Re: [PATCH V3 11/17] dt-bindings: arm: mediatek: Add #reset-cells property for MT8192-sys-clock

[Rex-BC Chen <rex-bc.chen@mediatek.com>]

On Mon, 2022-04-25 at 15:44 +0800, Krzysztof Kozlowski wrote:
> On 25/04/2022 04:37, Rex-BC Chen wrote:
> > > > +  '#reset-cells':
> > > > +    maximum: 2
> > > 
> > > Why this is a maximum? Usually this is const, so how do you use
> > > it
> > > (with
> > > what values)?
> > > 
> > 
> > We need to let the driver compatible with previous setting in
> > drivers/clk/mediatek/reset.c
> 
> Then it should be enum [1, 2], because '0' is not valid.
> 
> > There are two use cases in our reset driver:
> > (Refer to [1])
> > 
> > 1. #reset-cells = <1>
> >    When we input the argument, the older driver
> > use is to calculate  
> >    bank and bit by different method. From the implementation of
> >    reset_xlate(), we can see if the argument number is 1, it will
> >    return directly.
> 
> I understand this is an old binding with older compatibles, so this
> should be restricted per each variant (allOf:if:then)... but wait,
> old
> binding did not allow any reset-cells! You add an entirely new
> binding
> property and already want to support some older (deprecated?) way.
> 
> > 
> > 2. #reset-cells = <2>
> >    The input arguments is offset and bit. When we input two
> > arguments,
> >    we can use reset_xlate() to calculate the corresponding id to
> > assert
> >    and deassert.
> > 
> > [1]:
> > 
https://urldefense.com/v3/__https://lore.kernel.org/all/20220422060152.13534-10-rex-bc.chen@mediatek.com/__;!!CTRNKA9wMg0ARbw!0U0Yrp6WQxZ0YNMjaLJbAdq6Zyc524B4CY57-TP7QJ5FoSkCM72VI7mHJyWa1SZCnYTK$
> >  
> > 
> > If it's acceptable, I will add this in commit message.
> 
> 
> Best regards,
> Krzysztof

Hello Krzysztof,

I will let #reset-cells = <1> in next version and abandon the
modification of reset_xlate().

Thanks!

BRs,
Rex