Bug#928146: libdrm-amdgpu1: GPU hang

[Tuxicoman <debian@jesuislibre.net>]

Package: libdrm-amdgpu1
Severity: important

Dear Maintainer,

Sometimes my AMD GPU (Radeon 580) freezes and I have to force reboot the
system.
This appears when I play video games.

Here is a trace from dmesg:

[41030.750400] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout,
signaled seq=13431126, emitted seq=13431128
[41030.750405] [drm] GPU recovery disabled.
[41206.115499] INFO: task kworker/u8:0:6943 blocked for more than 120 seconds.
[41206.115504]       Tainted: G           OE     4.19.0-4-amd64 #1 Debian
4.19.28-2
[41206.115506] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this
message.
[41206.115508] kworker/u8:0    D    0  6943      2 0x80000000
[41206.115523] Workqueue: events_unbound commit_work [drm_kms_helper]
[41206.115525] Call Trace:
[41206.115532]  ? __schedule+0x2a2/0x870
[41206.115535]  ? __switch_to_asm+0x40/0x70
[41206.115537]  schedule+0x28/0x80
[41206.115539]  schedule_timeout+0x26d/0x390
[41206.115611]  ? dce110_timing_generator_get_position+0x5b/0x70 [amdgpu]
[41206.115673]  ? dce110_timing_generator_get_crtc_scanoutpos+0x70/0xb0
[amdgpu]
[41206.115676]  dma_fence_default_wait+0x238/0x2a0
[41206.115679]  ? dma_fence_release+0x90/0x90
[41206.115681]  dma_fence_wait_timeout+0xdd/0x100
[41206.115684]  reservation_object_wait_timeout_rcu+0x173/0x280
[41206.115748]  amdgpu_dm_do_flip+0x112/0x340 [amdgpu]
[41206.115811]  amdgpu_dm_atomic_commit_tail+0x750/0xdb0 [amdgpu]
[41206.115818]  ? wait_for_completion_timeout+0x3b/0x1a0
[41206.115828]  commit_tail+0x3d/0x70 [drm_kms_helper]
[41206.115833]  process_one_work+0x1a7/0x3a0
[41206.115836]  worker_thread+0x30/0x390
[41206.115838]  ? create_worker+0x1a0/0x1a0
[41206.115841]  kthread+0x112/0x130
[41206.115843]  ? kthread_bind+0x30/0x30
[41206.115845]  ret_from_fork+0x35/0x40

As I encounter such crashes often, I can give you more log info if you tell me
how to collect it

Thanks for the help.



-- System Information:
Debian Release: buster/sid
  APT prefers testing
  APT policy: (500, 'testing')
Architecture: amd64 (x86_64)
Foreign Architectures: i386

Kernel: Linux 4.19.0-4-amd64 (SMP w/4 CPU cores)
Kernel taint flags: TAINT_OOT_MODULE, TAINT_UNSIGNED_MODULE
Locale: LANG=fr_BE.utf8, LC_CTYPE=fr_BE.utf8 (charmap=UTF-8), LANGUAGE=fr_BE.utf8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)
LSM: AppArmor: enabled