On Tue, 10 Mar 2026 11:05:33 +0100, Paolo Abeni [off-list ref] wrote:

On 3/5/26 4:19 AM, Xuan Zhuo wrote:

quoted

The commit a2fb4bc4e2a6a03 ("net: implement virtio helpers to handle UDP
GSO tunneling.") introduces support for the UDP GSO tunnel feature in
virtio-net.

The virtio spec says:

    If the \field{gso_type} has the VIRTIO_NET_HDR_GSO_UDP_TUNNEL_IPV4 bit or
    VIRTIO_NET_HDR_GSO_UDP_TUNNEL_IPV6 bit set, \field{hdr_len} accounts for
    all the headers up to and including the inner transport.

The commit did not update the hdr_len to include the inner transport.

I observed that the "hdr_len" is 116 for this packet:

    17:36:18.241105 52:55:00:d1:27:0a > 2e:2c:df:46:a9:e1, ethertype IPv4 (0x0800), length 2912: (tos 0x0, ttl 64, id 45197, offset 0, flags [none], proto UDP (17), length 2898)
        192.168.122.100.50613 > 192.168.122.1.4789: [bad udp cksum 0x8106 -> 0x26a0!] VXLAN, flags [I] (0x08), vni 1
    fa:c3:ba:82:05:ee > ce:85:0c:31:77:e5, ethertype IPv4 (0x0800), length 2862: (tos 0x0, ttl 64, id 14678, offset 0, flags [DF], proto TCP (6), length 2848)
        192.168.3.1.49880 > 192.168.3.2.9898: Flags [P.], cksum 0x9266 (incorrect -> 0xaa20), seq 515667:518463, ack 1, win 64, options [nop,nop,TS val 2990048824 ecr 2798801412], length 2796

116 = 14(mac) + 20(ip) + 8(udp) + 8(vxlan) + 14(inner mac) + 20(inner ip) + 32(innner tcp)

Fixes: a2fb4bc4e2a6a03 ("net: implement virtio helpers to handle UDP GSO tunneling.")
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
---
 include/linux/virtio_net.h | 31 ++++++++++++++++++++-----------
 1 file changed, 20 insertions(+), 11 deletions(-)

diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
index 23f42bb8ece0..906157779955 100644
--- a/include/linux/virtio_net.h
+++ b/include/linux/virtio_net.h

@@ -209,18 +209,29 @@ static inline int virtio_net_hdr_to_skb(struct sk_buff *skb,

 static inline void virtio_net_set_hdrlen(const struct sk_buff *skb,
 					 struct virtio_net_hdr *hdr,
+					 struct skb_shared_info *sinfo,
 					 bool little_endian,
 					 bool feature_hdrlen)
 {
 	u16 hdr_len;

 	if (feature_hdrlen) {
-		hdr_len = skb_transport_offset(skb);
-
-		if (hdr->gso_type == VIRTIO_NET_HDR_GSO_UDP_L4)
-			hdr_len += sizeof(struct udphdr);
-		else
-			hdr_len += tcp_hdrlen(skb);
+		if (sinfo->gso_type & (SKB_GSO_UDP_TUNNEL |
+				       SKB_GSO_UDP_TUNNEL_CSUM)) {
+			hdr_len = skb_inner_transport_offset(skb);
+
+			if (hdr->gso_type == VIRTIO_NET_HDR_GSO_UDP_L4)
+				hdr_len += sizeof(struct udphdr);
+			else
+				hdr_len += inner_tcp_hdrlen(skb);
+		} else {
+			hdr_len = skb_transport_offset(skb);
+
+			if (hdr->gso_type == VIRTIO_NET_HDR_GSO_UDP_L4)
+				hdr_len += sizeof(struct udphdr);
+			else
+				hdr_len += tcp_hdrlen(skb);
+		}

I already mentioned the following a couple of times, with no replies...
perhaps the 3rd attempt will be more successful.

I think it's a bad adding tunnel specific check here. It sounds like
layering violation. Instead you could have tunnel-specific variant, say
virtio_net_set_tnl_hdrlen(), and use it in virtio_net_hdr_tnl_from_skb().

Or you could try what i suggested in:

https://lore.kernel.org/netdev/CAF6piCLkv6kFqoq7OQfJ=Su9AVHSQ9J7DzaumOSf5xuf9w-kyA@mail.gmail.com/#t (local)

I saw your suggestions, though I didn't reply directly. I proceeded to adjust
the code based on my understanding of your feedback, but it appears my
understanding was incorrect. After re-reading your suggestions, I think
virtio_net_hdr_from_skb does not need to be changed. Instead, within
virtio_net_hdr_tnl_from_skb, we should directly call the device header length
function based on whether the feature is negotiated.

Thanks.

Thanks,

Paolo