Re: [PATCH 2/2] arm64: dts: qcom: sm8650: Add initial Valve Deckard devicetree

[Konrad Dybcio]

On 9/22/26 1:28 PM, Neil Armstrong wrote:
> Add an initial DeviceTree for the Valve "Steam Frame"
> codename Deckard. The Steam Frame is a virtual reality headset
> based on the Qualcomm Snapdragon 8 Gen 3 SoC (SM8650).
> 
> Providing initial support for:
> - System reserved memory, regulators, gpio keys, PMIC functions..
> - PMIC ADC Channels
> - UFS Storage
> - USB Gadget
> - WCN7850 Wifi7 + Bluetooth
> - I2C + basic peripherals
> - SDCard slot
> 
> The bootloader starts Linux in EL2, thus a bunch of EL1 & DSP related
> nodes needs to be removed and the reserved-memory mapping must
> be defined.
> 
> Signed-off-by: Neil Armstrong <neil.armstrong@xxxxxxxxxx>
> ---

[...]

> +/* Remove nodes */
> +/delete-node/ &ipa;
> +/delete-node/ &gpu_zap_shader;
> +/delete-node/ &lpass_wsa2macro;
> +/delete-node/ &lpass_rxmacro;
> +/delete-node/ &lpass_txmacro;
> +/delete-node/ &lpass_wsamacro;
> +/delete-node/ &lpass_vamacro;
> +/delete-node/ &lpass_tlmm;
> +/delete-node/ &remoteproc_adsp;
> +/delete-node/ &remoteproc_cdsp;
> +/delete-node/ &remoteproc_mpss;
> +/delete-node/ &reserved_memory;
> +/delete-node/ &swr0;
> +/delete-node/ &swr1;
> +/delete-node/ &swr2;
> +/delete-node/ &swr3;

Hm, maybe we should think about sm8650-audioreach (simialrly to
how we have a split for the kodiak variants)


> +
> +/ {
> +	model = "Valve Deckard";
> +	compatible = "valve,deckard", "qcom,sm8650";
> +
> +	aliases {
> +		serial0 = &uart15;
> +		serial1 = &uart14;
> +	};
> +
> +	chosen {
> +		stdout-path = "serial0:115200n8";
> +
> +		framebuffer {
> +			compatible = "simple-framebuffer";
> +			width = <4320>;
> +			height = <2160>;
> +			stride = <(4320 * 4)>;
> +			format = "a8r8g8b8";
> +			memory-region = <&framebuffer_mem>;
> +
> +			/* Bootloader will enable the node */
> +			status = "disabled";

Is there a reason not to enable it? Does the bootloader prefer to
only enable it sometiems?

[...]

> +	reserved-memory {
> +		ranges;
> +
> +		#address-cells = <2>;
> +		#size-cells = <2>;

it's usually:

addr
size
ranges
<newline>

[...]

> +
> +		aop_mem: aop-region@81c00000 {
> +			no-map;
> +			reg = <0x0 0x81c00000 0x0 0x60000>;

0.dtsi

> +		};
> +
> +		ramoops: ramoops-region@84a00000 {
> +			compatible = "ramoops";
> +			reg = <0 0x84a00000 0 0x100000>;
> +			record-size = <0x40000>;
> +			console-size = <0x40000>;
> +			pmsg-size = <0x1000>;
> +			ecc-size = <16>;
> +		};
> +
> +		aop_cmd_db_mem: aop-cmd-db-region@81c60000 {
> +			compatible = "qcom,cmd-db";
> +			reg = <0x0 0x81c60000 0x0 0x20000>;
> +			no-map;
> +		};

Likewise many of these reservations also exist there

[...]

> +	/delete-node/ smp2p-adsp;
> +	/delete-node/ smp2p-cdsp;
> +	/delete-node/ smp2p-modem;

I think these shouldn't hurt even if your usecase is different
> +
> +	wcn7850-pmu {
> +		compatible = "qcom,wcn7850-pmu";
> +
> +		pinctrl-names = "default";
> +		pinctrl-0 = <&wlan_en>, <&bt_default>;

nit:

property-n
property-names

please

otherwise looks pretty good

Konrad