Re: [PATCH net-next 10/10] bnxt_en: Implement .ndo_features_check().

[Michael Chan <michael.chan@broadcom.com>]

[-- Attachment #1: Type: text/plain, Size: 1580 bytes --]

On Sat, Apr 24, 2021 at 2:52 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Sat, 24 Apr 2021 16:14:31 -0400 Michael Chan wrote:
> > +     features = vlan_features_check(skb, features);
> > +     if (!skb->encapsulation)
> > +             return features;
> > +
> > +     switch (vlan_get_protocol(skb)) {
> > +     case htons(ETH_P_IP):
> > +             l4_proto = ip_hdr(skb)->protocol;
> > +             break;
> > +     case htons(ETH_P_IPV6):
> > +             l4_proto = ipv6_hdr(skb)->nexthdr;
> > +             break;
> > +     default:
> > +             return features;
> > +     }
> > +
> > +     /* For UDP, we can only handle 1 Vxlan port and 1 Geneve port. */
> > +     if (l4_proto == IPPROTO_UDP) {
> > +             struct bnxt *bp = netdev_priv(dev);
> > +             __be16 udp_port = udp_hdr(skb)->dest;
> > +
> > +             if (udp_port != bp->vxlan_port && udp_port != bp->nge_port)
> > +                     return features & ~(NETIF_F_CSUM_MASK |
> > +                                         NETIF_F_GSO_MASK);
> > +     }
> > +     return features;
>
> This is still written a little too much like a block list.
>
> What if, for example it's a UDP tunnel but with extension headers?
> Is there any particular case that is served by not writing it as:
>
>         if (l4_proto == UDP && (port == bp->vxl_port ||
>                                 port == bp->nge_port))
>                 return features;
>         return features & ~(CSUM | GSO);
> ?

Sure, I can change it to check for port == instead of port != to make
it more straight forward.

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]