Bug 1120 - Kernel hangs when APIC/ACPI enabled
Summary: Kernel hangs when APIC/ACPI enabled
Status: CLOSED CODE_FIX
Alias: None
Product: ACPI
Classification: Unclassified
Component: Config-Interrupts (show other bugs)
Hardware: i386 Linux
: P2 high
Assignee: Len Brown
URL:
Keywords:
: 1245 (view as bug list)
Depends on:
Blocks: 845
  Show dependency tree
 
Reported: 2003-08-17 09:06 UTC by Jonathan Heaney
Modified: 2004-03-03 22:00 UTC (History)
2 users (show)

See Also:
Kernel Version: 2.6.0-test3-bk3
Subsystem:
Regression: ---
Bisected commit-id:


Attachments
Output of acpidmp (17.38 KB, application/x-gzip)
2003-08-20 15:11 UTC, Jonathan Heaney
Details
dmidecode output (2.61 KB, application/x-gzip)
2003-08-20 15:12 UTC, Jonathan Heaney
Details

Description Jonathan Heaney 2003-08-17 09:06:07 UTC
Distribution: Gentoo x86
Hardware Environment: AMD KT266A/AthlonXP
Software Environment: GCC-3.2.3
Problem Description: System locks up at "mice: PS/2 mouse device common for all
mice" during kernel boot.  Problem first appeared in 2.6.0-test3-bk3 (_may_ have
been in bk2, I went from bk2->bk3). Unresolved as of bk5.

During boot (prior to lockup), these lines appear, that I have never seen before
with any other kernels:

PCI VIA IRQ Fixup for 0000:00:11:2 from 15 to 11
PCI VIA IRQ Fixup for 0000:00:11:3 from 15 to 11

These are for the 2 USB headers that I've enabled.

I use the Bios to force the USB devices to IRQ 15 (which they share with my
Intel PCI Ethernet card).  I have an all-SCSI system, neither IDE port is
enabled in Bios so IRQ 15 is free.  IRQ 11 is vga card.  This setup has always
worked (had Linux on it since the box was built).

The only ACPI feature I'm using is Button.

This may be similar to bug 1115
http://bugzilla.kernel.org/show_bug.cgi?id=1115

And should block bug 845.
Comment 1 Jonathan Heaney 2003-08-17 09:07:08 UTC
Added block to bug 845
Comment 2 Jonathan Heaney 2003-08-17 14:42:35 UTC
I really should be a bit more careful when filing bug reports...

1.  System is VIA KT266A board, not AMD.
2.  I went from bk1->bk3, NOT bk2->bk3, and from bk3->bk5
3.  My mouse is a USB mouse, but USB stuff is compiled as modules.
Comment 3 Jonathan Heaney 2003-08-20 08:02:14 UTC
This bug is still present in 2.6.0-test3-bk6 and bk7.
Comment 4 Len Brown 2003-08-20 10:27:44 UTC
Yes, we're having problems with some via chips sets.

Please confirm that cmdline "pci=noacpi" boots.
If it doesn't, try "acpi=off".

Can you attache the output of dmidecode?
http://www.nongnu.org/dmidecode/

Also, if you're able to boot with ACPI enabled, attach a copy 
of /proc/acpi/dsdt  If you can't boot with ACPI enabled, then attach a copy of 
acpidmp output. 
http://developer.intel.com/technology/iapc/acpi/downloads/pmtools-
20010730.tar.gz
Comment 5 Jonathan Heaney 2003-08-20 15:11:19 UTC
Created attachment 680 [details]
Output of acpidmp

As requested...

I've not tried the kernel boot parameters method for disabling acpi, I just
rebuilt my kernel without APIC or ACPI.  When it boots I don't get the PCI VIA
IRQ Fixup lines for my USB ports.
Comment 6 Jonathan Heaney 2003-08-20 15:12:17 UTC
Created attachment 681 [details]
dmidecode output

Again, as requested.
Comment 7 Jonathan Heaney 2003-08-23 08:36:43 UTC
Just tried 2.6.0-test4, still the same problem, but I tried adding 'pci=noacpi'
boot option as advised, and the kernel boots OK with no PCI VIA IRQ Fixup messages.
Comment 8 Nicolas Mailhot 2003-08-31 14:14:18 UTC
Please check if bug #10 fix works for you
Comment 9 Greg Kroah-Hartman 2003-09-18 11:32:54 UTC
*** Bug 1245 has been marked as a duplicate of this bug. ***
Comment 10 Jonathan Heaney 2003-09-24 17:36:03 UTC
I've tried the via patch indicated in the other threads with kernel
2.6.0-test5-bk8, but it still will not boot unless I use the 'pci=noacpi' option.

btw Mobo is an Epox 8KHA+ with KT266A chipset.
Comment 11 Shaohua 2003-11-19 19:05:20 UTC
any chance to get the dmesg through serial port?

>PCI VIA IRQ Fixup for 0000:00:11:2 from 15 to 11
wield, though your IRQ15 is free, ACPI will not use it. I guess your box 
didn't use ACPI as IRQ router.
Comment 12 Jonathan Heaney 2003-11-20 05:03:13 UTC
Well I thought I'd been keeping an eye on the changelogs, but somewhere along
the line this seems to have been fixed without me noticing.  I'm using the
mm-sources kernel (2.4.60-test9 based), and it boots fine without the pci=noacpi
option.  dmesg output shows "PCI: Using ACPI for IRQ routing".

There was a new Bios released for my motherboard at the weekend - the only fix
listed for it is w.r.t. the AMD Applebred CPU though, so I don't know if this
has anything to do with it.

Anyway, it works now, so I'm not touching anything :-)

Note You need to log in before you can comment on or make changes to this bug.