An easy way to virtualize the running system

Clone this repo:
  1. a32c773 Merge by Andy Lutomirski · 3 months ago master
  2. 45f2b58 Merge pull request #17 from zevweiss/resolvconf-fix by Andrew Lutomirski · 3 months ago
  3. 8ff05f5 virtme-run: add --name flag by Zev Weiss · 11 months ago
  4. 4816957 virtme-init: make resolv.conf world-readable. by Zev Weiss · 3 months ago
  5. 17888c7 Merge pull request #15 from fzago-cray/forupstream by Andrew Lutomirski · 3 months ago

What is virtme?

Virtme is a set of simple tools to run a virtualized Linux kernel that uses the host Linux distribution or a simple rootfs instead of a whole disk image.

Virtme is tiny, easy to use, and makes testing kernel changes quite simple.

Some day this might be useful as a sort of sandbox. Right now it's not really configurable enough for that.

Virtme is hosted at in utils/kernel/virtme/virtme.git (web | git). It's mirrored on github.

How to use virtme

You'll need a Linux kernel that has these options (built-in or as modules)


For networking support, you also need CONFIG_VIRTIO_NET.

For script support, you need CONFIG_VIRTIO_CONSOLE.

For disk support, you need CONFIG_SCSI_VIRTIO.

That kernel needs to be sane. Your kernel is probably sane, but allmodconfig and allyesconfig generate insane kernels. Sanity includes:


You may also have better luck if you set:


An easy, somewhat-reliable way to generate a working config is to append the prereqs.config file to your .config and then run make defconfig.

Your host system will need to satisfy some prerequisites:

  • You need Python 3.3 or higher.
  • QEMU 1.6 or higher is recommended. QEMU 1.4 and 1.5 are partially supported using a rather ugly kludge.
    • You will have a much better experience if KVM is enabled. That means that you should be on bare metal with hardware virtualization (VT-x or SVM) enabled or in a VM that supports nested virtualization. On some Linux distributions, you may need to be a member of the “kvm” group. Using VirtualBox or most VPS providers will fall back to emulation.
  • Depending on the options you use, you may need a statically linked busybox binary somewhere in your path.

Once you have such a kernel, run one of:

  • virtme-run --kdir PATH_TO_KERNEL_TREE
  • virtme-run --installed-kernel
  • virtme-run --installed-kernel VERSION
  • virtme-run --kimg PATH_TO_KERNEL_IMAGE

Note that the --kdir and --kimg modes do not support modules yet.

You can then do things like cd /home/username and you will have readonly access to all your files.

Virtme gives you console input and output by default. Type ctrl-a x to exit. Type ctrl-a c to access the QEMU monitor.

For now, the virtme console is a serial console -- virtconsole seems to be unusably buggy. I don't know of any way to keep the tty state in sync between the host and guest, so resizing the host window after starting the guest may confuse guest libraries like readline.


If you want graphical output instead of console output, pass --graphics. Note that this is the opposite of QEMU's default behavior.

Architecture support

By default, virtme will use whatever architecture would be shown by uname -m. You can override this with --arch. Note that you may need to do some poorly documented fiddling for now to get non-native architectures working, and you will almost certainly need to set --root to a root that matches the architecture.

In general, the easiest way to configure a working kernel is to run:

virtme-configkernel --arch=ARCH --defconfig


x86 (both x86_64 and i386) is fully supported, although some odd KVM configurations may cause problems.


ARM is supported using qemu‘s vexpress-a15 machine. There is no built-in KVM support for ARM right now, although it might work by accident -- I don’t own a real KVM-capable ARM machine to test it on.

If you use any mode other than --kdir, you‘ll need to manually set QEMU’s -dtb option. I‘m not sure why -- I assumed that QEMU would provide its own device tree, but this doesn’t seem to be the case.


Aarch64 works out of the box if you have a new enough version of QEMU.


PPC64 appears to be reasonably functional.


Other architectures may or may not work. Adding support is trivial, so ping me if you need another architecture. Unrecognized architectures use a set of maybe-acceptable defaults.

Upcoming features

In the near term, the high-priority features are:

  • Support for modular virtfs and 9p for non-installed kernels.


Please see DCO.txt