Mount point
A mount point is a location in a filesystem hierarchy at which a mounted filesystem, bind-mounted subtree or file becomes accessible.
On Linux, the target can be a directory or a file. A bind mount exposes an existing directory subtree or a single file at another location; a mount point need not represent an entire disk or an entire filesystem. [1]
Mounting over a directory hides its previous contents from ordinary pathname lookup through that location; it does not delete those contents. The same filesystem can be mounted more than once. The pathname is an attachment location, not a unique identifier for the underlying filesystem or storage device. [2]
/proc/PID/mountinfo describes mounts in the target process's mount namespace. Its root field identifies the root within the mounted filesystem, whereas mount point is expressed relative to that process's root directory. These fields are not interchangeable. A mount ID may be reused after unmount; a parent outside the process's root may have no matching record. The list is not a permanent identity catalogue. [3]
Mount boundaries also constrain operations. Linux rename() cannot move a name between distinct mounts, even when they expose the same filesystem; this can produce EXDEV. Matching filesystem or device identity alone is therefore insufficient to establish that a rename can cross two paths. This restriction is separate from file and directory synchronization for crash durability. [4]
Relations
Citations
- Linux man-pages: mount(2), target location and bind operations
- util-linux: mount(8), covered directory contents and repeated attachment
- Linux man-pages: proc_pid_mountinfo(5), process scope and fields 4-5
- Linux man-pages: rename(2), EXDEV and distinct mounts
Sources & caveats
- Documentation synthesis from rolling primary manuals checked on 15 November 2026. The operational details are Linux-specific; no privileged mounting, rename test or recovery experiment was performed.