View/print PDF file
UNIX at Fermilab

Chapter Contents

Chapter 7: The UNIX File System
  7.1 Directory Structure
    7.1.1 Pathnames
    7.1.2 The Home Directory
    7.1.3 Command Line Directory Shortcuts
    7.1.4 Directories and Executables
  7.2 Files
    7.2.1 Filenames
    7.2.2 Filename Expansion and Wildcard Characters
  7.3 Manipulating Files
    7.3.1 List Directory Contents: ls
    7.3.2 List File Contents: cat, less, more, head, and tail
    7.3.3 Copy a File: cp
    7.3.4 Move or Rename a File: mv
    7.3.5 Reference a file: ln
    7.3.6 Remove a File: rm
    7.3.7 Copy to/Restore from Archive or Tape: tar
    7.3.8 Compress or Expand a File: gzip, gunzip
  7.4 Information About Files
    7.4.1 Find a File: find
    7.4.2 Search for a Pattern: grep
    7.4.3 Count a File: wc
    7.4.4 Dump a File: od
    7.4.5 Determine File Type: file
  7.5 Manipulating Directories
    7.5.1 Print Working Directory: pwd
    7.5.2 List Directory Contents: ls
    7.5.3 Change Directory: cd
    7.5.4 Make a Directory: mkdir
    7.5.5 Copy a Directory
    7.5.6 Move (Rename) a Directory: mv or mvdir
    7.5.7 Remove a Directory: rmdir
  7.6 File and Directory Permissions
    7.6.1 File Access Permissions
    7.6.2 Directory Permissions
  7.7 Temporary Directories

Chapter 7: The UNIX File System

The UNIX file system has a hierarchical or tree-like structure with the directory called root (/) as its source. The system is essentially composed of files and directories. In this chapter we describe techniques for manipulating files and directories, and commands designed to provide information about them.

7.1 Directory Structure

The UNIX system automatically puts you at a specific location in the file system when you log in. This is called your login directory. Typically, this is the same as your home directory. The name of your home directory is usually the same as your login name. Within this directory you can create files and additional directories (sometimes called subdirectories) in which to group the files. You can move and delete your own files and directories and control access to them.

The root of the file system is called root and is written as a slash (/). In other words, to change to the root directory, type:

% cd / 

There is only one directory tree on a system even if several devices are mounted in that tree. (All devices are viewed as files.) The current directory or working directory is the directory that you are currently working in, which is also the directory that commands refer to by default. Files in your current directory can, therefore, be specified by their filenames only.

7.1.1 Pathnames

Wherever you can use a filename, you can also use a pathname, which is how you point to files that are not in your current directory. You can refer to files in other directories using either a relative path name, that is a path specified relative to your current directory, or with an absolute path name, that is a path specified relative to the root of the file system.

Absolute path names are preceded with /, the root directory. If a pathname does not begin with / it is assumed to be a relative path name. Relative path names begin with a directory or filename, a . (pronounced "dot") which refers to the current directory, or .. (pronounced "dot dot") which refers to the directory immediately above the current directory. The character / also separates components of the pathname, which are directory names, except for the last one, which can be either a simple filename or a directory name.

In summary, every file has a pathname, and its absolute pathname is of the form:

/rootdir/dir2/... /filename

The following is the form of a relative pathname of a file:

dir_n/dir_n+1/... /filename

An example of an absolute path name is:


If my current directory is /usr/smith, then I can refer to the file afile in subdirectory project1 with a relative pathname like this:


Or, if my current directory is /usr/smith/project1, I can refer to a file named fileb in /usr/smith/project2 as:


Note that you cannot necessarily tell if fileb is an ordinary file or a directory name. Many commands will accept a directory name, and if it is a directory name, the command in which it is used may perform the action on all files in the directory. This behavior can be dangerous!

7.1.2 The Home Directory

Your home directory is the top of your personal branch in the file system, and is usually designated by your username, i.e. /<path>/<username>.

Tilde (~)

In most UNIX shells other than sh, the tilde (~) stands for the home directory. Used alone, it specifies your home directory. Followed by a different user's login name, it expands into the pathname of the home directory of that user. This is a convenient way to refer to a user's directory, because it is independent of where the system manager may place the directory on the disk.

The use of tilde (~) to refer to a home directory is limited. It isn't available in the Bourne shell, and isn't available in FORTRAN.

Of the following three examples, the first refers to the file def from your own home directory, the second to the home directory of user jones, and the third refers to file data1 in the subdirectory project1 of jones' home directory.




To change to jones' home directory you'd enter:

% cd ~jones 


FullFUE provides the command logdir which returns the full path of the specified user. logdir by itself returns the path of the invoking user. For example:

% logdir <username> 

is equivalent to:

% echo <~username> 

To change directories you'd enter (note the use of backquotes to use the output of the enclosed string):

% cd `logdir <username>` 

In contrast to the tilde, logdir can be used within commands, scripts, FORTRAN and C programs, and other programs in all shells.


