ROCK64 | No LED device nodes with Linux 6.12

[sanlupkim]

Creating a bug report/issue

• [ x] I have searched the existing open and closed issues

Required Information

• DietPi version | cat /boot/dietpi/.version G_DIETPI_VERSION_CORE=9 G_DIETPI_VERSION_SUB=14 G_DIETPI_VERSION_RC=2 G_GITBRANCH='master' G_GITOWNER='MichaIng' G_LIVE_PATCH_STATUS[0]='applied'
• Distro version | echo $G_DISTRO_NAME $G_RASPBIAN bookworm
• Kernel version | uname -a Linux rock64-60 6.12.35-current-rockchip64 #1 SMP PREEMPT Fri Jun 27 10:11:46 UTC 2025 aarch64 GNU/Linux
• SBC model | echo $G_HW_MODEL_NAME or (EG: RPi3) ROCK64 (aarch64)
• Power supply used | (EG: 5V 1A RAVpower) Any
• SD card used | (EG: SanDisk ultra) SanDisk 32G

Additional Information (if applicable)

• Software title | (EG: Nextcloud) headless
• Was the software title installed freshly or updated/migrated? upgrade from 9.12 to 9.14.2
• Can this issue be replicated on a fresh installation of DietPi? Yes, it's kernel related (?)

• Bug report ID | echo $G_HW_UUID

Steps to reproduce

Run kernel 6.12. On 6.6.5 leds (red and white) working. krernel 6.6.5:

ls -l /sys/class/leds/
total 0
lrwxrwxrwx 1 root root 0 Jan 21  2016 led-0 -> ../../devices/platform/leds/leds/led-0
lrwxrwxrwx 1 root root 0 Jan 21  2016 led-1 -> ../../devices/platform/leds/leds/led-1

krernel 6.12.35: directory empty. No dir /sys/devices/platform/leds/leds/

Expected behaviour

during kernel boot red blinking (heartbeat), white led - blinking when read/write on SD card.

Actual behaviour

red and white led constantly on. directory /sys/class/leds/ is empty No directory /sys/devices/platform/leds/leds/

Extra details

dmesg error: [ 20.703474] platform leds: deferred probe pending: leds-gpio: Failed to get GPIO '/leds/led-0'

[MichaIng]

Many thanks for your reports. I will have a look next dev cycle. Sources to look at:

• Mainline Linux 6.12 device tree: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/arch/arm64/boot/dts/rockchip/rk3328-rock64.dts?h=linux-6.12.y (the LEDs are defined there)
• Mainline Linux 6.6 device tree: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/arch/arm64/boot/dts/rockchip/rk3328-rock64.dts?h=linux-6.6.y
• Armbian patches for Linux 6.12: https://github.com/armbian/build/tree/main/patch/kernel/archive/rockchip64-6.12
• Armbian patches for Linux 6.6: https://github.com/armbian/build/tree/main/patch/kernel/archive/rockchip64-6.6

[sanlupkim]

I've check latest armbian distro (trixie) with edge-kernels. The conclusion is that bug appeared in 6.12 kernel:

• Linux rock64 6.11.0-edge-rockchip64 #1 SMP PREEMPT Sun Sep 15 14:57:56 UTC 2024 aarch64 GNU/Linux - leds OK
• Linux rock64 6.12.11-current-rockchip64 #1 SMP PREEMPT Thu Jan 23 16:23:05 UTC 2025 - leds NOT working
• latest edge kernel - 6.16.4-edge-rockchip64 - sitll NOT working with message kernel: platform leds: deferred probe pending: leds-gpio: Failed to get GPIO '/leds/led-0' HTH

[MichaIng]

Since I expect LED nodes at a different path, is this empty as well?

ls -l /sys/class/leds/

And do the device tree nodes exist?

ls -l /proc/device-tree/leds/

[sanlupkim]

On kernel > 6.11 /sys/class/leds/ is empty. /proc/device-tree/leds/ exist and looks like no difference between 6.11 and 6.12, i.e.:

 cat /proc/device-tree/leds/led-1/linux,default-trigger
 heartbeat

[MichaIng]

So the device tree nodes exist, associated with a driver via compatible = "gpio-leds"; the same way like on most other SBCs, but the driver fails to set up the sysfs nodes. You pasted a related kernel error:

[ 20.703474] platform leds: deferred probe pending: leds-gpio: Failed to get GPIO '/leds/led-0'

It menions led-0 while you showed the device tree node led-1 only. Does /proc/device-tree/leds/led-0 exist as well and have the expected mmc0 default trigger?

[sanlupkim]

Oh ,sorry led-1 was a example. led-0 also exist and has mmc0 trigger.

