[PATCH v5 03/11] media: dt-bindings: media: add bindings for rockchip rk3568 vicap

[Michael Riesch]

Hi Krzysztof, Sakari,

Thanks for your feedback! Also, sorry for the delayed response, but as
the e-mail address indicates, there has been a job change in between
that kept me busy :-)

On 3/7/25 10:49, Sakari Ailus wrote:
> Hi Krzysztof, Michael,
> 
> On Fri, Mar 07, 2025 at 08:51:54AM +0100, Krzysztof Kozlowski wrote:
>> On Thu, Mar 06, 2025 at 05:56:04PM +0100, Michael Riesch wrote:
>>> Add documentation for the Rockchip RK3568 Video Capture (VICAP) unit.
>>>
>>> Signed-off-by: Michael Riesch <michael.riesch at wolfvision.net>
>>
>> subject: only one media prefix, the first
>>
>> A nit, subject: drop second/last, redundant "bindings". The
>> "dt-bindings" prefix is already stating that these are bindings.
>> See also:
>> https://elixir.bootlin.com/linux/v6.7-rc8/source/Documentation/devicetree/bindings/submitting-patches.rst#L18

Ack. Plain "media: dt-bindings: add rockchip rk3568 vicap" it is, then.

>>
>>> ---
>>>  .../bindings/media/rockchip,rk3568-vicap.yaml      | 169 +++++++++++++++++++++
>>>  MAINTAINERS                                        |   1 +
>>>  2 files changed, 170 insertions(+)
>>>
>>
>> ...
>>
>>> +  clocks:
>>> +    items:
>>> +      - description: ACLK
>>> +      - description: HCLK
>>> +      - description: DCLK
>>> +      - description: ICLK
>>> +
>>> +  clock-names:
>>> +    items:
>>> +      - const: aclk
>>> +      - const: hclk
>>> +      - const: dclk
>>> +      - const: iclk
>>> +
>>> +  rockchip,cif-clk-delaynum:
>>> +    $ref: /schemas/types.yaml#/definitions/uint32
>>> +    minimum: 0
>>> +    maximum: 127
>>> +    description:
>>> +      Delay the DVP path clock input to align the sampling phase, only valid
>>> +      in dual edge sampling mode. Delay is zero by default and can be adjusted
>>> +      optionally.
>>
>> default: 0

Ack.

> 
> And this is technically specific to the DVP port (0). Should (or could?) it
> be located there?

"Should"? Yes, makes sense to me.
"Could"? I guess, as we are referencing port-base here it should be
feasible. Not an expert opinion, mind you.

> 
>>
>>> +
>>> +  iommus:
>>> +    maxItems: 1
>>> +
>>> +  resets:
>>> +    items:
>>> +      - description: ARST
>>> +      - description: HRST
>>> +      - description: DRST
>>> +      - description: PRST
>>> +      - description: IRST
>>> +
>>> +  reset-names:
>>> +    items:
>>> +      - const: arst
>>> +      - const: hrst
>>> +      - const: drst
>>> +      - const: prst
>>> +      - const: irst
>>> +
>>> +  rockchip,grf:
>>> +    $ref: /schemas/types.yaml#/definitions/phandle
>>> +    description: Phandle to general register file used for video input block control.
>>> +
>>> +  power-domains:
>>> +    maxItems: 1
>>> +
>>> +  ports:
>>> +    $ref: /schemas/graph.yaml#/properties/ports
>>> +
>>> +    properties:
>>> +      port at 0:
>>> +        $ref: /schemas/graph.yaml#/$defs/port-base
>>> +        unevaluatedProperties: false
>>> +        description: The digital video port (DVP, a parallel video interface).
>>> +
>>> +        properties:
>>> +          endpoint:
>>> +            $ref: video-interfaces.yaml#
>>> +            unevaluatedProperties: false
>>> +
>>> +            properties:
>>> +              bus-type:
>>> +                enum: [5, 6]
>>> +
>>> +            required:
>>> +              - bus-type
>>> +
>>> +      port at 1:
>>> +        $ref: /schemas/graph.yaml#/properties/port
>>> +        description: Internal port connected to a MIPI CSI-2 host.
>>> +
>>> +        properties:
>>> +          endpoint:
>>> +            $ref: video-interfaces.yaml#
>>> +            unevaluatedProperties: false
>>
>> Hm, does it actually work? graph/port does not allow any other
>> properties. You should use graph/port-base and probably still narrow
>> lanes for both of port at 0 and port at 1.
> 
> I'd list the relevant properties for both DVP and CSI-2, either as
> mandatory or with defaults (could be reasonable for DVP signal polarities
> but not e.g. on number of CSI-2 lanes).

Not sure whether we are on the same page here. As pointed out in the
last round of feedback
(https://lore.kernel.org/all/0b19c544-f773-435e-9829-aaaa1c6daf7a@wolfvision.net/),
port at 1 is not MIPI CSI, but some internal interface.

I tried to clarify this by changing the description of this port to
"Internal port connected to a MIPI CSI-2 host." The host (see
rockchip,rk3568-mipi-csi.yaml) has a port that is actually MIPI CSI and
one port that is the other end of port at 1 here.

As to port at 1 here, I am not aware of any properties that can be set. Not
even very peculiar ones similar to rockchip,cif-clk-delaynum. Should I
have overlooked something, I think we can relax the constraints, but we
should start strict, right?

> 
>>
>>
>>> +
>>> +required:
>>> +  - compatible
>>> +  - reg
>>> +  - interrupts
>>> +  - clocks
>>> +  - ports
>>> +
>>> +additionalProperties: false
>>> +
>>> +examples:
>>> +  - |
>>> +    #include <dt-bindings/clock/rk3568-cru.h>
>>> +    #include <dt-bindings/interrupt-controller/arm-gic.h>
>>> +    #include <dt-bindings/interrupt-controller/irq.h>
>>> +    #include <dt-bindings/power/rk3568-power.h>
>>> +    #include <dt-bindings/media/video-interfaces.h>
>>> +
>>> +    parent {
>>
>> soc {

Ack.

Best regards,
Michael

>>
>>> +        #address-cells = <2>;
>>> +        #size-cells = <2>;
>>
>> Best regards,
>> Krzysztof
>>
>