I've been running Proxmox on a new AMD Ryzen 1700x server for the last few weeks and because I'm using GPU passthrough I've been using the testing repository because of certain AMD patches that are available in the newer kernel, this has been working great!
Yesterday I decided to run an apt upgrade & apt dist-upgrade because I have an event coming up where I need to use the server.
After a reboot suddenly the GPU passthrough VM wasn't working anymore. After trying a lot of things (older kernel, config files, new VM, re-checking everything), I couldn't even boot the Windows 10 ISO. After more testing I pinpointed the problem down to the VM being Q35 type. All my Q35 type VM's where not booting correctly anymore, normal i440 type still worked just fine.
I finally did a "apt remove *qemu*" which removed packages "proxmox-ve pve-container pve-ha-manager pve-manager pve-qemu-kvm qemu-server spiceterm".
I then decided to comment out the test repository, apt update again and re-install those exact packages but now from the stable branch. Immediately after that I tried to start the VM (Windows 10) again and immediately it started working again.
After a reboot and now still running the 4.13.13-6-pve kernel, everything is still working. So I believe something in the current test repository is braking Q35 guests.
I couldn't find how to submit a bug report and/or if I should but I wanted to make sure to share this if anyone else runs into the problem.
Yesterday I decided to run an apt upgrade & apt dist-upgrade because I have an event coming up where I need to use the server.
After a reboot suddenly the GPU passthrough VM wasn't working anymore. After trying a lot of things (older kernel, config files, new VM, re-checking everything), I couldn't even boot the Windows 10 ISO. After more testing I pinpointed the problem down to the VM being Q35 type. All my Q35 type VM's where not booting correctly anymore, normal i440 type still worked just fine.
I finally did a "apt remove *qemu*" which removed packages "proxmox-ve pve-container pve-ha-manager pve-manager pve-qemu-kvm qemu-server spiceterm".
I then decided to comment out the test repository, apt update again and re-install those exact packages but now from the stable branch. Immediately after that I tried to start the VM (Windows 10) again and immediately it started working again.
After a reboot and now still running the 4.13.13-6-pve kernel, everything is still working. So I believe something in the current test repository is braking Q35 guests.
I couldn't find how to submit a bug report and/or if I should but I wanted to make sure to share this if anyone else runs into the problem.
