On 28/02/2025 10:51, David Jander wrote:

quoted

quoted

quoted

Second sentence is both redundant and really not relevant to this
binding. It's not this binding which decides about sharing.  

Good point. I think I should drop the whole property, since it is indeed
irrelevant. If extra supplies need to be specified, they always can be, right?  

You should specify all supplies now, because hardware should be fully
described by binding and DTS.

In the case of the hardware I use for testing all of this, there are several
tmc5240 chips which have their "SLEEPN" pin tied together controlled by a
single GPIO pin that needs to be pulled high before any of these chips can be
talked to. The usual way I know of solving this is by specifying a common
"virtual" supply of type "regulator-fixed" with an enable gpio.

No, that is not usual way. Representing pin as fake supply is hack and
not correct hardware description.

But this isn't strictly a supply that has to do with this chip or driver, so I
don't think it should be specified in the schema. I do need to use it in my
particular case though. Is there a better way of doing this?

I speak about voltage and current supplies. These you must specify.

quoted

What's more, the necessary supplies (according to datasheet) should be
required, not optional.

Do you mean that they should be in the binding definition as well? I.e. add
all of Vs, Vdd1v8 and Vcc_io here?

Yes, all expected supplies must be in the binding.


Best regards,
Krzysztof