Bug 42174

Created attachment 52708 [details]
intel_reg_dumper_19VGA1_15VGA2.not_working:  15" L150P plugged into VGA2 does not work

Created attachment 52712 [details]
intel_reg_dumper_15VGA1_19VGA2: 15" L150P display works when plugged into VGA1

Created attachment 52714 [details]
dmesg_15VGA1_19VGA2:  good

Created attachment 52715 [details]
dmesg_19VGA1_15VGA2.not_working: bad

Does the 15" display work independently on VGA-2 if nothing is plugged into VGA-1?

We have very little control over what the ADD2 card actually does, we request an output mode, it chooses what to signal to the display. As far as I can see, the register settings are consistently swapped over as the displays are swapped, so internally things look correct.

With no display plugged into VGA1, the 15" display did not work when plugged into VGA2.

http://www.smolts.org/client/show/pub_ccbe78ba-ae75-4723-b25a-6184f43c1794

When you're at the "non-working" state, what happens if you try to change the video modes (using xrandr) on the not-working monitor? Does any of the modes work?

yes.. looks like I can get some of the modes to work using xrandr.

i.e.  xrandr --output VGA2 --mode 800x600 --rate 75

The table below shows the modes I tried using the 15" display attached to VGA2.  BAD indicates that the screen was blank, blinking green light and a "Input Signal Out of Range" message.  GOOD indicates that there was a picture and solid green light.

   1024x768       60.0_BAD     75.1_GOOD     70.1_BAD
   832x624        74.6_BAD
   800x600        72.2_GOOD    75.0_GOOD     60.3        56.2
   640x480        72.8         75.0          60.0        59.9
   720x400        70.1

I tried the following modes with the 19" plugged into VGA1 and the 15" plugged into VGA2.

   1024x768       60.0_BAD     75.1_GOOD

(In reply to comment #9)
> yes.. looks like I can get some of the modes to work using xrandr.
> 
> i.e.  xrandr --output VGA2 --mode 800x600 --rate 75
> 
> The table below shows the modes I tried using the 15" display attached to VGA2.
>  BAD indicates that the screen was blank, blinking green light and a "Input
> Signal Out of Range" message.  GOOD indicates that there was a picture and
> solid green light.

The "picture and solid green light" is what you were expecting to see?

Please do this:
- go back to the non-working configuration
- obtain the output of "xrandr --verbose" and paste it here.

Created attachment 52760 [details]
xrandr_non-working: 19"VGA1 1280x1024 60.0     15"VGA2 1024x768 60.0>

>The "picture and solid green light" is what you were expecting to see?
Yes

>Please do this:
>- go back to the non-working configuration
>- obtain the output of "xrandr --verbose" and paste it here.
The xrandr_non-working attachment is with the --verbose flag
# xrandr
Screen 0: minimum 320 x 200, current 2304 x 1024, maximum 8192 x 8192
VGA1 connected 1280x1024+0+0 (normal left inverted right x axis y axis) 376mm x 301mm
   1280x1024      60.0*+   75.0     72.0  
   1152x864       75.0  
   1024x768       75.1     70.1     60.0  
   832x624        74.6  
   800x600        72.2     75.0     60.3     56.2  
   640x480        72.8     75.0     66.7     60.0  
   720x400        70.1  
TV1 disconnected (normal left inverted right x axis y axis)
TV2 disconnected (normal left inverted right x axis y axis)
VGA2 connected 1024x768+1280+0 (normal left inverted right x axis y axis) 304mm x 228mm
   1024x768       60.0*+   75.1     70.1  
   832x624        74.6  
   800x600        72.2     75.0     60.3     56.2  
   640x480        72.8     75.0     60.0     59.9  
   720x400        70.1

From looking at the logs and from the "good" and "bad" definitions you see, the first pattern I can observe is that the "-HSync -VSync" modes don't work...

It would be interesting to test 720x400 since it is the only "-HSync +VSync" mode.

I'll try to find out why this is happening.

15" Display: Mode 720x400 "-HSync +VSync" on VGA2 does not work

   1024x768       60.0_BAD     75.1_GOOD     70.1_BAD
   832x624        74.6_BAD
   800x600        72.2_GOOD    75.0_GOOD     60.3        56.2
   640x480        72.8         75.0          60.0        59.9
   720x400        70.1_BAD

While you're at a non-working mode, what happens if you set the 18th bit of the 0xe1140 register to 1?

If you use "./intel_reg_read 0xe1140", you'll probably get a value of "0xa0000004". In this case, you'll have to "./intel_reg_write 0xe1140 0xa0040004".

Used 1024x768@60Hz on VGA2
No change after setting the register.  Blank Screen, blinking green light and 
"Input Signal Out of Range Message"

# ./intel_reg_read 0xe1140
0xE1140 : 0xA0000004

# ./intel_reg_write 0xe1140 0xa0040004
Value before: 0xA0000004
Value after: 0xA0040004

I have some good news. I tried setting the Sync_Polarity to High 

I can get the 2nd display to work by setting 0xe1140 to 0xA000001C

HDMIB register and setting it to 

# ./intel_reg_write 0xe1140 0xa000001C
Value before: 0xA0000004
Value after: 0xA000001C

Ok, two more questions:

- If you connect another monitor on the sdvo output and ask for a mode that uses -VSync/-HSync, does it work?
- If you connect the 15" monitor on the non-sdvo output, do the -VSync/-HSync modes work?

If the problems only happen in the sDVO output, I'll have to ask:
- Connect a monitor on a non-sdvo, save the "xrandr --verbose" output.
- Connect the same monitor on the sdvo, save the "xrandr --verbose" output.

