Re: [PATCH 4/5] dt-bindings: mips: Document mti,mips-cm

["Jiaxun Yang" <jiaxun.yang@flygoat.com>]

在2024年5月7日五月 下午5:50，Conor Dooley写道：
> On Tue, May 07, 2024 at 10:01:52AM +0100, Jiaxun Yang wrote:
>> Add devicetree binding documentation for MIPS Coherence Manager.
>> 
>> Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
>> ---
>>  .../devicetree/bindings/mips/mips-cm.yaml          | 37 ++++++++++++++++++++++
>>  1 file changed, 37 insertions(+)
>> 
>> diff --git a/Documentation/devicetree/bindings/mips/mips-cm.yaml b/Documentation/devicetree/bindings/mips/mips-cm.yaml
>> new file mode 100644
>> index 000000000000..b92b008d7758
>> --- /dev/null
>> +++ b/Documentation/devicetree/bindings/mips/mips-cm.yaml
Hi Cornor,

Thanks for your comments.

>
> Filename matching the compatible please.
Ok.

>
>> @@ -0,0 +1,37 @@
>> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
>> +%YAML 1.2
>> +---
>> +$id: http://devicetree.org/schemas/mips/mips-cm.yaml#
>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
>> +
>> +title: MIPS Coherence Manager
>> +
>> +description: |
>> +  Defines a location of the MIPS Coherence Manager registers.
>> +
>> +maintainers:
>> +  - Jiaxun Yang <jiaxun.yang@flygoat.com>
>> +
>> +properties:
>> +  compatible:
>> +    const: mti,mips-cm
>
> Is it actually only available on mips? Google seems to report there
> being Coherence Managers on their RISC-V offerings too.

I think for MIPS's RISC-V system, it is only used by SBI and transparent
to kernel, so it won't present in DT. 

Register fields for RISC-V system is totally different with MIPS one, and
there is no driver to be reused. In MIPS system CM code is highly coupled
with arch code, so for RISC-V if we want to expose it to kernel we'll need
a new set of driver and a new binding.

>
>> +  reg:
>> +    description: |
>
> The | isn't needed, there's no formatting to preserve.
Ok.

>
>> +      Base address and size of an unoccupied memory region, which will be
>> +      used to map the MIPS CM registers block.
>
> This sounds like it should actually be a memory-region that references
> some reserved memory, not a reg, given the description. I think the
> commit message here is lacking any information about what the intentions
> are for this binding.
So it's actually a register block that can be remapped to anywhere in
MMIO address space. DeviceTree usually passes firmware's mapping location
to kernel.

There are some other similar bindings like mti,mips-cdmm and mti,mips-cpc,
I just copied phraseology from them, should I try to explain it more here?

>
>> +    maxItems: 1
>> +
>> +required:
>> +  - compatible
>> +  - reg
>> +
>> +additionalProperties: false
>> +
>> +examples:
>> +  - |
>> +    cm@1fbf8000 {
>
> And a generic node name here please. I actually don't quite know what to
> suggest though, but "coherency-manager" would likely be better than
> "cm".
Ok

Thanks!
- Jiaxun
>
> Thanks,
> Conor.
>
>> +      compatible = "mti,mips-cm";
>> +      reg = <0x1bde8000 0x8000>;
>> +    };
>> +...
>> 
>> -- 
>> 2.34.1
>> 
>
> 附件:
> * signature.asc

-- 
- Jiaxun