On Tue, Jul 06, 2021 at 07:52:27AM +0200, Christoph Hellwig wrote:

On Mon, Jul 05, 2021 at 06:26:03PM +0800, Ming Lei wrote:

quoted

+static void loop_free_idle_workers(struct work_struct *work)
+{
+	struct loop_device *lo = container_of(work, struct loop_device,
+			idle_work.work);
+	struct loop_worker *pos, *worker;
+
+	spin_lock(&lo->lo_work_lock);
+	list_for_each_entry_safe(worker, pos, &lo->idle_worker_list,
+				idle_list) {
+		if (time_is_after_jiffies(worker->last_ran_at +
+						LOOP_IDLE_WORKER_TIMEOUT))
+			break;
+		list_del(&worker->idle_list);
+		rb_erase(&worker->rb_node, &lo->worker_tree);
+		css_put(worker->blkcg_css);
+		kfree(worker);
+	}
+	if (!list_empty(&lo->idle_worker_list))
+		loop_set_timer(lo);
+	spin_unlock(&lo->lo_work_lock);
+}
+
+

Double empty line.  But that whole fact that the loop driver badly
reimplements work queues is just fucked up.  We need to revert this
shit.

Can you explain why work queues are bad for fixing this blkcg/memcg
issue?


Thanks, 
Ming