Subject: Re: ACPI SCI causing 90k interrupts/sec?!
To: Thor Lancelot Simon <tls@rek.tjls.com>
From: Frank van der Linden <fvdl@netbsd.org>
List: port-i386
Date: 09/06/2003 20:15:58
On Fri, Sep 05, 2003 at 04:42:45PM -0400, Thor Lancelot Simon wrote:
> Since upgrading to 1.6W, some of my systems have been much slower.  I
> just realized that the ACPI SCI is interrupting 60,000 times per second
> on one of them and 90,000 times per second on another.  One system is
> an AMD760MPX-chipset motherboard, the other is a very similar AMD760MP
> chipset board.

This problem is fixed now, I added some code to make sure that the
SCI interrupt is always active-low, unless there is an ACPI interrupt
source override (only checked with MPACPI).

> On another note, given the current "systat vmstat" display on i386 it is
> quite difficult to conclusively say just _what_ is causing a given
> interrupt.  Systat prints out things like "ioapic0 pin 9" but the device
> probe lines print out "apicN pin M" and irq only, not which ioapic and
> which ioapic pin -- and ACPI doesn't even print that.

I changed that for all PCI devices, I've been meaning to do that for
a while. Unfortunately, it's more work to do it for ISA devices, since
the interrupt string is printed in the (MI) drivers itself there. I
didn't feel like implementing a new isa_intr_string function and
going through ISA device code to use it.

- Frank