Linux Boots to a Black Screen After a Graphics Driver Update: How to Get Back In

When a Linux machine stops at a blank screen or a cursor after a driver update, it’s an admission to the GRUB console that is the point of first action. Boot into the recovery mode first, then start a tinkering session with the kernel boot parameters, should get you into a usable state from which to diagnose for real.

has documented a recovery line of reduced size but more than sufficient to prove implication: boot into recovery from GRUB, add a pinch of nomodeset or noresume to the kernel, then check the logs to separate good repair, bad config from a reboot that confirms the fix. But not before the toolkit for marking each vendor is rolled out.

Reach the Recovery Mode

A recovering Linux partition may need kernel parameters like nomodeset or noresume. Each mode of fixing starts with the nominative.

Holding Shift on BIOS systems or Esc on UEFI systems during boot can expose the GRUB menu.

For machines using UEFI, it's Escape that sends you crashing up to the head of the line.

On recovering that might be Ubuntu, unmark the as-needed keyboard options and start scanning the menu.

In the advanced mode menu, open up a root shell, then use a shell to demote the previous kernel parmeter settings to increase the scope of what can be edited.

Temporary Boot Edits

The most gentle mode moves start with the nominative.

Once a machine recovers itself, run the command journalctl -b-1 | grep kernel in a root shell to filter out lines that refer to the use of modules: drm, nouveau, amdgpu, and i915 can fail, and so can GDM, Wayland, Xorg, or DRM.

Black screens can also be fixed by calling Ctrl+Alt+F2 to reach a terminal, logging in, then using the terminal to edit GRUB's boot line. Add nomodeset and noresume before rebooting to increase chances of reaching a login prompt.

Vendor documentation has shown that nvidia-bug-report.sh outputs useful diagnostic bundles for NVIDIA, while the gpu-manager.log if output after a reboot may also contain useful diagnostics on bootlaoded modules for AMD and Intel.

Roll Back the Change

Once a recovery shell is reached, decide whether the scope of the repair is one module, or the whole driver stack. NVIDIA, AMD, and Intel require different repair flows.

NVIDIA drivers can be rolled back from the recovery shell, or checked using nvidia-bug-report.sh to then either un/reinstall from binary. AMD drivers can be uninstalled, then reinstalled, although un/reinstalling can require that the effect of the parameter noresume be expected.

Running nvidia-bug-report.sh can help diagnose whether the issue lies with the NVIDIA driver, kernel, or DKMS configuration. Run this to rule out NVIDIA if it is not the source of the issue.

The Ask Ubuntu report says adding noresume can help in recovery. This parameter may be OPTIONAL here, and could be avoided in favour of DDR3 if your PROCESSOR is integrated with Intel’s on-die graphics and the issue lies with the onboard graphics.

For NVIDIA: nvidia-bug-report.sh nvidia-settings —reinitialize-host-software

For AMD: /etc/init.d/gdm3 stop sudo aticonfig --initial

For Intel: /etc/init.d/gdm3 restart

Logs:

journalctl -b-1 | grep kernel /var/log/Xorg.0.log

Vendor-Specific Flaws

A machine that fails to start due to errors in the NVIDIA driver can show symptoms such as a black screen, high GPU temperatures, or a machine that fails to boot at all. This is by far the most common failure mode.

AMD machines may fail to boot at all, or may fail to reach a login prompt. The error log shows of interest is /var/log/dm-raid.log, which may show errors with 'amd/81xx' or similar.

A machine that fails to boot due to errors in the Intel driver can have symptoms that aren't limited to black screens. Broad symptoms such as overfull logs, instability from blackouts, or instances where logging in might reset your machine can be indicative of an Intel integrated card issue. But the symptom you're most likely to see with Intel integrated graphics is a black screen.

An instance where black simbols like repeating terminal menus, scrolling, or messages that you've read too slowly show up instead of a desktop could also be caused by Intel Integrated GPUs, because integrated means you're using a CPU that's limited, and GPUs that were made for shared use. Wayland might also be involved, otherwise if you're using an installed machine's driver.

Reboot and Confirm

Once those diagnostics run, complete the package repair, driver reinstall, or configuration change, the final step is to reboot and confirm the fix.

For UX design reasons, permanent boot parameters like nomodeset, noresume, or rhgb should be removed if it was added to help get into a usable console. Once a machine recovers, it should boot cleanly, and without the need for temporary parameters to set internal error log levels to MEDIUM.

The system should boot within a reasonable time after being power cycled. If a black screen occurs, consider capturing the diagnostic bundle nvidia-bug-report.sh.

Tell me: What's your favorite thing about Ubuntu?

Sources

• Ubuntu.Fan — ubuntu.fan
• Ask Ubuntu — askubuntu.com
• NVIDIA Developer Forums — forums.developer.nvidia.com
• KDE UserBase — userbase.kde.org
• Fedora Discussion — discussion.fedoraproject.org