Re: [Qemu-devel] [PATCH v2] RFC: qxl: allow to specify head limit to qxl driver

2015-06-12 Thread Frediano Ziglio
 
 On Do, 2015-06-11 at 10:38 +0100, Frediano Ziglio wrote:
  libvirt has this as a video card parameter (actually set to 1 but not
  used). This parameter will allow to limit setting a use can do (which
  could be confusing).
 
 Hmm.  So heads is '1' by default but ignored today.  When this starts to
 be actually applied that will break existing multihead setups I suspect.
 

Yes, unfortunately libvirt always set the default to 1 even if not used by 
driver.
However the Qemu default (with no parameter) is still unlimited so it's a 
libvirt issue more then Qemu.

  This patch rely on some change in spice-protocol which are not still
  accepted. See
  http://lists.freedesktop.org/archives/spice-devel/2015-June/020221.html.
 
 Spice-server changes, right?
 Needs #ifdefs so qemu continues to build with old spice-server versions.
 The spice-server changes need to be upstream first.
 

Yes, I'll do.

 Also: when we pass on the limit to spice-server anyway spice-server can
 take care to apply the limit both ways and not call the
 -client_monitors_config() callback with more than $limit monitors.  Has
 the advantage to reduce the test matrix:  Limit either works or doesn't.
 There will be no spice-server/qemu version combination where the limit
 is applied one way only.
 
   Gerd
 

Is always Qemu that decide to apply the limitation, default (even for 
spice-server) is still unlimited. As Qemu compiled with option on won't run 
with older version there will never this problem. However I would agree that 
the patch for Qemu keeping all limits on one place would be smaller.

Frediano



Re: [Qemu-devel] [PATCH v2] RFC: qxl: allow to specify head limit to qxl driver

2015-06-11 Thread Gerd Hoffmann
On Do, 2015-06-11 at 10:38 +0100, Frediano Ziglio wrote:
 libvirt has this as a video card parameter (actually set to 1 but not
 used). This parameter will allow to limit setting a use can do (which
 could be confusing).

Hmm.  So heads is '1' by default but ignored today.  When this starts to
be actually applied that will break existing multihead setups I suspect.

 This patch rely on some change in spice-protocol which are not still
 accepted. See
 http://lists.freedesktop.org/archives/spice-devel/2015-June/020221.html.

Spice-server changes, right?
Needs #ifdefs so qemu continues to build with old spice-server versions.
The spice-server changes need to be upstream first.

Also: when we pass on the limit to spice-server anyway spice-server can
take care to apply the limit both ways and not call the
-client_monitors_config() callback with more than $limit monitors.  Has
the advantage to reduce the test matrix:  Limit either works or doesn't.
There will be no spice-server/qemu version combination where the limit
is applied one way only.

  Gerd