workqueue: Better document teardown for delayed_work

destroy_workqueue() does not ensure that non-pending work submitted with
queue_delayed_work() gets cancelled. The caller has to ensure that
manually.

Add this information about delayed_work in destroy_workqueue()'s
docstring.

Add a TODO for destroy_workqueue() to wait for all delayed_work.

Signed-off-by: Philipp Stanner <phasta@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index cf620328273757e4e9a06f8f29ca7ead436db47b..2cb8276a27a99c951883d09ed2dbd5f488bda8e2 100644 (file)
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -5837,6 +5837,17 @@ static bool pwq_busy(struct pool_workqueue *pwq)
  * @wq: target workqueue
  *
  * Safely destroy a workqueue. All work currently pending will be done first.
+ *
+ * This function does NOT guarantee that non-pending work that has been
+ * submitted with queue_delayed_work() and similar functions will be done
+ * before destroying the workqueue. The fundamental problem is that, currently,
+ * the workqueue has no way of accessing non-pending delayed_work. delayed_work
+ * is only linked on the timer-side. All delayed_work must, therefore, be
+ * canceled before calling this function.
+ *
+ * TODO: It would be better if the problem described above wouldn't exist and
+ * destroy_workqueue() would cleanly cancel all pending and non-pending
+ * delayed_work.
  */
 void destroy_workqueue(struct workqueue_struct *wq)
 {