Before You Start
About this tutorial
Welcome to "Intermediate administration," the third of four tutorials designed to prepare you for the Linux Professional Institute's 101 (release 2) exam. This tutorial (Part 3) is ideal for those who want to improve their knowledge of fundamental Linux administration skills. We'll cover a variety of topics, including system and Internet documentation, the Linux permissions model, user account management, and login environment tuning.
If you are new to Linux, we recommend that you start with Part 1 and Part 2. For some, much of this material will be new, but more experienced Linux users may find this tutorial to be a great way of "rounding out" their foundational Linux system administration skills.
By the end of this series of tutorials (eight in all covering the LPI 101 and 102 exams), you will have the knowledge you need to become a Linux Systems Administrator and will be ready to attain an LPIC Level 1 certification from the Linux Professional Institute if you so choose.
System and network documentation
Types of Linux system documentation
There are essentially three sources of documentation on a Linux system: manual pages, info pages, and application-bundled documentation in /usr/share/doc. In this section, we'll explore each of these sources before looking "outside the box" for more information.
Manual pages, or "man pages", are the classic form of UNIX and Linux reference documentation. Ideally, you can look up the man page for any command, configuration file, or library routine. In practice, Linux is free software, and some pages haven't been written or are showing their age. Nonetheless, man pages are the first place to look when you need help.
To access a man page, simply type man followed by your topic of inquiry. A pager will be started, so you will need to press q when you're done reading. For example, to look up information about the ls command, you would type:
$ man ls
Knowing the layout of a man page can be helpful to jump quickly to the information you need. In general, you will find the following sections in a man page:
||Name and one-line description of the command
||How to use the command
||In-depth discussion on the functionality of the command
||Suggestions for how to use the command
||Related topics (usually man pages)
man page sections
The files that comprise manual pages are stored in /usr/share/man (or in /usr/man on some older systems). Inside that directory, you will find that the manual pages are organized into the following sections:
Multiple man pages
Some topics exist in more than one section. To demonstrate this, let's use the whatis command, which shows all the available man pages for a topic:
$ whatis printf
printf (1) - format and print data
printf (3) - formatted output conversion
In this case, man printf would default to the page in section 1 ("User Programs"). If we were writing a C program, we might be more interested in the page from section 3 ("Library functions"). You can call up a man page from a certain section by specifying it on the command line, so to ask for printf(3), we would type:
$ man 3 printf
Finding the right man page
Sometimes it's hard to find the right man page for a given topic. In that case, you might try using man -k to search the NAME section of the man pages. Be warned that it's a substring search, so running something like man -k ls will give you a lot of output! Here's an example using a more specific query:
$ man -k whatis
apropos (1) - search the whatis database for strings
makewhatis (8) - Create the whatis database
whatis (1) - search the whatis database for complete words
All about apropos
The example on the previous panel brings up a few more points. First, the apropos command is exactly equivalent to man -k. (In fact, I'll let you in on a little secret. When you run man -k, it actually runs apropos behind the scenes.) The second point is the makewhatis command, which scans all the man pages on your Linux system and builds the database for whatis and apropos. Usually this is run periodically by root to keep the database updated:
For more information on "man" and friends, you should start with its man page:
$ man man
By default, the man program will look for man pages in /usr/share/man, /usr/local/man, /usr/X11R6/man, and possibly /opt/man. Sometimes, you may find that you need to add an additional item to this search path. If so, simply edit /etc/man.conf in a text editor and add a line that looks like this:
From that point forward, any man pages in the /opt/man/man* directories will be found. Remember that you'll need to rerun makewhatis to add these new man pages to the whatis database.
One shortcoming of man pages is that they don't support hypertext, so you can't jump easily from one to another. The GNU folks recognized this shortcoming, so they invented another documentation format: "info" pages. Many of the GNU programs come with extensive documentation in the form of info pages. You can start reading info pages with the info command:
Calling info in this way will bring up an index of the available pages on the system. You can move around with the arrow keys, follow links (indicated with a star) using the Enter key, and quit by pressing q. The keys are based on Emacs, so you should be able to navigate easily if you're familiar with that editor. For an intro to the Emacs editor, see the developerWorks tutorial, Living in Emacs.