The environment variable HOME is automatically set to the absolute pathname of your home directory. Environment variables are discussed in section 9.5 Shell Variables and Environment Variables and HOME is described in section 9.6 Some Important Variables. To see the value of HOME, enter:

% echo $HOME 

From some other directory, you can change to your home directory or one of its subdirectories using a command like the following:

% cd $HOME1 


% cd $HOME/mysubdir 

7.1.3 Command Line Directory Shortcuts

Current directory
Parent directory of the current directory ("up" one directory)
Your home directory (all shells but sh)
Environment variable whose value is your home directory
Home directory of another user (all shells but sh)
Root directory

7.1.4 Directories and Executables

It is appropriate at this point to mention the relationship between directories and commands. A command is simply the name of an executable file, located in some directory. To execute a command, the shell first needs to find the executable file. The shell therefore needs to be given a set of directories to search. This information is provided via the environment variable PATH which is a list of search directories. You can display it with the command:

% echo $PATH  

PATH is explained more thoroughly in section 9.6 Some Important Variables. Standard UNIX commands are generally grouped in a few standard directories (e.g., /usr/bin), and your default PATH contains these. See section 9.6 Some Important Variables to learn how to run executables that you create and store in your own directories.

The utility which is useful in cases where a command may be ambiguous, for example due to aliasing (see section 9.7 The Alias Command), and you want to know exactly which executable file or files the command runs. which lists the files that would be executed if the specified command(s) had been run. The syntax for which is:

% which <command> [<command2> ...] 

Each argument is expanded if it is aliased, and your path is searched for the executable files associated with the commands. See the man page for more information.

7.2 Files

An ordinary file contains ASCII characters or binary data and is considered by the UNIX system to be merely a sequence of bytes. No structure is imposed on the file and no meaning attached to its contents by the system; the meaning depends on the program that reads the file.

A directory file contains an entry for each file in that directory. The directory entry for a particular file contains the file name and inode number. The inode number is a volume data structure used by the file system. It has an associated entry in the inode table which contains other information about the file such as the owner, file protection, modification date.

A hidden file is an ordinary file whose name begins with a period (called "dot"). The reason they are called hidden is that the ls (list files) command does not list them by default. Use the -a option with ls to see them. Hidden files do not appear in filename expansion of *, either. Filename expansion is discussed in section 7.2.2, below.

UNIX does not support file versions. If you edit a file and save it with the same name, your earlier version is overwritten. Similarly, if you copy or rename (move) a file to a filename that already exists, the original file is overwritten.

7.2.1 Filenames

A full file specification has only two parts, the directory specification and the file name. A filename is composed of from 1 to 14 characters in old UNIX implementations and a much larger number in more recent versions (up to 255, typically). Although you can use any character in a filename except /, UNIX assigns special meaning to many characters (metacharacters), so they should be avoided (see section 9.2 Special Characters (Metacharacters)). It is safe to use the upper- and lowercase letters, numbers, dash (-), underscore (_), period (.), and comma (,). As mentioned in the previous section, files beginning with a dot (.) are hidden files. The "filenames" . and .. (single and double dot) are reserved. The . refers to the current directory, and the .. refers to the current directory's parent directory. No two files in the same directory can have the same name, but files in different directories can have the same name.

Dashes are fairly common in UNIX filenames simply because it's easier to type my-file than my_file.

Filenames are case sensitive. This means MYFILE is different from Myfile is different from myfile is different from myFile, etc.

You cannot distinguish a directory file from an ordinary file by its name, although some people make their own convention by beginning directory filenames with a capital letter, or ending them in .d.

Filename extensions are not required in UNIX. You can include a period and an extension in a filename to help describe the contents of the file, but it will not have special meaning to UNIX itself. However, programs can make use of extensions, for example the FORTRAN compiler expects certain extensions. Note, you can have more than one period in a filename, for example, lex.yy.c.

7.2.2 Filename Expansion and Wildcard Characters

The UNIX shells have a number of special characters which can be used on the command line when specifying filenames and directory names. They allow the shell to expand the argument into a set of filenames. These characters are called wildcards. Filename references that contain these characters are called ambiguous file references. Filename expansion is also called globbing.

The question mark (?) causes the shell to generate filenames which match any single character in that position. For example, out? matches out1 but not out12.

The asterisk (*) causes the shell to generate filenames which match any number of characters (including zero characters) in that position. For example, myfile matches myf*. The * alone means all files (except those that begin with dot (.), which is a special case).

