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

Jason Gunthorpe jgunthorpe at obsidianresearch.com
Mon Nov 5 14:08:17 PST 2007


On Mon, Nov 05, 2007 at 01:51:58PM -0800, 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?

I think at the end of the thread it was left sort of
hanging.. Certainly, for instance, the qlogic vnic driver does exactly
the same thing as was proposed for ipoib when it's hw csum mode is
used, so it isn't an unconditionally bad thing.

IMHO, what was never addressed was to produce a patch set that made
the csum offload process work properly and not be incompatible with
existing code. Ie negotiated, and tagged with appropriate CHECKSUM_*
marks in all the right places so that forwarding works. Basically make
IPoIB CM mode use all the same optimization tricks as we see in a vnic
protocol or in something like the xen networking bypass.

Unless someone is stepping up to do that work it doesn't seem like
there is anything that needs discussion face to face?

Jason



More information about the general mailing list