AMD Ryzen 5700G (Renoir) + ASUS B450-I: Second suspend hangs after resume - xhci_hcd & asus_wmi issues

General system Kernel, boot, graphics & hardware
1 
========= System =============== 
      
- CPU: AMD Ryzen 7 5700G (Renoir APU)
- Motherboard: ASUS ROG STRIX B450-I (BIOS 5901)
- OS: EndeavourOS (Arch-based)
- Kernel: 6.12.63-1-lts
- Mesa: 25.3.4
- Display: Wayland (GNOME Shell 49.3)

========== Bug ===================

First suspend/resume cycle works correctly. Second suspend hangs with cursor update failures. PC fan keeps running.

========== Logs in Attachement ===================

- after-resume-1_with-612-kernel.log 
- after-failed-second-suspend_with-612-kernel.log 
- suspend-log-0x1C_with-618-kernel.txt

========== Key Errors ===================

- xhci_hcd 0000:01:00.0: xHC error in resume, USBSTS 0x401, Reinit
- gnome-shell: Cursor update failed: drmModeAtomicCommit: Invalid argument
- System hangs at "Performing sleep operation 'suspend'" 

========== Update ======================

Testing with Kernel 6.18 shows:
- New error: "asus_wmi: failed to register LPS0 sleep handler" (not present in 6.12 LTS)
- amdgpu.dcfeaturemask=0x1C fixes cursor errors in 6.12 LTS but has NO effect in 6.18
- Suspect: ASUS board-specific LPS0 handler failure interferes with suspend
- Hardware: ASUS ROG STRIX B450-I GAMING (BIOS 5901)

====== Workaround =============

Use suspend-to-disk (hibernation) instead of suspend-to-RAM.

20260203-bugzilla-kernel-log-files-suspend-issue.pdf (44.4 KB)

2

Update: Looks resolved after some kernel/Mesa updates!

Wanted to follow up on this after a few months of using hibernation as a
workaround. My original report was on kernel 6.12.63-1-lts / Mesa 25.3.4.
I’m now on kernel 6.18.42-1-lts / Mesa 26.0, and just tested 3 consecutive
suspend/resume cycles - all successful, no hang. Previously the second
cycle would reliably hang after “PM: suspend entry (deep)” and needed a
hard reset.

I did retest a few months ago on 6.18.17-1-lts and it was still broken
then, so the fix landed somewhere between 6.18.17 and 6.18.42 (kernel),
or possibly needed the Mesa 26.0 upgrade too - I can’t pinpoint the exact
patch.

The cursor error and xhci_hcd/asus_wmi messages still show up in the logs,
but they no longer block suspend/resume - just cosmetic/non-fatal now.

Going to keep testing over the next days before fully switching back from
hibernate to suspend as my daily driver, but wanted to share in case
others with similar hardware (Ryzen 5700G + ASUS B450) want to retest too.

Also posted this update on the AMD GitLab issue:

Thanks to everyone who read/engaged with the original thread!