Map Linux sg Devices to Disks, Optical Drives and Tapes

You will finish with a read-only way to relate generic SCSI-generic nodes such as /dev/sg0 to higher-level Linux device names such as /dev/sda and /dev/sr0. You will also know when the result is incomplete and why a newer discovery tool is usually a better choice.

Allow about fifteen minutes. You need a shell and the sg3-utils package. The examples below use sg_map 1.12 (20171010) from package version 1.46-3ubuntu4. The command only inspects device nodes, but opening SCSI devices may require elevated privileges on your host. It does not alter disks or persistent configuration.

1. Confirm the installed command

Check which executable will run, then ask it for its version. These are ordinary read-only commands:

$ command -v sg_map
/usr/bin/sg_map
$ sg_map -V
Version string: 1.12 20171010

The option is an uppercase -V, not the GNU-style --version. On this installation, --version is rejected as an unknown switch. That small difference is an easy scripting trap.

2. Run the default numeric mapping

With no selection option, sg_map scans numeric generic nodes and reports mappings for disks, optical drives and tapes:

$ sudo sg_map
/dev/sg0  /dev/sda
/dev/sg1  /dev/sdb

Your output will reflect the hardware and device names on the current machine. A line containing only /dev/sgN means that the generic device was found but no corresponding disk, optical or tape name was identified. A scanner is one example of a SCSI device that can have no alternate SCSI node.

Privileges are host-specific. If the unprivileged command reports messages such as Permission denied, rerun it with sudo if your account is authorised:

$ sg_map
Error opening /dev/sg0 : Permission denied
Stopping because no sg devices found

On the installed command, even this no-device result returned status 0 during the check. Do not use a zero status alone as proof that every mapping was read successfully.

3. Limit the result to one device family

Use one of the -s options when you are looking for a particular kind of node. For disks:

$ sudo sg_map -sd
/dev/sg0  /dev/sda
/dev/sg1  /dev/sdb

Use -sr for optical nodes named /dev/srN, -scd for older /dev/scdN names, and -st for tape nodes. These options filter the names that are displayed; they do not rescan or reclassify a device.

For a quick check of the command's accepted options, use:

$ sg_map -h
Usage: sg_map [-a] [-h] [-i] [-n] [-sd] [-scd or -sr] [-st] [-V] [-x]

4. Add SCSI address and type details

Use -x when a path is not enough to identify the physical target. The command adds host, bus, SCSI ID, LUN and SCSI type fields after each active generic node:

$ sudo sg_map -x
/dev/sg0  1 0 0 0  0  /dev/sda
/dev/sg1  2 0 0 0  0  /dev/sdb

The five numeric fields are, in order, host number, bus, SCSI ID, LUN and SCSI type. In the example, both entries have type 0, which is the disk type. Treat the numbers as identifiers for investigation, not as a stable naming scheme. Device numbering can change after hardware changes, adapter removal or reboot.

If you need the device's vendor, product and revision strings as well, add -i:

$ sudo sg_map -i -x
/dev/sg0  1 0 0 0  0  /dev/sda  ATA     DEVICE-NAME  REVISION

The inquiry text is hardware-dependent, so do not write a parser that assumes one fixed spacing or product length. Check the actual output on the host you are documenting.

5. Avoid the obsolete alphabetical scan

The default is the numeric scan, equivalent to -n, which looks for names such as /dev/sg0. The -a option instead assumes alphabetical names such as /dev/sga and /dev/sgb:

$ sudo sg_map -n
$ sudo sg_map -a

Linux has deprecated alphabetical sg node names since the 2.2 kernel series. Use -a only when you are dealing with a deliberately old environment that actually has those nodes. It is not a compatibility switch for modern /dev/sg0 devices.

6. Treat the result as a diagnostic snapshot

sg_map is deprecated and its manual page says it has not been actively updated for years. On modern Linux, the project recommends using lsscsi or udev facilities instead. The older utility scans conventional names and can be misleading when devices have been added or removed, when naming is sparse, or when udev assigns names differently from old major and minor number assumptions.

Use the output to answer a focused question, such as which current /dev/sgN node corresponds to a disk. If a mapping matters to an automation job, record a persistent identifier from the appropriate udev or SCSI tooling rather than storing /dev/sg0 as if it were permanent.

There is nothing to undo after these examples: the command reads device information and exits. If you used sudo, it granted access only to that invocation. It did not change ownership, permissions, udev rules or device configuration.