Bug 14487

Summary: no DVI on Gigabyte GA-MA69GM-S2H
Product: xorg Reporter: Adrian Bunk <bunk>
Component: Driver/radeonhdAssignee: Egbert Eich <eich>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium CC: brice.goglin, petrhlavka
Version: git   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Xorg.0.log
none
Xorg.0.log
none
Xorg.0.log with commit 47f8a6f784acc69da001b318ba3d80a9fa4c883d reverted
none
Logs from ASUS M2A-VM HDMI
none
Test patch.
none
Test patch (with typo fixed).
none
Xorg log with patched driver
none
Xorg.0.log with patch from comment 8 applied
none
Logs with manually specified modeline
none
Xorg.0.log - DDIA git
none
Xorg.0.log
none
Xorg.0.log - 1680x1050 none

Description Adrian Bunk 2008-02-13 14:17:19 UTC
Created attachment 14296 [details]
Xorg.0.log

Analog output works fine with the onboard 690G, but DVI output isn't working.

When starting X with the monitor connected through the DVI cable I'm _sometimes_ seeing something like a quarter of my X background picture fullscreen for half a second, but there's never any usable picture and the monitor tells me that it doesn't get any input.

Tested with both the 1.1.0 package from Debian unstable and current -git.
Comment 1 Petr Hlavka 2008-02-14 03:00:15 UTC
Hi, I have the same issue on ASUS M2A-VM HDMI. VGA output works without any problems. With DVI I get black screen (no input signal), according to log the X started successfully.

Tested with latest git, the DVI output never worked standalone for me. It stopped working completely after revision which enabled HDMI convetors (in December).
Comment 2 Egbert Eich 2008-02-15 09:56:44 UTC
Petr, the patch you are referring to was just a very simple. It marked HDMI connectors DVI. 
It should be able to undo the commit 47f8a6f784acc69da001b318ba3d80a9fa4c883d and test if this really made the difference. 
If it does could you please provide the log for the driver with and  without this patch. (generated with -logverbose 7).
Comment 3 Adrian Bunk 2008-02-17 03:07:32 UTC
Reverting commit 47f8a6f784acc69da001b318ba3d80a9fa4c883d made a difference:

X now immediately gives up stating "no screens found".
Comment 4 Adrian Bunk 2008-02-17 03:09:54 UTC
Created attachment 14363 [details]
Xorg.0.log

Xorg.0.log (commit a9af866ae712a0048d374dc640e482d1f4ce8859)
Comment 5 Adrian Bunk 2008-02-17 03:11:39 UTC
Created attachment 14364 [details]
Xorg.0.log with commit 47f8a6f784acc69da001b318ba3d80a9fa4c883d reverted

Xorg.0.log (commit a9af866ae712a0048d374dc640e482d1f4ce8859 with commit 47f8a6f784acc69da001b318ba3d80a9fa4c883d reverted)
Comment 6 Petr Hlavka 2008-02-17 13:26:49 UTC
Created attachment 14365 [details]
Logs from ASUS M2A-VM HDMI

To be clear, without commit 47f8a6f784, the DVI output works for me only when one monitor is connected using both outputs (VGA+DVI). With this commit, X starts when monitor is connected only through DVI, however the monitor wents off or stays black and sometimes the screen blinks for a half of second appox. every 10 seconds. Logs from various combination of monitor connection and presence of commit 47f8a6f784 are attached.
Comment 7 Egbert Eich 2008-02-19 04:28:26 UTC
Created attachment 14409 [details] [review]
Test patch.

This patch makes my ASUS M2A-VM HDMI work.
Adrian, could you please test this on your hardware?
Please note: this is for RS690 only. This patch is not expected to work anywhere else.
Comment 8 Egbert Eich 2008-02-19 08:52:31 UTC
Created attachment 14417 [details] [review]
Test patch (with typo fixed).

