On Tue, 2 Nov 2010, Thomas Renninger wrote:
>
> if idle= boot param is passed, cpuidle drivers should not register
> and only arch specific idle routines should get active.
>
> Ideally cpuidle subsystem would be made aware of this.
> But acpi parses sleep state tables before registering for cpuidle
> which should not happen if idle= is passed.
>
> -> Check for boot_option_idle_override in cpuidle drivers, not
> cpuidle subsystem.
>
> Compare with:
> arch/x86/kernel/process.c
> and
> drivers/acpi/processor_idle.c
idle=halt and idle=nomwait will not
set boot_option_idle_override,
so checking that option in intel_idle
is a NOP for those incantations.
idle_setup() is a confusing mess.
That function and all these idle workaround flags
need to be re-whacked.
thanks,
-Len Brown, Intel Open Source Technology Center
> Signed-off-by: Thomas Renninger <trenn@xxxxxxx>
> CC: lenb@xxxxxxxxxx
> CC: linux-acpi@xxxxxxxxxxxxxxx
> CC: stable@xxxxxxxxxx
> CC: linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx
>
> diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c
> index 41665d2..e50389b 100644
> --- a/drivers/idle/intel_idle.c
> +++ b/drivers/idle/intel_idle.c
> @@ -410,6 +410,9 @@ static int __init intel_idle_init(void)
> {
> int retval;
>
> + if (boot_option_idle_override)
> + return -ENODEV;
> +
> retval = intel_idle_probe();
> if (retval)
> return retval;
>
_______________________________________________
linux-pm mailing list
linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx
https://lists.linux-foundation.org/mailman/listinfo/linux-pm