You can also specify an info page on the command line:
$ info diff
For more information on using the info reader, try reading its info page. You should be able to navigate primitively using the few keys I've already mentioned:
$ info info
There is a final source for help within your Linux system. Many programs are shipped with additional documentation in other formats: text, PDF, PostScript, HTML, to name a few. Take a look in usr/share/doc (or /usr/doc on older systems). You'll find a long list of directories, each of which came with a certain application on your system. Searching through this documentation can often reveal some gems that aren't available as man pages or info pages, such as tutorials or additional technical documentation. A quick check reveals there's a lot of reading material available:
$ cd /usr/share/doc
$ find . -type f | wc -l
Whew! Your homework this evening is to read just half (3791) of those documents. Expect a quiz tomorrow. ;-)
The Linux Documentation Project
In addition to system documentation, there are a number of excellent Linux resources on the Internet. The Linux Documentation Project is a group of volunteers who are working on putting together the complete set of free Linux documentation. This project exists to consolidate various pieces of Linux documentation into a location that is easy to search and use.
An LDP overview
The LDP is made up of the following areas:
- Guides - longer, more in-depth books, such as The Linux Programmer's Guide
- HOWTOs - subject-specific help, such as the DSL HOWTO
- FAQs - Frequently Asked Questions with answers, such as the Brief Linux FAQ
- Man pages - help on individual commands (these are the same manual pages you get on your Linux system when you use the man command).
If you aren't sure which section to peruse, you can take advantage of the search box, which allows you to find things by topic.
The LDP additionally provides a list of Links and Resources such as Linux Gazette (see links in Resources) and Linux Weekly News, as well links to mailing lists and news archives.
Mailing lists provide probably the most important point of collaboration for Linux developers. Often projects are developed by contributors who live far apart, possibly even on opposite sides of the globe. Mailing lists provide a method for each developer on a project to contact all the others, and to hold group discussions via e-mail. One of the most famous development mailing lists is the Linux Kernel Mailing List.
More about mailing lists
In addition to development, mailing lists can provide a method for asking questions and receiving answers from knowledgeable developers, or even other users. For example, individual distributions often provide mailing lists for newcomers. You can check your distribution's Web site for information on the mailing lists it provides.
If you took the time to read the LKML FAQ at the link on the previous panel, you might have noticed that mailing list subscribers often don't take kindly to questions being asked repeatedly. It's always wise to search the archives for a given mailing list before writing your question. Chances are, it will save you time, too!
Internet "newsgroups" are similar to mailing lists, but are based on a protocol called NNTP ("Network News Transfer Protocol") instead of e-mail. To participate, you need to use an NNTP client such as slrn or pan. The primary advantage is that you only take part in the discussion when you want, instead of having it continually arrive in your inbox. :-)
The newsgroups of primary interest start with comp.os.linux. You can browse the list on the LDP site.
Vendor and third-party Web sites
Web sites for the various Linux distributions often provide updated documentation, installation instructions, hardware compatibility/incompatibility statements, and other support such as a knowledge base search tool. For example:
Hardware and software vendors
Many hardware and software vendors have added Linux support to their products in recent years. At their sites, you can find information about which hardware supports Linux, software development tools, released sources, downloads of Linux drivers for specific hardware, and other special Linux projects. For example:
The Linux permissions model
One user, one group
In this section, we'll take a look at the Linux permissions and ownership model. We've already seen that every file is owned by one user and one group. This is the very core of the permissions model in Linux. You can view the user and group of a file in a ls -l listing:
$ ls -l /bin/bash
-rwxr-xr-x 1 root wheel 430540 Dec 23 18:27 /bin/bash
In this particular example, the /bin/bash executable is owned by root and is in the wheel group. The Linux permissions model works by allowing three independent levels of permission to be set for each filesystem object -- those for the file's owner, the file's group, and all other users.
