USB on Virt. Boxes

Michael Torrie torriem at
Wed Oct 9 08:57:12 MDT 2013

On 10/09/2013 02:38 AM, Dan Egli wrote:
> Heh, yea, "from the host" is the point I meant. When I setup these
> machines, I'd like each virtual machine to have a unique MAC. My experience
> of VMware server from about three years ago showed that the GUEST eth0 had
> the HOST eth0's MAC address unless specifically overridden. That's what I
> was asking about. I was wondering if a new machine automatically generated
> a new MAC or if it had to be done (or COULD be done) separately.

They all generate the MAC address, but you can sure change it.
Additionally, with VirtualBox or KVM, you could generate the xml files
that define the virtual machine yourself, and use any MAC address you want.

> Well, USB is very important since much of the work that these boxes will be
> doing will involve heavy IO between the system and a USB device. I can't be
> sure since I haven't actually seen the device yet, but I imagine it's
> probably USB 2.0, although 3.0 is a distinct possibility. How does KVM
> handle USB 3.0 (or the new 3.1 for that matter)?
> Yea, sounds like KVM is good, except for the USB issue above. Any
> workarounds for USB 2.0 with KVM? If not, what would be your next
> recommendation?

So apparently qemu-kvm (the actual virtual machine) has supported
pass-through USB 2.0 for some time.  It's just that libvirt and
virt-manager (the GUI) lacked support.  As of virt-manager 0.9.1, this
is now present in the GUI.

Sadly Red Hat is still on 0.9.0 on RHEL 6, which is what I'm using.  But
if you were using something not quite so conservative, I bet it's all
there.  I might just try building a more recent version of virt-manager
to see if it's working (I also have a need for USB 2.0 support).

More information about the PLUG mailing list