On 3/23/22 4:17 AM, Daniel P. Berrangé wrote:
> On Wed, Mar 23, 2022 at 12:01:19PM +0400, Marc-André Lureau wrote:
> > Hi
> >
> > On Wed, Mar 23, 2022 at 12:47 AM Jonathon Jongsma <jjongsma(a)redhat.com>
wrote:
> > >
> > > This patch series enables support for the qemu-vdagent character device
which
> > > enables copy/paste support between guest and client when using vnc
graphics.
> > >
> > > The guest must be configured with something like the following:
> > >
> > > <channel type='qemu-vdagent'>
> > > <source clipboard='on' mouse='on'/>
> > > <target type='virtio'
name='com.redhat.spice.0'/>
> > > </channel>
> > >
> > > Copy/paste sync requires a vnc client that has support for copy/paste
commands.
> > > Currently virt-viewer does not work, but the version of tigervnc provided
by
> > > fedora (executable name 'vncviewer') does work.
> > >
> > > More details about this device on Gerd's blog:
> > >
https://www.kraxel.org/blog/2021/05/qemu-cut-paste/
> > >
> > > OPEN QUESTIONS:
> > >
> > > - I'm not fully convinced that the `<source>` element is the
right place for
> > > the new `clipboard` / `mouse` configuration options, but I'm not
sure that
> > > adding a new xml element to the `<channel>` is the right approach
either.
> > > Suggestions welcome.
> >
> > I am not convinced <channel> is appropriate either for -chardev
> > qemu-vdagent. That's why in my dbus display series it's an
> > implementation detail
> >
(
https://patchew.org/Libvirt/20211222194345.766352-1-marcandre.lureau@redh...),
> > and it's not exposed to the domain XML directly. However I realize
> > that my approach is probably broken, as it looks like I totally missed
> > the associated virtserialport! :)
> >
> > As you say "source" is strange to put those extra channel options
too.
> >
> > And the only really valid associated device is a virtserialport of
> > name "com.redhat.spice.0", so what's the point in exposing that
detail
> > to the user?
You could say the same thing about the normal spice vdagent, right? And we
allow the user to specify a different virtio name for the spicevmc channel.
spice-vdagent does offer a --virtio-serial-port-path option that can be used
to connect to a different name within the guest, after all.
>
> At the QEMU level we have one serial port with chardv
>
> -chardev qemu-vdagent,id=ch1,name=vdagent,clipboard=on
> -device virtio-serial-pci \
> -device virtserialport,chardev=ch1,id=ch1,name=com.redhat.spice.0
>
>
> but IIUC, we can have multiple frontends - ie both VNC and DBus
> display, and in that case the clipboard=on property on the
> chardev will apply to both VNC & DBus displays. We can't
> control clipboard independantly for the displays, so it is
> right to put the control in the XML for the <channel> for
> both VNC and DBus.
>
In addition to <channel>, there are several other xml elements that
essentially map to a qemu character device: <serial>, <console>, etc. Would
it make more sense to introduce a new element like this for the
qemu-vdagent? This might make it easier to introduce new configuration
options that wouldn't really fit in the general <channel>. For example,
something like:
<qemuvdagent>
<clipboard copypaste="yes"/>
<target type='virtio' name='com.redhat.spice.0'/>
</qemuvdagent>
The devices in libvirt represent the guest exposed devics, and that's
still a virtio-serial port, so <channel> is the right place.
The chardev is merely a backend and doesn't need it own top level
XML element. This does mean yuo can use qemu-vgagent chardev with
any of serial, parallel, channel, etc but that's harmless from
libvirt's POV.
With regards,
Daniel
--
|: