On Mon, Jul 22, 2024 at 08:57:30PM -0600, Rob Herring wrote:
> On Fri, 19 Jul 2024 15:17:15 +0200, Johan Hovold wrote:
> > This series fixes some issues with the current x1e80100 PCIe support,
> > adds the PCIe5 nodes and enables the modem on the CRD.
> >
> > The fixes should go into 6.11, but the modem support depends on them so
> > I decided to send everything in one series.
> > Johan Hovold (7):
> > arm64: dts: qcom: x1e80100-crd: fix PCIe4 PHY supply
> > arm64: dts: qcom: x1e80100: fix PCIe domain numbers
> > arm64: dts: qcom: x1e80100-crd: fix up PCIe6a pinctrl node
> > arm64: dts: qcom: x1e80100-crd: disable PCIe6A perst pull down
> > arm64: dts: qcom: x1e80100-crd: fix missing PCIe4 gpios
> > arm64: dts: qcom: x1e80100: add PCIe5 nodes
> > arm64: dts: qcom: x1e80100-crd: enable SDX65 modem
> >
> > arch/arm64/boot/dts/qcom/x1e80100-crd.dts | 110 +++++++++++++++++--
> > arch/arm64/boot/dts/qcom/x1e80100.dtsi | 125 +++++++++++++++++++++-
> > 2 files changed, 224 insertions(+), 11 deletions(-)
> My bot found new DTB warnings on the .dts files added or changed in this
> series.
> New warnings running 'make CHECK_DTBS=y qcom/x1e80100-crd.dtb' for 20240719131722.8343-1-johan+linaro@xxxxxxxxxx:
>
> arch/arm64/boot/dts/qcom/x1e80100-crd.dtb: pci@1c00000: Unevaluated properties are not allowed ('vddpe-3v3-supply' was unexpected)
> from schema $id: http://devicetree.org/schemas/pci/qcom,pcie-x1e80100.yaml#
I was surprised about this as this property is already used by this DT
in the upcoming 6.11-rc1 (for the NVMe).
I found this thread were Abel tried to add the property to the new
dedicated x1e80100 schema, but received some push back:
https://lore.kernel.org/lkml/20240604235806.GA1903493-robh@xxxxxxxxxx/
Looking at the back story for this, the alternate name 'vpcie3v3' has
been used by some non-Qualcomm controllers since 2016, while Qualcomm DT
has been using 'vddpe-3v3' since 2018. And it's been clearly documented
as part of the bindings the whole time.
Earlier this year, 'vddpe-3v3' was incorrectly removed from the Qualcomm
binding, which results in a checker warnings for a bunch old Qualcomm
DTs. I've just sent a patch to restore this Qualcomm name here:
https://lore.kernel.org/lkml/20240723151328.684-1-johan+linaro@xxxxxxxxxx/
If we want to replace the Qualcomm specific name with the alternate name
then this would need to be done by deprecating the current name, while
adding backward compatibility support for the old name to the driver.
I'm not sure anyone cares enough about this inconsistency to actually
pursue this.
Johan
