[Pm-utils] second suspend/resume cycle hangs on resume

[Nigel Cunningham]

Hi Brian.

On Mon, 2008-12-15 at 22:28 +0000, Brian J. Murrell wrote:
> On Tue, 16 Dec 2008 07:59:39 +1100, Nigel Cunningham wrote:
> 
> > Hi.
> 
> Hi,
>  
> > How about if you just
> > 
> > echo mem > /sys/power/state
> 
> That has an interesting effect.  That does basically the same thing as 
> when I boot the kernel with init=/bin/bash and do a pm-suspend, and 
> that's to suspend fine, and resume fine except no backlight.  A second 
> suspend/resume (flying blind of course, but the suspend does happen as it 
> powers down and the power led blinks) fails per normal.  So the pm-utils 
> scripts seem to be doing something for for the first resume at least in 
> that the backlight comes back on.

Okay; It's probably a quirk that's (rightly) being applied.

> > Two things:
> > 
> > 1) I notice that it says in the LKML message that you're still using a
> > 2.6.24 based kernel.
> 
> Either there, or some other list I updated to say that I was using a 
> kernel.org 2.6.28-rc8 kernel currently with the same behaviour.
> 
> > With issues like this, it's always best to try with
> > a very recent kernel too, just in case the problem has been fixed since
> > Hardy was released but the fix was not backported to Hardy kernels.
> 
> Indeed.  Checked that one off.  :-)

Okay. Terrific.

> > 2) The reference to the module _might_ be due to using a framebuffer
> > console. If you're booting with vga=791 or such like, try temporarily
> > removing the vga=791.
> 
> I have no vga= arguments in my kernel command line.

Sorry for putting you wrong there - now that I've read the menuconfig
page about the i915 module, I know that it's the DRM module for X.
Without X running, it _should_ be unloadable. It shouldn't matter
anyway, as booting to init S should not cause it to be loaded, and thus
exclude it from the list of potential trouble makers.

The only remaining thing I can think of is to ask you to send us your
dmesg, preferably including the output from the first cycle. If you were
to turn on Verbose Power Management debugging messages as well (if
they're not already on - in the Power management and ACPI options menu
in the kernel config) and make the kernel log buffer nice and big so no
output is lost (General setup -> Kernel log buffer size), that would be
great.

Regards,

Nigel