## System details

ASUS Zenbook 14 UM3406KA; AMD Ryzen AI 7 350 / Radeon 860M (Krackan, amdgpu); Omarchy 4.0.4-1; kernel 7.2.5-3-omarchy; linux-firmware-amdgpu 20260810-2; BIOS UM3406KA.305 (2025-07-29).

## What's wrong?

Hibernation does not resume correctly on this Zenbook. In the captured incident, systemd entered hibernation at **2026-10-02 01:15:07 UTC-5**. The restored session logged waking from ACPI S4 at **08:43:21**, immediately followed by AMDGPU MES/queue and device-restore failures. GPU reset/recovery also failed. Systemd nevertheless recorded that the hibernate operation returned at **08:43:51**. A subsequent fresh boot began at **08:44:35**; the reason for that reboot was not established by this diagnosis.

Expected: return to the previous session with the GPU and other devices restored and functional.

Observed failure sequence: `MES ... ADD_QUEUE` does not respond, restoration of `gfx_v11_0` fails with `-110`, the SDMA ring times out, and GPU recovery fails with `-62`. The same restoration also logs a MediaTek Wi-Fi timeout, AMD NPU PSP/SMU errors, and later ACPI AML loop timeouts.

The exact method used to request hibernation, the wake trigger, applications/peripherals connected, visible screen/input symptom, and reproduction frequency were not captured in the read-only inspection. They should not be inferred from the kernel logs. No new reproduction or workaround test was performed.

### Full hardware and package details

- Laptop: ASUS Zenbook 14 `UM3406KA_UM3406KA`; motherboard `UM3406KA`.
- CPU: AMD Ryzen AI 7 350 w/ Radeon 860M; 16 logical CPUs.
- GPU: `63:00.0` AMD/ATI Krackan, identified by lspci as `Radeon 840M / 860M Graphics`, PCI `1002:1114`, revision `c2`; ASUS subsystem `1043:1194`. Driver in use: `amdgpu`. No NVIDIA display controller was detected.
- RAM available to Linux: `31941408 KiB` (approximately 30.46 GiB).
- BIOS: `UM3406KA.305`, dated `2025-07-29`.
- OS: Omarchy 4.0.4, based on Arch Linux.
- Kernel running during the failed resume and the current boot: `7.2.5-3-omarchy`, `#1 SMP PREEMPT_DYNAMIC Mon, 14 Sep 2026 19:55:01 +0000`.
- Stock Arch kernel is also installed, but has not been tested for this incident.
- NVMe: SK hynix PVC10 (DRAM-less), PCI `1c5c:1f69`, at `61:00.0`; approximately 953.9 GiB.
- Root: LUKS-encrypted Btrfs on `/dev/mapper/root`; root subvolume `@`.
- Boot: Limine and a UKI at `/boot/EFI/Linux/omarchy_linux-omarchy.efi`.

Installed package versions at diagnosis:

```text
omarchy 4.0.4-1
linux-omarchy 7.2.5-3
linux 7.2.3.arch1-3
linux-firmware 20260810-2
linux-firmware-amdgpu 20260810-2
linux-firmware-mediatek 20260810-2
linux-firmware-other 20260810-2
amd-ucode 20260810-2
systemd 261.2-1
mkinitcpio 41.1-1
limine 12.8.0-1
limine-mkinitcpio-hook 1.38.0-1.1
btrfs-progs 7.1-1
nvidia-utils 610.57.04-1
```

### Hibernation configuration already checked

- Persistent swap: `/swap/swapfile`, `31941404 KiB` usable (about 30.46 GiB), priority 0; active and unused at inspection.
- Additional swap: `/dev/zram0`, `31940604 KiB`, priority 100. This is RAM-backed and is not counted as persistent hibernation storage.
- Image-size setting: `/sys/power/image_size = 12991971328` bytes (about 12.10 GiB). The journal reports snapshot preallocation of `13489856 kbytes`; this is an allocation message, not a measurement of bytes ultimately written to swap. No insufficient-swap error was found in the captured attempt.
- `resume=/dev/mapper/root resume_offset=1965756` agrees across `/proc/cmdline`, Limine, and the UKI's embedded command line.
- `/sys/power/resume = 253:0`, matching `/sys/block/dm-0/dev`.
- `/sys/power/resume_offset = 1965756`.
- Read-only `btrfs inspect-internal map-swapfile -r /swap/swapfile` returned **1965756**.
- Resume parameters originate in `/etc/limine-entry-tool.d/resume.conf`. `/etc/kernel/cmdline` contains the base root/cipher parameters, with the resume parameters added by the Limine drop-in.
- Other relevant boot parameters: `zswap.enabled=0`, `rtc_cmos.use_acpi_alarm=1`, `initramfs_async=0`, `quiet splash loglevel=0 systemd.show_status=false rd.udev.log_level=0`.
- `/sys/power/state`: `freeze mem disk`.
- `/sys/power/mem_sleep`: `[s2idle]` (no `deep` option).
- `/sys/power/disk`: `[platform] shutdown reboot suspend test_resume`.
- Firmware reports ACPI sleep states `S0 S4 S5`.

