On Fri, 2006-05-19 at 12:05, Jason Gunthorpe wrote: > On Fri, May 19, 2006 at 10:19:50AM -0400, Hal Rosenstock wrote: > > > > I think it is also prudent to not use a GRH if the DGID's prefix is > > > fe80::/64 (link local scope). > > > > Agreed. This saves on the overhead in the local subnet case. > > > > However, wouldn't the HopLimit returned from the SA in this case be 0 or > > 1 so why would that check be needed ? > > > > If it is, I think the DGID prefix checking may be more than just this. > > I think the SA should work like you are describing, but how sure are > we that all existing ones do? I'm only suggesting it in case there are > some broken SA's out there.
Are you aware of any currently that don't ? In this case, IMO the penalty of including the GRH should be paid (until those SMs are fixed) rather than adding any additional prefix checking into the end node. -- Hal _______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
