About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
When umount /mnt/data prints umount: /mnt/data: target is busy., something still uses the filesystem: a shell sitting in a directory on it, a process with a file open, another filesystem mounted below it, or a swap file. The umount manual says a filesystem "cannot be unmounted when it is 'busy'". Name the holder first, release it by its own proper means, then run a plain umount again.
What the error means
The words of the message come from the program's source code, not from a manual page. In util-linux v2.42.4, libmount/src/context_umount.c turns the kernel's EBUSY result into the text target is busy, and sys-utils/umount.c prints it with the format "%s: %s.". The first %s is the name you gave umount (for example /mnt/data) and the second is the message, so the line reads umount: /mnt/data: target is busy. The umount: prefix is added by the program name; only the target name varies.
The umount manual describes what counts as busy: "Note that a filesystem cannot be unmounted when it is 'busy' - for example, when there are open files on it, or when some process has its working directory there, or when a swap file on it is in use." It adds: "The offending process could even be umount itself - it opens libc, and libc in its turn may open for example locale files."
So the error is not a fault in the disk. It is the system refusing to detach a filesystem that is in use. The job is to find the user of the filesystem, not to force the detach.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| A shell or process has its working directory on the mount, including your own shell | Prompt shows a path under the mount | fuser -vm lists the process with the access letter c (current directory) |
| A service or job has files open on the mount | Application, backup or report job running | lsof on the mount point lists the files and the owning process |
| A program runs from the mount or maps a library from it | A process whose command is a path under the mount | fuser -vm shows access letter e (executable) or m (mapped file) |
| Another filesystem is mounted below it | Extra mount lines under the mount point | findmnt -R prints all submounts |
| A swap file on the mount is in use | Swap shown under the mount path | swapon --show lists a file under the mount |
| A network share whose server is unreachable | umount hangs |
The server does not answer; the umount manual ties -f to this case |
| The holder is invisible to your tools | fuser and lsof list nothing, yet the mount stays busy |
Run both as root. fuser "may only be able to gather partial information unless run with privileges" and "will not see block devices mounted by processes in a different mount namespace" |
Prove the cause before changing anything
All of these are read-only. Run them as root; the fuser manual warns that results are partial without privileges.
1. Record the mount as it is now. findmnt is described as finding a filesystem, and its --target option "accepts any file or directory and then findmnt displays the filesystem for the given path":
findmnt -o TARGET,SOURCE,FSTYPE,OPTIONS --target /mnt/data
Keep the output. The findmnt manual says its default output "is subject to change", so the command names the columns: mount point, source device, filesystem type and mount options, which you need if you must remount. The -o option defines the output columns; the column names are in the util-linux source at tag v2.42.4.
2. List nested mounts. The findmnt manual says -R will "Print recursively all submounts for the selected filesystems", and -n will "Do not print a header line":
findmnt -nR /mnt/data
Without the header, one line is the mount itself, and more than one line means other filesystems are mounted below the one you want to detach.
3. Name the processes using the filesystem. In fuser, -v gives a ps-style listing and -m treats the name as a mounted filesystem:
fuser -vm /mnt/data
The access column uses the letters in the fuser manual: c current directory, e executable being run, f open file, F open file for writing, r root directory, m mapped file or shared library. A shell with c is someone standing in the directory.
4. List the open files. The lsof manual says that when the name is a mounted-on directory, "lsof will list all the files open on the file system":
lsof /mnt/data
5. Look for swap on the mount.
swapon --show
Any path under /mnt/data in the list is a swap file on that filesystem.
If steps 3 and 4 both list nothing but the mount is still busy, suspect a process in another mount namespace, a nested mount from step 2, or a hung network share. Forcing is not offered for any of those.
Fix
Work from the cheapest release to the most disruptive. Run steps 3 and 4 again after each one.
Move shells out of the mount
Ask each user whose shell shows access c to run cd / or cd ~. In your own shell, run cd / before the next umount. Nothing is saved or changed on disk.
Stop the service that holds files
Find the unit that owns the process first. The systemctl manual says status accepts "a PID to show the unit containing that PID". Use a PID that fuser -vm printed; the unit name is on the first lines of the output:
systemctl status 2417
report.service below is the example unit name; use the name this command shows.
First save the current state of the service:
systemctl is-active report.service > /root/report.service-state.2026-11-22
The systemctl manual says is-active checks "whether any of the specified units are active" and prints the current state unless --quiet is used.
Disruptive step: the next command stops report.service. Every client of that service loses its connection, and work in progress that the service does not save on stop is lost.
Stop the unit by its own name, then repeat fuser -vm /mnt/data:
systemctl stop report.service
The systemctl manual says stop deactivates the unit. If the saved state was active, the unit was running before and must be started again when you are done.
To undo it: if /root/report.service-state.2026-11-22 says active, run systemctl start report.service, then systemctl is-active report.service. The result must match the saved file. If the saved file says anything else, the unit was not running before and nothing needs to be started.
Stop using a swap file on the mount
First save the current swap list:
swapon --show > /root/swap-list.2026-11-22
Disruptive step: the next command stops the system paging to the named file. Check that the memory and the remaining swap can carry the load before you run it.
The swapon manual says swapoff "disables swapping on the specified devices and files". Name the exact file shown in the list:
swapoff /mnt/data/swapfile
To undo it: run swapon /mnt/data/swapfile, then swapon --show and compare with the saved list.
Unmount
A plain umount is the right command once steps 3 and 4 list nothing and findmnt -R showed no submounts. If it showed submounts, use the next subsection instead.
