workqueue: Add pool_workqueue to pending_pwqs list when unplugging multiple inactive works

In unplug_oldest_pwq(), the first inactive work item on the
pool_workqueue is activated correctly. However, if multiple inactive
works exist on the same pool_workqueue, subsequent works fail to
activate because wq_node_nr_active.pending_pwqs is empty — the list
insertion is skipped when the pool_workqueue is plugged.

Fix this by checking for additional inactive works in
unplug_oldest_pwq() and updating wq_node_nr_active.pending_pwqs
accordingly.

Fixes: 4c065db ("workqueue: Enable unbound cpumask update on ordered workqueues")
Cc: [email protected]
Cc: Carlos Santa <[email protected]>
Cc: Ryan Neph <[email protected]>
Cc: Lai Jiangshan <[email protected]>
Cc: Waiman Long <[email protected]>
Cc: [email protected]
Signed-off-by: Matthew Brost <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
Acked-by: Waiman Long <[email protected]>

Author: mbrost05

kernel/workqueue.c

@@ -1849,8 +1849,20 @@ static void unplug_oldest_pwq(struct workqueue_struct *wq)
 	raw_spin_lock_irq(&pwq->pool->lock);
 	if (pwq->plugged) {
 		pwq->plugged = false;
-		if (pwq_activate_first_inactive(pwq, true))
+		if (pwq_activate_first_inactive(pwq, true)) {
+			/*
+			 * While plugged, queueing skips activation which
+			 * includes bumping the nr_active count and adding the
+			 * pwq to nna->pending_pwqs if the count can't be
+			 * obtained. We need to restore both for the pwq being
+			 * unplugged. The first call activates the first
+			 * inactive work item and the second, if there are more
+			 * inactive, puts the pwq on pending_pwqs.
+			 */
+			pwq_activate_first_inactive(pwq, false);
+
 			kick_pool(pwq->pool);
+		}
 	}
 	raw_spin_unlock_irq(&pwq->pool->lock);
 }