Presented by Chris Wilson at AfNOG 2008.
Original author unknown, probably Ayitey Bulley, presented at AfNOG 2007.
You've seen how to get incoming email onto your server from the Internet. How do your users get access to their mail on your server?
POP3 is the oldest protocol. You can list messages in your mailbox, download them and delete them, and that's about it. Normally used when you want to keep all your mail on a local computer (not on the server).
Advantages:
Disadvantages:
Example clients:
IMAP was designed to replace POP3. It has many more features and is designed to help clients (MUAs) to manage mail on the store, and synchronise local copies with the store. Mail is normally kept on the server, at least until your mailbox is nearly full, when users should move it to offline or archive storage to make space.
Advantages:
Disadvantages:
Example clients:
Webmail has been used for years to provide users with access to mail, without installing any software (clients) or configuration. Users read their mail in their browser. The server displays lists of messages and message contents as web pages, and allows download of attachments through the browser. You have probably used webmail if you use Gmail, Yahoo or Hotmail from an Internet cafe.
Advantages:
Disadvantages:
Microsoft Exchange is a mail server (groupware) that is closely integrated with Microsoft's desktop mail client, Outlook. Users are familiar with the interface and it's popular. It integrates shared and private calendars, address books and To-Do lists.
Advantages:
Disadvantages
Because Exchange offers nothing more than IMAP for open clients such as FreeBSD, and requires a Windows server, we will not cover it further.
What services do we want to provide?
For scalability, we are going to arrange for exim to deliver all
local mail in Maildir format. This creates a subdirectory called
"Maildir" in the user's home directory, which in turn
contains three subdirectories: new,
cur and tmp.
Messages are written into tmp,
moved to new when
delivery is complete, and moved to cur
when read. Each message has a long unique filename based on the
hostname and the time of day.
Because each message is stored in a separate file, it is much faster for the pop3 daemon to start up every time a user connects. It also allows for safe delivery onto a shared (NFS) disk backend.
Exim does not include any software for retrieving mail from a mailbox, so we need to install additional software. Courier is a mail system which includes a number of packages. In fact it has its own MTA, but we will ignore this (it is still under heavy development, and does not have the flexibility needed for an ISP environment). The components we are interested in are the IMAP/POP3 servers and 'sqwebmail', the webmail server.
You can get the entire courier system as one package (including the MTA), or just the components. We will get the authlib, pop3/imap and webmail components separately.
Remember: in the command examples given below, commands shown with the prompt "$" should be run as your normal non-root userid. Only those commands with prompt "#" need to be run as root.
Edit /usr/exim/configure,
find the local_delivery
transport and modify it as follows:
local_delivery: driver = appendfile directory = $home/Maildir maildir_format maildir_use_size_file delivery_date_add envelope_to_add return_path_add # group = mail # mode = 0660
Optionally you could add further parameters to this transport which let you impose quotas on your users, for example to limit all users to 10 megabytes of storage each:
maildir_tag = ,S=$message_size quota_size_regex = ,S=(\d+) quota = 10M quota_warn_threshold = 90%
This quota mechanism relies on users not meddling with the quota information which is stored within their maildir; in other words, users with shell access would be able to bypass their quota if they knew what they were doing.
Remember to HUP your exim daemon (with the hupexim command)
after changing the configuration. Now test out your new
configuration by delivering to some local account on your machine:
$ /usr/exim/bin/exim -bt localuser localuser@pcnn.e1.ws.afnog.org router = localuser, transport = local_delivery $ /usr/exim/bin/exim localuser Here is a test . $ cd /home/localuser/Maildir $ ls cur new tmp $ ls new 102078119.7969.pcnn.e1.ws.afnog.org,S=426 $ cat new/* Return-path: <root@pcnn.e1.ws.afnog.org> ... Here is a test
Note: once you have changed to Maildir delivery, you will find that any local Unix MUA (which looks for new messages in /var/mail/username) will no longer see your incoming mail. How to fix this depends on which MUA you are using. Some examples:
/usr/local/etc/Muttrc
and add:
set spoolfile="~/Maildir/"
You can get the entire courier system as one package (including the MTA), or just the components. We will get the authlib, pop3/imap and webmail components separately.
