KDE (plasma, kwin, ...) on Wayland causes repeated GPU hangs

I wanted to try out KDE on Wayland, but upon logging in it immediately causes a GPU hang. It recovers after a few seconds and I get to use the desktop for anywhere between another few seconds to several minutes, then the next GPU hang inevitably hits. It repeats like that.

The device is a Dell XPS 13 9310, so Intel graphics. plasma packages are version 5.23.3-1, wayland is 1.19.0-2. The kernel is 5.15.2-2-MANJARO.

[44248.661314] i915 0000:00:02.0: [drm] GPU HANG: ecode 12:1:85dffffb, in plasmashell [201251]
[44248.662338] i915 0000:00:02.0: [drm] Resetting rcs0 for stopped heartbeat on rcs0
[44248.662372] i915 0000:00:02.0: [drm] plasmashell[201251] context reset due to GPU hang
[44253.654926] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:6!
[44253.654934] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:4!
[44253.656224] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:8!
[44253.658105] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:a!
[44253.659134] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:c!
[44253.659316] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:e!
[44253.661303] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:10!
[44253.662762] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:12!
[44253.663457] Fence expiration time out i915-0000:00:02.0:plasmashell[201251]:14!
[44259.529381] Asynchronous wait on fence 0000:00:02.0:kwin_wayland[201177]:f0 timed out (hint:intel_atomic_commit_ready [i915])
[44263.590880] i915 0000:00:02.0: [drm] GPU HANG: ecode 12:1:85dffffb, in plasmashell [201251]
[44263.591901] i915 0000:00:02.0: [drm] Resetting rcs0 for stopped heartbeat on rcs0
[44263.591934] i915 0000:00:02.0: [drm] plasmashell[201251] context reset due to GPU hang
[44348.703081] Asynchronous wait on fence 0000:00:02.0:kwin_wayland[201177]:1102 timed out (hint:intel_atomic_commit_ready [i915])
[44353.620632] i915 0000:00:02.0: [drm] GPU HANG: ecode 12:1:85dffffb, in plasmashell [201553]
[44353.621655] i915 0000:00:02.0: [drm] Resetting rcs0 for stopped heartbeat on rcs0
[44353.621687] i915 0000:00:02.0: [drm] plasmashell[201553] context reset due to GPU hang
[44441.504513] Asynchronous wait on fence 0000:00:02.0:kwin_wayland[201177]:23a0 timed out (hint:intel_atomic_commit_ready [i915])
[44446.630000] i915 0000:00:02.0: [drm] GPU HANG: ecode 12:1:85dffffb, in kwin_wayland [201177]
[44446.631026] i915 0000:00:02.0: [drm] Resetting rcs0 for stopped heartbeat on rcs0
[44446.631061] i915 0000:00:02.0: [drm] kwin_wayland[201177] context reset due to GPU hang

I’ve disabled i915.enable_psr and enabled i915.enable_fbc as per recommendations in the Arch wiki page for the 9310, and I feel like it improved a bit but it’s hard to quantify. I have not tried other environments such as GNOME or Sway. This is with a newly-created user with default settings, and the session was started with the ly display manager.

Is there anything I could do or is the combination of KDE, Wayland and Intel just not working well?

5.15.2-2-MANJARO kernel does not work with intel graphics yet

(see https://forum.manjaro.org/t/kernel-5-15-issue-in-recent-stable-update/91031/10)

manjaro patched kernel 5.15.5-2 (should be in testing branch)

It seems to work well with 5.15.5-2 from testing. Thanks both.

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.