EDD problem -- regression -- crash on booting by probing EDD

EDD problem -- regression -- crash on booting by probing EDD

[werner]

[-- Attachment #1: Type: text/plain, Size: 506 bytes --]

There is a regression in 2.6.25.   Booting stops with: Probing EDD (edd=off to disable) message.

It dont work the kernel boot parameter edd=off .  With pci=off the computer continues to boot, but later it dont find hard disks etc.

Thats happening on 2.6.25   definitive, but was verified to go back until 2.6.24-rc6-git8, but it wasnt yet at 2.6.24-rc6-git2 which worked correctly.


Werner Landgraf

SYS Linux

http://linux.softpedia.com/get/System/Operating-Systems/Linux-Distributions/SYS-34168.shtml

EDD problem -- regression -- crash on booting by probing EDD

[werner]

[-- Attachment #1: Type: text/plain, Size: 1029 bytes --]

The problem below , I continued to check, and it seems really be a kernel bug (inclusively that the option edd=off don't work)

Because of this, I had to release an update of my SYS distro DVD, with i=installation using the normal 2.6.25 kernel, and a=alternatively 2.4.24-rc6-git2 what dont have that problem.  ftp://ftp5.gwdg.de/pub/linux/install/sys/  

The problem happened with an ACER Aspire 7004WSMi laptop.   

Werner Landgraf


=====================================================================================
There is a regression in 2.6.25.   Booting stops with: Probing EDD (edd=off to disable) message.

It dont work the kernel boot parameter edd=off .  With pci=off the computer continues to boot, but later it dont find hard disks etc.

Thats happening on 2.6.25   definitive, but was verified to go back until 2.6.24-rc6-git8, but it wasnt yet at 2.6.24-rc6-git2 which worked correctly.


Werner Landgraf

SYS Linux

http://linux.softpedia.com/get/System/Operating-Systems/Linux-Distributions/SYS-34168.shtml

Re: EDD problem -- regression -- crash on booting by probing EDD

[Andrew Morton]

> On Sun, 20 Apr 2008 19:33:32 -0300 (GFT) "werner" <werner@copaya.yi.org> wrote:
> There is a regression in 2.6.25.

Rafael bcc'ed.

>   Booting stops with: Probing EDD (edd=off to disable) message.
> 
> It dont work the kernel boot parameter edd=off .  With pci=off the computer continues to boot, but later it dont find hard disks etc.

Tim and Matt cc'ed.

> Thats happening on 2.6.25   definitive, but was verified to go back until 2.6.24-rc6-git8, but it wasnt yet at 2.6.24-rc6-git2 which worked correctly.
> 

hm, that's a nice narrow window.  Unfortunately for the next few days I
can't personally get in and find out what we changed during it.

It would be useful if you could add `initcall_debug' to your kernel
boot command line, please. That will help us to confirm that the
bug is indeed in edd_init(), rather than in something which was
called later on.

Thanks.

Re: EDD problem -- regression -- crash on booting by probing EDD

[Matt Domsch]

On Tue, Apr 22, 2008 at 06:28:40AM -0700, Andrew Morton wrote:
> > On Sun, 20 Apr 2008 19:33:32 -0300 (GFT) "werner" <werner@copaya.yi.org> wrote:
> > There is a regression in 2.6.25.
> 
> Rafael bcc'ed.
> 
> >   Booting stops with: Probing EDD (edd=off to disable) message.
> > 
> > It dont work the kernel boot parameter edd=off .  With pci=off the computer continues to boot, but later it dont find hard disks etc.
> 
> Tim and Matt cc'ed.
> 
> > Thats happening on 2.6.25   definitive, but was verified to go back until 2.6.24-rc6-git8, but it wasnt yet at 2.6.24-rc6-git2 which worked correctly.
> > 
> 
> hm, that's a nice narrow window.  Unfortunately for the next few days I
> can't personally get in and find out what we changed during it.

I've scanned the diff between these, and can't see anything that would
obviously cause a problem.  Nothing that touches the EDD codepath
directly at least.  I'll be curious to see the additional initcall
debug data.

-- 
Matt Domsch
Linux Technology Strategist, Dell Office of the CTO
linux.dell.com & www.dell.com/linux

Re: EDD problem -- regression -- crash on booting by probing EDD

[werner]

[-- Attachment #1: Type: text/plain, Size: 4034 bytes --]

As I tested yesterday extensively, the problem wasn't present at 2.6.24-rc5-git6 (SYS Linux 0.17 install DVD) nor at 2.6.24-rc6-git2 (SYS 0.19) ,  but at 2.6.24-rc6-git8 (SYS 0.20-rc1 and 0.20-rc2) and 2.6.25 (0.21-r3).

