[openib-general] Problem is routing CM REQ

[Sean Hefty]

>What your #4 and #5 are talking about is not just that, but also PR
>queries that can unambigously identify the LID selections of the
>router in advance. That is hugely different! IMHO, just because a
>reversible path exists and will be used by the router shouldn't be
>taken to mean that the it is the only one or that the SA can tell you
>which of many possible choices it will be.

Yes - I was trying to define a routed path as reversible with respect to a
connection.  It makes things easier.  :)  This is where we've been
disconnecting.

I was wanting a packet sent from the remote GID to the local GID to come back
over the local DLID/SLID path specified in the path record if reversible is
true.  I give.  This was too strong of an assumption, since the response path
could travel a different DLID/SLID path and still qualify as reversible.

So, it seems that with respect to connections between subnets, path records
should be treated as if they were not reversible.  Using my model then would
require 4 queries...  (I need to read back through the discussion and see if the
different ideas can be condensed/summarized.)

>If so then I'd expect also for a SGID=off-subnet query to return the
>remote LIDs to make CM work properly with existing conforming
>implementations (that use 3 PR queries to get non-reversable paths
>;>).

I think it makes more sense to push interaction with a remote SA to the end node
to give them greater control over the query and avoid the local SA indirection.

- Sean