Re: [PATCH 0/2] dt-bindings: ufs: qcom: Split SC7280 and similar into separate file

[Krzysztof Kozlowski <krzk@kernel.org>]

On 30/07/2025 15:53, Nitin Rawat wrote:
> 
> 
> On 7/30/2025 6:05 PM, Krzysztof Kozlowski wrote:
>> The binding for Qualcomm SoC UFS controllers grew and it will grow
>> further.  It already includes several conditionals, partially for
>> difference in handling encryption block (ICE, either as phandle or as IO
>> address space) but it will further grow for MCQ.
>>
>> See also: lore.kernel.org/r/20250730082229.23475-1-quic_rdwivedi@quicinc.com
>>
>> The question is whether SM8650 and SM8750 should have their own schemas,
>> but based on bindings above I think all devices here have MCQ?
>>
>> Best regards,
>> Krzysztof
>>
> 
> 
> Hi Krzysztof,
> 
> If I understand correctly, you're splitting the YAML files based on MCQ 
> (Multi-Circular Queue) support:

Not entirely, I don't know which devices support MCQ. I split based on
common parts in the binding.

> 
> -qcom,sc7280-ufshc.yaml includes targets that support MCQ
> -qcom,ufs-common.yaml includes common properties
> -qcom,ufs.yaml includes targets that do not support MCQ
> 
> 
> In future, if a new property applies to both some MCQ and some
> non-MCQ targets, we would need to update both YAML files. In the current 

No

> implementation, we handle such cases using if-else conditions to include 
> the new property.

Hm?

> 
> For reference, only SM8650 and SM8750 currently support MCQ, though more 
> targets may be added later.

Are you sure? Are you claiming that SM8550 hardware does not support MCQ?

> 
> Regarding the patch 
> lore.kernel.org/r/20250730082229.23475-1-quic_rdwivedi@quicinc.com, 
> instead of using two separate YAML files, we could use if-else 
> conditions to differentiate the reg and reg-name properties between MCQ 
> targets (SM8650 and SM8750) and non-MCQ targets (all others).

It's a mess already and you want to make it messy. I already responded
on that.

Best regards,
Krzysztof