Bug 1129

Summary: DVI output with default modeline doesn't work with external panel
Product: xorg Reporter: Adam Sampson <ats-fd>
Component: Driver/RadeonAssignee: Xorg Project Team <xorg-team>
Status: RESOLVED DUPLICATE QA Contact:
Severity: major    
Priority: high CC: Andreas.Kleisun, chrisj, christer, colohan+freedesktop, luigiwalser, mharris
Version: 6.8.0   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Output of "lspci -v" as root
none
Minimal xorg.conf
none
Xorg.0.log after cold boot with RGB disconnected
none
Xorg.0.log after attaching RGB cable
none
Tool to set XV_SWITCHCRT to 1 none

Description Adam Sampson 2004-08-18 16:13:13 UTC 
I've just bought an Asus-branded Radeon 9200SE 128MiB RGB+DVI card;
manufacturer's code A9200SE/TD/P/128M/A, part number 90-C1VBG6-GUAY1Z, serial
number 47CM1I3257, which I'm using with a Dell 2001FP LCD monitor. I'm having
some trouble getting it to work with the Xorg X server (using the "radeon"
driver) on my Duron-based PC running Linux 2.6.8.1 with the radeon DRM module
(the one that came with Linux 2.6.8.1) built in.

I've tried both the Xorg 6.7.0 release and the CVS tag XORG-6_7_99_902; both
behave the same way. The logs below are from the CVS version.

The first problem is that DVI output doesn't work at all in X: while I can view
the text-mode console fine through the monitor's DVI input, once I start X the
screen goes blank and switches off (indicating no input). RGB output, however,
works beautifully in both text mode and X; the picture quality's very nice. I've
tried starting X with the RGB cable disconnected; I've tried rebooting the
machine with the RGB cable disconnected; I've tried various values for the
"MonitorLayout" option. Nothing seems to produce a picture on the DVI output.

The second problem is that XVideo overlays don't appear to be visible. (I'm
using MPlayer and VLC to test this.) The player appears to be playing correctly,
but rather than the video I just get a plain dark blue window. Since I can only
use the RGB output at the moment, I guessed that it might be sending the scaler
output to the wrong CRTC, so I tried the "OverlayOnCRTC2" option, but it didn't
make a difference.

(Using OpenGL output in MPlayer works, but it's too slow for comfortable
viewing; in general, OpenGL stuff works, but it seems much slower than my old
Matrox G400 was. I haven't experimented with this very much, but I thought it
might be worth mentioning...)

I'll attach the output of "lspci -v", my xorg.conf, and /var/log/Xorg.0.log from
two different test runs: the first being the first time I start Xorg after a
cold boot with just the DVI cable attached, and the second being when I start it
again after connecting the RGB cable too.

If there's anything you'd like me to try, or any more information you'd like me
to provide, please get in touch. Thanks!
Comment 1 Adam Sampson 2004-08-18 16:13:58 UTC 
Created attachment 671 [details]
Output of "lspci -v" as root
Comment 2 Adam Sampson 2004-08-18 16:14:28 UTC 
Created attachment 672 [details]
Minimal xorg.conf
Comment 3 Adam Sampson 2004-08-18 16:15:11 UTC 
Created attachment 673 [details]
Xorg.0.log after cold boot with RGB disconnected
Comment 4 Adam Sampson 2004-08-18 16:15:31 UTC 
Created attachment 674 [details]
Xorg.0.log after attaching RGB cable
Comment 5 Adam Sampson 2004-08-20 13:40:32 UTC 
Created attachment 690 [details]
Tool to set XV_SWITCHCRT to 1

After some experimentation, I was able to get XVideo output working on the RGB
output by writing this little program, which sets the XV_SWITCHCRT attribute to
1. XV_SWITCHCRT was left at 0 even with OverlayOnCRTC2 enabled in xorg.conf.
Perhaps it should default to 1 if that option is specified?
Comment 6 Alex Deucher 2004-08-31 18:49:05 UTC 
try the "monitorlayout" option to force the output types.  see the radeon man
page.  the overlayoncrt2 option has been removed and replaced with the Xv
attribute. the driver should detect which output is active and associate it with
crtc1 if you are using one monitor. the Xv attribute is really only for
switching the overlay when clone mode in effect.
Comment 7 Adam Sampson 2004-09-01 01:07:42 UTC 
I've tried the following with only the DVI output connected:

Option "MonitorLayout" "TMDS, NONE"
Option "MonitorLayout" "TMDS, CRT"
Option "MonitorLayout" "TMDS, TMDS"
  The monitor stays turned on, but the screen is completely black.

Option "MonitorLayout" "NONE, TMDS"
  The monitor turns off.

Option "MonitorLayout" "LVDS, NONE"
  X spins on startup, necessitating a reboot to make the machine usable again.

I guessed that OverlayOnCRTC2 had been retired, but it's still listed in the
radeon manpage, and it's clearly detecting the wrong output on my system.
Comment 8 Thomas J. Moore 2005-04-09 15:57:22 UTC 
Just to let you know that at least one other is having this problem (no DVI
output - can't verify invis. XVideo), with several generic 128M radeon 9200
(non-SE) cards, since Xorg 6.8 at least.  My initial workaround was to use
UseFBDev, but that disables some of the recent DRI enhancements, so I've
reverted to using analog D-SUB exclusively.   I tried comparing the kernel's
initialization with the X server's, but saw no significant differences.  One
thing I did notice, though, is that disabling the X analog D-SUB output by
either detaching the connector at X startup or using MonitorLayout, you still
get some sort of funky video display on the CRT output (like maybe what's
supposed to be displaying on the DVI output).  Of course MonitorLayout seems to
have its own little quirks, notably that DDC info seems to get screwed up when
using it.  Note that I use the same monitor (HP L2335) for both outputs.
Comment 9 Adam Sampson 2005-09-06 12:22:30 UTC 
Aha! Having discovered by accident that the DVI output did work for lower
resolutions, I found this page:

http://suif.stanford.edu/~csapuntz/rv280-linux-dvi.html

... which suggests that it's a problem with either the chipset or the monitor
not being able to handle the default 1600x1200 modeline, and gives an
alternative to use (which works fine for me in 6.8.2 without the given patch).
This solves all the problems I was having, so I'd be happy to see this resolved
as FIXED/NOTABUG.

I'm not sure if it'd be possible to detect an LCD being used with an rv280 and
use the alternate modelines automatically -- it sounds like that'd be hard to
make bulletproof -- but it might be worth documenting this in the radeon man page.

Thanks to all the people who've commented on this bug; hopefully the same fix
will work for others.
Comment 10 T. Hood 2005-09-23 09:53:24 UTC 
*** Bug 945 has been marked as a duplicate of this bug. ***
Comment 11 T. Hood 2005-09-23 10:13:35 UTC 
*** Bug 1724 has been marked as a duplicate of this bug. ***
Comment 12 T. Hood 2005-09-26 08:16:45 UTC 
*** Bug 4193 has been marked as a duplicate of this bug. ***
Comment 13 T. Hood 2005-09-26 08:17:26 UTC 
*** Bug 3627 has been marked as a duplicate of this bug. ***
Comment 14 T. Hood 2005-09-26 08:19:39 UTC 
Please open a separate bug report to track the xv problem, if it persists.
Comment 15 Adam Jackson 2005-12-25 17:48:33 UTC 

*** This bug has been marked as a duplicate of 5386 ***

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.