diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c
index bba806ce57d35dcd9f4f746d534a24a6cf32a873..6e36b88ca7c9c296fa5306e3376835f4259a7760 100644
--- a/drivers/net/ethernet/realtek/r8169.c
+++ b/drivers/net/ethernet/realtek/r8169.c
@@ -6074,7 +6074,6 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 	struct device *d = tp_to_dev(tp);
 	dma_addr_t mapping;
 	u32 opts[2], len;
-	bool stop_queue;
 	int frags;
 
 	if (unlikely(!rtl_tx_slots_avail(tp, skb_shinfo(skb)->nr_frags))) {
@@ -6116,6 +6115,8 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	txd->opts2 = cpu_to_le32(opts[1]);
 
+	netdev_sent_queue(dev, skb->len);
+
 	skb_tx_timestamp(skb);
 
 	/* Force memory writes to complete before releasing descriptor */
@@ -6128,16 +6129,16 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	tp->cur_tx += frags + 1;
 
-	stop_queue = !rtl_tx_slots_avail(tp, MAX_SKB_FRAGS);
-	if (unlikely(stop_queue))
-		netif_stop_queue(dev);
+	RTL_W8(tp, TxPoll, NPQ);
 
-	if (__netdev_sent_queue(dev, skb->len, skb->xmit_more)) {
-		RTL_W8(tp, TxPoll, NPQ);
-		mmiowb();
-	}
+	mmiowb();
 
-	if (unlikely(stop_queue)) {
+	if (!rtl_tx_slots_avail(tp, MAX_SKB_FRAGS)) {
+		/* Avoid wrongly optimistic queue wake-up: rtl_tx thread must
+		 * not miss a ring update when it notices a stopped queue.
+		 */
+		smp_wmb();
+		netif_stop_queue(dev);
 		/* Sync with rtl_tx:
 		 * - publish queue status and cur_tx ring index (write barrier)
 		 * - refresh dirty_tx ring index (read barrier).