Re: [PATCHv1 2/8] Documentation: arm: Link idle-states binding to code

From: Sudeep Holla
Date: Fri May 10 2019 - 09:04:05 EST

On Fri, May 10, 2019 at 04:59:40PM +0530, Amit Kucheria wrote:
> The enable-method needs to be psci for the psci_cpuidle_ops to be
> correctly registered.
> Add a note to the binding documentation on where to find the declaration
> of the enable-method since it is a macro and escapes any attempts to
> grep for it.
> Cc: Sudeep Holla <sudeep.holla@xxxxxxx>
> Signed-off-by: Amit Kucheria <amit.kucheria@xxxxxxxxxx>
> ---
> Documentation/devicetree/bindings/arm/idle-states.txt | 7 +++++++
> 1 file changed, 7 insertions(+)
> diff --git a/Documentation/devicetree/bindings/arm/idle-states.txt b/Documentation/devicetree/bindings/arm/idle-states.txt
> index 45730ba60af5..3a42335a6f3d 100644
> --- a/Documentation/devicetree/bindings/arm/idle-states.txt
> +++ b/Documentation/devicetree/bindings/arm/idle-states.txt
> @@ -239,6 +239,10 @@ processor idle states, defined as device tree nodes, are listed.
> # On ARM v8 64-bit this property is required and must
> be:
> - "psci"
> + (This assumes that the enable-method is "psci"
> + in the cpu node[6] that then uses the
> + CPUIDLE_METHOD_OF_DECLARE macro to setup the
> + psci_cpuidle_ops callbacks)

I don't prefer to refer some Linux implementation macros in DT bindings
as they may disappear any day. Further, the use of CPUIDLE_METHOD_OF_DECLARE
is restricted to ARM32 platforms only. So better to move it down without
the reference to the above macro or any kernel implementation details if

> # On ARM 32-bit systems this property is optional

Something like:
"This assumes that the "enable-method" property is set to "psci" in
in the cpu node[6] and use this property to set up the CPU idle
management in OS PM implementations"

With something on these line, you can add:

Acked-by: Sudeep Holla <sudeep.holla@xxxxxxx