The whole compiled packages of these kernels still are at copaya.yi.org/tgz , but I extracted  the important parts, see enclosed here.    The .tar.bz2 file contains  the config  and the protocoll of the compilation (inclusive warning / error messages). 

As you can see, the diff between the config of 2.6.24-rc6-git2 and -git8 is nothing.  That means: THE ERROR ENTERED INTO THE KERNEL BETWEEN THESE TWO GITs !!      

You may also analyze the error/warning messages. Please observe that the error message just at the beginning was already subject of another reclamation mine together with another person, and seems to have be corrected.

I inform still about the following:    As I make many experients and also install Linux often for firnds and neighbours, I see steady-steady some kernel errors, but have normally no time to track them, in the house of the persons.  That error here, I saw already several times, and never it helped the edd=off option.    Yesterday it happened that a friend brought a laptop (ACER Aspire 7004WSMi) to me, after he was unsuccessfull to install  SYS 0.21-r3 dvd with 2.6.25-rc7 . Then I tracked that problem during about 5 hours, inclusively trying a lot of different kernel options.  edd=off don't work at all, with no of the affected kernels.  With 2.6.24-rc6-git2 and earlier kernels the problem didnt appear, even without edd=off option, and with the later kernels it appears, even with edd=off .    With pci=off , also the recent kernels boots -- but this is useless because the system don't find any hard disk.  acpi=off or hwprobe=off have no influence on that problem.

Pls search that error and repair it !
**********  YOU SHOULD MAKE AN EXTRA KERNEL OPTION, WHAT IGNORES THE HARDWARE AND CONTINUES TO BOOT, like pci=off,  BUT WHAT DETECTS CD ROMS, HARD DISKS SATA / IDE / RAID , AND WHAT PERMITS TO USE THE COMPUTER NORMALLY AT LEAST IN VESA-FB OR FBDEV MODE !!!!! *************

As there has no other solution, I released last night SYS 0.21-r4 DVD , what contains both, the 2.6.25 , and alternatively 2.6.24-rc6-git2 , so that people can use them computer anyway / anyhow :    ftp://ftp5.gwdg.de/pub/linux/install/sys/  

I can put the suggested debug, generally, in the kernel command line of my SYS install DVD. However, the person gone with his laptop so that at the moment I cannot test it; I have to wait if the same problem occurs on another computer.  


Werner Landgraf

Linux  SYS   

http://linux.softpedia.com/get/System/Operating-Systems/Linux-Distributions/SYS-34168.shtml
 


annexed: 1 file
======================================================================================
On 22/Abr/2008 11:21 Matt Domsch wrote ..
> On Tue, Apr 22, 2008 at 06:28:40AM -0700, Andrew Morton wrote:
> > > On Sun, 20 Apr 2008 19:33:32 -0300 (GFT) "werner" <werner@copaya.yi.org> wrote:
> > > There is a regression in 2.6.25.
> > 
> > Rafael bcc'ed.
> > 
> > >   Booting stops with: Probing EDD (edd=off to disable) message.
> > > 
> > > It dont work the kernel boot parameter edd=off .  With pci=off the computer
> continues to boot, but later it dont find hard disks etc.
> > 
> > Tim and Matt cc'ed.
> > 
> > > Thats happening on 2.6.25   definitive, but was verified to go back until 2.6.24-rc6-git8,
> but it wasnt yet at 2.6.24-rc6-git2 which worked correctly.
> > > 
> > 
> > hm, that's a nice narrow window.  Unfortunately for the next few days I
> > can't personally get in and find out what we changed during it.
> 
> I've scanned the diff between these, and can't see anything that would
> obviously cause a problem.  Nothing that touches the EDD codepath
> directly at least.  I'll be curious to see the additional initcall
> debug data.
> 
> -- 
> Matt Domsch
> Linux Technology Strategist, Dell Office of the CTO
> linux.dell.com & www.dell.com/linux

[-- Attachment #2: YYY.tar.bz2 --]
[-- Type: application/x-tbz, Size: 224838 bytes --]

Re: EDD problem -- regression -- crash on booting by probing EDD

[devzero]

Hi !

