free().  Except for
css_released(), cgroup core didn't guarantee the order of invocation.
css_offline() or css_free() could be called on a parent css before its
children.  This behavior is unexpected and led to bugs in cpu and
memory controller.

The previous patch updated ordering for css_offline() which fixes the
cpu controller issue.  While there currently isn't a known bug caused
by misordering of css_free() invocations, let's fix it too for
consistency.

css_free() ordering can be trivially fixed by moving putting of the
parent css below css_free() invocation.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
kernel/cgroup.c

index d01c604..3abce1e 100644 (file)
@@ -4378,14 +4378,15 @@ static void css_free_work_fn(struct work_struct *work)
 
        if (ss) {
                /* css free path */
+               struct cgroup_subsys_state *parent = css->parent;
                int id = css->id;
 
-               if (css->parent)
-                       css_put(css->parent);
-
                ss->css_free(css);
                cgroup_idr_remove(&ss->css_idr, id);
                cgroup_put(cgrp);
+
+               if (parent)
+                       css_put(parent);
        } else {
                /* cgroup free path */
                atomic_dec(&cgrp->root->nr_cgrps);