> > > Christoph, > > > > We had a smaller volume move to cache the device attributes on the IB > > device structure, and I just realized it was dropped on the floor. > > Ira, that was a reviewer comment you got when worked on OPA and I > > missed the fact it didn't reach to acceptance > > http://marc.info/?t=142309310600006&r=1&w=2 > > I vote for 1st and most doing this and taking things from there. > > That's also workable. > > But Ira's patch is missing ib_query_device conversion to simply take the > cached > device attributes. > > I'm fine with either way to go.
My patch eventually became the get_port_immutable call because the OPA MAD flag was going to be based per port rather than per device. This fit nicely which other devices which required different core support based on the port rather than the device. For port attributes there are values which can change and some which can not. Thus the addition of a set of immutable values which the core layer could safely store in the device structure. I could see a similar patch to Christophs to remove get_port_immutable. At the time I was just following the design pattern for the rest of the infrastructure. However, for query_port there are values which may change so we need to keep that call. In my quick review of the device attributes I don't see anything which can change after the device is loaded by the core. So it seems Christoph patch will work. Ira
