Re: [PATCH] create CONFIG_SUSPEND_UP_POSSIBLE

[Len Brown <lenb@kernel.org>]

On Tuesday 31 July 2007 02:38, Pavel Machek wrote:
> Hi!
> 
> > Without this change, it is possible to build CONFIG_HIBERNATE
> > on all !SMP architectures, but not necessarily their SMP versions.
> 
> Did you want to say "CONFIG_SUSPEND"?

Yes.

> > I don't know for sure if the architecture list under SUSPEND_UP_POSSIBLE
> > is correct.  For now it simply matches the list for
> > SUSPEND_SMP_POSSIBLE.
> 
> I do not think it is.
> 
> > Signed-off-by: Len Brown <len.brown@intel.com>
> > ---
> >  Kconfig |    7 ++++++-
> >  1 file changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/kernel/power/Kconfig b/kernel/power/Kconfig
> > index 412859f..ccf6576 100644
> > --- a/kernel/power/Kconfig
> > +++ b/kernel/power/Kconfig
> > @@ -72,6 +72,11 @@ config PM_TRACE
> >  	CAUTION: this option will cause your machine's real-time clock to be
> >  	set to an invalid time after a resume.
> >  
> > +config SUSPEND_UP_POSSIBLE
> > +	bool
> > +	depends on (X86 && !X86_VOYAGER) || (PPC64 && (PPC_PSERIES ||
> 
> At least ARM can do suspend, too... probably others. I was under
> impression that SUSPEND is "supported" by all the architectures, just
> some of them veto it at runtime (using pm_ops or how was it renamed).

The reason this entire thread started is because Linus, Jeff and others
said that they didn't want code magically compiled into their kernel
that they did not explicitly ask for -- even if the savings were small
and that kernel was already something rather beefy, such as ACPI+SMP.

The current code is simply broken, because it allows SUSPEND
on IA64 if UP, but not on SMP.  It should really be neither.

Further, the only requirement to enable CONFIG_SUSPEND by default
is PM && CONFIG_SMP=n, and that doesn't make much sense -- particularly
since many of the non X86 architectures are small and according
to the logic above, if SUSPEND (and PM_SLEEP, which it enables)
adds any code to those kernels when they don't need it,
then it would be a regression.

thus my patch, if there is a list for SMP, there should
be a list for !SMP.  If the list is incomplete, we need to fix it.
 
-Len