ip6_tunnel: fix GRE6 segmentation

Commit 6c11fbf97e69 ("ip6_tunnel: add MPLS transmit support")
moved assiging inner_ipproto down from ipxip6_tnl_xmit() to
its callee ip6_tnl_xmit(). The latter is also used by GRE.

Since commit 38720352412a ("gre: Use inner_proto to obtain inner
header protocol") GRE had been depending on skb->inner_protocol
during segmentation. It sets it in gre_build_header() and reads
it in gre_gso_segment(). Changes to ip6_tnl_xmit() overwrite
the protocol, resulting in GSO skbs getting dropped.

Note that inner_protocol is a union with inner_ipproto,
GRE uses the former while the change switched it to the latter
(always setting it to just IPPROTO_GRE).

Restore the original location of skb_set_inner_ipproto(),
it is unclear why it was moved in the first place.

Fixes: 6c11fbf97e69 ("ip6_tunnel: add MPLS transmit support")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tested-by: Vadim Fedorenko <vfedorenko@novek.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
index 288bafd..28ca70a 100644 (file)
--- a/net/ipv6/ip6_tunnel.c
+++ b/net/ipv6/ip6_tunnel.c
@@ -1239,8 +1239,6 @@ route_lookup:
        if (max_headroom > dev->needed_headroom)
                dev->needed_headroom = max_headroom;
 
-       skb_set_inner_ipproto(skb, proto);
-
        err = ip6_tnl_encap(skb, t, &proto, fl6);
        if (err)
                return err;
@@ -1377,6 +1375,8 @@ ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
        if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
                return -1;
 
+       skb_set_inner_ipproto(skb, protocol);
+
        err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
                           protocol);
        if (err != 0) {