My DATA_HDD (/dev/sda2, NTFS) refused a read-write mount. It mounted read-only, so I backed everything up to another drive first and only then cleared the dirty bit. Along the way I also restored ~4 GB of Kali files (CTF tooling, docs, media) from an old Timeshift snapshot into ~/kali-restore-tmp/.
journalctl: ntfs3(sda2): volume is dirty and "force" flag is not set!
It is recommended to use chkdsk.Windows Fast Startup / hibernation / an unclean shutdown left the dirty bit set, and the Linux ntfs3 driver blocks RW mounts to protect your data. No I/O errors — the sibling partitions on the same disk mounted fine, so this was purely the dirty flag.
udisksctl mount -b /dev/sda2 --options ro
df -h /run/media/$USER/DATA_HDD # 199G, 78G used
ls /run/media/$USER/DATA_HDDTarget: a folder on the second drive (GAMES_HDD/DATA_HDD_backup/). Lesson learned: on one spinning HDD, rsync -avh with per-file logging was too slow (aborted at 16G) and tar re-reads everything on resume (killed at 52G). What worked was quiet cp, one copy at a time — parallel copies on the same spindle just halve the speed:
mkdir -p "/run/media/$USER/GAMES_HDD/DATA_HDD_backup"
nohup cp -a "/run/media/$USER/DATA_HDD/Redmi-Note-10-Pro" \
"/run/media/$USER/GAMES_HDD/DATA_HDD_backup/" > /tmp/redmi.log 2>&1 &
# resume-friendly variant for big trees (skip existing without re-reading):
nohup cp -aun "/run/media/$USER/DATA_HDD/CACHYOS_MIGRATION" \
"/run/media/$USER/GAMES_HDD/DATA_HDD_backup/" > /tmp/cachyos.log 2>&1 &cp -aun = archive + update + no-clobber: skips existing files without re-reading their data, the fastest resume on a spinning disk.
udisksctl unmount -b /dev/sda2
sudo ntfsfix /dev/sda2
udisksctl mount -b /dev/sda2
# best long-term: boot Windows, run `chkdsk E: /f`, and disable Fast Startupntfsfix lives at /usr/bin/ntfsfix; ntfs-3g wasn't installed, the kernel used ntfs3.root:root 700 by design. Read via sudo, or grant yourself access with sudo setfacl -R -m u:$USER:rX ....