RE: [PATCH v5 1/6] dt-bindings: media: platform: visconti: Add Toshiba Visconti Video Input Interface bindings

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



> -----Original Message-----
> From: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> Sent: Wednesday, February 1, 2023 6:46 PM
> To: ishikawa yuji(石川 悠司 ○RDC□AITC○EA開)
> <yuji2.ishikawa@xxxxxxxxxxxxx>
> Cc: krzysztof.kozlowski@xxxxxxxxxx; hverkuil@xxxxxxxxx; mchehab@xxxxxxxxxx;
> iwamatsu nobuhiro(岩松 信洋 □SWC◯ACT)
> <nobuhiro1.iwamatsu@xxxxxxxxxxxxx>; robh+dt@xxxxxxxxxx;
> krzysztof.kozlowski+dt@xxxxxxxxxx; rafael.j.wysocki@xxxxxxxxx;
> broonie@xxxxxxxxxx; linux-media@xxxxxxxxxxxxxxx;
> linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> devicetree@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH v5 1/6] dt-bindings: media: platform: visconti: Add Toshiba
> Visconti Video Input Interface bindings
> 
> On Mon, Jan 30, 2023 at 09:06:25AM +0000, yuji2.ishikawa@xxxxxxxxxxxxx wrote:
> > On Monday, January 23, 2023 4:26 AM, Laurent Pinchart wrote:
> > > On Tue, Jan 17, 2023 at 06:01:27PM +0100, Krzysztof Kozlowski wrote:
> > > > On 17/01/2023 16:58, Laurent Pinchart wrote:
> > > > > On Tue, Jan 17, 2023 at 04:42:51PM +0100, Krzysztof Kozlowski wrote:
> > > > >> On 17/01/2023 16:26, Laurent Pinchart wrote:
> > > > >>>
> > > > >>>> +
> > > > >>>> +          clock-lanes:
> > > > >>>> +            description: VIIF supports 1 clock line
> > > > >>>
> > > > >>> s/line/lane/
> >
> > Sorry for a late reply.
> > I'll fix the description.
> >
> > > > >>>
> > > > >>>> +            const: 0
> > > > >>>
> > > > >>> I would also add
> > > > >>>
> > > > >>>           clock-noncontinuous: true
> > > > >>>           link-frequencies: true
> > > > >>>
> > > > >>> to indicate that the above two properties are used by this device.
> > > > >>
> > > > >> No, these are coming from other schema and there is never need
> > > > >> to mention some property to indicate it is more used than other case.
> > > > >> None of the bindings are created such way, so this should not be
> exception.
> > > > >
> > > > > There are some bindings that do so, but that may not be a good
> > > > > enough reason, as there's a chance I wrote those myself :-)
> > > > >
> > > > > I would have sworn that at some point in the past the schema
> > > > > wouldn't have validated the example with this omitted. I'm not
> > > > > sure if something changed or if I got this wrong.
> > > >
> > > > You probably think about case when using
> > > > additionalProperties:false, where one has to explicitly list all
> > > > valid properties. But not for unevaluatedProperties:false.
> > >
> > > Possibly, yes.
> > >
> > > > > video-interfaces.yaml defines lots of properties applicable to
> > > > > endpoints. For a given device, those properties should be
> > > > > required
> > > >
> > > > required:
> > > >  - foo
> > > >
> > > > > (easy, that's defined in the bindings), optional,
> > > >
> > > > by default (with unevaluatedProperties:false) or explicitly
> > > > mention
> > > > "foo: true (with additionalProperties:false)
> > > >
> > > > >  or forbidden. How do
> > > >
> > > > foo: false (with unevaluatedProperties:false) or by default (with
> > > > additionalProperties:false)
> > >
> > > I think we should default to the latter. video-interfaces.yaml
> > > contains lots of properties endpoint properties, most bindings will
> > > use less than half of them, so having to explicitly list all the
> > > ones that are not used with "foo: false" would be quite
> > > inconvenient. Furthermore, I expect more properties to be added to
> > > video-interfaces.yaml over time, and those shouldn't be accepted by default
> in existing bindings.
> > >
> >
> > I caught up with this discussion after some exercise on JSON schema
> validator.
> > I'll remove "unevaluatedProperties: false" at the "endpoint" and add
> "aditionalProperties: false" instead.
> > Furthermore, I'll explicitly declare required properties (required: ["foo"]) and
> optional properties (properties: {foo: true}) for Visconti.
> > Is this correct understanding?
> 
> Looks very good to me !
> 
> > Are these changes also applied to "port", which is the parent node of
> > the "endpoint" ?
> 
> That shouldn't be needed, as the "port" node should only have "endpoint"
> children and no other properties (except for reg, and possibly #address-cells and
> #size-cells of course).

All right. I'll apply the change to "endpoint".

> > > > > we differentiate between the latter two cases ?
> 
> --
> Regards,
> 
> Laurent Pinchart

Regards,
Yuji Ishikawa




[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux