My old notes on KVM | skybert.net

These are more old notes on using KVM on Debian, written around 2011. Please see this page on kvm for my most recent notes on KVM.

Setting up bridged networking

Be aware that you probably cannot do this using a wireless card as most of these don't support bridging. I had the problem that everything looked right until I did ifup eth0 on the virtual guest. Then my host connection to the internet died.

When using wired network on the host machine, the following configuration in/etc/network/interfaces works well for me. Note that I've set up a static IP for the guest here as this is what I need for running this in production:

# The primary network interface
iface eth0 inet dhcp
      post-up /etc/init.d/firewall start
      post-up ifup br0

# bridge for kvm
iface br0 inet static
      address 192.168.1.110
      network 192.168.1.0
      netmask 255.255.255.0
      broadcast 192.168.1.255
      gateway 192.168.1.1
      bridge_ports eth0
      bridge_stp off
      bridge_fd 0
      bridge_maxwait 0
      post-up sysctl -p /etc/sysctl.conf

The lines I've added to/etc/sysctl.conf to get it working with the KVM guests is:

# for KVM guests
net.bridge.bridge-nf-call-ip6tables = 0
net.bridge.bridge-nf-call-iptables = 0
net.bridge.bridge-nf-call-arptables = 0

kvm quits, just saying "Aborted"

Runningstrace -f in front of the kvm command showed that it couldn't allocate the memory:

# strace -f kvm  -net nic -net tap -m 1024 my.img
[pid  8260] mmap(NULL, 1073881088, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = -1 ENOMEM (Cannot allocate memory)

This could be verified by looking atfree -m, which yielded only 111MB free. Freeing up some memory so that kvm had the 1024 MB I allocated for it, solved the problem.

KVM fails to bring up the qemu window

This is one of those where the error message doesn't have anything to do with the actual error:

# kvm \
  -net nic,macaddr=00:00:00:00:00:04 \
  -net tap \
  -m 1024 /opt/virtual/palantir.img \
  -cdrom /home/torstein/tmp/debian-6.0.3-amd64-netinst.iso \
  -boot d
[..]
~~~~~~~~~~~~~~~~~~~~~~~~~~| DirectFB 1.2.10 |~~~~~~~~~~~~~~~~~~~~~~~~~~
(c) 2001-2008  The world wide DirectFB Open Source Community
(c) 2000-2004  Convergence (integrated media) GmbH
----------------------------------------------------------------

(*) DirectFB/Core: Single Application Core. (2011-10-17 12:04)
(*) Direct/Memcpy: Using libc memcpy()
(!) Direct/Util: opening '/dev/fb0' and '/dev/fb/0' failed
--> No such file or directory
(!) DirectFB/FBDev: Error opening framebuffer device!
(!) DirectFB/FBDev: Use 'fbdev' option or set FRAMEBUFFER environment variable.
(!) DirectFB/Core: Could not initialize 'system_core' core!
--> Initialization error!
Could not initialize SDL - exiting

As the error message hints at, there's indeed no framebuffer device on my box, for that I had to pass vga=something on the LILO/GRUB kernel parameter line. However, what's more interesting to me, is why it tries touse the frame buffer instead of the nice window it usually pops up.

And the reason? It was because my root user didn't have access to the local X session. Duh! Hence, simply doing:

$ xhost +

in a normal user shell remedied it. Man! This bugger made me scratch my head, I was even reading through the Linux kerenl documentation on framebuffers :-)

Another error message for the same reason

I've also discovered that this error message also has the same remedy:

Invalid MIT-MAGIC-COOKIE-1 keyCould not initialize SDL