Understanding "ls -l"
Let's take a look at our ls -l output and inspect the first column of the listing:
$ ls -l /bin/bash
-rwxr-xr-x 1 root wheel 430540 Dec 23 18:27 /bin/bash
This first field -rwxr-xr- contains a symbolic representation of this particular files' permissions. The first character (-) in this field specifies the type of this file, which in this case is a regular file. Other possible first characters:
||character special device
||block special device
$ ls -l /bin/bash
-rwxr-xr-x 1 root wheel 430540 Dec 23 18:27 /bin/bash
The rest of the field consists of three character triplets. The first triplet represents permissions for the owner of the file, the second represents permissions for the file's group, and the third represents permissions for all other users:
Above, the r means that reading (looking at the data in the file) is allowed, the w means that writing (modifying the file, as well as deletion) is allowed, and the x means that "execute" (running the program) is allowed. Putting together all this information, we can see that everyone is able to read the contents of and execute this file, but only the owner (root) is allowed to modify this file in any way. So, while normal users can copy this file, only root is allowed to update it or delete it.
Who am I?
Before we take a look at how to change the user and group ownership of a file, let's first take a look at how to learn your current user id and group membership. Unless you've used the su command recently, your current user id is the one you used to log in to the system. If you use su frequently, however, you may not remember your current effective user id. To view it, type whoami:
# su drobbins
What groups am I in?
To see what groups you belong to, use the groups command:
drobbins wheel audio
In the above example, I'm a member of the drobbins, wheel, and audio groups. If you want to see what groups other user(s) are in, specify their usernames as arguments:
$ groups root daemon
root : root bin daemon sys adm disk wheel floppy dialout tape video
daemon : daemon bin adm
Changing user and group ownership
To change the owner or group of a file or other filesystem object, use chown or chgrp, respectively. Each of these commands takes a name followed by one or more filenames.
# chown root /etc/passwd
# chgrp wheel /etc/passwd
You can also set the owner and group simultaneously with an alternate form of the chown command:
# chown root:wheel /etc/passwd
You may not use chown unless you are the superuser, but chgrp can be used by anyone to change the group ownership of a file to a group to which they belong.
Recursive ownership changes
Both chown and chgrp have a -R option that can be used to tell them to recursively apply ownership and group changes to an entire directory tree. For example:
# chown -R drobbins /home/drobbins
chown and chgrp can be used to change the owner and group of a filesystem object, but another program, called chmod, is used to change the rwx permissions that we can see in an ls -l listing. chmod takes two or more arguments: a "mode", describing how the permissions should be changed, followed by a file or list of files that should be affected:
$ chmod +x scriptfile.sh
In the above example, our "mode" is +x. As you might guess, a +x mode tells chmod to make this particular file executable for both the user and group and for anyone else.
If we wanted to remove all execute permissions of a file, we'd do this:
$ chmod -x scriptfile.sh
So far, our chmod examples have affected permissions for all three triplets -- the user, the group, and all others. Often, it's handy to modify only one or two triplets at a time. To do this, simply specify the symbolic character for the particular triplets you'd like to modify before the + or - sign. Use u for the "user" triplet, g for the "group" triplet, and o for the "other/everyone" triplet:
$ chmod go-w scriptfile.sh
We just removed write permissions for the group and all other users, but left "owner" permissions untouched.
In addition to flipping permission bits on and off, we can also reset them altogether. By using the = operator, we can tell chmod that we want the specified permissions and no others:
$ chmod =rx scriptfile.sh
Above, we just set all "read" and "execute" bits, and unset all "write" bits. If you just want to reset a particular triplet, you can specify the symbolic name for the triplet before the = as follows:
$ chmod u=rx scriptfile.sh
Up until now, we've used what are called symbolic modes to specify permission changes to chmod. However, there's another common way of specifying permissions: using a 4-digit octal number. Using this syntax, called numeric permissions syntax, each digit represents a permissions triplet. For example, in 1777, the 777 sets the "owner", "group", and "other" flags that we've been discussing in this section. The 1 is used to set the special permissions bits, which we'll cover later (see " The elusive first digit" at the end of this section). This chart shows how the second through fourth digits (777) are interpreted:
Numeric permission syntax
Numeric permission syntax is especially useful when you need to specify all permissions for a file, such as in the following example:
$ chmod 0755 scriptfile.sh
$ ls -l scriptfile.sh
-rwxr-xr-x 1 drobbins drobbins 0 Jan 9 17:44 scriptfile.sh
In this example, we used a mode of 0755, which expands to a complete permissions setting of -rwxr-xr-x.
