[XCSSA] Xen and security

[xcssa@xcssa.org]

On Tuesday 09 January 2007 21:03, xcssa-admin@xcssa.org wrote:
> Background:
>
> Xen works by requiring that the guest OS be slightly re-written
> to avoid unsafe operations (e.g. updating page table entries) by
> using a new method (e.g. a hypervisor call to Xen).
>
> I've been reading this:
> http://www.cl.cam.ac.uk/research/srg/netos/papers/2003-xensosp.pdf

Not any more if you're running Xen 3 and using a new AMD or Intel proc with 
Pacifica or VT microcode technology respectively:
http://en.wikipedia.org/wiki/Xen


> My question is, does this mean that a guest OS can violate the
> security model if it chooses to be malicious?  I'm sort of thinking
> that guest OS code running at ring > 0 might generate an exception and
> be terminated or hose itself, but wouldn't be able to affect any other
> guest OSes or the host.  Can anyone think of a way that it could?

Got me.  I thought exceptions and interrupts (even of Dom0) should be handled 
by the hypervisor (the only true ring 0 instance).. no?

Is this Travis?

Tweeks