Dear Maintainer,

after upgrading to qemu-system-x86 1:11.0.0+ds-2 my uefi windows 11
installation does not boot. It hangs in an endless loop trying to start.

The command line used is:
executing sudo /usr/bin/kvm  -name srv-vm-11 -rtc base=localtime -netdev vde,id=net0,sock=/var/run/vde2/tap0.ctl -device virtio-net-pci,netdev=net0,mac=56:54:00:00:02:0b -m 16G -global driver=cfi.pflash01,property=secure,value=on -drive if=pflash,format=raw,unit=0,file=/usr/local/vms/srv-vm-11/uefi-srv-vm-11/OVMF_CODE_4M.secboot.fd,readonly=on -drive if=pflash,format=raw,unit=1,file=/usr/local/vms/srv-vm-11/uefi-srv-vm-11/uefi_vars.fd -machine q35,accel=kvm:tcg -cpu EPYC-Rome-v4,fma4=off,tbm=off,xop=off,fsrm=off,erms=off,pcid=off,misalignsse=off,sse4a=off,cr8legacy=off,xsaves=off,mmxext=off,fxsr-opt=off,osvw=off,sha-ni=off,svm=off,vmx=off,cmov,cx8,fpu,fxsr,syscall,sse,sse2,cx16,lahf-lm,popcnt,sse3,ssse3,sse4.1,sse4.2,avx,avx2,bmi1,bmi2,f16c,fma,abm,movbe,xsave,xsaveopt,adx,aes,fsgsbase,rdrand,rdseed,rdtscp,x2apic,ibpb,stibp,virt-ssbd,amd-ssbd -smp threads=2,cores=6,dies=1,sockets=1 -device qemu-xhci -device usb-mouse -device usb-tablet -drive file=/usr/local/vms/srv-vm-11/srv-vm-11.img,format=raw,index=0,id=sda,media=disk,if=none,aio=io_uring,discard=unmap -drive file=/usr/local/vms/srv-vm-11/Win11_24H2_German_x64.iso,format=raw,index=2,id=sr0,media=cdrom,if=none,aio=io_uring -drive file=/usr/local/vms/srv-vm-11/virtio-win-0.1.266.iso,format=raw,index=3,id=sr1,media=cdrom,if=none,aio=io_uring -device virtio-scsi-pci,id=scsi0 -device scsi-hd,drive=sda,rotation_rate=1 -device scsi-cd,drive=sr0 -device scsi-cd,drive=sr1 -device virtio-rng-pci -device virtio-balloon-pci -chardev socket,id=virtiocon0,path=/run/srv-vm-11.con,server=on,wait=off -device virtio-serial -device virtconsole,chardev=virtiocon0 -device VGA,vgamem_mb=32,edid=on -display gtk -monitor unix:/run/srv-vm-11.mon,server,nowait -pidfile /run/srv-vm-11.pid -run-with user=1000:1000 -enable-kvm -chardev socket,id=chrtpm,path=/usr/local/vms/srv-vm-11/tpm0-srv-vm-11/swtpm-sock -tpmdev emulator,id=tpm0,chardev=chrtpm -device tpm-tis,tpmdev=tpm0

The behaviour does not change in dependence of the version of ovmf -
2025.02-9 from testing, 2025.02-8+deb13u1 from stable both show the
same boot problem. The version from stable was used in the command above.

Downgrading qemu-system-x86 to stable 1:10.0.8+ds-0+deb13u1+b2 solves
the problem.

Hi

I'm having the same problem here with version 1:11.1.1+ds-1. I'm also
stuck in an endless boot loop. Downgrading to the Testing branch didn't
help.


Greetings

     Holger...

Since I don't have a way to reproduce this behavior -
all my windows guests, including a collection of
guests used for testing new qemu, and multiple
(windows, various versions) guests used in production -
all works just fine with either version of qemu, --
you have to diagnose the problem and find out what
is going on there, why exactly your guests are
not booting correctly.

Please note also these versions of qemu are used
in multiple other places too, with no issues like
that.

I'm not saying the issue does not exist, I'm only
saying I don't have a way to reproduce it.

Thanks,

/mjt

I've since found the cause. This bug report describes the problem.

https://www.deepl.com/de/translator

I downgraded Qemu to version 10. After that, the VM booted again. I
disabled Virtual Machine Platform/Hyper-V in the VM as described in the
bug report. Unfortunately, WSL no longer works after that, but that's
okay. It works again with the version of Qemu in the “unstable” repository.

Holger...







Am 28.09.26 um 14:17 schrieb Michael Tokarev:

The link in the first email was incorrect.

https://gitlab.com/qemu-project/qemu/-/work_items/3215

Hi All,

I was able to stop the endless loop during the windows boot by replacing
the cpu flage mentioned in the bug report by
-cpu host,ibpb,stibp,virt-ssbd,amd-ssbd,pdpe1gb,topoext
After this change the vm starts also with gemu 1:11.1.0+ds-2.

