Skip to content

Latest commit

 

History

History
146 lines (79 loc) · 3.6 KB

dd.md

File metadata and controls

146 lines (79 loc) · 3.6 KB
Raw

dd

POSIX 7.

Low level file operations.

Mnemonic: Duplicate Data.

Funny mnemonic: Data Destroyer.

Reason: the main use case for dd is to manipulate block devices like hard disks at a low level without considering file structure. This makes operations such as disk copy very fast, but can make a wipe a disk by changing two characters of a command.

if of

  • if = input file. If not given, stdin.
  • of = output file. If not given, stdout.

Same as echo a | cat:

echo a | dd

Same as cat a:

echo a > a
dd if=a

Same as cp a b:

dd if=a of=b

status

Stop printing status line at the end of the transfer:

echo a | dd status=none

bs

How many input and output bytes to read/write at once.

Also defines the block size for count, skip and seek.

obs and ibs for output and input separately.

Default values: 512 B.

count

Copy up to count blocks (defined by bs):

[ "$(printf '1234' | dd status=none bs=2 count=1)" = '12' ] || exit 1
[ "$(printf '1234' | dd status=none bs=1 count=3)" = '123' ] || exit 1

size suffixes

  • c: 1 (char)
  • w: 2 (word)
  • kB: 1000
  • K: 1024
  • MB: 1000*1000
  • M: 1024*1024

and so on for G, T, P, E, Z and Y!

[ `echo -n 123 | dd status=none bs=1c count=1` = 1 ] || exit 1
[ `echo -n 123 | dd status=none bs=1w count=1` = 12 ] || exit 1

The larger the chunk size, the potentially faster file transfers will be.

Nowadays, 4M is a good value for large file transfers.

skip

Skip first n input blocks (defined by bs or ibs):

[ `echo -n 123 | dd status=none bs=1 skip=1` = 23 ] || exit 1

seek

Skip first n output blocks (defined by bs or obs):

TODO minimal example.

conv

Comma separated list of options.

Most useful ones:

  • ucase: convert to uppercase:

      [ `echo -n abc | dd status=none conv=ucase` = ABC ] || exit 1

  • noerror: ignore read errors, which would otherwise stop the transfer.

    Prefer a specialized tool like ddrescue to this method.

    Application: part of your hard disk is broken, but you are willing to risk a faulty full transfer anyways.

iflag oflag

TODO

verbose

Monitor progress

Not easy: http://askubuntu.com/questions/215505/how-do-you-monitor-the-progress-of-dd

Applications

Copy one hard disk to another:

sudo dd bs=4M if=/dev/sda of=/dev/sdb

Zero an entire block device located at /dev/sdb:

sudo dd bs=4M if=/dev/zero of=/dev/sdb

As of 2013 with mainstream system specs, this took around 6 minutes on a 2GiB flash device (5.0 MiB/s).

If you are really serious about preventing forensic data recovery, use a program for with a more advanced algorithm, or just destroy the hard disk.

If you just want to hide content from very simple retrieval, doing this command for a few seconds is likely to destroy the partition table, and that should be enough.

ddrescue

Like dd, but behaves smartly for disk errors.

This can really make a difference when backing up faulty hard disks, so always use it in that case.

There are two versions of ddrescue with incompatible interfaces:

  • GNU. Executable name: ddrescue. Ubuntu package: gddrescue.
  • Kurt Garloff's. Executable name: dd_rescue. Ubuntu package: ddrescue.

We recommend sticking to the GNU version, as it is maintained by an organization rather than an individual.

Basic usage:

ddrescue -fv /dev/from /dev/to ddrescue.log
  • -f: write even if destination exists already. Required to copy device files to other device files.
  • -v: print the transfer status while it is going on.

TODO: what is trimming and splitting? Why do they take so long at the end? Splitting can be prevented with: -n.