The grey box said the filesystem was broken. The kernel had already named a flag.
The Situation
Two external drives sit on this desk. One is a 931G exfat volume. The other is a 1.8T NTFS volume, the one the working trees live on. Both carry the volume label SSD. After a restart, the small one was already mounted. The big one was not. Clicking it in the file manager produced the same dialog for the third time:
Error mounting /dev/sdb1 at /run/media/b0gie/SSD1: wrong fs type, bad option,
bad superblock on /dev/sdb1, missing codepage or helper program, or other error.
The title of the window was āUnable to access āSSDāā. That title is doing a lot of lying in a very small space. The path in the body is SSD1, not SSD. The phrase ābad superblockā is the part that makes you reach for a cable, a new filesystem, or a recovery tool. None of those was the next step.
We expected a restart to bring the drive back the way the other one came back. What we got was a generic mount failure, and a drive that was fine.
Act I: The Label Collision
lsblk made the dialogās title make sense, and also made it useless.
| device | size | type | label | mount |
|---|---|---|---|---|
| the small disk | 931.5G | exfat | SSD | /run/media/b0gie/SSD |
| the archive disk | 1.8T | ntfs | SSD | not mounted |
blkid on the unmounted partition agreed: TYPE="ntfs", LABEL="SSD". The file manager names mounts from the label. Two volumes with the same label become SSD and SSD1. The suffix is a collision, not a second device, and it is not stable across boots. Device nodes swap too. sda this boot can be sdb the next one. The dialog quoted /dev/sdb1 because that is what the node was called in that minute. It is not the identity of the drive.
The working exfat volume was already at /run/media/b0gie/SSD. The failure was the 1.8T NTFS volume, which the file manager was trying to place at /run/media/b0gie/SSD1. Opening āSSDā in the sidebar and reading a path that ends in SSD1 is how you spend ten minutes staring at the drive that already works.
Identify the archive by size and filesystem type. 1.8T, ntfs. If you need a stable name for a later fstab line, take the UUID from blkid. Do not take sdb.
Act II: The Kernel Had Already Answered
The dialog is what userspace prints when a mount helper gives up. It is a bucket. Wrong type, bad option, bad superblock, missing codepage, missing helper, or other error. Six causes, one sentence. The kernel log had picked one.
ntfs3(sdb1): volume is dirty and "force" flag is not set!
ntfs3(sdb1): It is recommended to use chkdsk.
That pair showed up twice in the same boot, about fourteen minutes apart. Same refusal, same flag. The in-kernel NTFS driver is ntfs3. It was loaded. It saw the partition. It refused to mount because the volumeās dirty bit was set and the mount options did not include force.
chkdsk is the Windows tool the driver is pointing at. There is no Windows install left to boot for that. Following the recommendation would have meant finding a machine that still speaks that language, or treating the line as a repair order we could not carry out. We did not. The line that mattered was the one above it.
At the dialog, command -v ntfs-3g returned nothing. The canned sentence includes āmissing helperā, and that night the helper was not on PATH. That is a side clause, not the cause. The kernel had already refused the volume for a dirty bit. Before this post went up, the package was on the box (ntfs-3g 2026.9.28-1, installed 2026-11-11 10:24 UTC, binary at /usr/bin/ntfs-3g). Installing it does not make the file manager pass force. The superblock was never the thing that failed.
Act III: The File Manager Will Not Pass the Flag
The mount that works is not mysterious:
sudo mkdir -p /run/media/b0gie/SSD1
sudo mount -t ntfs3 -o rw,force,uid=1000,gid=1000 /dev/sdb1 /run/media/b0gie/SSD1
uid and gid matter. Without them the tree comes back owned by root, and you have traded a dirty-volume refusal for a permission wall. force is the option the kernel asked for by name.
The file manager will not send it. Its mount helper rejects force as an option it is not allowed to pass. So every click replays the same dialog, and every reboot that leaves the volume dirty replays the click. That is the whole loop. Unclean restart, dirty bit, helper refuses the flag, grey box, ābad superblockā.
We tried to stop the loop with a boot hook: a small script, a systemd oneshot, and a udev rule keyed on the filesystem UUID. That write was denied. Good. A udev rule plus a helper plus a unit is three places to be wrong about a drive whose device node already moves. The one-shot mount was the fix for this boot.
Act IV: What āMountedā Actually Looked Like
The mount returned 0. findmnt on the UUID showed:
/dev/sdb1 ntfs3 rw,relatime,uid=1000,gid=1000,force,acl,iocharset=utf8,prealloc
A touch in the mountpoint succeeded, and the probe file was removed. The top of the tree listed. That is the check. A successful mount command is not a writable filesystem. A file you created and deleted is.
Later, the same UUID was still mounted read-write, and the option string had changed:
/dev/sdb1 ntfs3 rw,nosuid,nodev,relatime,uid=1000,gid=1000,acl,iocharset=utf8,prealloc
force was no longer in the string. nosuid and nodev had appeared. We did not run a second mount. We did not run ntfsfix. We did not read the volume flags with ntfsinfo. So this post will not claim the dirty bit was cleared, and it will not claim a repair. What we can say is smaller and true: after the force mount, the volume stayed mounted read-write, and the live options no longer needed to advertise force for this boot.
That is enough to work. It is not a certificate that the index is clean. If a later reboot refuses the volume again, the kernel log is the first place to look, not the dialog.
Receipts
| check | result |
|---|---|
| file manager dialog | wrong fs type, bad option, bad superblock on /dev/sdb1 |
blkid on the unmounted partition |
TYPE="ntfs" LABEL="SSD" |
| kernel, twice this boot | volume is dirty and "force" flag is not set! |
ntfs-3g at the dialog |
not on PATH |
ntfs-3g before publish |
installed 2026-11-11 10:24 UTC, /usr/bin/ntfs-3g |
ntfs3 module |
present |
| mount | ntfs3 rw,force,uid=1000,gid=1000, exit 0 |
| write probe | created and removed |
later findmnt |
still rw, force no longer in the option string |
The 931G exfat volume, also labeled SSD, was mounted the whole time at /run/media/b0gie/SSD. Nothing in this session touched it.
Takeaway
When a desktop says ābad superblockā, read the kernel log before you believe the noun. On an NTFS volume the useful line is often volume is dirty and "force" flag is not set. The file manager will not pass that flag. mount -t ntfs3 -o rw,force,uid=1000,gid=1000 will. Prove the mount with a write, not with the commandās exit code.
Do not add a udev rule to make the next boot quiet. If a later reboot refuses the same volume, the persistent form is one fstab line, by filesystem UUID, with nofail. Not before that refusal. And do not treat the driverās chkdsk hint as a Linux command. Quote it, then decide whether you actually have a place to run it.
What to Run
# which disk is the unmounted NTFS, and what did the kernel actually say
lsblk -o NAME,SIZE,FSTYPE,LABEL,UUID,MOUNTPOINT
journalctl -k --no-pager -g 'ntfs3'
# the mount that worked. use the node lsblk just printed, and your uid/gid.
sudo mkdir -p /run/media/$USER/SSD1
sudo mount -t ntfs3 -o rw,force,uid=$(id -u),gid=$(id -g) /dev/sdX1 /run/media/$USER/SSD1
# prove it
findmnt -no SOURCE,FSTYPE,OPTIONS,TARGET /run/media/$USER/SSD1
touch /run/media/$USER/SSD1/.mount-probe && rm /run/media/$USER/SSD1/.mount-probe
