4

20.04.1 LTS . Software updater (GUI application) showed kernel 5.8.0-34 as an update on 7 Jan 2021. During the update, the build for the (proprietary) nvidia driver from Ubuntu (not Nvidia) failed. I NEED the proprietary full function Nvidia driver and the Nvidia proprietary GUI that interfaces to the proprietary driver. Is there a workaround or a yet later production revision of either the Nvidia proprietary .deb package or the kernel to work with the current production Nvidia package?

Yasha Karant
  • 961
  • 1
    What is the output of dkms status? Also, how did you install the NVIDIA driver from Ubuntu? If you use the graphics-drivers PPA the driver works fine with 5.8.0-34-generic kernel. – Terrance Jan 07 '21 at 20:11
  • Exactly the same situation here. Right now I am back to 5.4.0-59 to be able to use the system. – Jonatã Jan 08 '21 at 03:11
  • @Terrance I did not look at dkms status -- how do I supply that here? The install was done "automatically" when Software Updater showed a new production kernel (5.8); during the update (I had the scrolling screen part of Software Updater open to watch the details), the build of the Nvidia driver (presumably from Ubuntu packages) failed. How does one get/install (actual commands, please) the graphics-driver PPA for Nvidia proprietary current production and then install (build) this? – Yasha Karant Jan 09 '21 at 04:34
  • To add the PPA it is simply sudo add-apt-repository ppa:graphics-drivers/ppa, then it should automatically do the update of the Repos after that command. Then you should be able to now install the sudo apt install nvidia-driver-460 which just came out today. Run all of this from a terminal window. After any of the NVIDIA driver installations or updates I recommend a reboot for the new driver to take effect. – Terrance Jan 09 '21 at 04:49
  • @terrance Thank you for that clarification. You mentioned PPA in your initial response. In the future for other issues, if one mentions say FOOBAR, how does one translate FOOBAR to a PPA -- is the syntax always ppa:FOOBAR/ppa . In my case, FOOBAR would have been nvidia, not graphics-drivers. If there is no "standard", but one must "know" the specific ppa, how does one discover this? You clearly have resolved this issue -- may one ask the method you used? There was no "hint" from the scrolling output "box" under Software Updater to use "graphics-drivers". – Yasha Karant Jan 09 '21 at 05:35
  • Not all PPAs are ppa:foobar/ppa but a good majority of them are. Sometimes PPAs contain non-opensource software or they contain software that Ubuntu doesn't have permission for them to include them in the initial installation ISO package. So, knowing which ones to use is done through research or experience from other users. – Terrance Jan 10 '21 at 17:57

7 Answers7

3

I got the same error and just found out the solution. First thing I did was to uninstall the 5.8 kernel and rever to the 5.4.0-60. That however didn't solved the issue. So I realized that the nvidia-smi utils was stating that I had a version mismatch. So I removed all nvidia drivers and reinstalled with the correct version, that would be (on my case) 450. What I reinstalled was:

  • linux-modules-nvidia-450-5.4.0-6-generic
  • nvidia-dkms-450
  • nvidia-kernel-common-450
  • nvidia-kernel-source-450
  • nvidia-utils-450

After a reboot, the nvidia-smi util was working, but still no X. Then I found out that the xserver-xorg-video-nvidia-450 was removed, so I reinstalled and now everything is back to normal.

You may need to install other packages, such as:

  • libnvidia-common-450
  • libnvidia-gl-450

Hope that helps

Mateus Interciso
  • 31
  • Version mismatches are caused by installing a newer or older version of the driver but the existing driver is still in use until a reboot happens. – Terrance Jan 09 '21 at 01:55
2

I had the same issue. It seems like the linux-modules-nvidia-* packages have dependency issues. The packages do not list the corresponding kernel header packages in their dependencies, but only the kernel images. This obviously leads to a build fail.
Installing the header files for the current kernel fixed the problem for me.

For Example: I had the linux-modules-nvidia-390-generic-hwe-20.04 installed, so I installed the linux-generic-hwe-20.04 package, which always installs the newest HWE kernel image and header files. (see here) Imho that package should be the dependency for the nvidia package.

Someone
  • 21
1

I followed the instructions above from Terrance. As given, these instructions did not work, but did enable a solution. The detailed output appears below; my guess is that the "package" explained by Terrance is not full production in that my experience is that "production" packages resolve and install all needed dependencies; this one did not. I attempted to post in this answer the full details (17kByte). Below is the actual command set that did work on 20.04.1 LTS; however, this Ask Ubuntu system will not let me post the output so that others may compare as to what the output for a working Nvidia driver system appears to be.

sudo add-apt-repository ppa:graphics-drivers/ppa

sudo apt install libnvidia-compute-460 nvidia-compute-utils-460 libnvidia-ifr1-460 libnvidia-decode-460 libnvidia-encode-460 nvidia-utils-460 nvidia-driver-460
Yasha Karant
  • 961
  • That is strange that sudo apt install nvidia-driver-460 didn't install all these packages for you. That install step installed them all for me. Anyway, good answer here and I am sure it will help others on their way as well! =) – Terrance Jan 10 '21 at 17:58
0

Having a issues post upgrade myself. Unfortunately, the timing was terrible as I had just dropped in a new laptop drive with a fresh install and picked up the update. Even after getting the 460 driver to work, I started having serious performance issues with one of the two displays when connecting my external monitor. The performance issues were significant and multiple crashes had me simply replace the original system back to kernel 5.4.0 and nvidia 450 to get the system usable again. Serious quality control issue with this update.

tca
  • 31
  • Sure would be nice to have a way to track this issue. I have to halt any updates until I can rely on 5.8 + nv460 – tca Jan 12 '21 at 19:40
0

i faced a similar same problem 3 days ago after having applied UBUNTU 20.04 LTS updates (kernel 5.8.0.36)

I could solve the problem by using the generic GPU driver instead of NVIDIA's which is no longer supported

please see here for details : https://askubuntu.com/questions/1305862/after-an-update-and-reboot-in-ubuntu-20-04-i-can-no-longer-complete-the-boot

Gilles Gaido
  • 173
0

I faced same situation with this whole kernel thing: changing from 5.4 to 5.8 broke my virtualbox. I went back to 5.4 until everything gets fixed.

Lucas Aimaretto
  • 423
  • 6
  • 18
0

After an automatic update to 5.8.n, my Thinkpad W540 showed a white screen. I was able to boot with 5.4.n which still works perfectly. I followed the advice from Terence to install nvidia-driver-460. After rebooting, I was able to see 5.8.n, but the screen resolution was set to 640 x 320 and there was no other option available to change these settings.

I tried everything I could think of. I attempted to install new drivers through a hugely magnified unscrollable terminal screen working blind and typing into space off-screen and found that I had no internet access. No NVIDIA drivers and no Wi-Fi.

I stopped looking for more problems as it would take too long to fix. I recommend that all users experiencing problems with 5.8.n interrupt normal boot and select the previous reliable kernel version (5.4.n for me) until this car-crash update is fixed by the people who caused it.

Sorry to be critical, but fair's fair. 5.4.n is great so the developers can get it VERY right but not this time.

BeastOfCaerbannog
  • 14,585
birdbrain
  • 1