eDP laptop screen does not come up until a suspend/resume cycle happens.

[Robert Hooker]

Starting with the precise 3.2.0-32 kernel update, this laptop's eDP screen refuses to come up until the lid is closed and opened again triggering a suspend/resume cycle. I bisected this down to

http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-precise.git;a=commit;h=d3f281cf0493885d92ffc8e2c9e7e4e6d838d6a5

which came in through the upstream 3.2.29 update that also exibits the problem. 3.2.32 is also affected, and reverting it fixes the issue. In the working case, it's using 6bpc (PIPEACONF: 0xc0000050 (enabled, active, 6bpc)) and in the broken case it's using 8bpc (PIPEACONF: 0xc0000000 (enabled, active, 8bpc)). I can't attach full logs because it is an unreleased machine, but I've attached a few sanitized ones.

[Robert Hooker (sarvatt)]

Kernel with the revert commit that was used for testing.

http://kernel.ubuntu.com/~sarvatt/lp1070978/

[Robert Hooker (sarvatt)]

This bug affects quantal as well, 3.5.0-18 has the same symptom, and reverting the above commit from it (which was introduced in the 3.5.3 stable update) also fixes this specific machine.

[Robert Hooker (sarvatt)]

This is fixed in the 3.7 series after the modesetting rework.

[Robert Hooker (sarvatt)]

Created attachment 69076
dmesg after reproducing it with drm.debug=0xe

This commit causes a lenovo v490u to have a black screen after booting

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=4344b813f105a19f793f1fd93ad775b784648b95

It's working properly in 3.7, but 3.6 is broken without reverting that, and that commit was backported to the stable kernels which are now broken also. Suspending/resuming the machine after booting brings the display back up.

(working with revert)
[ 12.179335] [drm:intel_choose_pipe_bpp_dither], clamping display bpc (was -1) to eDP (6)
[ 12.179337] [drm:intel_choose_pipe_bpp_dither], setting pipe bpc to 8 (max display bpc 6)

(failing)
[ 14.264823] [drm:intel_choose_pipe_bpp_dither], setting pipe bpc to 8 (max display bpc 8)

[Robert Hooker (sarvatt)]

Created attachment 69077
intel_reg_dumper output (working)

[Robert Hooker (sarvatt)]

Created attachment 69078
intel_reg_dumper output (failing)

[Daniel-ffwll (daniel-ffwll)]

Please double-check whether your 3.6 tree contains:

commit 0c96c65b48fba3ffe9822a554cbc0cd610765cd5
Author: Jani Nikula <email address hidden>
Date: Wed Sep 26 18:43:10 2012 +0300

    drm/i915: use adjusted_mode instead of mode for checking the 6bpc force flag

This should fix fallout from the cited commit.

[Robert Hooker (sarvatt)]

No luck unfortunately, tried 3.2.32 and 3.6.3 that both contain that commit and they have the same problem.

[Robert Hooker (sarvatt)]

Created attachment 69090
3.6.3 dmesg

[Jani-nikula (jani-nikula)]

Related to https://bugzilla.kernel.org/show_bug.cgi?id=47641

Robert, please add the hex dump of your EDID from /sys/class/drm/*/edid, where * is your display.

[Robert Hooker (sarvatt)]

Created attachment 69821
edid

00ffffffffffff000dae771400000000
02160103801f11780a2b359757548f29
23505400000001010101010101010101
010101010101da1d56e250002030442d
470035ae10000018e01856e250002030
442d470035ae10000018000000fe0043
4d4e0a202020202020202020000000fe
004e3134304247452d4c41320a200051

[Cybjit (cybjit)]

Seeing similar this on Dell laptops in bug #1077064

[Robert Hooker (sarvatt)]

Another eDP machine (dell e6510) broken by this commit on

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1077064

[Daniel-ffwll (daniel-ffwll)]

Revert merged to -fixes:

commit 2f4f649a69a9eb51f6e98130e19dd90a260a4145
Author: Jani Nikula <email address hidden>
Date: Mon Nov 12 14:33:44 2012 +0200

    drm/i915: do not ignore eDP bpc settings from vbt

Please yell if this is still an issue for you.

[Robert Hooker (sarvatt)]

The upstream bug report hasn't imported the new comments yet, but they found this was related to https://bugzilla.kernel.org/show_bug.cgi?id=47641 also so it broke dithering on a HP EliteBook 2540p. Intel has reverted this commit with http://cgit.freedesktop.org/~danvet/drm-intel/commit/?id=2f4f649a69a9eb51f6e98130e19dd90a260a4145 (which is not yet in Linus' tree yet) because it broke more than it fixed and they need another workaround for the one machine it did fix.