[Bug 221933] New: Screen brightness dimmer since kernel 7.2 (amdgpu)

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=221933

            Bug ID: 221933
           Summary: Screen brightness dimmer since kernel 7.2 (amdgpu)
           Product: Drivers
           Version: 2.5
          Hardware: All
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P3
         Component: Video(DRI - non Intel)
          Assignee: drivers_video-dri@kernel-bugs.osdl.org
          Reporter: 3lefji@gmail.com
        Regression: No

Hi everyone,

Follow-up on a brightness issue I've been having since the kernel 7.2 update
(still present in 7.2.2). Sharing what I found in case it helps others or
points to the root cause.

System info:
- Laptop: Lenovo IdeaPad Gaming 3
- CPU: AMD Ryzen 5 5500H (iGPU handles the internal display backlight)
- GPU: Nvidia RTX 2050 (hybrid AMD/Nvidia, PRIME render offload setup)
- OS: Fedora 44, GNOME 50, Wayland
- Kernel: 7.2.2-cachyos
- Backlight device: amdgpu_bl1

Issue: since kernel 7.2, the same brightness percentage looks much dimmer than
before. I used to sit around 30% in dim rooms, now I need about 60% for the
same visual result. Raw values:
```
cat /sys/class/backlight/amdgpu_bl1/brightness -> 45755
cat /sys/class/backlight/amdgpu_bl1/max_brightness -> 65535
```
So roughly 70% reported but visually much closer to what used to be ~30-35%.
Looks like a mapping/curve change in the amdgpu backlight handling, similar to
reports I found for other kernel versions affecting amdgpu_bl on AMD laptops
(both integrated and hybrid setups).

I tried the common workaround of adding amdgpu.backlight=vendor to
GRUB_CMDLINE_LINUX. Brightness mapping did go back to normal after that change,
BUT it caused severe system instability right after the password screen:
extreme lag, barely able to move the mouse, a strange fade/scanline artifact on
the lock screen, GPU idle at 32°C while CPU sat at 13% doing nothing. Had to
revert the kernel parameter to get a usable system again.

So amdgpu.backlight=vendor is not a safe workaround at least on this hybrid
AMD/Nvidia + Wayland + GNOME 50 combination.

Has anyone found a fix that doesn't cause this kind of freeze, or is this a
case where we just need to wait for an upstream amdgpu fix? Happy to test
patches or provide more logs if useful.

Thanks!

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.