ERROR : Deprecated CPU topology since qemu update 9.0 (~2 weeks)

English Virtualization
1

this problem is supposed to be fixed in the next qemu release, which might or might not be anytime soon.
in the meantime, and since 2 weeks, i can’t launch my VM with GPu passthrough, and i can’t find any older version of qemu on the tumbleweed repos that would allow me to do a downgrade.
i’m just stuck since 2 weeks with no way of running my VM.

i’ve been extremely disapointed by tumbleweed recently,
i’ve encountered a ton of problems since months, from visual glitches solved by mesa updates (taking a few days to get fixed), to a broken breeze dark theme (happening since ~May and still not fixed)

sorry for the rant but this had to be said.
if anyone has a solution for this problem, please let me know

2

@kyral have you tried starting manually with virsh, then connecting? I just deleted and added my gpu back with virsh edit <machine> after detatching.

3

yeah, i’ve tried removing the problematic arguments, but whenever i start the VM with virsh or virt-manager, the args come back and the error appears, some people mentionned that workaround that you’ve mentionned but it’s no longer working

4

@kyral You have to detach with virsh…
In my case the PCI ID for the GT710 is 8:00:0 and 8:00:1, so;

systemctl stop libvirtd
systemctl start libvirtd

virsh nodedev-detach pci_0000_08_00_0
Device pci_0000_08_00_0 detached
virsh nodedev-detach pci_0000_08_00_1
Device pci_0000_08_00_1 detached

virsh start Windows_11_Pro
Domain 'Windows_11_Pro' started

Then use virt-manager or client to connect.

You can use virt-manager to add back in if deleted, but you need to do above each time your wanting to start the vm with gpu.

5

i’ll try that, it’s a PITA to do it everytime though

6

@kyral you may not have to all the time, the few restarts of the vm I did worked fine.

7

so there’s no other way ? downgrading the package isn’t a solution ?

8

@kyral it’s an option as always to downgrade, you could use tumbleweed-cli, or just pick a snapshot from history https://download.opensuse.org/history/ and rollback the relevant packages. I only have the gpu passthrough as a test, not critical for me, so can live without the need.

9

well as far as the history goes, the only qemu package i found, is aready part of the broken builds, i tried downgrading from there but it doesn’t fix the issue, this problem appeared on qemu 9.0

10

@kyral You can go back revisions on the Build Service, checkout and rebuild locally… Show Virtualization / qemu - openSUSE Build Service

11

i’ve tried doing that but i failed, i guess i’ll just wait for an update, this is too complicated for me…

12

well i’ve tried the latest qemu 9.0.2 that fixes the deprecated topo error, (from the Virtualization Staging repo)
the error no longer happens, but i’m still getting this weird error when i try to run the vm with virsh:

qemu-system-x86_64: GLib: ../glib/gmem.c:177: failed to allocate 18446744065192157996 bytes

it tries to allocate 1,84467E+07 terabytes, which makes absolutely no sense

13

fixed with qemu 9.0.2 and linux 6.10

14

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.