On Thu, 2011-06-30 at 11:55 +0100, Stefano Stabellini wrote:
> On Thu, 30 Jun 2011, Ian Campbell wrote:
> > On Wed, 2011-06-29 at 03:06 +0100, Madjik wrote:
> > > Hi,
> > >
> > > I'm currently trying xen 4.1.1 with the 3.0.0-rc5 (heard that it now
> > > contains the xen code to run as front or back xen domain).
> > >
> > > Everything seems to run fine, except that I am trying since a few hours
> > > to mount a .vhd file in the dom0.
> > Everything needed to run a guest is upstream but not everything needed
> > to run any guest, if you see what I mean. In particular the upstream
> > kernel contains blkback but not blktap so using a vhd would not be
> > expected to work today.
> > I'm afraid blktap is unlikely to be accepted upstream as it is. There is
> > a plan to remove the kernel component and have guests communicate
> > directly with the tapdisk process in userspace (via /dev/xen/gtndev et
> > al). There is no real downside to this since everything goes through
> > tapdisk anyway. I don't know how advanced that plan is currently.
> > With the 4.1 release you might be able to use the qdisk (qemu based
> > backend) to run a VHD but it is known not to perform all that well (and
> > I don't know off-hand how you enable it). Performance is, AIUI, better
> > with the upstream qemu port.
> If you want to run 3.0.0-rc5 as dom0 your best bet would be to convert
> your vhd file to a raw file, setup LVM, then dd the raw file into a new
> LVM volume. This in order to get the best performances.
> Otherwise the easiest thing to do would be to clone a kernel tree that
> contains blktap2, I believe Daniel has one based on 3.0, see the thread
> "xen 4.1.1 + 3.0.0-rc5 dom0 + blktap2".
Um, that's this thread ;-)
Xen-devel mailing list