>As I tested yesterday extensively, the problem wasn't present at 2.6.24-rc5-git6 (SYS \
>Linux 0.17 install DVD) nor at 2.6.24-rc6-git2 (SYS 0.19) ,  but at 2.6.24-rc6-git8 \
>(SYS 0.20-rc1 and 0.20-rc2) and 2.6.25 (0.21-r3).

any chance to git-bisect that further, to find the single commit which broke your kernel?

regards
roland



List:       linux-kernel
Subject:    Re: EDD problem -- regression -- crash on booting by probing EDD
From:       "werner" <werner () copaya ! yi ! org>
Date:       2008-04-23 0:20:44
Message-ID: 1208910044.20701 () copaya ! yi ! org
[Download message RAW]

As I tested yesterday extensively, the problem wasn't present at 2.6.24-rc5-git6 (SYS \
Linux 0.17 install DVD) nor at 2.6.24-rc6-git2 (SYS 0.19) ,  but at 2.6.24-rc6-git8 \
(SYS 0.20-rc1 and 0.20-rc2) and 2.6.25 (0.21-r3).

The whole compiled packages of these kernels still are at copaya.yi.org/tgz , but I \
extracted  the important parts, see enclosed here.    The .tar.bz2 file contains  the \
config  and the protocoll of the compilation (inclusive warning / error messages). 

As you can see, the diff between the config of 2.6.24-rc6-git2 and -git8 is nothing.  \
That means: THE ERROR ENTERED INTO THE KERNEL BETWEEN THESE TWO GITs !!      

You may also analyze the error/warning messages. Please observe that the error \
message just at the beginning was already subject of another reclamation mine \
together with another person, and seems to have be corrected.

I inform still about the following:    As I make many experients and also install \
Linux often for firnds and neighbours, I see steady-steady some kernel errors, but \
have normally no time to track them, in the house of the persons.  That error here, I \
saw already several times, and never it helped the edd=off option.    Yesterday it \
happened that a friend brought a laptop (ACER Aspire 7004WSMi) to me, after he was \
unsuccessfull to install  SYS 0.21-r3 dvd with 2.6.25-rc7 . Then I tracked that \
problem during about 5 hours, inclusively trying a lot of different kernel options.  \
edd=off don't work at all, with no of the affected kernels.  With 2.6.24-rc6-git2 and \
earlier kernels the problem didnt appear, even without edd=off option, and with the \
later kernels it appears, even with edd=off .    With pci=off , also the recent \
kernels boots -- but this is useless because the system don't find any hard disk.  \
acpi=off or hwprobe=off have no influence on that problem.

Pls search that error and repair it !
**********  YOU SHOULD MAKE AN EXTRA KERNEL OPTION, WHAT IGNORES THE HARDWARE AND \
CONTINUES TO BOOT, like pci=off,  BUT WHAT DETECTS CD ROMS, HARD DISKS SATA / IDE / \
RAID , AND WHAT PERMITS TO USE THE COMPUTER NORMALLY AT LEAST IN VESA-FB OR FBDEV \
MODE !!!!! *************

As there has no other solution, I released last night SYS 0.21-r4 DVD , what contains \
both, the 2.6.25 , and alternatively 2.6.24-rc6-git2 , so that people can use them \
computer anyway / anyhow :    ftp://ftp5.gwdg.de/pub/linux/install/sys/  

I can put the suggested debug, generally, in the kernel command line of my SYS \
install DVD. However, the person gone with his laptop so that at the moment I cannot \
test it; I have to wait if the same problem occurs on another computer.  


_________________________________________________________________________
In 5 Schritten zur eigenen Homepage. Jetzt Domain sichern und gestalten! 
Nur 3,99 EUR/Monat! http://www.maildomain.web.de/?mc=021114

Re: EDD problem -- regression -- crash on booting by probing EDD

[Randy Dunlap]

On Tue, 22 Apr 2008 21:20:44 -0300 (GFT) werner wrote:

> As I tested yesterday extensively, the problem wasn't present at 2.6.24-rc5-git6 (SYS Linux 0.17 install DVD) nor at 2.6.24-rc6-git2 (SYS 0.19) ,  but at 2.6.24-rc6-git8 (SYS 0.20-rc1 and 0.20-rc2) and 2.6.25 (0.21-r3).
> 
> The whole compiled packages of these kernels still are at copaya.yi.org/tgz , but I extracted  the important parts, see enclosed here.    The .tar.bz2 file contains  the config  and the protocoll of the compilation (inclusive warning / error messages). 
> 
> As you can see, the diff between the config of 2.6.24-rc6-git2 and -git8 is nothing.  That means: THE ERROR ENTERED INTO THE KERNEL BETWEEN THESE TWO GITs !!      

