Could someone explain to me why we need a special NVM interface vs just using a filesystem on the same medium?
I worked at an embedded system shop for 3 years and this was one of the things that irked me when I worked on code outside of what my team usually worked on. We had a filesystem, accessible with all of the standard C file API... but there was still a lot of stuff being put into this mystical NVM (usually user controlled settings) that then made things like hard resets hairy because even as a developer I wasn't sure what would be wiped when I flashed different software onto the device.
This may have just been bad practice, or a result of carrying over the idea of NVM after filesystems became available. But I don't know why anyone would try to propagate that practice these days.
I don't really understand your question here. NVM is non-volatile memory. Stick a filesystem on it and it's still non-volatile memory.
Do you mean they were writing configuration files directly to a known address in Flash instead of using a filesystem? It's much simpler and less resource-intensive and doesn't require writing and maintaining a driver. There are plenty of cases where direct-writes are better, and plenty of cases where a filesystem is better, and a nice big gray area in between.
I worked at an embedded system shop for 3 years and this was one of the things that irked me when I worked on code outside of what my team usually worked on. We had a filesystem, accessible with all of the standard C file API... but there was still a lot of stuff being put into this mystical NVM (usually user controlled settings) that then made things like hard resets hairy because even as a developer I wasn't sure what would be wiped when I flashed different software onto the device.
This may have just been bad practice, or a result of carrying over the idea of NVM after filesystems became available. But I don't know why anyone would try to propagate that practice these days.