On Fri, Jan 14, 2011 at 11:13:06AM -0700, Eric Blake wrote:
> On 01/13/2011 08:45 AM, Cole Robinson wrote:
> > @@ -3751,6 +3753,12 @@ qemuBuildCommandLine(virConnectPtr conn,
> > goto error;
> >
> > virCommandAddArg(cmd, str);
> > +
> > + if (sound->model == VIR_DOMAIN_SOUND_MODEL_ICH6) {
> > + virCommandAddArgList(cmd,
> > + "-device", "hda-duplex", NULL);
>
> Suppose I want a guest with two sound cards, both using model='ich6'.
> Am I correct that the qemu command line needs exactly one "-device
> hda-duplex" to enable sound from either card to get back to the host,
> but two "-device intel-hda,..." to implement the two cards in the guest?
> If so, this doesn't quite do the right thing (you instantiate the
> hda-duplex device twice); you'd need a bool variable set after the first
> time you emit this extra device.
The 'intel-hda' device should have an 'id' set on the command
line. The codec devices should then use that id value to associate
themselves with the intel-hda device.
QEMU generally allows you to omit all these ids and does some
'sensible' behaviour, but libvirt must never rely on that
because we need to guarentee a stable guest ABI. Everything
must thus be fully specified
Regards,
Daniel
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list