### Initramfs and systemd

The base `/etc/mkinitcpio.conf` mentions systemd hooks, but Omarchy's drop-ins override that configuration. The **actual active UKI** was inspected with `lsinitcpio`; it contains `hooks/encrypt`, `hooks/resume`, and `usr/lib/systemd/systemd-hibernate-resume`.

Effective hook configuration:

```text
HOOKS=(base udev plymouth keyboard autodetect microcode modconf kms keymap consolefont block encrypt filesystems fsck btrfs-overlayfs)
HOOKS+=(resume)
MODULES+=(thunderbolt)
```

Runtime hook order reported by `lsinitcpio -a`:

```text
Early: udev
Normal: udev plymouth keymap consolefont encrypt resume
Late: plymouth btrfs-overlayfs
```

Thus the encryption hook runs before the resume hook. The UKI contains the AMDGPU and NVMe modules. No missing-resume-hook or mismatched-offset problem was found.

`systemd-analyze cat-config systemd/sleep.conf` showed no active custom sleep settings. The hibernate service executes `/usr/lib/systemd/systemd-sleep hibernate`. One effective vendor drop-in exists:

```text
/usr/lib/systemd/system/systemd-hibernate.service.d/10-nvidia-no-freeze-session.conf
[Service]
Environment="SYSTEMD_SLEEP_FREEZE_USER_SESSIONS=false"
```

That file is owned by `nvidia-utils 610.57.04-1`, despite this laptop having only an AMD display controller. The same setting affects the suspend service. Systemd warns about it in the log below. Its contribution to this failure is unknown; it has not been changed or tested.

Logind drop-ins set `HandlePowerKey=ignore` and `InhibitDelayMaxSec=15`. The effective configuration retains the default lid-suspend behavior. System-sleep hook files present include `keyboard-backlight`, `unmount-fuse`, and `nvidia`; their presence alone does not establish a cause.

