On Sat, Jun 08, 2024 at 05:36:36PM +0200, Johan Hovold wrote:

On Wed, Jun 05, 2024 at 11:43:16AM +0300, Dmitry Baryshkov wrote:

quoted

On Wed, May 29, 2024 at 06:29:55PM +0200, Johan Hovold wrote:

quoted

Rework the pm8008 binding by dropping internal details like register
offsets and interrupts and by adding the missing regulator and
temperature alarm properties.

Note that child nodes are still used for pinctrl and regulator
configuration.

Also note that the pinctrl state definition will be extended later and
could eventually also be shared with other PMICs (e.g. by breaking out
bits of qcom,pmic-gpio.yaml).

Obviously we want to adapt this style of bindings for the other PMICs
too. My main concern here are PMICs which have two kinds of controlled
pins: GPIOs and MPPs. With the existing bindings style those are
declared as two subdevices. What would be your suggested way to support
MPPs with the proposed kind of bindings?

As far as I understand newer PMICs do not have MPP blocks and we do not
necessarily want to convert the existing bindings.

Well, I definitely want to do so.

That said, if there is ever a need to describe two separate gpio blocks
this can, for example, be done using subnodes on those PMICs.

This creates an asymmetry between older and newer PMICs. Wouldn't it be
better to always use gpios subnode for GPIO pins? This way older PMICS
will use the same approach _plus_ mpps {} subnode instead of having
either nothing or two subnodes.

The same approach probably applies to some other subdevices: temp-alarm
vs adc-tm, etc.

-- 
With best wishes
Dmitry