[ltp] belt & suspenders suspend-to-ram script for acpid

[Eric Jorgensen]

On Thu, 22 Jun 2006 12:17:53 +0100
Richard Neill <rn214@hermes.cam.ac.uk> wrote:

> 
> 
> Eric Jorgensen wrote:
> >    This is tailored to my X21, so there are a few specific items
> >    that
> > are specific to the X20/X21/T20 and similar machines. Users of
> > better thinkpads will have to substitute the appropriate video bits.
> > 
> 
> Thanks very much. I wonder whether this also applies to the very
> similar  X22. My X22 sleeps fine, but goes totally dead  on resume:
> the HDD spins  up, but that's it. This only occurs if I have 3D-accel
> working.
> 
> Can you tell me:
> 
>    Which kernel and xorg versions you have
>    Could you post your "Device" section of xorg.conf


   The X21 is just an X20 with a faster cpu - and has no 3d accel at
all. You have a much newer ATi graphics chip. 

   At any rate, 2.6.17.1 (also working in various other late 2.6
kernels, back to 12 or so), and here ya go: 

Section "Device"
        Identifier      "ATI Technologies, Inc. Rage Mobility P/M AGP
2x"        Driver          "ati"
        BusID           "PCI:1:0:0"
EndSection

 
> > there is logic to prevent going to sleep if the thinkpad has
> > only just awakened in the last minute. This resolves the most
> > annoying thing my thinkpad has ever done -- requiring me to wake it
> > up three times
> 
> I think it's a race condition caused by multiple scripts in 
> /etc/acpi/events (and maybe some other clients of acpid) triggering on
> 
> Fn-F4. They all try to sleep, one gets there first, and then the
> others  take effect on wake up.


   Maybe for some people. Even if i have only one event script and one
action script, it attempts to sleep three times. 

   My diagnostics show that on my X21, acpid is receiving three
sequential lid events from the kernel. The events are numbered, and if
i record the events, there are three distinct lid events and they
absolutely come in threes.

   Additionally, the sleep script has always completely executed and
exited before it gets called again - even if i put 'sleep 100' at the
end. 

   I don't have the evidence anymore because my harddrive died. Never
buy toshiba harddrives. Don't let the 16mb of cache tempt you. I had two
of the things croak on me. 

   fwiw, I never use Fn-F4 or any other method to suspend - I just close
the lid. 

   In any case, this brute-force approach would resolve the race
condition as well. I'm not surprised that I'm not the first person to
end up with this sort of workaround, just dismayed that it wasn't in the
wiki.