Thanks and Regards,
Axel

So this is basically -cpu host

What was the flag which made it looping (which you removed?)?

Thanks,

/mjt

Hi

Out of curiosity, I just quickly installed Windows 11 Pro on my laptop
using Virt-Manager. I left everything at the default settings and didn't
change a thing. After the installation, I enabled “Virtual Machine
Platform” in the Windows 11 settings.

Result: An immediate boot loop after the reboot.

That doesn't happen on my desktop computers. Everything works fine.


If you'd like more information, please let me know. Attached is an inxi
-F output for the hardware being used.


$ inxi -F
System:
   Host: laptop Kernel: 7.2.8+deb14-amd64 arch: x86_64 bits: 64
   Console: pty pts/2 Distro: sidux 2008-04 Πόντος - kde-full -
(200812222147)
Machine:
   Type: Laptop System: TUXEDO product: TUXEDO InfinityBook S 17 Gen6 v:
N/A serial: N/A
   Mobo: NB01 model: NS50_70MU serial: N/A Firmware: UEFI vendor: INSYDE
v: 1.07.16NRTR4
     date: 05/26/2022
Battery:
   ID-1: BAT0 charge: 45.9 Wh (71.7%) condition: 64/73.9 Wh (86.6%)
volts: 7.85 min: 7.7
CPU:
   Info: quad core model: 11th Gen Intel Core i7-11370H bits: 64 type:
MT MCP cache: L2: 5 MiB
   Speed (MHz): avg: 646 min/max: 400/4800 cores: 1: 646 2: 646 3: 646
4: 646 5: 646 6: 646
     7: 646 8: 646
Graphics:
   Device-1: Intel TigerLake-LP GT2 [Iris Xe Graphics] driver: i915 v:
kernel
   Display: x11 server: X.org v: 1.21.1.24 driver: X: loaded:
modesetting unloaded: fbdev,vesa
     dri: iris gpu: i915 tty: 160x56 resolution: 1920x1080
   API: EGL v: 1.5 drivers: iris,swrast platforms: gbm,surfaceless,device
   API: OpenGL v: 4.6 vendor: mesa v: 26.2.3-2 note: console (EGL
sourced) renderer: Mesa Intel
     Iris Xe Graphics (TGL GT2), llvmpipe (LLVM 22.1.8 256 bits)
   Info: Tools: api: eglinfo,glxinfo x11: xdriinfo, xdpyinfo, xprop, xrandr
Audio:
   Device-1: Intel 500 Series Family On-Package High Definition Audio
driver: snd_hda_intel
   API: ALSA v: k7.2.8+deb14-amd64 status: kernel-api
   Server-1: PulseAudio v: 17.0 status: active (root, process)
Network:
   Device-1: Intel Wi-Fi 6 AX201 driver: iwlwifi
   IF: wlp0s20f3 state: up mac: bc:09:1b:87:b4:32
   Device-2: Realtek RTL8111/8168/8211/8411 PCI Express Gigabit Ethernet
driver: r8169
   IF: enp53s0 state: down mac: 80:fa:5b:9f:b9:3c
   IF-ID-1: virbr0 state: down mac: 52:54:00:dd:cc:3b
Bluetooth:
   Device-1: Intel AX201 Bluetooth driver: btusb type: USB
   Report: hciconfig ID: hci0 state: up address: BC:09:1B:87:B4:36 bt-v: 5.2
RAID:
   Device-1: bpool type: zfs status: ONLINE level: linear raw: size:
1.88 GiB free: 1.79 GiB
     zfs-fs: size: 1.75 GiB free: 1.67 GiB
   Components: Online: 1: nvme0n1p2
   Device-2: rpool type: zfs status: ONLINE level: linear raw: size: 396
GiB free: 337 GiB
     zfs-fs: size: 383.69 GiB free: 324.64 GiB
   Components: Online: 1: nvme0n1p3
Drives:
   Local Storage: total: raw: 931.52 GiB usable: 917.77 GiB used: 58.68
GiB (6.4%)
   ID-1: /dev/nvme0n1 vendor: Samsung model: SSD 970 EVO Plus 500GB
size: 465.76 GiB
   ID-2: /dev/nvme1n1 vendor: Western Digital model: WD Blue SN570 500GB
size: 465.76 GiB
Partition:
   ID-1: / size: 349.75 GiB used: 25.11 GiB (7.2%) fs: zfs logical:
rpool/laptop/debian-root
   ID-2: /boot size: 1.75 GiB used: 83.2 MiB (4.7%) fs: zfs logical:
bpool/laptop/debian-boot
   ID-3: /boot/efi size: 2 GiB used: 482.2 MiB (23.6%) fs: vfat dev:
/dev/nvme0n1p1
   ID-4: /home size: 357.66 GiB used: 33.02 GiB (9.2%) fs: zfs logical:
rpool/laptop/debian-home
Swap:
   ID-1: swap-1 type: partition size: 512 MiB used: 0 KiB (0.0%) dev:
/dev/zd0
Sensors:
   System Temperatures: cpu: 30.0 C mobo: N/A
   Fan Speeds (rpm): N/A
Info:
   Memory: total: 32 GiB available: 31.13 GiB used: 2.52 GiB (8.1%)
igpu: 64 MiB
   Processes: 400 Uptime: 1m Init: systemd Shell: Bash inxi: 3.3.41

