cgroup: use ->subtree_control when testing no internal process rule
authorTejun Heo <tj@kernel.org>
Tue, 23 Feb 2016 15:00:51 +0000 (10:00 -0500)
committerTejun Heo <tj@kernel.org>
Tue, 23 Feb 2016 15:00:51 +0000 (10:00 -0500)
No internal process rule is enforced by cgroup_migrate_prepare_dst()
during process migration.  It tests whether the target cgroup's
->child_subsys_mask is zero which is different from "subtree_control"
write path which tests ->subtree_control.  This hasn't mattered
because up until now, both ->child_subsys_mask and ->subtree_control
are zero or non-zero at the same time.  However, with the planned
addition of implicit controllers, this will no longer be true.

This patch prepares for the change by making
cgorup_migrate_prepare_dst() test ->subtree_control instead.

Signed-off-by: Tejun Heo <tj@kernel.org>
kernel/cgroup.c

index 2b11436..ac5451e 100644 (file)
@@ -2527,11 +2527,11 @@ static int cgroup_migrate_prepare_dst(struct cgroup *dst_cgrp,
        lockdep_assert_held(&cgroup_mutex);
 
        /*
-        * Except for the root, subtree_ss_mask must be zero for a cgroup
+        * Except for the root, subtree_control must be zero for a cgroup
         * with tasks so that child cgroups don't compete against tasks.
         */
        if (dst_cgrp && cgroup_on_dfl(dst_cgrp) && cgroup_parent(dst_cgrp) &&
-           dst_cgrp->subtree_ss_mask)
+           dst_cgrp->subtree_control)
                return -EBUSY;
 
        /* look up the dst cset for each src cset and link it to src */