Re: [PATCH net-next 1/2] net: enetc: declare NETIF_F_IP_CSUM and do it in software

On Thu, Oct 07, 2021 at 09:47:20AM +0300, Ioana Ciornei wrote:

On Wed, Oct 06, 2021 at 05:24:18PM -0700, Jakub Kicinski wrote:

quoted

On Wed,  6 Oct 2021 23:13:07 +0300 Ioana Ciornei wrote:

quoted

This is just a preparation patch for software TSO in the enetc driver.
Unfortunately, ENETC does not support Tx checksum offload which would
normally render TSO, even software, impossible.

Declare NETIF_F_IP_CSUM as part of the feature set and do it at driver
level using skb_csum_hwoffload_help() so that we can move forward and
also add support for TSO in the next patch.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>

Did you choose NETIF_F_IP_CSUM intentionally?
It'll only support IPv4, and since you always fall back to SW
I'd think NETIF_F_HW_CSUM makes more sense.

Somewhat intentionally, yes.

If I would use NETIF_F_HW_CSUM, as I understand it, the GSO path, added
in the next patch, would have to compute the checksum not only for IPv6
but also for any other protocols other than UDP and TCP (which currently
it supports).
I just didn't look into that at the moment.

Now that I think of it, you have a point with declaring NETIF_F_HW_CSUM.

In the non-GSO case, skb_checksum_help() will be able to handle anything
that we throw at it.

On the GSO case, only skbs with TCP over IPv4 or IPv6 (depending on what
we declare in the features - TSO/TSO6) will be received - which the csum
code can handle.

Anyhow, I'll change this to use NETIF_F_HW_CSUM instead of
NETIF_F_IP_CSUM and I'll also look into IPv6 for the TSO part.

Thanks,
Ioana