As with most software packages under FreeBSD, you have a choice of installing directly from source, or using the ports system. If you install from source you have the most control over which version is installed and which compilation options are used.
However installing
from packages is easier, gives you a record of which files where
installed where, and installs the files in the "normal"
places you'd expect for a FreeBSD system. In particular, the commands
get installed in /usr/local/bin
and /usr/local/sbin,
which is already in your $PATH. We will use the ports in this lab.
The courier packages now share a single authentication library, courier-authlib. This package is responsible for looking up usernames and passwords - it can retrieve this information from various locations, including Unix system accounts (authpam), SQL databases (authmysql and authpgsql), LDAP databases (authldap), and local file databases (authuserdb). Having a separate package means that the same authentication configuration can now be shared by both POP3/IMAP and Webmail.
# cd /usr/ports/security/courier-authlib # make (select AUTH_USERDB & AUTH_VCHKPW from the popup menu) # make install # make clean (optional step - deletes temporary files in 'work' subdir)
courier-authlib runs a pool of authentication daemons which
perform the actual work; courier-imap and sqwebmail communicate with
these daemons via a socket. So the next thing we need to do is to
start the daemons. First you need to edit /etc/rc.conf:
# vi /etc/rc.conf add the following line: courier_authdaemond_enable="YES"
Courier-authlib itself has a single configuration file,
/usr/local/etc/authlib/authdaemonrc.
For the purposes of this exercise, we will turn on
authentication debugging.
# cd /usr/local/etc/authlib # vi authdaemonrc change this line: DEBUG_LOGIN=0 to: DEBUG_LOGIN=1
To save resources, you can also configure the authdaemond process not to try any authentication mechanisms which you know you don't need. For example, if all your authentication is only via PAM for Unix system passwords, then you can remove all the others:
... authmodulelist="authpam" ...
Now we are ready to start the authentication daemons:
# /usr/local/etc/rc.d/courier-authdaemond start Starting courier_authdaemond. (if you don't see this, check your /etc/rc.conf) # ps auxwww | grep authdaemond root 36787 0.0 0.2 1220 720 p1 S 10:40AM 0:00.00 /usr/local/sbin/courierlogger -pid=/usr/local/var/spool/authdaemon/pid -start /usr/local/libexec/courier-authlib/authdaemond root 36788 0.0 0.2 1464 880 p1 S 10:40AM 0:00.00 /usr/local/libexec/courier-authlib/authdaemond root 36789 0.0 0.2 1464 880 p1 S 10:40AM 0:00.00 /usr/local/libexec/courier-authlib/authdaemond root 36790 0.0 0.2 1464 880 p1 S 10:40AM 0:00.00 /usr/local/libexec/courier-authlib/authdaemond root 36791 0.0 0.2 1464 880 p1 S 10:40AM 0:00.00 /usr/local/libexec/courier-authlib/authdaemond root 36792 0.0 0.2 1464 880 p1 S 10:40AM 0:00.00 /usr/local/libexec/courier-authlib/authdaemond root 36793 0.0 0.2 1464 880 p1 S 10:40AM 0:00.00 /usr/local/libexec/courier-authlib/authdaemond
ps shows one courierlogger process, and six authdaemond processes (one master, five workers). If you didn't see "Starting courier_authdaemond" then you made a typing error.
You can test the authentication system by itself; the "authtest" command sends requests down the authentication socket, and displays the responses which come back. Test using any Unix login account which already exists on your system.
# authtest brian -- find an account called 'brian' # authtest brian foo -- check 'brian' has password 'foo' # authenumerate -- list all accounts
Try it also with a non-existent username, and with both the right password and a wrong password for an account, to confirm that passwords are being validated properly.
