workqueue: warn about flush_scheduled_work()
Alan Stern [Fri, 12 Feb 2010 08:39:21 +0000 (17:39 +0900)]
This patch (as1319) adds kerneldoc and a pointed warning to
flush_scheduled_work().

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Tejun Heo <tj@kernel.org>

kernel/workqueue.c

index 5bfb213..0225fea 100644 (file)
@@ -845,6 +845,30 @@ int schedule_on_each_cpu(work_func_t func)
        return 0;
 }
 
+/**
+ * flush_scheduled_work - ensure that any scheduled work has run to completion.
+ *
+ * Forces execution of the kernel-global workqueue and blocks until its
+ * completion.
+ *
+ * Think twice before calling this function!  It's very easy to get into
+ * trouble if you don't take great care.  Either of the following situations
+ * will lead to deadlock:
+ *
+ *     One of the work items currently on the workqueue needs to acquire
+ *     a lock held by your code or its caller.
+ *
+ *     Your code is running in the context of a work routine.
+ *
+ * They will be detected by lockdep when they occur, but the first might not
+ * occur very often.  It depends on what work items are on the workqueue and
+ * what locks they need, which you have no control over.
+ *
+ * In most situations flushing the entire workqueue is overkill; you merely
+ * need to know that a particular work item isn't queued and isn't running.
+ * In such cases you should use cancel_delayed_work_sync() or
+ * cancel_work_sync() instead.
+ */
 void flush_scheduled_work(void)
 {
        flush_workqueue(keventd_wq);