Linux Tape Drive Management with mt Command

Magnetic tape drives remain a standard solution for high-capacity, cold-storage data archiving. In Linux, interacting with these sequential-access storage devices relies on the kernel's SCSI tape subsystem and user-space management through the mt (magnetic tape) utility. This guide explains how the Linux operating system interacts with physical tape hardware, how device nodes handle sequential operations, how the mt command controls drive mechanics via system calls, and the primary commands used for tape navigation.

Linux Tape Device Nodes

Linux interacts with tape hardware primarily through the st (SCSI Tape) kernel driver. When a tape drive is connected, the kernel exposes it as special character device files located in /dev/:

Rewinding devices (/dev/st0, /dev/st1, etc.): When a process completes writing or reading to this device file and closes the file descriptor, the kernel automatically rewinds the tape back to the beginning (BOT).
Non-rewinding devices (/dev/nst0, /dev/nst1, etc.): When an operation completes, the tape head remains at its current position. This is essential for appending multiple backup archives or sessions to a single tape.

Because tape drives are sequential rather than block-based (like HDDs or SSDs), data cannot be addressed randomly via standard filesystem mounts. Instead, data is written in sequential blocks separated by logical markers called filemarks.

How the `mt` Command Operates

The mt command does not read or write payload data itself; instead, it sends management and transport instructions directly to the drive mechanism.

Under the hood, mt opens the designated tape device file and issues ioctl (input/output control) system calls using standard tape operations defined in the <sys/mtio.h> header. These ioctl commands are translated by the Linux st driver into SCSI Stream Commands (SSC) sent across the bus (SAS, Fibre Channel, or USB) directly to the tape drive controller.

Essential `mt` Operations

By default, mt looks for a device designated by the TAPE environment variable. If this variable is not set, you specify the device using the -f flag. Always use non-rewinding nodes (/dev/nst*) with mt to prevent unintended rewinds after command execution.

1. Checking Drive Status

To check hardware readiness, error conditions, compression status, and tape head location:

mt -f /dev/nst0 status

The output shows the current file number, block number, density code, and tape flags (such as BOT for Beginning of Tape or ONLINE).

2. Positioning the Tape

Because tapes are linear, locating a specific backup archive requires positioning the head relative to filemarks:

Forward Space File (fsf): Moves the tape forward by a specified number of filemarks.
```
mt -f /dev/nst0 fsf 1
```
Backward Space File (bsf): Moves the tape backward across filemarks.
```
mt -f /dev/nst0 bsf 1
```
Rewind: Returns the tape immediately to the physical beginning.
```
mt -f /dev/nst0 rewind
```

3. Writing Filemarks

When appending data using streaming utilities like tar or dd, Linux writes an End-of-File (EOF) marker when the stream closes. You can also manually write filemarks to segment archives:

mt -f /dev/nst0 weof 1

4. Ejecting and Unloading Media

To safely unload tape tension and eject the cartridge from the drive:

mt -f /dev/nst0 offline

Workflow Integration

In practice, mt works alongside archiving tools. A standard archival workflow involves:

Checking drive status with mt -f /dev/nst0 status.
Writing an archive using tar -czf /dev/nst0 /path/to/data.
Verifying the tape head advanced past the new filemark using mt status.
Writing a second archive to the same tape using tar -czf /dev/nst0 /path/to/second_dataset.
Rewinding the tape using mt -f /dev/nst0 rewind.