Hi,
I'm a little confused.

Is this supposed to be "between 2.6.25-rc6-git2 and 2.6.25-rc6-git8"?
You say 2.6.24-rc6-git*.  I just want to make sure of which patch
interval we are looking at.

And the shouting (all CAPS) isn't helping.

> You may also analyze the error/warning messages. Please observe that the error message just at the beginning was already subject of another reclamation mine together with another person, and seems to have be corrected.
> 
> I inform still about the following:    As I make many experients and also install Linux often for firnds and neighbours, I see steady-steady some kernel errors, but have normally no time to track them, in the house of the persons.  That error here, I saw already several times, and never it helped the edd=off option.    Yesterday it happened that a friend brought a laptop (ACER Aspire 7004WSMi) to me, after he was unsuccessfull to install  SYS 0.21-r3 dvd with 2.6.25-rc7 . Then I tracked that problem during about 5 hours, inclusively trying a lot of different kernel options.  edd=off don't work at all, with no of the affected kernels.  With 2.6.24-rc6-git2 and earlier kernels the problem didnt appear, even without edd=off option, and with the later kernels it appears, even with edd=off .    With pci=off , also the recent kernels boots -- but this is useless because the system don't find any hard disk.  acpi=off or hwprobe=off have no influence on that problem.
> 
> Pls search that error and repair it !
> **********  YOU SHOULD MAKE AN EXTRA KERNEL OPTION, WHAT IGNORES THE HARDWARE AND CONTINUES TO BOOT, like pci=off,  BUT WHAT DETECTS CD ROMS, HARD DISKS SATA / IDE / RAID , AND WHAT PERMITS TO USE THE COMPUTER NORMALLY AT LEAST IN VESA-FB OR FBDEV MODE !!!!! *************
> 
> As there has no other solution, I released last night SYS 0.21-r4 DVD , what contains both, the 2.6.25 , and alternatively 2.6.24-rc6-git2 , so that people can use them computer anyway / anyhow :    ftp://ftp5.gwdg.de/pub/linux/install/sys/  
> 
> I can put the suggested debug, generally, in the kernel command line of my SYS install DVD. However, the person gone with his laptop so that at the moment I cannot test it; I have to wait if the same problem occurs on another computer.  
> 
> 
> Werner Landgraf
> 
> Linux  SYS   
> 
> http://linux.softpedia.com/get/System/Operating-Systems/Linux-Distributions/SYS-34168.shtml
>  
> 
> 
> annexed: 1 file
> ======================================================================================
> On 22/Abr/2008 11:21 Matt Domsch wrote ..
> > On Tue, Apr 22, 2008 at 06:28:40AM -0700, Andrew Morton wrote:
> > > > On Sun, 20 Apr 2008 19:33:32 -0300 (GFT) "werner" <werner@copaya.yi.org> wrote:
> > > > There is a regression in 2.6.25.
> > > 
> > > Rafael bcc'ed.
> > > 
> > > >   Booting stops with: Probing EDD (edd=off to disable) message.
> > > > 
> > > > It dont work the kernel boot parameter edd=off .  With pci=off the computer
> > continues to boot, but later it dont find hard disks etc.
> > > 
> > > Tim and Matt cc'ed.
> > > 
> > > > Thats happening on 2.6.25   definitive, but was verified to go back until 2.6.24-rc6-git8,
> > but it wasnt yet at 2.6.24-rc6-git2 which worked correctly.
> > > > 
> > > 
> > > hm, that's a nice narrow window.  Unfortunately for the next few days I
> > > can't personally get in and find out what we changed during it.
> > 
> > I've scanned the diff between these, and can't see anything that would
> > obviously cause a problem.  Nothing that touches the EDD codepath
> > directly at least.  I'll be curious to see the additional initcall
> > debug data.
> > 
> > -- 
> > Matt Domsch
> > Linux Technology Strategist, Dell Office of the CTO
> > linux.dell.com & www.dell.com/linux
> 


---
~Randy

Re: EDD problem -- regression -- crash on booting by probing EDD

[Randy Dunlap]

On Tue, 22 Apr 2008 21:20:44 -0300 (GFT) werner wrote:

