2006-10-21 23:18:41

by john stultz

[permalink] [raw]
Subject: PAE broken on Thinkpad

Yea. So I know I probably shouldn't run a PAE kernel on my 1Gig laptop,
but in trying to do so I found it won't boot.

All I get on the screen is this (hand copied)
Uncompressing Linux... Ok, booting the kernel.










Int 14: CR2 c1000000 err 00000002 EIP c065d950 CS 00000060 flags 00010006
Stack: c08b3000 00038000 00001000 00000020 c068f5c0 ffffffc0 fffffc2e 00000001


>From gdb that EIP looks to be in __alloc_bootmem_core (string.h:372)

.config attached.

thanks
-john


Attachments:
laptop.config (37.90 kB)

2006-10-22 01:52:51

by Arjan van de Ven

[permalink] [raw]
Subject: Re: PAE broken on Thinkpad

On Sat, 2006-10-21 at 16:18 -0700, john stultz wrote:
> Yea. So I know I probably shouldn't run a PAE kernel on my 1Gig laptop,
> but in trying to do so I found it won't boot.


which CPU do you have? Not all laptop processors support PAE at all...
(for example the pentiumM generations before NX was added)

check /proc/cpuinfo the flags line to see if "pae" is there


--
if you want to mail me at work (you don't), use arjan (at) linux.intel.com
Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org

2006-10-22 02:01:08

by Andi Kleen

[permalink] [raw]
Subject: Re: PAE broken on Thinkpad

Arjan van de Ven <[email protected]> writes:

> On Sat, 2006-10-21 at 16:18 -0700, john stultz wrote:
> > Yea. So I know I probably shouldn't run a PAE kernel on my 1Gig laptop,
> > but in trying to do so I found it won't boot.
>
>
> which CPU do you have? Not all laptop processors support PAE at all...
> (for example the pentiumM generations before NX was added)

It shouldn't have crashed in bootmem then, just paniced early.

-Andi

2006-10-22 02:01:38

by Andi Kleen

[permalink] [raw]
Subject: Re: PAE broken on Thinkpad

john stultz <[email protected]> writes:

> Yea. So I know I probably shouldn't run a PAE kernel on my 1Gig laptop,
> but in trying to do so I found it won't boot.

You don't say what version?


> Int 14: CR2 c1000000 err 00000002 EIP c065d950 CS 00000060 flags 00010006
> Stack: c08b3000 00038000 00001000 00000020 c068f5c0 ffffffc0 fffffc2e 00000001
>
>
> >From gdb that EIP looks to be in __alloc_bootmem_core (string.h:372)

string.h??

-Andi

2006-10-22 02:30:43

by john stultz

[permalink] [raw]
Subject: Re: PAE broken on Thinkpad

On Sun, 2006-10-22 at 04:01 +0200, Andi Kleen wrote:
> john stultz <[email protected]> writes:
>
> > Yea. So I know I probably shouldn't run a PAE kernel on my 1Gig laptop,
> > but in trying to do so I found it won't boot.
>
> You don't say what version?

Sorry, the current -git.

-john


2006-10-22 02:50:22

by Andi Kleen

[permalink] [raw]
Subject: Re: PAE broken on Thinkpad

On Sunday 22 October 2006 04:30, john stultz wrote:
> On Sun, 2006-10-22 at 04:01 +0200, Andi Kleen wrote:
> > john stultz <[email protected]> writes:
> >
> > > Yea. So I know I probably shouldn't run a PAE kernel on my 1Gig laptop,
> > > but in trying to do so I found it won't boot.
> >
> > You don't say what version?
>
> Sorry, the current -git.

Normally the early exception handler should print a backtrace, i wonder
why that didn't work.

Can you change the

static int current_ypos = 25

in arch/x86_64/kernel/early_printk.c to

static int current_ypos = 0

and see if that displays the backtrace?

-Andi

2006-10-22 22:03:14

by john stultz

[permalink] [raw]
Subject: Re: PAE broken on Thinkpad

On Sun, 2006-10-22 at 04:50 +0200, Andi Kleen wrote:
> On Sunday 22 October 2006 04:30, john stultz wrote:
> > On Sun, 2006-10-22 at 04:01 +0200, Andi Kleen wrote:
> > > john stultz <[email protected]> writes:
> > >
> > > > Yea. So I know I probably shouldn't run a PAE kernel on my 1Gig laptop,
> > > > but in trying to do so I found it won't boot.
> > >
> > > You don't say what version?
> >
> > Sorry, the current -git.
>
> Normally the early exception handler should print a backtrace, i wonder
> why that didn't work.
>
> Can you change the
>
> static int current_ypos = 25
>
> in arch/x86_64/kernel/early_printk.c to
>
> static int current_ypos = 0
>
> and see if that displays the backtrace?

Unfortunately not. :( I just get those two lines up at the top of the
screen.

thanks
-john

2006-10-22 22:19:29

by Andi Kleen

[permalink] [raw]
Subject: Re: PAE broken on Thinkpad


>
> Unfortunately not. :( I just get those two lines up at the top of the
> screen.

Have to debug it the hard way then. Can you add lots of
early_printks to let's say setup_arch and then around the
following calls in start_kernel and see how far it gets?

-Andi

2006-10-22 22:35:32

by Andi Kleen

[permalink] [raw]
Subject: Re: PAE broken on Thinkpad II

On Monday 23 October 2006 00:19, Andi Kleen wrote:
>
> >
> > Unfortunately not. :( I just get those two lines up at the top of the
> > screen.
>
> Have to debug it the hard way then. Can you add lots of
> early_printks to let's say setup_arch and then around the
> following calls in start_kernel and see how far it gets?

Or alternatively if some earlier kernel worked bisect it please

-Andi