On Thu, Nov 10, 2022 at 06:43:34PM +0100, Miquel Raynal wrote: > Hi Rob, > > robh@xxxxxxxxxx wrote on Thu, 10 Nov 2022 08:05:45 -0600: > > > On Thu, Nov 10, 2022 at 09:50:34AM +0100, Miquel Raynal wrote: > > > Hi Rob, > > > > > > robh@xxxxxxxxxx wrote on Wed, 9 Nov 2022 22:00:55 -0600: > > > > > > > On Fri, Nov 04, 2022 at 05:38:33PM +0100, Miquel Raynal wrote: > > > > > Add a schema for the ONIE tlv NVMEM layout that can be found on any ONIE > > > > > compatible networking device. > > > > > > > > > > Signed-off-by: Miquel Raynal <miquel.raynal@xxxxxxxxxxx> > > > > > --- > > > > > .../bindings/nvmem/layouts/nvmem-layout.yaml | 1 + > > > > > .../nvmem/layouts/onie,tlv-layout.yaml | 115 ++++++++++++++++++ > > > > > 2 files changed, 116 insertions(+) > > > > > create mode 100644 Documentation/devicetree/bindings/nvmem/layouts/onie,tlv-layout.yaml > > > > > > > > > > diff --git a/Documentation/devicetree/bindings/nvmem/layouts/nvmem-layout.yaml b/Documentation/devicetree/bindings/nvmem/layouts/nvmem-layout.yaml > > > > > index f64ea2fa362d..8512ee538c4c 100644 > > > > > --- a/Documentation/devicetree/bindings/nvmem/layouts/nvmem-layout.yaml > > > > > +++ b/Documentation/devicetree/bindings/nvmem/layouts/nvmem-layout.yaml > > > > > @@ -19,6 +19,7 @@ description: | > > > > > > > > > > oneOf: > > > > > - $ref: kontron,sl28-vpd.yaml > > > > > + - $ref: onie,tlv-layout.yaml > > > > > > > > > > properties: > > > > > compatible: true > > > > > diff --git a/Documentation/devicetree/bindings/nvmem/layouts/onie,tlv-layout.yaml b/Documentation/devicetree/bindings/nvmem/layouts/onie,tlv-layout.yaml > > > > > new file mode 100644 > > > > > index 000000000000..1d91277324ac > > > > > --- /dev/null > > > > > +++ b/Documentation/devicetree/bindings/nvmem/layouts/onie,tlv-layout.yaml > > > > > @@ -0,0 +1,115 @@ > > > > > +# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause) > > > > > +%YAML 1.2 > > > > > +--- > > > > > +$id: http://devicetree.org/schemas/nvmem/layouts/onie,tlv-layout.yaml# > > > > > +$schema: http://devicetree.org/meta-schemas/core.yaml# > > > > > + > > > > > +title: NVMEM layout of the ONIE tlv table > > > > > + > > > > > +maintainers: > > > > > + - Miquel Raynal <miquel.raynal@xxxxxxxxxxx> > > > > > + > > > > > +description: > > > > > + Modern networking hardware implementing the Open Compute Project ONIE > > > > > + infrastructure shall provide a non-volatile memory with a table whose the > > > > > + content is well specified and gives many information about the manufacturer > > > > > + (name, country of manufacture, etc) as well as device caracteristics (serial > > > > > + number, hardware version, mac addresses, etc). The underlaying device type > > > > > + (flash, EEPROM,...) is not specified. The exact location of each value is also > > > > > + dynamic and should be discovered at run time because it depends on the > > > > > + parameters the manufacturer decided to embed. > > > > > + > > > > > +properties: > > > > > + compatible: > > > > > + const: onie,tlv-layout > > > > > + > > > > > + product-name: true > > > > > > > > This is a node? If so, you need: > > > > > > > > type: object > > > > additionalProperties: false > > > > > > I thought referencing a schema under a property would be enough? > > > > > > Indeed in nvmem.yaml we create the property nvmem-layout and make it > > > reference nvmem-layout.yaml. Then, in nvmem-layout.yaml: > > > > > > oneOf: > > > - $ref: kontron,sl28-vpd.yaml > > > - $ref: onie,tlv-layout.yaml > > > > > > we reference the different layouts that may apply (very much like what > > > you proposed to list the mtd partition parsers, if I got it right). > > > > > > Isn't it enough? > > > > No. It is enough to allow the property, but nothing defines what it must > > be (a node) and what the node contains in the case of empty nodes. Try > > adding 'product-name = "foo";' and it won't warn. > > There was a misunderstanding on my side. I thought your comment was > about the nvmem-layout node. Actually you were commenting about all the > sub-nodes defining nvmem-cells inside, so I'm fully aligned with your > response. > > However, if I understood it correctly, you basically said that: > > property: > $ref: foo.yaml > > is not the same as: > > property: > type: object > $ref: foo.yaml > > If that's the case, then should we consider dropping this patch (which > you agreed with in the first place)? >From a json-schema standpoint, they may not be the same. You can't know without knowing what's in foo.yaml. They are the same only if foo.yaml contains 'type: object'. It does for us because the tools will add 'type: object' to the top-level of every schema file. properties: property: properties: foo: {} These would all pass validation with the above: property; property = "bar"; More generally, json-schema's behavior is if a keyword doesn't apply for an instance, just silently ignore it. So while 'properties' keyword only makes sense on an object or maxItems on an array, json-schema doesn't care. The dtschema tools do a bit to counteract that. Rob
