Bug 13373

Summary: Nouveau will not display DVI-A-1
Product: xorg Reporter: Adam Petaccia <adam>
Component: Driver/nouveauAssignee: Nouveau Project <nouveau>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium    
Version: unspecified   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Xorg-DVIA1 20071123
none
Xorg DVI-A-1 20071124
none
Xorg log 2007 11 26
none
Xorg 2007 11 26 (2nd attempt)
none
2007 11 27 Attempt
none
xrandr output
none
2007 12 18
none
2007 12 18 - VT Switch
none
Successful attempt 2007 12 24 none

Description Adam Petaccia 2007-11-23 18:34:37 UTC
stillunknown asked me if starting up nouveau on DVI-A-1 worked (from cold start).  It doesn't.  Curiously enough, DVI-D-1 works, as well as DVI-A-0 (DVI-D-0 works as well).  This could explain a lot concerning why dual-head wasn't working for me.
Comment 1 Adam Petaccia 2007-11-23 18:36:10 UTC
Created attachment 12707 [details]
Xorg-DVIA1 20071123

This is from a cold start.  Replacing DVI-D-1 with DVI-A-1 doesn't work.  Interestingly enough, the monitor does indicate that it's getting a signal, but nothing is displayed.  This is from the latest git.
Comment 2 Maarten Maathuis 2007-11-24 04:17:24 UTC
I took out a potentional problem, please try again.
Comment 3 Adam Petaccia 2007-11-24 05:51:08 UTC
Created attachment 12708 [details]
Xorg DVI-A-1 20071124

Xorg comes up now.  Running xrandr --auto when I plug the second monitor in gives me "Input not supported", but thankfully I can still VT switch and kill the X server.
Comment 4 Maarten Maathuis 2007-11-24 06:24:05 UTC
I have a vague idea what may be causing this, i'm working on it, but i need to make sure i introduce no holes into the system.
Comment 5 Maarten Maathuis 2007-11-24 06:48:48 UTC
Give it another spin.
Comment 6 Adam Petaccia 2007-11-26 05:10:48 UTC
Created attachment 12722 [details]
Xorg log 2007 11 26

DVI-A-1 comes up fine, but when I run xrandr --auto, my monitor just shows "Input Not Supported", and pulling the cable reveals that there is no signal to DVI-D-0.
Comment 7 Maarten Maathuis 2007-11-26 10:31:24 UTC
Please try again, i fixed something important (for your situation).
Comment 8 Adam Petaccia 2007-11-26 11:13:34 UTC
Created attachment 12727 [details]
Xorg 2007 11 26 (2nd attempt)

Exact same results as last time.
Comment 9 Maarten Maathuis 2007-11-26 11:44:54 UTC
I spotted an even bigger bug, will have to think about how i fix this.
Comment 10 Maarten Maathuis 2007-11-26 13:13:24 UTC
It should now set the vpll properly, which means your analog output should at least stay alive (i hope ;-)).
Comment 11 Adam Petaccia 2007-11-27 05:20:10 UTC
Created attachment 12736 [details]
2007 11 27 Attempt

Exact same result.  However, I did switch to a console and try to run xrandr.  I'll attach that to see if its helpful.
Comment 12 Adam Petaccia 2007-11-27 05:20:47 UTC
Created attachment 12737 [details]
xrandr output

I had to run this by switching to VT1 and running DISPLAY=:0 xrandr.  Don't know if this is helpful at all.
Comment 13 Maarten Maathuis 2007-11-27 08:35:30 UTC
Maybe i "fixed" that first thing incorrectly, it still seems like your getting the wrong clock. So it's a wrong sel_clk or the order of clocks is the other way around.
Comment 14 Maarten Maathuis 2007-11-30 08:00:47 UTC
There were a few significant bugs, please try again.

And yes, it did take this long :-(
Comment 15 Adam Petaccia 2007-12-02 06:27:57 UTC
Just recording that DVI-D-0 doesn't work now.
Here's a little chart of functionality:

• Cold Plugging
   ∘ ./ DVI-A-0+DVID1
   ∘ DVI-D-0
      ‣ No Signal
   ∘ ./ DVI-A-0
   ∘ ./ DVI-D-1
• Hot Plug
   ∘ DVI-D-1 hotplug DVI-A-0
      ‣ DVI-A-0 is corrupted
   ∘ DVI-A-0 hotplug DVI-D-0
      ‣ DVI-D-0: No Signal
   ∘ DVI-A-1 hotplug DVI-D-1
      ‣ DVI-D-1: No Signal
Comment 16 Maarten Maathuis 2007-12-02 10:25:11 UTC
This was with todays version?
Comment 17 Adam Petaccia 2007-12-02 13:39:55 UTC
~ 8 AM Eastern Time.
Comment 18 Maarten Maathuis 2007-12-05 09:27:45 UTC
There were some changes to tmds modesetting, so please try again.
Comment 19 Adam Petaccia 2007-12-06 06:13:29 UTC
Cold Plug
  ∘ ./ DVI-A-0+DVI-D-1 (RandR works, too)
  ∘ ./DVI-A-1
  ∘ X DVI-D-0
    ‣ Weird signal
  ∘ X DVI-D-1
    ‣ No Signal
Hot Plug
  ∘ DVI-D0 hotplug DVI-A-1
    ‣ weirder colors
  ∘ DVI-A-0 hotplug DVI-D-1
    ‣ No Signal
  ∘ DVI-A-1 hotplug DVI-D-0
    ‣ No Signal

Note that these results aren't consistant.  I've gotten DVI-D-0 to have a very odd looking output, or none.
Comment 20 Maarten Maathuis 2007-12-06 09:52:54 UTC
Please post xlogs of at least the problematic output too.
Comment 21 Maarten Maathuis 2007-12-06 12:32:46 UTC
I have discovered that a 7600gt only has one dual link transmitter, explaining why DVI-D-1 is less problematic.
Comment 22 Maarten Maathuis 2007-12-06 14:42:24 UTC
I was wrong about that, the 7600gt actually has two dual link outputs.
But a new xorg log will reveal what is actually happening ;-)
Comment 23 Maarten Maathuis 2007-12-08 09:42:03 UTC
There's been a slew of fixes, and i've had positive feedback about dual link dvi on a 7600gs. So give it another shot.
Comment 24 Maarten Maathuis 2007-12-17 10:49:03 UTC
Please try again.
Comment 25 Adam Petaccia 2007-12-18 08:32:10 UTC
Created attachment 13184 [details]
2007 12 18

Still wrong colors; this is a log before a VT switch.
Comment 26 Adam Petaccia 2007-12-18 08:32:54 UTC
Created attachment 13185 [details]
2007 12 18 - VT Switch

Log after the VT switch, with colors fixed.
Comment 27 Maarten Maathuis 2007-12-23 03:03:49 UTC
You have a reasonable chance now, we were not setting the default color palette on a modeset. So let me know if that works.
Comment 28 Adam Petaccia 2007-12-24 13:34:47 UTC
Created attachment 13346 [details]
Successful attempt 2007 12 24

It works now, with latest git.  No VT switch required.
Comment 29 Maarten Maathuis 2007-12-24 13:39:50 UTC
Does this cover the known bugs for you (=can i close this bug)?

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.