[openib-general] APIC error

[Jason Gunthorpe]

On Tue, Apr 04, 2006 at 02:04:43PM -0700, Makia Minich wrote:

> Using earlier BIOS versions, I don't see the issue at all, but on the latest
> BIOS versions I'm dead in the water.  So, I was hoping that someone else has
> seen this issue and perhaphs might know of a workaround.

You might want to check that the BIOS hasn't changed the APIC
configuration in some strange way between the two revs. The error code
you are getting is documented as 'that a message received by this APIC
was not accepted by this or any other APIC'. Which probably means the
encoded destinations bits in the MSI address don't match what the
APICs have been programmed for...

I'd suggest dumping APIC registers ID (0x20), LOG_DEST (0xD0),
and DEST_FORMAT (0xE0) for your two BIOS versions. [ie add some
printks to the verify_local_APIC function]. If they are different
then perhaps a patch the msi code could be devised.

Also it is probably worth capturing the MSI address with lspci -v
and comparing that between revs.

Do any IO-APIC based interrupts work (cat /proc/interrupts)?
The other thing you can try is to capture the IO APIC configuration
debug messages (ie the stuff from io_apic.c:print_IO_APIC) and see if
the debug output changes between bios revisions. That may give a clue
what is going on.

Good Luck,
Jason