net: hns3: fix for tunnel type handling in hns3_rx_checksum
authorYunsheng Lin <linyunsheng@huawei.com>
Mon, 6 May 2019 02:48:45 +0000 (10:48 +0800)
committerDavid S. Miller <davem@davemloft.net>
Tue, 7 May 2019 17:37:13 +0000 (10:37 -0700)
According to hardware user manual, the tunnel packet type is
available in the rx.ol_info field of struct hns3_desc. Currently
the tunnel packet type is decided by the rx.l234_info, which may
cause RX checksum handling error.

This patch fixes it by using the correct field in struct hns3_desc
to decide the tunnel packet type.

Fixes: 76ad4f0ee747 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/hisilicon/hns3/hns3_enet.c

index 944e0aecb816fcbeec17bcfe7fb5bc6e4759646b..14312ff01233831ff58b3204445315f5ba423302 100644 (file)
@@ -2463,7 +2463,7 @@ static int hns3_gro_complete(struct sk_buff *skb)
 }
 
 static void hns3_rx_checksum(struct hns3_enet_ring *ring, struct sk_buff *skb,
-                            u32 l234info, u32 bd_base_info)
+                            u32 l234info, u32 bd_base_info, u32 ol_info)
 {
        struct net_device *netdev = ring->tqp->handle->kinfo.netdev;
        int l3_type, l4_type;
@@ -2490,7 +2490,7 @@ static void hns3_rx_checksum(struct hns3_enet_ring *ring, struct sk_buff *skb,
                return;
        }
 
-       ol4_type = hnae3_get_field(l234info, HNS3_RXD_OL4ID_M,
+       ol4_type = hnae3_get_field(ol_info, HNS3_RXD_OL4ID_M,
                                   HNS3_RXD_OL4ID_S);
        switch (ol4_type) {
        case HNS3_OL4_TYPE_MAC_IN_UDP:
@@ -2694,7 +2694,7 @@ static int hns3_add_frag(struct hns3_enet_ring *ring, struct hns3_desc *desc,
 
 static int hns3_set_gro_and_checksum(struct hns3_enet_ring *ring,
                                     struct sk_buff *skb, u32 l234info,
-                                    u32 bd_base_info)
+                                    u32 bd_base_info, u32 ol_info)
 {
        u16 gro_count;
        u32 l3_type;
@@ -2703,7 +2703,7 @@ static int hns3_set_gro_and_checksum(struct hns3_enet_ring *ring,
                                    HNS3_RXD_GRO_COUNT_S);
        /* if there is no HW GRO, do not set gro params */
        if (!gro_count) {
-               hns3_rx_checksum(ring, skb, l234info, bd_base_info);
+               hns3_rx_checksum(ring, skb, l234info, bd_base_info, ol_info);
                return 0;
        }
 
@@ -2743,7 +2743,7 @@ static int hns3_handle_bdinfo(struct hns3_enet_ring *ring, struct sk_buff *skb)
 {
        struct net_device *netdev = ring->tqp->handle->kinfo.netdev;
        enum hns3_pkt_l2t_type l2_frame_type;
-       u32 bd_base_info, l234info;
+       u32 bd_base_info, l234info, ol_info;
        struct hns3_desc *desc;
        unsigned int len;
        int pre_ntc, ret;
@@ -2757,6 +2757,7 @@ static int hns3_handle_bdinfo(struct hns3_enet_ring *ring, struct sk_buff *skb)
        desc = &ring->desc[pre_ntc];
        bd_base_info = le32_to_cpu(desc->rx.bd_base_info);
        l234info = le32_to_cpu(desc->rx.l234_info);
+       ol_info = le32_to_cpu(desc->rx.ol_info);
 
        /* Based on hw strategy, the tag offloaded will be stored at
         * ot_vlan_tag in two layer tag case, and stored at vlan_tag
@@ -2796,7 +2797,8 @@ static int hns3_handle_bdinfo(struct hns3_enet_ring *ring, struct sk_buff *skb)
        skb->protocol = eth_type_trans(skb, netdev);
 
        /* This is needed in order to enable forwarding support */
-       ret = hns3_set_gro_and_checksum(ring, skb, l234info, bd_base_info);
+       ret = hns3_set_gro_and_checksum(ring, skb, l234info,
+                                       bd_base_info, ol_info);
        if (unlikely(ret)) {
                u64_stats_update_begin(&ring->syncp);
                ring->stats.rx_err_cnt++;