[drm:amdgpu_dm_atomic_commit_tail [amdgpu]] *ERROR* Waiting for fences timed out or interrupted!

[Christian Pernegger]

Hello,

one of my computers has been crashing while gaming rather a lot
lately, with kernel messages pointing to amdgpu. First line see
subject, rest in the attached log.
SSH still works, attempts to shutdown/reboot don't quite finish.

Radeon VII in an Asus Pro WS X570-Ace. Ubuntu 18.04.3 HWE, mesa-aco.
This one was with kernel 5.3.0-24-generic [hwe-edge], mesa
19.3.0+aco+git1575452833-3409c06e26d-1bionic1, vesa20_* from
linux-firmware-20191022, running Ori and the Blind Forest: Definitive
Edition via Proton/WINED3D11 under Steam Remote Play. I've had similar
crashes sporadically even with 5.0 [plain hwe] and linux-firmware
completely stock, and with native games (e.g. Crusader Kings II)
running locally.
It used to be maybe once every other week, though, that was tolerable,
now Ori usually triggers it in under an hour. Turning off ACO via
RADV_PERFTEST=llvm makes it worse (not bad enough to make it trigger
quickly and reliably. though), going back to kernel 5.0 helps (as in
an hour or two might go by without a crash, but the performance impact
is severe).

All very vague. Which is why this isn't pretending to be a bug report,
just a "has anyone seen this?" kind of shout-out. If it's worthy of
following up, I'd be happy to provide further info, just tell me what.

Cheers,
C.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: amdgpu.journal
Type: application/octet-stream
Size: 51233 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20191206/a48de9bf/attachment-0001.obj>