This test patch supersedes the previous one which had a typo.
Comment 9 Petr Hlavka 2008-02-21 23:53:38 UTC
Created attachment 14499 [details]
Xorg log with patched driver

With patch applied, monitor connected through DVI, the X won't start because no available modes for monitor are found. I will try to setup a custom modeline, but I hasn't been successful yet.
Comment 10 Adrian Bunk 2008-02-22 09:45:44 UTC
Created attachment 14505 [details]
Xorg.0.log with patch from comment 8 applied

X didn't start with the patch from comment 8.
Comment 11 Petr Hlavka 2008-02-24 03:03:10 UTC
Created attachment 14536 [details]
Logs with manually specified modeline

I've done some new tests with patch from comment 8 and ASUS M2A-VM HDMI.
- using only dvi connection, monitor autodetection stops working. So I've manually specified the modeline, X stared, however I'm still getting only black screen (or no input signal).
- using dvi+vga connection, no manual modeline, the behaviour seems to be same for me as with unpatched version.

Still, the only (may be weird) way how to get the dvi output, is reverting the commit 47f8a6f784 and using dvi+vga connection.
Comment 12 Egbert Eich 2008-02-27 06:00:59 UTC
Please forget about the attached patch.
The problem has to be solved differently. Appearantly DFP3 refers to a digital output that not even ATI knew it was used.
Support for this will be added shortly.

Connector table Gigabyte:
(II) RADEONHD(0): Connector[0] {RHD_CONNECTOR_VGA, "VGA CRT1", RHD_DDC_0, RHD_HPD_NONE, { RHD_OUTPUT_DACA, RHD_OUTPUT_NONE } }
(II) RADEONHD(0): Connector[1] {RHD_CONNECTOR_TV, "SVIDEO TV1", DDC_NONE, RHD_HPD_NONE, { RHD_OUTPUT_DACA, RHD_OUTPUT_NONE } }
(II) RADEONHD(0): Connector[2] {RHD_CONNECTOR_DVI, "HDMI Type A DFP3", RHD_DDC_1, RHD_HPD_NONE, { RHD_OUTPUT_LVTMA, RHD_OUTPUT_NONE } }

Connector table ASUS:
(II) RADEONHD(0): Connector[0] {RHD_CONNECTOR_VGA, "VGA CRT1", RHD_DDC_0, RHD_HPD_NONE, { RHD_OUTPUT_DACA, RHD_OUTPUT_NONE } }
(II) RADEONHD(0): Connector[1] {RHD_CONNECTOR_TV, "SVIDEO TV1", DDC_NONE, RHD_HPD_NONE, { RHD_OUTPUT_DACA, RHD_OUTPUT_NONE } }
(II) RADEONHD(0): Connector[2] {RHD_CONNECTOR_DVI, "HDMI Type A DFP2", RHD_DDC_2, RHD_HPD_NONE, { RHD_OUTPUT_LVTMA, RHD_OUTPUT_NONE } }
(II) RADEONHD(0): Connector[3] {RHD_CONNECTOR_DVI, "HDMI Type A DFP3", RHD_DDC_1, RHD_HPD_NONE, { RHD_OUTPUT_TMDSA, RHD_OUTPUT_NONE } }
Comment 13 Egbert Eich 2008-03-02 23:57:51 UTC
*** Bug 14234 has been marked as a duplicate of this bug. ***
Comment 14 Petr Hlavka 2008-03-26 15:20:08 UTC
Created attachment 15485 [details]
Xorg.0.log - DDIA git

I've tried the latest git with DDIA support on M2A-VM HDMI with LCD monitor connected by DVI output and still getting only black screen.
Comment 15 Egbert Eich 2008-03-26 16:11:42 UTC
OK, this is the LVTMA block that hasn't been changed with this fix. Here with my digital display it is working ok. I guess I need to look around for different displays to test.
Comment 16 Egbert Eich 2008-04-08 04:02:34 UTC
For quicker reference: the mode line used was:
RADEONHD(0): Modeline "1680x1050"x60.0  146.25  1680 1784 1960 2240  1050 1053 1059 1089 -hsync +vsync (65.3 kHz)

Adrian, Some additions have been made to better set the electrical parameters on the LVTMA block on RS690. 
Could you please retest? Thanks!

