[iris/anv] GPU hangs in Unigine Heaven & Valley benchmarks with 2x AA setting
System information
System:
Host: gp62-linux418 Kernel: 6.1.6-arch1-3 arch: x86_64 bits: 64
compiler: gcc v: 12.2.1 Desktop: KDE Plasma v: 5.26.5 tk: Qt v: 5.15.8
wm: kwin_wayland dm: SDDM Distro: Arch Linux
CPU:
Info: quad core model: Intel Core i7-6700HQ bits: 64 type: MT MCP
arch: Skylake-S rev: 3 cache: L1: 256 KiB L2: 1024 KiB L3: 6 MiB
Speed (MHz): avg: 1325 high: 2600 min/max: 800/3500 cores: 1: 900 2: 900
3: 2600 4: 900 5: 901 6: 900 7: 2600 8: 900 bogomips: 41621
Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx
Graphics:
Device-1: Intel HD Graphics 530 vendor: Micro-Star MSI driver: i915
v: kernel arch: Gen-9 ports: active: eDP-1 empty: DP-1,HDMI-A-1,HDMI-A-2
bus-ID: 00:02.0 chip-ID: 8086:191b
Device-2: NVIDIA GM107M [GeForce GTX 960M] vendor: Micro-Star MSI
driver: nouveau v: kernel arch: Maxwell pcie: speed: 8 GT/s lanes: 16
bus-ID: 01:00.0 chip-ID: 10de:139b temp: 39.0 C
Device-3: Chicony USB 2.0 Webcam Device type: USB driver: uvcvideo
bus-ID: 1-11:5 chip-ID: 04f2:b521
Display: wayland server: X.org v: 1.21.1.6 with: Xwayland v: 22.1.7
compositor: kwin_wayland driver: X: loaded: modesetting
alternate: fbdev,intel,nouveau,nv,vesa dri: iris,nouveau gpu: i915,nouveau
display-ID: 0
Monitor-1: eDP-1 res: 1920x1080 size: N/A
API: OpenGL v: 4.6 Mesa 22.3.3 renderer: Mesa Intel HD Graphics 530 (SKL
GT2) direct render: Yes
Describe the issue
When running Unigine Heaven 4.0 or Unigine Valley 1.0 with 2x anti aliasing (which is part of the Basic preset for both) on the Intel GPU (both with the iris driver as well as with zink+anv) GPU hang events occur before the benchmark either crashes (on iris) or freezes (with zink+anv).
I also tried running them on Plasma with X11 as well as using dwm on X11, the issue occurs in those cases as well.
Changing the AA setting to disabled, 4x or 8x makes the issue go away, and changing the other settings (quality, tesselation, resolution) has no effect on the issue.
Running those two benchmarks on Nouveau via DRI_PRIME=1
works fine.
Regression
Only tried Heaven and Valley today so no idea unfortunately.
Log files as attachment
- dmesg.log (iris) dmesg-zink.log
- backtrace.log (iris)
- Output of
/sys/class/drm/card0/error
: error.bz2 (iris) error-zink.bz2
The above logs (both iris and zink) were obtained by running Heaven using dwm on X11 with the drm-tip kernel (built on 19 Jan 2023 05:47:24 +0000) plus Mesa 22.3.3, although as mentioned before I can reproduce it in Plasma X11 and Wayland as well, and using kernels 5.15.88 and 6.1.6 also did not have any effect.
Any extra information would be greatly appreciated
I originally filed this issue at drm/intel#7901, and I was told that it might fit better here.