Re: [PATCH 03/11] dt-bindings: reset: mobileye,eyeq5-reset: add EyeQ6L and EyeQ6H

From: Théo Lebrun
Date: Thu Apr 11 2024 - 10:06:15 EST


Hello,

On Thu Apr 11, 2024 at 8:14 AM CEST, Krzysztof Kozlowski wrote:
> On 10/04/2024 19:12, Théo Lebrun wrote:
> > Add bindings for EyeQ6L and EyeQ6H reset controllers.
> >
> > Some controllers host a single domain, meaning a single cell is enough.
> > We do not enforce reg-names for such nodes.
> >
> > Signed-off-by: Théo Lebrun <theo.lebrun@xxxxxxxxxxx>
> > ---
> > .../bindings/reset/mobileye,eyeq5-reset.yaml | 88 ++++++++++++++++++----
> > MAINTAINERS | 1 +
> > 2 files changed, 74 insertions(+), 15 deletions(-)
> >
> > diff --git a/Documentation/devicetree/bindings/reset/mobileye,eyeq5-reset.yaml b/Documentation/devicetree/bindings/reset/mobileye,eyeq5-reset.yaml
> > index 062b4518347b..799bcf15bed9 100644
> > --- a/Documentation/devicetree/bindings/reset/mobileye,eyeq5-reset.yaml
> > +++ b/Documentation/devicetree/bindings/reset/mobileye,eyeq5-reset.yaml
> > @@ -4,11 +4,13 @@
> > $id: http://devicetree.org/schemas/reset/mobileye,eyeq5-reset.yaml#
> > $schema: http://devicetree.org/meta-schemas/core.yaml#
> >
> > -title: Mobileye EyeQ5 reset controller
> > +title: Mobileye EyeQ reset controller
> >
> > description:
> > - The EyeQ5 reset driver handles three reset domains. Its registers live in a
> > - shared region called OLB.
> > + EyeQ reset controller handles one or more reset domains. They live in shared
> > + regions called OLB. EyeQ5 and EyeQ6L host one OLB each, each with one reset
> > + instance. EyeQ6H hosts 7 OLB regions; three of those (west, east,
> > + accelerator) host reset controllers. West and east are duplicates.
> >
> > maintainers:
> > - Grégory Clement <gregory.clement@xxxxxxxxxxx>
> > @@ -17,27 +19,83 @@ maintainers:
> >
> > properties:
> > compatible:
> > - const: mobileye,eyeq5-reset
> > + enum:
> > + - mobileye,eyeq5-reset
> > + - mobileye,eyeq6l-reset
> > + - mobileye,eyeq6h-we-reset
> > + - mobileye,eyeq6h-acc-reset
> >
> > - reg:
> > - maxItems: 3
> > + reg: true
>
> Same mistakes. Please open existing bindings with multiple variants,
> e.g. some Qualcomm, and take a look how it is done there.

Thanks for the pointer to good example, that is useful! So if we take
one random binding matching
Documentation/devicetree/bindings/clock/qcom,*.yaml and that contains
the "reg-names" string, we see:

reg:
items:
- description: LPASS qdsp6ss register
- description: LPASS top-cc register

reg-names:
items:
- const: qdsp6ss
- const: top_cc

I don't understand one thing; this doesn't tell you:

You can provide 2 MMIO blocks, which must be qdsp6ss and top_cc.

But it tells you:

Block zero must be qdsp6ss.
Block one must be top_cc.

If we do that I do not get the point of reg-names; we put more
information in our devicetree that is in any case imposed.

This is why I went with a different approach looking like:

reg:
minItems: 2
maxItems: 2
reg-names:
minItems: 2
maxItems: 2
items:
enum: [ d0, d1 ]

I know this is not perfect, but at least you don't enforce an order for
no reason. If "items: const..." approach should be taken, then I'll
remove reg-names which bring no benefit.

Thanks Krzysztof,

--
Théo Lebrun, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com