Bug: 570.124.04 - freeze on monitor wakeup - "Flip event timeout"

Graphics / Linux Linux
1

Started right after I upgraded to 570.124.04.

Was not present in 570.86.16.

To replicate -

  1. Lock the system with two monitors. Wait until monitors are switched off.
  2. Press any key.
  3. Symptoms: Primary monitor wakes up but system is frozen. The other monitor does not wake up.

I tried Ctrl+Alt+F1 to reboot, unfortunately that did not work either; it froze the system completely (Capslock no longer lit up on keyboard).

System configuration -

GPU: RTX 3090
Monitors: Two
Operating System: Arch Linux
Desktop Environment: KDE (Wayland)
Display Manager: ly (text-based, instead of SDDM)

Journal excerpt -

repeat of this many times -
Mar 02 12:42:05 workstation kwin_wayland[1190]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
Mar 02 12:42:05 workstation kwin_wayland[1190]: kwin_wayland_drm: Please report this at Linux - NVIDIA Developer Forums
Mar 02 12:42:05 workstation kwin_wayland[1190]: kwin_wayland_drm: With the output of ‘sudo dmesg’ and ‘journalctl --user-unit plasma-kwin_wayland --boot 0’
Mar 02 12:42:06 workstation kwin_wayland[1190]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
Mar 02 12:42:06 workstation kwin_wayland[1190]: kwin_wayland_drm: Please report this at Linux - NVIDIA Developer Forums
Mar 02 12:42:06 workstation kwin_wayland[1190]: kwin_wayland_drm: With the output of ‘sudo dmesg’ and ‘journalctl --user-unit plasma-kwin_wayland --boot 0’
Mar 02 12:42:06 workstation kwin_wayland[1190]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
Mar 02 12:42:06 workstation kwin_wayland[1190]: kwin_wayland_drm: Please report this at Linux - NVIDIA Developer Forums
Mar 02 12:42:06 workstation kwin_wayland[1190]: kwin_wayland_drm: With the output of ‘sudo dmesg’ and ‘journalctl --user-unit plasma-kwin_wayland --boot 0’

interspersed sometimes with kernel: [drm:nv_drm_atomic_comit [nvidia_drm]] ERROR -
Mar 02 12:42:35 workstation kwin_wayland[1190]: kwin_wayland_drm: With the output of ‘sudo dmesg’ and ‘journalctl --user-unit plasma-kwin_wayland --boot 0’
Mar 02 12:42:36 workstation kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
Mar 02 12:42:36 workstation kwin_wayland[1190]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
Mar 02 12:42:36 workstation kwin_wayland[1190]: kwin_wayland_drm: Please report this at Linux - NVIDIA Developer Forums

3 Likes
2

Having the same problem. I’ve submitted a bug report to NVIDIA. I’m also getting: nvidia-modeset: ERROR: GPU:0: Error while waiting for GPU progress

Running 6.13.5-arch1-1 with 570.124.04-2 and when I turn off my monitor and turn it back on the screen is black and my second monitor that I didn’t turn off is frozen and the system completely freezes.

nvidia-bug-report.log (2.5 MB)

1 Like
3

I have the same issue when returning back from stand-by. It started happening on 570.124.04, but it was not an issue on 570.86.16.

Here’s my journalctl dump:

kwin_wayland[1353]: kf.windowsystem: static bool KX11Extras::mapViewport() may only be used on X11
kwin_wayland[1353]: kwin_scene_opengl: 0x501: GL_INVALID_VALUE error generated. <levels>, <width> and <height> must be 1 or greater.
kwin_wayland[1353]: kwin_scene_opengl: Invalid framebuffer status:  "GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT"
kwin_wayland[1353]: kwin_scene_opengl: 0x502: GL_INVALID_OPERATION error generated. Framebuffer name must be generated before being bound.
kwin_wayland[1353]: kwin_scene_opengl: 0x506: GL_INVALID_FRAMEBUFFER_OPERATION error generated. Operation is not valid because a bound framebuffer is not framebuffer complete.
kwin_wayland[1353]: kwin_scene_opengl: 0x502: GL_INVALID_OPERATION error generated.
kwin_wayland[1353]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
kwin_wayland[1353]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
kwin_wayland[1353]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
kwin_wayland[1353]: kwin_wayland_drm: The main thread was hanging temporarily!
4

Workaround

It seems like the bug occurs only when two monitors are active.

