Skip to content

dGPU keeps shutting down causing short freezes and does not wake up from sleep

Brief summary of the problem:

Last 3 months I was using dGPU as primary for GNOME Wayland via ENV{DEVNAME}=="/dev/dri/card0", TAG+="mutter-device-preferred-primary".

I used sudo radeontop -b 3 workaround to prevent permanent freeze on dGPU shutdown and have a working wake-up from sleep (s2idle).

On 5.17.1 does not seem to help with sleep anymore so here I will share all the necessary information to fix these issues preferably without any more workarounds.

Hardware description:

  • Machine Type: Laptop System: ASUSTeK product: ROG Strix G513QY_G513QY v: 1.0 (AMD Advantage Edition)
  • CPU: 8-core model: AMD Ryzen 9 5900HX with Radeon Graphics
  • iGPU: 08:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Cezanne [1002:1638] (rev c4)
  • dGPU (RX6800M): 03:00.0 Display controller [0380]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 22 [Radeon RX 6700/6700 XT / 6800M] [1002:73df] (rev c3)
  • iGPU is connected to internal laptop screen via eDP, dGPU is set as primary for GNOME Wayland session
  • System Memory: 64 GB

System information:

  • Distro name and Version: Manjaro rolling
  • Kernel version: Linux 5.17.1-3-MANJARO SMP PREEMPT Thu Mar 31 12:27:24 UTC 2022 x86_64 GNU/Linux

How to reproduce the issue:

Short freezes

  • Set dGPU as primary for GNOME Wayland via ENV{DEVNAME}=="/dev/dri/card0", TAG+="mutter-device-preferred-primary"
  • Login into GNOME Wayland
  • Tail the system journal via sudo journalctl -f
  • Stop UI interactions for 10 seconds then resume
  • Notice a short freeze 1-2 seconds and log messages in the journal about dGPU resuming

Does not wake up from sleep (s2idle)

  • Set dGPU as primary for GNOME Wayland via ENV{DEVNAME}=="/dev/dri/card0", TAG+="mutter-device-preferred-primary"
  • Login into GNOME Wayland
  • Press the power button to go into Sleep
  • After a few seconds press the power button again to resume
  • Notice the keyboard and indicators light up but the screen stays blank

Results of a more extensive testing

I tested multiple kernel versions from 5.15, 5.16 and 5.17 series and no longer think it's a regression, they all work the same. So, I did an extensive testing on kernel 5.17.1-3-MANJARO. Here is what I did and found:

  • Added pm_debug_messages=1 amd_pmc.dyndbg=+p to kernel parameters
  • Logged into Gnome Wayland session with dGPU as primary (iGPU has eDP output connected to laptop screen)
  • Ran sudo journalctl -f, pressed Win button (Gnome workspace overview) and watched
  • 5.17.1_journal_minimal-load.zip - every 7 seconds dGPU seems to resume and go back to sleep again
  • Entered sleep, waited 10 seconds, tried to resume -> blank screen -> force shutdown
  • Logged in again the same way and additionally ran sudo radeontop -b 3. Entered overview and watched
  • 5.17.1_journal_radeontop-dgpu.zip - with radeontop running dGPU not longer goes back and forth sleep/resume
  • Entered sleep, waited 10 seconds, tried to resume -> blank screen -> force shutdown
  • Logged in again the same way, run instead glxgears (XWayland, used dGPU for OpenGL) and watched
  • 5.17.1_journal_glxgears.zip - with glxgears running dGPU also does not go back and forth resuming
  • Entered sleep, waited 10 seconds, resumed successfully. Entered sleep, waited 8 hours, resumed successfully. I still ended up with blank screen while resuming again after a few hours.
  • My BIOS does not have any options for enabling NVME passwords

Attached files:

Log files

Edited by Dmytro Voloshyn