### Exact kernel messages

Source: `journalctl -b -1 -k --no-pager -o short-iso` at diagnosis, boot ID `f9da7de716af40a2aaa14d0d66fe94d8`. Selected complete log lines are preserved below; only the hostname is replaced with `<host>`. Times use UTC-5.

```text
2026-10-02T01:15:07-05:00 <host> kernel: PM: hibernation: hibernation entry
2026-10-02T08:43:21-05:00 <host> kernel: PM: hibernation: Allocated 3372464 pages for snapshot
2026-10-02T08:43:21-05:00 <host> kernel: PM: hibernation: Allocated 13489856 kbytes in 49.52 seconds (272.41 MB/s)
2026-10-02T08:43:21-05:00 <host> kernel: PM: hibernation: Normal pages needed: 4009420 + 1024, available pages: 4146437
2026-10-02T08:43:21-05:00 <host> kernel: ACPI: PM: Waking up from system sleep state S4
2026-10-02T08:43:21-05:00 <host> kernel: nvme nvme0: 16/0/0 default/read/poll queues
2026-10-02T08:43:21-05:00 <host> kernel: mt7921e 0000:62:00.0: Message 00020007 (seq 9) timeout
2026-10-02T08:43:21-05:00 <host> kernel: mt7921e 0000:62:00.0: PM: dpm_run_callback(): pci_pm_restore returns -110
2026-10-02T08:43:21-05:00 <host> kernel: mt7921e 0000:62:00.0: PM: failed to restore async: error -110
2026-10-02T08:43:21-05:00 <host> kernel: amdgpu 0000:63:00.0: MES failed to respond to msg=ADD_QUEUE
2026-10-02T08:43:21-05:00 <host> kernel: amdgpu 0000:63:00.0: failed to map legacy queue
2026-10-02T08:43:21-05:00 <host> kernel: amdgpu 0000:63:00.0: failed to map gfx queue
2026-10-02T08:43:21-05:00 <host> kernel: amdgpu 0000:63:00.0: resume of IP block <gfx_v11_0> failed -110
2026-10-02T08:43:21-05:00 <host> kernel: amdgpu 0000:63:00.0: amdgpu_device_ip_resume failed (-110).
2026-10-02T08:43:21-05:00 <host> kernel: amdgpu 0000:63:00.0: PM: dpm_run_callback(): pci_pm_restore returns -110
2026-10-02T08:43:21-05:00 <host> kernel: amdgpu 0000:63:00.0: PM: failed to restore async: error -110
2026-10-02T08:43:22-05:00 <host> kernel: amdxdna 0000:64:00.1: [drm] *ERROR* PSP is not ready, ret 0xffffff92
2026-10-02T08:43:22-05:00 <host> kernel: amdxdna 0000:64:00.1: [drm] *ERROR* release tmr failed, ret -110
2026-10-02T08:43:23-05:00 <host> kernel: amdxdna 0000:64:00.1: [drm] *ERROR* smu cmd 8 timed out
2026-10-02T08:43:23-05:00 <host> kernel: amdxdna 0000:64:00.1: [drm] *ERROR* Set hard dpm level 0 failed, ret -110
2026-10-02T08:43:23-05:00 <host> kernel: amdgpu 0000:63:00.0: ring sdma0 timeout, signaled seq=614453, emitted seq=614457
2026-10-02T08:43:23-05:00 <host> kernel: amdgpu 0000:63:00.0: Starting sdma0 ring reset
2026-10-02T08:43:23-05:00 <host> kernel: amdgpu 0000:63:00.0: reset sdma queue (0:0:0)
2026-10-02T08:43:23-05:00 <host> kernel: amdgpu 0000:63:00.0: failed to wait on sdma queue reset done
2026-10-02T08:43:23-05:00 <host> kernel: amdgpu 0000:63:00.0: failed to reset legacy queue
2026-10-02T08:43:23-05:00 <host> kernel: amdgpu 0000:63:00.0: Ring sdma0 reset failed
2026-10-02T08:43:23-05:00 <host> kernel: amdgpu 0000:63:00.0: GPU reset begin!. Source:  1
2026-10-02T08:43:24-05:00 <host> kernel: amdxdna 0000:64:00.1: [drm] *ERROR* smu cmd 4 timed out
2026-10-02T08:43:24-05:00 <host> kernel: amdxdna 0000:64:00.1: [drm] *ERROR* Power off failed, ret -110
2026-10-02T08:43:28-05:00 <host> kernel: amdgpu 0000:63:00.0: SMU: No response msg_reg: 31 resp_reg: 0
2026-10-02T08:43:28-05:00 <host> kernel: amdgpu 0000:63:00.0: Failed to power ungate VPE!
2026-10-02T08:43:28-05:00 <host> kernel: amdgpu 0000:63:00.0: [drm] *ERROR* DPM enable vpe failed, ret = -62.
2026-10-02T08:43:28-05:00 <host> kernel: amdgpu 0000:63:00.0: MODE2 reset
2026-10-02T08:43:33-05:00 <host> kernel: amdgpu 0000:63:00.0: SMU: No response msg_reg: 31 resp_reg: 0
2026-10-02T08:43:33-05:00 <host> kernel: amdgpu 0000:63:00.0: Failed to mode2 reset!
2026-10-02T08:43:33-05:00 <host> kernel: amdgpu 0000:63:00.0: Mode2 reset failed!
2026-10-02T08:43:33-05:00 <host> kernel: amdgpu 0000:63:00.0: ASIC reset failed with error, -62 for drm dev, 0000:63:00.0
2026-10-02T08:43:33-05:00 <host> kernel: amdgpu 0000:63:00.0: GPU reset end with ret = -62
2026-10-02T08:43:33-05:00 <host> kernel: amdgpu 0000:63:00.0: GPU Recovery Failed: -62
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: ring sdma0 timeout, signaled seq=614457, emitted seq=614457
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: Starting sdma0 ring reset
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: reset sdma queue (0:0:0)
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: failed to wait on sdma queue reset done
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: failed to reset legacy queue
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: Ring sdma0 reset failed
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: GPU reset begin!. Source:  1
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: [drm] *ERROR* DPM enable vpe failed, ret = -95.
2026-10-02T08:43:35-05:00 <host> kernel: amdgpu 0000:63:00.0: Failed to disallow df cstate
2026-10-02T08:43:51-05:00 <host> kernel: ACPI Error: Aborting method \_SB.A018 due to previous error (AE_AML_LOOP_TIMEOUT) (20260408/psparse-543)
2026-10-02T08:43:51-05:00 <host> kernel: ACPI Error: Aborting method \_SB.ALIB due to previous error (AE_AML_LOOP_TIMEOUT) (20260408/psparse-543)
2026-10-02T08:43:51-05:00 <host> kernel: ACPI Error: Aborting method \_SB.PCI0.GPPC.NHI0.PPS3 due to previous error (AE_AML_LOOP_TIMEOUT) (20260408/psparse-543)
2026-10-02T08:43:51-05:00 <host> kernel: ACPI Error: Aborting method \_SB.PCI0.GPPC.NHI0._PS3 due to previous error (AE_AML_LOOP_TIMEOUT) (20260408/psparse-543)
2026-10-02T08:43:51-05:00 <host> kernel: PM: hibernation: hibernation exit
```

