cgroupns: Close race between cgroup_post_fork and copy_cgroup_ns

In most code paths involving cgroup migration cgroup_threadgroup_rwsem
is taken.  There are two exceptions:

- remove_tasks_in_empty_cpuset calls cgroup_transfer_tasks
- vhost_attach_cgroups_work calls cgroup_attach_task_all

With cgroup_threadgroup_rwsem held it is guaranteed that cgroup_post_fork
and copy_cgroup_ns will reference the same css_set from the process calling
fork.

Without such an interlock there process after fork could reference one
css_set from it's new cgroup namespace and another css_set from
task->cgroups, which semantically is nonsensical.

Cc: stable@vger.kernel.org
Fixes: a79a908f ("cgroup: introduce cgroup namespaces")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Tejun Heo <tj@kernel.org>

kernel/cgroup.c

@@ -2962,6 +2962,7 @@ int cgroup_attach_task_all(struct task_struct *from, struct task_struct *tsk)
 	int retval = 0;
 
 	mutex_lock(&cgroup_mutex);
+	percpu_down_write(&cgroup_threadgroup_rwsem);
 	for_each_root(root) {
 		struct cgroup *from_cgrp;
 
@@ -2976,6 +2977,7 @@ int cgroup_attach_task_all(struct task_struct *from, struct task_struct *tsk)
 		if (retval)
 			break;
 	}
+	percpu_up_write(&cgroup_threadgroup_rwsem);
 	mutex_unlock(&cgroup_mutex);
 
 	return retval;
@@ -4343,6 +4345,8 @@ int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from)
 
 	mutex_lock(&cgroup_mutex);
 
+	percpu_down_write(&cgroup_threadgroup_rwsem);
+
 	/* all tasks in @from are being moved, all csets are source */
 	spin_lock_irq(&css_set_lock);
 	list_for_each_entry(link, &from->cset_links, cset_link)
@@ -4371,6 +4375,7 @@ int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from)
 	} while (task && !ret);
 out_err:
 	cgroup_migrate_finish(&preloaded_csets);
+	percpu_up_write(&cgroup_threadgroup_rwsem);
 	mutex_unlock(&cgroup_mutex);
 	return ret;
 }