Interrupt storm on acpi0 occurred after madt.c change

[John Baldwin]

On Saturday 31 January 2004 02:22 pm, Peter Schultz wrote:
> John Baldwin wrote:
> > On Friday 30 January 2004 11:40 am, Sangwoo Shim wrote:
> >>Hi.
> >>I've experienced interrupt storm on acpi0 after your v1.10 madt.c commit.
> >>(which is located at src/sys/i386/acpica/madt.c);
> >>What really makes me frustrating is this storm wasn't here before 1.10
> >>change, which is exactly stand against your commit log message. :-(
> >>I've recompiled kernel with v1.9 of madt.c and the problem didn't occur.
> >>
> >>I'm using ASUS P2B-D with two p3-650 processors.
> >>BIOS version is 1014 beta 3 (which is the last published one);
> >>
> >>My kernel config is stripped GENERIC, which includes
> >>option	CPU_FASTER_5X86_FPU
> >>option	CPU_ENABLE_SSE
> >>device	acpi		# Mine hangs with acpi loaded as module.
> >>
> >>Attached file is my acpidump -t -d output.
> >>Thanks.
> >
> > Can you provide a dmesg?  IRQ20 on your box is a PCI interrupt and PCI
> > interrupts are level/lo, not level/hi.  To top it off, ACPI states that
> > the SCI is supposed to always be level/lo.  You must have a very broken
> > BIOS to get a storm now.
>
> I have this problem as well, here's my verbose dmesg and ASL:
>
> http://bis.midco.net/pmes/dmesg.txt
> http://bis.midco.net/pmes/tyan_s1832dl.asl
>
> I've not reverted to v1.9 of madt.c, but I'm sure it would make all the
> difference it did for the others.

Ugh, does Linux work ok on this motherboard?  Unfortunately, I have seen 
BIOS's that program level/hi and some want level/lo, and apparently the 
tyan's want the broken level/hi, but I can't tell the difference as they both 
use the same exact MADT entry.  BIOS writers suck.

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org