Samsung PM9B1 NVMe I/O timeouts and brief system stalls on MSI Modern 15H

[Hogwarts-coder10]

Samsung PM9B1 NVMe I/O timeouts and brief system stalls on MSI Modern 15H

Problem

My MSI Modern 15H occasionally freezes/stalls for a few seconds and then resumes normally.

The kernel repeatedly reports NVMe I/O timeouts:

nvme nvme0: I/O tag 3 (1003) QID 5 timeout, completion polled
nvme nvme0: I/O tag 273 (7111) QID 3 timeout, completion polled

The issue occurs during normal desktop usage.

Hardware

• MSI Modern 15H
• Intel Core Ultra 7 155H
• 16 GB RAM
• Samsung MZVL41T0HBLB-00BTW (PM9B1, DRAM-less)
• 1 TB NVMe
• Firmware: HXC70W1Q

NVMe health

SMART reports:

critical_warning: 0
available_spare: 100%
percentage_used: 2%
media_errors: 0
num_err_log_entries: 0
temperature: 50 C

The NVMe error log contains no recorded errors.

PCIe

The NVMe is connected at:

PCIe 4.0 x4
16 GT/s

AER reports no Correctable, Non-Fatal, or Fatal errors.

Kernel testing

The issue occurs on both:

• Arch kernel 7.1.8-arch1-3
• Arch LTS

I also tested with:

nvme_core.default_ps_max_latency_us=0

This successfully disabled APST:

Autonomous Power State Transition Enable (APSTE): Disabled

However, the NVMe I/O timeouts continued.

Firmware / ACPI messages

The kernel reports:

ACPI FADT declares the system doesn't support PCIe ASPM, so disable it

and:

pci ...: can't override BIOS ASPM; OS doesn't have ASPM control

The NVMe controller also reports:

nvme nvme0: D3 entry latency set to 8 seconds

Storage layout

The system uses:

NVMe -> LUKS2 -> Btrfs

lsblk --discard shows:

nvme0n1    DISC-GRAN 512B
nvme0n1p2  DISC-GRAN 512B
root       DISC-GRAN 0B

fstrim.timer is currently disabled.

Reproduction on Arch LTS

The issue is also reproducible on the Arch LTS kernel.

During a single LTS boot, the following NVMe timeouts occurred:

Aug 23 17:22:07 omarchy kernel: nvme nvme0: I/O tag 3 (1003) QID 5 timeout, completion polled
Aug 23 17:22:45 omarchy kernel: nvme nvme0: I/O tag 273 (7111) QID 3 timeout, completion polled
Aug 23 17:38:27 omarchy kernel: nvme nvme0: I/O tag 7 (b007) QID 6 timeout, completion polled

Question

Could this be an interaction between the Samsung PM9B1, Intel PCIe root port, MSI firmware/ACPI, and the Linux/Omarchy storage configuration?

I can provide the full dmesg, lspci -vv, SMART output, and NVMe error log if needed.

[Hogwarts-coder10]

Aug 23 17:46:38 omarchy kernel: nvme nvme0: I/O tag 128 (2080) QID 2 timeout, completion polled
Aug 23 17:47:35 omarchy kernel: nvme nvme0: I/O tag 256 (c100) QID 11 timeout, completion polled

these were on Linux Kernel 7.1.8-arch1-3

[Hogwarts-coder10]

Additional NVMe diagnostics

NVMe:

• Model: SAMSUNG MZVL41T0HBLB-00BTW (PM9B1 DRAM-less)
• Firmware: HXC70W1Q
• Kernel: 7.1.8-arch1-3
• Temperature: 43 °C
• Percentage used: 2%
• Media errors: 0
• Critical warning: 0

NVMe error log:

• No recorded NVMe errors; error entries report successful completion.

PCIe device:

• Link: 16 GT/s x4
• ASPM: Disabled
• AER: No uncorrectable/correctable errors
• HeaderLog: all zero

PCIe root port:

• Intel Core Ultra 200H/200V Series PCIe Root Port Apps 2x scaled with GDK_SCALE set to 1 #11
• Link: 16 GT/s x4
• ASPM: Disabled
• AER: No reported errors

The timeout still occurs despite APST being disabled:
nvme_core.default_ps_max_latency_us=0

The issue also reproduces on the LTS kernel.

It's a combination of PM9B1 + firmware HXC70W1Q + Intel Core Ultra 200H/200V PCIe root port + repeated host-side NVMe timeouts + clean NVMe error log/AER.

[Hogwarts-coder10]

Additional timeout

Aug 24 11:07:06 omarchy kernel: nvme nvme0: I/O tag 198 (00c6) QID 3 timeout, completion polled