A pair of brackets ([ ]) surrounding a list of characters causes the shell to match filenames containing the individual characters in that position. The brackets define a character class and each definition can only replace a single character in a filename. In other words, it is like a question mark that will only allow certain characters. For example, memo1 and memoa match memo[14a]), but memo3 and memo1a do not. A hyphen can be used to define a range of characters, for example [a-z] represents all lowercase characters. Thus memo[a-z] matches memoa but not memo2 or memoB.

matches any single character in a filename


matches any string of characters (including the empty string) in a filename

[ ]

matches any single character from the set enclosed in the brackets


% ls out*
lists all files beginning with out
% ls out?
lists all files with 4-character names beginning with out
% ls out[ab]*
lists all files beginning with out followed by a or b (e.g., outa4)
% ls *out*
lists all files containing out

Filename expansion may surprise you with the results. For example, ls b* would list all files starting with b in the current directory, but it would also list the contents of all directories whose names start with b because of the way ls behaves for a directory argument. If you want to be sure of what filename expansion will result in, you can use the echo command to check it before executing a command.2 For example, say you have a few matching files in your directory for the command:

% echo *out*  

You would obtain output something like this:

fout fout275 inandout out1 out2 out 

Filename expansion in csh can be turned off by setting the noglob variable:

% set noglob 

To turn it back on, type unset noglob.

7.3 Manipulating Files

This section describes the basic file manipulation commands:

Section 7.5 Manipulating Directories describes the commands you can use to change and manipulate directories.

7.3.1 List Directory Contents: ls

The ls command, which stands for list, is used to list the contents of a directory. ls has many options, some of which are system-dependent, so only a few of them are described here. For a complete description of the command, refer to the man pages for ls.

ls by itself lists the names of the files and subdirectories in the current directory (in multicolumn format on some platforms), sorted alphabetically.

The format is:

% ls [<options>] [<filenames>] 

where some of the options are:

List all entries, including those that begin with . (dot).
List in the long format, giving mode, number of links, owner, group, size in bytes, and (by default) time of last modification, by default sorted by filename.
List in columns (default on some platforms)
Put a / after the name of each file that is a directory, an * after the name of each file that is executable, and an @ after the name of each file that is a symbolic link.
Recursively list subdirectories encountered.
Sort by time stamp (latest first) instead of by name. The default time stamp is the last modification time (see -u).
Use the time of last access for sorting if used with the -t option or printed in the date column if used with the -l option. -ult both sorts by and displays last access date.
If the argument is a directory, list the directory itself, not its contents. Use with -l to get the status (e.g., permissions) of a directory.

If the argument is a directory, ls displays the contents of the directory. Note that this can happen unintentionally as a result of filename expansion. This behavior can be prevented with the -d option. The -t option is useful when looking at recent files:

% ls -lt  

will result in the long output sorted by reverse modification date rather than by filename.

The following is a sample output of ls -l.
total 251  
drwxr-xr-x 3 nicholls g020c 512   May 2  08:53 Tools  
drwxr-xr-x 2 nicholls g020c 512   May 2  09:01 bin  
-rw-r--r-- 1 nicholls g020c 446   May 4  14:09 defaults  
-rw-r--r-- 1 nicholls g020c 95418 May 1  17:42 intro.lpr  
-rw-r--r-- 1 nicholls g020c 0     May 10 17:51 lsout  
-rw-r--r-- 1 nicholls g020c 6683  May 1  16:46 man.lpr  
-rw-r--r-- 1 nicholls g020c 12258 May 9  16:16 out  

The first line indicates the number of blocks used. The rest of the lines report on (sub)directories or files in the directory being reported on. The first column of the output is called the mode. The character in this first column indicates the type of file, and for our purposes here, they are:

ordinary file

The next 9 characters are interpreted as three fields of three characters each, indicating the read (r), write (w), and execute (x) permissions for owner (sometimes called user), group, and other, in that order (see section 7.6.1 File Access Permissions for a discussion of permissions).

Next is the number of links to the file or directory. This refers to the number of different names established for it. Normally files have 1, and directories have 1 each for the directory itself, its parent directory, and each of its subdirectories. In the sample output above, notice that the directory Tools has 3 and bin has 2. Evidently, Tools has one subdirectory and bin has none.

The next fields are the login name of the owner, the group to which the owner belongs, the size of the file in bytes, the date and time the file was last modified, and, finally, the filename (which can be a directory name).

7.3.2 List File Contents: cat, less, more, head, and tail

UNIX has a number of commands that can be used for displaying the contents of a file at the terminal.


cat, which stands for "concatenate and print," is the standard UNIX file display; it simply prints the file to the screen.3 When piped to less (see section 6.4.4 Filters which describes less as a filter), cat displays the file contents a screen at a time, and some simple commands may be executed at the supplied prompt.

% cat <filename>... [| less] 

As its name suggests, cat is in fact quite useful for copying and concatenating files. Output can be piped to a file rather than to the screen, using standard output redirection (see section 6.4.2 Standard Input and Output Redirection). The following example concatenates the three specified files and copies them sequentially to a single file called allthreefiles:

% cat fileone filetwo filethree > allthreefiles 

less, more

A shortcut for cat <filename> |less is to use less as a file browser:

% less <filename> 

And wherever you use less, you can alternatively use more, although it is not as functional as less. You cannot move backwards through the file with more.

head, tail

head displays the first <n> lines of the specified file or files. If more than one file is specified, the filename is displayed before each set of file contents of <n> lines. <n> defaults to 10 lines.

