If you have a system running the amd64 bsd.mp kernel could you please
try out the diff below. Please provide me with a dmesg.


----- Forwarded message from Brad <brad@comstyle.com> -----

Date: Sat, 14 Jan 2006 18:05:18 -0500
From: Brad <brad@comstyle.com>
To: hackers@openbsd.org
Subject: amd64 ioapic interrupt routing diff
User-Agent: Mutt/1.4.2i

Please test this diff out on systems running the amd64 MP kernel
whether it is truely MP or just a UP box.

I ran into a problem on one of krw's VIA systems, the following NetBSD
PR was very similar to his situation, though a different VIA chipset
and amd64 instead of i386. The interrupt code is shared between i386/amd64
on NetBSD so I expected the same behavior. This diff does not fully resolve
the issue on krw's system but it does allow his USB mouse to work for a very
short period of time before the system again freezes where as without this
it does not work at all.

http://www.netbsd.org/cgi-bin/query-...l?number=22728

I have other things I am working on but the changes are much too invasive for
the release and will require more extensive testing.


Fix for ioapic irq routing.

From NetBSD


Index: intr.c
================================================== =================
RCS file: /cvs/src/sys/arch/amd64/amd64/intr.c,v
retrieving revision 1.11
diff -u -p -r1.11 intr.c
--- intr.c 25 Sep 2005 20:48:18 -0000 1.11
+++ intr.c 11 Jan 2006 00:52:20 -0000
@@ -266,6 +264,14 @@ intr_allocate_slot(struct pic *pic, int
*/
if (legacy_irq != -1) {
ci = &cpu_info_primary;
+ /* must check for duplicate pic + pin first */
+ for (slot = 0 ; slot < MAX_INTR_SOURCES ; slot++) {
+ isp = ci->ci_isources[slot];
+ if (isp != NULL && isp->is_pic == pic &&
+ isp->is_pin == pin ) {
+ goto duplicate;
+ }
+ }
slot = legacy_irq;
isp = ci->ci_isources[slot];
if (isp == NULL) {
@@ -292,7 +298,7 @@ intr_allocate_slot(struct pic *pic, int
goto other;
}
}
-
+duplicate:
if (pic == &i8259_pic)
idtvec = ICU_OFFSET + legacy_irq;
else {

----- End forwarded message -----