Not booting with Linux 6.6.48 but still booting with 6.6.42

Hi,

Since the last system updates QubesOS doesn’t boot anymore with the new default Linux. After showing GRUB, instead of starting the boot process, the screen simply shows the blinking underline character and nothing else.

When I’m interrupting GRUB, I can navigate to the alternative boot entries and at least the second entry is still working (Linux 6.6.42).

  • Not booting: Qubes, with Xen 4.17.4 and Linux 6.6.48-1.qubes.fc37.x86_64
  • Still correctly booting: Qubes, with Xen 4.17.4 and Linux 6.6.24-1.qubes.fc37.x86_64

My current setup:

Brand: ASUS
CPU: AMD Ryzen 7 5800X 8-Core Processor
Chipset: Advanced Micro Devices, Inc. [AMD] Starship/Matisse Root Complex [1022:1480]
Graphics: NVIDIA Corporation GP107 [GeForce GTX 1050 Ti] [10de:1c82] (rev a1) (prog-if 00 [VGA controller])
QubesOS version: R4.2.3
BIOS: 4002

Any suggestions on how to proceed here? What could be the root cause? How could this even be debugged?

Thanks!

2 Likes

Same issue with a 22-core Intel CPU and nVidia GTX1080.

The older kernel(s) work, so I’m sticking with those. If the next kernel doesn’t work, then I’ll worry about it. ymmv

2 Likes

Based on this:

The bug should be fixed in newer kernel.

2 Likes

Thanks, @quantum and @apparatus . I really hope you are right.

On the one hand it indeed sounds logical. On the other hand it felt more severe than “just” a GPU issue because I couldn’t even reach another TTY during the blinking underline (e.g. by pressing CTRL + ALT + F2). I would expect that this should still be possible with such GPU issues.

Let’s see what will happen with the next update then. I’m planning to report back here.

1 Like

Tried the latest kernel, 6.10, but it has the same issue - boot stalls, so something else going on. Back to 6.6.42.

2 Likes

Thanks for testing and reporting this here, @quantum .

Does anyone have a suggestion on how to troubleshoot this? Are there some kind of log files that we could check?

It seems to me that the OS is not started at that point yet but that the error occurs when GRUB tries to initiate the startup of the OS. So maybe we should try to collect some GRUB logs?

1 Like

Issue was reported here:

Building recent kernels with the reverted commit works for me.
I’m not sure when this will be fixed, for now I’ve been including a revert commit patch in my kernel builds and it’s worked perfectly on all recent releases.

2 Likes

Issue is also present in kernel-latest 6.11.2-1

1 Like

Not booting with 6.6.54. Expected a nouveau issue would have been fixed asap, given the number of nVidia Linux users, so this delay is surprising.

3 Likes

Still not working with 6.6.54 on one of my Nvidia desktops but the other one with an older Nvidia card has never had a problem

3 Likes

Same issue with GTX1060.
6.6.54 also doesn’t work. We will be locked out of the system if the next update doesn’t solve the problem either.

1 Like

You won’t be locked out, the current kernel (6.6.42 in your case) that Qubes OS is using when you update dom0 won’t be removed during update.

1 Like

You can update the dnf.conf to save more than 3 kernel versions. I save 5

1 Like

Sorry for the beginner question, but what is the easiest way to install 6.6.42 and and force qubes 4.2 to use that?

Right now I have:

6.1.75 (The only one I have that boots)
6.6.54 That came with 4.2.2 and does not boot
6.11.2 That came with kernel-latest today and does not boot

All the previous instructions I can find for doing this are for Qubes 4.0 or 4.1 and I think the conf locations have changed.

Odd selection of kernels.

sudo qubes-dom0-update --action=reinstall (kernel version)

Or

sudo qubes-dom0-update kernel-6.6.42-1.qubes.fc37

1 Like

Thank you!

I didn’t make those selections, its just what was there.

I assumed 6.1.75 is leftover from my 4.1 upgrade.

As I mentioned the other two are what came with 4.2.2 and kernel-latest

If anyone else needs this, the actual command needed is:

sudo qubes-dom0-update kernel-6.6.42-1.qubes.fc37

1 Like

:+1:
You might want to increase the number of kernels you hold onto from 3 to 5 (in dnf.conf).

Update for those interested:
https://lore.kernel.org/nouveau/20241114004603.3095485-1-airlied@gmail.com/

I tested this patch earlier today and it works. It will probably be part of the next kernel release if it’s processed fast enough.

1 Like

4.2.3 ISO includes faulty kernel, anyone trying to install the first time may not find out what’s wrong @marmarek. 4.2.2 works fine.

also until this issue is fixed, better to ignore new kernels in dom0 adding exclude=kernel at the end of /etc/dnf/dnf.conf