> As I tested yesterday extensively, the problem wasn't present at 2.6.24-rc5-git6 (SYS Linux 0.17 install DVD) nor at 2.6.24-rc6-git2 (SYS 0.19) ,  but at 2.6.24-rc6-git8 (SYS 0.20-rc1 and 0.20-rc2) and 2.6.25 (0.21-r3).
> 
> The whole compiled packages of these kernels still are at copaya.yi.org/tgz , but I extracted  the important parts, see enclosed here.    The .tar.bz2 file contains  the config  and the protocoll of the compilation (inclusive warning / error messages). 
> 
> As you can see, the diff between the config of 2.6.24-rc6-git2 and -git8 is nothing.  That means: THE ERROR ENTERED INTO THE KERNEL BETWEEN THESE TWO GITs !!      
> 
> You may also analyze the error/warning messages. Please observe that the error message just at the beginning was already subject of another reclamation mine together with another person, and seems to have be corrected.
> 
> I inform still about the following:    As I make many experients and also install Linux often for firnds and neighbours, I see steady-steady some kernel errors, but have normally no time to track them, in the house of the persons.  That error here, I saw already several times, and never it helped the edd=off option.    Yesterday it happened that a friend brought a laptop (ACER Aspire 7004WSMi) to me, after he was unsuccessfull to install  SYS 0.21-r3 dvd with 2.6.25-rc7 . Then I tracked that problem during about 5 hours, inclusively trying a lot of different kernel options.  edd=off don't work at all, with no of the affected kernels.  With 2.6.24-rc6-git2 and earlier kernels the problem didnt appear, even without edd=off option, and with the later kernels it appears, even with edd=off .    With pci=off , also the recent kernels boots -- but this is useless because the system don't find any hard disk.  acpi=off or hwprobe=off have no influence on that problem.
> 
> Pls search that error and repair it !
> **********  YOU SHOULD MAKE AN EXTRA KERNEL OPTION, WHAT IGNORES THE HARDWARE AND CONTINUES TO BOOT, like pci=off,  BUT WHAT DETECTS CD ROMS, HARD DISKS SATA / IDE / RAID , AND WHAT PERMITS TO USE THE COMPUTER NORMALLY AT LEAST IN VESA-FB OR FBDEV MODE !!!!! *************
> 
> As there has no other solution, I released last night SYS 0.21-r4 DVD , what contains both, the 2.6.25 , and alternatively 2.6.24-rc6-git2 , so that people can use them computer anyway / anyhow :    ftp://ftp5.gwdg.de/pub/linux/install/sys/  
> 
> I can put the suggested debug, generally, in the kernel command line of my SYS install DVD. However, the person gone with his laptop so that at the moment I cannot test it; I have to wait if the same problem occurs on another computer.  

We need that initcall_debug information.  And it would be helpful
to have a list of hardware that is installed on the problem system:
lspci -v


> Werner Landgraf
> 
> Linux  SYS   
> 
> http://linux.softpedia.com/get/System/Operating-Systems/Linux-Distributions/SYS-34168.shtml
>  
> 
> 
> annexed: 1 file
> ======================================================================================
> On 22/Abr/2008 11:21 Matt Domsch wrote ..
> > On Tue, Apr 22, 2008 at 06:28:40AM -0700, Andrew Morton wrote:
> > > > On Sun, 20 Apr 2008 19:33:32 -0300 (GFT) "werner" <werner@copaya.yi.org> wrote:
> > > > There is a regression in 2.6.25.
> > > 
> > > Rafael bcc'ed.
> > > 
> > > >   Booting stops with: Probing EDD (edd=off to disable) message.
> > > > 
> > > > It dont work the kernel boot parameter edd=off .  With pci=off the computer
> > continues to boot, but later it dont find hard disks etc.
> > > 
> > > Tim and Matt cc'ed.
> > > 
> > > > Thats happening on 2.6.25   definitive, but was verified to go back until 2.6.24-rc6-git8,
> > but it wasnt yet at 2.6.24-rc6-git2 which worked correctly.
> > > > 
> > > 
> > > hm, that's a nice narrow window.  Unfortunately for the next few days I
> > > can't personally get in and find out what we changed during it.
> > 
> > I've scanned the diff between these, and can't see anything that would
> > obviously cause a problem.  Nothing that touches the EDD codepath
> > directly at least.  I'll be curious to see the additional initcall
> > debug data.
> > 
> > -- 
> > Matt Domsch
> > Linux Technology Strategist, Dell Office of the CTO
> > linux.dell.com & www.dell.com/linux
> 


---
~Randy

Re: EDD problem -- regression -- crash on booting by probing EDD

[Matt Domsch]

