Created attachment 295535 [details] dmesg In a recent update, the values of /sys/class/backlight/intel_backlight/brightness no longer correspond with my screen brightness. Steps to reproduce: Change the value of /sys/class/backlight/intel_backlight/brightness to an integer from 0-255 (where 255 is the max_brightness). Expected results: At value 0, the screen is off, and at values 1-255, the screen gradually goes from dim to full brightness. Actual results: At value 0, the screen is off, and at values 1-255, the screen is the same brightness. My system: My system requires the i915.enable_dpcd_backlight=1 kernel parameter for this feature to work. But I performed a normal Arch system upgrade to 5.11.2-arch1-1 on 2/21/2021, which has caused the behavior above. My linux version before the upgrade was 5.10.16.arch1-1. I tried many values of the other kernel parameters acpi_backlight but they did not resolve anything. My hardware is Dell XPS 17 9700 with an NVIDIA Optimus graphics hardware. The optimus setup uses i915 for the intel modesetting and nvidia for the main nvidia setting.
Created attachment 295537 [details] systemd boot config with linux kernel parameters
Created attachment 295539 [details] Xorg.conf when in optimus modesetting i915 mode
Created attachment 295541 [details] Xorg.conf when in nvidia mode
Update: I put version 5.10.16.arch1-1 as the kernel that was working, but I cannot guarantee this is true. It may have broken sooner than this, but it was certainly in the month of February, during which I upgraded several times starting from 5.10.11.
Created attachment 295543 [details] lspci -k
can you check both vanilla 5.10 and 5.11 kernel and see if this is an upstream regression?
If this is an upstream regression, it is probably a graphics problem. It would be great if you can use git bisect to find out the offending commit.
Thanks Zhang, I will check this weekend.
Confirmed after downgrading only the kernel package that the problem does not exist in 5.10.16. I will git bisect it this weekend.
Hi Zhang. You were right. It came from the intel-drm project. I singled out the offending commit to be 98e497e203a5, "[1/2] drm/i915/dpcd_bl: uncheck PWM_PIN_CAP when detect eDP backlight capabilities". I have found a similar bug report and will tag my information there. I have updated the URL for this bug to be that bug report. I also noticed that the kernel is well into the 5.12 branch and the affected codefile has gone through a lot of refactoring, so I'm going to rebuild that branch and see if they have already fixed it.
Not fixed in 5.12 yet. Anywho feel free to close this bug as a dupe.
Thanks for update. Bug closed.