Bug 61929 - [HSW eDP regression 3.8.2] eDP blank until xrandr is re-run
Summary: [HSW eDP regression 3.8.2] eDP blank until xrandr is re-run
Status: CLOSED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: Other All
: high major
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-03-07 01:51 UTC by shui yangwei
Modified: 2017-10-06 14:46 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
eDP black screen dmesg (121.38 KB, text/plain)
2013-03-07 01:51 UTC, shui yangwei
no flags Details

Description shui yangwei 2013-03-07 01:51:36 UTC
Created attachment 76066 [details]
eDP black screen dmesg

Environment:
----------------------------
Kernel: (linux-3.8.y)19b00d2dc9bedf0856e366cb7b9c7733ded659e4
Some additional commit info:
Author: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Date:   Mon Mar 4 06:04:08 2013 +0800

    Linux 3.8.2

lspci:
----------------------------
00:00.0 Host bridge: Intel Corporation Device 0c04 (rev 06)
00:02.0 VGA compatible controller: Intel Corporation Device 0416 (rev 06)

BIOS:
----------------------------
BR_Production_QM87_5MB_BIOS-112.1_ME-9.0.0.1310.v3_Undertest

Reproduce step:
----------------------------
1. Plugged DP and eDP
2. reboot
3. xinit&
4. Then you will find eDP black screen, but if I use xrandr clone mode, eDP will display OK.

Description:
----------------------------
I have used an older kernel, there's not this issue. I will give detail description about which branch, other platform test results when all of our testing finished. If you need some special messages, please comments.
Comment 1 shui yangwei 2013-03-12 06:36:32 UTC
I have retest it on all branches, because our nightly kernel can reproduce this bug, so I trying to figure out which branch caused this issue:

drm-intel-next-queued:  correct
drm-intel-fixes: correct

up-stream branches:
---------------------
drm-next: reproduced this bug

Kernel Info:
---------------------
Kernel: (drm-intel-nightly)b81e059ec5a7128622ab5d74d78e9b4f361b54ae
Some additional commit info:
Merge: 35f8bad 210561f
Author: Dave Airlie <airlied@redhat.com>
Date:   Wed Feb 20 11:40:49 2013 +1000


Description:
--------------------
This issue is caused by merging up stream branches, but internal branches are good, so I'm surprised that why nightly kernels also have this issue.
Comment 2 Gordon Jin 2013-03-12 16:44:38 UTC
(In reply to comment #1)
> This issue is caused by merging up stream branches, but internal branches
> are good, so I'm surprised that why nightly kernels also have this issue.

what do you mean by "internal branches"?
Comment 3 shui yangwei 2013-03-13 01:18:35 UTC
(In reply to comment #2)
> (In reply to comment #1)
> > This issue is caused by merging up stream branches, but internal branches
> > are good, so I'm surprised that why nightly kernels also have this issue.
> 
> what do you mean by "internal branches"?

internal branches: drm-intel-next-queued and drm-intel-fixes 
Both of them don't exist this bug. I retest 3.8 kernel, it doesn't have this issue.
Comment 4 Daniel Vetter 2013-03-18 17:52:12 UTC
I've moved branches around a bit around the merge window, so could very well be that just testing drm-intel-* branches doesn't test all relevant things.

Can you try to figure out a working baseline and then bisect from there to -nightly, please?
Comment 5 shui yangwei 2013-03-21 11:54:32 UTC
After bisect, I find the bad commit which I reported, is exactly the first bad.(this commit is on drm-next branch, and merged to nightly)

Kernel Info:
---------------------
Kernel: (drm-intel-nightly)b81e059ec5a7128622ab5d74d78e9b4f361b54ae
Some additional commit info:
Merge: 35f8bad 210561f
Author: Dave Airlie <airlied@redhat.com>
Date:   Wed Feb 20 11:40:49 2013 +1000
Comment 6 Daniel Vetter 2013-03-21 14:05:44 UTC
I'm confused. Can you please cite the exact git sha1? There are already a few different sha1s mentioned in this bug report by you, and to me it's not clear which one it is.
Comment 7 shui yangwei 2013-03-22 05:16:21 UTC
(In reply to comment #6)
> I'm confused. Can you please cite the exact git sha1? There are already a
> few different sha1s mentioned in this bug report by you, and to me it's not
> clear which one it is.

Description:
-----------------
I have tested the parents:
35f8badc1cf652381fa3f82c1fbea39f4dbe87fd(on drm-next)
and 
210561ffd72d00eccf12c0131b8024d5436bae95(on drm-intel-fixes)

Both of the parents git sha1 are good, but if I tested with the merged commit in drm-next which I listed below, this bug will come up. I guess it's the problem by merging.

--------------------------
commit b81e059ec5a7128622ab5d74d78e9b4f361b54ae
Merge: 35f8bad 210561f
Author: Dave Airlie <airlied@redhat.com>
Date:   Wed Feb 20 11:40:49 2013 +1000

    Merge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

    So here's my promised pile of fixes for 3.9. I've dropped the core prep
    patches for vt-switchless suspend/resume as discussed on irc. Highlights:
    - Fix dmar on g4x. Not really gfx related, but I'm fed up with getting
      blamed for dmar crapouts.
    - Disable wc ptes updates on ilk when dmar is enabled (Chris). So again,
      dmar, but this time gfx related :(
    - Reduced range support for hsw, using the pipe CSC (Ville).
    - Fixup pll limits for gen3/4 (Patrick Jakobsson). The sdvo patch is
      already confirmed to fix 2 bug reports, so added cc: stable on that one.
    - Regression fix for 8bit fb console (Ville).
    - Preserve lane reversal bits on DDI/FDI ports (Damien).
    - Page flip vs. gpu hang fixes (Ville). Unfortuntely not quite all of
      them, need to decide what to do with the currently still in-flight ones.
    - Panel fitter regression fix from Mika Kuoppala (was accidentally left on
      on some pipes with the new modset code since 3.7). This also improves
      the modeset sequence and might help a few other unrelated issues with
      lvds.
    - Write backlight regs even harder ... another installement in our eternal
      fight against the BIOS and backlights.
    - Fixup lid notifier vs. suspend/resume races (Zhang Rui). Prep work for
      new ACPI stuff, but closing the race itself seems worthwile on its own.
    - A few other small fixes and tiny cleanups all over.
Comment 8 Daniel Vetter 2013-04-08 10:19:36 UTC
Is this still an issue on latest dinq?
Comment 9 shui yangwei 2013-04-09 03:29:33 UTC
(In reply to comment #8)
> Is this still an issue on latest dinq?

I have retested latest dinq, and there's not this issue, I listed the information of the kernel which I tested below:
-----------------------------------------
Kernel: (drm-intel-next-queued)99fb11e2d60385a2e10f063afd4cbb06032ef92a
Some additional commit info:
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Apr 4 22:20:34 2013 +0200

    drm/i915: set CB tuning also for the reduce clock
Comment 10 Daniel Vetter 2013-04-09 08:37:54 UTC
Thanks for retesting, I'm closing this as resolved now.
Comment 11 Elizabeth 2017-10-06 14:46:53 UTC
Closing old verified.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.