You are not logged in.
I apologize if my description is not easy to understand. But I am not sure how to describe the problem, other than to say that it looks like the screen image has slid about half an inch to the left, and then quickly goes back to its regular place. This starts out as soon as I boot up, even before it gets to graphical screen.
I noticed this with 5.15.10.arch1-1. But if I boot with 5.10.87-1-lts the screen is fine.
[I also tried to boot with an Alpine linux 5.15 kernel, and see the same screen shifting/shaking on my terminal. So I'm guessing this is a problem with the kernel.]
When booted with the problematic Arch 5.15.10 kernel, dmesg shows the following error (that is not present with the lts kernel)
i915 0000:00:02.0: [drm] *ERROR* CPU pipe A FIFO underrun
So far I have tried the following things
1) adding intel_idle.max_cstate=4 to kernel command line
2) adding i915.enable_psr=0 to kernel command line.
Neither of these (or together) does not change the shaking/shifting.
The laptop monitor is running at 2160x1440 60.00*+
Any thoughts will be appreciated.
Last edited by serat (2021-12-21 16:00:55)
Offline
Which graphics card/gpu/cpu generation exactly? E.g. it works fine on my HD 630/Skylake laptop
From the symptoms could also be a modesetting race anomaly, in which case the fact that you didn't see this (yet) on the LTS kernel doesn't have to mean much since this will be up to random chance. If you want to lower that risk, try explicitly setting up https://wiki.archlinux.org/title/Kernel … _KMS_start
However there where some frame buffer changes in 5.15 that had quite a fallout in the earlier revisions so maybe some remnants of that.
Last edited by V1del (2021-12-21 08:28:02)
Offline
You could also try to disable https://wiki.archlinux.org/title/Intel_ … nable_fbc)
Offline
@V1del:
I doubt it is a race problem. With LTS kernel I have never seen it, and with the 5.15 kernel, I always see it. So probably not a random occurrence.
@seth:
"enable_fbc = -1" according to systool -m i915 -av
This is a gemini lake laptop.
The weird thing is that "cat /sys/module/intel_idle/parameters/max_cstate" for both the working LTS and non-working 5.15 kernel showed 9
But if I set kernel parameter intel_idle.max_cstate=2 for the 5.15 kernel, then it stops with its slide/shake behavior. No idea why this setting works or is needed, but only for 5.15 kernel.
Anyway, to summarize, setting kernel parameter intel_idle.max_cstate to something less than or equal 2 makes the laptop stop the slide/shake. But I'm guessing that this has power usage implications?
Offline
"-1" is "auto", the idea is to disable it ("0")
Offline
Thank you, @seth.
I tried enable_fbc=0, but it does not work and still keeps shaking/sliding.
The only way to make it usable was to add intel_idle.max_cstate=2 to command line, which for some reason is needed 5.15 but not for the LTS kernel.
I'll make it as solved, and maybe it can help someone in the future.
Offline