[PATCH v2 0/7] net: Support checksum in UDP tunnels

[PATCH v2 0/7] net: Support checksum in UDP tunnels

[Tom Herbert]

This patch series adds support for using checksums in UDP tunnels. With
this it is possible that two or more checksums may be set within the
same packet and we would like to do that efficiently.

This series also creates some new helper functions to be used by various
tunnel protocol implementations.

v2: Fixed indentation in tcp_v6_send_check arguments.

Re: [PATCH v2 0/7] net: Support checksum in UDP tunnels

[Jesse Gross]

On Sun, Jun 1, 2014 at 9:00 PM, Tom Herbert <therbert@google.com> wrote:
> This patch series adds support for using checksums in UDP tunnels. With
> this it is possible that two or more checksums may be set within the
> same packet and we would like to do that efficiently.

Do you also plan to implement double checksum offload in the absence
of TSO? I know it is functionally correct now but it's not offloaded.

Re: [PATCH v2 0/7] net: Support checksum in UDP tunnels

[Tom Herbert]

On Wed, Jun 4, 2014 at 3:28 PM, Jesse Gross <jesse@nicira.com> wrote:
> On Sun, Jun 1, 2014 at 9:00 PM, Tom Herbert <therbert@google.com> wrote:
>> This patch series adds support for using checksums in UDP tunnels. With
>> this it is possible that two or more checksums may be set within the
>> same packet and we would like to do that efficiently.
>
> Do you also plan to implement double checksum offload in the absence
> of TSO? I know it is functionally correct now but it's not offloaded.

We'd need HW support for that (like NETIF_F_HW_CSUM2). I think there
is some justification in light of tunneling protocols over UDP that
carry sensitive data, like virtual network ID, and don't otherwise
have any provision at L3 to detect data corruption (e.g. vxlan).

To implement, I think this is just another pair of checksum offsets
(csum_start, csum_offset) (CHECKSUM_PARTIAL2?) that a driver would
implement. Both csum_start and csum_offset could also be stored in 8
bits (right shifted to give values up to 510) if there are concerns
about space in TX descriptors.

Are any vendors interested in this functionality?

Thanks,
Tom

Re: [PATCH v2 0/7] net: Support checksum in UDP tunnels

[Jesse Gross]

On Wed, Jun 4, 2014 at 4:11 PM, Tom Herbert <therbert@google.com> wrote:
> On Wed, Jun 4, 2014 at 3:28 PM, Jesse Gross <jesse@nicira.com> wrote:
>> On Sun, Jun 1, 2014 at 9:00 PM, Tom Herbert <therbert@google.com> wrote:
>>> This patch series adds support for using checksums in UDP tunnels. With
>>> this it is possible that two or more checksums may be set within the
>>> same packet and we would like to do that efficiently.
>>
>> Do you also plan to implement double checksum offload in the absence
>> of TSO? I know it is functionally correct now but it's not offloaded.
>
> We'd need HW support for that (like NETIF_F_HW_CSUM2). I think there
> is some justification in light of tunneling protocols over UDP that
> carry sensitive data, like virtual network ID, and don't otherwise
> have any provision at L3 to detect data corruption (e.g. vxlan).
>
> To implement, I think this is just another pair of checksum offsets
> (csum_start, csum_offset) (CHECKSUM_PARTIAL2?) that a driver would
> implement. Both csum_start and csum_offset could also be stored in 8
> bits (right shifted to give values up to 510) if there are concerns
> about space in TX descriptors.
>
> Are any vendors interested in this functionality?

My guess is that you won't get too many vendors to respond here but
without going into specifics, I know that a few are planning this. I
guess there isn't too much benefit until the hardware exists (which is
a different situation from GSO) but I was just wondering if you were
planning on building out the infrastructure.