device: 0.874
performance: 0.849
hypervisor: 0.835
semantic: 0.814
architecture: 0.798
socket: 0.781
graphic: 0.756
x86: 0.738
virtual: 0.726
network: 0.694
ppc: 0.667
files: 0.628
risc-v: 0.614
boot: 0.604
VMM: 0.604
PID: 0.598
kernel: 0.597
arm: 0.589
register: 0.547
vnc: 0.500
KVM: 0.487
i386: 0.483
permissions: 0.467
peripherals: 0.466
TCG: 0.426
mistranslation: 0.406
user-level: 0.380
debug: 0.342
assembly: 0.311

LXD with QEMU 6.2.0 (and 7.0.0-rc3) breaks during stateful migration
Description of problem:

Steps to reproduce:
```
sudo snap install --lxd
sudo lxd init --auto
lxc init images:ubuntu/20.04/cloud v1 --vm
Creating v1
lxc config device override v1 root size.state=2GiB
Device root overridden for v1
lxc config set v1 migration.stateful=true
lxc start v1
sleep 10
lxc exec v1 -- uptime
 22:05:54 up 0 min,  0 users,  load average: 0.07, 0.02, 0.00
lxc snapshot v1 --stateful
Error: Migration call failed
lxc snapshot v1 --stateful
Error: Monitor is disconnected
```

The first attempt at `lxc snapshot v1 --stateful` caused this in the `lxc info v1 --show-log` log output:

```
qemu-system-x86_64: qemu_savevm_state_complete_precopy_non_iterable: bdrv_inactivate_all() failed (-1)
```

The second attempt caused this:

```
qemu-system-x86_64: ../block.c:6757: bdrv_inactivate_recurse: Assertion `!(bs->open_flags & BDRV_O_INACTIVE)' failed.
```

Which crashed QEMU completely and caused the VM to die.
Nothing relevant showed up in dmesg, so this wasn't caused by an obvious seccomp or apparmor policy issue.
Additional information:
Originally reported by Stephane Graber at https://github.com/lxc/lxd/issues/9875