Upgrade from Fedora 42 to 43 results in black screen during boot/single user mode recovery

Ask Fedora
, ,
1

This is the first time I have run into this one before.
After upgrading from F42 to F43, the first boot it was fine, went right into SDDM then into KDE without an issue.

However, after running another dnf update just to make sure and doing some cleanup of old kernels and stuff, after rebooting it would start and just show a black screen. Could not switch TTY either, just hung.

I booted into single user mode (adding single to end of kernel line in grub), and got the shell and checked last boot with journalctl -b-1 and didn’t really see anything that would cause it. A couple permission issues, service saying it couldn’t find a group which I know existed, and some hardware USB messages and some selinux messages, which selinux is in permissive mode.

I tried to tweak the udev rules and edit the selinux perms and got rid of a few messages but would still boot to a blank screen. Finally just for the heck of it, I tried to do a regular exit from the single user debug mode and when it did that, it continued to boot successfully into SDDM then I was able to log into KDE.

So oddly enough to get into my system now, I have to go into single user mode, then exit that, and I can do it as soon as the login prompt occurs, then it boots successfully. Otherwise a regular boot results in a hang.

I do have 3 nvidia video cards in the system, one of which is blacklisted and used as a pass-through into libvirt VM’s. The other two drive 3 monitors. This worked well under F42 with the occasional screen refresh issue, so I wouldn’t think this would cause any issue, especially since I can continue the boot and it does go in from single user mode.

Now I do know there are another 282 packages to install/upgrade since did this, so I will do that and if it fixes it, I’ll post here. Otherwise, has anyone run into a scenario like this before?

1 Like
3

There’s another report of a similar issue after applying recent updates here:

However, I know you’ve examined the logs and I guess you didn’t see anything like this user did?

sddm-helper-start-wayland[1626]: "/usr/bin/kwin_wayland: symbol lookup error: /usr/bin/kwin_wayland: undefined symbol: _ZN21KGlobalAccelInterface10keyPressedEi\n"
4

Please try running sudo dnf distro-sync --refresh and ensure the system is 100% up to date.

5

Nope. Closest thing us a sddm-helper[3341]: Detected locale “C” with character encoding “ANSI_X3.4-1968”, which is not UTF-8.

The boot that goes blank doesn’t even reach sddm

6

I ran the updates I was missing like I mentioned, then I checked the refresh and the only change was a downgrade to cockpit which I did. I had some updates to it from 45drives, I removed those.

I do know that the boot which hangs stops here:

…
Nov 03 09:31:26 Dilbert-KVM nm-dispatcher[2743]: Failed to reload iscsi.service: Transport endpoint is not connected
Nov 03 09:31:26 Dilbert-KVM nm-dispatcher[2743]: See system logs and ‘systemctl status iscsi.service’ for details.
Nov 03 09:31:31 Dilbert-KVM kernel: logitech-hidpp-device 0003:046D:4076.0006: HID++ 4.1 device connected.
Nov 03 09:31:36 Dilbert-KVM systemd-udevd[1037]: 0000:03:00.0: Worker [1100] processing SEQNUM=4461 is taking a long time.
Nov 03 09:31:36 Dilbert-KVM systemd-udevd[1037]: 0000:35:00.0: Worker [1124] processing SEQNUM=4661 is taking a long time.
Nov 03 09:31:36 Dilbert-KVM systemd-udevd[1037]: 0000:34:00.0: Worker [1118] processing SEQNUM=4645 is taking a long time.

The successful boot which I have to go into single user mode, then continues, doesn’t show any results from nm-dispatcher, and doesn’t show the 3 processing taking a long time. The 3 udevd entries are the 3 nvidia cards, instead, it shows a failback error message:
Nov 03 09:32:21 Dilbert-KVM kernel: nvidia 0000:35:00.0: [drm] fb1: nvidia-drmdrmfb frame buffer device
Nov 03 09:32:21 Dilbert-KVM systemd[1]: nvidia-fallback.service - Fallback to nouveau as nvidia did not load was skipped because of an unmet condition check (ConditionPathExists=!/sys/module/nvidia).
Nov 03 09:32:21 Dilbert-KVM kernel: logitech-hidpp-device 0003:046D:4069.0007: HID++ 4.5 device connected.

I checked and the nvidia drivers are installed, akmod and kmod are there, but looks like I need to try to reinstall nvidia drivers. Just odd though it works fine through single user into GUI vs boot direct into GUI.

7

Tried to reinstall the drivers and get the same result.
I also check on the boots where I am able to get in via single user, the nvidia kernel module is loading.

I do see a lot of messages showing up about black screens. Only thing is mine is locking and won’t switch to tty CTRL-ALT switch won’t work, either system is hung or USB dies so keyboard won’t work, will have to try a PS2 keyboard if I can find one.

8

So interestingly, I applied the latest set of updates today and it actually rebooted.
It hung on a libvirt error I had with an invalid storage pool but continued to boot.

I removed the invalid storage pool entry and rebooted and then it went back to the same issue, locking up. Going into single user mode and escaping out is the work around for now.

Almost seems like a race condition.