# ls -l /proc/device-tree/leds/
total 0
-r--r--r-- 1 root root 10 Sep 13 16:14 compatible
drwxr-xr-x 2 root root  0 Sep 13 16:14 led-0
drwxr-xr-x 2 root root  0 Sep 13 16:14 led-1
-r--r--r-- 1 root root  5 Sep 13 16:14 name

[MichaIng]

Let's see what happens if we disable led-0, since there was no error for led-1:

sudo apt install device-tree-compiler
sudo mkdir -p /boot/overlay-user
cat << '_EOF_' | sudo tee /boot/overlay-user/disable-led-1.dtso > /dev/null
/dts-v1/;
/plugin/;
/ {
	fragment@0 {
		target = <&power_led>;
		__overlay__ {
			status = "disabled";
		};
	};
};
_EOF_
sudo dtc -I dts -O dtb -o /boot/overlay-user/disable-led-1.dtbo /boot/overlay-user/disable-led-1.dtso
G_SUDO G_CONFIG_INJECT 'user_overlays=' 'user_overlays=disable-led-1' /boot/dietpiEnv.txt

[sanlupkim]

(You mean apt install device-tree-compiler.)

Now error is about led-1: [ 22.756108] platform leds: deferred probe pending: leds-gpio: Failed to get GPIO '/leds/led-1'

[MichaIng]

Now error is about led-1:

:( okay, "failed to get GPIO" sounds like something with the gpios properties is wrong.

leds {
	compatible = "gpio-leds";

	power_led: led-0 {
		gpios = <&rk805 1 GPIO_ACTIVE_LOW>;
		linux,default-trigger = "mmc0";
	};

	standby_led: led-1 {
		gpios = <&rk805 0 GPIO_ACTIVE_LOW>;
		linux,default-trigger = "heartbeat";
	};
};

Many other RK3328 SBCs use other GPIO nodes/banks (one of these: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/arch/arm64/boot/dts/rockchip/rk3328.dtsi?h=linux-6.12.y#n1109) and define pinctrl-0 as well. But some do the same: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/arch/arm64/boot/dts/rockchip/rk3328-roc-cc.dts?h=linux-6.12.y#n92

The rk805 node has the gpio-controller property as well, so formally all correct, and a question about it is actually wired. But nothing in all these regards changed within the last years in the device tree, also not the parent SoC one.

[MichaIng]

Does the rk805 node exist, actually?

ls -l /proc/device-tree/i2c@ff160000/pmic@18

[sanlupkim]

Yes, it is

# ls -l /proc/device-tree/i2c@ff160000/pmic@18
total 0
-r--r--r-- 1 root root  4 Sep 13 21:44 '#clock-cells'
-r--r--r-- 1 root root  4 Sep 13 21:44 '#gpio-cells'
-r--r--r-- 1 root root 21 Sep 13 21:44  clock-output-names
-r--r--r-- 1 root root 15 Sep 13 21:44  compatible
-r--r--r-- 1 root root  0 Sep 13 21:44  gpio-controller
-r--r--r-- 1 root root  4 Sep 13 21:44  interrupt-parent
-r--r--r-- 1 root root  8 Sep 13 21:44  interrupts
-r--r--r-- 1 root root  5 Sep 13 21:44  name
-r--r--r-- 1 root root  4 Sep 13 21:44  phandle
-r--r--r-- 1 root root  4 Sep 13 21:44  pinctrl-0
-r--r--r-- 1 root root  8 Sep 13 21:44  pinctrl-names
-r--r--r-- 1 root root  4 Sep 13 21:44  reg
drwxr-xr-x 9 root root  0 Sep 13 21:44  regulators
-r--r--r-- 1 root root  0 Sep 13 21:44  rockchip,system-power-controller
-r--r--r-- 1 root root  4 Sep 13 21:44  vcc1-supply
-r--r--r-- 1 root root  4 Sep 13 21:44  vcc2-supply
-r--r--r-- 1 root root  4 Sep 13 21:44  vcc3-supply
-r--r--r-- 1 root root  4 Sep 13 21:44  vcc4-supply
-r--r--r-- 1 root root  4 Sep 13 21:44  vcc5-supply
-r--r--r-- 1 root root  4 Sep 13 21:44  vcc6-supply
-r--r--r-- 1 root root  0 Sep 13 21:44  wakeup-source

[sanlupkim]

Problem could be related to rk80x I found some differences. On 6.12 (led NOT working) there IS directory /sys/devices/virtual/devlink/i2c:1-0018--platform:leds:

ls -1 /sys/devices/virtual/devlink/
i2c:1-0018--platform:ff100000.syscon:io-domains
i2c:1-0018--platform:ff300000.gpu
i2c:1-0018--platform:ff3c0000.hdmi
i2c:1-0018--platform:ff520000.mmc
i2c:1-0018--platform:ff540000.ethernet
i2c:1-0018--platform:leds
i2c:1-0018--platform:sdmmc-regulator

On kernel 6.11 there are lines in dmesg:

[    2.549651] rk808-regulator rk808-regulator.2.auto: there is no dvs0 gpio
[    2.550368] rk808-regulator rk808-regulator.2.auto: there is no dvs1 gpio
[    2.551003] rk808-regulator rk808-regulator.2.auto: max buck steps per change: 8