Bug 456352 - Defective AMI BIOS present in multiple LGA 775 motherboards by Foxconn, ASUS, and MSI causes ACPI failures
Summary: Defective AMI BIOS present in multiple LGA 775 motherboards by Foxconn, ASUS,...
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel
Version: 9
Hardware: All
OS: Linux
Target Milestone: ---
Assignee: Kernel Maintainer List
QA Contact: Fedora Extras Quality Assurance
Depends On:
TreeView+ depends on / blocked
Reported: 2008-07-23 02:43 UTC by Ryan Farmer
Modified: 2013-01-10 07:56 UTC (History)
1 user (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Last Closed: 2008-08-06 18:36:40 UTC
Type: ---

Attachments (Terms of Use)
Clip of kernel messages on boot, with shoddy Foxconn motherboard (39.59 KB, text/plain)
2008-07-23 02:43 UTC, Ryan Farmer
no flags Details
Clip from /var/log/messages with (39.02 KB, text/plain)
2008-07-23 03:53 UTC, Ryan Farmer
no flags Details

Description Ryan Farmer 2008-07-23 02:43:57 UTC
Description of problem:

Multiple ACPI problems and warnings on boot, PC will suspend and resume, but on
the next reboot, it hangs and requires a hard reset, Thermal Zone and Fan
control support also appear to be missing, I had to set the BIOS to take control
of the fans, or else they run at 100% at all times.

The website for this motherboard:


I am using latest BIOS, which seems to allow CPU Frequency scaling to function.

Version-Release number of selected component (if applicable):


Running a 2.6.26-git9 kernel appears to fix (some?) of this in that ACPI: Failed
To Attach Device messages are gone and PC suspends, resumes, and reboots, still
get checksum error relating to tbautils.

Behaviour is the same with ACPI 1, 2, or 3. 

How reproducible:

Steps to Reproduce:

1. Install Foxconn G33M or G33M-S motherboard.
2. Boot Fedora
3. Try a suspend, resume reboot.
Actual results:

Depdning on your kernel, either resume no worky, reboot no worky, and in all
cases, ugly kernel error messages.

Expected results:

Suspend, resume, reboot all worky, kernel gives no error messages.

Additional info:

I've attached a lot of all my kernel related messages.

I've attempted to ask Foxconn about this, and they recommended I remove all of
my RAM and see if the problem continues. (OK, not as useful as funny, but still)

Comment 1 Ryan Farmer 2008-07-23 02:43:57 UTC
Created attachment 312410 [details]
Clip of kernel messages on boot, with shoddy Foxconn motherboard

Comment 2 Ryan Farmer 2008-07-23 03:53:40 UTC
Created attachment 312417 [details]
Clip from /var/log/messages with

System log messages with kernel, I installed this kernel because
it appears to be more verbose.

Comment 3 Matthew Garrett 2008-07-27 00:06:44 UTC
Linux correctly reports that the table has an incorrect checksum, but that
should be purely cosmetic. The only obvious remaining bug is that the machine
fails to reboot after a suspend to RAM? Does booting with


as a kernel command option work around this?

Comment 4 Matthew Garrett 2008-07-27 02:58:49 UTC
Man I suck at bugzilla. Reopening since I closed it by accident.

Comment 5 Ryan Farmer 2008-07-27 03:08:09 UTC
reboot-b has no effect.

I've sent you my ACPI dump.

Foxconn is apparently working on some kind of a patch, and has stated that they
will make an official announcement over why this happened on Monday.

They've been frantically working on getting the BIOS to boot up Linux and
correctly interface with it on their test machines.

From what I can tell an improved BIOS ROM is on the way?

I will post link to the ROM when they put it on their site so we can close this bug.

Comment 6 Matthew Garrett 2008-07-27 03:16:51 UTC
Just to confirm, you tried 


and not


, right? It needs to be passed as a kernel option from grub.

Comment 7 Ryan Farmer 2008-07-27 03:53:18 UTC
yes, I put reboot=b in menu.lst, same behavior

Get an error on resume that I couldn't make out before.

Something like:

ATA 6 revalidation error.

Comment 8 Ryan Farmer 2008-07-27 07:27:58 UTC
Just to go over what Matthew has stated on his blog, paraphrasing:

"There's no way altering the DSDT fixes any of the errors, especially not the

It does indeed fix all of that, or at least no errors in my log when I change to
_OSI and Store Zero (What Vista and XP get), and remove the _OS reference to
"Linux", so I don't know, I'm even more uncomfortable with this, he says this
shouldn't happen, but it does.

Comment 9 Ryan Farmer 2008-07-29 08:10:45 UTC

I have been in touch with Foxconn several times today, this bug also affects:

Asus P5K-E
MSI P965 series

It is the BIOS that is defective, not the chipset.

American Megatrends (AMI) shipped the defective BIOS, if yours is an Award BIOS,
this is not an issue.


I've been told to expect the Foxconn fix within a matter of days, users of ASUS
or MSI motherboards are encouraged to contact them for the fix and reference
that Foxconn has confirmed that the AMI BIOS is defective.

Comment 10 Ryan Farmer 2008-07-29 10:36:03 UTC
I would also like to say before anyone asks that despite what original post
says, I never was able to get it to reboot properly after suspend no matter what
I did, I was deliriously tired (and typing angry) when I wrote this up, trying
to figure out what was wrong, if the problem would go away with the newest kernels.

Comment 11 Ryan Farmer 2008-08-01 21:19:52 UTC
BIOS fix for Foxconn motherboards very soon, I am beta testing a new BIOS that
resolves all the party poopers as long as you're using kernel 2.6.26

MSI and ASUS appear to want to have nothing to do with this, thats my hunch
anyway, kindly pester them if you're their customer would be my suggestion, but
we all know that would make me a hypocrite. :P

Comment 12 Matthew Garrett 2008-08-06 14:49:41 UTC
I've been in communication with the AMI engineers. The issue appears to be that Linux doesn't clear the WAK_STS flag on resume, which causes the BIOS to think that a reset is actually a resume. I've got a test board now, so will check this later today. Should be a trivial patch.

Comment 13 Matthew Garrett 2008-08-06 18:36:40 UTC
Fix applied to Rawhide and sent upstream. Should land within the next day or so. Tested successfully on a G33M board.

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