On Mon, 2007-02-05 at 11:56, Michael Arndt wrote: > Hi, > > > A router should not allow a SMP to cross a subnet boundary. SMPs are > > restricted to the local subnet. > > I work on a discovering mechanism for switchless InfiniBand Architectures > like Rings, Tori or maybe Hyper-Cubes. There is just one single subnet, no > switches or routers. Please ignore the background and focus to the problem > about the second packet. Maybe you have some ideas even you are not involved > in the hole project. That would be nice.
Guess you don't mean IB router when you say router in your description. I also have no theories without more information: Is the sender a normal node ? Is normal node mean standard OpenIB without changes ? How was the SMI changed ? On which nodes ? Only the intermediate one ? Aside from the initial path being [0][1][1], what are the hop count and hop pointer ? What are DrDLID and DrSLID as well as the LIDs in the LRH of the SMP ? -- Hal > Thanks Michael _______________________________________________ openib-general mailing list openib-general@openib.org http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
