Re: [PATCH 2/2] arm64: dts: qcom: ipq5424: Move PHYs and PERST# to Root Port node

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [Critical] Legacy device trees fail to probe with -EBUSY because the 'perst' GPIO descriptor is requested twice without being released.
--

commit f286f42842ba6bb45e869489bbe2830fdd9be93c
Author: Kathiravan Thirumoorthy <kathiravan.thirumoorthy@oss.qualcomm.com>

arm64: dts: qcom: ipq5424: Move PHYs and PERST# to Root Port node

This commit updates the device tree to define PHYs and the PERST# 
(reset-gpios) under the PCIe Root Port node instead of the main controller 
node. This aligns with the new binding style.

> diff --git a/arch/arm64/boot/dts/qcom/ipq5424-rdp466.dts b/arch/arm64/boot/dts/qcom/ipq5424-rdp466.dts
> index de71b72ae6dcc..be8657239c46b 100644
> --- a/arch/arm64/boot/dts/qcom/ipq5424-rdp466.dts
> +++ b/arch/arm64/boot/dts/qcom/ipq5424-rdp466.dts
> @@ -86,8 +86,6 @@ &pcie2 {
>  	pinctrl-0 = <&pcie2_default_state>;
>  	pinctrl-names = "default";
>  
> -	perst-gpios = <&tlmm 31 GPIO_ACTIVE_LOW>;
> -

[Severity: Critical]
This is a pre-existing issue, but does moving these properties expose a
probe failure on legacy device trees?

When parsing a legacy DT, qcom_pcie_parse_ports() successfully requests the
"perst" GPIO:

drivers/pci/controller/dwc/pcie-qcom.c:qcom_pcie_parse_ports() {
...
	pcie->reset = devm_gpiod_get_optional(dev, "perst",
					      GPIOD_OUT_HIGH);
	if (IS_ERR(pcie->reset))
		return PTR_ERR(pcie->reset);
...
}

But it then returns -ENODEV due to the lack of a "phys" property on the child
node. The probe function then falls back to qcom_pcie_parse_legacy_binding(),
which attempts to request the same "perst" GPIO again:

drivers/pci/controller/dwc/pcie-qcom.c:qcom_pcie_parse_legacy_binding() {
...
	reset = devm_gpiod_get_optional(dev, "perst", GPIOD_OUT_HIGH);
	if (IS_ERR(reset))
		return PTR_ERR(reset);
...
}

Because the first request was never released using devm_gpiod_put(), will
the second devm_gpiod_get_optional() fail with -EBUSY and fatally abort the
probe, breaking backwards compatibility for older platforms?

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260602-move_perst_to_rp-v1-0-8ee5b80316da@oss.qualcomm.com?part=2