System Hangs at Boot

Ask Fedora
, , , ,
1

Problem Description

System was running fine for over a week without reboot. After a routine reboot, the system hangs during boot showing service startup messages (cups.service, sshd.service, gdm.service starting) but never reaches the login screen. Keyboard becomes unresponsive at some point during boot. No new journal logs are being created - all logs are from Nov 10 (last successful boot).

What I’m expecting to achieve

I want to boot my system back on GNOME Wayland without reinstalling the whole thing.

System Information

Laptop: Lenovo LOQ 15AHP9 (Model 83DX)

Hardware:

  • CPU: AMD Ryzen 7 8845HS (16 cores)
  • GPU 1: NVIDIA GeForce RTX 4060 Max-Q / Mobile (AD107M)
  • GPU 2: AMD Radeon (HawkPoint1 integrated)
  • RAM: 16GB
  • Storage: 2x NVMe (nvme0n1: 931.5G, nvme1n1: 476.9G) [But sometimes they switch for some random reason unrelated to the main problem]
  • Filesystem: Btrfs with subvolumes (root00, home00)

Software:

  • OS: Fedora 41 Workstation Edition
  • Desktop: GNOME 47 (Wayland)
  • Kernels Tried: 6.11.4-301, 6.16.7-100, 6.16.8-100, 6.17.7-100 (After updating the whole system)
  • NVIDIA Driver: 580.105.08 (From RPM Fusion)
  • Display Managers Tested: GDM, SDDM, LightDM

Troubleshooting Steps Taken

1. Chroot Recovery from Live USB

Since system wouldn’t boot, used Fedora 41 Live USB for all troubleshooting. Then I mounted the whole system and did chroot

sudo chroot /mnt

Key finding: No new journal entries after Nov 10. System crashes before journald can write logs.

journalctl --list-boots
# Shows last successful boot: Nov 10 20:05:18
# Current failed boots: No entries

Initial Boot Screen Errors (from photo):

When I boot normally, some errors in the screen are:

[FAILED] Failed to start akmods.service
[FAILED] Failed to start auditd.service  
ACPI BIOS Error (bug): Could not resolve symbol [\_SB.PCI0.GPP0.PEGP.GPS.NVD1], AE_NOT_FOUND

Maybe there’s more, but journalctl doesn’t show me logs from today. Only boots from more than a week ago:

Last successful boot (Nov 10) showed normal startup with these services working fine:

  • akmods.service: OK
  • auditd.service: OK
  • gdm.service: Started successfully
  • NVIDIA modules loaded correctly

3. Service Management

Disabled problematic services to simplify boot:

systemctl disable auditd
systemctl disable firewalld
systemctl disable akmods.service
systemctl disable tuned.service
systemctl disable jellyfin.service
systemctl mask systemd-backlight@backlight:nvidia_wmi_ec_backlight.service

Nothing happens…

4. Kernel Parameters Tested

Multiple boot parameter combinations tried:

Initial (failing):

ro rootflags=subvol=root00 rhgb quiet

Attempted fixes:

# Added ACPI workarounds
grubby --update-kernel=ALL --args="acpi_osi=Linux nomodeset"

# Blacklisted nouveau
grubby --update-kernel=ALL --args="rd.driver.blacklist=nouveau modprobe.blacklist=nouveau"

# Enabled NVIDIA DRM
grubby --update-kernel=ALL --args="nvidia-drm.modeset=1 i915.modeset=1"

# Added debugging
grubby --update-kernel=ALL --args="systemd.log_level=debug drm.debug=0x1e"

Nothing happens…

Final kernel args:

ro rootflags=subvol=root00 rhgb quiet rd.driver.blacklist=nouveau,nova_core modprobe.blacklist=nouveau,nova_core i915.modeset=1 nvidia-drm.modeset=1 systemd.log_level=debug drm.debug=0x1e

Still cannot boot and the debug options didn’t work.

5. NVIDIA Driver Work

Verified NVIDIA modules built and present:

ls /lib/modules/6.17.7-100.fc41.x86_64/extra/nvidia/
# All 5 modules present: nvidia.ko, nvidia-drm.ko, nvidia-modeset.ko, nvidia-uvm.ko, nvidia-peermem.ko

lsmod | grep nvidia
# Driver loads successfully in terminal mode
# nvidia 580.105.08 confirmed working

nvidia-smi
# Shows RTX 4060, driver working correctly

Nouveau blacklisted:

cat /etc/modprobe.d/blacklist-nouveau.conf
blacklist nouveau
options nouveau modeset=0

Rebuilt initramfs multiple times:

dracut --force --no-hostonly --kver 6.17.7-100.fc41.x86_64
# Final size: 151MB (correct, includes NVIDIA modules)

6. System Upgrade

I upgraded entire system to latest packages:

dnf upgrade --refresh -y
# Installed kernel 6.17.7-100.fc41.x86_64
# Updated NVIDIA to 580.105.08
# 682 packages updated

7. Display Manager Attempts

GDM (default):

