Bug 216899 - amdgpu ring timeout after gfxoff on Navi 24 beige_goby
Summary: amdgpu ring timeout after gfxoff on Navi 24 beige_goby
Status: RESOLVED ANSWERED
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: Intel Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-01-07 16:28 UTC by Dylan Simon
Modified: 2023-01-08 09:39 UTC (History)
0 users

See Also:
Kernel Version: 6.1.2-gentoo
Subsystem:
Regression: No
Bisected commit-id:


Attachments
boot messages and pci info (124.70 KB, text/plain)
2023-01-07 16:28 UTC, Dylan Simon
Details

Description Dylan Simon 2023-01-07 16:28:12 UTC
Created attachment 303547 [details]
boot messages and pci info

I have a new Powercolor AMD Radeon RX 6400 ITX and whenever the screen is blanked (DPMS or X blanking -- I haven't tried non-X) with gfxoff enabled, the whole system hangs (or at least network and alt-sysrq are unresponsive, though there seems to be some activity):

Jan  7 01:12:29 talula kernel: [drm:amdgpu_job_timedout] *ERROR* ring gfx_0.0.0 timeout, signaled seq=23406, emitted seq=23407
Jan  7 01:12:29 talula kernel: [drm:amdgpu_job_timedout] *ERROR* Process information: process X pid 3404 thread X:cs0 pid 3412
Jan  7 01:12:29 talula kernel: amdgpu 0000:75:00.0: amdgpu: GPU reset begin!
Jan  7 01:12:32 talula kernel: amdgpu 0000:75:00.0: amdgpu: SMU: I'm not done with your previous command: SMN_C2PMSG_66:0x00000029 SMN_C2PMSG_82:0x00000000
Jan  7 01:12:32 talula kernel: amdgpu 0000:75:00.0: amdgpu: Failed to disable gfxoff!

Setting amdgpu_gfxoff to 0 in debugfs prevents this, and everything works correctly.  I've tried this on 5.15.80 and checked I have the most recent firmware.  Happy to provide any additional info or try additional diagnostics.  Full boot messages and pci info for card attached.
Comment 1 Artem S. Tashkinov 2023-01-08 09:39:49 UTC
Please search for duplicates here https://gitlab.freedesktop.org/drm/amd/-/issues and if there are none, file a bug report over there.

Note You need to log in before you can comment on or make changes to this bug.