Regression: First hibernation attempt fails

Regression: First hibernation attempt fails

[Alan Jenkins]

Yet another issue in 2.6.30-rc1 which was not present in 2.6.29 :-).  I
wonder if it rings any bells?

On my laptop (EEE 4G), the first attempt at hibernation fails. 
Subsequent attempts succeed.

If I use s2disk as normal, I see "snapshotting system", then before it
gets to write out the image, it aborts and switches back to X.   I can't
see any error message, even if I check the log with "dmesg".

If I use "echo disk > /sys/power/state", echo reports the error "Cannot
allocate memory".

This only happens if I've logged in to X first.  In fact, I narrowed it
down to gnome-power-manager.  If I kill gnome-power-manager beforehand,
then it doesn't happen.

Alan

Re: Regression: First hibernation attempt fails

[Alan Jenkins]

On 4/9/09, Alan Jenkins <alan-jenkins-cCz0Lq7MMjm9FHfhHBbuYA@public.gmane.org> wrote:
> Yet another issue in 2.6.30-rc1 which was not present in 2.6.29 :-).  I
> wonder if it rings any bells?
>
> On my laptop (EEE 4G), the first attempt at hibernation fails.
> Subsequent attempts succeed.
>
> If I use s2disk as normal, I see "snapshotting system", then before it
> gets to write out the image, it aborts and switches back to X.   I can't
> see any error message, even if I check the log with "dmesg".
>
> If I use "echo disk > /sys/power/state", echo reports the error "Cannot
> allocate memory".
>
> This only happens if I've logged in to X first.  In fact, I narrowed it
> down to gnome-power-manager.  If I kill gnome-power-manager beforehand,
> then it doesn't happen.

Hmm... killing g-p-m doesn't seem to affect this any more, perhaps I
spoke too soon.  But not logging into X still avoids it.

Re: [linux-pm] Regression: First hibernation attempt fails

[Pavel Machek]

On Thu 2009-04-09 12:13:03, Alan Jenkins wrote:
> Yet another issue in 2.6.30-rc1 which was not present in 2.6.29 :-).  I
> wonder if it rings any bells?
> 
> On my laptop (EEE 4G), the first attempt at hibernation fails. 
> Subsequent attempts succeed.
> 
> If I use s2disk as normal, I see "snapshotting system", then before it
> gets to write out the image, it aborts and switches back to X.   I can't
> see any error message, even if I check the log with "dmesg".
> 
> If I use "echo disk > /sys/power/state", echo reports the error "Cannot
> allocate memory".
> 
> This only happens if I've logged in to X first.  In fact, I narrowed it
> down to gnome-power-manager.  If I kill gnome-power-manager beforehand,
> then it doesn't happen.

Anything interesting in dmesg?

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [linux-pm] Regression: First hibernation attempt fails

[Alan Jenkins]

Pavel Machek wrote:
> On Thu 2009-04-09 12:13:03, Alan Jenkins wrote:
>   
>> Yet another issue in 2.6.30-rc1 which was not present in 2.6.29 :-).  I
>> wonder if it rings any bells?
>>
>> On my laptop (EEE 4G), the first attempt at hibernation fails. 
>> Subsequent attempts succeed.
>>
>> If I use s2disk as normal, I see "snapshotting system", then before it
>> gets to write out the image, it aborts and switches back to X.   I can't
>> see any error message, even if I check the log with "dmesg".
>>
>> If I use "echo disk > /sys/power/state", echo reports the error "Cannot
>> allocate memory".
>>
>> This only happens if I've logged in to X first.  In fact, I narrowed it
>> down to gnome-power-manager.  If I kill gnome-power-manager beforehand,
>> then it doesn't happen.
>>     
>
> Anything interesting in dmesg?
>   

Nothing obvious. I've attached my latest at
<http://bugzilla.kernel.org/show_bug.cgi?id=13058>.

Regards
Alan

Re: [linux-pm] Regression: First hibernation attempt fails

[Pavel Machek]

On Fri 2009-04-10 12:08:13, Alan Jenkins wrote:
> Pavel Machek wrote:
> > On Thu 2009-04-09 12:13:03, Alan Jenkins wrote:
> >   
> >> Yet another issue in 2.6.30-rc1 which was not present in 2.6.29 :-).  I
> >> wonder if it rings any bells?
> >>
> >> On my laptop (EEE 4G), the first attempt at hibernation fails. 
> >> Subsequent attempts succeed.
> >>
> >> If I use s2disk as normal, I see "snapshotting system", then before it
> >> gets to write out the image, it aborts and switches back to X.   I can't
> >> see any error message, even if I check the log with "dmesg".
> >>
> >> If I use "echo disk > /sys/power/state", echo reports the error "Cannot
> >> allocate memory".
> >>
> >> This only happens if I've logged in to X first.  In fact, I narrowed it
> >> down to gnome-power-manager.  If I kill gnome-power-manager beforehand,
> >> then it doesn't happen.
> >
> > Anything interesting in dmesg?
> 
> Nothing obvious. I've attached my latest at
> <http://bugzilla.kernel.org/show_bug.cgi?id=13058>.