Hi All,

the flags were -cpu
EPYC-Rome-v4,fma4=off,tbm=off,xop=off,fsrm=off,erms=off,pcid=off,misalignsse=off,sse4a=off,cr8legacy=off,xsaves=off,mmxext=off,fxsr-opt=off,osvw=off,sha-ni=off,svm=off,vmx=off,cmov,cx8,fpu,fxsr,syscall,sse,sse2,cx16,lahf-lm,popcnt,sse3,ssse3,sse4.1,sse4.2,avx,avx2,bmi1,bmi2,f16c,fma,abm,movbe,xsave,xsaveopt,adx,aes,fsgsbase,rdrand,rdseed,rdtscp,x2apic,ibpb,stibp,virt-ssbd,amd-ssbd

Unfortunately I had no time to remove the flags one after the other.
The replacement is basically -cpu host, with the addition of some flags
recommended by the qemu documentation.

Thanks and Regards,
Axel



On Mon, 28 Sep 2026 18:06:42 +0300 Michael Tokarev <mjt@tls.msk.ru> wrote:

For me, win11 guest does not boot in qemu 11.0 with just
`-cpu EPYC-Rome-v4`, while boots okay in qemu 10.0.
Where this cpu type come from?

There's no boot loop, it just sits there spinning CPU at 100% and nothing
seems to happen.   Maybe this CPU type (the EPYC one) is not suitable for
my host (which is AMD Ryzen Zen3), and this is why it doesn't boot.

I dunno if this is the same problem as you're seeing, but here we go.

Holger, can you show which cpu/flags did you have which broke?  I guess
your issue is different because it started with a later version of qemu.

But at least I can try to bisect the boot hang which I have, maybe
something will show up.

Thanks,

/mjt

https://lore.kernel.org/qemu-devel/81a1f791-37ef-4d26-bcf7-024eab4cd810@tls.msk.ru/T/#u

Alex, this seems to be your original issue.

Now I want to hear more details from Holger.

Thanks,

/mjt

Am 01.10.26 um 10:04 schrieb Michael Tokarev:
Attached is win11.xml from Virt-manager. I haven't set any special flags.

Holger...

<cpu mode="host-passthrough" check="none" migratable="on"/>

Does it work with this setting?

I hoped to see the flags which does *not* work.

It would be good if you find the non-working version (the one
which broke for you in 11.x) and provided the qemu command line
(instead of the xml file - because I never used libvirt and don't
really know how to use it :))

This way, I - hopefully - will be able to find which change on
the qemu side broke things, and we can fix it.

Thanks,

/mjt

Am 01.10.26 um 13:01 schrieb Michael Tokarev:

No, that doesn't work. That's already how it's written in the XML file.
But I don't know anything about the QEMU command line.

Holger...

This is good, so there's a non-working config.  Ok.

libvirtd logfile have whole qemu command line as used to run this VM.
The logfile should contain info from previous startups too.

Usually it's /var/log/libvirt/qemu/MachineName.log (for system VMs) or
$HOME/.cache/libvirt/qemu/logs/MachineName.log (for user VMs)
(or something like this).

Please find the complete command line (qemu-system-x86_64...) of a
failing attempt and show it.

Please note the issue you're seeing is unlikely the one seen by Alex.

Thanks,

/mjt

Am 01.10.26 um 14:20 schrieb Michael Tokarev:
Attached win11.log

I couldn't find any more log files.

...

This is the one I was looking for.  I reproduced this issue here (it doesn't reboot
endlessly, but it doesn't boot either).  Looking at it..

Thanks!

/mjt

Control: retitle -1  qemu-system-x86: upgrade to qemu-system-x86 1:10.0.0+ds-2 breaks Windows 11 boot on an AMD EPYC CPU (PME issue)
Control: forwarded -1 https://lore.kernel.org/qemu-devel/81a1f791-37ef-4d26-bcf7-024eab4cd810@tls.msk.ru/T/#u
Control: tag -1 + confirmed upstream

Unfortunately it was too fast.  Nope, I hadn't reproduced the issue, - my
windows guest(s) run just fine with all combinations of options, so far.
(That issue which I "reproduced" yesterday was due to my windows guest being
in some wrong state after my previous experiments with the original issue
reported in this bug.)

So I don't have a reproducer for the second issue.

BTWW, let's rename this bug.

Thanks,

/mjt

So we shouldn't expect a quick fix. Am I right?


...



Am 02.10.26 um 12:23 schrieb Michael Tokarev:

Hi

When I set the CPU to “Skylake Server v4 or v5” in Virt-Manager, the
Windows 11 VM boots without any problems.


Holger