Because we enabled login debugging, you should find that each
authentication request generates detailled information in
/var/log/debug.log
showing how the request is passed to each module in turn. Have a look
in this file to confirm:
# less /var/log/debug.log
Further documentation for courier-authlib can be found on the web at
http://www.courier-mta.org/authlib/,
and is also installed in /usr/local/share/doc/courier-authlib/
Using ports, building courier-imap is straightforward:
# cd /usr/ports/mail/courier-imap # make (select AUTH_USERDB & AUTH_VCHKPW from the popup menu) # make install # make clean (optional step)
You can choose to run POP3, IMAP, or both. There is a configuration file for each one:
/usr/local/etc/courier-imap/pop3d /usr/local/etc/courier-imap/imapd
The default configuration is acceptable in most cases. However for a large server you may wish to increase the maximum number of concurrent connections from the default of 40, if you have fairly powerful hardware:
# cd /usr/local/etc/courier-imap # vi pop3d ... MAXDAEMONS=300 ... # vi imapd ... MAXDAEMONS=300 ...
Then, you need to enable the daemon(s) which you wish to run in
/etc/rc.conf
# vi /etc/rc.conf add the following line(s): courier_imap_pop3d_enable="YES" courier_imap_imapd_enable="YES"
And then run the startup script(s):
# /usr/local/etc/rc.d/courier-imap-pop3d.sh start Starting courier_imap_pop3d. # /usr/local/etc/rc.d/courier-imap-imapd.sh start Starting courier_imap_imapd.
Test using telnet: POP3 and IMAP are both text-based layer 7 protocols and you can drive them by hand.
# telnet localhost 110 Connected to localhost.ws.afnog.org Escape character is '^]'. +OK Hello there. user username +OK Password required. pass password +OK logged in. stat +OK 26 49857 retr 1 +OK 1073 octets follow. ... message . quit +OK Bye-bye. Connection closed by foreign host. # telnet localhost 143 Connected to localhost.ws.afnog.org. Escape character is '^]'. * OK [CAPABILITY IMAP4rev1 UIDPLUS CHILDREN NAMESPACE THREAD=ORDEREDSUBJECT THREAD=REFERENCES SORT QUOTA IDLE ACL ACL2=UNION STARTTLS] Courier-IMAP ready. Copyright 1998-2005 Double Precision, Inc. See COPYING for distribution information. a login username password a OK LOGIN Ok. a examine inbox * FLAGS (\Answered \Flagged \Deleted \Seen \Recent) * OK [PERMANENTFLAGS ()] No permanent flags permitted * 26 EXISTS * 0 RECENT * OK [UIDVALIDITY 989061119] Ok * OK [READ-ONLY] Ok a logout * BYE Courier-IMAP server shutting down a OK LOGOUT completed Connection closed by foreign host.
NOTE: The daemons will fail to login if the mail directory
does not exist, although current versions do now provide an error
message. Hence you need to have delivered at least one message to the
user, to create their mailbox, before they can login (or use the
'maildirmake' command to create it). Look for logging messages in
/var/log/maillog
and /var/log/debug.log.
If you wish, you can choose to allow pop3 over SSL (port 995) and imap over SSL (port 993). The advantage is that, for clients which support it, the traffic is encrypted. The disadvantage is higher CPU load on your server for the encryption of data.
To run IMAP and POP3 with SSL encryption (to protect passwords from eavesdroppers) you will need an encryption key and a certificate, just as we saw with Apache. You can use the same certificates that you generated earlier, just by copying them, as long as you access the host by the same name (e.g. localhost) that you placed in the certificate's Common Name.
If you want to generate a new certificate, follow these steps:
cd /usr/local/etc/courier-imap
# openssl genrsa -des3 -out server.key.generate 1024This creates a file called
server.key.generate, which
is protected by a password that you choose. You will have to enter
your chosen password twice. Please use a short temporary password.
