Re: [Qemu-devel] [PATCH] spice: add chardev (v3)

[Alon Levy]

On Thu, Jan 06, 2011 at 01:05:33PM +0100, Gerd Hoffmann wrote:
> On 12/17/10 16:01, Anthony Liguori wrote:
> >On 12/17/2010 07:39 AM, Alon Levy wrote:
> >>Adding a chardev backend for spice, for usage by spice vdagent in
> >>conjunction with a properly named virtio-serial device.
> >>
> >>Example usage:
> >>qemu -device virtio-serial -chardev spicevmc,name=vdagent,id=vdagent
> >>-devic
> 
> This example is incomplete btw ...

oops - the complete example is:

-device virtio-serial -device spicevmc,subtype=vdagent
-chardev spicevmc,id=vdagent,name=vdagent
-device virtserialport,chardev=vdagent,name=com.redhat.spice.0

> 
> >What doe this channel do?
> 
> It is a communication path between spice client and guest.
> 
> There used to be just one, for vdagent (spice guest agent) which
> sends mouse events for example.  There will be more in the future,
> and using chardevs allows us to handle that.
> 
> >I really don't feel comfortable with this. This is not connecting QEMU
> >to an existing interface that happens to fit our model.
> >
> >This is clearly a "library" that provides thin wrappers around internal
> >QEMU interfaces to implement code that belongs in QEMU outside of QEMU.
> 
> No.  This is about tunneling server/client connections through
> spice, so we can reuse the authentication and encryption provided by
> spice.
> 
> Assume you have your VM running on machine A and you are sitting in
> front of machine B.  You want use the smartcard attached to host B
> in your virtual machine.
> 
> Without spice you'll use this on machine A:
>    qemu -chardev socket,server,host=0.0.0.0,port=2001,id=ccid,nowait \
>       -usb -device usb-ccid -device ccid-card-passthru,chardev=ccid
> 
> and run "vscclient <machine A> 2001" on machine B.
> 
> With spice you'll use this on machine A:
>   qemu -chardev spicevmc,id=ccid,name=smartcard \
>       -usb -device usb-ccid -device ccid-card-passthru,chardev=ccid
> 
> and the spice client on machine B will forward the requests to
> vscclient (well, I think it is actually linked to libcaccard, which
> is effectively the same as vscclient just forwards the requests from
> the network to libcaccard too).
> 
> There is no hidden magic.
> 
> cheers,
>   Gerd
>