Re: [RFC] Xen Virtual Framebuffer

[Anthony Liguori <aliguori@us.ibm.com>]

John Byrne wrote:

> Anthony,
>
> Looking at Xen from the perspective of managing a farm of machines, 
> what I want is a single method for remotely accessing the 
> console/graphics for a guest domain, regardless of whether the domain 
> is PV/VT.

Yes, this is what the driving use-case for me has been.  I've done a lot 
of experiment with various VNC methods until I settled on this as the 
best approach.

> (A method of connect a VNC client to dom0 port 5900 + domid would be 
> just fine. [From the code, this seems to be way qemu does it for 
> VNC.]) The current model of "xm console" and Xvnc for PV domains is 
> annoying. Unless, I am misunderstanding what you are doing, once you 
> add the VNC support, your work provides what I want.

Yes, that's the goal.

> So, what are the prospects of getting it finished and in, in the near 
> future? Is there any chance of getting it into the 3.0.x code base?

The first step is to get the local display stuff working.  Then we can 
look at integrating that in Xen.  I have a Task list on the wiki and 
it's not really that much more work.

The second step is proof-of-concept VNC support which will probably be 
with libvncserver.  The display will be sluggish but it will be a start.

The final step would be to implement an accelerated X driver so that we 
could get the high-level operations info (CopyRect for window dragging, 
hw cursor, etc.) and then we can implement an efficient VNC server.

As for what goes into Xen when, that's up to Ian.  This is of course 
pretty early though so it's probably premature to really ask that until 
there's more progress made and some more discussion.

Regards,

Anthony Liguori

> Thanks,
>
> John Byrne
>
>
>
>
>
>
>