amdgpu: System with [AMD/ATI] Device [1002:73e3] have chance to fail entering s2idle
Brief summary of the problem:
I have a desktop with the [AMD/ATI]1002:73e3 installed. If there's no external monitor connected, the system can survice suspend/resume for >100 times w/o problem. If I connect the external monitor to the DP port on the motherboard (Intel GFX), there's no problem either. However, if I connect the DP cable to the AMD GFX card and run the suspend/resume test, it will have chance to freeze within 30 times suspend/resume test. The display is already blank, connect via ssh is down, but keyboard seems not yet suspended (keyboard light is still on). The system doesn't respond to anything at this moment until I power-cycle the machine. Tried kernel v6.3-rc2 and latest drm-tip kernel (date code 0323) w/o luck. Any suggestion about how to identify the problem? Thanks
Hardware description:
system-manufacturer: Dell Inc. system-product-name: Precision 3660 bios-version: 1.62.69 CPU: 13th Gen Intel(R) Core(TM) i9-13900 (32x) GPU: 0000:00:02.0 Display controller [0380]: Intel Corporation Device [8086:a780] (rev 04) 0000:03:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Device [1002:73e3]
System information:
- Distro name and Version: Ubuntu 22.04
- Kernel version: v6.1
How to reproduce the issue:
Connect the external monitor to the AMD GFX DP port. Manually suspend resume for 30 times. Or run rtcmake to trigger suspend.