When a process creates a new file, it specifies the permissions that it would like the new file to have. Often, the mode requested is 0666 (readable and writable by everyone), which is more permissive that we would like. Fortunately, Linux consults something called a "umask" whenever a new file is created. The system uses the umask value to reduce the originally specified permissions to something more reasonable and secure. You can view your current umask setting by typing umask at the command line:
On Linux systems, the umask normally defaults to 0022, which allows others to read your new files (if they can get to them) but not modify them.
To make new files more secure by default, you can change the umask setting:
$ umask 0077
This umask will make sure that the group and others will have absolutely no permissions for any newly created files. So, how does the umask work? Unlike "regular" permissions on files, the umask specifies which permissions should be turned off. Let's consult our mode-to-digit mapping table so that we can understand what a umask of 0077 means:
Using our table, the last three digits of 0077 expand to ---rwxrwx. Now, remember that the umask tells the system which permissions to disable. Putting two and two together, we can see that all "group" and "other" permissions will be turned off, while "user" permissions will remain untouched.
Introducing suid and sgid
When you initially log in, a new shell process is started. You already know that, but you may not know that this new shell process (typically bash) runs using your user id. As such, the bash program can access all files and directories that you own. In fact, we as users are totally dependent on other programs to perform operations on our behalf. Because the programs you start inherit your user id, they cannot access any filesystem objects for which you haven't been granted access.
For example, the passwd file cannot be changed by normal users directly, because the "write" flag is off for every user except root:
$ ls -l /etc/passwd
-rw-r--r-- 1 root wheel 1355 Nov 1 21:16 /etc/passwd
However, normal users do need to be able to modify /etc/passwd (at least indirectly) whenever they need to change their password. But, if the user is unable to modify this file, how exactly does this work?
Thankfully, the Linux permissions model has two special bits called suid and sgid. When an executable program has the suid bit set, it will run on behalf of the owner of the executable, rather than on behalf of the person who started the program.
Now, back to the /etc/passwd problem. If we take a look at the passwd executable, we can see that it's owned by root:
$ ls -l /usr/bin/passwd
-rwsr-xr-x 1 root wheel 17588 Sep 24 00:53 /usr/bin/passwd
You'll also note that in place of an x in the user's permission triplet, there's an s. This indicates that, for this particular program, the suid and executable bits are set. Because of this, when passwd runs, it will execute on behalf of the root user (with full superuser access) rather than that of the user who ran it. And because passwd runs with root access, it's able to modify the /etc/passwd file with no problem.
We've seen how suid works, and sgid works in a similar way. It allows programs to inherit the group ownership of the program rather than that of the current user.
Here's some miscellaneous yet important information about suid and sgid. First, suid and sgid bits occupy the same space as the x bits in a ls -l listing. If the x bit is also set, the respective bits will show up as s (lowercase). However, if the x bit is not set, it will show up as a S (uppercase).
Another important note: suid and sgid come in handy in many circumstances, but improper use of these bits can allow the security of a system to be breached. It's best to have as few suid programs as possible. The passwd command is one of the few that must be suid.
Changing suid and sgid
Setting and removing the suid and sgid bits is fairly straightforward. Here, we set the suid bit:
# chmod u+s /usr/bin/myapp
And here, we remove the sgid bit from a directory. We'll see how the sgid bit affects directories in just a few panels:
# chmod g-s /home/drobbins
Permissions and directories
So far, we've been looking at permissions from the perspective of regular files. When it comes to directories, things are a bit different. Directories use the same permissions flags, but they are interpreted to mean slightly different things.
