KWin Wayland "Applying output configuration failed" / "Failed to open drm node: Permission denied" on every S3 resume — Optimus RTX 3070 Ti Laptop GPU

Graphics / Linux Linux
1

System info

  • Laptop: Razer Blade 15 Advanced (Early 2022)
  • GPU: NVIDIA GeForce RTX 3070 Ti Laptop GPU (GA104) + Intel integrated (Optimus/hybrid)
  • Driver version: 610.43.03
  • OS: Fedora 44 (KDE Plasma, Wayland)
  • Kernel: 7.1.4-202.fc44.x86_64
  • Desktop: KWin Wayland (Plasma 6)

Summary

Note: not currently reproducing. This report is based on logs captured during a past episode. I don’t have live dmesg/journalctl attachments to provide.

Every time the system resumes from S3 suspend, KWin fails to reacquire the DRM device from the NVIDIA driver, producing:

kwin_wayland[14205]: Applying output configuration failed!
kwin_wayland[14205]: Failed to open drm node /dev/dri/card0: Permission denied
kwin_wayland[14205]: atomic commit failed: Permission denied

This does not happen on every ordinary resume, most suspend/resume cycles are fine. Instead, the system intermittently enters an episode where it starts re-suspending itself roughly every 10–20 seconds, indefinitely, and during these episodes every single resume shows the DRM/output-configuration failure above. A reboot always clears the episode; it does not recur until the next (unpredictable) occurrence, which can be days apart.

I can’t say with certainty the DRM failure causes the repeated re-suspending, or whether both are downstream of some other trigger, but within an episode, the two are tightly correlated in every capture I’ve taken: the display never becomes usable after a resume, and the system suspends again shortly after. This loop behavior was independently confirmed to not originate from KDE’s PowerDevil (killing that process mid-episode did not stop it), and not from idle/session timeout settings (all disabled/empty) during the episode.

Steps to reproduce

Not reliably reproducible on demand — the issue occurs intermittently (roughly every few days of normal use, trigger unknown). When it does occur:

  1. System spontaneously begins re-suspending every ~10–20 seconds
  2. Each resume in this state shows the DRM/output-configuration failure below
  3. A normal, isolated suspend/resume outside of one of these episodes does not show the failure
  4. Rebooting always ends the episode

Log excerpt (representative single cycle)

Jul 23 17:46:02 systemd-logind[1311]: Operation 'suspend' finished.
Jul 23 17:46:02 kwin_wayland[14205]: Applying output configuration failed!
Jul 23 17:46:12 systemd-logind[1311]: The system will suspend now!

Other errors seen consistently around the same resume window (may be separate/cosmetic issues, included for completeness):

kernel: ACPI Error: No handler for Region [VRTC] (...) [SystemCMOS] (20251212/evregion-131)
kernel: ACPI Error: AE_NOT_EXIST, While executing method \_WAK (20251212/hwesleep-47)
kernel: xhci_hcd 0000:3a:00.0: xHC error in resume, USBSTS 0x401, Reinit

What I’ve already tried / ruled out

  • Updated to latest available driver (595.80 → 610.43.03), issue persists identically on both.
  • Confirmed nvidia-suspend.service / nvidia-resume.service are enabled and running successfully each cycle (no errors from these services themselves).
  • Ruled out KDE-side idle/suspend policy as the cause of the re-suspend loop (all relevant powerdevilrc / powermanagementprofilesrc idle timers are unset).
  • Ruled out PowerDevil as sole cause of the loop (loop persisted after killing the process).
  • Ruled out stuck hardware sleep key / input-remapper interference (monitored raw input events during an active loop — no phantom key events).
  • Ruled out cron/systemd timers (systemctl list-timers clean).