Re: Interrupt routing with nvidia nforce2

[Lars Nordlund <lars.nordlund%hem.utfors.se@localhost>]

On Tue, 10 Jun 2008 22:23:42 +0100
Andrew Doran <ad%netbsd.org@localhost> wrote:

> Hi,
> 
> I have a Shuttle SN45 v1 that uses the nforce2 chipset. All of the ACPI
> interrupt routing information for the only PCI slot on the board appears
> correct, and as best I can tell it appears to match the hardware config.
> 
> When using the ioapic, interrupts for the slot never arrive. I suspect that
> some of the interrupt lines may not be physically wired up on the board, but
> hopefully it is a software problem.
> 
> Does anyone else have an nforce2 chipset and do you see similar problems?

Yes, I see something like this on an i386 machine here. It has an
nforce2 chipset and it causes interrupt timeout problems for this
device (an extra PCI SATA controller since there is no onboard SATA on
this mobo):

pdcsata0 at pci1 dev 8 function 0
pdcsata0: Promise PDC40718 SATA300 controller (rev. 0x02)
pdcsata0: interrupting at irq 11
pdcsata0: bus-master DMA support present

rendering the system unusable without its harddisk. Toggling the APIC
flag in the BIOS (unsure about exact naming) makes the timeouts go away.

Linux does not suffer from this. OpenSUSE works regardless of what I
set the APIC mode to.

I *think* FreeBSD and DragonflyBSD (same bootloader?) has trouble
booting depending on this flag. I get an instant reboot before the
kernel is loaded. It was however some time since I last tried them on
this machine and it was before I knew about this APIC voodoo. I can try
this again if needed.

Windows works in both APIC settings. But.., there is a big but here; It
only works with the APIC setting used when it was installed. If I flip
the mode in BIOS I have to reinstall windows to get it to work in the
other APIC mode.. Otherwise it will just hang during boot.


Best regards,
        Lars Nordlund