Roland Dreier wrote:
Fab> What happens during the handoff from the boot environment to Fab> the OS? Does the HCA get disabled and then the mthca driver Fab> starts fresh? Or does the mthca driver inherit a device that Fab> is already fully initialized. If it gets re-initialized, Fab> don't the ports go down when the boot agent shuts down (and Fab> the SM should get a GID out of service trap), followed by the Fab> ports going up when mthca starts? Or is the problem that the Fab> boot driver doesn't know when the handoff is, and thus can't Fab> disable the device?After the kernel takes over, mthca will reset the HCA and of course the SM will have to bring the port back up. But at the point that mthca is loaded, the system typically won't have a hostname set. The kernel will need to have the HCA port active with the mthca driver running before it can mount root and get to /etc/sysconfig/network or wherever the hostname is set.
Maybe we could use some parameter passing between the boot OS and the post boot OS? Are there mechanisms to do that? For OpenSM case the node description is only used for informational purposes. I think it will be very confusing if a node with a user given description will show up with no description after it is rebooted. For the gen2 stack we could use the following "hack": OpenSM scans all nodes for their description every time it does a full sweep. So we could cause an extra sweep after each node description change by faking trap 144 (HCA port capability mask change) and sending it over. However, this is non standard and each SM can treat it differently. If we limit our solution to the common case: On a regular non disk-less machine is it possible to have the node description be set before the QP0 is physically UP? EZ _______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
