Re: [PATCH v4] arm64: dts: qcom: sm7125-samsung: add initial device tree

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
- [Low] dts: rmtfs_mem node lacks unit address matching reg property
- [Low] dts: virtual i2c-gpio node placed inside memory-mapped soc bus
- [Low] dts: framebuffer node retains unit address without reg property
--

commit 6c9915e6435fc32203fc912dbc6609069cf97bc9
Author: Erikas Bitovtas <xerikasxx@gmail.com>

arm64: dts: qcom: sm7125-samsung: add initial device tree

Samsung Galaxy A52/A72 are devices released on atoll (SM7125) platform
in 2021. Add initial device tree for SM7125 Samsung platform with
support for various basic components.

> diff --git a/arch/arm64/boot/dts/qcom/sm7125-samsung-a52q.dts b/arch/arm64/boot/dts/qcom/sm7125-samsung-a52q.dts
> index 0000000000000..ecba5be8164ea
> --- /dev/null
> +++ b/arch/arm64/boot/dts/qcom/sm7125-samsung-a52q.dts

[ ... ]

> +&rmtfs_mem {
> +	reg = <0x0 0xf4f00000 0x0 0x202000>;
> +};

[Severity: Low]
Does appending this reg property cause a dtc warning about a missing
unit address?

In sm7125-samsung-common.dtsi, the base node is defined without one:

		rmtfs_mem: memory {
			compatible = "qcom,rmtfs-mem";
			qcom,use-guard-pages;

When a node has a reg property, its name should include the corresponding
unit address. Should the base node be updated or the name overridden here
to match the reg address (e.g., rmtfs_mem: memory@f4f00000)?

> diff --git a/arch/arm64/boot/dts/qcom/sm7125-samsung-common.dtsi b/arch/arm64/boot/dts/qcom/sm7125-samsung-common.dtsi
> index 0000000000000..7c5d062bd0809
> --- /dev/null
> +++ b/arch/arm64/boot/dts/qcom/sm7125-samsung-common.dtsi

[ ... ]

> +	chosen {
> +		#address-cells = <2>;
> +		#size-cells = <2>;
> +		ranges;
> +
> +		framebuffer@9c000000 {
> +			compatible = "simple-framebuffer";
> +			memory-region = <&cont_splash_mem>;
> +			width = <1080>;
> +			height = <2400>;
> +			stride = <(1080 * 4)>;
> +			format = "a8r8g8b8";
> +			clocks = <&gcc GCC_DISP_HF_AXI_CLK>;
> +		};
> +	};

[Severity: Low]
Is the @9c000000 unit address appropriate for the framebuffer node since
it relies on memory-region instead of a reg property?

A node must only have a unit address if it has a reg property.

Additionally, should chosen retain the #address-cells, #size-cells,
and ranges properties if it no longer has any children with a reg
property?

[ ... ]

> +&soc {
> +	i2c_17: i2c-17 {
> +		compatible = "i2c-gpio";
> +		sda-gpios = <&tlmm 51 (GPIO_ACTIVE_HIGH|GPIO_OPEN_DRAIN)>;
> +		scl-gpios = <&tlmm 52 (GPIO_ACTIVE_HIGH|GPIO_OPEN_DRAIN)>;
> +
> +		i2c-gpio,delay-us = <2>;

[Severity: Low]
Should this virtual i2c-gpio node be placed at the root level (/)
instead of inside the memory-mapped &soc bus?

The i2c_17 node is a bit-banged software I2C controller without
memory-mapped registers, so it lacks a reg property. dtc will flag a
warning that the node lacks a reg property but is in a bus.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-sm7125-samsung-v4-1-3233901b48bb@gmail.com?part=1