How to flush legacy ddt?

[shankerwangmiao]

I accidentally enabled dedup, setting dedup=on on some of my datasets in a zpool. I disabled dedup after about a hour. I want to flush the legacy ddt entries. I thus used find filtering files whose ctime fell within the time range, copied these files out, removed these files, and copied these files back. By doing so, I can successfully saw the decreasement of ddt entries using zdb -DDDD. However, after doing all these routine, there still around ~1k entries left, all the refcounts of which are 1. I used https://github.com/mrighele/zfs-find-deduped/blob/master/zfs_find_deduped.py, which first lists all the blocks in the ddt using zdb -DDDDD, and then lists all the blocks used by a filesystem using zdb -ddddd, and finds the files using the blocks in the ddt. The tool also cannot find any file that refers the blocks in the ddt. So is there any way to find out which files are referring the ddt entries, and is there any way to flush the legacy ddt?

System: Debian Bookworm; ZFS version: 2.2.2-4~bpo12+1