RE: [ACPI] MS-6368L ACPI IRQ problem still in 2.4.21

[Grover, Andrew]

> From: Marek Michalkiewicz [mailto:[e-mail]] 
> long time ago I noticed a problem with the ACPI IRQ not working
> if it is _not_ shared with some other PCI IRQ.  The problem still
> exists in the 2.4.21 kernel, confirmed on two machines with the
> MSI MS-6368L motherboard (VIA PLE133 chipset).
> 
> I need ACPI for just one thing: to run "shutdown -h now" after the
> power button is pressed.  (The box is a server which usually has
> no keyboard connected.)
> 
> As I can see in /proc/interrupts, the BIOS usually allocates IRQ9
> for ACPI (not shared with anything else), and the IRQ9 counter is
> always zero.  Pressing the power button has no effect at all.
> 
> There is an easy workaround: in BIOS setup, set IRQ9 to "Legacy ISA"
> instead of "PCI/ISA PnP" so that ACPI gets some other IRQ, shared
> with some other PCI devices (in my case, IRQ11 is shared by: acpi,
> usb-uhci, usb-uhci, eth0).  Then the power button works fine.
> 
> Is this a known problem?  Should I complain to MSI (BIOS fix),
> or is this a Linux bug?  Any patches I should try?  This looks
> a bit unusual to me - one would expect problems if an IRQ _is_
> shared, and some broken hardware/driver doesn't like sharing...
> 
> If there is no known fix, perhaps the "Legacy ISA" workaround
> (which I discovered by accident) should be documented somewhere?

Is this an SMP machine?
What do the INT_SRC_OVR lines in the dmesg say?

Thanks -- Andy