% head [-<n>] [<filename>...] 

tail displays the last lines of a file. Its syntax is a bit different:

% tail [+|-<n> lbc] [<filename>...] 

The option +<n> displays the file starting <n> lines down from the beginning of the file, -<n> displays the last <n> lines. l, b, or c requests display of <n> lines, blocks, or characters (default is l lines). If more than one file is specified, the filename is displayed before each set of file contents. <n> defaults to 10 lines.

tail is useful when you want to see how far a process got. To display the last line of a log file, enter:

% tail -1 <logfile> 

7.3.3 Copy a File: cp

The command cp (stands for copy) can be used to make a copy of a file, leaving the original version intact. You can copy a single file to another one (in the same or a different directory), or you can copy one or more files to a different directory, retaining the same filenames.

The syntax for these two situations varies slightly:

% cp [<options>] file1 targetfile 
the file <file1> is copied to <targetfile>, where <targetfile> may include a path
% cp [<options>] <file1> [<file2> ...] <targetdirectory> 
one or more files (<file1> <file2> ...) are copied to <targetdirectory>

If the target is a file, its contents are overwritten unless -i is specified, in which case you are prompted for confirmation.

Some options are:

If the target filename exists, you are prompted for confirmation before overwriting.
Used only with the <targetdirectory> form. Recursively copy a directory, its files, and its subdirectories to <targetdirectory>.

The first example below copies myfile to anotherfile, both in my current directory, prompting for verification in case anotherfile already exists.:

% cp -i myfile anotherfile  

New users may find it useful to define cpi as the alias for cp -i to use in place of cp so that prompting always occurs. Section 9.7 The Alias Command discusses aliases.

The second example copies files proj1 and proj2 to another directory named newproj which is parallel to the current directory (has same parent directory as current):

% cp proj1 proj2 ../newproj  

The third example copies the file oldproj/proj1 to my current directory (.), which is a parallel directory to oldproj (has same parent directory). The file proj1 keeps its name.

% cp ../oldproj/proj1 .  

7.3.4 Move or Rename a File: mv

The mv command (stands for move) allows you to rename a file in the same directory or move a file from one directory to another. If you move a file to a different directory, the file can be renamed or it can retain its original name. mv can also be used to move and rename directories.

% mv [<options>] <source1> [<source2> ...] <target>  

Depending on whether the <source(s)> and <target> are files or directories, different actions are taken. These are described in the table below. If <target> is a filename, only one <source> file may be specified.
new file name
Rename file to new name
existing file name
Overwrite existing file with source file contents; keep existing file name
new directory name
Rename directory to new name
existing directory name
Move directory such that it becomes a subdirectory of existing directory
one or more files
existing directory name
Move files to existing directory

An important option is:

If <target> exists, the user is prompted for confirmation before overwriting.

7.3.5 Reference a file: ln

The ln (link) command allows you to create a link in one directory to a file in the same or in a different directory, or to create a link to a different directory. Via links, a file or directory can appear to exist in multiple places, but only actually exist in one, thus conserving disk space. Links are often used to easily reference files or directories that would otherwise require a long path name.

The syntax for ln is similar to that for cp and mv, and in fact they are all run by the same executable.

The most commonly used options for the ln command are:

You are prompted before overwriting an existing filename.
This makes a symbolic, as opposed to an ordinary or hard, link. A symbolic link can point to a file that is in a different file system, whereas a hard link cannot.

A symbolic link displays the link and the file to which it is linked when you run ls -l; this is the only way to know the name that a file is linked to.

Note that when using the AFS file system, hard links can only be made between files that are in the same directory (the same volume, see section 8.5 AFS Volumes and Quota), so use the -s option even if you're in the same directory tree.

The syntax differs slightly for files and directories:

% ln [<options>] </path/to/file_name> </path/to/link_name> 
Create the link <link_name> to reference the file file_name. If <link_name> already exists (as a link or as a file), it gets overwritten (unless you use option -i).
% ln -s [<other options>] </path/to/file_name> </path/to/link_name> 
Create symbolic link named <link_name>, that links to <file_name> which exists in the same or another directory.
% ln -s [<other options>] </path/to/file_name1> [</path/to/file_name2> ...] <directory> 
Create a symbolic link in <directory> to each of the listed files. The files may all exist in different directories since the -s option is used. The link names will be the same as the filenames they link to. If files of the same name but in different directories are specified, only the first file specified of that name will have a link created.

