On Fri, Oct 06, 2006 at 04:44:17AM -0400, Daniel Veillard wrote:
On Thu, Oct 05, 2006 at 07:23:15PM +0100, Daniel P. Berrange wrote:
On Wed, Sep 27, 2006 at 06:36:38PM +0100, Daniel P. Berrange wrote:
Ok, looks like we have total agreement. I'll knock up a patch implementing option #4 and post it for review.
Attached is a patch which implements option 4. If no <driver> block is supplied, then we continue existing behavior of using file: for file backed disks, and phy: for block backed disks. Any of the following are then valid:
<driver name='file'/> -> file: <driver name='phy'/> -> phy: <driver name='tap'/> -> tap:aio: <driver name='tap' type='aio'/> -> tap:aio: <driver name='tap' type='qcow'/> -> tap:qcow:
The only supported names are 'file', 'phy' & 'tap'. If the name is 'tap' then it is valid to use an additional 'type' attributte. We don't do any checking on cotents of 'type' attribute, it is just passed straight through to xend, since the blktap driver has a wide variety of types available. When fetching XML from libvirt you are now guarenteed to be given a <driver> block in each disk.
Attached is an update version of the patch addressing the issues mentioned. Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|