Given that, here is my very hacky workaround without having to downgrade, until this is fixed.

I created this script to switch to one monitor on screen lock, and reset to my original config on unlock.

#!/bin/bash
# one_monitor_on_screenlock.sh

set -uexo pipefail

dbus-monitor --session "type='signal',interface='org.freedesktop.ScreenSaver',member='ActiveChanged'" | \
while read line; do
    if echo "$line" | grep -q "boolean true"; then
        kscreen-doctor output.DP-1.enable output.DP-4.disable
    elif echo "$line" | grep -q "boolean false"; then
        kscreen-doctor output.DP-4.enable output.DP-4.position.2560,0 output.DP-4.priority.1 output.DP-1.enable
    fi
done

Then I added it as a Login Script in KDE Settings → Autostart.

5

I have the exact same issue (Arch 6.13.5, Hyprland), except I have a single monitor (AW3821DW) and the moment it goes into DPMS ON (from the lock screen) I can’t have it turn on again. Requires a full reboot.

I get the following in the journal:

[drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000a00] Flip event timeout on head 0

I’ve reverted to 570.86.16 which is working properly.

6

I see this issue too, started with 570.124.04, did not happen with previous 570.86.16.
I’m on Gentoo with Linux 6.13.5-gentoo-dist and I have two monitors, 1 TFT connected via Displayport and a TV connected via HDMI. After I turn off all the screens, it seems everything is frozen and I can’t get any of the screens to turn on again. I can however ssh into the system and issue a soft reboot, so it is not actually frozen

7

I am also experiencing this but with three monitors.

[drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 2
[drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to apply atomic modeset.  Error code: -22

kwin_wayland_drm: atomic commit failed: Device or resource busy

kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
8

There is a new version, 570.124.06, in the cuda repository and I don’t have this issue on that version.

I’m sorry this isn’t true. Happened after the third time my monitors woke up. Although I did get this log event instead:

kwin_wayland_drm: No drm events for gpu "/dev/dri/card0" within last 1 second
9

Confirming. Same issue on latest Arch, with same logs on desktop with RTX4070 and one external monitor via DisplayPort.

mar 04 11:47:01 onyx kwin_wayland[751]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
mar 04 11:47:01 onyx kwin_wayland[751]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
mar 04 11:47:01 onyx kwin_wayland[751]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
mar 04 11:47:02 onyx kwin_wayland[751]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
mar 04 11:47:02 onyx kwin_wayland[751]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
mar 04 11:47:02 onyx kwin_wayland[751]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
mar 04 11:47:03 onyx kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
mar 04 11:47:03 onyx kwin_wayland[751]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
mar 04 11:47:03 onyx kwin_wayland[751]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
mar 04 11:47:03 onyx kwin_wayland[751]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
mar 04 11:47:04 onyx kwin_wayland[751]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
mar 04 11:47:04 onyx kwin_wayland[751]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
mar 04 11:47:04 onyx kwin_wayland[751]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
mar 04 11:47:05 onyx kwin_wayland[751]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
mar 04 11:47:05 onyx kwin_wayland[751]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
mar 04 11:47:05 onyx kwin_wayland[751]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
mar 04 11:47:06 onyx kwin_wayland[751]: kwin_wayland_drm: Pageflip arrived after all, 43546ms after the commit
mar 04 11:47:06 onyx kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
mar 04 11:47:06 onyx kwin_wayland[751]: kwin_wayland_drm: atomic commit failed: Permission denied
mar 04 11:47:09 onyx kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
mar 04 11:47:12 onyx kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0