ZimaOS 1.6.1 regression: shutdown leaves hardware unable to POST

[retocoder]

Summary

After upgrading from ZimaOS 1.5.4 to ZimaOS 1.6.1, my server started having a reproducible shutdown / next-power-on issue.

The OS appears to complete shutdown normally, but after the system powers off, the next manual power-on often fails before POST/BIOS. There is no BIOS logo, no video signal, and BIOS setup cannot be entered from the keyboard. Recovery requires a full AC power drain.

After rolling back to ZimaOS 1.5.4, the issue disappeared immediately on the same hardware configuration, including the NVIDIA RTX GPU installed.

Hardware

Motherboard: Gigabyte H610I DDR4
BIOS: F32, 2024-09-27
CPU: Intel Core i5-12400
RAM: 32 GB
GPU: NVIDIA RTX 2000 Ada 16 GB
NVMe: Transcend TS512GMTE220S
Storage: 2x WD 4TB HDD in RAID1

Affected version

ZimaOS: v1.6.1
Kernel: Linux 6.12.25
RAUC slot during issue: B

Kernel command line included:

intel_iommu=on
vfio_iommu_type1.allow_unsafe_interrupts=1
thunderbolt.host_reset=false
rauc.slot=B

Working version

ZimaOS: v1.5.4

After rollback to ZimaOS 1.5.4, shutdown and reboot work normally again.

Symptoms on ZimaOS 1.6.1

After running either shutdown or reboot from ZimaOS 1.6.1:

- systemctl poweroff completes
- system powers off
- next manual power-on does not reach POST/BIOS
- no BIOS logo
- no video signal
- BIOS setup cannot be entered from keyboard
- HDD LED does not reach normal boot activity
- sometimes the system enters power cycle
- sometimes it hangs before POST without power cycling
- full AC power drain is required to recover

This happened multiple times and was reproducible.

Important observations

1. Shutdown logs look clean

The previous boot journal shows a normal shutdown sequence. Docker, containerd, ZimaOS services, NetworkManager, auditd, and other system services stop normally. There is no visible kernel panic or oops before poweroff.

This suggests the issue happens after Linux userspace shutdown, during platform/firmware/power-state transition or next hardware initialization.

2. ACPI wake sources affected self-power-on, but did not fix the root issue

Initially, the following ACPI wake sources were enabled:

PEG1  - PCIe x16 root port / NVIDIA path
RP05  - PCIe root port with NVMe
GLAN  - Ethernet
XHCI  - USB controller

After disabling these wake sources from /proc/acpi/wakeup, the system no longer powered itself back on after shutdown.

However, the next manual power-on still failed before POST/BIOS and required full AC power drain.

3. NVIDIA initially looked suspicious, but was not the root cause

The system has an NVIDIA RTX 2000 Ada GPU. On ZimaOS 1.6.1, the NVIDIA driver was loaded:

NVIDIA driver: 580.105.08
GPU: NVIDIA RTX 2000 Ada
Persistence Mode: initially On
Runtime D3 status: Disabled by default
Video Memory: Active

Disabling NVIDIA persistence mode changed the symptom:

Before disabling persistence mode:
- next power-on could enter repeating power cycle

After disabling persistence mode:
- no repeating power cycle
- but system still hung before POST
- no video signal / no BIOS logo
- BIOS setup still not reachable

The NVIDIA state and driver details were captured in the diagnostic archive.

4. Issue also reproduced without the NVIDIA GPU installed

To isolate the GPU, I removed the RTX 2000 Ada card and changed BIOS display initialization from:

Initial Display Output = PCIe 1 Slot

to:

Initial Display Output = IGFX

The system booted without the RTX GPU installed.

However, after shutdown from ZimaOS 1.6.1, the same next-power-on problem still reproduced even without the NVIDIA GPU.

Therefore, the issue is not limited to the NVIDIA GPU. The GPU may affect or amplify the symptoms, but it does not appear to be the root cause.

5. Rollback to ZimaOS 1.5.4 fixed the issue immediately

After rolling back from ZimaOS 1.6.1 to ZimaOS 1.5.4, the problem disappeared immediately.

With the same hardware configuration, including the NVIDIA RTX 2000 Ada installed:

• shutdown is faster again
• reboot works normally
• system no longer powers itself back on
• system no longer enters power cycle
• system no longer hangs before POST
• no AC power drain is required after shutdown

This strongly suggests a regression in ZimaOS 1.6.1, likely related to kernel 6.12.25 / ACPI / S5 shutdown / platform power management on Gigabyte H610I DDR4.

Steps to reproduce

1. Boot ZimaOS 1.6.1 on Gigabyte H610I DDR4.
2. Let the system boot normally.
3. Run:
   sync
   systemctl poweroff
4. Wait until the system powers off.
5. Press the physical power button.
6. Observe that the machine does not reach POST/BIOS.

Observed result:

- no BIOS logo
- no video signal
- BIOS setup not reachable
- sometimes power cycle
- sometimes pre-POST hang
- full AC power drain required to recover

Expected result:

The system should boot normally after being powered on again.

Workaround

Rollback to:

ZimaOS 1.5.4

After rollback, the issue is no longer reproducible.

Diagnostic archive

I collected a diagnostic archive while the system was affected:

zima-report-20260529-175835.tar.gz

The archive contains:

- system version
- kernel version and command line
- hardware information
- storage layout
- PCI/USB inventory
- ACPI wakeup state
- journal logs from failed shutdown tests
- dmesg
- lspci -vvv
- NVIDIA state
- reproduction notes

Requested help

Please investigate whether ZimaOS 1.6.1 / kernel 6.12.25 introduced a regression in shutdown, ACPI S5, PCIe power management, or platform power-state handling on Gigabyte H610I DDR4 / Intel Alder Lake systems.

The same hardware works normally again after rollback to ZimaOS 1.5.4.

zima-report-20260529-175835.tar.gz

I don't know exactly if it's an issue in 1.6.1 itself or bad update from 1.5.4 to 1.6.1. Anyway nobody updates the system by reinstalling it from scratch. Number one feature in any NAS system should be stability in basics, so please help.