Re: [PATCH V2 1/2] dt-bindings: interrupt-controller: Add Loongson EIOINTC

[Krzysztof Kozlowski]

On 16/02/2023 02:46, Binbin Zhou wrote:
> On Tue, Feb 14, 2023 at 8:43 PM Krzysztof Kozlowski
> <krzysztof.kozlowski@xxxxxxxxxx> wrote:
>>
>> On 14/02/2023 13:40, Binbin Zhou wrote:
>>> On Tue, Feb 14, 2023 at 5:53 PM Krzysztof Kozlowski
>>> <krzysztof.kozlowski@xxxxxxxxxx> wrote:
>>>>
>>>> On 13/02/2023 13:15, Binbin Zhou wrote:
>>>>> Add Loongson Extended I/O Interrupt controller binding with DT schema
>>>>> format using json-schema.
>>>>>
>>>>> Signed-off-by: Binbin Zhou <zhoubinbin@xxxxxxxxxxx>
>>>>> ---
>>>>>  .../loongson,eiointc.yaml                     | 80 +++++++++++++++++++
>>>>>  1 file changed, 80 insertions(+)
>>>>>  create mode 100644 Documentation/devicetree/bindings/interrupt-controller/loongson,eiointc.yaml
>>>>>
>>>>> diff --git a/Documentation/devicetree/bindings/interrupt-controller/loongson,eiointc.yaml b/Documentation/devicetree/bindings/interrupt-controller/loongson,eiointc.yaml
>>>>> new file mode 100644
>>>>> index 000000000000..88580297f955
>>>>> --- /dev/null
>>>>> +++ b/Documentation/devicetree/bindings/interrupt-controller/loongson,eiointc.yaml
>>>>> @@ -0,0 +1,80 @@
>>>>> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
>>>>> +%YAML 1.2
>>>>> +---
>>>>> +$id: "http://devicetree.org/schemas/interrupt-controller/loongson,eiointc.yaml#";
>>>>> +$schema: "http://devicetree.org/meta-schemas/core.yaml#";
>>>>
>>>> Drop quotes from bopth.
>>>>
>>>>> +
>>>>> +title: Loongson Extended I/O Interrupt Controller
>>>>> +
>>>>> +maintainers:
>>>>> +  - Binbin Zhou <zhoubinbin@xxxxxxxxxxx>
>>>>> +
>>>>> +description: |
>>>>> +  This interrupt controller is found on the Loongson-3 family chips and
>>>>> +  Loongson-2K0500 chip and is used to distribute interrupts directly to
>>>>> +  individual cores without forwarding them through the HT's interrupt line.
>>>>> +
>>>>> +allOf:
>>>>> +  - $ref: /schemas/interrupt-controller.yaml#
>>>>> +
>>>>> +properties:
>>>>> +  compatible:
>>>>> +    enum:
>>>>> +      - loongson,eiointc-1.0
>>>>
>>>> Why not using SoC based compatible? It is preferred.
>>>
>>> Hi Krzysztof:
>>>
>>> So far, from the datasheet, I know that only the EXIOINTC of the
>>> Loongson-2K0500 is different from the other chips, and that is the
>>> "loongson,eio-num-vecs" below, which is 128, while all the others are
>>> 256.
>>> My original idea was to add this property to make compatible
>>> consistent, and also to make it easier to add new chips if they have
>>> different eio-num-vecs.
>>
>> We talk about different things. SoC based compatibles are preferred over
>> version ones. This was on the lists expressed many times. Please provide
>> a reason why you deviate from general recommendation. Flexibility and
>> genericness of bindings is not a reason - it's the opposite of the
>> argument, thus this will be a: NAK. :(
>>
>>
> Hi Krzysztof:
> 
> Allow me to give a brief overview of the current status of eiointc (DT-based):
>      Loongson-3A series supports eiointc;
>      Loongson-2K1000 does not support eiointc now;
>      Loongson-2K0500 supports eiointc, with differences from
> Loongson-3, e.g. only up to 128 devices are supported;
>      Loongson-2K2000 supports eiointc, similar to Loongson-3.
>      ....
> 
> As can be seen, there is now a bit of confusion in the chip's design of eiointc.
> 
> The design of eiointc is probably refined step by step with the chip.
> The same version of eiointc can be used for multiple chips, and the
> same chip series may also use different versions of eiointc. Low-end
> chips may use eiointc-2.0, and high-end chips may use eiointc-1.0,
> depending on the time it's produced.
> 
> So in the Loongson-2K series I have defined the current state as
> eiointc-1.0, using the dts property to indicate the maximum number of
> devices supported by eiointc that can be used directly in the driver.
> 
> If there are new changes to the design later on, such as the
> definition of registers, we can call it eiointc-2.0, which can also
> cover more than one chip.

Just go with SoC-based compatibles. If your version is not specific
enough, then it is not a good way to represent the hardware.

Best regards,
Krzysztof