Bug 98504 - graphical artefacts while resizing windows
Summary: graphical artefacts while resizing windows
Status: CLOSED DUPLICATE of bug 98402
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: XOrg git
Hardware: Other Linux (All)
: medium normal
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
Depends on:
Reported: 2016-10-30 15:43 UTC by Cristian Magherusan-Stanciu
Modified: 2017-01-10 17:26 UTC (History)
2 users (show)

See Also:
i915 platform: HSW
i915 features:

Video showing the issue in firefox (1.40 MB, video/x-matroska)
2016-10-30 15:43 UTC, Cristian Magherusan-Stanciu
no flags Details
gnome terminal under gnome shell wayland (302.01 KB, image/png)
2016-10-30 17:00 UTC, Alban Browaeys
no flags Details
journal dmesg output from a 4.8 run (74.05 KB, text/plain)
2016-11-03 16:45 UTC, Alban Browaeys
no flags Details

Description Cristian Magherusan-Stanciu 2016-10-30 15:43:35 UTC
Created attachment 127626 [details]
Video showing the issue in firefox

I noticed this on some applications after updating my Ubuntu 16.10 to the 4.9-rc2 kernel from the mainline PPA. I then updated to 4.9-rc3 and it behaves the same. This wasn't happening on 4.8 so I suspect it has to do with the kernel update.

It is always reproducing with Firefox and pinta(both based on GTK+) and Skype(based on Qt), but won't reproduce on other apps(like eog, also based on GTK+) so it is unlikely to be a toolkit bug, but I've also sometimes seen it in the lightdm lock screen, depending on the resolution.

On both Skype and Firefox it doesn't affect the topmost part of the windows, as you can see in the attached video.

I tried to measure the size of the blocks using gimp and they are roughly 480x160 pixels, and the blocks appear to be shuffled/interchanged always within the same row, so I guess this may be some kind of memory referencing issue.
Comment 1 Chris Wilson 2016-10-30 16:20:06 UTC
Looks like a regular userspace (stale buffer handle) bug. Please attach your Xorg.0.log so we have details about your hardware.
Comment 2 Alban Browaeys 2016-10-30 17:00:19 UTC
Created attachment 127628 [details]
gnome terminal under gnome shell wayland

I have this issue within gnome-shell wayland : 
 to repoduce there, best is start nautilus from gnome-terminal (and maximize it) and open a picture with eog from within the nautilus window

Haswell ULT (thinkpad Yoga S1)
00:02.0 VGA compatible controller: Intel Corporation Haswell-ULT Integrated Graphics Controller (rev 09)

i915 capabilities
gen: 7
pch: 3
is_mobile: no
is_i85x: no
is_i915g: no
is_i945gm: no
is_g33: no
hws_needs_physical: no
is_g4x: no
is_pineview: no
is_broadwater: no
is_crestline: no
is_ivybridge: no
is_valleyview: no
is_cherryview: no
is_haswell: yes
is_broadwell: no
is_skylake: no
is_broxton: no
is_kabylake: no
is_preliminary: no
has_fbc: yes
has_psr: yes
has_runtime_pm: yes
has_csr: no
has_resource_streamer: yes
has_rc6: yes
has_rc6p: no
has_dp_mst: yes
has_gmbus_irq: yes
has_hw_contexts: yes
has_logical_ring_contexts: no
has_l3_dpf: yes
has_gmch_display: no
has_guc: no
has_pipe_cxsr: no
has_hotplug: yes
cursor_needs_physical: no
has_overlay: no
overlay_needs_physical: no
supports_tv: no
has_llc: yes
has_snoop: no
has_ddi: yes
has_fpga_dbg: yes
has_pooled_eu: no

Linux kernel 4.9-rc2 and rc3

Gnome shell 3.22
 gtk3 3.22.1 and 3.22.2

I could test reverts , but cannot bisect so easely : I need to apply a few patches from lkml to cope with debian gcc defaults for PIC changes (affects ubuntu too) and a few other build niceties.
Comment 3 Alban Browaeys 2016-10-31 01:41:19 UTC
First bad commit in kernel :

commit a61007a83a4671da77210790997d5c8c92ed87ea
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Aug 18 17:17:02 2016 +0100

    drm/i915: Fix partial GGTT faulting

The previous one I noticed one issue (though not as big as this one : items are not aligned in "ls" output ... 
commit 03af84fe7f48937941fbb4dcd6cf66c68ae0edd4
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Aug 18 17:17:01 2016 +0100

    drm/i915: Choose partial chunksize based on tile row size

I mention as they could relate to each other. (mind this alignment issue is not ovbious and only happens by pressing "ls" : current folder and no options like "--colors"). This could be a previous commit issue (the bisect was not targeted at this item) but I can confirm this disappear with the above bad commit (when the output is not garbled as seen in screenshots, it does not exhibit this "smaller" bug).
Comment 4 yann 2016-11-03 10:18:46 UTC
Please attached kernel (ie dmesg) and xorg (ie Xorg.0.log) logs
Comment 5 Alban Browaeys 2016-11-03 16:45:48 UTC
Created attachment 127728 [details]
journal dmesg output from a 4.8 run

Sorry this is a stable kernel run dmesg . BUt if you are after data about the hw the output is the same for 4.8 and 4.9.

Well about Xorg .... wayland has no Xorg so I cannot add this output.

Mind I am not the initial reporter. Only I sorted the bug out : this issue is not Xorg related (I do not run Xorg) and the kernel commit revert fixes this bug  (but you have to revert on the kernel i915 code when the commit was added because it is not that easy with commits that follows)

Mind this is not the only major bug about intel gfx on 4.9 . I have no clue how you will manage to ship any close to stable a driver. The driver vma handling has been rewritten from top to bottom.

take the commit I mention as might be buggy too ... notice how this commit cannot do anything in the end ... the code section is enclosed by "not is_tiled" ... good ... then the commit add to this section a change to chunk_size that execute only when is_tiled is true !?

Then in a later commit the initial test "not is_tiled" is removed . So the commits were adding pile of dead code , then the initial guard was removed.
Comment 6 yann 2016-11-03 17:12:28 UTC
(In reply to Alban Browaeys from comment #5)
> Created attachment 127728 [details]
> journal dmesg output from a 4.8 run
> Sorry this is a stable kernel run dmesg . BUt if you are after data about
> the hw the output is the same for 4.8 and 4.9.
> Well about Xorg .... wayland has no Xorg so I cannot add this output.

Good catch, can you do then, for instance (from https://fedoraproject.org/wiki/How_to_debug_Wayland_problems#Information_to_include_in_your_bug_report), where command should be nautilus:

$ WAYLAND_DEBUG=1 command |& tee debug.out

and then attached debug.out
Comment 7 Chris Wilson 2016-11-07 11:46:51 UTC

*** This bug has been marked as a duplicate of bug 98402 ***
Comment 8 Jari Tahvanainen 2017-01-10 17:26:32 UTC
Closed as duplicate of closed+fixed.

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.