Bug 104347 - AMD RX 580: Hide/Show Chromium sometimes corrupts screen (see screenshot)
Summary: AMD RX 580: Hide/Show Chromium sometimes corrupts screen (see screenshot)
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/Gallium/radeonsi (show other bugs)
Version: 17.3
Hardware: x86-64 (AMD64) Linux (All)
: medium major
Assignee: Default DRI bug account
QA Contact: Default DRI bug account
: 106355 (view as bug list)
Depends on:
Reported: 2017-12-20 11:27 UTC by Arthur Borsboom
Modified: 2019-09-25 18:01 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:

Screenshot of scrambled window (720.89 KB, image/png)
2017-12-20 11:27 UTC, Arthur Borsboom
glxinfo (101.79 KB, text/plain)
2017-12-20 18:25 UTC, Arthur Borsboom
Chrome GPU detailed overview (65.23 KB, text/html)
2017-12-30 19:22 UTC, Arthur Borsboom

Description Arthur Borsboom 2017-12-20 11:27:45 UTC
Created attachment 136316 [details]
Screenshot of scrambled window

When I hide all Windows with a shortcut and then show another one, a window (not the whole desktop) looks a scrambled, see attachments. It restores itself usually after 3-5 seconds and/or hiding/showing of the window after a couple of times.

In dmesg there is no message of a GPU warning/error.

Kernel: Linux 4.14.5-1-ARCH
Driver: amdgpu
X11: xf86-video-amdgpu 1.4.0
Mesa: 17.3.0

If this is the wrong place to report this issue, can you please move it to the correct location?

If other information is needed, please let me know.
Comment 1 network723 2017-12-20 17:20:28 UTC
I had the same issue with my RX480, but it does not happen on hd7950 with amdgpu.
Comment 2 Michel Dänzer 2017-12-20 17:22:38 UTC
Please attach the corresponding glxinfo output.

Did this not happen with an older version of Mesa / LLVM? If so, can you bisect?
Comment 3 Arthur Borsboom 2017-12-20 18:24:45 UTC
> Did this not happen with an older version of Mesa / LLVM?

I do not know, since I have recently bought this AMD videocard to get rid of the out of kernel proprietary trouble with my previous Nvidia videocard. So I have no knowledge if this would happen with older versions of Mesa / LLVM.

> Please attach the corresponding glxinfo output.

See attachment.
If I can provide more information, please guide me. :)
Comment 4 Arthur Borsboom 2017-12-20 18:25:10 UTC
Created attachment 136330 [details]
Comment 5 network723 2017-12-20 19:40:26 UTC
(In reply to Michel Dänzer from comment #2)
> can you bisect?

I think it started happening around 2-3 months ago on my RX480, so it's quite a lot to bisect. I'd bisect myself, but I don't have polaris hardware anymore.
Comment 6 Michel Dänzer 2017-12-21 09:35:18 UTC
Arthur, can you try if it also happens with Mesa 17.2 (or maybe even older versions)?
Comment 7 Arthur Borsboom 2017-12-27 10:40:19 UTC
Hi Michel,

Momentarily I am unable to access my workstation for this purpose. When I will be able, I first have to find a way how to reproduce it on mesa 17.3, downgrade mesa to 17.2 and try to reproduce it there. Once I have the results I will report back to you.

This message is just to let you know, I have not forgotten you or your request. :)
Comment 8 Arthur Borsboom 2017-12-30 19:22:21 UTC
Created attachment 136453 [details]
Chrome GPU detailed overview
Comment 9 Arthur Borsboom 2017-12-30 19:22:49 UTC
Hi Michel,

Although I have not tested yet about mesa 17.2, I have noticed that this issue only appears in the full or small sized window of Chromium. So, until now it only has happened to Chromium; but this did not happen with my previous Nvidia card, so I guess there is a relationship with my (new) RX 580.

I also noticed that it only happens, once in a while (every 15-30 minutes) and when I active the Chromium window (bring it to the front). The garbled window auto recovers after a short time, I guess 5 seconds.

Does this new information help to narrow the search?
Do you still want me to try with mesa 17.2?
Or do you have another plan of attack?

Since I am on a rolling release these are currently my key components:
* Kernel: 4.14.8
* Kernel driver: amdgpu
* X11: xf86-video-amdgpu 1.4.0
* Mesa: mesa 17.3.1
* Chromium: 63.0.3239.108
* chrome://gpu/ gives: <see attachment: https://bugs.freedesktop.org/attachment.cgi?id=136453>

Maybe the chrome gpu page gives a clue?
Comment 10 Timothy Arceri 2018-05-02 23:31:09 UTC
*** Bug 106355 has been marked as a duplicate of this bug. ***
Comment 11 Norbert Klar 2018-05-04 12:13:24 UTC
Same problem here with RX 480 and newest MESA driver.
Comment 12 Ivan Chebykin 2018-05-07 11:19:47 UTC
Same problem occurs for me on RX 570 and linux 4.16 with mesa 18.0.1.
Comment 13 Norbert Klar 2018-05-07 18:11:36 UTC
Just tested with several RX 5xx and 4xx video card, and getting the same result.
Comment 14 Norbert Klar 2018-05-12 11:18:04 UTC
And looks like this bug not only happening with Chrome. Just installed Discord, and getting the same behaviour.
Comment 15 Roc Vallès Domènech 2018-05-13 07:15:29 UTC
I'm seeing this on a Vega64.

kernel 4.16.7-rt1
mesa 18.0.3
xf86-video-amdgpu 18.0.1
chromium 66.0.3359.170
Comment 16 Norbert Klar 2018-05-15 18:39:12 UTC
Found some articles from early 2017, and many RX4xx/5xx users report this problem. Any chance to get a fix?

Comment 17 Ivan Chebykin 2018-05-19 10:05:36 UTC
Hi, Norbert, your bug looks related to https://bugs.freedesktop.org/show_bug.cgi?id=105910.
Comment 18 Thomas Rohloff 2018-07-10 08:00:31 UTC
Same here.

OpenGL renderer string: Radeon RX 580 Series (POLARIS10, DRM 3.25.0, 4.17.3, LLVM 5.0.2)
OpenGL core profile version string: 4.5 (Core Profile) Mesa 18.1.3

There are also reports on i.e. reddit older than this bug report: https://www.reddit.com/r/archlinux/comments/70vqja/screen_has_recently_started_doing_this_on_chromium/
Comment 19 Michel Dänzer 2019-01-11 17:34:11 UTC
If this is still happening, https://gitlab.freedesktop.org/xorg/driver/xf86-video-amdgpu/merge_requests/21 might help.
Comment 20 GitLab Migration User 2019-09-25 18:01:36 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/1294.

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.