net: hns3: fix tx timeout issue

Currently, the driver knocks the ring doorbell before updating
the ring->last_to_use in tx flow. if the hardware transmiting
packet and napi poll scheduling are fast enough, it may get
the old ring->last_to_use in drivers' napi poll.
In this case, the driver will think the tx is not completed, and
return directly without clear the flag __QUEUE_STATE_STACK_XOFF,
which may cause tx timeout.

Fixes: 20d06ca2 ("net: hns3: optimize the tx clean process")
Signed-off-by: NJian Shen <shenjian15@huawei.com>
Signed-off-by: NJijie Shao <shaojijie@huawei.com>
Signed-off-by: NPaolo Abeni <pabeni@redhat.com>

drivers/net/ethernet/hisilicon/hns3/hns3_enet.c

@@ -2103,8 +2103,12 @@ static void hns3_tx_doorbell(struct hns3_enet_ring *ring, int num,
 	 */
 	if (test_bit(HNS3_NIC_STATE_TX_PUSH_ENABLE, &priv->state) && num &&
 	    !ring->pending_buf && num <= HNS3_MAX_PUSH_BD_NUM && doorbell) {
+		/* This smp_store_release() pairs with smp_load_aquire() in
+		 * hns3_nic_reclaim_desc(). Ensure that the BD valid bit
+		 * is updated.
+		 */
+		smp_store_release(&ring->last_to_use, ring->next_to_use);
 		hns3_tx_push_bd(ring, num);
-		WRITE_ONCE(ring->last_to_use, ring->next_to_use);
 		return;
 	}
 
@@ -2115,6 +2119,11 @@ static void hns3_tx_doorbell(struct hns3_enet_ring *ring, int num,
 		return;
 	}
 
+	/* This smp_store_release() pairs with smp_load_aquire() in
+	 * hns3_nic_reclaim_desc(). Ensure that the BD valid bit is updated.
+	 */
+	smp_store_release(&ring->last_to_use, ring->next_to_use);
+
 	if (ring->tqp->mem_base)
 		hns3_tx_mem_doorbell(ring);
 	else
@@ -2122,7 +2131,6 @@ static void hns3_tx_doorbell(struct hns3_enet_ring *ring, int num,
 		       ring->tqp->io_base + HNS3_RING_TX_RING_TAIL_REG);
 
 	ring->pending_buf = 0;
-	WRITE_ONCE(ring->last_to_use, ring->next_to_use);
 }
 
 static void hns3_tsyn(struct net_device *netdev, struct sk_buff *skb,
@@ -3563,9 +3571,8 @@ static void hns3_reuse_buffer(struct hns3_enet_ring *ring, int i)
 static bool hns3_nic_reclaim_desc(struct hns3_enet_ring *ring,
 				  int *bytes, int *pkts, int budget)
 {
-	/* pair with ring->last_to_use update in hns3_tx_doorbell(),
-	 * smp_store_release() is not used in hns3_tx_doorbell() because
-	 * the doorbell operation already have the needed barrier operation.
+	/* This smp_load_acquire() pairs with smp_store_release() in
+	 * hns3_tx_doorbell().
 	 */
 	int ltu = smp_load_acquire(&ring->last_to_use);
 	int ntc = ring->next_to_clean;