Systemd hibernation service:

```text
2026-10-02T01:15:07-05:00 <host> systemd[1]: Starting System Hibernate...
2026-10-02T01:15:07-05:00 <host> systemd-sleep[2861028]: User sessions remain unfrozen on explicit request ($SYSTEMD_SLEEP_FREEZE_USER_SESSIONS=0).
2026-10-02T01:15:07-05:00 <host> systemd-sleep[2861028]: This is not recommended, and might result in unexpected behavior, particularly
2026-10-02T01:15:07-05:00 <host> systemd-sleep[2861028]: in suspend-then-hibernate operations or setups with encrypted home directories.
2026-10-02T01:15:07-05:00 <host> systemd-sleep[2861028]: Performing sleep operation 'hibernate'...
2026-10-02T08:43:51-05:00 <host> systemd-sleep[2861028]: System returned from sleep operation 'hibernate'.
2026-10-02T08:43:51-05:00 <host> systemd[1]: systemd-hibernate.service: Deactivated successfully.
2026-10-02T08:43:51-05:00 <host> systemd[1]: Finished System Hibernate.
2026-10-02T08:43:51-05:00 <host> systemd[1]: systemd-hibernate.service: Consumed 49.681s CPU time over 1min 37.114s wall clock time, 3.8M memory peak.
```

No NVMe I/O error or timeout was found in the checked incident; its queue initialization appears above. `PM: Image not found (code -22)` appears on the subsequent fresh boot; that message alone does not establish that the earlier restoration failed to locate an image.

### Duplicate check and related reports

On 2026-10-02, the issue tracker was searched across open and closed reports using `hibernate`, `hibernation`, `UM3406KA`, `Zenbook`, `Krackan`, `ADD_QUEUE`, `gfx_v11_0`, and the exact phrase `GPU Recovery Failed`. Relevant issue bodies and selected comments were reviewed. **No exact duplicate of this Zenbook S4/MES ADD_QUEUE/failed-restore sequence was identified in those searches.** This does not prove the underlying driver bug is unique.

- [#4908](https://github.com/omacom/omarchy/issues/4908) (closed): same UM3406KA model; failure to enter sleep after an older `linux-firmware-other` update. Different phase and firmware versions. The maintainer requested upstream reporting of firmware issues.
- [#10924](https://github.com/omacom/omarchy/issues/10924) (open): same model, but I2C touchpad restoration after s2idle, rather than the AMDGPU S4 failure here.
- [#11208](https://github.com/omacom/omarchy/issues/11208) (open): Ryzen AI 7 350 / Radeon 860M; lock-screen input after s2idle on stock kernel 7.2.3, with a different log signature.
- [#13395](https://github.com/omacom/omarchy/issues/13395) (open): same `linux-omarchy 7.2.5-3` and AMD firmware package version, but an RDNA2 discrete GPU fails its pre-restore freeze callback (`pci_pm_freeze`, mode1 reset). Here the logged failure is the restored GPU's `pci_pm_restore`/MES path and mode2 recovery.
- [#4184](https://github.com/omacom/omarchy/issues/4184) and [its AMD/MES comment](https://github.com/omacom/omarchy/issues/4184#issuecomment-3820100143): broader AMD suspend/hibernate problems on different hardware; a possible related driver family of failures, not a confirmed duplicate.
- [#4372](https://github.com/omacom/omarchy/issues/4372): MES failure on HX370/890M under Chromium (`MISC`, not `ADD_QUEUE` at S4 restore).
- [#11413](https://github.com/omacom/omarchy/issues/11413): AMD Lucienne hibernate freezes, different hardware and failure signature.
- [#13119](https://github.com/omacom/omarchy/issues/13119): same Omarchy kernel version, but an MST external-monitor modeset regression on different GPU hardware.

### Current conclusion and verification limits

The strongest observed failure is in AMDGPU restoration after waking from S4, followed by unsuccessful GPU recovery. Wi-Fi, NPU, and ACPI errors suggest the platform/device power transition also warrants investigation. The origin has not been isolated to Omarchy, upstream Linux, firmware, or BIOS.

All diagnostic operations were read-only. No kernel or firmware update/downgrade, boot-parameter edit, service override, initramfs rebuild, driver unload, suspend/hibernate test, or BIOS change was performed. The alternative installed kernel and workarounds mentioned in related issues have not been tested on this machine.

No full Omarchy debug bundle or screenshot was collected or uploaded; the available evidence is the targeted inspection and the exact journal excerpts above. The observed symptom, reproduction context, and appropriate upstream-versus-Omarchy tracker remain to be confirmed. No tested fix is claimed.