Ubuntu not booting - suspect bootloader is broken

Hey all, I recently upgraded my ubuntu server to 25.04 (was previously 24.10) and things seemed okay for a bit.

I performed a regularly scheduled reboot yesterday (NOT the first one since upgrading) and am not not getting to a bootloader at all.

My bios screen will flash up as normal, then a cursor appears in the top corner for maybe a second, then just black screen and nothing. CTRL ALT DELETE doesn’t force a reboot, only way is hold power/reset.

Booting into a live usb, I’ve checked system is up to date, regen initramfs, update grub, but no change.

When trying to run grub-install, I get a warning about ext2(why does ubuntu still default to such an old filesystem?) and how embedding isn’t supported, but as there’s only 1 partition on the system I can only assume that’s how it was setup before.

There’s about 2MB before the primary partition but that doesn’t seem enough for a bootloader to me, and again it’s not actually partitioned. All the boot files are visible on the main parition under /boot

The error mentions using blocklists for ext2, but also makes it clear that that is an unreliable setup (assuming that’s how it was configured out the box, now that I’m experiencing this failure i’m inclined to believe that)

So what would be the best way to recover this? If someone can briefly explain blocklists and if that will get things up and running again that’s seeming like the most direct fix, but open to suggestions. I intend on replacing the system with a debian box at some point so reinstalling is on the table, but I would like to get the system booted as is for now, and then I can deal with the migration as it’s own job.

Are you running an Nvidia GPU?

ext4 is a superset of ext3 which is a superset of ext2. Each version adda new features that do not exist in the previous one. So by supporting ext2 for booting, you can support all future ext filesystems.

1 Like

Sorry should have mentioned, yes,

I’m getting the same behaviour plugged into on board or the gpu (on board was always default even before the upgrade) and removing the gpu hasn’t made a difference

Interesting on the ext stuff, seems i need to look into it some more

The nvidia dkms may not be compatible with the upgraded kernel. You probably just need to do a manual install of the nvidia driver. Alternatively, you can unblacklist you igpu and use that until the dkms is updated for your kernel.

What you are experiencing is an Nvidia thing and their unwilli ngness to contribute to open source drivers and her complicated way of keeping their driver proprietary while plugging into the existing opensource stack.

1 Like

Oh I’m well familiar with Nvidia’s BS, just didn’t realise it was so easy for everything to fall over even on a “stable” ubuntu install, but we live and learn.

I did try reinstalling the kernel which I thought would also reinstall the kernel modules, but I didn’t try the nvidia packages directly so I’ll give that a go this evening and see if it works.

I don’t believe the iGPU was ever blacklisted - the nvidia card is only used for compute, everything has always been set to output primarily on the iGPU, though I’m not rulling out that the upgrade changed this.

At work rn but I’ll report back once I get home :slight_smile:

Reinstalled the driver but i’m seeing the same behaviour.

Kernel version 6.8.0-60-generic and nvidia-dkms-570

Not sure what the next steps should be

Edit: since i installed the drivers manually im now trying to Ubuntu-drivers utility to see if that makes a difference

So I purged everything nvidia related, installed with ubuntu-drivers --gpgpu

Aaaand still behaving exactly the same. Interestingly this installed 535 drivers instead of 570 but both are compatible with my 1060 so doubt that’s a factor.

Would love to learn exactly what i need to do to fix it, but I’m edging closer to just reinstalling the box and being done with Ubuntu… Any other recommendations I try before that?

Unfortunately, I run all AMD systems since 2004. There must be some configuration thing that happened in the upgrade. Canonical handles and puts configuration files in different places than the other Linux distros, even than parent Debian.

Is there an option to boot into a previous kernel? Or maybe a snapshot you can boot to/restore from the command line?
If looking for alternatives to ubuntu I can suggest mint as a derivative.

yeah I wanted to try it out again in hopes it was better than last time I used it but lesson learned, time to backup the data and reinstall

1 Like

are you using the nvida open or closed drivers?

1 Like

Few hours later and we’re back where I should have been to begin with, on a debian base with everything working smoothly - and the setup was considerably easier than what I did for ubuntu, so that’s the last time I bother with that.

Thanks for all the suggestions everyone <3

1 Like

Closed - the open ones are technically supported on the 1060 I think but general advice I’ve seen is to not bother with open for anything without RT.

All stable on debian now with the closed driver provided by debian non-free repos

Yeah, after the 900 series, Nvidia started signing their firmware and encrypting it to prevent the nouveau open drivers from supporting all features. Pretyy scummy. At the same time, they use and sometimes contribute to nouveau for their Nvidia shield hardware, but that kit pretty much runs 900 series GPUs and stripped down 1000 and 2000 series gpu architecture.

1 Like

Relevant to the discussion.