gdm doesn't start after failed attempt to install amdgpu drivers until modprobe amdgpu (Ubuntu 22.04)

Question

Short version: After trying (and failing) to install the AMD radeon proprietary drivers, I uninstalled them and removed the packages it installed, but now:

• At boot, my graphics card is UNCLAIMED.
• GDM wont start until I do: sudo modprobe amdgpu (which also claims my graphics card)
• After login, DISPLAY is now :1 instead of :0 which I don't understand

What should I do to return to a working system that actually boots?

Details

Everything was working fine on my newly installed was working fine, except x11vnc.

To fix that I tried to install the (utter mess that is the) proprietary AMD radeon drivers using amdgpu-install_22.20.50200-1_all.deb, and it failed miserably. After modifying /etc/apt/sources.list.d/* multiple times, it finally ended up failing to build the kernel module because of a gcc version mismatch, which is when I gave up. To clean up the botched installation attempt, I ran:

sudo amdgpu-install --uninstall
sudo apt purge amdgpu-install
sudo apt autoremove

After a reboot, gdm no longer starts. I noticed that sudo lshw -c display showed:

$ sudo lshw -c display
  *-display UNCLAIMED       
       description: VGA compatible controller
       product: Baffin [Radeon RX 550 640SP / RX 560/560X]
       vendor: Advanced Micro Devices, Inc. [AMD/ATI]
       physical id: 0
       bus info: pci@0000:02:00.0
       version: ff
       width: 64 bits
       clock: 33MHz
       capabilities: pm pciexpress msi vga_controller bus_master cap_list
       configuration: latency=0
       resources: memory:d0000000-dfffffff memory:cfe00000-cfffffff ioport:b000(size=256) memory:fbb80000-fbbbffff memory:c0000-dffff
  *-graphics
       product: VESA VGA
       physical id: 1
       logical name: /dev/fb0
       capabilities: fb
       configuration: depth=32 resolution=1280,1024

So I compared /var/log/Xorg.0.log (which is now failing) with ~/.local/share/xorg/Xorg.0.log (which has not been rewritten since last time it succeeded booting) and noticed that the latter had a LoadModule: "amdgpu" line, so I tried sudo modprobe amdgpu, gdm started and now the graphics card is also "claimed":

$ sudo lshw -c display
  *-display                 
       description: VGA compatible controller
       product: Baffin [Radeon RX 550 640SP / RX 560/560X]
       vendor: Advanced Micro Devices, Inc. [AMD/ATI]
       physical id: 0
       bus info: pci@0000:02:00.0
       version: ff
       width: 64 bits
       clock: 33MHz
       capabilities: pm pciexpress msi vga_controller bus_master cap_list rom
       configuration: driver=amdgpu latency=0
       resources: irq:32 memory:d0000000-dfffffff memory:cfe00000-cfffffff ioport:b000(size=256) memory:fbb80000-fbbbffff memory:c0000-dffff

Once I log in, I how have DISPLAY=:1 where before the driver installation attempt it was DISPLAY=:0. This could be a side effect of first :0 trying to start which fails on the unrecognized card, and then after the modprobe it appears as if there now is a new (recognized) graphics card, and so it is given :1.

What can I do to clean up this mess? I don't want to hardcode loading of amdgpu, because I don't think it was hardcoded before...

I don't think it would be too much work to reinstall Ubuntu 22.04 from scratch, but needlessly to say, it would prefer to understand what went wrong and fix it, if possible.

Do you know what I should do to get amdgpu to load automatically and get DISPLAY :0 to work again?

Answer 1

So apparently the amd installer in its wisdom created:

/etc/modprobe.d/blacklist-amdgpu.conf containing:

blacklist amdgpu

Which is why the module didn't get loaded after reboot. Removed that file, and all is well again.