Commit 07cd1294 authored Jan 26, 2017 by Tejun Heo

cgroup: don't online subsystems before cgroup_name/path() are operational



While refactoring cgroup creation, a5bca215 ("cgroup: factor out
cgroup_create() out of cgroup_mkdir()") incorrectly onlined subsystems
before the new cgroup is associated with it kernfs_node.  This is fine
for cgroup proper but cgroup_name/path() depend on the associated
kernfs_node and if a subsystem makes the new cgroup_subsys_state
visible, which they're allowed to after onlining, it can lead to NULL
dereference.

The current code performs cgroup creation and subsystem onlining in
cgroup_create() and cgroup_mkdir() makes the cgroup and subsystems
visible afterwards.  There's no reason to online the subsystems early
and we can simply drop cgroup_apply_control_enable() call from
cgroup_create() so that the subsystems are onlined and made visible at
the same time.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Fixes: a5bca215 ("cgroup: factor out cgroup_create() out of cgroup_mkdir()") 
Cc: stable@vger.kernel.org # v4.6+

parent 7ce7d89f

kernel/cgroup.c

+5 −8

Original line number	Diff line number	Diff line
		@@ -5221,6 +5221,11 @@ err_free_css:
		return ERR_PTR(err);
		}

		/*
		* The returned cgroup is fully initialized including its control mask, but
		* it isn't associated with its kernfs_node and doesn't have the control
		* mask applied.
		*/
		static struct cgroup cgroup_create(struct cgroup parent)
		{
		struct cgroup_root *root = parent->root;
		@@ -5288,11 +5293,6 @@ static struct cgroup cgroup_create(struct cgroup parent)

		cgroup_propagate_control(cgrp);

		/* @cgrp doesn't have dir yet so the following will only create csses */
		ret = cgroup_apply_control_enable(cgrp);
		if (ret)
		goto out_destroy;

		return cgrp;

		out_cancel_ref:
		@@ -5300,9 +5300,6 @@ out_cancel_ref:
		out_free_cgrp:
		kfree(cgrp);
		return ERR_PTR(ret);
		out_destroy:
		cgroup_destroy_locked(cgrp);
		return ERR_PTR(ret);
		}

		static int cgroup_mkdir(struct kernfs_node parent_kn, const char name,

Admin message