Let's look at an example:

% ln -s /e741/run1/e_mu2/mydata r5742 

If r5742 is a directory, this creates a link called mydata in the directory r5742 that points to /e741/run1/e_mu2/mydata. You can now reference the data file as mydata (i.e., the same filename) as if it were in the directory r5742.

On the other hand, if r5742 is not an existing directory then it represents the name of the link being created. In this case, the command establishes a link called r5742 in the current working directory that points to the file /e741/run1/e_mu2/mydata. Running the command ls -l should display the following output:

lrw-r--r--   1 aheavey  g020          46 Aug 29 14:26 r5742 -> /e741/run1/e_mu2/mydata 

7.3.6 Remove a File: rm

The rm command (stands for remove) is used to remove the entries of one or more files.

% rm [<options>] <file>...  

Some commonly used options are:

Confirmation of removal of write-protected file occurs interactively, whether the standard input is a terminal or not. If used with the -r option, you are prompted about each directory before it is examined.
Causes rm to delete the contents of the specified directory, including all its subdirectories, and the directory itself (recursive). This option should be used cautiously.

The file list can include ambiguous file references, so rm should be used cautiously. You can use the echo utility with the same ambiguous file reference to see the list generated.

Removal of a file requires write permission to its directory, but neither read nor write permission to the file itself. If the file has no write permission and the standard input is a terminal, the set of permissions is printed and you are prompted for confirmation. If the answer begins with a y, the file is deleted. If the standard input is not the terminal, the files are deleted without confirmation.

New users may find it useful to define rmi as the alias for rm -i to use in place of rm so that prompting always occurs. Section 9.7 The Alias Command discusses aliases.

7.3.7 Copy to/Restore from Archive or Tape: tar

The tar utility (tape archive) can be used to create, add to, list and retrieve files from an archive file. Archive files are often stored on tape. The action taken by the tar command depends on the key, which is essentially a function option. The key must be specified on the command line as if it were the first option. It may be followed by function modifiers, and then by options and/or arguments. The keys and function modifiers must be grouped together before any arguments are listed. tar does not require, but does allow, a dash (-) before the list of keys and function modifiers. The keys and functions are:

create a new tar file
append specified files to tar file
list all files in the tar file, or all files in a specified file list
append new or changed files to tar file
unwind entire tar file or extract specified files from tar file and write each file to the directory as specified in the tar file relative to the current directory

The keys, function modifiers and options are discussed in the man pages. Be aware that they vary in some cases between UNIX flavors. The command syntax varies somewhat from key to key, so check the man pages for that information, too.

Creating a Tar File

When creating a tar file, we have a few recommendations for avoiding problems:

First, when possible, create the tar file on a machine of the same flavor as the target flavor. Occasionally a tar file doesn't unwind properly on a different platform.

Secondly, choose your working directory carefully. It is often convenient or desirable to be able to specify simple relative path names for the files to include in the tar file. For example:

% cd /path/to/dir 
% tar cvf /tmp/filename.tar . 

creates a tar file with all pathnames relative to /path/to/dir. In general you should not specify the pathname explicitly on the command line, unless it will be valid on any other system where the tar file may be unwound and used.

Thirdly, be careful choosing your target directory for the new tar file. Make sure that the target directory is outside of the directory tree that you're including in the tar file. Otherwise the tar file tries to include itself, and can grow infinitely large.

Unwinding a Tar File

To unwind a tar file, first cd to or create the target directory in which you want the tar file unwound, then unwind the product tar file:

% cd <target_directory>  
% tar xvf <path_to_unwound_tar_file>  

7.3.8 Compress or Expand a File: gzip, gunzip

Several utilities are available on UNIX systems for file compression. compress and pack are native UNIX utilities, and gzip is provided by FullFUE. We recommend you use gzip for file compression, and its associated utility gunzip for file expansion. gunzip recognizes and can expand files compressed with compress and pack as well as gzip.4

The file extensions gunzip recognizes include .gz, -gz, .z, -z, _z, and .Z. gunzip also recognizes the special extensions .tgz and .taz as shorthands for .tar.gz and .tar.Z, respectively. When compressing, gzip uses the .tgz extension if necessary instead of truncating a file with a .tar extension.

You will need to reference the man pages for details on syntax, options and usage. In their simplest forms gzip and gunzip can be used as follows, starting, for example with the original uncompressed file bigfile:

% gzip bigfile 

The result is bigfile.gz, whose size is reduced with respect to bigfile according to Lempel-Ziv coding (LZ77), the same compression scheme used by compress. Whenever possible, gzip replaces each file by one with the extension .gz, while keeping the same ownership modes and access and modification times. gzip will only attempt to compress regular files. In particular, it will ignore symbolic links. If the compressed file name is too long for its file system, gzip truncates it.

Compressed files can be restored to their original form using gunzip, or equivalently by using the -d option with gzip. If the original name saved in the compressed file is not suitable for its file system, a new name is constructed from the original one to make it "legal".

