On Fri, Oct 30, 2020 at 5:49 PM Xie He <xie.he.0...@gmail.com> wrote:
>
> On Fri, Oct 30, 2020 at 2:28 PM Willem de Bruijn
> <willemdebruijn.ker...@gmail.com> wrote:
> >
> > Yes, it might require holding off the other patches until net is
> > merged into net-next.
> >
> > Packet sockets are likely not the only way these packets are received?
> > It changes mac_len as computed in __netif_receive_skb_core.
>
> I looked at __netif_receive_skb_core. I didn't see it computing mac_len?
It's indirect:
skb_reset_network_header(skb);
if (!skb_transport_header_was_set(skb))
skb_reset_transport_header(skb);
skb_reset_mac_len(skb);
> I thought only AF_PACKET/RAW sockets would need this information
> because other upper layers would not care about what happened in L2.
I think that's a reasonable assumption. I don't have a good
counterexample ready. Specific to this case, it seems to have been
working with no one complaining so far ;)
> I see mac_len is computed in netif_receive_generic_xdp. I'm not clear
> about the reason why it calculates it. But it seems that it considers
> the L2 header as an Ethernet header, which is incorrect for this
> driver.
>
> > If there is no real bug that is fixed, net-next is fine.
