Question

Some solution could probably be applicable to Windows, however I am not familiar with the Windows OS, so this will be Linux focused.

As far as I understand, Unix file system all have the concept of inodes, which is where the file system metadata and the “file” is stored. Thus I am wondering if it is possible to use the inode number with some additional information to track files that are renamed or moved around?

What I was proposing to do was have an initial scan that would create a database of filename/path, their disk/drive that it is located on, their inode number, and finally some sort of checksum (sha-1).

This would enable the system to be able to use the inode number to quickly detect if a file got moved or renamed, then it would follow up with the checksum to check to see if its actually the same file.

I can see some possible problems with this scheme:

1. Files could be modified then moved/renamed, and this would fail to detect it because the checksum would not match.
2. Some (most?) application when they modify the file will create a new temporary file and then switch it with the current one, thus the inode wouldn’t match anyway even if the file is in the end unmodified….
3. Would need to store which device/file system it is on because inodes on each file system is unique to that file system.
4. Would need to deal with hard-links

I am wondering if there is any other gotchas that I am forgetting about here? I was hoping to be able to use the inodes to quickly track down which files got moved or renamed, then follow up with a checksum to confirm that it is actually the same file.

Answer 1

Look into inotify which allows you to get a call when anybody does anything to files in a specified file system or directory.