Interrupt routing issue on Compaq nx6325

Joerg Wunsch j at
Thu Dec 14 15:39:19 PST 2006

As John Baldwin wrote:

> > What in particular caught my eye in the Linux messages is:

> > Dec 12 23:33:26 remi kernel: ENABLING IO-APIC IRQs
> > Dec 12 23:33:26 remi kernel: ..TIMER: vector=0x31 apic1=0 pin1=2 apic2=-1 pin2=-1
> > Dec 12 23:33:26 remi kernel: ..MP-BIOS bug: 8254 timer not connected to IO-APIC
> > Dec 12 23:33:26 remi kernel: ...trying to set up timer (IRQ0) through the 8259A ...  failed.
> > Dec 12 23:33:26 remi kernel: ...trying to set up timer as Virtual Wire IRQ... works.

> > I didn't quite follow the referred to discussion of the Linux
> > kernel hackers, is the above their workaround, or is that
> > something entirely different?

> That's something different because of the way IRQ0 doesn't
> consistently work when using APIC.

Ah OK.

> I thought you meant that you had a problem with PCI devices not
> working (i.e. an "interrupt routing problem").

As disabling the APIC made the machine work, together with that Linux
messages, I thought of an interrupt routing problem.  Turns out it's
rather a weird CPU problem then.  (But still, Linux gets around that.)

As Ariff Abdullah wrote:

(power consumption)

> Just ignore large part of it. Most of the workarounds (disabling
> hlt, selective hlt) there does not applicable anymore with my above
> patch which:
> 1) HLT enabled, 2) APIC enabled 3) Piggybacking APIC
> timer through RTC interrupt (hw.apic.idle_cpu=1)

OK, that's what I'm using now.

The machine has been working for the entire day quite fine, but when I
came back at night, it had lost about half an hour of system time.  I
seen Bruce reported it losing one second per lid switch activation,
but in my case, the machine was left untouched physically.  It has
only been used for heavy compile jobs (JDK, OpenOffice) remotely, but
with the lid closed.

cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL                        NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)

More information about the freebsd-acpi mailing list