I broke X, probably by installing a proprietary NVIDIA driver. How can I get my X server back?

Question

When I start my computer, I get a text login--it doesn't even load lightdm. Dmesg gives me this:

[   42.367035] NVRM: API mismatch: the client has the version 304.43, but
[   42.367038] NVRM: this kernel module has the version 310.14.  Please
[   42.367040] NVRM: make sure that this kernel module and all NVIDIA driver
[   42.367041] NVRM: components have the same version.
[   42.619472] init: lightdm main process (1235) terminated with status 1
[   48.041960] vboxdrv: Found 1 processor cores.
[   48.044418] vboxdrv: TSC mode is 'synchronous', kernel timer mode is 'normal'.
[   48.044422] vboxdrv: Successfully loaded version 4.1.12_Ubuntu (interface 0x00190000).
[   48.459287] vboxpci: IOMMU not found (not registered)
[   50.544026] wlan0: no IPv6 routers present
[   52.729171] init: Failed to spawn vsftpd main process: unable to execute: No such file or directory
[   58.435781] init: plymouth-stop pre-start process (2306) terminated with status 1
[   59.169528] init: failsafe-x main process (1429) terminated with status 1

I don't know how much of this is relevant. I think this problem started around when I tried to install new proprietary NVIDIA drivers.

How can I get X back? I would install older drivers using the Additional Drivers program, except I can't start any version of X.

Answer 1

I had this same problem. It was caused by a failed installation of the 310 version of the nvidia drivers. I fixed it with

apt-get remove nvidia-experimental-310

After that I could start X.

Answer 2

Important! Read this entire post before trying!

Go to a terminal by pressing CTRL+ALT+F3. Login, and enter this command:

sudo apt-get purge nvidia-*

That will remove the proprietary drivers. Once it's done, backup your Xorg.conf, like so:

sudo mv /etc/X11/xorg.conf /etc/X11/xorg.conf.bak~

Then run this:

sudo update-initramfs -u

This updates the RAMFS, which seems to help (forget why though, sorry). Then reboot using this:

sudo reboot

Answer 3

I got a similar error as I was installing CUDA for Ubuntu 12.10 which is not yet fully supported by NVidia (they have 11.10 as the newer version.)

The problem was indeed a mix of libraries. I had to completely remove all libraries installed by default when installing Ubuntu and then make sure to reinstall the CUDA driver.

To get a list of the default nvidia packages installed:

dpkg -l | grep nvidia

To remove them:

apt-get remove --purge nvidia-...

Obviously the other solutions presented here would not work in case you were installing CUDA since reinstalling the default libraries & drivers would overwrite the CUDA installation!

---

In the newer versions of Ubuntu (since at least 13.10), CUDA can be installed using:

sudo apt-get install nvidia-cuda-dev
sudo apt-get install nvidia-cuda-doc
sudo apt-get install nvidia-cuda-dbg
sudo apt-get install nvidia-cuda-toolkit

These commands should automatically install the cuda library if not already installed.