r8169: make use of xmit_more

There was a previous attempt to use xmit_more, but the change had to be
reverted because under load sometimes a transmit timeout occurred [0].
Maybe this was caused by a missing memory barrier, the new attempt
keeps the memory barrier before the call to netif_stop_queue like it
is used by the driver as of today. The new attempt also changes the
order of some calls as suggested by Eric.

[0] https://lkml.org/lkml/2019/2/10/39Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/realtek/r8169_main.c

@@ -5641,6 +5641,8 @@ 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;
+	bool door_bell;
 	int frags;
 
 	if (unlikely(!rtl_tx_slots_avail(tp, skb_shinfo(skb)->nr_frags))) {
@@ -5684,13 +5686,13 @@ 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 */
 	dma_wmb();
 
+	door_bell = __netdev_sent_queue(dev, skb->len, netdev_xmit_more());
+
 	txd->opts1 = rtl8169_get_txd_opts1(opts[0], len, entry);
 
 	/* Force all memory writes to complete before notifying device */
@@ -5698,14 +5700,19 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	tp->cur_tx += frags + 1;
 
-	RTL_W8(tp, TxPoll, NPQ);
-
-	if (!rtl_tx_slots_avail(tp, MAX_SKB_FRAGS)) {
+	stop_queue = !rtl_tx_slots_avail(tp, MAX_SKB_FRAGS);
+	if (unlikely(stop_queue)) {
 		/* 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);
+	}
+
+	if (door_bell)
+		RTL_W8(tp, TxPoll, NPQ);
+
+	if (unlikely(stop_queue)) {
 		/* Sync with rtl_tx:
 		 * - publish queue status and cur_tx ring index (write barrier)
 		 * - refresh dirty_tx ring index (read barrier).