How to Set Up KVM Virtualization

KVM (Kernel-based Virtual Machine) turns Linux into a Type-1 hypervisor by loading two kernel modules—kvm and either kvm_intel or kvm_amd—that expose hardware virtualization extensions directly to guest VMs. Paired with QEMU for device emulation and libvirt as a management layer, the stack is production-grade and runs everything from desktop test VMs to large-scale cloud infrastructure. This guide walks through the full setup: checking CPU support, installing the software stack, configuring networking, and launching your first VM with virt-manager.

1. Verify Hardware Virtualization Support

KVM requires Intel VT-x or AMD-V extensions enabled in firmware. Check before installing anything.

grep -Eoc '(vmx|svm)' /proc/cpuinfo

A number greater than 0 means extensions are present. vmx is Intel, svm is AMD. If the output is 0, enter your UEFI/BIOS settings and enable the relevant virtualization option (often called Intel Virtualization Technology or AMD-V/SVM).

You can also run the dedicated check tool:

sudo apt install cpu-checker   # Debian/Ubuntu only
sudo kvm-ok

Expected output: INFO: /dev/kvm exists — KVM acceleration can be used. On Fedora/Arch this tool is not packaged; the /proc/cpuinfo check is sufficient.

2. Install KVM, QEMU, and libvirt

Debian / Ubuntu

sudo apt update
sudo apt install -y qemu-kvm libvirt-daemon-system libvirt-clients \
  bridge-utils virt-manager virtinst

Fedora / RHEL 9 / Rocky Linux 9

sudo dnf install -y @virtualization

The @virtualization group pulls in qemu-kvm, libvirt, virt-install, virt-manager, and supporting tools in one shot.

Arch Linux

sudo pacman -S qemu-full libvirt virt-manager virt-viewer \
  dnsmasq bridge-utils openbsd-netcat

On Arch, qemu-full includes all architecture targets. Use qemu-base if you only need x86_64 guests and want a smaller install.

3. Enable and Start libvirtd

libvirt exposes a management daemon that virt-manager and virsh connect to. Enable it at boot and start it now.

sudo systemctl enable --now libvirtd

Verify it is running:

sudo systemctl status libvirtd

The output should show active (running). On Fedora 38+ the socket-activated units (libvirtd.socket) are used by default—this is fine; libvirtd starts on demand.

4. Add Your User to the libvirt Group

Without group membership every libvirt command requires sudo. Adding your user to libvirt (and kvm on some distros) grants unprivileged access to the daemon and the /dev/kvm device.

sudo usermod -aG libvirt,kvm $(whoami)

Log out and back in, or use newgrp libvirt in the current shell, for the group change to take effect. Confirm with:

groups

5. Configure Networking

libvirt ships with a default NAT network (virbr0) that works out of the box for outbound internet access from guests. For guests that need to be addressable on your LAN—servers, for example—set up a bridge instead.

Option A: Use the Default NAT Network (Simplest)

Confirm the default network is active:

sudo virsh net-list --all

If default shows inactive, start and auto-start it:

sudo virsh net-start default
sudo virsh net-autostart default

Option B: Create a Bridged Network

A bridge gives guests a real IP on your physical network. The steps below use NetworkManager, which is present on Fedora, recent Ubuntu, and most desktop distros.

# Replace enp3s0 with your actual physical NIC (ip link to find it)
sudo nmcli connection add type bridge ifname br0 con-name br0
sudo nmcli connection add type ethernet ifname enp3s0 \
  master br0 con-name br0-slave
sudo nmcli connection modify br0 bridge.stp no
sudo nmcli connection up br0

After this, create a libvirt network definition that uses br0, or select the bridge directly in virt-manager when creating a VM.

6. Create a VM with virt-manager

virt-manager is a GTK application that wraps libvirt. It runs under both X11 and XWayland without any extra configuration.

1. Launch virt-manager from your application menu or run virt-manager in a terminal.
2. Click File → New Virtual Machine.
3. Choose Local install media (ISO image or CDROM) and browse to your ISO.
4. Set RAM and CPU. A minimum of 2048 MB RAM and 2 vCPUs is reasonable for most Linux guests.
5. Create a disk image—20 GB is a safe starting point. The default location is /var/lib/libvirt/images/.
6. On the final screen, tick Customize configuration before install if you want to change firmware (UEFI vs BIOS), add extra disks, or change the NIC model.
7. Click Begin Installation. A console window opens and the guest boots from the ISO.

7. Create a VM from the Command Line with virt-install

For scripting or headless servers, virt-install is the right tool.

sudo virt-install \
  --name ubuntu2404 \
  --ram 2048 \
  --vcpus 2 \
  --disk path=/var/lib/libvirt/images/ubuntu2404.qcow2,size=20,format=qcow2 \
  --cdrom /tmp/ubuntu-24.04-server-amd64.iso \
  --os-variant ubuntu24.04 \
  --network network=default \
  --graphics vnc,listen=127.0.0.1 \
  --noautoconsole

Use osinfo-query os | grep ubuntu to find the correct --os-variant string for your guest. Once the VM is created, connect to its console with:

sudo virsh console ubuntu2404

8. Verify the Setup

List all defined VMs and their states:

sudo virsh list --all

Check that KVM acceleration is actually in use (not software emulation):

sudo virsh domcapabilities | grep -i "kvm"

Inside a running guest, confirm the kernel sees a KVM clock source:

cat /sys/devices/system/clocksource/clocksource0/current_clocksource
# Expected: kvm-clock

Troubleshooting

Permission denied on /dev/kvm

Your user is not yet in the kvm group, or the group change has not been applied in the current session. Run newgrp kvm or log out completely and back in.

libvirtd fails to start

Check the journal for details:

sudo journalctl -u libvirtd -n 50 --no-pager

A common cause on Fedora is SELinux denials. Check with sudo ausearch -m avc -ts recent and, if needed, run sudo restorecon -Rv /var/lib/libvirt.

VM is slow or CPU usage is very high

Confirm the guest is using KVM acceleration, not TCG (software emulation). In virt-manager, open the VM details and check that the CPU type is not set to a software-only model. On the host, ls /dev/kvm should exist; if it does not, the KVM kernel modules are not loaded. Load them manually with sudo modprobe kvm_intel or sudo modprobe kvm_amd.

No network inside the guest

Verify IP forwarding is enabled on the host:

cat /proc/sys/net/ipv4/ip_forward
# Should be 1

If it is 0, enable it:

sudo sysctl -w net.ipv4.ip_forward=1
# To persist across reboots:
echo 'net.ipv4.ip_forward=1' | sudo tee /etc/sysctl.d/99-kvm-forward.conf