Here you go:
- Download this patch: nvidia-drm-hotplug-workqueue.patch (3.6 KB)
- bash NVIDIA-Linux-x86_64-545.29.06.run --apply-patch nvidia-drm-hotplug-workqueue.patch
- sudo bash NVIDIA-Linux-x86_64-545.29.06-custom.run
3 Likes
From what I can tell the patch seems to work. Did 5 reboots and it didnāt lock up. I also turned of the computer and turn everything back on again.
1 Like
dumb q, but will this hotplug be integrated into the next driver version (for those who donāt want to install via .run)?
1 Like
I had the same issue and I confirm that the patch works.
@aplattner i started getting these flip even timeouts again with 550 for some reason. But the driver seems to be patched now so it got to be something else. The timeouts disappeard with the patch for 545, I had no issues at all.
The issue may cause kde on wayland to freeze an adaptive sync screen 480960 ā Adaptive Sync screen may freeze when multiple display-related settings are changed at the same time
Can you please try with 550.54.14 and then attach a new bug report log if the problem still occurs?
Sure, here you go. Unfortunately it happens with 550.54.14 too.
nvidia-bug-report.log.gz (1.2 MB)
If the module is loaded from initramfs the problem does not occur thoā. So I use that as a āworkaroundā for now. However I removed it from initramfs to make the bug report.
When I enable fbdev=1 to disable simpledrm so that Wayland can run, I now get the following messages with the 550.54.14 drivers:
Feb 24 07:21:09 arch-desktop kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000c00] Failed to apply atomic modeset. Error code: -22
Feb 24 07:21:13 arch-desktop kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000c00] Flip event timeout on head 0
Attempting to log into a Wayland session just results in it crashing back to a frozen TTY console.
If I donāt enable fbdev=1 then Wayland session runs at 1fps.
You can also try turning off your secondary monitor while booting as a temporary fix
I am not using a secondary monitor. I only have one single 4k display.
Oh then it must be something else that triggers it for you. Do a bug report and attach it here
Hi @shelter
I have filed a bug 4513790 internally for tracking purpose.
I will try to duplicate issue locally with driver 550.54.14 and will get back to you if required any additional information.
1 Like
Thanks Amrit and @shelter. I tried installing CachyOS on a system with a similar display setup and wasnāt able to reproduce the problem, but Iāll keep trying.
From your bug report log, it looks like the console modeset is failing:
[ 5.706957] fbcon: nvidia-drmdrmfb (fb0) is primary device
[ 5.708002] [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to apply atomic modeset. Error code: -22
[ 5.708027] Console: switching to colour frame buffer device 240x67
[ 5.708785] nvidia 0000:01:00.0: [drm] fb0: nvidia-drmdrmfb frame buffer device
[ 5.721953] [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to apply atomic modeset. Error code: -22
If you donāt mind, could you please try booting with drm.debug=0xff added to your kernel parameters, reproduce the bug, and then generate a new bug report log? Iām curious why that modeset is failing.
I have had my gpu crash due to wayland in games (Baldurās Gate 3 and Helldivers 2). Initially you see some frames go backwards (fps graph on mangohud goes backwards) and then the entire screen gets pixelated gray squares (but the sound plays) and even persists when quitting the game. In addition in some games (Rogue trader, Baldurs Gate 3, Expeditions Rome) the fps is half of that on xorg. It seems thereās been a regression to some of the fixes from the October drivers that fixed many Wayland issues and made games have similar fps to x11).
Other times the game crashes the desktop. Idk how to post log otherwise I would have.
This is on a 2060 mobile. Archlinux. 545.29.06 drivers.
I am having the same issue.
RTX 4080 Suprim, i9 13900KF
I am using Fedora 39 with latest updates. I do not have a second monitor, I have one single monitor: AW3225QF 4k. I have tried adding fbdev=0 or 1 in grub cmdline, it did not work. Nothing helped currently.
What exactly happening is this: if I completely shutdown the system then boot cold, it USUALLY works fine. I do not get the error (Flip event timeout on head 0)
If I reboot the system, I usually get the error. For example sometimes KDE starts lagging when I move my windows around and I can clearly feel that in UI something is wrong. Especially FPS drops massively or KDE stutters. If I hot reboot, I definitely hit the error flip event time out error. Only way to fix for me is to cold boot and %90 times this works.
Iām dual booting with Win11, and everything works just fine.
I am attaching nvidia-bug-report
nvidia-bug-report.log.gz (347.9 KB)
Wow, took me 10 reboots and a system power off to trigger it again. This got to be a race conditionā¦
Updated log provided, hope it helps.
nvidia-bug-report.log.gz (624.8 KB)
EDIT:
I notied dmesg ran out of buffers, so it got stripped, hereās a full log:
dmesg.log.gz (102.9 KB)
Thanks for sharing the logs, we will review it and get back to you.
Just wanted to update that I also tried on my local setup to reproduce issue but no luck even after multiple restarts, shutdown and suspend/resume operations.
It would be great if someone has reliable repro steps and share with us.
I use these options:
options nvidia NVreg_UsePageAttributeTable=1 NVreg_InitializeSystemMemoryAllocations=0 NVreg_DynamicPowerManagement=0x02 NVreg_EnableGpuFirmware=1
options nvidia_drm modeset=1 fbdev=1
But perhaps they are in the bug report, I never checked. Iān not sure if EnableGpuFirmware actually does anything on the proprietary driver, it gets loaded, thatās all i know.
1 Like