Re: [PATCH v2] dt-bindings: usb: qcom,snps-dwc3: Fix glymur binding for non-MP controllers

[Krzysztof Kozlowski <krzk@kernel.org>]

On 03/09/2026 10:11, Krishna Kurapati wrote:
> 
> 
> On 9/3/2026 12:11 PM, Krzysztof Kozlowski wrote:
>> On 02/09/2026 16:10, Konrad Dybcio wrote:
>>> On 8/31/26 2:29 PM, Krzysztof Kozlowski wrote:
>>>> On 31/08/2026 11:03, Krishna Kurapati wrote:
>>>>>
>>>>>
>>>>> On 8/31/2026 2:07 PM, Krzysztof Kozlowski wrote:
>>>>>> On Fri, Aug 28, 2026 at 01:45:42PM +0530, Krishna Kurapati wrote:
>>>>>>> Currently for Glymur non-multiport USB controllers, number of interrupts
>>>>>>> are limited to a maximum of "5". Either hs_phy_irq or ss_phy_irq are to be
>>>>>>> present.
>>>
>>> [...]
>>>
>>>>>>>                  - qcom,eliza-dwc3
>>>>>>> +              - qcom,glymur-dwc3
>>>>>>>                  - qcom,ipq4019-dwc3
>>>>>>>                  - qcom,ipq8064-dwc3
>>>>>>>                  - qcom,kaanapali-dwc3
>>>>>>
>>>>>> This changes the ABI without an explanation.
>>>>>>
>>>>> As per the harware specification, the controllers do have both
>>>>> hs_phy_irq and ss_phy_irq. Hence wanted the binding to capture that
>>>>> info. Will update the commit message accordingly.
>>>>>
>>>>
>>>> So the change in 99df63d20dabda8d7ae01bcca7cdb1e92110a555 was
>>>> incomplete? Lack of context here is not helping.
>>>>
>>>> If they have these IRQs but it does not matter, then it is not a good
>>>> reason to change ABI and you should just add the clock as optional to
>>>> the end. Was this binding already released? I think yes.
>>>
>>> IIUC the intent of the patch is to allow specifying the currently
>>> missing hs_phy_irq on non-multiport controllers (i.e. usb3_[012]).
>>> Bundling the DT change would have made that easier.
>>
>> This does not answer my concerns. So repeating: patch is changing ABI
>> without an explanation. This means contributor needs to provide
>> explanation for the ABI change in terms of actual impact.
>>
>  From a functional standpoint, there would be no effect. I was trying to 
> make the bindings reflect that the hardware has hs_phy_irq.

If there is no functional impact, so before everything was fine and now
is fine as well, then usual expectations is to make it
backwards-compatible way, so grow the list with minItems.

Best regards,
Krzysztof