To restore bigfile.gz to its original name and size, enter:

% gunzip bigfile.gz 

7.4 Information About Files

This section gives a cursory overview of simple uses for two very powerful commands for dealing with files: find for searching for files and grep for searching for strings within files. We also describe wc which displays the size of a file, od which creates a dump of a file, and file which can determine file type.

7.4.1 Find a File: find

The find utility tests each file in the given pathname list to see if it meets the criteria specified by the expression supplied. It does this by recursively descending the directory hierarchy for each path name. The format is:

% find <path-name-list> <expression>  

<path-name-list> can contain file expansion metacharacters. Each element in <expression> is a separate boolean criterion. A space separating two criteria is a logical AND operator, a -o separating the criteria is a logical OR operator. A criterion can be negated by preceding it with an exclamation point (!). Criteria are evaluated from left to right unless parentheses are used to override this. Special characters must be quoted (use \) and there must be spaces on each side of the special character pair.

Some of the criteria that can be used within <expression> are:

-name <filename>
True if <filename> matches the name of the file being evaluated. Ambiguous file references can be used if enclosed in quotes.
-type <filetype>
True if the type of the file is <filetype>, where <filetype> is either d (directory) and f (ordinary file).
-atime <n>
True if the file has been accessed in <n> days.
-mtime <n>
True if the file has been modified in <n> days.
-newer <filename>
True if the file has been modified more recently than <filename> has.
Causes the matching path names to be displayed on the screen.
-exec <command> \;
True if <command> returns a zero exit status. <command> must be terminated with a quoted semicolon (note the \). An empty pair of braces ({}) within the command represents the filename of the file being evaluated.
-ok <command> \;
Same as -exec except the generated command line is displayed and executed only if the user responds by typing y.

In the previous list, +<n> means more than <n> , -<n> means less than <n>, <n> means exactly <n> .

Note that find doesn't do anything with the found files, it doesn't even display the names, unless instructed to.


% find . -name lostfile -print 
% find wwwork -name '*.html' -print 
% find . -name 'd*' -ok more {} \;  
% find . ! -name 'm*' -print 
% find . -exec grep -l "hello" {} \;  
% find ~\(-name a.t -o -name '*.o') -atime +7 -exec rm {} \;  

Note that using the find command takes up a lot of system resources.

In particular on AFS systems, you may accidentally end up searching servers all over the world if the top of the search is at the root directory (/). Generally you should be careful to only search the part of the UNIX tree that interests you. Here is an example:

Look for <filename> starting at the root directory (/), and exclude searches in the /afs and /nfs branches:

% find / \( -name /afs -prune \) -o \( -name /nfs -prune \) -o -name <filename> -print 

7.4.2 Search for a Pattern: grep

The grep utility searches the contents of one or more files for a pattern.

The format is:

% grep [<options>] <pattern> [<file> ...]  

Some of the options are:

Display only a count of lines that contain the pattern.
Ignore upper/lower case distinctions during comparisons.
Display only the name of each file that contains one or more matches.

The pattern can be a simple string or a regular expression (see section 6.4.5 Regular Expressions). You must quote regular expressions that contain special characters, spaces, or tabs (this can be done by enclosing the entire expression within single quotation marks).


% grep -i smith *  
% grep 'fr*og' abc  
% grep '^T' myfile  


% less myfile | grep '^T'
% grep 'file[0-9]' /usr/jones/junk  
% who | grep smith  
% ps -ef | grep smith 
% env | grep <string> 
% alias | grep <string> 

7.4.3 Count a File: wc

The wc command, which stands for word count, counts the number of lines, words, and characters there are in the named files, or in the standard input if the argument is absent. If there is more than one file, wc totals the count as well.

% wc [-lwc] [<names>]  

The options l, w, and c may be used in any combination to specify that a subset of lines, words, and characters be reported. The default is -lwc.

UNIX users frequently count things by piping them into wc. For example, to display the number of users logged into the system, you can execute:

% who | wc -l  

7.4.4 Dump a File: od

The od (octal dump) command can be used to examine the contents of a file in various formats: octal, decimal, hexadecimal, and ASCII. The default is octal.

The format is:

% od [<options>] [<file>] [<offset>] [|less] 

If <file> is not included, standard input is assumed. The options are:

Produces a character dump.
Produces a decimal dump.
Produces an octal dump.
Produces a hexadecimal dump.

The -c option prints non-printable characters as a printable character preceded by a backslash: \0 is null, \b is backspace, \f is form-feed, \n is new-line, \r is return, and \t is tab.

The <offset> specifies where in the file the dump is to begin, if different than the beginning of the file. It is of the form [+]<n>[.][b]. The + is only necessary if you have no file specified so that the command interpreter knows this is the offset not the file. Without . or b, <n> indicates the dump starts at (octal) byte <n> of the file. A . displays <n> in decimal, a b in 512-byte blocks.

