[Gllug] running Xen on another VM

damion.yates at gmail.com damion.yates at gmail.com
Thu Aug 21 23:50:57 UTC 2008


On Wed, 20 Aug 2008, Richard Jones wrote:

> On Wed, Aug 20, 2008 at 02:58:51PM +0100, Peter Corlett wrote:
> > On Wed, Aug 20, 2008 at 01:44:13PM +0100, Nahuel Marisi wrote:
> > [...]
> > > I was wondering if I can actually virtualize Xen in Virtualbox
> > > or other similar (preferably open source or free) VM.
> > 
> > In theory it should be possible to nest virtual servers
> > indefinitely - and you can do this on S/390 - but in practice it
> > doesn't really work on x86 and you'd be better off scrounging
> > another PC from somewhere and using that.
> 
> It's not possible even in theory on a PC.  Unlike mainframes,
> neither paravirtualization (as in Xen) nor hardware virtualization
> can be nested.

I'm not sure this is true.  Certainly in light of Intel VT and AMD-V.

UML in UML might be, which means Xen could run Xen with effort?

If I understand correctly, in UML the kernel is implemented entirely
in normal system calls, rather than CPU code itself, so this could
nest.

Xen would be much harder, but if all you're running is yet another
taylored kernel that you know is avoiding scary opscodes it could
work.
 
> Also, full emulators like Bochs don't emulate the hardware accurately
> enough that you can run Xen or KVM inside them.  I know, coz I tried
> it once.

Bochs is/was incomplete, but that's all.  There is no reason it
couldn't be fixed to simply BE a full CPU enough to run an instance of
Linux able to run vmware for example.  Of course this would be
stupidly slow.

Damion 

-- 
Damion Yates - damion.yates at gmail.com
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list