On Mon, 12 Sep 2005, Roland Dreier wrote: >I rewrote your patch to fix a few things: > > - I believe we always need to copy the redirect GID, even if the > redirect LID is non-zero; otherwise we will end up putting the GID > of the original port in our CM REQ. From my reading of the > description of ClassPortInfo in the IB spec, I think the redirect > GID must always be valid if redirection is being done.
OK. > - I moved the connect handling into a new function > srp_connect_target(), so that both the initial connect and future > reconnects get handling for port and DLID redirection. OK. > - I got rid of the change for CM REJ reason 25 -- that reject code > does not carry a ClassPortInfo in the ARI, but rather just the > GID at offset 0. OK. >By the way, it's not that critical for SRP, since it's not in the >upstream kernel yet, but I don't think we should use obfuscated email >addresses in Signed-off-by lines: > > > Signed-off-by: John Kingman <kingman <at> storagegear.com> > >Just put the '@' in there -- I think there are much bigger spam >sources to worry about. OK. >Anyway, here's the updated patch. Does this look OK to you? I believe that there are some errors that have already been noted. Will you be sending another patch? John _______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
