Re: [PATCH 15/36] dt-bindings: arm: Convert Actions Semi bindings to jsonschema

From: Rob Herring
Date: Sun Oct 07 2018 - 16:11:23 EST


On Sat, Oct 6, 2018 at 5:40 AM Andreas FÃrber <afaerber@xxxxxxx> wrote:
>
> Hi Rob,
>
> Am 05.10.18 um 18:58 schrieb Rob Herring:
> > Convert Actions Semi SoC bindings to DT schema format using json-schema.
>
> This sounds like the next Yanny vs. Laurel... I fail to see any json. ;)

Read the docs in patch 8.

> Also, it may help my understanding to be CC'ed on the cover letter, too?

Sorry, trying to avoid a huge Cc list.

> > Cc: "Andreas FÃrber" <afaerber@xxxxxxx>
> > Cc: Mark Rutland <mark.rutland@xxxxxxx>
> > Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
> > Cc: devicetree@xxxxxxxxxxxxxxx
> > Signed-off-by: Rob Herring <robh@xxxxxxxxxx>
> > ---
> > .../devicetree/bindings/arm/actions.txt | 56 -------------------
> > .../devicetree/bindings/arm/actions.yaml | 34 +++++++++++
> > 2 files changed, 34 insertions(+), 56 deletions(-)
> > delete mode 100644 Documentation/devicetree/bindings/arm/actions.txt
> > create mode 100644 Documentation/devicetree/bindings/arm/actions.yaml
> >
> > diff --git a/Documentation/devicetree/bindings/arm/actions.txt b/Documentation/devicetree/bindings/arm/actions.txt
> > deleted file mode 100644
> > index d54f33c4e0da..000000000000
> > --- a/Documentation/devicetree/bindings/arm/actions.txt
> > +++ /dev/null
> > @@ -1,56 +0,0 @@
> > -Actions Semi platforms device tree bindings
> > --------------------------------------------
> > -
> > -
> > -S500 SoC
> > -========
> > -
> > -Required root node properties:
> > -
> > - - compatible : must contain "actions,s500"
> > -
> > -
> > -Modules:
> > -
> > -Root node property compatible must contain, depending on module:
> > -
> > - - LeMaker Guitar: "lemaker,guitar"
> > -
> > -
> > -Boards:
> > -
> > -Root node property compatible must contain, depending on board:
> > -
> > - - Allo.com Sparky: "allo,sparky"
> > - - Cubietech CubieBoard6: "cubietech,cubieboard6"
> > - - LeMaker Guitar Base Board rev. B: "lemaker,guitar-bb-rev-b", "lemaker,guitar"
> > -
> > -
> > -S700 SoC
> > -========
> > -
> > -Required root node properties:
> > -
> > -- compatible : must contain "actions,s700"
> > -
> > -
> > -Boards:
> > -
> > -Root node property compatible must contain, depending on board:
> > -
> > - - Cubietech CubieBoard7: "cubietech,cubieboard7"
> > -
> > -
> > -S900 SoC
> > -========
> > -
> > -Required root node properties:
> > -
> > -- compatible : must contain "actions,s900"
> > -
> > -
> > -Boards:
> > -
> > -Root node property compatible must contain, depending on board:
> > -
> > - - uCRobotics Bubblegum-96: "ucrobotics,bubblegum-96"
> > diff --git a/Documentation/devicetree/bindings/arm/actions.yaml b/Documentation/devicetree/bindings/arm/actions.yaml
> > new file mode 100644
> > index 000000000000..af9345a228b4
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/arm/actions.yaml
> > @@ -0,0 +1,34 @@
> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/soc/arm/actions.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
>
> 404 for the schema. Where does one find an explanation?

In the docs. This could someday actually be hosted on devicetree.org,
but for now only the path, not the host, matters.

>
> > +
> > +title: Actions Semi platforms device tree bindings
> > +
> > +maintainers:
> > + - Andreas FÃrber <afaerber@xxxxxxx>
>
> Mani is now officially reviewer and the closest I have to a
> co-maintainer.

Okay, NP.

> I suggest we add him here in some form. I assume this is
> independent of MAINTAINERS patterns though, or will get_maintainers.pl
> parse this, too?

It is independent. Not really ideal, but the bindings don't live just
in the kernel. (And lots are missing a maintainer anyways).

> > +
> > +description: |
>
> Does the | have any meaning, or a stray typo?

Explained in other patch.

> > + The Actions Semi S500 is a quad-core ARM Cortex-A9 SoC. The Actions Semi
> > + S900 is a quad-core ARM Cortex-A53 SoC.
>
> You forgot the S700 as another quad-core Cortex-A53 SoC.
> Also, arm or Arm rather than ARM these days?

I was going to say it's just copied out from the old file, but that
doesn't seem to be the case here. I am sure I didn't spend time on
nice descriptions. :)

> > +
> > +properties:
> > + compatible:
> > + oneOf:
> > + - items:
> > + - enum:
> > + - lemaker,guitar-bb-rev-b
> > + - enum:
> > + - lemaker,guitar
> > + - allo,sparky
> > + - cubietech,cubieboard6
> > + - const: actions,s500
> > + minItems: 2
> > + maxItems: 3
> > + additionalItems: false
>
> Objection. You've managed to turn a perfectly human-comprehensible text
> into a machine-parseable representation incomprehensible for humans.
>
> First, there should remain some free-text explanation of the values
> defined here. Are comments allowed after the value or indented maybe?

Yes, both. Comments are the major reason for using YAML over JSON file format.

> Alternatively we could have a per-vendor file à la vendor-prefixes.txt,
> but that would seem inefficient.

You mean per board vendor files? We'd have to duplicate the SoC
compatibles then as you can't really split the schema to 2 files.

There's 2 SoC families doing per board vendor files: Atmel (only one
vendor though) and i.MX. This series makes those 2 a single file like
all other SoCs. I don't really care which way it is done, but I think
we should be consistent and I'm not going to convert everyone's
bindings to per board vendor files.

> Next, the above items construct is horrible. What about nested oneOf:
>
> + - items:
> + - oneOf:
> + - items:
> + - enum:
> + - lemaker,guitar-bb-rev-b
> + - const: lemaker,guitar
> + - items:
> + - enum:
> + - allo,sparky
> + - cubietech,cubieboard6
> + - const: actions,s500

That's not actually doing what you think. That's saying the first item
is a list of items (1 or 2 items). For example, you'd have to have
data like this to be valid:

[ [ allo,sparky ], actions,s500 ]

(Note that [] are more json style list format, but allowed in yaml too.)

Probably the best thing to do here is separate the 2 and 3 item cases.
That also would eliminate some combinations here that are allowed, but
not valid.

>
> This grouping is much clearer to me and hopefully to anyone adding
> further base boards for the module.
> We will have the same issue for the BPi-S64 module with S700 below.
>
> > + - items:
> > + - const: cubietech,cubieboard7
> > + - const: actions,s700
> > + - items:
> > + - const: ucrobotics,bubblegum-96
> > + - const: actions,s900
>
> Please make the board compatible an enum, even if only one is listed
> today. That makes it clearer where/how (and easier) to add new boards.

Yes, I did that in most cases, but not here some reason.

Rob