QEMU 11.1.0

QEMU emulates machines. Paired with KVM it provides the device model for nearly all Linux virtualization, and on its own it emulates foreign architectures.

What it does in a KVM setup

The split confuses people, and it is worth being precise. KVM is a kernel module that lets guest CPU instructions execute directly on the host processor. QEMU is the userspace program that emulates everything else: disk controllers, network cards, USB, graphics, firmware, and timers.

Used together, QEMU hands CPU execution to KVM and concentrates on devices. That combination is what people mean when they say “KVM.”

Our KVM, QEMU, and libvirt explainer covers the layering in full.

What is new in 11.1.0

We covered this when it shipped. The headline is RISC-V big-endian target support, plus expanded ISA coverage including the draft Zbr extension and Zvfbfa, KVM support for Zicbop and BFloat16, and the K230 board.

On ARM, the hvf accelerator gained nested virtualization and vGIC support. There is also Universal Flash Storage emulation, which removes a hardware dependency for anyone working on UFS drivers.

Running it directly

qemu-system-x86_64 -enable-kvm -m 4096 -smp 4 \
  -drive file=disk.qcow2,if=virtio \
  -netdev user,id=n0 -device virtio-net,netdev=n0

That works, and almost nobody manages virtual machines this way for long. libvirt provides the management layer, and virt-manager provides the interface.

The virtio point

The single most consequential choice when creating a guest. QEMU can emulate a real Intel network card, which works with any operating system and is slow because the guest issues hardware register writes that QEMU must interpret.

virtio devices are designed for virtualization. The guest knows it is a guest and talks to an efficient shared-memory interface. Disk and network throughput both improve substantially, and it is the reason a properly configured VM performs close to bare metal.

Foreign architectures

Without KVM, QEMU emulates the CPU too:

qemu-system-aarch64 -machine virt -cpu cortex-a72 -m 2048 ...
qemu-aarch64 ./arm-binary      # user-mode, run a single foreign binary

Slow, and it is how you test an ARM image on an x86 machine or run a Raspberry Pi image without a Pi.

Installing

sudo apt install qemu-system qemu-utils      # Debian and Ubuntu
sudo dnf install @virtualization             # Fedora
sudo pacman -S qemu-full                     # Arch

Distribution packaging lags QEMU point releases by varying amounts, and Debian stable in particular will be some way behind.

Verify Your Download

Source tarballs on download.qemu.org are published with GPG signatures.