On 08/10/2012 05:44 PM, Laine Stump wrote: > How about allowing multiple <vlan tag='n' [trunk='yes|no']/> at the > toplevel? Then you could have: > > <vlan tag='42'/> > > in the simplest case, or: > > <vlan tag='42' trunk='yes'/> > > if you wanted a trunk with a single tag, or: > > <vlan tag='42'/> > <vlan tag='43'/> > <vlan tag='44'/> > > if you want a trunk with multiple tags (trunk='yes' would be implicit). > > For either of these cases, the data definition can remain the same (as > long as we don't require that superfluous "trunk='yes|no'" attributes be > preserved across an iteration of parse/format.) I've implemented the data structure, parser, and formatter for the above proposal, and tied it into the domain and network config, as well as modifying the network driver to grab the most appropriate vlan object from the interface / portgroup / network configuration and make it available via a single function call: https://www.redhat.com/archives/libvir-list/2012-August/msg00790.html I haven't yet implemented actually *setting* the vlan for PCI passthrough with <interface type='hostdev'/>, but all of the xml2xml tests run properly, so hopefully it's all functional enough to hook into the openvswitch backend. -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list