On 6/29/22 07:54, Krzysztof Kozlowski wrote:
> On 28/06/2022 19:01, Fabrice Gasnier wrote:
>
>>>>>> + connector:
>>>>>> + type: object> + allOf:
>>>>>> + - $ref: ../connector/usb-connector.yaml#
>>>>>
>>>>> Full path, so /schemas/connector/...
>>>>>
>>>>> unevaluatedProperties: false
>>
>> Hi Krzysztof,
>>
>> I Just figured out usb-connector schema has "additionalProperties:
>> true". Adding "unevaluatedProperties: false" here seem to be useless.
>> At least at my end, this make any dummy property added in the example
>> below to be validated without error by the schema.
>
> No, it's expected. The common schema allows additional properties. You
> specific device schema (including common) should not allow anything more
> and this is expressed like you mentioned.
>
> However depending on the version of dtschema, the
> unevaluatedProperties:false might still be not implemented. AFAIK, Rob
> added it quite recently.
>
>>
>> Should this be updated in usb-connector.yaml instead ?
>
> No
>
>>
>> Shall I omit it here in the end ?
>
> You need to add here unevaluatedProperties: false (on the level of this
> $ref)
>
>>
>>>>
>>>> ack,
>>>>
>>>>>
>>>>>> +
>>>>>> + firmware-name:
>>>>>> + description: |
>>>>>> + Should contain the name of the default firmware image
>>>>>> + file located on the firmware search path
>>>>>> +
>>>>>> + wakeup-source: true
>>>>>> + power-domains: true
>>>>>
>>>>> maxItems
>>>>
>>>> Do you mean maxItems regarding the "power-domains" property ?
>>>
>>> Yes.
>>>
>>>> This will depend on the user platform, where it's used as an I2C device.
>>>> So I'm not sure this can / should be specified here.
>>>> Could please you clarify ?
>>>
>>> Then maybe this property is not valid here. Power domains usually are
>>> used for blocks of a SoC, having common power source and power gating.
>>> In your case it looks much more like a regulator supply.
>>
>> This property is used in our implementation to refer to SOC PM domain
>> for GPIO that is used to wakeup the system. This isn't only a regulator,
>> this PM domain serves various IPs such as I2C, GPIO, UART... (it manages
>> regulator and clocks used in low power).
>>
>> I can limit to 1 item if this is fine for you ?
>>
>> e.g. maxItems: 1
>
> Yes, it's good (assuming it is true :) ).
>
>>
>>>
>>>>
>>>>>
>>>>>> +
>>>>>> +required:
>>>>>> + - compatible
>>>>>> + - reg
>>>>>> + - interrupts
>>>>>> +
>>>>>> +additionalProperties: false
>>>>>> +
>>>>>> +examples:
>>>>>> + - |
>>>>>> + #include <dt-bindings/interrupt-controller/irq.h>
>>>>>> + i2c5 {
>>>>>
>>>>> Just "i2c"
>>>>
>>>> ack,
>>>>
>>>>>
>>>>>> + #address-cells = <1>;
>>>>>> + #size-cells = <0>;
>>>>>> +
>>>>>> + stm32g0@53 {
>>>>>
>>>>> Generic node name describing class of the device.
>>>>
>>>>
>>>> I wasn't aware of generic node name for an I2C device (not talking of
>>>> the controller). I may have missed it.
>>>>
>>>> Could you please clarify ?
>>>
>>> The class of a device is not a I2C device. I2C is just a bus. For
>>> example the generic name for Power Management IC connected over I2C
>>> (quite common case) is "pmic".
>>>
>>> For USB HCD controllers the generic name is "usb". For USB
>>> ports/connectors this is "connector". So what is your hardware?
>>> "interface" is a bit too unspecific to figure it out.
>>
>> Thanks, I better understand your point now.
>>
>> A common definition for the hardware here could be "USB Type-C PD
>> controller". I'll improve this schema title by the way.
>>
>> I had a quick look in various .dts files. I could find mainly:
>> - typec-portc@hh
>> - usb-typec@hh
>> - typec@hh
>>
>> Not sure if this has already been discussed in other reviews, it lacks
>> the "controller" idea in the naming IMHO.
>> Perhaps something like "typec-pd-controller" or
>> "usb-typec-pd-controller" could be used here ?
>>
>> Otherwise, I could adopt the shortest "typec" name if it's fine for you ?
