Win10 vm fails to start with pve kernel 5.3.18-1

[C.G.B. Spender]

Hi, as the title says. After rebooting with previous kernel 5.3.13-3 everything works as expected again.

The Start VM task for the Win10 vm errors on the latest kernel with "unable to read tail (got 0 bytes) " as it tries to start it on boot. Linux/FreeBSD vms and containers work normally. Trying to start the Win10 vm manually yields a stuck unkillable Start VM task which hangs the whole server on reboot. While stuck at the reboot there are some messages interspersed throughout the normal reboot output, like so:


The server then just hung on this screen for almost an hour at which point I lost patience and power cycled it.

config file for the relevant VM:

agent: 1
balloon: 2048
bios: ovmf
boot: dcn
bootdisk: scsi0
cores: 4
cpu: host
cpuunits: 1536
efidisk0: vm-storage:101/vm-101-disk-0.qcow2,size=128K
hostpci0: 06:00
hotplug: network,usb
machine: q35
memory: 4096
name: *********
numa: 0
onboot: 1
ostype: win10
protection: 1
scsi0: vm-storage:101/vm-101-disk-1.qcow2,discard=on,size=100G,ssd=1
scsihw: virtio-scsi-pci
smbios1: uuid=9fe652df-30ca-4947-bb0f-24239609e9fe
sockets: 1
startup: order=5,down=60
usb0: host=2-1.1,usb3=1
vga: qxl
vmgenid: 9210ad9c-7f77-41d4-a089-17889d9beaaf

Is there anything I can do to get the vm up and running on the latest kernel? Not that I necessarily need it, but trying to solve random issues like this is pretty much the whole point of having a homelab for me.

 

[NormalKid]

What you use - efi or legacy bios? Any other logs (etc/log/system.log or whatever in your system will be appreciated).

 

[C.G.B. Spender]

I use efi bios in the vm, the server itself uses legacy boot if that matters. I can provide more logs tomorrow (rest of household is using the nas the server runs atm and I don't wanna have to reboot now). Any other particular log that could be helpful apart from the file you mentioned? I've no idea where to start looking honestly.

 

[NormalKid]

I use efi bios in the vm, the server itself uses legacy boot if that matters. I can provide more logs tomorrow (rest of household is using the nas the server runs atm and I don't wanna have to reboot now). Any other particular log that could be helpful apart from the file you mentioned? I've no idea where to start looking honestly.

According to https://forum.proxmox.com/threads/web-interface-access-error-log-location.50105/post-233350 it’s /var/ log/syslog, start from there in any strange issue)

 

[NormalKid]

I use efi bios in the vm, the server itself uses legacy boot if that matters. I can provide more logs tomorrow (rest of household is using the nas the server runs atm and I don't wanna have to reboot now). Any other particular log that could be helpful apart from the file you mentioned? I've no idea where to start looking honestly.

There is more logs under /etc/pve, try out.

Did you update all or just kernel? What about storage configuration and status? (Ttry https://forum.proxmox.com/threads/after-nightly-backup-vm-dont-start-sos.16118/)

 

[GoblinHero]

I have exactly the same problem.
1. Everything is fully upgraded with apt-get dist-upgrade (kernel 5.3.18-1).
2. After reboot OPNSense (Freebsd based) VM does not start with "unable to read tail (got 0 bytes)" error in web interface. If I try to start VM from console - it just hang indefinitely. My second Win10 VM started normally.
3. After reboot with old 5.3.13-3 kernel everything works as usual.

Spoiler: Not working opnsense config

boot: c
bootdisk: virtio1
cores: 4
cpu: host,flags=+pcid
cpuunits: 2048
hostpci0: 05:00.0
hotplug: network
memory: 1500
name: fw.arda.link
net0: virtio=12:7F:89:7C:94:F6,bridge=vmbr0
numa: 0
onboot: 1
ostype: other
parent: i386
scsihw: virtio-scsi-pci
smbios1: uuid=becd4b4a-1863-4d3a-ae87-f5b9129838a7
sockets: 1
startup: order=1
virtio1: vm:vm-100-disk-0,size=3G
virtio2: /dev/disk/by-partlabel/optane58G_2_1G,size=1G

Spoiler: Working win10 config

boot: cd
bootdisk: virtio0
cores: 2
cpu: host
memory: 3072
name: Win10
net0: virtio=F6:CA:A7:3E:66:28,bridge=vmbr0
numa: 0
ostype: win10
scsihw: virtio-scsi-pci
smbios1: uuid=f008f000-9cc0-4f08-bbb4-fae2cbe207d5
sockets: 1
virtio0: vm:vm-150-disk-0,cache=writeback,size=64G
virtio1: /dev/disk/by-partlabel/optane58G_3_4G,cache=writeback,size=4G

 

[mstinsky]

The post I did shortly after this post seems to be the same issue: https://forum.proxmox.com/threads/pve-kernel-5-3-18-1-breaks-pci-passthrough.65104/#post-293922

 

[GoblinHero]

The post I did shortly after this post seems to be the same issue: https://forum.proxmox.com/threads/pve-kernel-5-3-18-1-breaks-pci-passthrough.65104/#post-293922

Thx. I had the same protection fault error in the log. So yes, seems like pci passtrough is the problem.

 

[C.G.B. Spender]

It would appear it is also h/w dependent. I just changed the server mobo and cpu (from sm x11ssl-f+xeon e3 1220v5 to sm x10dri-t+2x xeon e5 2630Lv4(supposedly; as they're engineering samples, shh) for the higher core count and more ram/pci(e) slots.

Ofc after booting up I forgot to manually boot into the previous kernel, thought what the hell let's see what happens, redid the passthrough settings for the new pci positions in the ui and wouldn't you know it the vm boots up with both usb and pci passthrough with 2 sockets and 40 cores like nothing ever happened.

Take from that what you will, ofc I'm not suggesting upgrade to fix the issue; the upgrade I did was a planned one, the new kernel just arrived before all the components did.