Bug 92575 - [4.2 regression] Massive graphics corruption on Iris Graphics 6100
Summary: [4.2 regression] Massive graphics corruption on Iris Graphics 6100
Status: CLOSED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: high major
Assignee: Elio
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-10-21 17:27 UTC by Kirys
Modified: 2016-08-29 15:12 UTC (History)
2 users (show)

See Also:
i915 platform: BDW
i915 features: display/Other


Attachments
Xorg log UXA buggy driver (93.53 KB, text/plain)
2015-10-22 18:51 UTC, Kirys
no flags Details
log that shows with kernel 4.2.3 (buggy one) (18.11 KB, text/plain)
2015-10-24 09:53 UTC, Kirys
no flags Details
log that shows with kernel 4.1.10 (good one) (16.49 KB, text/plain)
2015-10-24 09:53 UTC, Kirys
no flags Details

Description Kirys 2015-10-21 17:27:09 UTC
Hi all
I own an Intel NUC5i7RYH linked to 3 monitors (1 on HDMI and two on the display port using a splitter)
The cpu is an i7-5557U with Iris Graphics 6100.
Till 3 week ago (older fedora provided mesa driver) I was using the SNA driver without any issues till the update of the mesa and kernel.

Now the SNA is unusable, under SNA some monitors are not drawed at all for hours (the login progress bar stays also after the login is completed and neither the mouse shows) but sometimes (at random) it remember to update the screen and it shows that it should appear there...
And most of software (like firefox) shows massive graphics corruptions.
So I switched to UXA and as a result the performance dropped so much that a video played on youtube make the machine unusable and any kde animation is choppy!

Distribution Fedora 22
Hardware: NUC5i7RYH
Kernel: 4.2.3-200.fc22.x86_64
X.Org X Server 1.17.2
GLX version: 1.4
OpenGL core profile version string: 3.3 (Core Profile) Mesa 10.6.9 (git-8957b69)
OpenGL core profile shading language version string: 3.30
OpenGL version string: 3.0 Mesa 10.6.9 (git-8957b69)
OpenGL shading language version string: 1.30
OpenGL ES profile version string: OpenGL ES 3.0 Mesa 10.6.9 (git-8957b69)
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.00
libdrm 2.4.61
Comment 1 Kenneth Graunke 2015-10-21 21:56:06 UTC
Reassigning to the 2D driver..

As a temporary measure, you could try AccelMethod "blt" instead of UXA.
Comment 2 Chris Wilson 2015-10-22 09:09:38 UTC
Start with attaching a good Xorg.0.log (if you still have the driver from ~3 weeks ago) and today's broken Xorg.0.log. Then see which kernels you still have available, if you know which you had ~3 weeks ago, retest with that.
Comment 3 Kirys 2015-10-22 18:50:55 UTC
Due to a very busy week for I can only attach the current UXA log,
This week-end I'll try to make all the test (I'll check dnf log for previous version of the x.org driver)
Comment 4 Kirys 2015-10-22 18:51:54 UTC
Created attachment 119091 [details]
Xorg log UXA buggy driver

Xorg log UXA buggy driver
Comment 5 Kirys 2015-10-24 09:52:08 UTC
I've discovered that the issue is related to the kernel version not the driver itself (both updated at the same day according to the dnf log)
With kernel 4.2.3 I have all the issues I've reported while with kernel 4.1.10 everything seems to work fine (but I'll make some more tests).
I'll attach both logs.
Comment 6 Kirys 2015-10-24 09:53:23 UTC
Created attachment 119166 [details]
log that shows with kernel 4.2.3 (buggy one)
Comment 7 Kirys 2015-10-24 09:53:53 UTC
Created attachment 119167 [details]
log that shows with kernel 4.1.10 (good one)
Comment 8 Jani Nikula 2015-10-26 08:47:31 UTC
Kirys, can you bisect the issue?
Comment 9 Kirys 2015-10-26 10:43:07 UTC
What should I do?
Comment 10 Jani Nikula 2015-10-26 13:12:35 UTC
(In reply to Kirys from comment #9)
> What should I do?

