Re: [off-topic] Fully virtualisable CPUs

Bryn Paul Arnold Jones (
Fri, 6 Sep 1996 23:32:59 +0100 (BST)

On Fri, 6 Sep 1996, Alan Cox wrote:

> > Sorry about being off-topic, but what makes a CPU fully virtualisable
> > (that it can run several OSes in different VMs)? And what make the x86
> > not so? Is there any improvement in the latter x86 CPUs (Pentium, PPro)
> > that makes virtualising easier?
> Not that off topic I think..
> With all processors trying to do something that needs supervisor access
> causes an exception. The difference with a processor you can virtualise is
> that any instructions which allow you to query things like the supervisor
> state and other supervisor mode instructions also causes a trap. That means
> the underlying code (even a user mode program) can catch all of these traps
> and the program it is managing thinks it is in supervisor mode when in
> reality its in user mode but being 'helped' along.
> On the intel chips instructions that report the CPU state don't cause
> an exception. Pity as if they did we could conceivably just boot windows95
> or nt or OS/2 or whatever in a window (and vice versa).
> Alan

Hmm, that's not a very big difference (well it may be in silicon, but I
don't think it's to big), they still could do that in the next round of
P5/P6/P...'s ;).


PGP key pass phrase forgotten,   \ Overload -- core meltdown sequence 
again :( and I don't care ;)      |            initiated.
                                 / This space is intentionally left   
                                |  blank, apart from this text ;-)