Hunting for missing ACPI interrupt

[David Murray <dave@xxxxxxxxxxxxxxxxxx>]

Hi All,

I'm wondering if anyone out there can offer any hints for finding a 
missing interrupt.

I've got box here where the front-panel power button has never worked - 
it doesn't shut the system down.  I thought I'd look into it - how hard 
can it be to make an off switch work, I thought.  (People familiar with 
ACPI will be laughing out loud at this point.)

Anyway, it turns out the power button doesn't work because the ACPI 
interrupt never appears, at least according to vmstat -i.  Building the 
ACPI module with debug shows that it seems to be doing all the right 
things - it figures out where the registers are in i/o space, enables 
ACPI, installs a handler and unmasks the power button as an interrupt 
source.  And pressing the button *does* toggle the corresponding status 
bit, so the hardware really ought to be generating an interrupt at this 
point.

So, either I still have some ACPI (or BIOS or hardware) misconfiguration 
which means the interrupt isn't being generated as advertised, or I have 
some sort of interrupt handling misconfiguration which means the 
interrupt's not getting through to the handler.  Can anyone make any 
suggestions as to how I might diagnose which it is?

I did take a look through the FreeBSD ACPI mailing list (I'm assuming 
the Dragonfly code's still pretty similar) but found only lots of 
problems with interrupt storms, not interrupt droughts.

Thanks in advance!


--
David Murray