  • Installed, enabled, configured for Wayland
  • Crashes immediately after boot
  • Keeps restarting (restart counter hits 5)
  • Shows: Failed to start gdm.service - GNOME Display Manager
  • Error: Failed to list cached users: GDBus.Error:org.freedesktop.DBus.Error.NameHasNoOwner

SDDM:

dnf install sddm
systemctl enable sddm
  • Same behavior as GDM
  • Login screen appears briefly
  • After entering credentials, blinks and returns to login
  • All session types fail: GNOME, GNOME Classic, GNOME on Wayland

LightDM:

dnf install lightdm lightdm-gtk
systemctl enable lightdm
  • Failed to start with exit-code
  • X server startup issues with hybrid GPU

8. Terminal Mode + startx

Only working solution:

systemctl set-default multi-user.target
# Boot to text mode, then:
startx

This works but:

  • Uses X11 instead of Wayland
  • Crashes when changing display settings
  • Not stable for dual-GPU laptop

9. KDE Plasma Attempt

dnf group remove "GNOME Desktop Environment" -y
dnf install @kde-desktop-environment --allowerasing -y
systemctl enable sddm
  • Still hangs at boot
  • SDDM fails to start
  • Same Wayland/GPU initialization issue

10. Wayland Configuration

Environment variables set:

cat /etc/environment
__GLX_VENDOR_LIBRARY_NAME=nvidia
__NV_PRIME_RENDER_OFFLOAD=1
__VK_LAYER_NV_optimus=NVIDIA_only
MOZ_ENABLE_WAYLAND=1
GBM_BACKEND=nvidia-drm

GDM Wayland enabled:

cat /etc/gdm/custom.conf
#WaylandEnable=false  # Commented out = Wayland enabled

Files Examined

Xorg logs (from working startx session):

/home/martin/.local/share/xorg/Xorg.2.log:
[   557.367] (WW) AMDGPU(0): flip queue failed: Invalid argument
[   557.367] (WW) AMDGPU(0): Page flip failed: Invalid argument

No useful logs from failed boots - System never gets far enough to write to journal.

My Theory

This is a Wayland compositor initialization failure with hybrid NVIDIA+AMD GPU setup. Here’s why:

  1. X11 works (startx), Wayland doesn’t - Clear indication Wayland-specific issue
  2. NVIDIA driver itself is functional - Terminal mode proves this
  3. Started after week-long uptime - Maybe this is because state/cache corruption, not hardware
  4. AMDGPU page flip errors - Hybrid GPU handoff failing in Wayland
  5. Display manager crashes before session starts - Compositor can’t initialize

Wayland has been a headache for several years, but there’s no going back since it’s been standarized in all distros. I really hope this issue brings too light something that can make Wayland work in the future

Questions

  1. Is this a known issue with NVIDIA 580.x + Wayland on AMD+NVIDIA hybrid laptops?
  2. Should I try the open-source nvidia-open kernel modules instead?
  3. Is there a way to force proper GPU selection for Wayland compositor?
  4. How can I capture actual compositor crash logs when journal isn’t writing?
  5. Should I file a bug with NVIDIA, Fedora, or upstream GNOME/Mutter?

Any help appreciated. System is completely stuck without working graphical environment.

2

I’ll be honest with all of you, I don’t have at the moment the luxury of doing a whole backup and reinstall the whole system. I’ve dedicated this whole day to troubleshoot using other similar posts and even asked Claude AI for suggestions. At the end, I feel defeated, so I’ve decided to compile all of the things I’ve done so far.

3

Common recovery steps:

Follow with permissive SELinux mode, check failed units, proceed with graphical boot:

sudo setenforce 0
sudo systemctl --failed
sudo systemctl isolate graphical.target

Then collect and analyze the logs, possibly over SSH.

Also there’s a newer driver in the testing repo you may want to try:

sudo dnf upgrade \*nvidia\* --enable-repo=rpmfusion-\*-testing
tail -f /var/log/akmods/akmods.log
tail -f /var/cache/akmods/nvidia/.last.log

BTW, that error message from GDM indicates failing to access D-Bus.
This is beyond any GUI related issues.

1 Like
4

I will be honest with you.

  1. You are on fedora 41 which WILL become EOL this month. A version upgrade is strongly recommended.

  2. Fedora 43 does not support X11 and you said you have issues with that so you probably should not consider using startx to launch the desktop

As far as support for the nvidia drivers there have been some issues and bug reports with the nvidia 580.105 releases

I would recommend that you downgrade the drivers back to the 580.95.05 version and see if that solves the issue.
sudo dnf downgrade akmod-nvidia xorg-x11-drv-nvidia

1 Like
5

Are you sure you are not already using the open kernel modules? The RPMFusion packages should automatically select the open kernel modules when they detect a compatible GPU.

You can check with modinfo nvidia | grep license.

For the open-kernel modules you’ll see:

license:        Dual MIT/GPL
6

Guys, thank you so much with the suggestions. I had to make a quick decision tonight and installed Silverblue 43. I’m now dealing with compatibility problems, but at least I have a working computer now.