workqueue: fix a possible race condition between rescuer and pwq-release
authorLai Jiangshan <laijs@cn.fujitsu.com>
Fri, 18 Apr 2014 15:04:16 +0000 (11:04 -0400)
committerJiri Slaby <jslaby@suse.cz>
Mon, 9 Jun 2014 13:53:36 +0000 (15:53 +0200)
commitd445003e2c54b48d280f255a40cbad97f0c26019
treeb4c2c6cd7b1055c7ca6a360d5886cc513638ac4c
parent3b3b224978f8426000d0c26ecdde037356f3d873
workqueue: fix a possible race condition between rescuer and pwq-release

commit 77668c8b559e4fe2acf2a0749c7c83cde49a5025 upstream.

There is a race condition between rescuer_thread() and
pwq_unbound_release_workfn().

Even after a pwq is scheduled for rescue, the associated work items
may be consumed by any worker.  If all of them are consumed before the
rescuer gets to them and the pwq's base ref was put due to attribute
change, the pwq may be released while still being linked on
@wq->maydays list making the rescuer dereference already freed pwq
later.

Make send_mayday() pin the target pwq until the rescuer is done with
it.

tj: Updated comment and patch description.

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
kernel/workqueue.c