Mounting storage devices

What is mounting?

Each hardware storage device is separated into partitions. A partition is a fixed space on a storage device where data can be written to and read from. A storage device can also contain only one partition spanning all of its available space.

Mounting is the process of making a partition accessible to userspace software. For userspace software to be able to read and write from and to partitions, they need to be mounted with a filesystem. The first filesystem to be mounted and the most important one, is the so called root filesystem. The root filesystem is already mounted read-only when init is started. It begins at / with a tree-like structure and every subfolder to it can be used to mount another filesystem.

As an example:
say there are the physical storage medium sda and sdb, with corresponding partitons sda1 and sda2. If the root filesystem lives on sda1, sda1 is mounted at /, that means all data that is written to or read from files in the folder / or any subfolder, e.g. /test, is saved on sda1. Suppose now, that additionally sdb1 is mounted at /mnt/drive. Now anything that is written to /mnt/drive and any of its subdirectories lands on sdb1 while everything else below / lands on sda1.

/                   # sda1 is mounted here
├── dev             # goes into sda1
├── mnt             # goes into sda1
│   └── drive       # sda2 is mounted here
│       ├── data1   # goes to sda2
│       └── data2   # goes to sda2
├── package         # goes into sda1
├── test            # goes into sda1
└── usr             # goes into sda1
    └── lib         # goes into sda1
			

Accessing storage media

All partitions are accesible via a corresponding device node at /dev. There are multiple naming conventions, differing between Linux and the BSDs. As an example, SSDs are named sdaY, sdbY, ... on Linux. Here, a, b, ... is enumerating the different drives, while Y enumerates the different partitions of each drive.

The problem is the following:
On most UNIX like systems the enumeration of drives is not stable, that is, it can change depending on the order in which the UEFI/BIOS detects the hardware. As an example: On a Linux system running on a machine with two SSD's R and S; R can be available at /dev/sda and S at /dev/sdb and after a reboot, R might be available at /dev/sdb and S at /dev/sda.

There are multiple approaches to this problem:

• Address each partition by its partition UUID, short PARTUUID. media-video/vlc This is assigned upon partitioning the storage device and stable thereafter.
• Address each file system by its UUID. This is assigned upon creating a files sytem on a partition and stable thereafter.
• Address each file system by its label. A label can optionally be assigned to each file system.
• Address each storage device by its hardware ID. Hardware ID's is unique between multiple different hardware devices and should be unique between multiple pieces of the same hardware.

After that, the UUID, PARTUUID or label can be given to a tool like findfs (on Linux) which will return the device node path /dev/sdXY

When using some userspace device managers, there may be symlinks at /dev/disk/by-id, /dev/disk/by-label and some more. by-id, by-label, by-partuuid and by-uuid can be considered stable and safe to be used for mounting.

How does it work?

There are the two main approaches on mounting partitions:

1. Mount each partition individually by issuing mount /dev/XY /path/to/mountpoint for all devices X and partitions Y.
2. Create a configuration file /etc/fstab which contains instructions on how and where to mount each partition. Then run mount -a to have mount parse /etc/fstab and mount the partitions accordingly.