Re: [PATCH net-next] virtio-net: avoid unnecessary checksum calculation on guest RX

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Nov 25, 2025 at 08:00:55PM +0000, Jon Kohler wrote:
> 
> 
> > On Nov 25, 2025, at 12:57 PM, Paolo Abeni <pabeni@redhat.com> wrote:
> > 
> > CC netdev
> 
> Thats odd, I used git send-email --to-cmd='./scripts/get_maintainer.pl,
> but it looks like in MAINTAINERS, that only would have hit
> VIRTIO CORE AND NET DRIVERS, which does not include netdev@
> 
> Should that have ?
> L: netdev@vger.kernel.org <mailto:netdev@vger.kernel.org>
> 
> Said another way, should all changes to include/linux/virtio_net.h
> be cc’d to netdev DL?
> 
> I suspect the answer is yes, I’ll send a patch for that in the
> interest of not having this issue again :)

I think yes. But only virtio net not core. I guess we should
split net from core then.

> > 
> > On 11/25/25 6:51 PM, Jon Kohler wrote:
> >> Commit a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP
> >> GSO tunneling.") inadvertently altered checksum offload behavior
> >> for guests not using UDP GSO tunneling.
> >> 
> >> Before, tun_put_user called tun_vnet_hdr_from_skb, which passed
> >> has_data_valid = true to virtio_net_hdr_from_skb.
> >> 
> >> After, tun_put_user began calling tun_vnet_hdr_tnl_from_skb instead,
> >> which passes has_data_valid = false into both call sites.
> >> 
> >> This caused virtio hdr flags to not include VIRTIO_NET_HDR_F_DATA_VALID
> >> for SKBs where skb->ip_summed == CHECKSUM_UNNECESSARY. As a result,
> >> guests are forced to recalculate checksums unnecessarily.
> >> 
> >> Restore the previous behavior by ensuring has_data_valid = true is
> >> passed in the !tnl_gso_type case.
> >> 
> >> Cc: Paolo Abeni <pabeni@redhat.com>
> >> Fixes: a2fb4bc4e2a6 ("net: implement virtio helpers to handle UDP GSO tunneling.")
> >> Signed-off-by: Jon Kohler <jon@nutanix.com>
> >> ---
> >> include/linux/virtio_net.h | 2 +-
> >> 1 file changed, 1 insertion(+), 1 deletion(-)
> >> 
> >> diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
> >> index b673c31569f3..570c6dd1666d 100644
> >> --- a/include/linux/virtio_net.h
> >> +++ b/include/linux/virtio_net.h
> >> @@ -394,7 +394,7 @@ virtio_net_hdr_tnl_from_skb(const struct sk_buff *skb,
> >> tnl_gso_type = skb_shinfo(skb)->gso_type & (SKB_GSO_UDP_TUNNEL |
> >>     SKB_GSO_UDP_TUNNEL_CSUM);
> >> if (!tnl_gso_type)
> >> - return virtio_net_hdr_from_skb(skb, hdr, little_endian, false,
> >> + return virtio_net_hdr_from_skb(skb, hdr, little_endian, true,
> >>        vlan_hlen);
> >> 
> >> /* Tunnel support not negotiated but skb ask for it. */
> > 
> > virtio_net_hdr_tnl_from_skb() is used also by the virtio_net driver,
> > which in turn must not use VIRTIO_NET_HDR_F_DATA_VALID on tx.
> 
> Ah! Good eye, I’ll see what trouble I can get into and send a v2
> > 
> > I think you need to add another argument to
> > virtio_net_hdr_tnl_from_skb(), or possibly implement a separate helper
> > to take care of csum offload - the symmetric of
> > virtio_net_handle_csum_offload().
> > 
> > Also you need to CC netdev, otherwise the patch will not be processed by
> > patchwork.
> > 
> > /P
> 
> No problems on cc’ing netdev, just didn’t realize this one header didn’t
> auto cc the list. Will keep an eye on that, and happy to send a patch to
> MAINTAINERS file for discussion.