Re: [PATCH net-next v2 10/10] bnxt_en: Implement .ndo_features_check().

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

[-- Attachment #1: Type: text/plain, Size: 1346 bytes --]

On Mon, Apr 26, 2021 at 10:00 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Mon, 26 Apr 2021 09:45:17 -0700 Michael Chan wrote:
> > On Mon, Apr 26, 2021 at 9:35 AM Michael Chan <michael.chan@broadcom.com> wrote:
> > > On Mon, Apr 26, 2021 at 9:29 AM Jakub Kicinski <kuba@kernel.org> wrote:
> > > > On Sun, 25 Apr 2021 13:45:27 -0400 Michael Chan wrote:
> > > > This should have been "features & ~(CSUM | GSO)" if you actually
> > > > accepted my feedback.
> >
> > Sorry, hit send too early.  If it is not UDP, it could be GRE or IP
> > over IP for example, right?  Why do we need to turn off offload?
>
> All supported protocols can be included in the allow list.
> That's one of the costs of NETIF_F_IP_CSUM, the driver needs
> to make sure the device can understand the packet.

Only UDP encapsulations have the 2 port limitations.  The rest are supported.

>
> > > I mentioned extension headers as an example,
> >
> > Extension headers (Geneve for example) are supported.
>
> I thought the Geneve things were called options. I meant IPv6 extension
> headers, which the device may also support, but then the right thing to
> do is something like a call to ipv6_skip_exthdr() to retrieve the L4
> proto.

You're right and I misunderstood you.  I will double check for ipv6
extension headers support and will send a follow up patch.  Thanks.

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