On Tue, Jan 11, 2005 at 04:37:38PM -0800, Grant Grundler wrote: > I'll try to get some netperf numbers for IPoIB next (with msi_x=1)...
I forgot to mention why with "msi_x=1" and not with "msi=1": ionize:~# modprobe ib_mthca msi_x=1 ib_mthca: Mellanox InfiniBand HCA driver v0.06-pre (November 8, 2004) ib_mthca: Initializing Mellanox Technology MT23108 InfiniHost (0000:81:00.0) GSI 60 (level, low) -> CPU 0 (0x0000) vector 63 ACPI: PCI interrupt 0000:81:00.0[A] -> GSI 60 (level, low) -> IRQ 63 MSI INIT SUCCESS ionize:~# ... [ mess with tg3 driver for a bit ] ionize:/home/grundler# rmmod ib_mthca ionize:/home/grundler# modprobe ib_mthca msi=1 ib_mthca: Mellanox InfiniBand HCA driver v0.06-pre (November 8, 2004) ib_mthca: Initializing Mellanox Technology MT23108 InfiniHost (0000:81:00.0) ACPI: PCI interrupt 0000:81:00.0[A] -> GSI 60 (level, low) -> IRQ 63 Can't enable MSI. Device already had MSI-X vectors assigned ionize:/home/grundler# i.e. one can't switch from MSI-X to MSI for a particular device. Roland and I both suspect this is probably something to do with core MSI support. grant _______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
