[openib-general] question regarding GRH flag in ib_ah_attr

[Hal Rosenstock]

On Fri, 2006-05-12 at 13:55, Sean Hefty wrote:
> Jason Gunthorpe wrote:
> > How about this, how do you see this scenario:
> > 
> > 1) Client gets a DGID from 'someplace'
> > 2) Client sends a SA query to resolve the DGID to a Path Record
> > 3) Client configures a QP based on the Path Record
> > 
> > Now, the question I'm interested in is this:
> >   During step #3 what test should the client apply to determine if a 
> >   GRH should be used with the QP.
> 
> This is the scenario that I need to resolve.
> 
> What would happen if the GRH flag were always set?

That would work but there would be additional overhead (especially for
small packets this would be more noticeable) in the local subnet case.

> Set only if the GID prefixes of the SGID/DGID were different?

That's one way although it is more complex than what Jason has been
proposing for this (SA response with HopLimit>=2). I'm not yet sure that
the latter is sufficient as I think there may be other factors as to
whether a packet is forwarded off subnet. One is the prefix scope (but I
would think link local scopes should be limited in HopLimit except for
multicasts (Jason cited that multicasts were required to have HopLimit
0xFF) but they require GRHs anyhow) so maybe I'm wrong about this and
HopLimit>=2 is sufficient.

-- Hal

> - Sean