Installing passthrough for virt-manager

I'm on MX25 Linux. I need a step by step guide to installing passthrough for virt-manager. My GPU specs are nvidia-smi

You can do GPU passthrough for your RTX 4070 on MX Linux (Debian-based) using the standard KVM/VFIO approach; the main steps are: enable IOMMU/virtualization in BIOS, configure the kernel with VFIO options, isolate the GPU from the host using vfio-pci, and then attach the GPU to your VM in virt‑manager while also handling NVIDIA’s anti-VM behavior.[7][9]

Below is a concise, practical step‑by‑step focused on MX25 + NVIDIA + virt‑manager.

1. Check hardware & install KVM stack

Reboot into BIOS/UEFI and enable:
- Intel: VT‑d / Intel IOMMU / “Intel Virtualization for Directed I/O”.
- AMD: SVM / AMD‑V and IOMMU / AMD‑Vi.[7]

Boot MX25 and install KVM+virt‑manager and UEFI firmware:

sudo apt update
sudo apt install qemu-kvm libvirt-daemon-system libvirt-clients virt-manager ovmf

Add your user to libvirt and kvm groups, then re‑login:
```
sudo usermod -aG libvirt,kvm $USER
```
This is the typical setup on Debian/MX for virt‑manager.[9][7]

2. Enable IOMMU in the kernel

Find if you are on GRUB (most MX installs are). Edit:
```
sudo nano /etc/default/grub
```
In the line starting with GRUB_CMDLINE_LINUX_DEFAULT, add:
- For Intel CPU:
  - intel_iommu=on iommu=pt
- For AMD CPU:
  - amd_iommu=on iommu=pt[8][7]
Example:
```
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash intel_iommu=on iommu=pt"
```
Update GRUB and reboot:
```
sudo update-grub
sudo reboot
```
After reboot, verify IOMMU:
```
dmesg | grep -e IOMMU -e DMAR
```
You should see lines indicating IOMMU enabled / DMAR active.[8][7]

3. Inspect IOMMU groups and get GPU IDs

You want the RTX 4070 (and its audio function) in their own IOMMU group or in a group where everything can be passed to the VM.

Install pciutils if missing and list groups:

sudo apt install pciutils
find /sys/kernel/iommu_groups/ -type l

Identify the GPU and its audio device:
```
lspci -nn | grep -E "NVIDIA|VGA|Audio"
```
On a 4070, you’ll see something like:
- 01:00.0 VGA compatible controller: NVIDIA Corporation ... [10de:XXXX]
- 01:00.1 Audio device: NVIDIA Corporation ... [10de:YYYY]
Note both IDs in the [vendor:device] form (for example, 10de:2786 and 10de:22bc).[7][8]

4. Bind GPU to vfio-pci (hide it from host Xorg)

Because your Xorg and desktop currently use the 4070, you need to move it to vfio-pci so only the VM can use it.[8][7]

Create a modprobe file:
```
sudo nano /etc/modprobe.d/vfio.conf
```

Add (using your own IDs from lspci):

options vfio-pci ids=10de:XXXX,10de:YYYY
options vfio-pci disable_vga=1

Ensure required modules load early:

echo -e "vfio\nvfio_pci\nvfio_iommu_type1\nvfio_virqfd" | sudo tee /etc/modules-load.d/vfio.conf

Blacklist the regular NVIDIA stack for the host:

sudo nano /etc/modprobe.d/blacklist-nvidia-gpu-passthrough.conf

Add:

blacklist nvidia
blacklist nvidia_drm
blacklist nvidia_modeset
blacklist nouveau

This prevents the host Xorg from grabbing the 4070.[7][8]

Rebuild initramfs and reboot:
```
sudo update-initramfs -u
sudo reboot
```
After reboot, confirm vfio-pci is in use:
```
lspci -nnk -d 10de:XXXX
```
The “Kernel driver in use” should show vfio-pci.[8][7]

5. Prepare Windows VM in virt‑manager

You should ideally have a second GPU or iGPU for the Linux host’s display; otherwise you’ll be headless on the host and manage it over SSH. Most passthrough guides assume two GPUs.[3][7]

Start virt‑manager, create a new VM:
- OS type: Windows 10/11 (depending on ISO).
- Firmware: choose UEFI (OVMF) in the Overview → Firmware field.[7]
- Storage: place VM image on a fast disk.
Before first boot, go to:
- “Add Hardware” → “PCI Host Device”
- Add:
  - The RTX 4070 (VGA device).
  - The NVIDIA audio device on the same bus (01:00.1 etc.).[10][7]
For better performance:
- Set CPU model to “host‑passthrough” in CPU settings.
- Use VirtIO disk and network drivers for Windows (via VirtIO ISO).

6. Fix NVIDIA drivers detecting a VM

Recent NVIDIA Windows drivers may refuse to work if they detect they are inside a VM. You can hide KVM from the guest via XML changes.[2][7]

In virt‑manager, open the VM → “Overview” → “XML” or “Edit XML” (enable XML editing in preferences if needed).[2][7]
Inside the <features> section, add:
```
<kvm>
  <hidden state='on'/>
</kvm>
```
This is equivalent to setting kvm=off and helps NVIDIA drivers load correctly in the guest.[2][7]
Ensure the <hyperv> feature is either disabled or minimally used if you run into error 43‑style driver issues.

7. Boot VM and install drivers

Boot the VM; it should show output on the monitor connected to the 4070 (if directly cabled) or via SPICE/VNC (though performance is on the passed‑through GPU).[7]
In Windows, install:
- Chipset drivers, if needed.
- Latest NVIDIA Game Ready or Studio drivers for the RTX 4070.[1][7]
Check Device Manager:
- No yellow exclamation on the NVIDIA GPU.
- NVIDIA Control Panel opens normally.

If the driver refuses to load or you get code 43–type symptoms:

Double‑check the <kvm hidden='on'> feature is present.
Confirm the host is not loading any NVIDIA driver for that card (lspci -nnk shows vfio-pci only).[2][7]

8. Quick comparison: host vs VM setup focus

Aspect	Host (MX25) setup goal	VM (Windows) setup goal
IOMMU/BIOS	Enable VT‑d / IOMMU / AMD‑Vi	Not applicable
Kernel params	`intel_iommu=on` or `amd_iommu=on iommu=pt`	Not applicable
Driver binding	RTX 4070 bound to `vfio-pci`, not `nvidia`	NVIDIA Windows driver sees bare‑metal GPU
virt‑manager config	OVMF firmware, PCI passthrough devices added	OS type set to Windows, CPU host‑passthrough
Anti‑VM handling	`<kvm hidden='on'/>` in XML	NVIDIA driver loads without error 43‑style issue

[9][2][8][7]

If you paste the output of lspci -nn | grep -E "NVIDIA|VGA|Audio" and confirm whether you have an iGPU or second card for the host, a tailored set of exact lines for your vfio.conf and GRUB entry can be provided.

1
2
3
4
5
6
7
8
9
10