Ok, next theory was that some module stuff is responsible (IIRC, from
the lists). Can you try with everything-built-in (no modules) kernel?
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [linux-pm] Regression: First hibernation attempt fails

[Alan Jenkins]

Pavel Machek wrote:
> On Fri 2009-04-10 12:08:13, Alan Jenkins wrote:
>   
>> Pavel Machek wrote:
>>     
>>> On Thu 2009-04-09 12:13:03, Alan Jenkins wrote:
>>>   
>>>       
>>>> Yet another issue in 2.6.30-rc1 which was not present in 2.6.29 :-).  I
>>>> wonder if it rings any bells?
>>>>
>>>> On my laptop (EEE 4G), the first attempt at hibernation fails. 
>>>> Subsequent attempts succeed.
>>>>
>>>> If I use s2disk as normal, I see "snapshotting system", then before it
>>>> gets to write out the image, it aborts and switches back to X.   I can't
>>>> see any error message, even if I check the log with "dmesg".
>>>>
>>>> If I use "echo disk > /sys/power/state", echo reports the error "Cannot
>>>> allocate memory".
>>>>
>>>> This only happens if I've logged in to X first.  In fact, I narrowed it
>>>> down to gnome-power-manager.  If I kill gnome-power-manager beforehand,
>>>> then it doesn't happen.
>>>>         
>>> Anything interesting in dmesg?
>>>       
>> Nothing obvious. I've attached my latest at
>> <http://bugzilla.kernel.org/show_bug.cgi?id=13058>.
>>     
>
> Ok, next theory was that some module stuff is responsible (IIRC, from
> the lists). Can you try with everything-built-in (no modules) kernel?
> 									Pavel
>   

It's still present in v2.6.30-rc1-136-g62b8e68, which includes "module:
try_then_request_module must wait".

I don't think it's due to a missing module, because it doesn't happen if
I run "s2disk" from a console without logging in to KDE. I've put some
more details in Bugzilla about how it is reproduced (and what makes it
go away), it's a bit confusing.

But so long as I get the conditions right, it reproduces 100%. I'm going
to just try bisecting it - unless anyone has any more suggestions.

Thanks
Alan

[BISECTED] Regression: First hibernation attempt fails

[Alan Jenkins]

Alan Jenkins wrote:
> Pavel Machek wrote:
>   
>> On Fri 2009-04-10 12:08:13, Alan Jenkins wrote:
>>   
>>     
>>> Pavel Machek wrote:
>>>     
>>>       
>>>> On Thu 2009-04-09 12:13:03, Alan Jenkins wrote:
>>>>   
>>>>       
>>>>         
>>>>> Yet another issue in 2.6.30-rc1 which was not present in 2.6.29 :-).  I
>>>>> wonder if it rings any bells?
>>>>>
>>>>> On my laptop (EEE 4G), the first attempt at hibernation fails. 
>>>>> Subsequent attempts succeed.
>>>>>
>>>>> If I use s2disk as normal, I see "snapshotting system", then before it
>>>>> gets to write out the image, it aborts and switches back to X.   I can't
>>>>> see any error message, even if I check the log with "dmesg".
>>>>>
>>>>> If I use "echo disk > /sys/power/state", echo reports the error "Cannot
>>>>> allocate memory".
>>>>>
>>>>> This only happens if I've logged in to X first.  In fact, I narrowed it
>>>>> down to gnome-power-manager.  If I kill gnome-power-manager beforehand,
>>>>> then it doesn't happen.
>>>>>         
>>>>>           
>>>> Anything interesting in dmesg?      
>>>>         

>>> Nothing obvious. I've attached my latest at
>>> <http://bugzilla.kernel.org/show_bug.cgi?id=13058>.
>>>       

Bisection succeeded:

1faa16d22877f4839bd433547d770c676d1d964c
"block: change the request allocation/congestion logic to be sync/async
based"

If I revert the commit, it fixes my bug.

Regards
Alan