Re: pcmcia + acpi + suspend2 seems impossible...
On 11/15/05, Eric van der Paardt <evdp@bishopfixtures.com> wrote:
> > -----Original Message-----
> > From: Martin Hauser [mailto:mh@alla.franken.de]
> > Sent: Tuesday, November 15, 2005 2:20 AM
> > To: debian-laptop@lists.debian.org
> > Subject: Re: pcmcia + acpi + suspend2 seems impossible...
> -snip-
> > > So far I haven't been able to get all three of these together in one
> > > kernel. Suspend2 doesn't appear to be a problem -- though after I
> > > apply the suspend2 patch I generally have to uncheck a bunch of
> > > extraneous module options in menuconfig. Suspend2 + pcmcia almost
> > > works for me; currently my main problem is that, *SOMETIMES*, after
> > > suspension, I get this:
> > >
> > > kernel: unregister_netdevice: waiting for eth0 to become free.
> Usage
> > > count = 5
> > >
> > > after which eth0 (a pcmcia NIC) is dead till reboot, and in fact
> even
> > > rebooting is blocked by this persistent message, which fills up my
> > > kernel log at intervals of 1 to 15 seconds.
> > >
> >
> > Hmm, interesting. Probably you should get your 'hibernate' script to
> > forcefully unload the pcmcia modules and the cardbus modules +
> probably
> > hack it so it does and /etc/init.d/pcmcia stop before it goes to
> sleep.
> > That way it'd probably be working back once you're resumed. You then
> > just push back in the modules, restart pcmcia and there you go. If you
> > need detailed instructions on how to do that, let me know.
>
> I have the same laptop running Debian and can confirm that you have to
> write some shell scripts to unload the PCMCIA drivers for suspend and
> resume to work. In addition to the PCMCIA drivers I also had to remove
> all the audio drivers for sound to work after resume.
>
> For me the magic combination was down the network, unload the WLAN,
> unload the audio, unload the PCMCIA, and then suspend. On resume it
> was; load audio, load PCMCIA, then load WLAN, and then fire up DHCP...
> Net result was a machine that suspended, but took nearly as long as
> shutting it down and restarting it.
>
> I don't recall using any kernel patches to do any of this, but as this
> is my sons laptop it has been a long while... so I could very well be
> wrong... as I do remember being leery to mess with ACPI as it's horribly
> broke on this machine and has been known to render them useless.
>
> E
>
>
hmm. Don't have the laptop on me right now, but I know I bring down
the PCMCIA interface before hibernation using the "hibernate" script.
I think the "unregister_netdevice" issue is a problem with the latest
kernel, as I haven't really seen it with earlier (2.6.12) kernels.
I'm wondering whether either one of you could send me a .config for a
working kernel -- Eric, sounds like you don't necessarily have access
to yours all the time -- but It would be nice to see if soneone else's
config also broke on my system, or whether the problem is in my
configuration.
A bummer about ACPI. I actually find that, with the exception of
rendering cardbus *unusable* on my machine, acpi seems to work pretty
well for me, at least with the most recent kernel. I was reminded of
why I like it today during lecture, when I went to move my laptop off
a piece of paper it was sitting on, accidentally hit the power switch,
and powered off accidentally! It would be nice if the power button
could be mapped to something else, like e.g. a dialog "do you really
want to turn off your computer?"
Anyway thanks, I will certainly add some additional protection to the
suspend script & see if that helps.
Matt
Reply to: