io_uring: small optimisation of tctx_task_work

There should be no completions stashed when we first get into
tctx_task_work(), so move completion flushing checks a bit later
after we had a chance to execute some task works.
Signed-off-by: NPavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/c6765c804f3c438591b9825ab9c43d22039073c4.1647897811.git.asml.silence@gmail.comSigned-off-by: NJens Axboe <axboe@kernel.dk>

fs/io_uring.c

@@ -2475,10 +2475,6 @@ static void tctx_task_work(struct callback_head *cb)
 	while (1) {
 		struct io_wq_work_node *node1, *node2;
 
-		if (!tctx->task_list.first &&
-		    !tctx->prior_task_list.first && uring_locked)
-			io_submit_flush_completions(ctx);
-
 		spin_lock_irq(&tctx->task_lock);
 		node1 = tctx->prior_task_list.first;
 		node2 = tctx->task_list.first;
@@ -2492,10 +2488,13 @@ static void tctx_task_work(struct callback_head *cb)
 
 		if (node1)
 			handle_prev_tw_list(node1, &ctx, &uring_locked);
-
 		if (node2)
 			handle_tw_list(node2, &ctx, &uring_locked);
 		cond_resched();
+
+		if (!tctx->task_list.first &&
+		    !tctx->prior_task_list.first && uring_locked)
+			io_submit_flush_completions(ctx);
 	}
 
 	ctx_flush_and_put(ctx, &uring_locked);