Question: may running sdmon harm the storage subsystem?

[Eugeny1]

This is a question, not bug report or issue report. At least for now.

I assume the one(s) who designed the tool worked out its side effects, and my question is: may running this tool (CMD56 and related commands) cause any harm to the system because of e.g. interference with standard SD-card driver workflow?

One time I had volume going read-only after running the tool. I do not have logs saved, and unable to see what the actual cause for the issue was, but if monitoring the card regularly the volume going read-only (or failing in any other way) is unacceptable.

Thanks!

[Ognian]

Hi,
the tool "should" not cause any harm to the filesystem, BUT the sd card driver (in the linux kernel) is the one who is responsible for that (not this tool). Before making this tool public I had problems with the raspberry pi linux kernel sd card driver (the whole filesystem was gone). Actually only day's after posting the problem on a kernel list, the responsible developer asked about how to reproduce it, and fixed the driver... this is the story of how this repo got public... so actually on a raspberry pi it should be quite save as long as there are no big changes in the sd card driver...
Hope this helps
Ognian

[Eugeny1]

Thank you. I think I have seen that issue report, and solution. Is it this one? It was back in 2018, and hopefully there is a fix in my linux, which is openWRT.

[Eugeny1]

IMHO your tool must be a part of standard Linux. You'd also add support for MMC health checking. Looking into the sources I'd say the EXT_CSD for SD and MMC are being only read on their initialization, but that will not help if system is expected to run for years without reboot / failure. Thus, to have the analytics it should be required to have up-to-date data from the EXT_CSD. Thanks.

[Ognian]

yes I think the issue was that one ... since the tool is licensed as GPL it can be part of linux, but that is not my decision...
I think someone already added it to FreeBSD as a package, so this is also a step in that direction...

[Eugeny1]

pYYBAGNVUESAWgCnABxrTf4tSGM159.pdf
Page 37, regarding CMD62/63.

Edit: small comment on Sandisk requiring signing NDA for implementing the health info read. This is nonsense for simple reason. If one is unable to identify the card's health and considers his/her application important, this requirement, and thus lack of monitoring, leads to customer churn. I have already made a decision not to use Sandisk cards, and use Kingston instead, simply because of ability replacing "fully industrial" card with "less industrial" one if I need. This is just a wrong business decision.