Hibernation oops with 2.6.31 won't fit the screen - how to debug?

Hibernation oops with 2.6.31 won't fit the screen - how to debug?

[Ondrej Zary]

Hello,
hibernation worked great with 2.6.30 - I had over a month of uptime (it never 
worked so well before - always crashed after some time). But in 2.6.31, it's 
broken again - and completely. The first hibernation works but it always 
crashes on the second one.

The problem is that the oops is so long that it does not fit the screen - even 
1280x1024 is not enough. Scrolling does not work (as it used to do on 2.4 
kernels), netconsole does not catch it. So how can I debug this?

-- 
Ondrej Zary

Re: Hibernation oops with 2.6.31 won't fit the screen - how to debug?

[Frans Pop]

Ondrej Zary wrote:
> hibernation worked great with 2.6.30 - I had over a month of uptime (it
> never worked so well before - always crashed after some time). But in
> 2.6.31, it's broken again - and completely. The first hibernation works
> but it always crashes on the second one.
> 
> The problem is that the oops is so long that it does not fit the screen -
> even 1280x1024 is not enough. Scrolling does not work (as it used to do
> on 2.4 kernels), netconsole does not catch it. So how can I debug this?

One way to get at least an indication of what broke the hibernation is to 
run a bisection to find the change that broke hibernation. That could 
provide sufficient clues for a next step.

Would that be an option for you?

Cheers,
FJP

Re: Hibernation oops with 2.6.31 won't fit the screen - how to debug?

[Jiri Slaby]

On 09/19/2009 11:44 PM, Frans Pop wrote:
> Ondrej Zary wrote:
>> hibernation worked great with 2.6.30 - I had over a month of uptime (it
>> never worked so well before - always crashed after some time). But in
>> 2.6.31, it's broken again - and completely. The first hibernation works
>> but it always crashes on the second one.
>>
>> The problem is that the oops is so long that it does not fit the screen -
>> even 1280x1024 is not enough. Scrolling does not work (as it used to do
>> on 2.4 kernels), netconsole does not catch it. So how can I debug this?
> 
> One way to get at least an indication of what broke the hibernation is to 
> run a bisection to find the change that broke hibernation. That could 
> provide sufficient clues for a next step.

And if it is reproducible with some debugging mode (see
Documentation/power/basic-pm-debugging.txt, section Test modes of
hibernation), you may not need to do a full hibernation/thaw cycle.

Could you attach at least the part of the oops you see?

Re: Hibernation oops with 2.6.31 won't fit the screen - bisected

[Ondrej Zary]

On Saturday 19 September 2009 23:44:54 Frans Pop wrote:
> Ondrej Zary wrote:
> > hibernation worked great with 2.6.30 - I had over a month of uptime (it
> > never worked so well before - always crashed after some time). But in
> > 2.6.31, it's broken again - and completely. The first hibernation works
> > but it always crashes on the second one.
> >
> > The problem is that the oops is so long that it does not fit the screen -
> > even 1280x1024 is not enough. Scrolling does not work (as it used to do
> > on 2.4 kernels), netconsole does not catch it. So how can I debug this?
>
> One way to get at least an indication of what broke the hibernation is to
> run a bisection to find the change that broke hibernation. That could
> provide sufficient clues for a next step.
>
> Would that be an option for you?

I bisected it and found this:

c7121843685de2bf7f3afd3ae1d6a146010bf1fc is first bad commit
commit c7121843685de2bf7f3afd3ae1d6a146010bf1fc
Author: Magnus Damm <magnus.damm@gmail.com>
Date:   Tue Jul 28 14:09:55 2009 -0700

    clocksource: Save mult_orig in clocksource_disable()

    To fix the common case where ->enable() does not set up
    mult, make sure mult_orig is saved in mult on disable.

    Also add comments to explain why we do this.

    Signed-off-by: Magnus Damm <damm@igel.co.jp>
    Cc: johnstul@us.ibm.com
    Cc: lethal@linux-sh.org
    Cc: akpm@linux-foundation.org
    LKML-Reference: <20090618152432.10136.9932.sendpatchset@rx1.opensource.se>
    Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

:040000 040000 c965062bc79af46cdb3522fc7ab1cc81d2f84de3 26616c88a8128b25fd15c3da39ac6251cd9723b6 M      include


Reverting this commit fixes the problem.

-- 
Ondrej Zary

Re: Hibernation oops with 2.6.31 won't fit the screen - how to debug?

[Pavel Machek]

On Sat 2009-09-19 21:55:00, Ondrej Zary wrote:
> Hello,
> hibernation worked great with 2.6.30 - I had over a month of uptime (it never 
> worked so well before - always crashed after some time). But in 2.6.31, it's 
> broken again - and completely. The first hibernation works but it always 
> crashes on the second one.
> 
> The problem is that the oops is so long that it does not fit the screen - even 
> 1280x1024 is not enough. Scrolling does not work (as it used to do on 2.4 
> kernels), netconsole does not catch it. So how can I debug this?

I see you solved that already but... vga=6 and mdelay(1000) in
printk/oops code is canonical way to debug that.
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html