What I'm trying to check:
- Is the problem on the monitor (and its edid)?
- Is the problem on the way we read EDID from the monitors on the sdvo card?

>- If you connect another monitor on the sdvo output and ask for a mode that
>uses -VSync/-HSync, does it work?

No. I connected the 19" display to VGA2 and "1024x768 (0x47) 75.0MHz -HSync -VSync" did not work.  I was able to get it to work by updating 0xE1140 to 0xA000001C. 

1024x768 (0x47)   75.0MHz -HSync -VSync *current
        h: width  1024 start 1048 end 1184 total 1328 skew    0 clock   56.5KHz
        v: height  768 start  771 end  777 total  806           clock   70.1Hz

[root@sbtest1 intel]# ./intel_reg_read 0xe1140
0xE1140 : 0xA0000004
[root@sbtest1 intel]# ./intel_reg_write 0xe1140 0xa000001C
Value before: 0xA0000004
Value after: 0xA000001C   <--- This made it work

>- If you connect the 15" monitor on the non-sdvo output, do the -VSync/-HSync
>modes work?

Yes.  The 15" display using -VSync/-HSync works on VGA1

>If the problems only happen in the sDVO output, I'll have to ask:
>- Connect a monitor on a non-sdvo, save the "xrandr --verbose" output.
>- Connect the same monitor on the sdvo, save the "xrandr --verbose" output.

See Attachments: 
xrandr_vga1-15_vga2-19
xrandr_vga1-19_vga2-15

Created attachment 52969 [details]
xrandr_vga1-15_vga2-19

Created attachment 52970 [details]
xrandr_vga1-19_vga2-15

Paulo,
As an experiment, I set the SDVO HDMIB to always use +vsync and +hsync and was able to get all the modes on both my displays to work when attached to VGA2.  When booting to the command line,  the 19" display on VGA1 blinked twice before the 15" display on VGA2 synced up which took about 12 seconds.  Once it was up, all the modes including the -vsync and/or -hsync modes worked.

These are the changes I made for the experiment.

--- linux-3.0.x86_64.orig/drivers/gpu/drm/i915/intel_sdvo.c     2011-10-28 11:58:37.508722513 -0500
+++ linux-3.0.x86_64/drivers/gpu/drm/i915/intel_sdvo.c          2011-11-01 21:44:34.852123713 -0500
@@ -1072,9 +1072,13 @@ static void intel_sdvo_mode_set(struct d
                        sdvox |= intel_sdvo->color_range;
                if (INTEL_INFO(dev)->gen < 5)
                        sdvox |= SDVO_BORDER_ENABLE;
-               if (adjusted_mode->flags & DRM_MODE_FLAG_PVSYNC)
+
+                /* HDMIB is not working with -vsync -hsync
+                 * Always set SDVO_VSYNC_ACTIVE_HIGH and SDVO_HSYNC_ACTIVE_HIGH
+                 */
+               if (1 || adjusted_mode->flags & DRM_MODE_FLAG_PVSYNC)
                        sdvox |= SDVO_VSYNC_ACTIVE_HIGH;
-               if (adjusted_mode->flags & DRM_MODE_FLAG_PHSYNC)
+               if (1 || adjusted_mode->flags & DRM_MODE_FLAG_PHSYNC)
                        sdvox |= SDVO_HSYNC_ACTIVE_HIGH;
        } else {
                sdvox = I915_READ(intel_sdvo->sdvo_reg);

Hello again

I need you to provide some more information:

First of all, please revert the patch you mentioned in comment #21. It is a temporary fix for the problem, you can use it in your production systems while we don't write a "proper" fix for the bug, but we don't want it interfering with our tests.

Then please get to the "bad state" (where the monitor on sdvo is out of sync). Then you go to the console (without Xorg, press ctrl+alt+f1) and login as root. Go to where you compiled intel-gpu-tools, and go to the "test/" directory. There, please run "./testdisplay -i > testdisplay-output.txt" and attach the "testdisplay-output.txt" file to this bug.

In addition, I'll attach a kernel patch to this bug report. Please apply this patch, reboot and test if the patch fixes the problem. If not, please boot with drm.debug=0x06 and send the "dmesg" output to me.

Created attachment 53752 [details] [review]
proposed patch

Patch mentioned by the comment #22

Thanks Paulo!
We are working on getting the information you requested.

The testdisplay-output.txt is the result of running "./testdisplay -i > testdisplay-output.txt"

Created attachment 54122 [details]
testdisplay-output.txt         VGA1 - Good        VGA2 - No Sync

Created attachment 54127 [details]
dmesg.att_53752_06

The 53752: proposed patch did not fix the problem.
I applied the patch and VGA2 did not have sync after reboot.  It doesn't look like intel_sdvo_get_mode_from_dtd is getting called.

*** Bug 43333 has been marked as a duplicate of this bug. ***

Based on everything that was studied/discussed (not only in this bug report), this is supposed to be the "final" patch proposal:

http://lists.freedesktop.org/archives/intel-gfx/2011-December/014052.html

Please test it and report the results. It is supposed to be equivalent to the other patches already tested by you, so we shouldn't expect any problems.

Thank you,
Paulo

The patch allows the second display to work. As noted in comment 21, it takes a few seconds after the the 1st display comes up before the second display syncs.

Closing bug. Patch is in Linus's tree and also queued for Stable trees.

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=ba68e086223a5f149f37bf8692c8cdbf1b0ba3ef

If the patch does not fix the bug, feel free to reopen.