We recommend that you always pipe the output of od to less (or more) so that you can manipulate it. Large files can be unwieldy, and you may not be able to stop the output once it's going!

7.4.5 Determine File Type: file

The file utility can be used to determine the file type of a file according to its contents. It bases its guesses on a list of "magic numbers" recorded in a "magic file", /etc/magic. Some of the file types are:

file determines the filetype by looking at the beginning of the file and comparing it to entries in the magic file. The command format is:

% file <filename>...  

7.5 Manipulating Directories

This section describes the commands you can use to organize and use the UNIX directory structure. It describes how to make and remove directories, and move from one directory to another. Listing the contents of a directory (files and subdirectories) was described in a previous section. Section 7.6.2 Directory Permissions explains the meaning of access permissions as applied to directories.

7.5.1 Print Working Directory: pwd

The pwd command (for print working directory) displays the path name of your working (current) directory. The command format is:

% pwd  

7.5.2 List Directory Contents: ls

The ls command, which stands for list, is used to list the contents of a directory. ls has many options, some of which are system-dependent. A few of them are described in section 7.3.1 List Directory Contents: ls. For a complete description of the command, refer to the man pages for ls.

7.5.3 Change Directory: cd

When you first log in to the system, you are placed in your home directory, which is then also your current working directory. You can use the cd command (for change directory) to change your current working directory. The command format is:

% cd [<directory>]  

You can specify a complete path or a relative path. You can use .. (for the parent directory) in your pathname. You must have execute permission (which provides search permission in this case) on a directory to cd to it.

If <directory> is not specified, you are returned to your home directory.

The following examples illustrate moving to different directories:

% cd 
% cd Tools  
% cd /usr/jones  
% cd ~jones/ourfiles  
% cd ../Tools  

7.5.4 Make a Directory: mkdir

The mkdir command (for make directory) is used to create a directory. The command format is:

% mkdir <dirname> ...  

If a pathname is not specified, the directory is created as a subdirectory of the current working directory. Directory creation requires write access to the parent directory. The owner ID and group ID of the new directory are set to those of the creating process.


% mkdir progs  
% mkdir /usr/nicholls/Tools/Less  

7.5.5 Copy a Directory

The most straightforward way of copying a directory and its contents is to pipe the output of the ls command (see section 7.3.1 List Directory Contents: ls) into the file copy facility cpio (see the man pages). However this technique does not copy subdirectories.

First, create the destination directory using mkdir (see section 7.5.4 Make a Directory: mkdir), if it doesn't already exist. Secondly, from the source directory, run the command (shown with recommended options; see man pages for option information):

% ls | cpio -dumpV <destination_dir> 

The <destination_dir> must be specified relative to the source directory.

If you need to copy a directory structure, then use the tar utility instead. It is described in section 7.3.7 Copy to/Restore from Archive or Tape: tar. The following sequence of commands (shown on a single line) copies a structure from the source directory to the destination directory. The <destination_dir> is taken as relative to the <source_dir>:

% cd <source_dir>; tar cf - . | (cd <destination_dir>; tar xfBp -) 

The "-" is used for the name of the tar file (argument to the f option) so that tar writes to the standard output or reads from the standard input, as appropriate.

7.5.6 Move (Rename) a Directory: mv or mvdir

See section 7.3.4 Move or Rename a File: mv for information on mv. To move a directory (<olddirname>) and its contents to a different position in the directory tree, use the command format:

% mvdir <olddirname> <newdirpath> 

If <newdirpath> exists already, then the directory gets moved to <newdirpath>/<olddirname>. Note that the two arguments cannot be in the same path. For example:

% mvdir x/y x/z 

is ok, but

% mvdir x/y x/y/z 

is not ok.

7.5.7 Remove a Directory: rmdir

You can remove a directory with the rmdir command. The directory must contain no files or subdirectories, and you must have write permission to the parent directory.

% rmdir <dirname> ...  

You can use an absolute or relative pathname.

You can also use rm -r as described in Section 7.3.6 Remove a File: rm. rm -r will delete a directory, all subdirectories, and all files. This command should be used with extreme caution.

For example, the following command deletes the directory temp, all subdirectories of temp and all files contained in those directories, prompting before each removal, and confirming removal of write-protected files (-i):

% rm -ir /usr/jones/temp  

7.6 File and Directory Permissions

7.6.1 File Access Permissions

The UNIX file system allows you to control read, write, and execute access to your files on the basis of user (owner), group, and other (everyone else).5 In this section we will consider only the standard UNIX file permissions.

Note that in the AFS file system, file permissions are mediated by Access Control Lists (ACLs) that are set on a directory level. The standard UNIX file permissions don't apply in this case except for the owner permissions, which apply to all users. AFS file permissions are treated in section 8.6 File and Directory Permissions.

