Re: [PATCH v2] arm64: dts: rockchip: add pwm-fan for NanoPC-T6

[Hugh Cole-Baker <sigmaris@gmail.com>]

Hello Dragan,

I haven't been able to test patches for a while, but now I have access to
the NanoPC board again, and I've done some tests:

On 10/11/2025 00:23, Dragan Simic wrote:
> Hello Hugh,
> 
> Thanks for the v2!  Please, see a couple of nitpicks below.
> 
> On Sunday, November 09, 2025 20:20 CET, Hugh Cole-Baker <sigmaris@gmail.com> wrote:
>> FriendlyELEC offers an optional heatsink and fan addon for the NanoPC-T6
>> and T6 LTS, which plugs in to the fan connector on the board driven by
>> pwm1. Add the fan as an active cooling device for the SoC package.
>>
>> Signed-off-by: Hugh Cole-Baker <sigmaris@gmail.com>
>> ---
>> Changes from v1: https://lore.kernel.org/linux-rockchip/20251026194858.92461-1-sigmaris@gmail.com/
>> * add the fan to the base board dtsi instead of overlay (Heiko)
>> * just use 2 trip points for warm and hot temperatures (Dragan, Alexey)
>>
>> References:
>> FriendlyELEC heatsink with fan addon:
>> https://www.friendlyelec.com/index.php?route=product/product&product_id=305
>> Vendor DT with trip points and PWM duty cycle values:
>> https://github.com/friendlyarm/kernel-rockchip/blob/4944602540b62f5aad139fe602a76cf7c3176128/arch/arm64/boot/dts/rockchip/rk3588-nanopi6-rev01.dts#L75-L90
> 
> I think it would be better to move these references to the patch
> description, so they become directly available in the repository.
> It might be the best to use the "... [n]" form for the references,
> which puts them as close to the backed contents as possible.
> 
> Oh, and I think that "arm64: dts: rockchip: Enable active cooling
> on NanoPC-T6" as the patch subject would read nicer. :)
> 
>>  .../boot/dts/rockchip/rk3588-nanopc-t6.dtsi   | 39 +++++++++++++++++++
>>  1 file changed, 39 insertions(+)
>>
>> diff --git a/arch/arm64/boot/dts/rockchip/rk3588-nanopc-t6.dtsi b/arch/arm64/boot/dts/rockchip/rk3588-nanopc-t6.dtsi
>> index fafeabe9adf9e..9164a0ee6228e 100644
>> --- a/arch/arm64/boot/dts/rockchip/rk3588-nanopc-t6.dtsi
>> +++ b/arch/arm64/boot/dts/rockchip/rk3588-nanopc-t6.dtsi
>> @@ -11,6 +11,7 @@
>>  #include <dt-bindings/input/input.h>
>>  #include <dt-bindings/pinctrl/rockchip.h>
>>  #include <dt-bindings/soc/rockchip,vop2.h>
>> +#include <dt-bindings/thermal/thermal.h>
>>  #include <dt-bindings/usb/pd.h>
>>  #include "rk3588.dtsi"
>>  
>> @@ -89,6 +90,14 @@ usr_led: led-1 {
>>  		};
>>  	};
>>  
>> +	fan: pwm-fan {
>> +		compatible = "pwm-fan";
>> +		cooling-levels = <0 35 64 100 150 255>;
>> +		fan-supply = <&vcc5v0_sys>;
>> +		pwms = <&pwm1 0 50000 0>;
>> +		#cooling-cells = <2>;
>> +	};
>> +
>>  	sound {
>>  		compatible = "simple-audio-card";
>>  		pinctrl-names = "default";
>> @@ -591,6 +600,36 @@ &i2s6_8ch {
>>  	status = "okay";
>>  };
>>  
>> +&package_thermal {
>> +	polling-delay = <1000>;
>> +
>> +	trips {
>> +		package_warm: package-warm {
>> +			temperature = <50000>;
>> +			hysteresis = <2000>;
>> +			type = "active";
>> +		};
>> +
>> +		package_hot: package-hot {
>> +			temperature = <60000>;
>> +			hysteresis = <2000>;
>> +			type = "active";
>> +		};
>> +	};
> 
> It should be better to use 55 and 65 oC as the trip thresholds,
> because people often report around 50 oC as the observed idle-state
> temperature of their RK3588 SoCs, so increasing the first threshold

Were these people using the heatsink accessory on the NanoPC-T6, or were
they using some other board with worse heat dissipation? I recorded the
package temperature on my NanoPC-T6 with the FriendlyELEC heatsink and fan
and even though it's not idle but actually running Home Assistant,
OpenLDAP, PostgreSQL, Prometheus, Grafana and Nginx, the temperature ranges
between 46 and 48°C, and the fan never even spins up.

Then I ran stress-ng and recorded the temperature under CPU load, with this
version of the patch and another version with 55 & 65°C trip points. [1]

I suspect if people are seeing 50°C at idle, they're either using a less
effective heatsink, or in a very warm ambient temperature. When talking
about the NanoPC-T6 fan specifically, we kinda have to assume the
FriendlyELEC heatsink is used, as there's no mounting for the fan without
the heatsink.

> to 55 oC should be beneficial by preventing the fan from spinning
> when the SoC is actually idle.  The second threshold is usually set
> to be 10 oC higher, so it should end up at 65 oC.

I'll send a v3 with 55 and 65°C trip points, but I doubt it makes any
difference to the fan activating while idle; from my testing it seems to
just let the SoC get a bit hotter under heavy load.

[1]: https://gist.github.com/sigmaris/2d5590271cf26da8ec6cfc7ef8e3e8bc

>> +	cooling-maps {
>> +		map0 {
>> +			trip = <&package_warm>;
>> +			cooling-device = <&fan THERMAL_NO_LIMIT 1>;
>> +		};
>> +
>> +		map1 {
>> +			trip = <&package_hot>;
>> +			cooling-device = <&fan 2 THERMAL_NO_LIMIT>;
>> +		};
>> +	};
>> +};
>> +
>>  &pcie2x1l0 {
>>  	reset-gpios = <&gpio4 RK_PB3 GPIO_ACTIVE_HIGH>;
>>  	vpcie3v3-supply = <&vcc_3v3_pcie20>;
>