Re: [Qemu-devel] [PATCH 00/22] QAPI Round 1

[Michael Roth]

On 03/09/2011 07:14 AM, Avi Kivity wrote:
> On 03/09/2011 03:12 PM, Anthony Liguori wrote:
> > On 03/09/2011 02:51 AM, Avi Kivity wrote:
> > > On 03/08/2011 09:19 PM, Anthony Liguori wrote:
> > > > > Both the guest and the management agent (and both can listen for
> > > > > events). I don't see why guest->qemu RPC is problematic for
> > > > > migration - at least when qemu terminates it.
> > > >
> > > >
> > > > If it's terminated in QEMU, it's fine, but then it's not QMP
> > > > anymore. Let me think about whether there's a way to achieve this
> > > > without a guest->qemu RPC.
> > >
> > > Why not?
> > >
> > > { execute: 'write-keystore' arguments: { 'key': 'foo', 'value': 'bar'
> > > } }
> >
> > This is coming from the guest?
>
> Yes.
>
> > QMP doesn't do bidirectional RPC today. It could, but that is a
> > fundamental change in the protocol.
>
> Could use a separate channel for talking to qemu.

That's a possibility. Might be tricky to do right though...we wouldn't 
want to have a guest get a normal QMP session over the channel, so we'd 
likely end up with a separate QMP server that uses a different 
guest-specific dispatch table for commands.

Would be nice to not have to rely on multiple channels 
though...particularly in the case of isa-serial channels where available 
ports might be scarce. But I wouldn't consider that a showstopper either.