Re: [PATCH v11 4/7] arm64: dts: qcom: ipq9574: Reorder reg and reg-names

[Varadarajan Narayanan <quic_varada@quicinc.com>]

On Thu, Mar 06, 2025 at 12:56:02PM +0100, Krzysztof Kozlowski wrote:
> On 06/03/2025 12:49, Krzysztof Kozlowski wrote:
> > On 20/02/2025 10:42, Varadarajan Narayanan wrote:
> >> The 'reg' & 'reg-names' constraints used in the bindings and dtsi are
> >> different resulting in dt_bindings_check errors. Re-order the reg entries,
> >
> > Why?

Initially ipq9574 had 5 reg entries. ipq5332 has 6. To be able to use ipq9574 as
fallback for ipq5332 had to add the sixth entry to ipq9574. Then it becomes
similar to sdx55. Hence to avoid duplication, changed ipq9574 to use sdx55 reg
definition. Because of this the erg entries' order changed.

> >
> >> fix the node names and move the nodes to maintain sort order to address the
> >
> > Fixing (how?) node name looks like separate problem.

Because the reg entries order changed, the "parf" register became the first
entry. This resulted in the address in pcie@xxx to not match with the first reg
entry and this was changed. Since the nodes have to be located per address sort
order, had to move the node to an appropriate slot per the address sort order.

> >> following errors/warnings.
> >>
> >> 	arch/arm64/boot/dts/qcom/ipq9574-rdp449.dtb: pcie@20000000: reg-names:0: 'parf' was expected
>
> How can I reproduce this error?
>
> Isn't this error which you intentionally added and now you claim you
> fix? In the same patchset?
>
> This really looks like breaking things just to call it "look, I fixed
> something" two patches later in the same set.

True. But had to do these to have ipq9574 as fallback compatible. Have asked for
suggestions to handle this better. Will follow the approach that is acceptable
to the community.

Thanks
Varada