Re: [Qemu-devel] [PATCH for mst/pci] output nc->name in NIC_RX_FILTER_CHANGED event

[Amos Kong <akong@redhat.com>]

On Wed, Jun 26, 2013 at 12:02:45PM +0200, Markus Armbruster wrote:
> Amos Kong <akong@redhat.com> writes:
> 
> > netclient 'name' entry in event is useful for management to know
> > which device is changed. n->netclient_name is not always set.
> > This patch changes to use nc->name. If we don't assign 'id',
> > qemu will set a generated name to nc->name.
> >
> > Signed-off-by: Amos Kong <akong@redhat.com>
> > ---
> >  hw/net/virtio-net.c | 11 +++--------
> >  1 file changed, 3 insertions(+), 8 deletions(-)
> >
> > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > index c88403a..e4d9752 100644
> > --- a/hw/net/virtio-net.c
> > +++ b/hw/net/virtio-net.c
> > @@ -200,14 +200,9 @@ static void rxfilter_notify(NetClientState *nc)
> >      VirtIONet *n = qemu_get_nic_opaque(nc);
> >  
> >      if (nc->rxfilter_notify_enabled) {
> > -        if (n->netclient_name) {
> > -            event_data = qobject_from_jsonf("{ 'name': %s, 'path': %s }",
> > -                                    n->netclient_name,
> > -                                    object_get_canonical_path(OBJECT(n->qdev)));
> > -        } else {
> > -            event_data = qobject_from_jsonf("{ 'path': %s }",
> > -                                    object_get_canonical_path(OBJECT(n->qdev)));
> > -        }
> > +        event_data = qobject_from_jsonf("{ 'name': %s, 'path': %s }",
> > +                           nc->name,
> > +                           object_get_canonical_path(OBJECT(n->qdev)));
> >          monitor_protocol_event(QEVENT_NIC_RX_FILTER_CHANGED, event_data);
> >          qobject_decref(event_data);
> 
> Is this on top of "[PATCH v3 0/2] mac programming over macvtap"?

It's based on mst's pci tree:

  https://git.kernel.org/cgit/virt/kvm/mst/qemu.git/log/?h=pci
  https://git.kernel.org/cgit/virt/kvm/mst/qemu.git/commit/?h=pci&id=1c0fa6b709d02fe4f98d4ce7b55a6cc3c925791c
 
> Yes, qdev IDs are optional, and therefore can serve as reliable
> identifier only when the user / management application always specifies
> one, and even then, you're still screwed for auto-created devices.
> Easily avoided for NICs, but yes, the problem is real.
> 
> However, I don't agree with the solution "use NetCientState name",
> because that's too ad hoc, and not general.  Can we please use QOM
> paths?

path is always provided in event, libvirt can use it.

-- 
			Amos.