Re: [PATCH] dt-bindings: mmc: move compatible property to its specific binding

[<Dharma.B@microchip.com>]

Hi Martin,

On 08/01/25 2:04 am, Martin Blumenstingl wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> 
> Hi Dharma,
> 
> On Tue, Jan 7, 2025 at 4:34 AM <Dharma.B@microchip.com> wrote:
>>
>> On 20/12/24 1:41 am, Conor Dooley wrote:
>>> On Thu, Dec 19, 2024 at 09:40:41AM +0530, Dharma Balasubiramani wrote:
>>>> Move the `compatible` property into its specific binding to make the MMC
>>>> slot more generic and modular.
>>> This makes no sense, as presented. What's the real reason for this
>>> change? You want to ref mmc-slot.yaml but the compatible is causing a
>>> driver to probe?
>>
>> We don’t have the configuration for that driver enabled. Wouldn’t
>> including the compatible in the DTS files without the actual driver be
>> redundant?
>>
>> Is it the correct approach to add the compatible just to fix the dt
>> binding errors?
> Let me try to summarize what I understand so far:
> - your are trying to convert the dt-binding of atmel-hsmci from .txt to .yaml
> - while doing so Rob asked to reference the mmc-slot schema
> - after referencing the mmc-slot schema you now get warnings in when
> validating the .dts because your .dts doesn't specify compatible =
> "mmc-slot"
> 
> Is that correct?

Yes.
> 
> There aren't many MMC controllers with multiple slot support out there.
> When I wrote the dt-bindings for amlogic,meson-mx-sdio I *think* (it's
> been some years) Ulf pointed out another dt-binding
> (Documentation/devicetree/bindings/mmc/cavium-mmc.txt) and driver
> (drivers/mmc/host/cavium-thunderx.c) that already used the mmc-slot
> compatible string.
> 
>> related discussion:
>> https://lore.kernel.org/lkml/63473475-f29e-4a65-a0aa-1f1e4112b57d@microchip.com/
> Rob has suggested two approaches in that thread:
> - don't mark the "compatible" property as required (in
> Documentation/devicetree/bindings/mmc/mmc-slot.yaml)
> - add the compatible string where needed (I attached a diff with an
> example where I picked one random Atmel board and added the compatible
> string)
> 
> Your patch is different from these suggestions as it forbids the
> compatible property in the generic mmc-slot binding.
> What's the problem with Rob's suggestions? If they cannot be
> implemented then please document why that is.

Thanks for the comprehensive explanation.

I misinterpreted the Rob's suggestion [1].

"One issue is 'compatible' is required. Either that would have to be 
dropped as required."

Instead of just dropping it from "required:", I removed the property 
itself and moved it to another binding.

I will send a v2 by removing it from the required, will it be fine?

> 
> 
> Best regards,
> Martin


-- 
With Best Regards,
Dharma B.