Re: [PATCH v5] dt-bindings: mfd: twl: Add missing sub-nodes for TWL4030 & TWL603x

[Rob Herring <robh@kernel.org>]

On Wed, Sep 03, 2025 at 04:46:43PM +0200, Andreas Kemnade wrote:
> Am Wed, 3 Sep 2025 00:55:25 +0200
> schrieb Jihed Chaibi <jihed.chaibi.dev@gmail.com>:
> 
> > > > +                  - ti,twl4030-power-idle-osc-off  
> > >
> > > this allows quite weird combinations like
> > >  "ti,twl4030-power-idle", "ti,twl4030-power-idle".
> > > I would propose to rather clean this up to things used in
> > > twl4030-power.c and at the same time available in dts, also
> > > taking the brush in the dts. I do not expect that these specific
> > > compatibles are in use anywhere. I looked around earlier.
> > >
> > > Regards,
> > > Andreas  
> > 
> > Hi Andreas,
> > 
> > Thank you for the feedback. I've done a deeper investigation into
> > the 'power:compatible' strings to see if the schema could be made
> > stricter.
> > 
> > While cleaning up the list, I found an existing DTSI file
> > (logicpd-torpedo-som.dtsi) that uses the combination:
> > 'compatible = "ti,twl4030-power-idle-osc-off", "ti,twl4030-power-idle";'
> > 
> > Since this "idle, idle" combination is already in use, it seems we
> > cannot make the schema stricter without breaking this existing
> > board.
> > 
> well the only maybe fallback line  I see here is
> ti,twl4030-power-idle-osc-off -> ti,twl4030-power-idle ->
> ti,twl4030-power.
> But you allow "twl,twl4030-power-idle", "ti,twl4030-power-idle"
> That absolutely makes no sense.

Actually, the above would be prevented. String entries have to be unique 
normally except a few cases which use non-unique-string-array type. 

> Then the question is whether there is the need for fallback compatibles.
> They are needed if there is one piece of software which does only know
> the fallback and can use the hardware in some limited mode, e.g.
> u-boot using some mmc controller only without some high speed mode.
> Looking around, I do not find anything in u-boot or barebox for the
> twl4030-power compatibles.
> 
> And if we define "ti,twl4030-power-idle" as a fallback for
> "ti,twl4030-power-idle-osc-off", then it is a fallback for everyone
> using "ti,twl4030-power-idle-osc-off", so then the dts would need to be
> corrected.
> 
> There is one exception: "ti,twl4030-power-omap3-evm" is still used but
> not everybody knows it (e.g. pm34xx.c), so there is a reason for a
> fallback compatible:"ti,twl4030-power-idle"
> 
> And the rest, time for the brush and lets not totally mess up
> ti,twl.yaml.

This is all pretty ancient h/w, so I wouldn't worry about it too much. 

Rob