net: vxlan: lwt: Use source ip address during route lookup.
authorpravin shelar <pshelar@ovn.org>
Sat, 6 Aug 2016 00:45:36 +0000 (17:45 -0700)
committerDavid S. Miller <davem@davemloft.net>
Mon, 8 Aug 2016 21:16:23 +0000 (14:16 -0700)
LWT user can specify destination as well as source ip address
for given tunnel endpoint. But vxlan is ignoring given source
ip address. Following patch uses both ip address to route the
tunnel packet. This consistent with other LWT implementations,
like GENEVE and GRE.

Fixes: ee122c79d42 ("vxlan: Flow based tunneling").
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Acked-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/vxlan.c

index da4e3d6632f647f95b6fdbc0d22abf098afdbe01..b81223422cf07e35225d16bb4d255f83f3e92d0f 100644 (file)
@@ -1811,7 +1811,7 @@ static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
        fl4.flowi4_mark = skb->mark;
        fl4.flowi4_proto = IPPROTO_UDP;
        fl4.daddr = daddr;
-       fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
+       fl4.saddr = *saddr;
 
        rt = ip_route_output_key(vxlan->net, &fl4);
        if (!IS_ERR(rt)) {
@@ -1847,7 +1847,7 @@ static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
        memset(&fl6, 0, sizeof(fl6));
        fl6.flowi6_oif = oif;
        fl6.daddr = *daddr;
-       fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
+       fl6.saddr = *saddr;
        fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
        fl6.flowi6_mark = skb->mark;
        fl6.flowi6_proto = IPPROTO_UDP;
@@ -1920,7 +1920,8 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
        struct rtable *rt = NULL;
        const struct iphdr *old_iph;
        union vxlan_addr *dst;
-       union vxlan_addr remote_ip;
+       union vxlan_addr remote_ip, local_ip;
+       union vxlan_addr *src;
        struct vxlan_metadata _md;
        struct vxlan_metadata *md = &_md;
        __be16 src_port = 0, dst_port;
@@ -1938,6 +1939,7 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
                dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
                vni = rdst->remote_vni;
                dst = &rdst->remote_ip;
+               src = &vxlan->cfg.saddr;
                dst_cache = &rdst->dst_cache;
        } else {
                if (!info) {
@@ -1948,11 +1950,15 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
                dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
                vni = vxlan_tun_id_to_vni(info->key.tun_id);
                remote_ip.sa.sa_family = ip_tunnel_info_af(info);
-               if (remote_ip.sa.sa_family == AF_INET)
+               if (remote_ip.sa.sa_family == AF_INET) {
                        remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
-               else
+                       local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
+               } else {
                        remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
+                       local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
+               }
                dst = &remote_ip;
+               src = &local_ip;
                dst_cache = &info->dst_cache;
        }
 
@@ -1992,15 +1998,14 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
        }
 
        if (dst->sa.sa_family == AF_INET) {
-               __be32 saddr;
-
                if (!vxlan->vn4_sock)
                        goto drop;
                sk = vxlan->vn4_sock->sock->sk;
 
                rt = vxlan_get_route(vxlan, skb,
                                     rdst ? rdst->remote_ifindex : 0, tos,
-                                    dst->sin.sin_addr.s_addr, &saddr,
+                                    dst->sin.sin_addr.s_addr,
+                                    &src->sin.sin_addr.s_addr,
                                     dst_cache, info);
                if (IS_ERR(rt)) {
                        netdev_dbg(dev, "no route to %pI4\n",
@@ -2043,13 +2048,12 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
                if (err < 0)
                        goto xmit_tx_error;
 
-               udp_tunnel_xmit_skb(rt, sk, skb, saddr,
+               udp_tunnel_xmit_skb(rt, sk, skb, src->sin.sin_addr.s_addr,
                                    dst->sin.sin_addr.s_addr, tos, ttl, df,
                                    src_port, dst_port, xnet, !udp_sum);
 #if IS_ENABLED(CONFIG_IPV6)
        } else {
                struct dst_entry *ndst;
-               struct in6_addr saddr;
                u32 rt6i_flags;
 
                if (!vxlan->vn6_sock)
@@ -2058,7 +2062,8 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
 
                ndst = vxlan6_get_route(vxlan, skb,
                                        rdst ? rdst->remote_ifindex : 0, tos,
-                                       label, &dst->sin6.sin6_addr, &saddr,
+                                       label, &dst->sin6.sin6_addr,
+                                       &src->sin6.sin6_addr,
                                        dst_cache, info);
                if (IS_ERR(ndst)) {
                        netdev_dbg(dev, "no route to %pI6\n",
@@ -2104,7 +2109,8 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
                        return;
                }
                udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
-                                    &saddr, &dst->sin6.sin6_addr, tos, ttl,
+                                    &src->sin6.sin6_addr,
+                                    &dst->sin6.sin6_addr, tos, ttl,
                                     label, src_port, dst_port, !udp_sum);
 #endif
        }