Re: [PATCH 3/6] acpi : remove pointless cpuidle device state_count init

["Rafael J. Wysocki" <rjw-KKrjLPT3xs0@public.gmane.org>]

On Sunday, September 16, 2012, Daniel Lezcano wrote:
> On 09/07/2012 11:50 PM, Rafael J. Wysocki wrote:
> > On Friday, September 07, 2012, Daniel Lezcano wrote:
> >> The cpuidle core takes care of filling this field from drv->state_count.
> > 
> > I'm not quite sure this is always valid.  If dev has already been
> > initialized and dev->state_count is different from 0,
> > cpuidle_enable_device() doesn't actually change it.
> > 
> > Have you considered all of the possible scenarios?
> 
> Ok, there is the scenario where the ACPI supports _CST.
> At runtime, ACPI may notify the OS a C-state changed for a specific cpu
> and this is done through 'acpi_processor_cst_has_changed' followed by
> 'acpi_processor_setup_cpuidle_cx'.
> 
> So at the end, we could have different cpus with one cpu with less
> C-states than the other, if I understood correctly ACPIspec30.pdf =>
> page 262 :)
> 
> In conclusion, we should keep this variable filled from there and keep
> this in mind in cpuidle.c in order to not break acpi in the future.
> Maybe a comment in cpuidle.c would help ... especially with a single
> C-state registration.

Sure, adding a comment in there sounds like a good idea.

Thanks,
Rafael