It’s a simplification for understanding it. It’s actually a lot more complicated (and involves virtual file systems). This is likely the origin of Kubernetes ConfigMaps being able to mount “volumes” that are individual files.
This is actually documented under --bind in the mount command man page:
One can also remount a single file (on a single file). It’s also possible to use a bind mount to create a mountpoint from a regular directory, for example:
mount --bind foo foo
I tried the example of bind mounting /etc/passwd. First, I had to create a file (not a directory) at the target:
touch /mnt/mypasswd
Then give the --bind option (or -o bind equivalent) to the mount command:
mount --bind /etc/passwd /mnt/mypasswd
The mount command shows something that might confuse someone who doesn’t understand bind mounts:
/dev/sda2 on /mnt/mypasswd type ext4 (rw,relatime)
This gets even more confusing when listing block devices with lsblk where this shows up as a “partition” type.
sda 8:0 0 476.9G 0 disk
├─sda1 8:1 0 1G 0 part /boot/efi
└─sda2 8:2 0 475.9G 0 part /mnt/mypasswd
The sda2 is actually created just for that single file mount even though it has identical size it has a different device minor:major number. This stuff is in /proc/self/mountinfo as well.
# grep passwd /proc/self/mountinfo
154 29 8:2 /etc/passwd /mnt/mypasswd rw,relatime shared:1 - ext4 /dev/sda2 rw
The findmnt command will show which mounts are files.
# findmnt |grep passwd
├─/mnt/mypasswd /dev/sda2[/etc/passwd] ext4 rw,relatime
So bind mounting is like filtering out specific directories—or a single file—from a base, existing file system and only making it available as a mountable file system itself.