First save the mount details, submounts included:
findmnt -R --list -o TARGET,SOURCE,FSTYPE,OPTIONS /mnt/data > /root/findmnt-data.2026-11-22
Disruptive step: the next command detaches /mnt/data. Every program that uses a path under it stops seeing the filesystem.
umount /mnt/data
To undo it: if /etc/fstab has a line for /mnt/data, run mount /mnt/data. The mount manual says that with only a directory given, "mount looks for a mountpoint (and if not found then for a device) in the /etc/fstab file". Otherwise mount the saved device on the saved mount point with the type and options in /root/findmnt-data.2026-11-22. Then run findmnt -o TARGET,SOURCE,FSTYPE,OPTIONS --target /mnt/data and compare with that file.
Unmount a tree of nested mounts
The umount manual describes -R as "Recursively unmount each specified directory. Recursion for each directory will stop if any unmount operation in the chain fails for any reason."
First save the whole tree:
findmnt -R --list -o TARGET,SOURCE,FSTYPE,OPTIONS /mnt/data > /root/findmnt-tree.2026-11-22
Disruptive step: the next command detaches /mnt/data and every filesystem mounted below it, each of which is listed in /root/findmnt-tree.2026-11-22. Every program that uses a path in that tree stops seeing those filesystems.
umount -R /mnt/data
To undo it: mount the filesystems from the saved file in order, starting with /mnt/data and then each one below it. For each line that has an entry in /etc/fstab, run mount with that mount point; otherwise use the source, type and options on that line of /root/findmnt-tree.2026-11-22. Then run findmnt -R --list -o TARGET,SOURCE,FSTYPE,OPTIONS /mnt/data and compare it with the saved file.
Example (illustrative): On web01, umount /mnt/data prints umount: /mnt/data: target is busy. The engineer runs findmnt -nR /mnt/data, which prints one line, the mount itself, so nothing is mounted below it. fuser -vm /mnt/data lists two processes: a shell belonging to user ops2 with access c, and a process of report.service with access F. lsof /mnt/data shows the same service writing a file under /mnt/data/reports. The engineer asks ops2 to run cd /, saves systemctl is-active report.service (the file says active), stops the unit, repeats both checks (now empty) and runs umount /mnt/data. After the maintenance the unit is started and is-active prints active again.
Confirm the fix
umountreturns with no message.findmnt --target /mnt/datano longer shows the filesystem as a separate mount; with a path that is not a mount point, findmnt reports the filesystem that contains the path.fuser -vmandlsofon the old device or mount point show nothing you did not expect.- Any service you stopped is started again once the maintenance is done and reports the saved state.
What to watch afterwards
- Add the mount to your maintenance checklist: stop the services that use it, in the order that releases it, before the unmount.
- Run
fuser -vmandlsofon every mount you plan to detach before the maintenance window, not during it. - Keep the saved
findmntoutput with the change record, so a remount uses the original device and options.
Read next
Related articles you may want next.
- TroubleshootingFiles Deleted but Linux Disk Space Is Still Used: Find Open File ReferencesAttribute retained deleted files to exact processes and choose a supported reopen or controlled restart.Read article
- TroubleshootingNFS "Stale file handle": what changed on the server and how to remountNFS Stale file handle means the server no longer honours a handle the client holds. Find what changed on the server, then unmount and remount safely.Read article
- VMwareDatastore "Is in Use" and Will Not Unmount: Record the Consumers Before You Change AnythingvSphere 8.0 datastore won't unmount, resource in use? Record inventory objects, host settings and open files per host, with evidence strength, before any change.Read article
Risks and limits
- These commands have not been run or lab-tested.
- Forced and lazy unmounts are not offered above, because neither can be undone safely. The umount manual describes
-fas "Force an unmount (in case of an unreachable NFS system)" and adds that it "does not guarantee that umount command does not hang". It describes-las "Detach the filesystem from the file hierarchy now, and clean up all references to this filesystem as soon as it is not busy anymore", names its recommended use as preventing "hangs on shutdown due to an unreachable network share", and says that for network filesystems "Remounts of the share will not be possible". For that reason neither option is offered here for a local filesystem that a program is using. fuser -kkills the processes it lists, with SIGKILL unless another signal is named. The fuser manual's own example,fuser -km /home, "kills all processes accessing the file system /home in any way". Killed processes cannot be restored, so no kill command is given here.- Mounts inside containers and other mount namespaces are not covered; the fuser manual says it will not see block devices mounted by processes in a different mount namespace.
- A busy mount that belongs to a shared datastore or a network share needs the owner of that service to agree before anything is stopped.
Sources
- umount(8) — util-linux 2.43 development manual page, 2026-08-03, accessed 22 November 2026
- findmnt(8) — util-linux 2.43 development manual page, 2026-08-03, accessed 22 November 2026
- mount(8) — util-linux 2.43 development manual page, 2026-08-03, accessed 22 November 2026
- swapon(8) — util-linux 2.43 development manual page, 2026-05-24, accessed 22 November 2026
- fuser(1) — psmisc manual page, 2022-11-02, accessed 22 November 2026
- lsof(8) — lsof 4.99.7 manual page, no date shown, accessed 22 November 2026
- systemctl — systemd documentation, version 262, no date shown, accessed 22 November 2026
- libmount/src/context_umount.c at v2.42.4 — util-linux source code at a fixed tag, no date shown, accessed 22 November 2026
- misc-utils/findmnt.c at v2.42.4 — util-linux source code at a fixed tag (column names TARGET, SOURCE, FSTYPE, OPTIONS), no date shown, accessed 22 November 2026
- sys-utils/umount.c at v2.42.4 — util-linux source code at a fixed tag, no date shown, accessed 22 November 2026
This article was originally published on .