Re: [PATCH v18 12/23] dt-bindings: media: i2c: add MAX9296A, MAX96716A, MAX96792A

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] dt-bindings: validation bypass via endpoint@X naming
--

commit a82d316b34b24310779166fe2f87bfb5386cac5d
Author: Cosmin Tanislav <demonsingur@gmail.com>

dt-bindings: media: i2c: add MAX9296A, MAX96716A, MAX96792A

The patch introduces a new device tree binding for the MAX9296A, MAX96716A,
and MAX96792A GMSL2/3 to CSI-2 deserializers. It defines the required device
properties, I2C alias pools, and endpoint graph layouts for inputs and
outputs.

> diff --git a/Documentation/devicetree/bindings/media/i2c/maxim,max9296a.yaml b/Documentation/devicetree/bindings/media/i2c/maxim,max9296a.yaml
> new file mode 100644
> index 0000000000000..de6bfcec5dc63
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/media/i2c/maxim,max9296a.yaml

[ ... ]

> +  ports:
> +    $ref: /schemas/graph.yaml#/properties/ports
> +
> +    patternProperties:
> +      '^port@[0-1]$':
> +        $ref: /schemas/graph.yaml#/properties/port
> +        description: GMSL Input ports 0-1
> +
> +      '^port@[2-3]$':
> +        $ref: /schemas/graph.yaml#/$defs/port-base
> +        unevaluatedProperties: false
> +        description: CSI-2 Output ports 0-1
> +        properties:
> +          endpoint:

[Severity: Medium]
Does defining constraints under the literal endpoint key allow a validation
bypass for unit-addressed nodes?

Because the port@[2-3] block references /schemas/graph.yaml#/$defs/port-base,
the generic port-base schema evaluates both the literal endpoint node and any
^endpoint@[0-9a-f]+$ nodes. If a device tree defines endpoint@0, it will
satisfy unevaluatedProperties: false, but the local schema block skips it
entirely. This silently allows invalid or incomplete node configurations
(e.g., missing the mandatory data-lanes property) to pass dt_binding_check.

Would it be better to apply constraints using
patternProperties: "^endpoint(@[0-9a-f]+)?$", or explicitly map
^endpoint@[0-9a-f]+$ to false if only a single unaddressed endpoint is
intended here?

> +            $ref: /schemas/media/video-interfaces.yaml#
> +            unevaluatedProperties: false
> +
> +            properties:
> +              data-lanes:
> +                minItems: 1
> +                maxItems: 4
> +
> +              lane-polarities:
> +                minItems: 2
> +                maxItems: 5
> +
> +              link-frequencies:
> +                maxItems: 1
> +
> +            required:
> +              - data-lanes

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260930-gmsl2-3_serdes-v18-0-43117818f499@analog.com?part=12