[Spice-devel] [PATCH 28/35] vdi port: redesign.

[Gerd Hoffmann]

> To check this I'll have to compile the vdi_port driver, since back
> porting the spice_vmc.c file from qemu-kvm-rhel6 would be too much
> work (there is not virtio-serial in current qemu used in upstrea
> spice). But I'll try to do it and send the patch over.

qemu on freedesktop.org has both vdi-port and spice-vmc drivers.  With 
vdi-port not being intented to be submitted upstream, but it is useful 
to have for the time being as the spice-vmc windows bits are not ready yet.

> Actually, the movement of write to the plug doesn't look right - the
> interface is basically a table that is filled by the user, in this
> case qemu.

Which plug?

Data structures we have (for all interface types):

   (1) Implementation -- static data for the (qemu) implementation of
       the interface (the function pointers).
   (2) Instance -- data for the actual instance of the interface.

The whole idea is best explained with qxl where you can actually have 
multiple instances:  There is a single qxl implementation, and there is 
one instance per qxl device (i.e. for the multihead case you have 
multiple instances).

The instance is used everythere to refer to the interface:  libspice 
passes it when calling the interface callbacks.  qemu passes it when 
calling spice_server_$interface_$action() functions.  It holds a pointer 
to the implementation, a pointer to the private libspice data and 
anything else needed (qxl id for example).


Oh, and I think the current wakeup + read/write callback scheme has it 
advantages.  Especially spice-server can call read() whenever it is 
ready to accept data.  With the data rates we have today spice-server 
will grab the data instantly after wakeup, but that will change once we 
xfer bulky clipboard data.

cheers,
   Gerd