Graphics artifacts with i3-wm on amdgpu

[diegoviola]

I'm submitting a…

[x] Bug
[ ] Feature Request
[ ] Documentation Request
[ ] Other (Please describe in detail)

Current Behavior

I am currently experiencing graphics artifacts/corruption when opening windows on i3-wm. I've asked for help on #radeon thinking this was a driver issue but MrCooper suggested this to be a bug in i3.

https://bugs.freedesktop.org/show_bug.cgi?id=105910#c10

Here's some of our conversation from today:

https://people.freedesktop.org/~cbrill/dri-log/?channel=radeon&date=2019-01-04

I am running i3wm on a Ryzen 5 2400G desktop using the integrated vega graphics. I am also using the modesetting xorg driver.

The artifacts occur when I run a program and when the program hasn't drawn anything yet, the artifacts then are gone after the program finished loading. I've seen this bug mostly with Qt applications.

My system details:

• Arch Linux
• mesa 18.3.1-1
• Linux 4.20.0-arch1-1-ARCH

Expected Behavior

No image corruption/artifacts when opening new windows.

Reproduction Instructions

Start an application on i3wm and observe corruption when the application hasn't drawn anything yet.

Environment

Output of i3 --moreversion 2>&-:

i3 version: 
Binary i3 version:  4.16 (2018-11-04) © 2009 Michael Stapelberg and contributors
Running i3 version: 4.16 (2018-11-04) (pid 536)to abort…)
Loaded i3 config: /home/diego/.config/i3/config (Last modified: Mon 24 Dec 2018 02:33:19 AM -02, 999901 seconds ago)

The i3 binary you just called: /usr/bin/i3
The i3 binary you are running: i3

https://github.com/diegoviola/dotfiles/blob/master/.config/i3/config

- Linux Distribution & Version: Arch Linux
- Are you using a compositor (e.g., xcompmgr or compton): No.

[i3bot]

I don’t see a link to logs.i3wm.org. Did you follow https://i3wm.org/docs/debugging.html? (In case you actually provided a link to a logfile, please ignore me.)

[diegoviola]

MrCooper | dviola: AFAICT i3 seems to copy from a pixmap to its parent window; my guess would be it never initializes the contents of that
         | pixmap
MrCooper | so its contents are undefined
MrCooper | that indeed seems to be the case, if I interpret the code in i3-wm/src/x.c:x_push_node() correctly
MrCooper | according to /usr/include/cairo/cairo.h, cairo doesn't initialize the contents of a surface created with
         | cairo_xcb_surface_create
MrCooper | I'd file an i3 bug about this

[Airblader]

Thanks for the report. I haven't looked yet, but this sounds pretty reasonable and likely. Related, but for a different codepath I think is #3481.

[cdaringe]

i'm observing this as well

$ i3 --moreversion 2>&- | pbcopy
Binary i3 version:  4.16.1-1-g341fcf99 (2019-01-27, branch "gaps") © 2009 Michael Stapelberg and contributors
(Getting version from running i3, press ctrl-c to abort…)
Running i3 version: 4.16.1-1-g341fcf99 (2019-01-27, branch "gaps") (pid 1350)
Loaded i3 config: /home/cdaringe/.i3/config (Last modified: Sat 20 Apr 2019 09:07:34 PM PDT, 1732 seconds ago)

The i3 binary you just called: /usr/bin/i3
The i3 binary you are running: i3

config: https://github.com/cdaringe/dotfiles/blob/master/env/.i3/config

in the event that @diegoviola's datas don't provide sufficient debuggins info, i am also willing to start collecting logs later

[shiro]

Also experiencing this on an AMD GPU

$ i3 --moreversion 2>&-
Binary i3 version:  4.17.1 (2019-08-30) © 2009 Michael Stapelberg and contributors
Running i3 version: 4.17.1 (2019-08-30) (pid 4637)abort…)
Loaded i3 config: /home/shiro/.config/i3/config (Last modified: Wed 06 Nov 2019 01:11:42 AM CET, 384 seconds ago)

The i3 binary you just called: /usr/bin/i3
The i3 binary you are running: /usr/bin/i3

Corruption is definitely the same as the post above, is apparent and causes serious eye pain.

Same result on the latest git version of i3-gaps also.

[vladimirardelean]

I have the same problem on Debian and as far as i see on the net is very present in all amd apu's with vega graphics, the artifacts are very bad when accelerated graphics is needed, this is clearly a amdgpu open source driver issue. I wonder why there is not a clear fix as this happened from ryzen 2nd generation apu's.

[Alexey104]

this is clearly a amdgpu open source driver issue

I use Nvidia, and I have always had this issue in i3 with both free/proprietary drivers and with/without any compositors. Similar artifacts also appear when resizing windows.