net/mlx4_en: Don't attempt to TX offload the outer UDP checksum for VXLAN
authorOr Gerlitz <ogerlitz@mellanox.com>
Thu, 30 Oct 2014 13:59:27 +0000 (15:59 +0200)
committerDavid S. Miller <davem@davemloft.net>
Thu, 30 Oct 2014 23:48:58 +0000 (19:48 -0400)
For VXLAN/NVGRE encapsulation, the current HW doesn't support offloading
both the outer UDP TX checksum and the inner TCP/UDP TX checksum.

The driver doesn't advertize SKB_GSO_UDP_TUNNEL_CSUM, however we are wrongly
telling the HW to offload the outer UDP checksum for encapsulated packets,
fix that.

Fixes: 837052d0ccc5 ('net/mlx4_en: Add netdev support for TCP/IP
     offloads of vxlan tunneling')
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/mellanox/mlx4/en_tx.c

index 34c137878545fc672dad1a3d86e11c034c0ac368..454d9fea640ecb0aa849b8d8106695d017ca3309 100644 (file)
@@ -836,8 +836,11 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
         * whether LSO is used */
        tx_desc->ctrl.srcrb_flags = priv->ctrl_flags;
        if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
-               tx_desc->ctrl.srcrb_flags |= cpu_to_be32(MLX4_WQE_CTRL_IP_CSUM |
-                                                        MLX4_WQE_CTRL_TCP_UDP_CSUM);
+               if (!skb->encapsulation)
+                       tx_desc->ctrl.srcrb_flags |= cpu_to_be32(MLX4_WQE_CTRL_IP_CSUM |
+                                                                MLX4_WQE_CTRL_TCP_UDP_CSUM);
+               else
+                       tx_desc->ctrl.srcrb_flags |= cpu_to_be32(MLX4_WQE_CTRL_IP_CSUM);
                ring->tx_csum++;
        }