Bug 60709

Please attach your dmesg output.  Are you using an hdmi display with radeon.audio=1?  Can you bisect?

Yes i am using radeon.audio=1. Disabling this makes 3.10.5 work without green screen.
Attaching dmesg from 3.10.3 - thought it was .2, sorry.

I'll try to bisect - may take some time, i'll try asap.

Created attachment 107140 [details]
dmesg 3.10.3 with radeon.audio=1

Torsten: please provide output of
avivotool regs hdmi
using 3.9 and 3.10

Created attachment 107141 [details]
hdmi regs 3.9.11 with radeon.audio=1

Created attachment 107142 [details]
hdmi regs 3.10.5 with radeon.audio=0

Whoops, your avivotool is too old to support "hdmi" (or you mispelled it). Could you update avivotool, please?

After updating I'll need output of "avivotool regs hdmi" from 3.9 and 3.10, both using radeon.audio=1.

Created attachment 107145 [details]
hdmi regs 3.9.11 with radeon.audio=1

Created attachment 107146 [details]
hdmi regs 3.10.5 with radeon.audio=1

I am lost now a little bit.

To make sure rc1 did fail also i've installed all mainline ppa kernels from 3.9.11 to 3.10.5 including rc ones, to possible get a smaller bisect window.

Just bootet into 3.10rc1 and it worked.
Now i've bootet 3.10.3 and 3.10.5 - works - no more green stuff.

Did power off and power on the box, no difference - green is gone.

I did see this green one a few times until today everytime i boot 3.10.3 or 3.10.5 - now its gone *magic*.

Still those:

[   56.600787] HDMI: ELD buf size is 0, force 128
[   56.600803] HDMI: invalid ELD data byte 0

ones but HDMI sound is working and green is gone.

Any ideas?

I'll try to boot windows - maybe its there again after rebooting.

Created attachment 107147 [details]
dmesg 3.10.5 green

Created attachment 107148 [details]
hdmi regs 3.10.5 with radeon.audio=1 and green

Created attachment 107149 [details]
camera image

So that is REALLY the trick.
After booting to windows (vista 32bit, latest stable Ati Radeon driver), its green again when booting to 3.10.5.
I don't even know now what i have done that it did work a few minutes ago with the same kernel.

So now the question is - what should i do now? Still bisecting?

AddOn:

RC1 did fail after vista did run.
My first bisect kernel between rc1 and 3.9 did boot well (building next one).

After this i can boot to 3.10.5 without any green, so it seems the 3.9.x kernel does something that makes 3.10.5 happy.

I'll go on bisecting - but this will take time, because i have to boot to windows everytime to get the green thing back - and must boot 3.10.5 to make sure its really green before i can test a new bisect kernel.

I'll really appreciate any hints which make this not so cumbersome like it feels to me now.

Some interesting effect running my second bisect kernel which is bad (green).
Turned my monitor off while next kernel is building, after turning it on again, green is gone.

Done with bisect - did take some time but got it:

d3418eacad403033e95e49dc14afa37c2112c134 is the first bad commit
commit d3418eacad403033e95e49dc14afa37c2112c134
Author: Rafał Miłecki <zajec5@gmail.com>
Date:   Thu Apr 18 09:23:12 2013 -0400

    drm/radeon/evergreen: setup HDMI before enabling it
    
    Closed source driver fglrx seems to enable infoframes and audio packets
    at the end, which makes sense, do the same.
    
    Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
    Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

:040000 040000 67e10b90afeefe688d0c291e603c80dd7a9e810a 1105ef4336989eee87c971b035fedcc092c42417 M	drivers

Fix posted:
[FIX][PATCH] drm/radeon: fix WREG32_OR macro setting bits in a register
http://lists.freedesktop.org/archives/dri-devel/2013-August/043835.html

This is a duplicate of #60687 (I'm afraid I don't have privileges to close this)

*** This bug has been marked as a duplicate of bug 60687 ***