When you get a chance, can you try out this patch? I tested that it worked
for a local subnet connection by commenting out the hop_limit check. So,
I'm interested to know if you run into any problems. If you do run into
issues, madeye may be able to help.
I've reworked this patch, and added a couple more based on 2.6.21-rc4. The
patches are available from
git://git.openfabrics.org/~shefty/rdma-dev.git ib_router
I started to put checks to validate the CM REQ data against the received LRH
data, but dropped them.
A couple of other notes:
* If the LID in the CM REQ is permissive, then the LID/SL data from the LRH is
used instead. The user of the ib_cm is responsible for setting the LID to
permissive to control this, which should allow this functionality to work
alongside the more general ib_remote_sa type of solution.
* The CM architecture allows for primary and alternate paths where one is subnet
local, and the other is routed. I don't think anything breaks this at the
moment, and I believe that the changes to the ib_cm to support routing are done.
(Support for non-reversible paths is still missing.)
I'm still developing the ib_remote_sa solution. All of my testing done so far
is limited to a single subnet, with hacks thrown into the code to force
execution of the new code paths.
- Sean
_______________________________________________
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