Unless you remove the password, you will have to
enter it every time you start Courier. However, there is no easy way
to generate a key without a password.# openssl rsa -in server.key.generate -out server.keyEnter the same password as before, to decrypt the key. This creates a new file,
server.key, that contains the
same key but without any password protection.# openssl req -new -key server.key -out server.csrIt will ask you for country, address, etc. These will appear in the certificate when you inspect it in the browser. The Common Name must be the same name that you use to access the mail server (e.g.
localhost or
pc28.sse.ws.afnog.org). A certificate request file
(CSR) will be generated. The CSR is the file that you would send to
a Certificate Authority if you want to buy a real certificate.# openssl x509 -req -days 3650 \ -in server.csr -signkey server.key \ -out server.crtThis generates a signed certificate file (also called a cert) which the IMAP and POP servers can send to the client. Your clients should not trust this certificate, because it is not signed by a professional certificate authority. The clients should display a warning message whenever you connect to a server using a self-signed certificate like this.
Now you need to place the certificate and key file in the correct place
so that Courier IMAPd and POP3d can find them. Start by changing to the
directory that contains your certificates. If you are reusing the Apache
certificates, you should be in /usr/local/etc/apache22.
If you just generated new certificates in
/usr/local/etc/courier-imap, stay in that directory.
# cat server.crt server.key \ > /usr/local/share/courier-imap/imapd.pem # cat server.crt server.key \ > /usr/local/share/courier-imap/pop3d.pem
Now that you have an SSL certificate and key file, you can enable the SSL daemons in /etc/rc.conf:
# vi /etc/rc.conf courier_imap_pop3d_ssl_enable="YES" # pop3 over ssl, port 995 courier_imap_imapd_ssl_enable="YES" # imap over ssl, port 993
Then you start the servers:
# /usr/local/etc/rc.d/courier-imap-pop3d-ssl.sh start Starting courier_imap_pop3d_ssl. # /usr/local/etc/rc.d/courier-imap-imapd-ssl.sh start Starting courier_imap_imapd_ssl.
You can't use a regular telnet to test it, because all your communication needs to be encrypted, but openssl has an SSL client you can use to make an encrypted connection for testing:
# openssl s_client -connect localhost:995
Unlike many other webmail solutions, which use POP3 or IMAP to talk to the mail store, sqwebmail reads and writes Maildir directories directly. This makes it efficient in the case where POP/IMAP and webmail run on the same box, or where there is an NFS-shared mailstore.
sqwebmail is feature-rich, very customisable through HTML templates and stylesheets, supports multiple languages, and is simple to install (it runs as a single CGI). Note however that it is still under very active development and hence subject to change quite frequently.
You should have already installed Apache, but if not, please install it from ports:
# cd /usr/ports/www/apache22 # make install clean (You may see a menu at this point, leave all the options at their default, press [Tab] and [Enter] to highlight the [OK] button and continue). # vi /etc/rc.conf apache_enable="YES" # /usr/local/etc/rc.d/apache22 start Syntax OK Starting apache22.
Check your Apache install is working by pointing a web browser at http://localhost/. You should get a page saying "It works!".
Now install sqwebmail:
# cd /usr/ports/mail/sqwebmail # make WITH_CHARSET=all # make install # make clean (optional step)
The option "WITH_CHARSET=all" allows sqwebmail to view messages in a wide range of character sets. This increases the size of the binary by about one megabyte with the extra translation tables which are included.
The make command will display a menu with various options for the port. Press Tab and then Enter to accept the defaults.
One other change is required: add the following line to
/etc/crontab to
periodically clean out old sessions:
0 * * * * bin /usr/local/share/sqwebmail/cleancache.pl
Sqwebmail comes in two parts: a small CGI stub which sends HTTP
requests down a socket; and a pool of daemons which perform the
actual work. The CGI stub is installed in /usr/local/www/cgi-bin-dist
by default, and there are some graphics installed in
/usr/local/www/data-dist/sqwebmail/*.
These files need to be copied to a location where Apache can find them:
# cp -r /usr/local/www/cgi-bin-dist/sqwebmail \ /usr/local/www/apache22/cgi-bin # cp -r /usr/local/www/data-dist/sqwebmail \ /usr/local/www/apache22/data
sqwebmail's main configuration file is
/usr/local/etc/sqwebmail/sqwebmaild
- however you almost certainly don't need to change it.
As usual, you will need to enable the sqwebmail daemon in
/etc/rc.conf, and
then call its startup script.
