BusyBox: A Swiss Army Knife for Linux

Nicholas Wells

Issue #78, October 2000

Learn how to save disk space with this open-source tool for embedded systems.

Linux is being used in more and more tight places—devices or situations where multiple gigabytes of storage or dozens of megabytes of RAM are simply not available. Many Linux Journal articles have addressed this growing embedded systems space. Here, I will describe a terrific open-source tool for people needing a suite of utilities for use in these tight places.

BusyBox is a single-binary application that includes the capabilities of more than 70 standard Linux utilities. The BusyBox project was started by Bruce Perens when he was leading the Debian project. Many others have contributed code since then; the project is currently maintained by Erik Andersen, who also maintains a web page for it at http://busybox.lineo.com/. (BusyBox is sponsored by embedded Linux vendor Lineo, Inc. See http://www.lineo.com/.)

The corresponding standard utilities that BusyBox replaces occupy about 1.5MB of disk space in a standard Linux system. The BusyBox program uses only 260KB of disk space; it can even be compiled to include only a handful of utilities that you select, still within a single binary. We'll see how to do this later.

How It Works

Linux distributions normally include packages with many separate utilities, often in sets named textutils, shellutils or something similar. These utilities are generally very full-featured. For example, the command to list files, ls, supports over 50 command-line options. Because these utilities are so full-featured, they are sometimes larger than you might expect. For example, the ls command (dynamically linked) occupies 48KB of disk space. While you wouldn't notice that on a standard PC, many of those 50 options are just wasted space on an embedded system or boot disk. BusyBox combines numerous utilities in a single binary, limiting the features of each utility to those most commonly needed. The ls command in BusyBox has a mere 12 options—more than enough for most of us.

When you enter the name of a utility in a shell (at the Linux command line), the shell locates the binary with that name and executes it. When you install BusyBox, it creates links in your file system so that instead of having a real ls command, you have a symbolic link named “ls” that refers to the BusyBox binary. A directory listing of ls would then appear like this:

lrwxrwxrwx 1 nwells users 12 May 17 14:47 ls ->/bin/busybox

Typing ls at the command prompt causes the shell to launch the BusyBox program, which in turn examines the name under which you invoked it. Seeing that you entered ls in this example, the BusyBox program proceeds to act like ls, interpreting any additional options according to that command.

What Does BusyBox Include?

If this talk about 70 utilities in one is making you curious, here are a few samples of what BusyBox includes. The web site given above has a complete list with documentation:

  • chmod, chown, chroot, lsmod, rmmod, insmod

  • dd, df, du, mkfs.minix, fsck.minix, mount, sfdisk

  • find, grep, sed, gzip, gunzip, tar

  • kill, killall, ps

  • cp, cut, mv, rm, ln, more, tr

  • nslookup, ping, telnet

  • init, syslogd, sh

Many of these utilities are reduced versions to fit in the 260KB total space mentioned above. For example, the sh shell doesn't support if/then/else statements or while loops, but it has enough built-in commands to handle simple startup shell scripts. Also, commands like init and tar have greatly reduced functionality, but they are sufficient to get the job done for many applications where you don't have space for the full-blown utility.

Compiling BusyBox

Adding BusyBox to your system is simple. First, obtain the source code by visiting ftp.lineo.com/pub/busybox and downloading the most recent version of the compressed tar file. As of this writing, the most recent version is the file busybox-0.45.tar.gz, but newer versions will probably be available by the time you read this. Work on BusyBox seems to progress at a steady pace.

After you've downloaded the tar file, place it in a working directory and use this command to untar the file:

tar xvzf busybox-0.45.tar.gz

Change to the newly created busybox directory (I write the name in lowercase this time to match the binary name) and enter the make command:

cd busybox-0.45
make
Now you're ready to test a few BusyBox commands. The busybox binary is located in the main busybox directory after you use the make command. In order to use a particular feature of BusyBox, execute the busybox binary with a command name as a parameter. For example, to use the ls command, from within the busybox-0.45 directory, enter:
 ./busybox ls
Or to use the lsmod command, enter:
 ./busybox lsmod
Additional command options can be placed after the command name. For example, to use the du command to view only the contents of the /etc directory tree, type this: ./busybox du /etc

