vrf: plug skb leaks
authorNikolay Aleksandrov <nikolay@cumulusnetworks.com>
Wed, 19 Aug 2015 03:12:29 +0000 (06:12 +0300)
committerDavid S. Miller <davem@davemloft.net>
Thu, 20 Aug 2015 19:59:42 +0000 (12:59 -0700)
Currently whenever a packet different from ETH_P_IP is sent through the
VRF device it is leaked so plug the leaks and properly drop these
packets.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Acked-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/vrf.c

index ed208317cbb50fe4848a4eeef53805be7fa4be99..4aa06450fafafcc3c8fb7fa128275c11ef2f3632 100644 (file)
@@ -97,6 +97,12 @@ static bool is_ip_rx_frame(struct sk_buff *skb)
        return false;
 }
 
+static void vrf_tx_error(struct net_device *vrf_dev, struct sk_buff *skb)
+{
+       vrf_dev->stats.tx_errors++;
+       kfree_skb(skb);
+}
+
 /* note: already called with rcu_read_lock */
 static rx_handler_result_t vrf_handle_frame(struct sk_buff **pskb)
 {
@@ -149,7 +155,8 @@ static struct rtnl_link_stats64 *vrf_get_stats64(struct net_device *dev,
 static netdev_tx_t vrf_process_v6_outbound(struct sk_buff *skb,
                                           struct net_device *dev)
 {
-       return 0;
+       vrf_tx_error(dev, skb);
+       return NET_XMIT_DROP;
 }
 
 static int vrf_send_v4_prep(struct sk_buff *skb, struct flowi4 *fl4,
@@ -206,8 +213,7 @@ static netdev_tx_t vrf_process_v4_outbound(struct sk_buff *skb,
 out:
        return ret;
 err:
-       vrf_dev->stats.tx_errors++;
-       kfree_skb(skb);
+       vrf_tx_error(vrf_dev, skb);
        goto out;
 }
 
@@ -219,6 +225,7 @@ static netdev_tx_t is_ip_tx_frame(struct sk_buff *skb, struct net_device *dev)
        case htons(ETH_P_IPV6):
                return vrf_process_v6_outbound(skb, dev);
        default:
+               vrf_tx_error(dev, skb);
                return NET_XMIT_DROP;
        }
 }