For a directory, if the "read" flag is set, you may list the contents of the directory; "write" means you may create files in the directory; and "execute" means you may enter the directory and access any sub-directories inside. Without the "execute" flag, the filesystem objects inside a directory aren't accessible. Without a "read" flag, the filesystem objects inside a directory aren't viewable, but objects inside the directory can still be accessed as long as someone knows the full path to the object on disk.
Directories and sgid
And, if a directory has the "sgid" flag enabled, any filesystem objects created inside it will inherit the group of the directory. This particular feature comes in handy when you need to create a directory tree to be used by a group of people that all belong to the same group. Simply do this:
# mkdir /home/groupspace
# chgrp mygroup /home/groupspace
# chmod g+s /home/groupspace
Now, any users in the group mygroup can create files or directories inside /home/groupspace, and they will be automatically assigned a group ownership of mygroup as well. Depending on the users' umask setting, new filesystem objects may or may not be readable, writable, or executable by other members of the mygroup group.
Directories and deletion
By default, Linux directories behave in a way that may not be ideal in all situations. Normally, anyone can rename or delete a file inside a directory, as long as they have write access to that directory. For directories used by individual users, this behavior is usually just fine.
However, for directories that are used by many users, especially /tmp and /var/tmp, this behavior can be bad news. Since anyone can write to these directories, anyone can delete or rename anyone else's files -- even if they don't own them! Obviously, it's hard to use /tmp for anything meaningful when any other user can type rm -rf /tmp/* at any time and destroy everyone's files.
Thankfully, Linux has something called the sticky bit. When /tmp has the sticky bit set (with a chmod +t), the only people who are able to delete or rename files in /tmp are the directory's owner (typically root), the file's owner, or root. Virtually all Linux distributions enable /tmp's sticky bit by default, but you may find that the sticky bit comes in handy in other situations.
The elusive first digit
And to conclude this section, we finally take a look at the elusive first digit of a numeric mode. As you can see, this first digit is used for setting the sticky, suid, and sgid bits:
Here's an example of how to use a 4-digit numeric mode to set permissions for a directory that will be used by a workgroup:
# chmod 1775 /home/groupfiles
As homework, figure out the meaning of the 1755 numeric permissions setting. :)
Linux account managment
In this section, we'll look at the Linux account management mechanism, starting with the /etc/passwd file, which defines all the users that exist on a Linux system. You can view your own /etc/passwd file by typing less /etc/passwd.
Each line in /etc/passwd defines a user account. Here's an example line from my /etc/passwd file:
As you can see, there is quite a bit of information on this line. In fact, each /etc/passwd line consists of multiple fields, each separated by a :.
The first field defines the username (drobbins), and the second field contains an x. On ancient Linux systems, this field contained an encrypted password to be used for authentication, but virtually all Linux systems now store this password information in another file.
The third field (1000) defines the numeric user id associated with this particular user, and the fourth field (1000) associates this user with a particular group; in a few panels, we'll see where group 1000 is defined.
The fifth field contains a textual description of this account -- in this case, the user's name. The sixth field defines this user's home directory, and the seventh field specifies the user's default shell -- the one that will be automatically started when this user logs in.
/etc/passwd tips and tricks
You've probably noticed that there are many more user accounts defined in /etc/passwd than actually log in to your system. This is because various Linux components use user accounts to enhance security. Typically, these system accounts have a user id ("uid") of under 100, and many of them will have something like /bin/false listed as a default shell. Since the /bin/false program does nothing but exit with an error code, this effectively prevents these accounts from being used as login accounts -- they are for internal use only.
