Bug 62564 - [HD 4000] triple monitor configuration - resolutions above 1024x768 hangs/locks kernel and/or xserver
[HD 4000] triple monitor configuration - resolutions above 1024x768 hangs/loc...
Status: RESOLVED FIXED
Product: DRI
Classification: Unclassified
Component: DRM/Intel
unspecified
x86 (IA32) Linux (All)
: medium major
Assigned To: Intel GFX Bugs mailing list
Intel GFX Bugs mailing list
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2013-03-20 16:55 UTC by nadrimajstor
Modified: 2013-03-26 17:58 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
xrandr --verbose (24.38 KB, text/plain)
2013-03-20 16:55 UTC, nadrimajstor
no flags Details
dmesg (215.87 KB, text/plain)
2013-03-20 17:23 UTC, nadrimajstor
no flags Details
Xorg.0.log (43.82 KB, text/plain)
2013-03-20 17:23 UTC, nadrimajstor
no flags Details
lines on the left side (272.58 KB, image/jpeg)
2013-03-26 17:58 UTC, nadrimajstor
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description nadrimajstor 2013-03-20 16:55:07 UTC
Created attachment 76831 [details]
xrandr --verbose

System environment:
-- chipset: QM77
-- system architecture: i686
-- xf86-video-intel: 2.20.0
-- xserver: 1.11.3
-- mesa: 9.0.1
-- libdrm version: 2.4.32-1ubuntu1
-- kernel version: 3.5.0-26-generic
-- Linux distribution: Ubuntu 12.04.2
-- Motherboard: kino-qm770-r10
-- Display connector: DVI, HDMI, HDMI

Setting native monitor resolutions above 1024x768 for all three monitors always produce this continuously repeating error:
...
[drm:intel_dp_start_link_train], training pattern 1 signal levels 00000000
[drm:intel_dp_start_link_train], training pattern 1 signal levels 02000000
[drm:intel_dp_start_link_train], training pattern 1 signal levels 04000000
...

Kernel 3.2.x hangs during boot.
Kernel 3.5.x - 3.9 will boot fine by automatically setting resolution to 1024x768
However, no matter which combination of connectors I use issuing commands such as:
xrandr --output VGA1 --auto --output DP2 --auto --output DP3 --auto
or
xrandr --output HDMI1 --auto --output DP2 --auto --output DP3 --auto
will always trigger this error.

Same monitor configuration works fine in Win7 so I assume that hardware is ok. :|
Comment 1 Daniel Vetter 2013-03-20 17:08:17 UTC
Infinite loops in DP link training should be fixed with

commit 3b4f819d5eac94ba8fe5e8c061f6dabfe8d7b22c
Author: Takashi Iwai <tiwai@suse.de>
Date:   Mon Mar 11 18:40:16 2013 +0100

    Revert "drm/i915: try to train DP even harder"

Which is only yet available in the drm-intel-fixes branch of

http://cgit.freedesktop.org/~danvet/drm-intel

but it's on the way to upstream and tagged for stable backporting. Can you please retest with that patch/revert?
Comment 2 nadrimajstor 2013-03-20 17:23:03 UTC
Created attachment 76833 [details]
dmesg
Comment 3 nadrimajstor 2013-03-20 17:23:41 UTC
Created attachment 76834 [details]
Xorg.0.log
Comment 4 nadrimajstor 2013-03-26 16:01:01 UTC
(In reply to comment #1)
> Infinite loops in DP link training should be fixed with
> 
> commit 3b4f819d5eac94ba8fe5e8c061f6dabfe8d7b22c
> Author: Takashi Iwai <tiwai@suse.de>
> Date:   Mon Mar 11 18:40:16 2013 +0100
> 
>     Revert "drm/i915: try to train DP even harder"
> 
> Which is only yet available in the drm-intel-fixes branch of
> 
> http://cgit.freedesktop.org/~danvet/drm-intel
> 
> but it's on the way to upstream and tagged for stable backporting. Can you
> please retest with that patch/revert?

Retested. I can confirm - this patch/revert solved this issue.
(tried on ubuntu 12.04.2 with kernel 3.5.0-26-generic #42~precise1)
Comment 5 Daniel Vetter 2013-03-26 16:21:44 UTC
Thanks for the feedback. Just to double-check: Everything works fine again, screens all work as expected?
Comment 6 nadrimajstor 2013-03-26 17:55:54 UTC
(In reply to comment #5)
> Thanks for the feedback. Just to double-check: Everything works fine again,
> screens all work as expected?

Actually there are some additional quirks.
I've got inconstant results on:
- changing VTs
- hot-plugging monitors
- waking up from power-saving
- changing resolution
Ranging from:
- "no-signal" on one of the outputs
- lines on left side of screen (see picture)
- some outputs sometime have correct signal sometime they not.

I'm thinking this must be related to the fact that chipset have only 2 PLLs (which I don't mind - I intend use two 1920x1080 HDMI TVs + monitor)
I'll have to do a bit of research, go trough existing bug reports and eventually file another bug if needed.
However, any help is much appropriated.
Comment 7 nadrimajstor 2013-03-26 17:58:22 UTC
Created attachment 77072 [details]
lines on the left side