Bug 71226

Created attachment 88624 [details]
Broken colors on HDMI-0

Right after
xrandr --output HDMI-0 --auto

Created attachment 88625 [details]
Good colors on HDMI-0

After simple
xrandr --output HDMI-0 --off
xrandr --output HDMI-0 --auto
routine from the broken case.

Created attachment 88626 [details]
Broken colors on eDP

After leaving perfectly working display for ~15 minutes, coming back and leaving screensaver I got that broken colors on panel display.

Created attachment 88627 [details]
Good colors on eDP

Reboot fixed colors on my panel. Usually waiting for some times also does (I don't remember if screensaver/power has to be involved).

I'm having a very similar problem on my BARTS XT. 

It seems to only show up in apps that use OpenGL and only after a suspend/resume cycle. I'm not using a desktop compositor, so i never see it affect the whole screen like Rafał's screenshots.

Colors appear to invert or brighten randomly then go back to normal a couple seconds later. Similar to the attached jpegs.

Linux Brimstone 3.12.3-1-ARCH #1 SMP PREEMPT Wed Dec 4 21:45:42 CET 2013 x86_64 GNU/Linux

OpenGL version string: 3.0 Mesa 10.1.0-devel (git-92ceb32)
OpenGL shading language version string: 1.30

local/ati-dri-git 60031.92ceb32-1 (mesagit)

Also of note is that this does not happen with my REDWOOD (HD5670) card in the same machine with the same software versions and configuration.

(In reply to comment #5)
> It seems to only show up in apps that use OpenGL and only after a
> suspend/resume cycle.

If it only happens in OpenGL apps, that indicates a problem in the 3D rendering pipeline.


> I'm not using a desktop compositor, so i never see it affect the whole screen
> like Rafał's screenshots.

Actually, my assumption from Rafał's description was that he's only seeing the problem on one of several connected displays at a time, which would indicate a problem in the display pipeline instead (assuming all displays are showing the same contents or are all updating their contents while the problem occurs). Rafał, is my assumption correct?

(In reply to comment #7)
> Actually, my assumption from Rafał's description was that he's only seeing
> the problem on one of several connected displays at a time, which would
> indicate a problem in the display pipeline instead (assuming all displays
> are showing the same contents or are all updating their contents while the
> problem occurs). Rafał, is my assumption correct?

That's right. I got wrong colors on HDMI-0 while everything was fine on eDP at the same time.

(In reply to comment #8)
> 
> That's right. I got wrong colors on HDMI-0 while everything was fine on eDP
> at the same time.

That's usually indicative of the monitor not dealing with the HDMI packets from the GPU properly.  E.g., the monitor may think it's getting YUV data rather than RGB.  Disabling hdmi mode (driving the display like DVI) should avoid the problem if that's the issue (e.g., radeon.audio=0).

Does this help?

http://www.spinics.net/lists/stable/msg32892.html

It will be in the 3.14 kernel.