So, user accounts themselves are defined in /etc/passwd. Linux systems contain a companion file to /etc/passwd that's called /etc/shadow. This file, unlike /etc/passwd, is readable only by root and contains encrypted password information. Let's look at a sample line from /etc/shadow:
Each line defines password information for a particular account, and again, each field is separated by a :. The first field defines the particular user account with which this shadow entry is associated. The second field contains an encrypted password. The remaining fields are described in the following table:
||# of days since 1/1/1970 that the password was modified
||# of days before password will be allowed to be changed (0 for "change anytime")
||# of days before system will force user to change to a new password (-1 for "never")
||# of days before password expires that user will be warned about expiration (-1 for "no warning")
||# of days after password expiration that this account is automatically # disabled by the system (-1 for "never disable")
||# of days that this account has been disabled (-1 for "this account is enabled")
||Reserved for future use
Next, we take a look at the /etc/group file, which defines all the groups on a Linux system. Here's a sample line:
The /etc/group field format is as follows. The first field defines the name of the group; the second field is a vestigial password field that now simply holds an x, and the third field defines the numeric group id of this particular group. The fourth field (empty in the above example) defines any users that are members of this group.
You'll recall that our sample /etc/passwd line referenced a group id of 1000. This has the effect of placing the drobbins user in the drobbins group, even though the drobbins username isn't listed in the fourth field of /etc/group.
A note about associating users with groups: on some systems, you'll find that every new login account is associated with an identically named (and usually identically numbered) group. On other systems, all login accounts will belong to a single users group. The approach that you use on the system(s) you administrate is up to you. Creating matching groups for each user has the advantage of allowing users to more easily control access to their own files by placing trusted friends in their personal group.
Adding a user and group by hand
Now, I'll show you how to create your own user and group account. The best way to learn how to do this is to add a new user to the system manually. To begin, first make sure that your EDITOR environment variable is set to your favorite text editor:
# echo $EDITOR
If it isn't, you can set EDITOR by typing something like:
# export EDITOR=/usr/bin/emacs
You should now find yourself in your favorite text editor with the /etc/passwd file loaded up on the screen. When modifying system passwd and group files, it's very important to use the vipw and vigr commands. They take extra precautions to ensure that your critical passwd and group files are locked properly so they don't become corrupted.
Now that you have the /etc/passwd file up, go ahead and add the following line:
testuser:x:3000:3000:LPI tutorial test user:/home/testuser:/bin/false
We've just added a "testuser" user with a UID of 3000. We've added him to a group with a GID of 3000, which we haven't created just yet. Alternatively, we could have assigned this user to the GID of the users group if we wanted. This new user has a comment that reads LPI tutorial test user; the user's home directory is set to /home/testuser, and the user's shell is set to /bin/false for security purposes. If we were creating an non-test account, we would set the shell to /bin/bash. OK, go ahead and save your changes and exit.
Now, we need to add an entry in /etc/shadow for this particular user. To do this, type vipw -s. You'll be greeted with your favorite editor, which now contains the /etc/shadow file. Now, go ahead and copy the line of an existing user account (one that has a password and is longer than the standard system account entries):
Now, change the username on the copied line to the name of your new user, and ensure that all fields (particularly the password aging ones) are set to your liking:
Now, save and exit.
Setting a password
You'll be back at the prompt. Now, it's time to set a password for your new user:
# passwd testuser
Enter new UNIX password: (enter a password for testuser)
Retype new UNIX password: (enter testuser's new password again)
Now that /etc/passwd and /etc/shadow are set up, it's now time to get /etc/group configured properly. To do this, type:
Your /etc/group file will appear in front of you, ready for editing. Now, if you chose to assign a default group of users for your particular test user, you do not need to add any groups to /etc/groups. However, if you chose to create a new group for this user, go ahead and add the following line:
Now save and exit.
Creating a home directory
We're nearly done. Type the following commands to create testuser's home directory:
# cd /home
# mkdir testuser
# chown testuser.testuser testuser
# chmod o-rwx testuser
Our user's home directory is now in place and the account is ready for use. Well, almost ready. If you'd like to use this account, you'll need to use vipw to change testuser's default shell to /bin/bash so that the user can log in.
