Re: [PATCH v2 01/14] dt-bindings: phy: Document the Qualcomm IPQ5332 UNIPHY USB Super-speed PHY

[George Moussalem]

On 8/17/26 16:38, Konrad Dybcio wrote:
> On 8/17/26 2:37 PM, Konrad Dybcio wrote:
>> On 8/13/26 12:48 PM, Krzysztof Kozlowski wrote:
>>> On 13/08/2026 12:28, George Moussalem wrote:
>>>>>> +examples:
>>>>>> +  - |
>>>>>> +    #include <dt-bindings/clock/qcom,ipq5332-gcc.h>
>>>>>> +
>>>>>> +    phy@4b0000 {
>>>>>> +        compatible = "qcom,ipq5332-uniphy-usb-ss-phy";
>>>>>> +        reg = <0x4b0000 0x800>;
>>>>>
>>>>> That is the same address as PCI PHY, so you have one device, not two.
>>>>
>>>> yes, AFAIK they're separate PHYs (pcie-28lp vs usb-ss-22ull) sharing the
>>>> same MMIO register space in IPQ5332. To enable USB3, the compatible
>>>> string, clocks, resets, and related USB3 properties override the PCIe
>>>> properties. In IPQ5018, the PHYs have different register address spaces,
>>>> hence representing it as a new devicetree node.
>>>>
>>>
>>> Same MMIO space is same device, IMO, especially that you have the same
>>> registers (looking at the driver).
>>
>> It's indeed one device, fittingly named "PCIE+USB COMBO PHY"
> 
> so the implementation, matching prior art for similar cases should be:
> 
> - phy node with #phy-cells = <1>
> - consumer takes a reference through phys = <&phy FOO_PCIE> or FOO_USB3
> - new driver in drivers/phy/qualcomm to coordinate the two modes

Thanks for confirming it's one device. I'm already working on a new
version using phy-cells and having the consumer select the phy type.

However: I'm adjusting the existing phy-qcom-uniphy-pcie-28lp.c as
that's the driver currently used for PCIe only in both IPQ5018 and IPQ5332.
So I'm adjusting it to register 2 PHYs and provide both PCIe and USB
functionality. Any objections/concerns?

> 
> Konrad

Best regards,
George