> I wanted to add some clarification. > > We have two types of IB devices: > 1)Devices that can operate as an InfiniBand adapter only > 2)Devices that can operate as an InfiniBand adapter or as an Ethernet NIC > > As per the current implementation of OFED stack, the driver architecture > of #2 is very different from #1 because it needs to make sure InfiniBand > and Ethernet functions can share the device without interfering with > each other. > > I was thinking that we can fix /proc/interrupts issue for case#1 first > and worry about #2 later because the design to fix /proc/interrupts > for mlx4 case is going to be different and independent just as the > driver design is different and independent for the two cases today.
I disagree. A verbs consumer of mlx4 doesn't have to worry about the internal design of the driver being different from mthca, and I would hope that carries over to indentifying interrupts. It's much better for users if we can just come up with a solution that handles both of your cases at once, rather than an ad hoc solution for a subset of drivers. > Please let me know if you still think we need a common solution for > both cases mentioned above. Any suggestions at a high level for such > a common solution? I already suggested adding MSI-X vector information to /sys/devices/... to match the existing "irq" file there. That would allow userspace to figure out which interrupt belonged where. Jason's idea of adding the PCI device name to the interrupt name seems viable to me as well. - R. _______________________________________________ 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