Account admin utils
Now that you know how to add a new account and group by hand, let's review the various time-saving account administration utilities available under Linux. Due to space constraints, we won't cover a lot of detail describing these commands. Remember that you can always get more information about a command by viewing the command's man page. If you are planning to take the LPIC 101 exam, you should spend some time getting familiar with each of these commands.
||By default, any files that a user creates are assigned to the user's group specified in /etc/passwd. If the user belongs to other groups, he or she can type newgrp thisgroup to set current default group membership to the group thisgroup. Then, any new files created will inherit thisgroup membership.
||The chage command is used to view and change the password aging setting stored in /etc/shadow.
||A general-purpose group administration tool.
||Used to add/delete/modify groups in /etc/group
||Used to add/delete/modify users in /etc/passwd. These commands also perform various other convenience functions. See the man pages for more information.
||Used to convert passwd and group files to "new-style" shadow passwords. Virtually all Linux systems already use shadow passwords, so you should never need to use these commands.
Tuning the user environment
Your shell has many useful options that you can set to fit your personal preferences. So far, however, we haven't discussed any way to have these settings set up automatically every time you log in, except for re-typing them each time. In this section we will look at tuning your login environment by modifying startup files.
First, let's add a friendly message for when you first log in. To see an example message, run fortune:
No amount of careful planning will ever replace dumb luck.
Now, let's set up fortune so that it gets run every time you log in. Use your favorite text editor to edit a file named .bash_profile in your home directory. If the file doesn't exist already, go ahead and create it. Insert a line at the top:
Try logging out and back in. Unless you're running a display manager like xdm, gdm, or kdm, you should be greeted cheerfully when you log in:
mycroft.flatmonk.org login: chouser
Freedom from incrustations of grime is contiguous to rectitude.
The login shell
When bash started, it walked through the .bash_profile file in your home directory, running each line as though it had been typed at a bash prompt. This is called sourcing the file.
Bash acts somewhat differently depending on how it is started. If it is started as a login shell, it will act as it did above -- first sourcing the system-wide /etc/profile, and then your personal ~/.bash_profile.
There are two ways to tell bash to run as a login shell. One way is used when you first log in: bash is started with a process name of -bash. You can see this in your process listing:
$ ps u
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
chouser 404 0.0 0.0 2508 156 tty2 S 2001 0:00 -bash
You will probably see a much longer listing, but you should have at least one COMMAND with a dash before the name of your shell, like -bash in the example above. This dash is used by the shell to determine if it's being run as a login shell.
The second way to tell bash to run as a login shell is with the --login command-line option. This is sometimes used by terminal emulators (like xterm) to make their bash sessions act like initial login sessions.
After you have logged in, more copies of your shell will be run. Unless they are started with --login or have a dash in the process name, these sessions will not be login shells. If they give you a prompt, however, they are called interactive shells. If bash is started as interactive, but not login, it will ignore /etc/profile and ~/.bash_profile and will instead source ~/.bashrc.
Testing for interactivity
Sometimes bash sources your ~/.bashrc, even though it isn't really interactive, such as when using commands like rsh and scp. This is important to keep in mind because printing out text, like we did with the fortune command earlier, can really mess up these non-interactive bash sessions. It's a good idea to use the PS1 variable to detect whether the current shell is truly interactive before printing text from a startup file:
if [ -n "$PS1" ]; then
/etc/profile and /etc/skel
As a system administrator, you are in charge of /etc/profile. Since it is sourced by everyone when they first log in, it is important to keep it in working order. It is also a powerful tool in making things work correctly for new users as soon as they log into their new account.
However, there are some settings that you may want new users to have as defaults, but also allow them to change easily. This is where the /etc/skel directory comes in. When you use the useradd command to create a new user account, it copies all the files from /etc/skel into the user's new home directory. That means you can put helpful .bash_profile and .bashrc files in /etc/skel to get new users off to a good start.
