linux-kernel - Re: [PATCH 8/8] arm64: dts: allwinner: a523: Add NPU device node

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20250905161451.0d9615f6@donnerap>
Date: Fri, 5 Sep 2025 16:14:51 +0100
From: Andre Przywara <andre.przywara@....com>
To: Chen-Yu Tsai <wens@...nel.org>
Cc: Rob Herring <robh@...nel.org>, Krzysztof Kozlowski <krzk+dt@...nel.org>,
 Conor Dooley <conor+dt@...nel.org>, Stephen Boyd <sboyd@...nel.org>,
 Chen-Yu Tsai <wens@...e.org>, Jernej Skrabec <jernej@...nel.org>, Samuel
 Holland <samuel@...lland.org>, linux-sunxi@...ts.linux.dev,
 linux-clk@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
 devicetree@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 8/8] arm64: dts: allwinner: a523: Add NPU device node

On Sun, 31 Aug 2025 01:09:01 +0800
Chen-Yu Tsai <wens@...nel.org> wrote:

Hi,

> From: Chen-Yu Tsai <wens@...e.org>
> 
> The Allwinner T527 SoC has an NPU built in. Based on identifiers found
> in the BSP, it is a Vivante IP block. After enabling it, the etnaviv
> driver reports it as a GC9000 revision 9003.
> 
> The standard bindings are used as everything matches directly. There is
> no option for DVFS at the moment. That might require some more work,
> perhaps on the efuse side to map speed bins.
> 
> It is unclear whether the NPU block is fused out at the hardware level
> or the BSP limits use of the NPU through software, as the author only
> has boards with the T527.

I happen to only have boards without the NPU, one A523, two A527s, one
T527, but the SKU without the NPU, and a H728.
So I can confirm that the clock gates and resets exist, but the whole NPU
MMIO frame behaves as read-as-zero/write ignore. At least it doesn't
crash, and the Linux driver just skips this NPU as it cannot identify it
(with all the ID registers being 0).

So I think it's fine to have this node in all the DTBs. We *could* have
something in U-Boot that probes for this RAZ/WI behaviour and slaps a
status = "disabled"; on it. In which case it might be beneficial to have a
status node in already. But I'd rather avoid the churn and reliance on
firmware, instead try to auto detect as much as possible.

> Signed-off-by: Chen-Yu Tsai <wens@...e.org>

Matches the binding and the manual:

Reviewed-by: Andre Przywara <andre.przywara@....com>

Cheers,
Andre

> ---
>  arch/arm64/boot/dts/allwinner/sun55i-a523.dtsi | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 
> diff --git a/arch/arm64/boot/dts/allwinner/sun55i-a523.dtsi b/arch/arm64/boot/dts/allwinner/sun55i-a523.dtsi
> index b6e82d53af54..1ab5b87ec78e 100644
> --- a/arch/arm64/boot/dts/allwinner/sun55i-a523.dtsi
> +++ b/arch/arm64/boot/dts/allwinner/sun55i-a523.dtsi
> @@ -850,6 +850,18 @@ mcu_ccu: clock-controller@...2000 {
>  			#clock-cells = <1>;
>  			#reset-cells = <1>;
>  		};
> +
> +		npu: npu@...2000 {
> +			compatible = "vivante,gc";
> +			reg = <0x07122000 0x1000>;
> +			interrupts = <GIC_SPI 199 IRQ_TYPE_LEVEL_HIGH>;
> +			clocks = <&mcu_ccu CLK_BUS_MCU_NPU_ACLK>,
> +				 <&ccu CLK_NPU>,
> +				 <&mcu_ccu CLK_BUS_MCU_NPU_HCLK>;
> +			clock-names = "bus", "core", "reg";
> +			resets = <&mcu_ccu RST_BUS_MCU_NPU>;
> +			power-domains = <&ppu PD_NPU>;
> +		};
>  	};
>  
>  	thermal-zones {