On 8/17/22 3:27 AM, Krzysztof Kozlowski wrote:

On 17/08/2022 11:15, Samuel Holland wrote:

quoted

quoted

quoted

+examples:
+  - |
+    audio-codec@2030000 {
+        compatible = "simple-mfd", "syscon";

This cannot be on its own. Both require device specific compatible.

Again, the device-specific compatible does not exist, because the binding for
the audio codec has not been written (and it will be quite nontrivial).

So I can:
  1) Leave the example as-is until the audio codec binding gets written,
     and fill in the specific compatible at that time.
  2) Remove the example, with the reasoning that the example really
     belongs with the MFD parent (like for the other regulator). Then
     there will be no example until the audio codec binding is written.
  3) Drop the analog LDOs from this series entirely, and some parts
     of the SoC (like thermal monitoring) cannot be added to the DTSI
     until the audio codec binding is written.

What do you think?

How about just removing the audio-codec node? The schema is about
regulators, not audio-codec.

That works for me. I put the extra node there to signify that this is a MFD
child and requires some parent node to work, but I suppose it is not that
helpful to have.

OTOH, if you have parent device schema, you could put the example only
there. But as I understand, you don't have, right?

Right.

quoted

The same question applies for the D1 SoC DTSI, where I use this same construct.

This is not correct and should be fixed. Either you add the schema with
compatible or please drop the device node from the DTSI.

That's what I was afraid of.

Regards,
Samuel

quoted

(And technically this does validate with the current schema.)