Re: [PATCH] ACPI/CPUIDLE: prevent setting pm_idle to NULL

[Thomas Gleixner <tglx@linutronix.de>]

On Mon, 28 Jul 2008, Andi Kleen wrote:

> > The problem here is that the acpi/cpuidle code can be in a state where
> > the _save/old variables _ARE_ NULL because they had not been
> > initialized with the original pm_idle before the module is removed or
> > the cst state changes. So all we have to do is to prevent pm_idle to
> > be set to NULL.
> 
> It still seems wrong to me to fall back to the cpuidle idle function
> instead of the earlier idle function just because cpuidle was loaded
> in a weird way. 

WTF are you talking about ? 

pm_idle_save is initialized with pm_idle (the original idle function
selected by the arch code).

When CST changes or the acpi/cpuidle modules are removed we need to
restore the original pm_idle function from pm_idle_save.

When pm_idle_save was not initialized, which can happen, then we wrote
NULL to pm_idle, which is obviously wrong. We simply want to avoid
that we write NULL into pm_idle.

That's all what the patch does. Nothing else. 

This problem was hidden by the magic

   if (!pm_idle)
      pm_idle = default_idle;

in the arch/x86 code which was removed when I refactored the pm_idle
initialization code.

Just get it. That "if (!pm_idle)" check in the arch code was just
papering over the fact that the acpi code set pm_idle to NULL under
certain conditions.

Thanks,

	tglx