Assigning to reporter for feedback. Please assign back to me when done.
Comment 17 Adrian Bunk 2008-04-08 11:08:42 UTC
(In reply to comment #16)
> For quicker reference: the mode line used was:
> RADEONHD(0): Modeline "1680x1050"x60.0  146.25  1680 1784 1960 2240  1050 1053
> 1059 1089 -hsync +vsync (65.3 kHz)

FYI: mine should be 1920x1200

> Adrian, Some additions have been made to better set the electrical parameters
> on the LVTMA block on RS690. 
> Could you please retest? Thanks!

It's becoming better, what happens now is roughly:
1. show correct screen for 1 second
2. show black screen for 3 seconds
3. goto 1.

The 1/3 seconds are estimated and not stable numbers, and sometimes only at about one quarter of the screen is displayed.

> Assigning to reporter for feedback. Please assign back to me when done.

done

Comment 18 Adrian Bunk 2008-04-08 11:12:03 UTC
Created attachment 15766 [details]
Xorg.0.log
Comment 19 Alex Deucher 2008-04-08 11:23:47 UTC
this looks similar to bug 15175.  Does the modeline in that bug help?
Comment 20 Petr Hlavka 2008-04-09 01:49:29 UTC
Created attachment 15777 [details]
Xorg.0.log - 1680x1050

I have similar issues with latest git and 1680x1050 monitor connected on DVI.
Comment 21 Adrian Bunk 2008-04-17 09:51:26 UTC
(In reply to comment #19)
> this looks similar to bug 15175.  Does the modeline in that bug help?

Yes, it does.

Except that I _always_ have effects similar to the problem I described in #15537 for the radeon driver.

And the mouse cursor is corrupted.

And there are various screen corruptions.

But after I discovered that the radeon driver also supports my hardware (why did noone here bother to tell me that the radeon driver might be worth a try???), and that except for #15537 the radeon driver works quite well, I've put radeonhd into /dev/null .
Comment 22 Luc Verhaegen 2008-04-19 21:24:12 UTC
Adrian: the screen corruption issues could be due to some other things, like running the radeon drm with our driver. Please verify that screen corruption goes away when a) no longer loading the drm b) using shadowfb.

I have heard someone on irc state that, since he put a fan on the northbridge cooler, his flickering went away. So this could very well be a temperature issue.  If you have any fan laying about, please place it somewhere very close to the northbridge heatsink, and see whether this makes a difference.
Comment 23 Adrian Bunk 2008-04-20 00:21:54 UTC
(In reply to comment #22)
> Adrian: the screen corruption issues could be due to some other things, like
> running the radeon drm with our driver. Please verify that screen corruption
> goes away when a) no longer loading the drm b) using shadowfb.

radeon works and radeonhd does not - this invariant is not affected by whether drm is loaded or not.

> I have heard someone on irc state that, since he put a fan on the northbridge
> cooler, his flickering went away. So this could very well be a temperature
> issue.  If you have any fan laying about, please place it somewhere very close
> to the northbridge heatsink, and see whether this makes a difference.

I don't havve any fan lying around.

It happens directly after booting a cold computer.
And if any of the it8716 sensors is located there it's definitely no temperature problem.
Comment 24 Egbert Eich 2008-04-20 09:35:59 UTC
The cursor issue is definitely due to a conflict between the two drivers.
I have seen it here but have not been able to look into it.
Comment 25 Adrian Bunk 2008-04-20 09:46:34 UTC
(In reply to comment #24)
> The cursor issue is definitely due to a conflict between the two drivers.
> I have seen it here but have not been able to look into it.

Conflict between which two drivers?

When I'm testing radeonhd the package that contains radeon is not installed on my system and the other way round.

Comment 26 Petr Hlavka 2008-05-19 14:56:03 UTC
I've tried after a month the latest git (2ea7cdd7) and finally, the dvi output is working without any problems on my ASUS M2A-VM now.

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.