Dear Maintainer,

After upgrading to kernel 7.0.7+deb14-amd64, the system becomes unstable and
eventually panics due to NVMe controller reset failures and filesystem I/O
errors.

System information:

* Machine: HP 250 G10 Notebook PC
* BIOS: F.24 11/20/2025
* Debian testing amd64
* Previous kernel 7.0.4+deb14-amd64 works correctly on the same hardware

What led to the situation:
After booting into kernel 7.0.7+deb14-amd64 and using the system normally, the
NVMe controller resets and the root filesystem becomes read-only. Shortly after
that, systemd crashes and the kernel panics.

What I did:

* Reinstalled linux-image-7.0.7+deb14-amd64
* SMART self-tests were executed and completed without error
* No NVMe media/data integrity errors are reported
* Booting back into 7.0.4 restores stability

Observed result:
The following messages appear before the panic:

nvme nvme0: Disabling device after reset failure: -19

I/O error, dev nvme0n1

EXT4-fs: Journal has aborted

Kernel panic - not syncing: Attempted to kill init!

The filesystem is remounted read-only and the system becomes unusable.

Expected result:
The system should remain stable under kernel 7.0.7, as it does under 7.0.4.

Additional information:
This may be related to NVMe APST/power-management handling. I am testing the
following kernel parameter as a workaround:

nvme_core.default_ps_max_latency_us=0

The issue appears reproducible on 7.0.7 and not reproducible on 7.0.4.

Regards,
NH Riyad

Hi,

Sorry for the delay in coming back to you. Can you please confirm if
the issue istill triggerable on 7.0.10-1 as in testing?

If youes, can you:

- Test 7.0.11 upstream? Does this expose the problem as well
- Provide fresh logs including the problem.

Given 7.0.4 was fine but 7.0.7 had the problem in case you can
reproduce the problem with recent kernel still, would you be able to
bisect the issue and we can get the breaking change?

Regards,
Salvatore

Hi

Thanks for the followup up.

In this case let's close the bug for now with 7.0.9-1. But as said if
you hit the problem again, we can proceed as per above.

Regards,
Salvatore