Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:7 as permitted sender) client-ip=2620:137:e000::3:7;
MIME-Version: 1.0
References: <20231104-gemini-largeframe-fix-v1-0-9c5513f22f33@linaro.org>
 <20231104-gemini-largeframe-fix-v1-4-9c5513f22f33@linaro.org> <036b481e-ac5b-4e77-b93a-4badaf19e185@lunn.ch>
In-Reply-To: <036b481e-ac5b-4e77-b93a-4badaf19e185@lunn.ch>
From:   Linus Walleij <linus.walleij@linaro.org>
Date:   Sun, 5 Nov 2023 21:56:23 +0100
Message-ID: <CACRpkdYp1kiGY0K7kNF+qadPyq1hu3G=2oc1gXnCt3DjtiJxag@mail.gmail.com>
Subject: Re: [PATCH net 4/4] net: ethernet: cortina: Handle large frames
To:     Andrew Lunn <andrew@lunn.ch>
Cc:     Hans Ulli Kroll <ulli.kroll@googlemail.com>,
        "David S. Miller" <davem@davemloft.net>,
        Eric Dumazet <edumazet@google.com>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>,
        =?UTF-8?B?TWljaGHFgiBNaXJvc8WCYXc=?= <mirq-linux@rere.qmqm.pl>,
        Vladimir Oltean <olteanv@gmail.com>,
        linux-arm-kernel@lists.infradead.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Sat, Nov 4, 2023 at 3:57=E2=80=AFPM Andrew Lunn <andrew@lunn.ch> wrote:

> > +              * Just bypass on bigger frames.
> > +              */
> > +             word1 |=3D TSS_BYPASS_BIT;
> > +     } else if (skb->ip_summed !=3D CHECKSUM_NONE) {
>
> I've never looked at how the network stack does checksums. But looking
> at this patch, it made me wounder, how do you tell the stack it needs
> to do a software checksum because the hardware cannot?

I read up on it: the documentation is in
Documentation/networking/checksum-offloads.rst
and in the header for skbuff, include/linux/skbuff.h

Actually we should check for =3D=3D CHECKSUM_PARTIAL which means
we need to do the checksum (!=3D CHECKSUM_NONE is not inclusive)
then I call a software fallback directly from the driver if I need to.

> Or for this
> driver, is it always calculating a checksum, which is then ignored?
> Maybe you can improve performance a little but disabling software
> checksum when it is not needed?

The ping was somehow working without proper checksum
before, but I think I'm doing the right thing now, also tested with
HTTP traffic, check out v2.

Thanks for pointing it out, the patch looks way better now.

Yours,
Linus Walleij