Re: [PATCH RFC] dt-bindings: thunderbolt: Add Qualcomm USB4 Host Router

[Konrad Dybcio]

On 9/18/25 5:23 PM, Mika Westerberg wrote:
> Hi,
> 
> On Thu, Sep 18, 2025 at 11:05:42AM +0200, Konrad Dybcio wrote:
>>> I it hard to change these DT bindings later on? If yes then I would
>>> definitely think forward and make this support MSI from the get-go.
>>
>> dt-bindings (attempt to) promise an ABI-like interface, so bindings
>> for *a given IP block* ("dt-bindings describe the hardware") should
>> not change, unless there's something critically wrong (e.g. "this
>> could have never really worked").
> 
> Then I think it is good to think few steps forward and make sure when
> Qualcomm adds MSI to their IP it can be easily desribed in the DT bindings.

Your reply makes me believe I glossed over an important detail..

If qcom,x1e80100-hr doesn't have MSIs, qcom,nextsoc-hr can add them
without any issues too, just that the X1E80100 binding can't be altered,
e.g. "binding" refers to the (compatible, allowed_props, required_props)
tuple

Konrad
> 
>> Adding new properties is always OK, marking the new properties  as
>> 'required' is not (unless it falls into the aforementioned case).
>>
>> It's also totally OK to add MSI properties to e.g. Apple Host Router
>> bindings specifically when they come around, as it's simply a different
>> piece of hardware. It's also OK to create a usb4-host-router.yaml down
>> the line, which will act as a common include and perform any
>> maintenance/code churn, so long as it doesn't end up in the bindings
>> for any specific hw block (e.g. this QC one) becoming more strict
>> than they were on HEAD^.
> 
> Okay thanks for the explanation.