Another important feature of BusyBox is that you can see a small on-line help screen for each utility that BusyBox replaces. Just use the --help option with the command. For example, to learn about the ls options that BusyBox supports, enter ./busybox ls --help

This displays the following help text:

BusyBox v0.45 (2000.05.17-20:38+0000) multi-call
binary -- GPL2
Usage: ls [-1acdelnpuxACF] [filenames...]
Options:
-a      do not hide entries starting with .
-c      with -l: show ctime (the time of last
        modification of file status information)
-d      list directory entries instead of contents
-e      list both full date and full time
-l      use a long listing format
-n      list numeric UIDs and GIDs instead of names
-p      append indicator (one of /=@|) to entries
-u      with -l: show access time (the time of last
        access of the file)
-x      list entries by lines instead of by columns
-A      do not list implied . and ..
-C      list entries by columns
-F      append indicator (one of */=@|) to entries

Installing BusyBox

BusyBox is most useful on a system without a set of regular Linux utility programs, but you'll probably be exploring it on a standard Linux system. Because of this, you should use the PREFIX variable when installing BusyBox. The installation process creates symbolic links for all the utilities that BusyBox supports. This allows you to enter ls instead of busybox ls. Suppose you had begun working with BusyBox in the /tmp directory (such that a directory called /tmp/busybox-0.45 was created by the tar command). Then, if you want to create symbolic links in the same area, use this command:

make PREFIX=/tmp/busybox-0.45 install

The /tmp/busybox-0.45 directory will then contain subdirectories named bin, sbin and usr, each with symbolic links to /bin/busybox. You'll also need to copy the busybox binary to /bin before using these symbolic links:

cp /tmp/busybox-0.45/busybox /bin
Now you're ready to explore the symbolic links in your BusyBox subdirectories. For example, change to the bin directory:
cd /tmp/busybox-0.45/bin
Then use the ls symbolic link with the --help option:
 ./ls --help
You see the same help text as indicated previously. This shows you that BusyBox is being used instead of the standard ls command on your Linux system.

BusyBox uses the GNU C library, or glibc, which can add substantially to the space requirements of an embedded system or boot disk. You might consider looking at alternate C libraries to save space. Examples include minix libc and newlibc. Another example that looks promising but doesn't yet support the full functionality of BusyBox is the uC-libc project from Rt-Control (see http://www.uclinux.org/). The maintainer of BusyBox, Erik Andersen, is currently working to enhance this mini C library so that it can be used to reduce the total size requirements for BusyBox.

Configuring BusyBox

The description of BusyBox so far is straightforward, but doesn't capture all that the program offers. Returning to the source code you un-tarred when you compiled BusyBox, load the file busybox.def.h into a text editor:

cd /tmp/busybox-0.45 vi busybox.def.h

The first part of this file (about the first 100 lines) contains #define statements for each utility capability that will be included in BusyBox. If you don't want to include the capabilities of one of these utilities, simply comment out that line. For example, if you don't need sed in the system you are using BusyBox on, comment out the line for sed using two forward slashes, like this:

//#define BB_SED
Commenting out a few of the larger utilities greatly reduces the size of the final busybox binary. For example, removing five complex programs (init, tar, sfdisk, gzip and gunzip) reduces the size of the busybox binary from 260KB to 155KB.

The second part of the busybox.def.h file (after a few explanatory comments) contains #define statements that activate or disable various features of BusyBox. Some of these features are intended to save memory, such as eliminating the use of the /proc file system, reducing the amount of on-line help provided and eliminating the use of regular expressions. Other #defines are specific to features of a single command. For example, you can eliminate the ability to create new tar files with the tar features of BusyBox. Unless you really need to shave off a couple more KB in the size of BusyBox, you shouldn't need to alter the #define options in the second section of the busybox.def.h file.

Conclusion

Embedded versions of Linux such as Lineo's Embedix may be the most obvious use of BusyBox, but you might come up with many other uses. For example, if you need to create an initrd file to boot up a system with unusual hardware, you can use BusyBox functionality to add the basic system utilities with a single easy-to-manage binary. Or, you might use BusyBox as part of a boot diskette or single-disk version of Linux, as the Linux Router Project and the Debian boot floppies.

Resources

Nicholas Wells is a writer who has published about ten books on Linux, KDE, and the Web with Sybex, IDG, etc. He enjoys studying languages (the human kind) and traveling to exotic places for Linux conferences.