On Wed, Apr 23, 2008 at 08:35:47AM -0700, Randy Dunlap wrote:
> On Tue, 22 Apr 2008 21:20:44 -0300 (GFT) werner wrote:
> 
> > As I tested yesterday extensively, the problem wasn't present at 2.6.24-rc5-git6 (SYS Linux 0.17 install DVD) nor at 2.6.24-rc6-git2 (SYS 0.19) ,  but at 2.6.24-rc6-git8 (SYS 0.20-rc1 and 0.20-rc2) and 2.6.25 (0.21-r3).
> > 
> > The whole compiled packages of these kernels still are at copaya.yi.org/tgz , but I extracted  the important parts, see enclosed here.    The .tar.bz2 file contains  the config  and the protocoll of the compilation (inclusive warning / error messages). 
> > 
> > As you can see, the diff between the config of 2.6.24-rc6-git2 and -git8 is nothing.  That means: THE ERROR ENTERED INTO THE KERNEL BETWEEN THESE TWO GITs !!      
> 
> Hi,
> I'm a little confused.
> 
> Is this supposed to be "between 2.6.25-rc6-git2 and 2.6.25-rc6-git8"?
> You say 2.6.24-rc6-git*.  I just want to make sure of which patch
> interval we are looking at.

It's definitely .24 given the dates for the SYS releases he notes. The
diffstat between the two is relatively short [1], but I can't see
anything that can be related.  This code runs so early, during real
mode startup, that my only suspects are:

 arch/x86/kernel/head_32.S                      |    2 +-
 arch/x86/kernel/smpboot_64.c                   |    2 +-
 include/asm-generic/tlb.h                      |    3 -
 include/asm-x86/byteorder.h                    |    4 +-
 include/asm-x86/msr.h                          |   74 +++++++++--------



[1] http://domsch.com/linux/edd-bug-2008-04-26.patch

but none of those are obviously the culprit to me...

Werner, the initcall_debug trace would be most helpful; without that
I'm not sure how much progress we can make on this.

Thanks,
Matt

-- 
Matt Domsch
Linux Technology Strategist, Dell Office of the CTO
linux.dell.com & www.dell.com/linux

Re: EDD problem -- regression -- crash on booting by probing EDD

[Randy Dunlap]

On Sat, 26 Apr 2008 18:17:04 -0500 Matt Domsch wrote:

> On Wed, Apr 23, 2008 at 08:35:47AM -0700, Randy Dunlap wrote:
> > On Tue, 22 Apr 2008 21:20:44 -0300 (GFT) werner wrote:
> > 
> > > As I tested yesterday extensively, the problem wasn't present at 2.6.24-rc5-git6 (SYS Linux 0.17 install DVD) nor at 2.6.24-rc6-git2 (SYS 0.19) ,  but at 2.6.24-rc6-git8 (SYS 0.20-rc1 and 0.20-rc2) and 2.6.25 (0.21-r3).
> > > 
> > > The whole compiled packages of these kernels still are at copaya.yi.org/tgz , but I extracted  the important parts, see enclosed here.    The .tar.bz2 file contains  the config  and the protocoll of the compilation (inclusive warning / error messages). 
> > > 
> > > As you can see, the diff between the config of 2.6.24-rc6-git2 and -git8 is nothing.  That means: THE ERROR ENTERED INTO THE KERNEL BETWEEN THESE TWO GITs !!      
> > 
> > Hi,
> > I'm a little confused.
> > 
> > Is this supposed to be "between 2.6.25-rc6-git2 and 2.6.25-rc6-git8"?
> > You say 2.6.24-rc6-git*.  I just want to make sure of which patch
> > interval we are looking at.
> 
> It's definitely .24 given the dates for the SYS releases he notes. The
> diffstat between the two is relatively short [1], but I can't see
> anything that can be related.  This code runs so early, during real
> mode startup, that my only suspects are:
> 
>  arch/x86/kernel/head_32.S                      |    2 +-
>  arch/x86/kernel/smpboot_64.c                   |    2 +-
>  include/asm-generic/tlb.h                      |    3 -
>  include/asm-x86/byteorder.h                    |    4 +-
>  include/asm-x86/msr.h                          |   74 +++++++++--------
> 
> 
> 
> [1] http://domsch.com/linux/edd-bug-2008-04-26.patch
> 
> but none of those are obviously the culprit to me...

Yeah, there are also some IDE changes in that -rc2..-rc8 range.
I'm wondering if any of them could interfere.

> Werner, the initcall_debug trace would be most helpful; without that
> I'm not sure how much progress we can make on this.

That should be helpful.

---
~Randy