[netcf-devel] Undefining a bridge does not recreate the original interface

Daniel P. Berrange berrange at redhat.com
Thu Nov 5 14:48:26 UTC 2009


On Thu, Nov 05, 2009 at 09:40:54AM -0500, Hugh O. Brock wrote:
> On Thu, Nov 05, 2009 at 09:25:41AM -0500, Cole Robinson wrote:
> > On 11/04/2009 07:35 PM, David Lutterkort wrote:
> > >> At the end of the day, I'm looking for the best way to avoid future
> > >> virt-manager bug reports like:
> > >>
> > >> - Help! I bridged eth1 in virt-manager, restarted the app, then deleted
> > >> the bridge and eth1 is no longer in my list of interfaces!
> > > 
> > > I think it all boils down to who should do device discovery. As I
> > > understand it, what you want is for virt-manager to have a way to list
> > > all physical interfaces on a machine, figure out which ones are not used
> > > by any bridge/bond etc. and offer those for addition into a new bridge.
> > > 
> > > For the first bit, libvirt provides the virNodeDevice* calls;
> > 
> > Yeah, I didn't think of that. It should be sufficient.
> > 
> >  for
> > > figuring out which devices are free, we'd need to add something to
> > > netcf, e.g. new error codes to indicate 'interface has no configuration'
> > > and 'interface is slave of another interface' when calling
> > > ncf_lookup_by_name. 
> > 
> > Not sure if I follow. If an interface has no configuration it won't show
> > up in ncf_list_interfaces, and an enslaved interface will be referenced
> > in some bridge/bond/foo xml, so all that information is already
> > available. I think using the nodedev APIs to enumerate physical
> > interfaces should be sufficient for virt-manager.
> 
> I still think netcf should refuse to repurpose a device to already
> exist. In Cole's case, if someone tries to enslave wlan0 to a bridge
> and it already exists with an IP address and so on, surely we should
> say "You have to undefine wlan0 first" and exit? It seems to me
> anything else violates the principle of least surprise...

And it better protects people against accidental mistakes if you give an
explicit error if the device is already configured.


Daniel
-- 
|: Red Hat, Engineering, London   -o-   http://people.redhat.com/berrange/ :|
|: http://libvirt.org  -o-  http://virt-manager.org  -o-  http://ovirt.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505  -o-  F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|


More information about the netcf-devel mailing list