Randomly occurring GPU hang on CherryView under Wayland (ecode 8:1:85dffffb)

• A clear subject describing the issue.

The hang occurs for no apparent reason and manifests itself in form of complete freeze for several seconds and rapid redrawing of the frame before the freeze and the one after, sometimes corrupted. Everything else, including the audio and the ability to move cursor, stays completely operational and it's even possible to restart the Wayland compositor blindly.

• Steps to reproduce the issue.

The hang doesn't seem to occur when there's no user-side activity. Other than that, it's completely random.

• How often does the steps listed above trigger the issue? For example: always, 1 out 3 times.

It happens always but takes different time to occur (usually from several hours to several minutes; half an hour in this case).

• Which platforms and features are affected (if you can).

The Wayland compositor (sway) yields absolutely no error messages in the debug mode when the hang happens, so other compositors may be affected too.

• The following information about your system:
  • system architecture: x86_64
  • kernel version: 5.6.0-300.fc32.x86_64
  • Linux distribution: Fedora 32 Beta
  • Machine or mother board model: see dmidecode.gz attached
  • Display connector: HDMI
  • GPU crash dump: see kernel.log.gz attached

---

Apr 02 19:57:53 machine-one kernel: Asynchronous wait on fence 0000:00:02.0:sway[1741]:8726 timed out (hint:intel_atomic_commit_ready+0x0/0x58 [i915])
Apr 02 19:57:57 machine-one kernel: i915 0000:00:02.0: GPU HANG: ecode 8:1:85dffffb, in Compositor [1873]
Apr 02 19:57:57 machine-one kernel: GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Apr 02 19:57:57 machine-one kernel: Please file a _new_ bug report at https://gitlab.freedesktop.org/drm/intel/issues/new.
Apr 02 19:57:57 machine-one kernel: Please see https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs for details.
Apr 02 19:57:57 machine-one kernel: drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Apr 02 19:57:57 machine-one kernel: The GPU crash dump is required to analyze GPU hangs, so please always attach it.
Apr 02 19:57:57 machine-one kernel: GPU crash dump saved to /sys/class/drm/card0/error
Apr 02 19:57:57 machine-one kernel: i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0
Apr 02 19:57:57 machine-one kernel: i915 0000:00:02.0: Compositor[1873] context reset due to GPU hang

dmidecode.gz kernel.log.gz