ati driver dual head 2nd monitor incorrect resolution

Peter Zijlstra a.p.zijlstra at
Tue Oct 12 06:10:02 PDT 2004

On Mon, 2004-10-11 at 17:06 -0400, Alex Deucher wrote:
> On Mon, 11 Oct 2004 12:49:04 -0700 (PDT), John H. <mistamaila at> wrote:
> > Ok
> > I upgraded to 6.8.1 on my fc2 machine.
> > I have a radeon 9600, and told my xorg.conf to use ati
> > driver.
> > 
> > If I have the following...
> >         Option      "MetaModes" "1600x1200-1024x768"
> > xinerama seems to work, but my second monitor, a flat
> > panel LCD at 1024x768, scrolls vertically once I log
> > in from xdm/gdm, so when I maximize apps in it I have
> > to scroll down or up to see them all.
> > 
> > If I do
> > Option      "MetaModes" "1024x768-1024x768"
> > 
> > Neither monitor needs to scroll(maximized apps work
> > fine) but obviously I want a higher resolution on the
> > first monitor, 1600x1200.
> > 
> > any ideas?
> > 
> use xinerama instead of mergedfb.  the scrolling is because with
> mergedfb, there is one big framebuffer that both crtcs point into.  it
> has be to big enough for both, so you get teh scrolling affect. you
> can lock the frame movement if you feel like editing the code in
> radeon_mergedfb.c, but then you risk having apps appear on a non
> visible part of the framebuffer.   The behavior you want is "regular"
> dualhead with xinerama.  to do that, set up  read the xinerama howto.
> In "regular" dualhead, each screen has it's own framebuffer and
> xinerama is used to stich them together as one single logical screen.

Any xinerama aware window manager will not put a window in the dead
space. Although with regular xinerama the dead space is not actually
backed by video memory it is still perfectly possible to put windows
there, so the problem doesn't really change one way or the other.

I attached a rudimentary patch to disable the panning under discussion.
I think I allready send this thing to Alex once. Unfortunatly I haven't
had any time the past few months to properly finish it.

The things that are lacking:
 - ban the mouse pointer from the dead space as well; now one can still
move the mouse in the undisplayed section; this gives rather weird
results with hardware assisted cursor blitting.
 - automagically enable the panning again when the virtual size exceeds
the physical size.

Kind regards,

Peter Zijlstra <a.p.zijlstra at>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: radeon_mergedfb_asyncpan_2.patch
Type: text/x-patch
Size: 11511 bytes
Desc: not available
URL: <>

More information about the xorg mailing list