Re: [Intel-wired-lan] bug with rx-udp-gro-forwarding offloading?

[Paolo Abeni <pabeni@redhat.com>]

On Tue, 2023-07-04 at 15:23 +0200, Ian Kumlien wrote:
> On Tue, Jul 4, 2023 at 2:54 PM Paolo Abeni <pabeni@redhat.com> wrote:
> > 
> > On Tue, 2023-07-04 at 13:36 +0200, Ian Kumlien wrote:
> > > Propper bug this time:
> > > cat bug.txt | ./scripts/decode_stacktrace.sh vmlinux
> > 
> > To be sure, is this with the last patch I shared? this one I mean:
> 
> The current modifications I have, on top of v6.4.1, is:
> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> index cea28d30abb5..8552caa197f9 100644
> --- a/net/core/skbuff.c
> +++ b/net/core/skbuff.c
> @@ -4272,6 +4272,11 @@ struct sk_buff *skb_segment_list(struct sk_buff *skb,
> 
>         skb_shinfo(skb)->frag_list = NULL;
> 
> +       /* later code will clear the gso area in the shared info */
> +       err = skb_header_unclone(skb, GFP_ATOMIC);
> +       if (err)
> +               goto err_linearize;
> +
>         while (list_skb) {
>                 nskb = list_skb;
>                 list_skb = list_skb->next;
> @@ -4328,6 +4333,9 @@ struct sk_buff *skb_segment_list(struct sk_buff *skb,
> 
>         skb->prev = tail;
> 
> +       if (WARN_ON_ONCE(!skb->next))
> +               goto err_linearize;
> +
>         if (skb_needs_linearize(skb, features) &&
>             __skb_linearize(skb))
>                 goto err_linearize;
> ---
> 
> > https://lore.kernel.org/netdev/92a4d42491a2c219192ae86fa04b579ea3676d8c.camel@redhat.com/
> > 
> > Could you please additionally enable CONFIG_DEBUG_NET in your build?
> 
> Sure, will do
> 
> > Could you please give a detailed description of your network topology
> > and the running traffic?
> 
> This machine has two "real interfaces" and two interfaces that runs as
> bridges for virtual machines
> eno1 - real internal
> eno2 - bridge - internal
> eno3 - real external
> eno4 - bridge - external
> 
> The bridges are used by three virtual machines, two of which are
> attached on both networks
> 
> Traffic seemed to be video streaming, at least at first, now I don't
> really know. I do have a few smart devices so I assume there is
> a bit of multicast traffic as well - but not really anything unusual as such.

In there any XDP program running on the host side? Possibly changing
the packet hdr?

Thanks!

/P