Re: "controller is down; will reset" on SK Hynix NVMe drive in Lenovo IdeaPad Pro 5

[Maria Ioannidou <mi@eworx.gr>]

I can confirm this issue on identical hardware, in case a second
report helps move the quirk forward.

Hardware:
   Laptop: Lenovo IdeaPad Pro 5 16APH8 (MTM 83AR001MGM)
   BIOS:   MDCN33WW (2024-01-24)
   Drive:  SKHynix_HFS001TEJ4X112N
   PCI ID: 1c5c:1d59

Kernel: 7.0.0-30-generic (Ubuntu 26.04)

Symptoms match the original report. After some idle time, writes
begin failing with I/O errors, ext4 remounts read-only, and the
system becomes unusable until reboot:

   systemd-journald: Failed to write entry to
   /var/log/journal/.../system.journal ... Input/output error

The failure is not confined to any userspace component. It
reproduces with no graphical session running (multi-user.target),
with the NVIDIA driver stack fully removed, and on a completely
fresh install of Ubuntu 26.04 onto a wiped disk - in that case
within roughly four minutes of first boot.

Diagnostics that made this misleading:
   - SMART self-assessment PASSED, Critical Warning 0x00,
     Available Spare 100%, Percentage Used 1%, Media and Data
     Integrity Errors 0
   - Sequential write throughput is normal (2.7 GB/s) when working
   - fsck from external boot media always completes clean, since a
     live USB barely touches the internal drive

Since a clean reinstall reproduces the fault, this initially
looked like failing hardware, and I was close to filing a warranty
claim before testing the APST hypothesis.

Booting with nvme_core.default_ps_max_latency_us=0 resolves it
completely. Stable for over 24 hours now, including several
reboots, suspend/resume cycles and repeated 4 GiB direct writes,
with no errors in dmesg.

I have not needed pcie_aspm=off or pcie_port_pm=off; disabling
APST alone has been sufficient here.

Happy to test a patch or provide further logs if useful.

-- 
Maria Ioannidou
Eworx Technical Team
mi@eworx.gr