my system is running Ubuntu and since around version 18.04 it is plagued with graphic corruption problems.
I use Plasma desktop with or without compositing and after a few minutes or hours, I start to see graphic corruptions.
I upgraded my system to Ubuntu 18.10 in the hope that newer Xorg and Mesa would help, but the problem remains.
And today I got a GPU hang:
[ 1746.029214] [drm] GPU HANG: ecode 6:0:0xff800fbf, in kwin_x11 , reason: hang on rcs0, action: reset
[ 1746.029216] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[ 1746.029216] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[ 1746.029217] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[ 1746.029217] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[ 1746.029218] [drm] GPU crash dump saved to /sys/class/drm/card0/error
[ 1746.029252] i915 0000:00:02.0: Resetting chip for hang on rcs0
so here is my report with the required info:
- uname -a
Linux vougeot 4.18.11-041811-generic #201809290731 SMP Sat Sep 29 11:33:39 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
This is the mainline kernel provided by Ubuntu: http://kernel.ubuntu.com/~kernel-ppa/mainline/
- Linux distribution: Ubuntu 18.10
- Machine or mother board model:
Manufacturer: Dell Inc.
Product Name: Latitude E6520
Vendor: Dell Inc.
Release Date: 11/14/2013
Base Board Information
Manufacturer: Dell Inc.
Product Name: 0NVF5K
model name : Intel(R) Core(TM) i7-2640M CPU @ 2.80GHz
- Display connector: internal LVDS panel
- dmesg and other attachment to follow below...
Created attachment 141812 [details]
Created attachment 141813 [details]
Content from /sys/class/drm/card0/error
Please note that during boot I see the following error message:
[ 2.455186] [drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun
[ 2.455228] [drm:intel_set_pch_fifo_underrun_reporting [i915]] *ERROR* uncleared pch fifo underrun on pch transcoder A
[ 2.455252] [drm:intel_pch_fifo_underrun_irq_handler [i915]] *ERROR* PCH transcoder A FIFO underrun
I never bothered about it previously because it does not prevent the system from booting and from displaying graphics.
Do you think it is related to my main graphic problem?
Created attachment 141814 [details]
Sample of corruption
Created attachment 141823 [details]
Other example of corruption
Could you please describe actions after which you observed those hangs?
I've tried to reproduce this issue but couldn't catch anything like that for several hours using following environment:
SandyBridge; Intel® HD Graphics 3000; Intel Core i5-2520M;
Ubuntu 18.10 kernel 4.18.11-041811-generic; Plasma desktop;
Is this issue still actual for you?
If so, could you please provide some description of actions which lead to this hang since I still faced no issues while use mentioned environment for several days.
>Could you please describe actions after which you observed those hangs?
I am using plasma with many (12) virtual desktops. As constantly running applications, I use thunderbird, firefox, akregator and many konsoles.
The crucial point to reproduce this bug is to run plasma with compositing activated.
I will attach a screenshot of my settings.
>I've tried to reproduce this issue but couldn't catch anything like that for several hours
Thanks for trying so hard!
Created attachment 142006 [details]
> Is this issue still actual for you?
Yes, even with recent updates:
Linux kernel 4.18.13-041813-generic
I even recorded a screenshot with more massive corruption.
Created attachment 142007 [details]
More massive corruption
The problem seems to come from glamor acceleration, since the problem seems to disappear when I disable graphic acceleration with the following configuration:
Option "AccelMethod" "none" # (instead of "glamor")
For the record, I am still seeing graphic corruption in Ubuntu 18.10 with those software versions:
Linux kernel 5.0.7
I am going to upgrade my system to Ubuntu 19.04...
I also tested Fedora 29 on the same laptop and I also see graphic corruption:
- in GNOME/Wayland
- in Plasma/Xorg
hi, I also tried to load about 12 desktops with some apps (web gl in browser, a lot of cmd's, etc) - didn't see your's corruptions (few gpu hangs in web content, without any stable steps, that's all).
Config for Xorg is the same with your's - glamor enabled, driver is modesetting.
I will try to work on this machine longer and update kernel.
Current setup is:
Thanks a lot for trying to reproduce this problem!
Sometimes it appears after a few minutes, sometimes it appears after a few hours. One way to make it appear quicker is to enable compositing in Plasma.
Since I seem to be the only one to have this problem, do you think it may be caused by a defective CPU/GPU? (it is very old by now)
I upgraded my Ubuntu system and I still see graphic corruption. Here is version info:
Linux kernel 5.0.7
I started to see transient corruption after a few hours of work. After ~20 hours, I get corruption in text for all newly started GTK3 applications (see screenshot). Qt5 applications still look fine or have occasional transient corruption.
Created attachment 143937 [details]
Screenshot for Ubuntu 19.04 that shows text corruption in GTK3 applications
Could you, please, try to use Mesa with patch https://gitlab.freedesktop.org/mesa/mesa/merge_requests/520?
If it didn't help - please, provide an apitrace (https://apitrace.github.io/) of corrupted app
I upgraded my system with disco-proposed that contains a more recent Mesa version:
I am still seeing graphic corruption.
Created attachment 143940 [details]
Graphic corruption in Ubuntu 19.04
Here is an example of corruption that I got only a few seconds after logging in. It occurred both with Mesa 19.0.1 and 19.0.2.
OpenGL triangles are clearly visible.
Sometimes there is no corruption but when a rectangle is displayed it is decomposed into several triangles and there is enough delay between triangles to notice that the rectangle is not displayed atomically.
My SandyBridge laptop died and my new laptop with a more recent CPU/GPU does not suffer from the same problem. Therefore, I will not be able to do further tests concerning this bug.
Should I close this bug?