Re: [Qemu-devel] [PATCH 2/3]: Add -uuid option to Qemu

[Andreas Färber]

Am 05.08.2008 um 15:53 schrieb Gleb Natapov:

> On Tue, Aug 05, 2008 at 03:43:39PM +0200, Gerd Hoffmann wrote:
> > Chris Lalancette wrote:
> > > As suggested by Anthony, this patch and the next one were split  
> > > apart.  This
> > > patch just adds the basic plumbing to qemu to be able to set the  
> > > uuid on the
> > > command-line, and retrieve the information via "info uuid".
> >
> > What is the point of CONFIG_UUID and the libuuid test in that patch?
> > I can't see any dependency on it in the source code.  IMHO the -uuid
> > switch should be present unconditionally.
> Here is the patch series where there is no CONFIG_UUID test (was  
> posted
> here two month ago):
> http://lists.gnu.org/archive/html/qemu-devel/2008-06/msg00443.html
> http://lists.gnu.org/archive/html/qemu-devel/2008-06/msg00444.html
>
> And this one checks for libuuid and uses it for random uuid  
> generation if
> available (UUID should be RFC compliant):
> http://lists.gnu.org/archive/html/qemu-devel/2008-06/msg00446.html

It seemed general consensus that we don't want values randomly  
generated at runtime in QEMU, rather having some constant, widely  
acceptable value and leaving it to either management tools or the user  
determining them. There was a long discussion about this topic before,  
concerning IPv4 subnets.

I don't really see a use case where we'd want a universally unique ID  
generated by QEMU at runtime either - it would seem better to generate  
it using some external tool (e.g., uuidgen or GUIDgen) and then pass  
the same value as command line argument (or config file option) to a  
certain virtual machine. After all it's supposed to be unique and not  
changing with each QEMU invocation.

Andreas