Re: [PATCH 09/18] dt-bindings: dma: ti: Add document for K3 BCDMA

[Rob Herring <robh@kernel.org>]

On Thu, Oct 8, 2020 at 3:40 AM Peter Ujfalusi <peter.ujfalusi@ti.com> wrote:
>
>
>
> On 07/10/2020 18.46, Rob Herring wrote:
> > On Wed, Oct 07, 2020 at 12:09:06PM +0300, Peter Ujfalusi wrote:
> >>
> >>
> >> On 06/10/2020 22.29, Rob Herring wrote:
> >>> On Wed, Sep 30, 2020 at 12:14:03PM +0300, Peter Ujfalusi wrote:
> >>>> New binding document for
> >>>> Texas Instruments K3 Block Copy DMA (BCDMA).
> >>>>
> >>>> BCDMA is introduced as part of AM64.
> >>>>
> >>
> >> ...
> >>
> >>>
> >>>> +  ti,sci:
> >>>> +    description: phandle to TI-SCI compatible System controller node
> >>>> +    allOf:
> >>>> +      - $ref: /schemas/types.yaml#/definitions/phandle
> >>>> +
> >>>> +  ti,sci-dev-id:
> >>>> +    description: TI-SCI device id of BCDMA
> >>>> +    allOf:
> >>>> +      - $ref: /schemas/types.yaml#/definitions/uint32
> >>>
> >>> We have a common definition for these.
> >>
> >> Yes, in arm/keystone/ti,k3-sci-common.yaml, but I could not get to use
> >> that as reference.
> >>
> >> I can not list it under the topmost allOf and drop the ti,sci and
> >> ti,sci-dev-id like this:
> >>
> >> allOf:
> >>   - $ref: /schemas/dma/dma-controller.yaml#
> >>   - $ref: /schemas/arm/keystone/ti,k3-sci-common.yaml#
> >>
> >> It results:
> >>   CHKDT   Documentation/devicetree/bindings/processed-schema-examples.json
> >>   DTEX    Documentation/devicetree/bindings/dma/ti/k3-bcdma.example.dts
> >>   SCHEMA  Documentation/devicetree/bindings/processed-schema-examples.json
> >>   DTC     Documentation/devicetree/bindings/dma/ti/k3-bcdma.example.dt.yaml
> >>   CHECK   Documentation/devicetree/bindings/dma/ti/k3-bcdma.example.dt.yaml
> >> Documentation/devicetree/bindings/dma/ti/k3-bcdma.example.dt.yaml:
> >> dma-controller@485c0100: 'ti,sci', 'ti,sci-dev-id' do not match any of
> >> the regexes: 'pinctrl-[0-9]+'
> >>         From schema: Documentation/devicetree/bindings/dma/ti/k3-bcdma.yaml
> >>
> >> If I remove the "additionalProperties: false" from the schema file, then
> >> it compiles fine.
> >
> > Yeah, you have to do 'unevaluatedProperties: false' which doesn't
> > actually do anything yet, but can 'see' into $ref's.
>
> I see, but even if I add the unevaluatedProperties: false I will have
> the same error as long as I have additionalProperties: false

Yes. I meant unevaluatedProperties instead of additionalProperties.

> If I remove the additionalProperties then it makes no difference if I
> have the unevaluatedProperties: false or I don't.

Not yet, but it will soon. Once I have the tree in a consistent state
in 5.10-rc1, there will be a meta-schema to check all this (which is
one of those must always be present).

Though, as of now 'unevaluatedProperties' doesn't do anything because
the underlying json-schema tool doesn't yet support it.

> >>>> +  ti,sci-rm-range-bchan:
> >>>> +    description: |
> >>>> +      Array of BCDMA block-copy channel resource subtypes for resource
> >>>> +      allocation for this host
> >>>> +    allOf:
> >>>> +      - $ref: /schemas/types.yaml#/definitions/uint32-array
> >>>> +    minItems: 1
> >>>> +    # Should be enough
> >>>> +    maxItems: 255
> >>>
> >>> Are there constraints for the individual elements?
> >>
> >> In practice the subtype ID is 6bits number.
> >> Should I add limits to individual elements?
> >
> > Yes:
> >
> > items:
> >   maximum: 0x3f
>
> Right, I can just omit the minimum.
>
> It would be nice if I could use definitions for these ranges to avoid
> duplicated lines by adding
>
> definitions:
>   ti,rm-range:
>     $ref: /schemas/types.yaml#/definitions/uint32-array
>     minItems: 1
>     # Should be enough
>     maxItems: 255
>     items:
>       minimum: 0
>       maximum: 0x3f
>
> to schemas/arm/keystone/ti,k3-sci-common.yaml
>
> and only have:
>
>   ti,sci-rm-range-bchan:
>     $ref:
> /schemas/arm/keystone/ti,k3-sci-common.yaml#/definitions/ti,rm-range
>     description: |
>       Array of BCDMA block-copy channel resource subtypes for resource
>       allocation for this host

Just do:

patternProperties:
  "^ti,sci-rm-range-[btr]chan$":
    ...

If this is common for other bindings, then you can put it in
ti,k3-sci-common.yaml.

> but it results:
> Documentation/devicetree/bindings/dma/ti/k3-bcdma.yaml:
> properties:ti,sci-rm-range-bchan: {'$ref':
> '/schemas/arm/keystone/ti,k3-sci-common.yaml#/definitions/ti,rm-range',
> 'description': 'Array of BCDMA block-copy channel resource subtypes for
> resource\nallocation for this host\n'} is not valid under any of the
> given schemas (Possible causes of the failure):
>         Documentation/devicetree/bindings/dma/ti/k3-bcdma.yaml:
> properties:ti,sci-rm-range-bchan: 'not' is a required property
>         Documentation/devicetree/bindings/dma/ti/k3-bcdma.yaml:
> properties:ti,sci-rm-range-bchan:$ref:
> '/schemas/arm/keystone/ti,k3-sci-common.yaml#/definitions/ti,rm-range'
> does not match 'types.yaml#[/]{0,1}definitions/.*'

We probably should allow for using 'definitions' which is pretty
common json-schema practice, but don't primarily in order to keep
folks within the lines. Things are optimized for not knowing
json-schema and trying to minimize errors I have to check for.
Supporting it would complicate the meta-schema and the tools' fixup
code. So far, the need for it has been pretty infrequent.

Rob