[ltp] Re: How to turn your T42p into a brick with ACPI...

[linux-thinkpad@linux-thinkpad.org]

On Sun, 10 Apr 2005, David A. Desrosiers wrote:

> 	Which modules? I don't (ever) run distribution-supplied
> kernels, so chances are I'm not running with 75 modules loaded, like
> the stock kernels like to do. My kernels are highly modular, and I've
> probably got 10-15 modules loaded at most times, more at peak
> peripheral times.
>
> 	I'm curious to see what modules everyone thinks needs to be
> unloaded before suspending. I'm only curious, because I've never had
> to unload anything before suspending with apm, except yenta_socket on
> a few of the more-recent 2.6.xx kernels (won't suspend if its loaded),
> but that's been solved now in the latest handful of kernels.

None, with the latest swsusp2, for me.  With older versions ~6 months,
ehci_hcd got in the way.  For anyone actually interested in swsusp2
who has problems, try suspending from init=/bin/sh boot with / ro
if you need to test.  The supplied hibernate script can also help
debugging if a module is holding off the suspend to disk.

Specific to your last comment David, a suspend-to-disk will always take
a finite amount of time, limited by the speed at which the machine can
write out core memory (in my case 1GB) to disk after doing a little
rearranging of caches.  And resume time will *always* be limited by
the time the BIOS takes to boot and hand over control to your init
fs, if that's where the resume occurs.  No, you don't need to wait
before closing the lid after asking for a suspend.  But as you don't
seem to want suspend to disk without supernatural disk write times,
I suggest you do as you've been saying, and stick with suspend to ram,
and plenteous reboots.