AMDGPU stalls since Xen 4.14.5-2 upgrade

[awokd]

Ever since the following upgrades:

python3-xen x86_64 2001:4.14.5-2.fc32 qubes-dom0-cached
60 k
xen x86_64 2001:4.14.5-2.fc32 qubes-dom0-cached
20 k
xen-hypervisor x86_64 2001:4.14.5-2.fc32 qubes-dom0-cached
7.8 M
xen-libs x86_64 2001:4.14.5-2.fc32 qubes-dom0-cached
625 k
xen-licenses x86_64 2001:4.14.5-2.fc32 qubes-dom0-cached
32 k
xen-runtime x86_64 2001:4.14.5-2.fc32 qubes-dom0-cached
19 M

My AMDGPU keeps crashing every 15 min with errors like:

dom0 kernel: radeon 0000:42:00.0: ring 4 stalled for more than 10113msec
dom0 kernel: radeon 0000:42:00.0: GPU lockup (current fence id
0x00000000001c4209 last fence id 0x00000000001c4219 on ring 4)
dom0 kernel: radeon 0000:42:00.0: Saved 1164 dwords of commands on ring 0.
dom0 kernel: radeon 0000:42:00.0: GPU softreset: 0x00000049

Swapped out the GPU with a spare in case it had gone bad, but didn't
help. Currently running Linux dom0 5.10.112-1.fc32.qubes.x86_64. Tried
booting with an earlier kernel, but no change. Any suggestions? Haven't
tried downgrading Xen yet.

--
- don't top post
Mailing list etiquette:
- trim quoted reply to only relevant portions
- when possible, copy and paste text instead of screenshots