Variables in bash can be marked so that they are set the same in any new shells that it starts; this is called being marked for export. You can have bash list all of the variables that are currently marked for export in your shell session:
declare -x EDITOR="vim"
declare -x HOME="/home/chouser"
declare -x MAIL="/var/spool/mail/chouser"
declare -x PAGER="/usr/bin/less"
declare -x PATH="/bin:/usr/bin:/usr/local/bin:/home/chouser/bin"
declare -x PWD="/home/chouser"
declare -x TERM="xterm"
declare -x USER="chouser"
Marking variables for export
If a variable is not marked for export, any new shells that it starts will not have that variable set. However, you can mark a variable for export by passing it to the export built-in:
$ export BAR
$ echo $FOO $BAR
$ echo $FOO $BAR
In this example, the variables FOO and BAR were both set, but only BAR was marked for export. When a new bash was started, it had lost the value for FOO. If you exit this new bash, you can see that the original one still has values for both FOO and BAR:
$ echo $FOO $BAR
Export and set -x
Because of this behavior, variables can be set in ~/.bash_profile or /etc/profile and marked for export, and then never need to be set again. There are some options that cannot be exported, however, and so they must be put in your ~/.bashrc and your profile in order to be set consistently. These options are adjusted with the set built-in:
$ set -x
The -x option causes bash to print out each command it is about to run:
$ echo $FOO
$ echo foo
This can be very useful for understanding unexpected quoting behavior or similar strangeness. To turn off the -x option, do set +x. See the bash man page for all of the options to the set built-in.
Setting variables with "set"
The set built-in can also be used for setting variables, but when used that way, it is optional. The bash command set FOO=foo means exactly the same as FOO=foo. Un-setting a variable is done with the unset built-in:
$ echo $FOO
$ unset FOO
$ echo $FOO
Unset vs. FOO=
This is not the same as setting a variable to nothing, although it is sometimes hard to tell the difference. One way to tell is to use the set built-in with no parameters to list all current variables:
$ set | grep ^FOO
$ set | grep ^FOO
$ unset FOO
$ set | grep ^FOO
Using set with no parameters like this is similar to using the export built-in, except that set lists all variables instead of just those marked for export.
Exporting to change command behavior
Often, the behavior of commands can be altered by setting environment variables. Just as with new bash sessions, other programs that are started from your bash prompt will only be able to see variables that are marked for export. For example, the command man checks the variable PAGER to see what program to use to step through the text one page at a time.
$ export PAGER
$ man man
With PAGER set to less, you will see one page at a time, and pressing the space bar moves on to the next page. If you change PAGER to cat, the text will be displayed all at once, without stopping.
$ man man
Unfortunately, if you forget to set PAGER back to less, man (as well as some other commands) will continue to display all their text without stopping. If you wanted to have PAGER set to cat just once, you could use the env command:
$ env PAGER=cat man man
$ echo $PAGER
This time, PAGER was exported to man with a value of cat, but the PAGER variable itself remained unchanged in the bash session.
Summary and resources
Congratulations on finishing Part 3 of this tutorial series! At this point, you should know how to locate information in system and Internet documentation, and you should have a good grasp of the Linux permissions model, user account management, and login environment tuning.
Be sure to check out the various Linux documentation resources covered in this tutorial -- particularly the Linux Documentation Project. You'll find its collection of guides, HOWTOs, FAQs, and man pages to be invaluable. Be sure to check out Linux Weekly News as well.
The Linux System Administrators guide (available from the "Guides" section at www.tldp.org) is a good complement to this series of tutorials -- give it a read! You may also find Eric S. Raymond's Unix and Internet Fundamentals HOWTO to be helpful.
Check out the other articles in this series:
In the "Bash by Example" article series, Daniel shows you how to use bash programming constructs to write your own bash scripts. This series (particularly Parts 1 and 2) will be good preparation for the LPIC Level 1 exam:
You can learn more about sed in the Sed by Example article series. If you're planning to take the LPI exam, be sure to read the first two articles of this series.
To learn more about awk, read the Awk by Example article series.
If you're not too familiar with the vi editor, I strongly recommend that you check out my Vi -- the cheat sheet method tutorial. This tutorial will give you a gentle yet fast-paced introduction to this powerful text editor. Consider this must-read material if you don't know how to use vi.