page allocation failure on wake from suspend
Brief summary of the problem:
Resuming my desktop computer from suspend (S3) often causes this error.
After waking the displays remain off. I can sshed in and reboot from the command line but the GPU did not recover and I had to power cycle.
- CPU: AMD Ryzen 9 3900X
- GPU: AMD RX 560
- System Memory: 16Gb
- Display(s): 2 × Dell P2415Q 24" 4K displays
- Type of Display Connection: 1 × DP, 1 × HDMI
- Distro name and Version: Arch Linux
- Kernel version: 5.9.1-arch1-1
- Custom kernel: N/A
- AMD package version: No package
How to reproduce the issue:
- Suspend computer
- Wake computer (press power button)
- Output of
journalctlfor the session that failed
amdgpu.log. Notable points in time:
- Nov 11 18:49:49 — Suspended
- Nov 12 09:31:37 — page allocation failure upon wake
- Nov 12 09:31:37 — amdgpu: Move buffer fallback to memcpy unavailable
- Nov 12 09:31:37 — amdgpu 0000:09:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] ERROR ring gfx test failed (-110)
- Nov 12 09:31:37 — [drm:amdgpu_device_ip_resume_phase2 [amdgpu]] ERROR resume of IP block <gfx_v8_0> failed -110
- Nov 12 09:31:37 — [drm:amdgpu_device_resume [amdgpu]] ERROR amdgpu_device_ip_resume failed (-110).
- Nov 12 09:31:37 — [drm:amdgpu_cs_ioctl [amdgpu]] ERROR Failed to process the buffer list -19!
- Nov 12 09:31:48 — amdgpu 0000:09:00.0: amdgpu: GPU reset begin!
- Nov 12 09:31:48 — BUG: kernel NULL pointer dereference, address: 0000000000000000
- Nov 12 09:33:16 — [drm:amdgpu_job_timedout [amdgpu]] ERROR ring sdma1 timeout, signaled seq=1715036, emitted seq=1715037amdgpu.log