Re: Resume from swsusp stopped working with 2.6.14 and 2.6.15-rc1

From: Pavel Machek
Date: Tue Nov 22 2005 - 13:47:24 EST


Hi!

> > > > > >> Bjorn, does it help if you change TIMEOUT in kernel/power/process.c to 30 * HZ?
> > > > > >
> > > > > > Funny, I thought that 6 seconds is way too much. Bjorn, please let us
> > > > > > know if 30 seconds timeout helps.
> > > > >
> > > > > It does.
> > > >
> > > > Ouch, yes, that's clear. It is stopping tasks during *resume*... So I
> > > > guess it gets wrong timing by design. Question is what to do with
> > > > that. Could we make keyboard driver pause the boot until it is done
> > > > resetting hardware? Or we can increase the timeout... would 10 seconds
> > > > be enough?
> > >
> > > Well, I think 10 seconds when suspending is a nice and resonable
> > > number. For resume though I think we should wait much longer, maybe
> > > even indefinitely - the only thing that timeout achieves is makes
> > > people fsck because the system can't recover from that state.
> >
> > I see your point, but it does not seem we need that changes this far. Your
> > patch is better, because we *could* hit that during suspend, just after
> > keyboard hotplug... right? And it will make resume faster for affected people.
>
> I disagree here. While my patch is a right thing to do (and as you
> know is already merged in mainline) it is not "better". Swsusp should
> not rely on the other subsystems being "nice" to it. Even with my
> patch there still could be moments when some thread is not suspended
> in 6 seconds when resuming causing unneeded resume failure and
> subsequent fsck.
>
> Please consider merging the patch below.

Well, I do not think this problem will surface again. It is first
failure in pretty long time. If it happens again, I'll take your
patch.
Pavel
--
Thanks, Sharp!
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/