KVM and the new virtio drivers

[Mark McLoughlin]

Hi,

(Replying to an old post just for the record)

On Wed, 2008-03-26 at 13:12 +0100, Dennis Jacobfeuerborn wrote:

> I'm trying to get a F9 guest running using the new virtio block device but 
> seem to run into trouble with LVM. I rebuilt the initrd in the guest adding 
> the options "--with=virtio_pci --with=virtio_blk" and after booting the 
> kernel correctly detects vd1 and vd2 but then fails to find any volume 
> groups. Using "media=disk" instead of "if=virtio" brings the guest up fine. 
> Is there a special option needed to make the kernel/initrd find the volume 
> group on these virtual block devices?

lvm got a patch to make it recognise "virtblk" in /proc/devices not long
after you posted:

* Wed Apr  2 2008 Jeremy Katz <katzj at redhat.com> - 2.02.33-11
- Adjust for new name for vio disks (from danpb)
- And fix the build (also from danpb)

> Are these drivers stable enough to maybe put them into the initrd by 
> default? I would be nice to be able to boot a F9 guest with the 
> paravirtualized drivers out of the box without having to build a custom 
> initrd image.

If you install the guest using virtio, then the drivers will
automatically be included in the initrd created during the install.

If you switch to virtio post-install, you need to use --with=virtio_blk
and --with=virtio_pci initially, but will then be automatically included
in any new initrds created while you're running the guest with virtio.

Cheers,
Mark.