>
> typec sounds good.
>
>>
>>>
>>>>
>>>>>
>>>>>> + compatible = "st,stm32g0-typec";
>>>>>> + reg = <0x53>;
>>>>>> + /* Alert pin on GPIO PE12 */
>>>>>> + interrupts = <12 IRQ_TYPE_EDGE_FALLING>;
>>>>>> + interrupt-parent = <&gpioe>;
>>>>>> +
>>>>>> + /* Example with one type-C connector */
>>>>>> + connector {
>>>>>> + compatible = "usb-c-connector";
>>>>>> + label = "USB-C";
>>>>>> +
>>>>>> + port {
>>>>>
>>>>> This does not look like proper schema of connector.yaml.
>>>>
>>>> This refers to graph.yaml [1], where similar example is seen [2].
>>>>
>>>> https://github.com/devicetree-org/dt-schema/blob/main/dtschema/schemas/graph.yaml#L79
>>>>
>>>> https://github.com/devicetree-org/dt-schema/blob/main/dtschema/schemas/graph.yaml#L207
>>>
>>> Just look at the usb-conector schema. It's different. You miss ports.
>>> Maybe other properties as well.
>>
>>
>> (I may miss something, and got confused around port/ports earlier)
>> The graph properties seems to allow both the 'port' and 'ports' syntax
>> thanks to the graph definition.
>> The "port" syntax is also used in other typec controller schemas.
>>
>> There's only one port in this example. Of course other example could use
>> two or more ports (like for USB HS / SS / aux) which would require using
>> the "ports" node (with port@0/1/2 childs).
>>
>> I can adopt the "ports" node if you prefer. As I see it just doesn't
>> bring much in the current example (The only drawback is this adds one
>> indentation/node level w.r.t. the bellow example, so not a big deal).
>
> The graph schema allows, but you include here usb-connector schema which
> requires to put it under "ports". You should not use it differently, so
> I expect here "ports" property, even with one port.
Hi Krzysztof,
This makes senses. I've updated this locally and also put this in .dts
file (not sent yet with this series as I lack some dependencies not yet
upstream).
I'm able to validate the schema, with your statement, by using
dt_binding_check.
/* Example with one type-C connector */
connector {
compatible = "usb-c-connector";
label = "USB-C";
ports {
#address-cells = <1>;
#size-cells = <0>;
port@0 {
reg = <0>;
con_usb_c_ep: endpoint {
remote-endpoint = <&usb_ep>;
};
};
};
};
Still when build the .dts (in my downstream, with W=1) I observe various
case, for a single port usage (e.g. USB HS only).
With above example I get:
---
Warning (graph_child_address): /soc/..../connector/ports: graph node has
single child node 'port@0', #address-cells/#size-cells are not necessary
Remove them as not necessary (suggested by this warning):
---
/* Example with one type-C connector */
connector {
compatible = "usb-c-connector";
label = "USB-C";
ports {
port {
con_usb_c_ep: endpoint {
remote-endpoint = <&usb_ep>;
};
};
};
};
Then I no longer get this warning upon build. But the dtbs_check complains:
---
connector: ports: 'port@0' is a required property
From schema: ..
Documentation/devicetree/bindings/connector/usb-connector.yaml
So It looks like to me there's something missing to handle the single
port case in usb-connector.yaml, when using the "ports".
Maybe usb-connector could be updated to handle "port" (w/o unit-addr) ?
I'm talking about:
required:
- port@0
So, I came up with:
--- a/Documentation/devicetree/bindings/connector/usb-connector.yaml
+++ b/Documentation/devicetree/bindings/connector/usb-connector.yaml
@@ -176,6 +176,9 @@ properties:
port number as described below.
properties:
+ port:
+ $ref: /schemas/graph.yaml#/properties/port
+
port@0:
$ref: /schemas/graph.yaml#/properties/port
description: High Speed (HS), present in all connectors.
@@ -189,8 +192,11 @@ properties:
description: Sideband Use (SBU), present in USB-C. This
describes the
alternate mode connection of which SBU is a part.
- required:
- - port@0
+ oneOf:
+ - required:
+ - port
+ - required:
+ - port@0
Do you agree on this approach ? (I can add a pre-cursor patch to this
series, to handle the single port case)
Please advise,
Best Regards,
Fabrice
>
> Best regards,
> Krzysztof
