nvidia driver in use
Experience, Information Technology

KDE Neon – Ubuntu – Will Not Boot Properly After Kernel 5.3.0 Upgrade

So, late on Friday you are typing away and see the little blue dot from the Software Updater built into KDE Neon. Maybe you are running Ubuntu and saw a message. Doesn’t matter how you made the decision to install updates late on a Friday, you are now screwed. Yeah. I know your pain. Why do you think I’m writing this post? I did this exact thing. I know better than install anything pushed out late on Friday, but I did it too.

Adding insult to injury is the subconscious panic which sets in when you see the startup screen hang. If you hit the escape key you see a bunch of boot text, but hitting return does nothing. You can see your hard drive flashing so you know BOINC and other things are running. Adding insult to injury, KDE Neon hides the grub boot menu by default. Neither holding down the left shift key nor hitting the escape key at startup brings up the grub menu where you could easily select the previous kernel.

Frantic searches on-line via a different computer give you lots of things you could do from the command line. All of them assume you can get to a command line.

Take a deep breath and relax. You can get to the command line. With that screen of boot log text up hit <alt><ctrl><F2>. Login at the text login prompt.

cd /etc/default

sudo nano grub

change the timeout_style to menu and the timeout to at least 5 but preferably 10 seconds.

 

grub changes

 

Save the file and exit.

sudo update-grub

Now you have 2 choices. The safe and orderly choice would be to reboot via sudo reboot before proceeding. I didn’t bother.

sudo dpkg -1 | grep linux-image

You will find two linux-image packages for 5.3.0. One is a generic hardware package and the other looks like full kernel. (That’s a dash one after dpkg in case the font makes it look like an l.)

sudo apt purge linux-image-5.3.0 that looks like a full kernel

If you didn’t reboot you will get some warning screen about removing the kernel you are running with a cancel option highlighted. Use the arrow key to choose the other one and hit enter. I don’t remember if I had to manually remove the generic hardware one as well or if that removal process removed it as well. I do remember it complained about dependency issue.

sudo dpkg -1 | grep linux-image

Now you will see an unsigned 5.3.0 kernel appear in the list.

sudo apt purge unsigned linux 5.3.0 package name

Now you need to reboot again.

All should be well.

What I believe caused this problem is the AGILE software development method and the fact the 12 year old boys never test with nvidia drivers or only test with the last driver, not all of them in the repository.

nvidia driver list

nvidia driver in use

As you can see there are quite a few nvidia drivers in the list. If you have an old enough card you cannot use the latest driver because nvidia dropped support for many of their older cards. Those of us running BOINC to help cure cancer, AIDS, and many other problems need the nvidia driver to expose the CUDA cores to BOINC. We can’t use the “default” driver because it keeps the CUDA cores to itself. My card has 384 CUDA cores. That’s a lot of number crunching power. As the server settings screen shows I’m running the 390 driver. It was the best one for my card. Many of you will be in the same boat because 340, though in the repo, doesn’t play well with current Ubuntu.