Hello,

I have an kernel version: kernel-5.8.18-300.fc33.x86_64.. This one is still working.

This week shutdown and the kernel has been updated/or new kernel has been installed: kernel-5.9.8-200.fc33.x86_64

Next day I boot my system and I got the message: "Oeps there went something wrong... ask you're system administrator", when I boot from the new kernel version.

But now, what is the best way to solve this:

• Must I try to install the the version 5.9.8-200 again? But how? And without damaging the "kernel-5.8.18-300.fc33.x86_64" version, which is still working.
• Or can I remove the new kernel, so that I downgrade to the version which is still working. And try to update again?
• Or... is there another solution, what I don't know?

Nico

Hmmm. I used solve that sort of thing by ignoring yum/dnf/whatever, and using rpm. Read man rpm.
They have all sorts of useful options: --nodeps, --force --ignore-this, --ignore-that, etc.

I had to use -ignorearch or somesuch because at the time I had an AMD 80586-133-P75 cpu which went in a 486 boartd, ran at 133Mhz, and roughly equalled a P75. But it read it as a '486

The package manager can remove the newer kernel if that's what you want. Boot from the older one and just remove the new one. You can try reinstalling the newer kernel and the firmware and see if that helps. I would suspect utdated firmware, which can happen if the kernel is put into the repo without all the firmware being updated to match. That should be resolved soon, though. This won't affect the older kernel, it will remain available and working unless you do something truly strange to it.

But... I'm not so familiar with Linux. I use to learn working with Linux.

I'm booting from the older working version. Now I'm searching how can SAVE remove the new ones, by the package manager?

In the meantime a second new kernel is 'installed' which is not working.
Is there a way to find the issue why the kernel is not working? Is there an advanced way to see the core dump maybe?

I don't know if it is good or not, but I found this: . Or is this dangerous that will also modify my working version?

When I do this, than my current version is not touched and stay working?

And to disable updating the kernell for now is this enough: ? And saw also that I have to add at the /etc/yum.conf the [main] section: exclude=kernel*. But I don't have a /etc/yum.conf!

I'd say the laziest way to downgrade is not to, but to upgrade again. Fedora often do that, and bugfix it fairly fast when the kernel pukes and everybody starts bellyaching.

I have found the issue: It is the NVIDIA card which I have: VGA controller NVIDIA GT215M - GeForce GTS 360M (rev a2) in my Thosiba Qosmio laptop.
My version of the driver was Nvidia driver 340.108 which I had installed a year ago when I started with Linux. I have installed Nvidia driver because I had little issue's on the screen when I switch with a window or moving a window than kind pixels were twinkeling. (I hope that you have an idea what I mean)

So I thought I need to replace the Nouveau driver. I found this:

• Nvidia.com
• inttf-nvidia-patcher
• fedora Nvidia Guide

That guy 'JR' did for me at least a very good job and the site is maintained till 15 january 2021. Now I can startup fedora version 5.9.11-200.fc33.x86_64 (and version 5.9.10-200.fc33.x86_64 as well).

The only thing were I'm afraid for, do I have to this after every fedora update now? I don't think that Nvidia is maintaining this driver...

Or does somebody have a better idea to maintain this?

I have used previous Nvidia drivers, and if you run an update, usually you don't have to update.

If you update Mesa, or the kernel modules, then you do. Now I know Nvidia had a habit or releasing new kernel very regularly, half of which ended in catastrophe, box not booting, etc. Perhaps they have improved. It's rare to need a kernel update anyhow, unless it's some important security fix.

Say what ?. Aren't we talking Fedora here ?.

In similar situations I simply boot the previous kernel rather than the latest - I have one machine that I don't allow to upgrade at all, but I enforce manual upgrades. Each to their own.

Fedora and I parted ways some time back.

The issue is likely sorted at this point, however this is a typical example of rolling/bleeding-edge distros disrupting work. Since no other member has posted anything about avoiding such a scenario in future, here's my two cents on preventing/limiting a recurrence.

1. Use a stable distro as host and move the experimental/rolling/bleeding edge OS as Guest OS. Choose your poison wisely (vmware/virtualbox/kvm/xen etc)
2. Assign as much of resources (cpu/disk/ram) as you can so the guest does not choke when you are using it. This depends on the use case though.
3. Turn off automatic updates and run them manually periodically (unless of course, its a spctre/meltdown type of scenario from few years back!).
4. For critical machines/workflows, test the kernel first, before putting it in action full time.
5. Frequent Backups of data.

It's a long time since I used Fedora, but normally when a new kernel is installed the old one is left in place just in case you have to boot from it. I did hear recently of Nvidia drivers having problem with a specific kernel, and the solution is obviously to stick with the previous one and await a new one. There's no rush to remove the difficult kernel if you aren't using it — it might just get re-installed in your next update!

1 members found this post helpful.