I'm no expert, so please forgive me if I'm missing something obvious here. But when I look at the two hooks for hugepages - alloc_hugepages.sh and unalloc_hugepages.sh - two things appear odd to me:

1. The comment block at the top instructs the user to place both files into both /prepare/begin/ and /release/end/. Is this really correct? Surely you would only want the alloc hook when starting up the VM, and would only want the unalloc hook when shutting it down?

2. The alloc hook defines a prepMemory function, but as far as I can tell (with my limited Bash knowledge), nothing ever calls it? Is it not important to prepare that memory for allocation?

Is there any way/command to my /etc/libvirt/hooks/qemu.d/win10/prepare/begin/start.sh intentionally prevent my VM to be started? I'm setting up some conditions tests before the VM can be started.

After upgrading to libvirt 7.0.0 (at least, the version published to Arch's community package repo on 1/20/2021) I started getting errors from the hugepages.sh script. It was trying to call functions for the start/begin and started/begin events.

I'm not sure what has changed as the libvirt docs mention those qemu events being in place for years now, and I never received these errors before.

Creating some empty functions with the appropriate names seems to be working on my machine.

how to generate this values?

i have a 48thread machine (dual cpu), i have pinning my VM like this

  <cputune>
    <vcpupin vcpu='0' cpuset='10'/>
    <vcpupin vcpu='1' cpuset='34'/>
    <vcpupin vcpu='2' cpuset='11'/>
    <vcpupin vcpu='3' cpuset='35'/>
    <vcpupin vcpu='4' cpuset='22'/>
    <vcpupin vcpu='5' cpuset='46'/>
    <vcpupin vcpu='6' cpuset='23'/>
    <vcpupin vcpu='7' cpuset='47'/>
  </cputune>

and use this in kernel parameters

isolcpus=10,11,22,23,34,35,46,47 nohz_full=10,11,22,23,34,35,46,47 rcu_nocbs=10,11,22,23,34,35,46,47

how adapt these values to my config?

greetings

Sorry if this is a very beginner question, I'm trying to set up a macOS VM in a notebook with a single graphics card, so I read this guide

When using the hook script I can't launch the VM and when I try to stop the service ´sudo systemctl stop libvirtd.service´, the libvirtd process uses 100% CPU and never stops the service and I have to force restart the system.

Are there logs that I can give you to help me troubleshoot this?

I'm trying to pass my Powercolor RX Vega 64 to my guest as shown in this screenshot:

When I use sudo vfioselect -o I get the following error:

/etc/modprobe.d/vfioselect.conf written successfully.
Bind: 30:00.0
30:00.1
Unbind:

./vfioselect: line 107: 30:00.0
30:00.1: syntax error in expression (error token is ":00.0
30:00.1")

When I start it without the -o switch I have no problems however.

EDIT: After a reboot the devices are actually still bound to amdgpu instead of vfio-pci. So even when it's not giving me an error message, it is actually not working.

Guess I'll go back to manually edit the module files myself.

EDIT 2: Sry, it actually works after issuing a sudo update-initramfs -u. There should be a notification that this update is not automatically performed by vfioselect.
There's still the syntax error when using the -o switch, though.