On Thu, 2012-05-31 at 20:47 +0100, Richard W.M. Jones wrote:
On Thu, May 31, 2012 at 02:42:11PM -0500, Chris Adams wrote:
> Once upon a time, Adam Williamson <awilliam(a)redhat.com> said:
> > On Tue, 2012-05-29 at 20:36 -0400, Jared K. Smith wrote:
> > > Yes, that's a possible culprit. I've had massive problems with
VT-d
> > > enabled on both a Thinkpad T510 and on a Thinkpad X220. I don't
> > > pretend to understand what advantages VT-d is *supposed* to give me,
> > > but it's the first thing I turn off in the BIOS. In fact, on the
> > > T510, I couldn't even get an installation to complete without turning
> > > it off.
> >
> > You need it to run as a KVM host with anything resembling speed.
>
> Isn't VT-d only for VMs directly talking to the hardware, bypassing the
> host? You can run VMs with decent speed using virtualized drivers
> without VT-d (and unless you have storage controllers and network
> interfaces dedicated to each VM, virtualized drivers is the only secure
> method).
That's right. VT-d allows the I/O port space to be partitioned like
memory, so that devices (SCSI disks etc) can be passed directly to
guests which can use them without (in theory) compromising the host,
eg. issuing DMA requests to overwrite host kernel memory.
VT-x is what Intel calls the regular hardware virtualization
extensions.
Gah. Sorry, I stopped reading after VT-. =)
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net