Sudden fatal error when running any VM with QEMU-KVM

[LVDave]

I'm running KUbuntu 20.04 as the host, and since yesterday, I'm getting the following error when attempting to start any VM I have set up.

https://imgur.com/P8zk50N

Up till now, everything worked fine. I've rebooted the system (Windows got me started doing THAT!), and noticing the "permission denied" on accessing
the kvm kernel module, I checked to be sure my user account was still a member of the "kvm" group.. It was. The only thing I could think of, was the
fact I wanted to test a flash drive I made yesterday to install OS using the Ventoy package on the flash drive. I ran "sudo kvm -hdb /dev/sdc" and it booted
the Ventoy menu. I didn't try running any vms then, until today. That "permission denied" appears to be a "smoking gun", but other than being a member
of the "kvm" group, I'm lost..

Thanks
Dave

[MAFoElffen]

There are three log files that you would want to look at for KVM problems...
1. $HOME/.virtinst/virt-install.log – virt-install tool log file.
2. $HOME/.virt-manager/virt-manager.log – virt-manager tool log file.
3. /var/log/libvirt/qemu/ – Log files for each running virtual machine.

Since you are having troubles with more than one VM, and it specifically notes, virt-manager, that is the log I would look at first.

You did say you installed a VM yesterday using sudo privileges, which gave that VM ownership as root, instead of a group permission, whic you may want to chenck the domain install lof for that VM...

Check the syslog, that kvm intitialized... (kvm [1]: Hyp mode initialized successfully)

And check your groups

$ groups
kvm libvirt

Then see if you can access these without problem or permission errors..

which qemu
which kvm
which virt-manager
kvm --version
virt-manager --version
sudo systemctl status libvirtd

[LVDave]

Thanks for the info.. Actually I didn't *install* a VM, I simply tested a Ventoy USB stick I'd just created, to carry the ISO files to be used FOR installing distros on bare-metal. In case you're not familiar with Ventoy, its a fantastic package that installs on a USB stick as two partitions, one called VentEFI, that boots the stick, the other partition can, when mounted, have ISO files copied to it, and when you boot the stick, you get a menu of the ISO files in the second partition and you can select which ISO you want to boot/install. I simply tested if the stick would boot to the menu. BUT... Now that I think about it, I did do the "kvm -hdb /dev/sdc" as root (sudo).. Am away from the machine with this problem right now but will check everything you listed when I get home... THANKS.. Kind of new to KVM...

[MAFoElffen]

If I do something with a VM booting off USB, then I set the details of that VM to use hardware USB pass-through... Then I boot from and write USB's from the VM... If you did that... Then I know the problem... LOL. (That is the only way a VM will boot from USB.)

It is a pain in the behind and does happen to me often. It happens even if you you take the USB out and plug it into the same USB port... I do not know why it is so picky, but it is. It's broken, until...

Open Virt-Manager, Go to the VM you had open to do that. Go to details view. Go to the USB Hardware from psychical, and remove/update. Everything will strat working again.

[LVDave]

Finally got time to try your suggestions..


Addressing the logs you mentioned, I don't have the following directories in $HOME..

$HOME/.virinst or $HOME/.virt-manager

I DO have the /var/log/libvirt/qemu with logs for each of my installed vms.. I did not see any log in the directory that might indicate it was a log for the
USB stick test I ran that apparently caused this problem.

As for groups, my user account is a member of kvm, libvirt

Here's a weird one..

which qemu is not found
which kvm is /usr/bin/kvm
which virt-manager is /usr/bin/virt-manager
kvm --version is QEMU emulator version 4.2.1 (Debian 1:4.2-3ubuntu6.17)
Copyright (c) 2003-2019 Fabrice Bellard and the QEMU Project developers
virt-manager --version is 2.2.1
systemctl status libvirtd is running


As for your comment where you said you knew where the problem was, let me clarify what I did with the USB kvm boot. I did not use Virt-manager at all with
this usb test. As I mentioned earlier in the thread, from the command line, I did "sudo kvm -hdb /dev/sdc". At no time did I use virt-manager to do this test. I recall testing a USB stick with an ISO before and did not have this malfunction.

[MAFoElffen]

hmmm. Let me look

The user's virt-manager.log is now located at $HOME/.cache/virt-manager/virt-manager.log

You had said that you now couldn't start "any" VM's. The logs of those VM's you tried to start and failed, might show "something" on why they are failing to start.

[LVDave]

Made an interesting discovery, where I can see all of my vms listed in virt-manager, if I did a "virsh list -all" I see NONE of them. Was going to try and
start one from virsh, if I can't see em, I can't start em.. Curious, is this getting to the point where I should just uninstall kvm and reinstall it?

[Doug S]

Made an interesting discovery, where I can see all of my vms listed in virt-manager, if I did a "virsh list -all" I see NONE of them. Was going to try and
start one from virsh, if I can't see em, I can't start em.. Curious, is this getting to the point where I should just uninstall kvm and reinstall it?

Did you mean to write "--all"?
For example, this is what I get:

doug@s19:~$ virsh list --all
 Id   Name      State
--------------------------
 1    desk-ii   running
 -    desk-ff   shut off
 -    desk-hh   shut off
 -    serv-xx   shut off

And I started the running VM via:

doug@s19:~$ virsh start desk-ii
Domain desk-ii started

[Dennis N]

Made an interesting discovery, where I can see all of my vms listed in virt-manager, if I did a "virsh list -all" I see NONE of them.

You may need to use sudo to run that. At least that's true on Fedora, which I use as my WM host. Notice:

[dmn@Tyana ~]$ virsh list --all
 Id   Name   State
--------------------

[dmn@Tyana ~]$ sudo virsh list --all
[sudo] password for dmn: 
 Id   Name                 State
-------------------------------------
 -    Ubuntu-2004-BIOS     shut off
 -    Ubuntu-2010-UEFI-2   shut off
 -    Ubuntu-2104-UEFI     shut off
 -    Zorin-16-UEFI        shut off

[MAFoElffen]

@Dennis N:

Well no... Not on Ubuntu (for Virsh), unless you are trying to do a virsh connect. Is default for Ubuntu:

ubuntu_system:~$ echo $LIBVIRT_DEFAULT_URI 
qemu:///system

But that doesn't exist outside of Ubuntu systems for virsh, unless you

export LIBVIRT_DEFAULT_URI=qemu:///system

Which will allow non-sudo users to do that.