Re: [Qemu-devel] [RFC v4 00/58] Memory API

[Michael S. Tsirkin]

On Wed, Jul 20, 2011 at 07:52:22PM +0300, Avi Kivity wrote:
> On 07/20/2011 07:10 PM, Michael S. Tsirkin wrote:
> >>  and vga could do it if the vnc display
> >>  is disconnected, or if the update rate is so high it can just assume
> >>  all memory is dirty.
> >
> >Yes, but it's not something *devices* should be doing.
> 
> No one knows about the frame buffer except the vga device.
> 
> >>  >What devices seem to need, instead, is enable/disable a region
> >>  >through a back channel.
> >>
> >>  What do you mean?
> >
> >At least for qxl, all it seems to do is enable/disable
> >logging when the region is enabled disabled.
> >I'm not sure I follow what does vmware do - something
> >similar only with a custom register instead of
> >the standard memory enable?
> 
> pv vga like vmware and qxl disables logging when the pv interface is active.

At least for qxl, logging seem to be disabled when region is
disabled:

    if (vga->map_addr && qxl->pci.io_regions[0].addr == -1)

-1 is a disabled region ...

> -- 
> error compiling committee.c: too many arguments to function