[PATCH] Radeon driver fixes

[Alex Deucher]

On Sat, 13 Nov 2004 14:04:31 +0100, Sjoerd Simons <sjoerd at luon.net> wrote:
> On Sat, Nov 13, 2004 at 12:22:32PM +1100, Benjamin Herrenschmidt wrote:
> 
> 
> > On Sat, 2004-11-13 at 01:08 +0100, Sjoerd Simons wrote:
> > > On Fri, Nov 12, 2004 at 04:29:05PM +1100, Benjamin Herrenschmidt wrote:
> > > > Here's a new version of the patch, changing the option to "VGAAccess",
> > > > and setting it OFF by default on PowerPC and ON by default on all other
> > > > archs (which mean it should do the right thing in most cases now). The
> > > > best would be to have a way to detect wether the card was actually in
> > > > a VGA mode without touching the VGA registers ... and based on that
> > > > decide wether to touch them or not... Is that possible at all ?
> > >
> > > Tested this patch on my albook (PowerBook5,2 Mobility Radeon 9600 M10).
> > > Dualhead works nicely now :)
> > >
> > > I've been testing the xvideo overlay. When running in clone mode, the overlay
> > > on the second head only works when the lcd is in it's native resolution. But
> > > unfortunately the panel has some ugly lines on it then, looks like a timing
> > > thing.  When not running in cloning mode the overlay on the second head only
> > > works in 24 bit dept, not in 16.
> >
> > What do you mean by "ulgy lines" when running in native resolution ? can
> > you be more explicit ? Is this related to overlay or happens in
> > general ? Have you tried option "LVDSProbePLL" "true" in your Device
> > section ? (My patch adds this option to help with the panels who need a
> > specific timing, though I may decide to make it default to true on LVDS
> > flat panels when there is no BIOS infos ...)
> 
> It's only when the panel and the second head (crt monitor) run on the panels
> native resolution. The lines are horizontal flickering lines which slowly move
> up. They seem to be there especially on the right of a window.
> 
> This is with LVDSProbePLL set to true. A closer look in my xorg log file
> indicates that the server notes the second head doesn't support the resolution
> and turns off MergeFB. But i still have an image on both heads.
> 

If mergedfb gets disabled, then crtc1 is probably driving both
outputs.  you can force mergedfb on by setting it to true and then
setting your crt2hsync and crt2vrefresh values in the config as per
the radeon man page.  Then crtc1 will drive one head and crtc2 will
drive the other.

Alex

> I've put my logfile at http://fire.luon.net/~sjoerd/Xorg.1.log
> 
>   Sjoerd
> --
> Ninety percent of everything is crap.
>                 -- Theodore Sturgeon
>