Re: [PATCH v3 3/3] arm64: qcom: sc7280: Move phy, perst to root port node

[Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>]

On 6/11/25 8:36 AM, Krzysztof Kozlowski wrote:
> On 10/06/2025 15:15, Konrad Dybcio wrote:
>> On 6/2/25 3:01 PM, Krzysztof Kozlowski wrote:
>>> On 08/05/2025 16:26, Konrad Dybcio wrote:
>>>> On 4/23/25 5:37 PM, Rob Herring wrote:
>>>>> On Sat, Apr 19, 2025 at 10:49:26AM +0530, Krishna Chaitanya Chundru wrote:
>>>>>> There are many places we agreed to move the wake and perst gpio's
>>>>>> and phy etc to the pcie root port node instead of bridge node[1].
>>>>>>
>>>>>> So move the phy, phy-names, wake-gpio's in the root port.
>>>>>> There is already reset-gpio defined for PERST# in pci-bus-common.yaml,
>>>>>> start using that property instead of perst-gpio.
>>>>>
>>>>> Moving the properties will break existing kernels. If that doesn't 
>>>>> matter for these platforms, say so in the commit msg.
>>>>
>>>> I don't think we generally guarantee *forward* dt compatibility though, no?
>>> We do not guarantee, comment was not about this, but we expect. This DTS
>>> is supposed and is used by other projects. There was entire complain
>>> last DT BoF about kernel breaking DTS users all the time.
>>
>> Yeah I get it.. we're in a constant cycle of adding new components and
>> later coming to the conclusion that whoever came up with the initial
>> binding had no clue what they're doing..
>>
>> That said, "absens carens".. if users or developers of other projects
>> don't speak up on LKML (which serves as the de facto public square for
>> DT development), we don't get any feedback to take into account when
>> making potentially breaking changes (that may have a good reason behind
>> them). We get a patch from OpenBSD people every now and then, but it's
>> a drop in the ocean.
>>
> I don't understand what you are commenting on. Do you reject what I
> asked for?

If the general consensus among kernel PCIe folks will come down to what
this patch does, I think it's fair to shift to a "correct" hw
description, especially if this is a requirement to resolve a blocker
on functionality (which the author didn't clarify whether is the case)

Konrad