Hi Geert, Krzysztof,

-----Original Message-----
From: John Madieu
Sent: Freitag, 24. April 2026 13:19
To: 'Geert Uytterhoeven' <geert@linux-m68k.org>
Subject: RE: [PATCH v5 01/14] ASoC: dt-bindings: sound: Add DT binding for
RZ/G3E sound

Hi Geert,

Thank you for the review.

quoted

I don't think both forms can really coexist on the same provider:
as `#sound-dai-cells` is declared once on the sound node, so on a given
board all references have to use the same cell count anyway. I might be
wrong. For your mixed example, the board would end up with `#sound-dai-
cells = <1>` and all consumer references would carry an index (the SSI1
one included).

The reason I kept both values in the enum was not to allow mixing within a
DTS, but to leave room for the two different kinds of board that will
plausibly use this IP:

  - a minimal board with one SSI wired to one codec, which can pick
    `#sound-dai-cells = <0>` and write the reference as
    `<&snd_rzg3e>`, without an index;

  - a board with multiple codecs (or one that wants to keep the door
    open for more), which picks `#sound-dai-cells = <1>` and writes
    all references as `<&snd_rzg3e N>`.

So the enum was meant as "pick one at board level" rather than "both
allowed at once".

This follows what the existing R-Car sound binding does, and the
convention is also spelled out as a comment on the sound node in the SoC
dtsi:

    /*
     * #sound-dai-cells is required
     *
     * Single DAI : #sound-dai-cells = <0>; <&rcar_sound>;
     * Multi  DAI : #sound-dai-cells = <1>; <&rcar_sound N>;
     */

That said, I'm happy to tighten this to `const: 1` if you (or
Krzysztof) think the flexibility isn't worth keeping. It would just mean
single-codec boards also have to write the index.
Would that be preferable?

Gentle ping on this topic, would you prefer me to fix it to '1' ?

Regards,
John.