amdgpu oopses and dies after ACPI-S3-Resuming when no monitors are active
Hi,
Description:
I started observing this some weeks ago, first time with Linux 6.7.4 but the situation stays the same even with current master (6.8-rc8+).
Repro steps:
- send the system to standby (
systemctl suspend
) which goes to ACPI S3 mode - turn off the power of all monitors
- resume the system with a keypress
- wait a couple of seconds and restore monitor power
Observed result:
- the screen remains either black or shows some shifted fragment of the last xscreensaver screen
- system no longer properly accessible, the only way out is rebooting via Magic-SysRQ (keyboard still works as long as you don't reconnect it, otherwise there is no way anymore)
- after reboot (when properly synced/unmounted), kernel messages can be retrieved (see attachment)
- no final messages in Xorg log to find - which makes sense since the journal reports "kernel: note: Xorg[3043] exited with irqs disabled"
Reproducibility:
- always (as described)
- never (if the monitors were working before resuming including one which takes ~5s to start)
Hardware description:
- CPU: Ryzen 5700G (IGP disabled in BIOS, although it does not make a difference)
- GPU: Advanced Micro Devices, Inc. [AMD/ATI] Navi 22 [Radeon RX 6700/6700 XT/6750 XT / 6800M/6850M XT] [1002:73df] (rev c1)
- System Memory: 32GB DDR4-2666
- Display(s): various, tried older WQHD, older FHD gaming display (does not make a difference)
- Type of Display Connection: DP (1x or 2x, does not make a difference)
System information:
- Distro name and Version: Debian Testing (latest)
- Kernel version: various 6.7+, see description
- Custom kernel: varios, including current Linux master (for 6.8.x, 5ad3cb0ed525b80c7f66c32b49a68c1f3510bec9)
- AMD official driver version: N/A
Log files (for system lockups / game freezes / crashes)
- Dmesg log (excerpt) drm-crash.txt