[PATCH net] udptunnels: Call handle_offloads after inserting vlan tag.

[PATCH net] udptunnels: Call handle_offloads after inserting vlan tag.

[Jesse Gross]

handle_offloads() calls skb_reset_inner_headers() to store
the layer pointers to the encapsulated packet. However, we
currently push the vlag tag (if there is one) onto the packet
afterwards. This changes the MAC header for the encapsulated
packet but it is not reflected in skb->inner_mac_header, which
breaks GSO and drivers which attempt to use this for encapsulation
offloads.

Fixes: 1eaa8178 ("vxlan: Add tx-vlan offload support.")
Signed-off-by: Jesse Gross <jesse@nicira.com>
---
 drivers/net/vxlan.c | 20 ++++++++++----------
 net/ipv4/geneve.c   |  8 ++++----
 2 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
index f8528a4..fceb637 100644
--- a/drivers/net/vxlan.c
+++ b/drivers/net/vxlan.c
@@ -1713,12 +1713,6 @@ static int vxlan6_xmit_skb(struct dst_entry *dst, struct sk_buff *skb,
 		}
 	}
 
-	skb = iptunnel_handle_offloads(skb, udp_sum, type);
-	if (IS_ERR(skb)) {
-		err = -EINVAL;
-		goto err;
-	}
-
 	skb_scrub_packet(skb, xnet);
 
 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
@@ -1738,6 +1732,12 @@ static int vxlan6_xmit_skb(struct dst_entry *dst, struct sk_buff *skb,
 		goto err;
 	}
 
+	skb = iptunnel_handle_offloads(skb, udp_sum, type);
+	if (IS_ERR(skb)) {
+		err = -EINVAL;
+		goto err;
+	}
+
 	vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
 	vxh->vx_flags = htonl(VXLAN_HF_VNI);
 	vxh->vx_vni = md->vni;
@@ -1798,10 +1798,6 @@ int vxlan_xmit_skb(struct rtable *rt, struct sk_buff *skb,
 		}
 	}
 
-	skb = iptunnel_handle_offloads(skb, udp_sum, type);
-	if (IS_ERR(skb))
-		return PTR_ERR(skb);
-
 	min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
 			+ VXLAN_HLEN + sizeof(struct iphdr)
 			+ (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
@@ -1817,6 +1813,10 @@ int vxlan_xmit_skb(struct rtable *rt, struct sk_buff *skb,
 	if (WARN_ON(!skb))
 		return -ENOMEM;
 
+	skb = iptunnel_handle_offloads(skb, udp_sum, type);
+	if (IS_ERR(skb))
+		return PTR_ERR(skb);
+
 	vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
 	vxh->vx_flags = htonl(VXLAN_HF_VNI);
 	vxh->vx_vni = md->vni;
diff --git a/net/ipv4/geneve.c b/net/ipv4/geneve.c
index 5a4828b..a566a2e 100644
--- a/net/ipv4/geneve.c
+++ b/net/ipv4/geneve.c
@@ -113,10 +113,6 @@ int geneve_xmit_skb(struct geneve_sock *gs, struct rtable *rt,
 	int min_headroom;
 	int err;
 
-	skb = udp_tunnel_handle_offloads(skb, csum);
-	if (IS_ERR(skb))
-		return PTR_ERR(skb);
-
 	min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
 			+ GENEVE_BASE_HLEN + opt_len + sizeof(struct iphdr)
 			+ (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
@@ -131,6 +127,10 @@ int geneve_xmit_skb(struct geneve_sock *gs, struct rtable *rt,
 	if (unlikely(!skb))
 		return -ENOMEM;
 
+	skb = udp_tunnel_handle_offloads(skb, csum);
+	if (IS_ERR(skb))
+		return PTR_ERR(skb);
+
 	gnvh = (struct genevehdr *)__skb_push(skb, sizeof(*gnvh) + opt_len);
 	geneve_build_header(gnvh, tun_flags, vni, opt_len, opt);
 
-- 
1.9.1

Re: [PATCH net] udptunnels: Call handle_offloads after inserting vlan tag.

[David Miller]

From: Jesse Gross <jesse@nicira.com>
Date: Thu,  9 Apr 2015 11:19:14 -0700

> handle_offloads() calls skb_reset_inner_headers() to store
> the layer pointers to the encapsulated packet. However, we
> currently push the vlag tag (if there is one) onto the packet
> afterwards. This changes the MAC header for the encapsulated
> packet but it is not reflected in skb->inner_mac_header, which
> breaks GSO and drivers which attempt to use this for encapsulation
> offloads.
> 
> Fixes: 1eaa8178 ("vxlan: Add tx-vlan offload support.")
> Signed-off-by: Jesse Gross <jesse@nicira.com>

Applied and queued up for -stable, thanks.