Eli Cohen wrote:
On Sun, 2008-03-23 at 08:36 +0200, Or Gerlitz wrote:
with IB_WC_LSO never being used over this patchset, can we just remove it?
It is used in cq.c. I don't think we can remove it even though IPoIB
does not use this completion type - other clients might want to.
How about adding this only when someone really wants it? I don't see
more clients using it other than IPoIB serving TCP traffic in the near
future.
@@ -660,6 +664,9 @@ struct ib_send_wr {
+ void *header;
+ int hlen;
+ int mss;
Can you add shorting documentation for the new fields?
Of course I can add - I just followed the "spirit" of other parts of the
code where there is not description. Is this an exception in this
regard?
yes, this is an exception since all the other fields has strict relation
to the IB spec, and not that I think that stateless offloading has to be
specc-ed before implemented or specc-ed at all, but its not sure not
obvious for someone looking on IB send WQE that <header,hlen,mss> is an
LSO material. If its easier for you to document it under
Documentation/infiniband that's fine as well.
Or.
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
