LDAP Authentication for Linux
2002 by metaconsultancy

LDAP is a directory server technology that allows information such as usernames and passwords for an entire site to be stored on a central server. This whitepapers describes how to set up a Linux workstation to use an LDAP server for user information and authentication.

Before proceeding, you will need a working LDAP server which can provide you with user information. If you need to set one up, consult our OpenLDAP whitepaper for instructions.

User information consists of such data as mappings between user id numbers and user names (used, for example, by ls -l), or home directory locations (used, for example, by cd ~). Lookups of such information are handled by the name service subsystem, configured in the file /etc/nsswitch.conf. Authentication (password checking), on the other hand, is handled by the PAM (plugable authentication module) subsystem, configured in the /etc/pam.d/ directory. While these two subsystems can (in fact must) be configured seperately, you will likely want both to use LDAP.


Begin by installing the shared library code necessary for the name service to use ldap.

# apt-get install libnss-ldap

Next, open the /etc/nsswitch.conf file, and tell the name service subsystem to use LDAP to obtain user information.

passwd:    files ldap
group:     files ldap
shadow:    files ldap		
Note that we do not eliminate the use of flat files, since some users and groups (e.g. root) will remain local. If your machines do not use flat files at all and your LDAP server goes down, not even root will be able to log in.

Finally, you need to tell then name service subsystem how to talk to your LDAP server. This is done in the file /etc/libnss-ldap.conf.

uri ldap://ldap.example.com/ ldap://ldap-backup.example.com/
base dc=example, dc=org
The uri directive specifies the domain name (or IP address) of your LDAP server. As our example illustrates, you can specify multiple LDAP servers, in which case they will be employed in failover fashion. The base directive specifies the root DN at which searches should start. For additional information on these and other configuration directives, man libnss-ldap.conf.

nss-ldap expects accounts to be objects with the following attributes: uid, uidNumber, gidNumber, homeDirectory, and loginShell. These attributes are allowed by the objectClass posixAccount.

There is a simple way to verify that your name service subsystem is using your LDAP server as instructed. Assign a file to be owned by a user that exists only in the LDAP database, not in /etc/passwd. If an ls -l correctly shows the username, then the name service subsystem is consulting the LDAP database; if it just shows the user number, something is wrong. For example, if the user john, with user number 1001, exists only in LDAP, we can try

# touch /tmp/test
# chown 1001 /tmp/test 
# ls -l /tmp/test
-rw-r-----     1 john     users         0 Jan  1 12:00 test
to determine whether the the name service is using LDAP.

Next we configure the PAM subsystem to use LDAP for passwords. Begin by installing the necessary PAM module.

# apt-get install libpam-ldap
The configuration file for the pam_ldap.so module is /etc/pam_ldap.conf.
uri ldaps://ldap.example.com/
base dc=example,dc=com
pam_password exop
The uri and base directives work the same way they do for /etc/libnss_ldap.conf and /etc/ldap/ldap.conf. Notice that we have used ldaps to ensure that connections over which passwords are exchanged are encrypted. The directive "pam_password exop" tells pam-ldap to change passwords in a way that allows OpenLDAP to apply the hashing algorithm specified in /etc/ldap/slapd.conf, instead of attempting to hash locally and write the result directly into the database.

pam-ldap assumes accounts to be ojbects with the following attributes: uid and userPassword. The attributes are allowed by the objectClass posixAccount.

We are now ready to configure individual services to use the LDAP server for password checking. Each service that uses PAM for authentication has its own configuration file /etc/pam.d/service. To configure a service to use LDAP for password-checking, you must modify its PAM configuration file.

To avoid an in-depth explanation of PAM, we will content ourselves with a few examples. Consider first the login program, which handles logins from the text console. A typical PAM stack which checks passwords both in /etc/passwd and in the LDAP database follows.

auth        required      pam_nologin.so
auth        sufficient    pam_ldap.so
auth        sufficient    pam_unix.so shadow use_first_pass
auth        required      pam_deny.so
After successful password authentication using the auth stack, login checks for the existance of an account using the account stack, so it is necessary to reference pam-ldap there, too.
account     sufficient    pam_unix.so
account     sufficient    pam_ldap.so
account     required      pam_deny.so
Other login-like programs include xdm and gdm (for graphical logins), ssh (for remote logins), su (for switching programs), and xlock and xscreensaver (for locked screens). Each has its own file in /etc/pam.d/.

Some applications not only authenticate passwords, but can also be used to change them. The prototypical example is of course passwd, the standard password-changing utility. Such programs can be configured to use LDAP by modifying their password stack.

password    required      pam_cracklib.so
password    sufficient    pam_ldap.so
password    sufficient    pam_unix.so
password    required      pam_deny.so

One convienient application of pam-ldap is to set up "black box" servers that can authenticate users for a particular service without having an account on the machine at all. Services such as netatalk, (Cyrus) imap, and (Postfix) smtp use PAM. By configuring their PAM stacks to use LDAP, while leaving LDAP out of the PAM stacks of services such as login and ssh, you can easily create a "black box" server.


To keep your computers from pounding your LDAP server every time a command such as ls -l /home is issued on a computer in your organization, it is a good idea to configure your workstations to cache some user data. As long as the data in the cache is sufficiently fresh, the workstations use in instead of asking your LDAP server again. The name server caching daemon (nscd) accomplishes exactly this task.

To install nscd on Debian, just

# apt-get install nscd

The configuration file for nscd is /etc/nscd.conf.

enable-cache            passwd          yes
positive-time-to-live   passwd          600
negative-time-to-live   passwd          20
suggested-size          passwd          211
check-files             passwd          yes