[ewg] Re: [ofa-general] Feedback on Developer's Summit

[Dror Goldenberg]

Roland Dreier wrote:
>  > As I said earlier on this thread,  the open issues I see with the stateless
>  > offload series are (A) the non interoperable checksum offload patch based on
>  > the IB ICRC sent by Michael (and if it is inter-operable, I'd like to be
>  > educated how) (B) LRO - a pure SW optimization, why it need to be in the
>  > ipoib driver and not at least shared with the network stack code above
>  > ipoib.
>
> For (A): as far as I'm concerned, turning off TCP/IP checksums is not
> something we want to do in IPoIB.  Is there anyone arguing in favor of
> it at this point?
>
>   

This issue is moot, but it's an option for devices that do not support 
csum offload and would like to still get better performance. Default is 
off. If you have a specific installation where it makes sense to turn 
off csum and rely on IB end to end ICRC then you can turn it on. On the 
other hand, it breaks the TCP/IP end to end data integrity validation 
and it violates csum calculation requirement. I can present what is 
there in the patch, motivation and let people choose. Therefore I think 
that a reasonable compromise will be to have it as an option and have it 
turned off by default.