Artifacts in X sessions, GPU fault 147

Submitted by Andrew Dorney

Assigned to Default DRI bug account

Link to original bug (#107095)

Description

Created attachment 140442 dmesg

I purchased a new "MSI Radeon RX 580 Armor MK2 8G OC" this week. I am getting intermittend visual artifacts on some windows, but not others.

Arch Linux

linux 4.17.3-1 mesa 18.1.3-1 xf86-video-amdgpu 18.0.1-2 xorg-server 1.20.0-9

When I play games, the card performs admirably. No artifacting, no stutter. I've tripled-to-quadrupled my FPS, so accelleration works great. Where I'm having problems is on GUIs.

Text and GUI elements often show squares or lines. The color of the squares or lines are the color of the window or desktop below the active window. When I move or force a redraw on the active window, the squares/lines disappear or reappear in a new place on the active window. It occurs regularly in Firefox, xterm, Konsole, and kdesu4 permission popups.

The problem occurs both when I am running only GUIs, and when I am running games and GUIs simultaneously, so I don't think it's a power state issue.

To narrow down if it was Plasma or not, I rebooted and started Fluxbox. I started only xterms, ran some commands, and got no errors whatsoever. Then I started moving them around the screen. That's when I got this in my dmesg:

[ 200.335982] amdgpu 0000:01:00.0: GPU fault detected: 147 0x0508c402 [ 200.335986] amdgpu 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x000FFAA1 [ 200.335988] amdgpu 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0A0C4002 [ 200.335991] amdgpu 0000:01:00.0: VM fault (0x02, vmid 5, pasid 32768) at page 1047201, read from 'TC3' (0x54433300) (196)

From then on, artifacts appeared on all windows per usual. Gaming and heavy 3D remained unaffected.

Should I RMA the card, or is this likely a software stack bug and I should wait for more Mesa/Kernel updates?

Please let me know if I can provide more information. Thanks!

Attachment 140442, "dmesg":
dmesg_2018-07-02.log