Specs
- OS: Fedora Kinoite 43
- GPU: NVIDIA GeForce RTX 4070 Ti 12GB
- Kernel: 6.19.11-200.fc43.x86_64
- Driver: 580.126.18 (via manually-signed
akmod-nvidia-openpackage via GitHub - CheariX/silverblue-akmods-keys: Fedora Silverblue with SecureBoot enabled and working akmods such as akmod-nvidia or akmod-VirtualBox. · GitHub , though the same issue happens underakmod-nvidia) - Secure Boot enabled
- Hard drive encryption via luks
- Using Wayland with KDE Plasma
- Up-to-date kernel & BIOS
Relevant kernel args: rd.driver.blacklist=nouveau,nova_core modprobe.blacklist=nouveau rd.driver.pre=nvidia,nvidia_modeset,nvidia_uvm,nvidia_drm nvidia-drm.modeset=1 nvidia-drm.fbdev=1
Issue
Whenever I try to shutdown my computer, the system hangs with a black screen indefinitely, never fully shutting down unless I hold the power button or perform the alt-sysrq-REISUB sequence. However, if I boot into a TTY instead of graphical desktop and shutdown via systemctl poweroff, the shutdown completes normally, leading me to suspect a bug in nvidia-drm and/or its interaction with kwin/wayland. Over the past few months I’ve been searching online and trying various solutions & kernel args to no avail, leading me to post here.
Upon each reboot I check the kernel logs via journalctl -k -b -1 and find slight variations of this sequence at the end:
Apr 13 19:01:17 fedora kernel: Lockdown: systemd-logind: hibernation is restricted; see man kernel_lockdown.7
Apr 13 19:01:17 fedora kernel: NVRM: Xid (PCI:0000:08:00): 56, CMDre 00000000 00000200 00000001 00000005 00210031
Apr 13 19:01:20 fedora kernel: nvidia-modeset: WARNING: GPU:0: Lost display notification (0:0x00000000); continuing.
Apr 13 19:01:24 fedora kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000800] Flip event timeout on head 0
Apr 13 19:01:27 fedora kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000800] Flip event timeout on head 1
There’s an Xid 56 code here, which seems to be some kind of display error.
This happens consistently on every shutdown, though sometimes there are no nv_drm_atomic_commit errors, e.g. just:
Apr 04 19:45:44 fedora kernel: NVRM: GPU at PCI:0000:08:00: GPU-01806c90-3368-4bf5-d5a3-249917ad8c01
Apr 04 19:45:44 fedora kernel: NVRM: Xid (PCI:0000:08:00): 56, CMDre 00000000 00000200 00000001 00000005 00210031
Apr 04 19:45:47 fedora kernel: nvidia-modeset: WARNING: GPU:0: Lost display notification (0:0x00000000); continuing.
Apr 04 19:45:47 fedora kernel: kauditd_printk_skb: 82 callbacks suppressed
What I’ve tried so far
Various other kernel args that I found online. The only one that kind-of works is nvidia-drm.fbdev=0 but that causes stuttering, phantom display devices, and general instability on Wayland.
Switching between akmod-nvidia and akmod-nvidia-open has the same behavior.
While I have two displays, unplugging/disabling either of them before shutdown/reboot doesn’t change the overall behavior. It just removes the head 1 lines but not the head 0 ones.
Kernel arg initcall_blacklist=simpledrm_platform_driver_init causes it to hang on boot, same with env var KWIN_DRM_NO_AMS=1.
Attached some debug files. Thanks in advance!
journalctl.log (121.0 KB)
nvidia-bug-report.log.gz (499.2 KB)