[TuxOnIce-devel] Help getting suspend to work

[Micha]

On Thu, 28 Feb 2008 09:01:21 +1100
Nigel Cunningham <ncunningham at crca.org.au> wrote:

> Hi Micha.
> 
> On Wed, 2008-02-27 at 17:54 +0200, Micha wrote:
> > On Wed, 27 Feb 2008 16:09:53 +0200
> > Micha <michf at post.tau.ac.il> wrote:
> > 
> > > On Thu, 21 Feb 2008 09:55:06 +1100
> > > Nigel Cunningham <nigel at nigel.suspend2.net> wrote:
> > > 
> > > > Hi Micha.
> > > > 
> > > > Micha wrote:
> > > > > Feb 20 21:50:25 vivalunalitshi kernel: Pageset1 has grown by 663
> > > > > pages. extra_pages_allowance is currently only 500.
> > > > 
> > > > In your /etc/hibernate/suspend2.conf,
> > > > 
> > > > add
> > > > 
> > > > ProcSetting extra_pages_allowance 1000
> > > > 
> > > > Your allowance for the memory drivers need isn't big enough at the 
> > > > moment. This will increase it.
> > > > 
> > > > Regards,
> > > > 
> > > > Nigel
> > > > 
> > > 
> > setting that value to 1000 still didn't work. I changed it to 1500 and that
> > got it to suspend rather reliably, although a bit strangely. It seems to
> > write caches, do some other things, all this will flashing the sleep led,
> > it then seems to turn off the machine for a few seconds, at which points it
> > wakes up again to resume freezing processes. Don't know if that is how it
> > is supposed to work.
> 
> We write the image in two parts. First, we save the page cache that we
> know won't be used/changed while we're hibernating, then we do an atomic
> copy of the remaining memory before saving it. The bit inbetween writing
> the caches and appearing to wake up again before powering off is where
> the atomic copy is being done. The machine isn't actually off at that
> time. Everything is just being stopped so we can get this atomic copy
> (well, it's not really atomic, but as close as possible). So the short
> answer is, yes - that's normal.
> 
> > Anyway the issue is that it seems to suspend ok but the machine doesn't
> > resume. The kernel starts to load, suspend starts up to load the image and
> > then it goes into a black screen with just a flashing cursor at the bottom
> > left, not error messages, no flashing leds, just stuck.
> 
> That will probably be because you have some driver loaded while
> hibernating that doesn't yet have properly implemented power management,
> and it's messing everything up. The easiest way to find out which driver
> is to start with a boot to single user mode, and try hibernating from
> there (preferably with as many modules as possible unloaded). If that
> works okay a couple of times, try adding modules and service until you
> can say "If I start X/load module Y, hibernation fails." I say a couple
> of times because sometimes we encounter drivers that appear to work okay
> after the first hibernation cycle, but the second cycle kills them
> properly.
> 

Seems to be the nvidia driver, since unloading everything but X fails to resume
and just killing X allows resuming.

anyone have success with getting the proprietary nvidia driver to suspend?

Thanks

> > I couldn't find anything regarding the resume in the log and very little
> > regarding the suspend. The only thing I see is the message saying that
> > resume failed, whether to remove the image.
> 
> Yes. That's normal, because nothing gets written to the filesystem after
> we start hibernating - to do so would introduce the possibility of
> corruption of the image and/or filesystem.
>  
> > I attached /var/log/messages in case it is of any yes.
> > 
> > I also read somewhere that the nvidia driver may be causing the issue, could
> > that be the problem?
> 
> It could be, but you could also have more than one thing playing up. I'd
> suggest starting with init S, as described above.
> 
> Regards,
> 
> Nigel
>