Obviously looks like an IPv6 problem, indeed : nc -6 -l 127.0.0.1 5900 fails, but -4 works.
I must confess I'm not very familiar with IPv6, and we're just starting to deploy it in my environment.  Can you at least refer me to useful reading references in order to solve this ?
I currently use a dirty fix by setting up a dummy interface with an IP address, but I'd like to avoid this ugly solution in the future.

Thanks !

Regards,

Eric Viseur
Etudiant Ingénieur Civil Electricien
+32 497 92 36 80
LinkedIn Profile


2013/7/2 Cole Robinson <crobinso@redhat.com>
On 07/02/2013 03:40 AM, Eric Viseur wrote:
> Hi,
>
> I'm currently setting up a virtualization system using Fedora 18, KVM and Open
> vSwitch.  I've created and successfully integrated my OVS bridge into
> libvirt.  However, if this bridge doesn't have an assigned IP address (and it
> won't in production mode), I can't start any VM, with Spice complaining with
> the following error :
>
> ((null):6425): Spice-Warning **:
> reds.c:2977:reds_init_socket: getaddrinfo(127.0.0.1,5900): Address family for
> hostname not supported
> qemu-kvm: failed to initialize spice server
>
> I don't understand from where it comes, since the local loop interface is up.
> Can anyone help me on this ?  Thanks !
>

That error makes me think it's something to do with ipv6 vs. ipv4. But it's
tough to say without more details. Does this fail?

nc -l 127.0.0.1 5900

Try with -4 and/or -6 switches. If so, there's something wrong with your host
network configuration independent of qemu.

If that doesn't fail, I'd grab the qemu command line from
/var/log/libvirt/qemu/$vmname.log, and try to reduce it to a minimal
reproducer. Maybe something like:

qemu-kvm -netdev bridge,br=<BRIDGENAME> -spice port=5900,addr=127.0.0.1

You would need to alter the whitelist in /etc/qemu/bridge.conf for that to
work. Then stuff all the details in a qemu bug report.

- Cole