Patch for setup_harddisks to use it for Xen-Domains "without disk"

[Maximilian Wilhelm]

Am Saturday, den 24 November hub Henning Sprang folgendes in die Tasten:

Hi!

> > The attached patch adds simple support for Xen Dom-Us which don't have
> > "a disk" as such (read: Not /dev/xvd[a-z] device) but get every
> > "partition" as a single /dev/xvd[a-z][0-9]+ device.

> Why not just use the solution for this described in the Wiki? That works
> without changing the FAI code.

Because I have to edit a script each time I want to setup a Dom-U.
Besides that you then have two ways of configuring disk stuff.

> Maybe, this patch might be interesting to make the file systems on these
> partitions, which, in the wiki solution requires explicitly write mkfs
> statements in the partition hook...

> Then again, not all people call their Xen disks xvd* - some use hd*,
> some sd* - all three work. At least currently, I also heard that this
> might change for some reason I did not fully understand - Fedora people
> say using hda/sda is not good, and they, and suse use xvda IIR. Still,
> the Debian tools for creating Xen DomU's use hda or sda, not xvda.

If you use a vanilla kernel the devices are called xvda.
I don't know what $distro does.

> And a last note - if you want more complex partitioning in xen domains,
> you can also just hand over the full disk, and give the domain a full
> hda, and partition it as usual.

Sure you can, but in general this might not be the best idea.
When you do partition a block device inside a Dom-U, you have to
fiddle a lot if you want to access the file system from the Dom-0 in
case your Dom-U is broken or you want to do $x.

Nearby if /dev/xvda exists, the script behaves as usual.

> You can then also try booting with pygrub - then you can load the kernel
> from inside the vm, which makes the vm self-containing, and the vm
> config file the only one additional part you need.

Sure.
As said above this is not changed.

Ciao
Max
-- 
	Follow the white penguin.