The idea is to find the kernel commit between v4.1 and v4.2 that introduced the regression (unless Chris has some better ideas at this point).

https://fedoraproject.org/wiki/User:Ignatenkobrain/Kernel/Bisection
https://wiki.ubuntu.com/Kernel/KernelBisection
Comment 11 Kirys 2015-10-26 14:28:28 UTC
well that's a lot of work ^_^'

The nuc is my home pc so is not something I can do during my work days, I'll try during weekends but it seems this will take a lot of time.
Comment 12 Jani Nikula 2015-10-27 10:14:04 UTC
(In reply to Kirys from comment #11)
> well that's a lot of work ^_^'
> 
> The nuc is my home pc so is not something I can do during my work days, I'll
> try during weekends but it seems this will take a lot of time.

It means roughly 13 kernel build, install, test cycles. You should probably try 'make localmodconfig' instead of using distro kernel config to minimize the build time.
Comment 13 Jani Nikula 2015-10-27 10:16:12 UTC
Oh, and while Chris is known to have pulled rabbits out of his hat, bisecting may yet be the fastest way to find out what's going on.
Comment 14 Kirys 2015-10-27 11:25:49 UTC
understood, I'll try to do it.
Even thought I've build "my kernel" before (due to a previous issue with intel driver many years ago aka bug 43709) first time I'll do this kind of work.
I'll follow the articles, but I can work on this only when I'm home so it'll take some time.
Bye
K.
Comment 15 Kirys 2015-10-31 08:07:47 UTC
the bisect tutorial (https://fedoraproject.org/wiki/User:Ignatenkobrain/Kernel/Bisection) for fedora doesn't seems to work on fedora 22 :(
http://forums.fedoraforum.org/showthread.php?t=307260
I'll try a "manual approach"
Cya
F.
Comment 16 cprigent 2015-11-17 17:58:09 UTC
Bug scrub:
Reassigned to reporter to check
Comment 17 Kirys 2015-11-28 22:29:16 UTC
As i said i was unable to follow the tutorial, and this is a really busy period for me.
I'll try to do this in the first week-end of december.
Comment 18 Kimmo Nikkanen 2015-12-01 11:43:02 UTC
Lowering the priority. 
This is not blocking our work and we haven't been able to reproduce the problem on our side.
Comment 19 cprigent 2015-12-01 17:15:11 UTC
Elio,
Can you reproduce it and provide details?
Thanks
Comment 20 Kirys 2015-12-01 18:39:29 UTC
Today I've updated (by mistake ^_^) to 4.2.6-200 and the problem seems to be reduced.
The massive corruption is gone, but there still random lock of updates on some monitors (I've 3 monitors connected to the pc) that happen always at login and seldom after it. The lock disapears if I go to text console (ctrl+alt+f2-6) and came back to the graphical one.
Home to make more test soon.
Cya
K
Comment 21 Tatsuyuki Ishi 2016-03-22 01:47:49 UTC
Currently screen update lockup is worse than a month ago. This is not only Skylake, but on most Intel cards. Confirmed on 965GM.

A little detail: This is mostly triggered by Qt and KDE. In my case KDE+SDDM is frequently triggering this bug.

Also see https://bugs.freedesktop.org/show_bug.cgi?id=92505 for some hints.
Comment 22 Jani Nikula 2016-06-17 16:46:31 UTC
(In reply to Tatsuyuki Ishi from comment #21)
> Currently screen update lockup is worse than a month ago. This is not only
> Skylake, but on most Intel cards. Confirmed on 965GM.

Please file new bugs on unrelated issues.
Comment 23 Jani Nikula 2016-06-17 16:47:18 UTC
Timeout, closing. I'll presume we've fixed this. Please reopen if the problem persists with latest kernels.
Comment 24 Elio 2016-08-23 21:43:35 UTC
Base Kernel from clean enviroment 4.04-301.fc22
After normal update, going to 4.4.14. 
No issues so far with 3 monitors, one by MST through DP dongle.
Comment 25 cprigent 2016-08-29 15:12:47 UTC
So closed.


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.