Lightdm failed to start X server (system76 thelio)

User Support Hardware Issues
1

i just installed qubes on a new system with a nvidia dGPU and an AMD iGPU. After a dom0 update, i rebooted and now lightdm will not start. After typing in my FDE password, i get a black screen with white text, stopping at

Failed to start lightdm.service

I have tried

  • ‘sudo qubes-dom0-update kernel-latest’
  • rd.qubes.hide_pci to hide the nvidia gpu
  • nomodeset
  • modprobe.blacklist=nouveau

but nothing had any effect. Next step is removing the GPU but is there something else i need to do to get this working?

2

Switch to another TTY using e.g. Ctrl+Alt+F2, log in and check the logs using e.g:

sudo dmesg | less
sudo systemctl -b -u lightdm

To see the errors.

3

I tried removing the nvidia graphics card but this doesn’t change anything.

dmesg and journalctl show nothing i can pick out as relevant.

On suggestion from another thread, i tried running lightdm --test-mode --debug. It shows that X is failing to start:

lightdm
lightdm1920×1621 398 KB

So i checked /var/log/Xorg.0.log and it seems like the error is happening because /dev/dri/card0 is missing. EDIT: actually idk what the real cause is.

xorg
xorg1920×2009 382 KB

4

Which CPU/iGPU do you have?
Search for amdgpu in the dmesg output to see if there are any errors.

5

It’s an AMD Ryzen 9 9950X, and according to this page the iGPU is a Radeon 2CU 2200MHz RDNA 2.

I did find some dmesg logs pertaining to this device:

dmesg
dmesg1920×1080 274 KB

Nothing for nvidia or nouveau.

6

Maybe there is a problem with newer kernel version that was installed during update.
Try to boot with the older kernel version that worked for you before.
Also, what Qubes OS version do you have and what kernel versions old/new do you have?

7

Very nice, changing the kernel version worked. In the grub menu, i selected “Advanced options for Qubes (with xen hypervisor)” → “Xen hypervisor, version 4.17.5” and i was given three options:

  • Qubes with xen 4.17.5 and Linux 6.6.77-1.qubes.fc37.x86_64
  • Qubes with xen 4.17.5 and Linux 6.12.47-1.qubes.fc37.x86_64
  • Qubes with xen 4.17.5 and Linux 6.15.10-1.qubes.fc37.x86_64

I tried each of them but only 6.6.77 allows me to boot into the OS as normal (lightdm launches fine).

I tried to set 6.6.77 as my default kernel using the method described in this guide, but it has no effect.

Seems like I’m still booting into 6.15 by default even after i follow the instructions and select 6.6 once.

PS I’m running Qubes 4.2.4.

8

Woops! All good, i have the old kernel set as default in grub now. My problem was just that i was misspelling one of the grub env variables in /etc/default/grub.

So what should i do now? My system now runs fine with the 6.6 kernel, but how should i report this kernel incompatibility so it can be fixed upstream?

9

Did you test your kernel versions with NVIDIA GPU removed?
You can try to boot from some other Linux OS (e.g. Fedora Live) and check if you’ll have the same issue with the same kernel versions there.
If it’s not an issue specific to Qubes OS or Xen, then it should be reported to the kernel mailing list or to Fedora.

10

After much trial and error, it turns out this problem had nothing to do with qubes or even with my GPUs.

See, when i first installed qubes os, i had to permanently enable the nomodeset kernel command line parameter to avoid a blue screen of death at boot. Thread. This was needed no matter which kernel version or Linux distro i used.

The breakthrough was when i finally tried switching to a new monitor, and removing the nomodeset flag (i.e, reverting to stock defaults but on a new monitor). This worked normally.

So evidently my hardware setup was misbehaving if:

  • using nomodeset plus a linux kernel newer than 6.6, OR
  • using stock kernel flags, plus my viotek monitor.

Apparently the hardware i have really doesn’t like something about this monitor. Fwiw it’s a Viotek GN32DB. i tried doing an EDID override but nothing worked. So i think my long term solution will just be to buy a new monitor…