Merge tag 'wq-for-6.16' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq

Pull workqueue updates from Tejun Heo:
 "Fix statistic update race condition and a couple documentation
  updates"

* tag 'wq-for-6.16' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
  workqueue: fix typo in comment
  workqueue: Fix race condition in wq->stats incrementation
  workqueue: Better document teardown for delayed_work

Author: torvalds

include/linux/workqueue.h

@@ -480,7 +480,7 @@ void workqueue_softirq_dead(unsigned int cpu);
  * executing at most one work item for the workqueue.
  *
  * For unbound workqueues, @max_active limits the number of in-flight work items
- * for the whole system. e.g. @max_active of 16 indicates that that there can be
+ * for the whole system. e.g. @max_active of 16 indicates that there can be
  * at most 16 work items executing for the workqueue in the whole system.
  *
  * As sharing the same active counter for an unbound workqueue across multiple

kernel/workqueue.c

@@ -3241,7 +3241,7 @@ __acquires(&pool->lock)
 	 * point will only record its address.
 	 */
 	trace_workqueue_execute_end(work, worker->current_func);
-	pwq->stats[PWQ_STAT_COMPLETED]++;
+
 	lock_map_release(&lockdep_map);
 	if (!bh_draining)
 		lock_map_release(pwq->wq->lockdep_map);
@@ -3272,6 +3272,8 @@ __acquires(&pool->lock)
 
 	raw_spin_lock_irq(&pool->lock);
 
+	pwq->stats[PWQ_STAT_COMPLETED]++;
+
 	/*
 	 * In addition to %WQ_CPU_INTENSIVE, @worker may also have been marked
 	 * CPU intensive by wq_worker_tick() if @work hogged CPU longer than
@@ -5837,6 +5839,17 @@ 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)
 {