Windows - NVMe Self-Test (kicking and aborting) not working properly

[csm10495]

Hey folks,

I'm having some trouble kicking device self-tests on NVMe devices using smartctl. It seems like it 'thinks' it starts the test, but then doesn't successfully do it.

If I use Samsung Magician to kick the tests, they work fine and report properly in smartctl. Aborting the tests also seem to fail.

Does it check if a test already ran in the given hour before kicking?

Attaching a video showing the behavior I'm seeing.

Info:
Smartctl: smartctl pre-7.4 2023-03-21 r5470 [x86_64-w64-mingw32-w10-22H2] (CircleCI)
Windows 10 Pro: 10.0.19045.2364 x64
Driver: <Builtin Windows 10 NVMe Driver: 'Standard NVM Express Controller'>
Drive: Samsung SSD 980 PRO 1TB, FW: 5B2QGXA7

Happy to provide other info, etc. Thanks!

[csm10495]

video showing the issue and comparing with magician

[Christian Franke]

Starting self-tests works for me with:
Smartctl: same CI build
Windows 10 Pro: 10.0.19045.2728 x64
Driver: same (stornvme.sys 10.0.19041.2075)
Drive: Samsung SSD 970 EVO Plus 500GB, FW: 2B2QEXM7

There actually is one strange behavior: Any self-test command is apparently silently ignored by the drive(?) for ~10 minutes after the last test finished. Not yet tested whether this could be reproduced under Linux.

I stopped using Samsung Magician because random garbage appeared in the SMART Error log of a Samsung 860 EVO 4TB SATA SSD after this tool has been run. Does not happen with smartctl.

Note that Samsung Magician might use vendor specific NVMe commands to start the tests. Smartctl does not (see ticket #1067), this might make a difference.

PS: Please avoid large attachments like videos in future submissions.