To determine the current permissions, use the long form of the ls command, ls -l. Referring to the example below, the nine characters immediately following the first field represent the one-bit flags known as the mode bits that control file access. A dash indicates a bit is not set, r stands for read access, w for write access, and x for execution access. The first set of three characters refer to owner permission, the middle three for group permission, and the last three for all other user classes.
total 251  
drwxr-xr-x 3 nicholls g020c 512   May 2  08:53 Tools  
drwxr-xr-x 2 nicholls g020c 512   May 2  09:01 bin  
-rw-r--r-- 1 nicholls g020c 446   May 4  14:09 defaults  
-rw-r--r-- 1 nicholls g020c 95418 May 1  17:42 intro.lpr  
-rw-r--r-- 1 nicholls g020c 0     May 10 17:51 lsout  
-rw-r--r-- 1 nicholls g020c 6683  May 1  16:46 man.lpr  
-rw-r--r-- 1 nicholls g020c 12258 May 9 16:16 out

In the example, ignoring the directory files (which have a d in position 1), the owner has rw access to the files, whereas group and others have read (r) access only.


The chmod command, which stands for change mode, is used to change access permissions of a file or directory:

% chmod <mode> <filename> ...     


% chmod <mode> <directory> ...  

In the absolute form of the mode where the level of protection is specified in octal format, <mode> looks like 741 or 554, for example, where each of the three octal numbers represents the sum of the permissions granted to its class: user, group, and other, in that order. The three types of permission have the values:

4 (100 octal)
2 (010 octal)
1 (001 octal)

For example, a mode of 741 means owner can read, write, and execute (4+2+1=7); group can read (4+0+0=4); and others can execute the file (0+0+1=1).

To give this permission to a file test, you would enter:

% chmod 741 test  

You can use an alternate form of <mode> in the chmod command in which <mode> is a three-character field specifying an action to be taken. The action is to add or subtract one or more permissions from one or more user classes. It takes the form:

<who> <operator> <permission(s)>  

These three positions within the field take the following characters:

represents the user class or classes; it takes any combination of u, g, o, and a for user (user is really the owner), group, other and all, respectively, where all includes the three individual classes
+ or - for adding or subtracting permissions, or = for setting a specific permission and resetting all other permissions for the specified user class(es)
any combination of r, w, and x for read, write, and execute, indicating the permissions to be permitted, denied, or reset.

Examples of the chmod command:

% chmod g-x progs 
% chmod g+rw,o-w out out1  
% chmod g=r myfile  

Note that classes of users or levels of protection not specified in a command are not modified in this form of the command (with the exception that = resets other permissions).


With the umask command you can specify a mask that the system uses to set access permissions when a file is created. In order to understand umask you need to know that access permission at file creation is application-dependent. Each command or application sets a file permission in its open command.6 The system then "subtracts" any user-defined mask, resulting in the final access permission for the file. You can set a umask by this command:

% umask [<ooo>]  

where <ooo> stands for three octal digits. The user-specified "mask", <ooo>, has the same positional structure as described above for chmod, but specifies permissions that should be removed (disallowed).

For example, a mask of 022 removes no permissions from owner, and removes write permission from group and others. Thus a file normally created with 777 would become 755 (this would appear as rwxr-xr-x in the format put out by the command ls -l). The following command could be put in your .cshrc or .profile.

% umask 022  

The meaning of permissions applied to directories is described in Section 7.6.2 Directory Permissions.

7.6.2 Directory Permissions

See section 8.6.2 Directory Permissions via Access Control Lists (ACLs) for AFS systems.

You can grant or deny permission for directories as well as files, and protection assigned to a directory file takes precedence over the permissions of individual files in the directory.

File access permissions of directory files are changed with the chmod command (see section 7.6.1 File Access Permissions).

7.7 Temporary Directories

By convention, there are directories named /tmp (and sometimes /usr/tmp) where programs and users can store temporary files. Many programs (e.g., compilers) write temporary files there or in the area specified by the environment variable TMPDIR. Since these are public areas, it is necessary to manage this space, which means that you cannot count on files being retained in these directories.

Many systems on site have fairly small /tmp areas and therefore you must be careful not to fill up this space. In general, you should only use /tmp for very temporary, small files. On many systems files in /tmp will disappear after a reboot or after existing for a week. You can set TMPDIR to a different location if there is not enough space in these areas.

Contact the administrators of the particular system to find out what the current policy is on the machine.

cd by itself is equivalent to cd $HOME.

echo is otherwise useful for sending messages to the terminal from a script and sending known data into a pipe.

There are better ways to display a file (see less and more, which follow cat in this section).

Under FullFUE on most systems (namely where gzip and gunzip are not installed in /usr/local/bin) you will need to run setup gtools in order to access them.

Note o is for other and not for owner as on VMS.

Normally only the loader creates files with execute permission.

View/print PDF file
CRL Product Home Page | CRL Manual Home Page | Computing Division | Fermilab at Work | Fermilab Home
This page generated on: 06/27/03 17:08:20