Heh, yeah, that's what I used to do too (and still do if a camera isn't 
Ahh. APM. That does explain the strange EIP register values: we're jumping

into the BIOS, and the BIOS is doing something unexpected.
However, that doesn't really explain the oops, because I don't think

anything changed in APM from 2.6.22->23, and in particular, I don't think

it has anything to do with the thing that caused problems for PPC.
We did have some APM *detection* changes, and maybe APM wasn't even

detected for you before, or it was detected differently. That would be due

to the bootup changes, I'm Cc'ing Peter Anvin (and the kernel mailing

list, in case somebody else see a pattern to this).
Can you please

 - try with APM turned off (APM really shouldn't be useful on any machines

   built in the last ten years or so), just to verify that things work

   without APM.

 - send the bootup "dmesg" output and a machine description (and please

   keep people Cc'd - sending things just to me is a sure-fire way to get

   things dropped eventually, if only because I'm a lazy clutz).

 - what was the last kernel that worked (and if you can bisect the

   problem, that would likely help enormously)
Thanks,
		Linus

-

Both "dmesg" (before and after the fix) are attached.  Also, here is the=20

machine specs:
Gigabyte Main Board

Name	GA-K8NSC -939

Type / Chipset	Socket 939 / nVidia nForce3 250 Gb Chipset

Dimensions 	+++ 35 mm (58 mm radius) around the socket to the next capacito=

r=20

or other parts of the

mainboard - for further information see Fit for Fan.

The overall size is 305 x 245mm.

Slots	+++ 5x pci, 1x AGP (1.5V 4x/8x), 4x DDR PC2100 - PC3200 (max. 4GB RAM)

Connectors=09

+++ 1x Floppy, 4x UDMA133, 2x SATA150, 2x seriell, 1x parallel, 2x PS/2,

onBoard 10/100/1000 Mbps base-T Ethernet LAN (Marvell 8001),

8x USB 2.0/1.1 (4x internal but mounting bracket isn't included),

8-Channel AC'97 RealTek ALC850 onboard (1x int. CD in, 1x int. S/PDIF, ext.=

=20

Line in, ext. Mic, ext. Line out, int. optional connector for 8-Channel Aud=

io=20

Combo Kit, not included)

and 3x fan connectors (1x used for chipset fan).

Vcore	++++ BIOS between 0.80 Volt and 1.7 Volt in steps of 0.025 and 0.05=20

Volt / Jumper not available

Multiplier	++++ BIOS between 4x and 25x (max multi depends on used CPU) /=20

Jumper not available

=46SB	++++ BIOS between 200 and 455 MHz (AGP 66 - 100 MHz) / Jumper not=20

available

VIO/VDD	++++ VDDR +0.1 - +0.2 / VAGP +0.1 - +0.3 / V HT-Link +0.1 - +0.3=20
Home page:

http://www.gigabyte.com.tw/Products/Motherboard/Products_Overview.aspx?P...

uctID=3D1881
BIOS upgrade (F8)

http://www.gigabyte.com.tw/Support/Motherboard/BIOS_Model.aspx?ProductID...

1881
Also lspci -v
root@fosters:/home/kevin# lspci -v

00:00.0 Host bridge: nVidia Corporation nForce3 250Gb Host Bridge (rev a1)

        Flags: bus master, 66MHz, fast devsel, latency 0

        Memory at f4000000 (32-bit, prefetchable) [size=3D32M]

        Capabilities: [44] HyperTransport: Slave or Primary Interface

        Capabilities: [c0] AGP version 2.0
00:01.0 ISA bridge: nVidia Corporation nForce3 250Gb LPC Bridge (rev a2)

        Subsystem: Giga-byte Technology Unknown device 0c11

        Flags...
[ message continues ]

Ok, this is the real reason. 
The APM code does:
        if (PM_IS_ACTIVE()) {

                printk(KERN_NOTICE "apm: overridden by ACPI.\n");

                apm_info.disabled = 1;

                return -ENODEV;

        }
and in previous kernels that would notice that you have ACPI enabled, and

APM gets shut out, and you never see your buggy APM BIOS.
In 2.6.23, this apparently doesn't happen for some reason.
And I think I see the problem: it's a config change. You don't have

PM_LEGACY enabled. Your config file diff shows:
	-CONFIG_PM_LEGACY=y

	+# CONFIG_PM_LEGACY is not set
I suspect we should make CONFIG_APM either depend on, or select,

PM_LEGACY. But as far as I can see, nothing has actually changed in this

area in the kernel, and this bug has been there before - just your config

change made it appear. 
Rafael? Stephen? Opinions? I'd think that making APM depend on

CONFIG_PM_LEGACY is the right thing to do these days..
		Linus

-

Speaking as the author of
	[PATCH] move pm_register/etc. to CONFIG_PM_LEGACY, pm_legacy.h
I agree.  arch/i386/kernel/apm.c clearly requires

include/linux/pm_legacy.h and the legacy PM API.
I would vote for a dependency rather than select, but don't have any

strong feelings on the matter...
	Jeff
-

It should be a select since a dependency would make it needlessly hard 
cu

Adrian
-- 
       "Is there not promise of rain?" Ling Tan asked suddenly out

        of the darkness. There had been need of rain for many days.

       "Only a promise," Lao Er said.

                                       Pearl S. Buck - Dragon Seed
-

Well, my experience with selects is such that I'd rather avoid them in the

future, if possible ...
Greetings,

Rafael

-

I agree and yes, I'd vote for a dependency too.
Greetings,

Rafael

-

On Thu, Oct 11, 2007 at 02:42:01PM -0400, Jeff Garzik wrote:

 > Linus Torvalds wrote:

 > >

 > > And I think I see the problem: it's a config change. You don't have

 > > PM_LEGACY enabled. Your config file diff shows:

 > >

 > > 	-CONFIG_PM_LEGACY=y

 > > 	+# CONFIG_PM_LEGACY is not set

 > >

 > > I suspect we should make CONFIG_APM either depend on, or select,

 > > PM_LEGACY. But as far as I can see, nothing has actually changed in this

 > > area in the kernel, and this bug has been there before - just your config

 > > change made it appear.

 > >

 > > Rafael? Stephen? Opinions? I'd think that making APM depend on

 > > CONFIG_PM_LEGACY is the right thing to do these days..

 >

 > Speaking as the author of

 >

 > 	[PATCH] move pm_register/etc. to CONFIG_PM_LEGACY, pm_legacy.h

 >

 > I agree.  arch/i386/kernel/apm.c clearly requires

 > include/linux/pm_legacy.h and the legacy PM API.

 >

 > I would vote for a dependency rather than select, but don't have any

 > strong feelings on the matter...
gti revert 987d4613e52e4f655278265aabbcc69237018b1d

should do the trick.
	Dave
--

http://www.codemonkey.org.uk

-

Well, I did have exactly one bug report about APM, and that was a user

who had inadvertently flipped his CONFIG_APM setting (from on to off in

his particular case -- it was a 10-year-old machine without ACPI.)
	-hpa

-