Pi 5 ACT LED: Incorrect mmc0 idle state & inverted brightness control (Kernel 6.12.20+rpt-rpi-2712)

[tome9111991]

Describe the bug

The green ACT LED (controlled via AON_GPIO_09) on the Raspberry Pi 5 exhibits non-standard and unexpected behavior with common triggers and direct sysfs brightness control with Kernel 6.12.20+rpt-rpi-2712 (and potentially other recent versions).

Issue 1: mmc0 Trigger Idle State

• When dtparam=act_led_trigger=mmc0 is set, the trigger correctly becomes active ([mmc0]).
• The LED correctly blinks during SD card activity.
• Problem: The LED remains solid green when the SD card is idle, contrary to the default-state = "off" set in the device tree.

Issue 2: cpu Trigger State

• When dtparam=act_led_trigger=cpu is set, the trigger correctly becomes active ([cpu]).
• Problem: The LED remains permanently off, even under high CPU load (e.g., sha256sum /dev/zero).

Issue 3: Manual Brightness Control Inverted

• When the trigger is set to none ([none] confirmed active), direct control via sysfs shows inverted logic:
  • echo 1 > /sys/class/leds/ACT/brightness turns the LED OFF.
  • echo 0 > /sys/class/leds/ACT/brightness turns the LED ON.

Conclusion/Hypothesis:
The inverted manual brightness control strongly suggests a driver bug related to handling the GPIO_ACTIVE_LOW property for AON_GPIO_09. This likely causes the incorrect idle state for mmc0 (brightness 0 interpreted as ON) and the lack of output for cpu (brightness 1 interpreted as OFF).

Steps to reproduce the behaviour

1. Use a Raspberry Pi 5 with a recent Raspberry Pi OS (Debian Bookworm base) and Kernel (e.g., 6.12.20+rpt-rpi-2712).
2. Edit /boot/firmware/config.txt and ensure any existing dtparam=act_led_... lines are commented out.
3. Test mmc0: Add dtparam=act_led_trigger=mmc0. Save and reboot. Observe LED is solid green when idle. Check with cat /sys/class/leds/ACT/trigger (shows [mmc0]).
4. Test cpu: Change the line to dtparam=act_led_trigger=cpu. Save and reboot. Observe LED is permanently off, even under load (e.g., sha256sum /dev/zero). Check with cat /sys/class/leds/ACT/trigger (shows [cpu]).
5. Test Manual Control: Change the line to dtparam=act_led_trigger=none. Save and reboot. Confirm trigger is [none] with cat /sys/class/leds/ACT/trigger. Then execute:
   • echo 1 | sudo tee /sys/class/leds/ACT/brightness -> Observe LED turns OFF.
   • echo 0 | sudo tee /sys/class/leds/ACT/brightness -> Observe LED turns ON.

Relevant Device Tree Snippet from bcm2712-rpi-5-b.dts (rpi-6.12.y branch):

       leds: leds {
           compatible = "gpio-leds";

           // [...] PWR LED definition omitted

           led_act: led-act { 
               label = "ACT";
               gpios = <&gio_aon 9 GPIO_ACTIVE_LOW>; 
               default-state = "off";
               linux,default-trigger = "mmc0";
           };
       };

Device (s)

Raspberry Pi 5

System

cat /etc/rpi-issue
Raspberry Pi reference 2024-11-19
Generated using pi-gen, https://github.com/RPi-Distro/pi-gen, 891df1e21ed2b6099a2e6a13e26c91dea44b34d4, stage2

vcgencmd version
2024/09/23 14:02:56
Copyright (c) 2012 Broadcom
version 26826259 (release) (embedded)

uname -a
Linux raspberrypi 6.12.20+rpt-rpi-2712 #1 SMP PREEMPT Debian 1:6.12.20-1+rpt1~bpo12+1 (2025-03-19) aarch64 GNU/Linux

[pelwell]

Are you saying that this behaviour has changed since 6.6?

[tome9111991]

not sure i have only test 6.12

[pelwell]

OK, thanks - no need to try 6.6.

[tome9111991]

dtparam=act_led_activelow=off

fix this issue

[pelwell]

Issue 1: This is expected behaviour - the green LED is doing dual duty as an indication that the power is on and that there is SD card activity. This is the same behaviour as found on Pi Zero et al. If you don't like, you've found a way to invert the logic.
Issue 2: It looks as though only 32-bit (ARCH=arm) builds use the ledtrig_cpu call that is necessary for the CPU trigger to work.
Issue 3: This is a side effect of the inversion for issue 1. There is no "inverted activity" trigger, so it is implemented by lying about the sense of the LED enable, which in turn turns the brightness upside down. It's not ideal, but it's not accidental.

[tome9111991]

Thank you for the detailed explanation! I understand the reasoning behind the intended Pi Zero-like dual-duty behavior for the ACT LED (solid green for power/idle, blinking off for activity) and the resulting inverted manual brightness control as a side effect.

Regarding the mmc0 trigger and the manual control inversion: I tested setting both dtparam=act_led_trigger=mmc0 AND dtparam=act_led_activelow=off in config.txt.
With this specific combination, the ACT LED now behaves in the 'traditional' way: it is correctly OFF when the SD card is idle and blinks (turns briefly ON) during I/O activity. The manual brightness control (with trigger none) also works non-inverted (1=ON, 0=OFF) when act_led_activelow=off is set.

So, it seems that adding dtparam=act_led_activelow=off not only corrects the manual brightness control logic but also restores the 'off-at-idle' behavior for the mmc0 trigger on my RPi 5 / Kernel 6.12.x system. This behavior is preferable for my use case as it provides a clearer distinction between idle and activity states. Thanks again for the insights!