[openib-general] ip over ib throughtput

[Andras.Horvath at cern.ch]

> That's what I would expect.  As far as I know, all FW before 3.3.2 has
> bugs with MSI and MSI-X.

well, my vendor :) tells me that 3.0.0 is the latest they have and
that's almost true since they've shipped some cards with 3.0.2. Anyway,
3.0.0 was pretty stable overnight in two HP rx2600's.

However, when I put a card in an rx4640 (four-way thing, same ZX1
chipset as the 2600 ) I got an oops upon modprobe. This is 2.6.10-mm2
(I've compared some diffs with 11-rc1 and nothing relevant seems to have
changed).

Here's the top of the oops (would you like me to send it all, more
details like kernel config etc? to the list, or private?)

TIA for any help,
Andras 

ACPI: PCI interrupt for device 0000:41:00.0 disabled
iosapic_unregister_intr(4294967295) unbalanced
Badness in iosapic_unregister_intr at arch/ia64/kernel/iosapic.c:636

Call Trace:
 [<a00000010000f280>] show_stack+0x80/0xa0
                                sp=e0000000064afbd0 bsp=e0000000064a11a0
 [<a00000010000f2d0>] dump_stack+0x30/0x60
                                sp=e0000000064afda0 bsp=e0000000064a1190
 [<a00000010004ac60>] iosapic_unregister_intr+0x340/0x3a0
                                sp=e0000000064afda0 bsp=e0000000064a1138
 [<a000000100009740>] acpi_unregister_gsi+0x20/0x40
                                sp=e0000000064afda0 bsp=e0000000064a1120
 [<a0000001002e7ae0>] acpi_pci_irq_disable+0x180/0x1c0
                                sp=e0000000064afda0 bsp=e0000000064a10e8
 [<a0000001004898a0>] pcibios_disable_device+0x20/0x40
                                sp=e0000000064afdb0 bsp=e0000000064a10d0
 [<a00000010028dd80>] pci_disable_device+0xc0/0xe0
                                sp=e0000000064afdb0 bsp=e0000000064a10a8
 [<a000000200192910>] mthca_init_one+0x230/0xd40 [ib_mthca]
                                sp=e0000000064afdc0 bsp=e0000000064a1050