# vi /etc/rc.conf ... sqwebmaild_enable="YES" # /usr/local/etc/rc.d/sqwebmail-sqwebmaild start Starting sqwebmaild.
If everything is working correctly, you should be able to point a web browser at http://localhost/cgi-bin/sqwebmail/sqwebmail and be presented with a login screen, where you can enter a username and password and login (e.g. sse and afnog).
If this does not work:
/var/log/httpd-access.log
and /var/log/httpd-error.log
/var/log/maillog
/var/log/debug.log
After you log in, you should by on the Folders screen. Click on the INBOX folder to check your mail. You may have some emails that you sent to yourself earlier for testing.
Check that you can send mail by clicking on the Create Message link, address the message to yourself (e.g. maina@pc28.sse.ws.afnog.org) type a message, and send it. Check that it appears in your Inbox. If not, check your Exim mail logs.
Further documentation for sqwebmail can be found at
http://www.courier-mta.org/sqwebmail/
and installed in /usr/local/share/doc/sqwebmail/
Squirrelmail is a lightweight and simple webmail server, that gives your users access to their email without installing any software or configuring anything.
Like most webmail systems, Squirrelmail uses IMAP to access the actual mailbox. This means that it doesn't need to understand mailbox, maildir or other mail storage formats, it can even work with mail stored on an Exchange server, and it doesn't have to run on the same host (or even in the same country) as the IMAP server.
Because webmail is inefficient on bandwidth, and IMAP is efficient, it makes sense to put webmail servers close to your users. If your email is stored off site, it might make sense to install a webmail server onsite that talks to your offsite server. If you have multiple campuses connected by slow links, consider a webmail server for each. If your customers use webmail in their offices, consider providing them with a local webmail server onsite.
Squirrelmail is written in the PHP scripting language. It has plenty of features and is easy to use and relatively fast and efficient (for webmail).
To install SquirrelMail from ports:
# cd /usr/ports/mail/squirrelmail # make install clean (This will install PHP, where you need to enable the Apache module. Go down the list of options with the arrow keys to: [ ] APACHE Build Apache module press Space to enable this option: [X] APACHE Build Apache module and then choose OK as before.)
SquirrelMail installs its files into /usr/local/www/squirrelmail, which Apache does not know anything about. To make Squirrelmail work, you need to modify the Apache configuration:
# vi /usr/local/etc/apache22/httpd.conf Add the following lines right at the end of the file: Alias /mail /usr/local/www/squirrelmail <Directory /usr/local/www/squirrelmail> Allow from all AddHandler php5-script php DirectoryIndex index.php </Directory>
SquirrelMail requires some configuration before it can be used:
# cd /usr/local/www/squirrelmail/config # perl conf.pl
This starts a menu system. The items that you might want to change are:
You must change the Domain, otherwise you will not be able to send mail. Exim will refuse to send it with the following error:
Requested action not taken: mailbox unavailable 550 relay not permitted
Now restart Apache:
# /usr/local/etc/rc.d/apache22 restart
and you should be able to access your webmail in Squirrelmail at https://localhost/mail. If SSL doesn't work for you, then try without encryption at http://localhost/mail.
Compare SquirrelMail to SqWebmail, list advantages and disadvantages of each.
Give your neighbour a mail account on your system. Let them check that they can collect mail using POP3, IMAP and Webmail. To receive mail using IMAP and POP3, you will need a client. You can find KMail on your graphical menu under Internet. Create an account for yourself using POP3, specifying your neighbour's machine as the server.
A number of behaviours of courier-imap and sqwebmail can be changed by means of "account options". These can be set globally, and overridden for individual accounts (although not for Unix system accounts). Try the following:
# vi /usr/local/etc/authlib/authdaemonrc change DEFAULTOPTIONS="" to DEFAULTOPTIONS="wbnodsn=1,wbnochangingfrom=1,disableshared=1" # /usr/local/etc/rc.d/courier-authdaemond restart
You can see account options for an account using "authtest username",
and list all accounts together with their options using
"authenumerate -o"
The available options are:
disableshared=1disablepop3=1
disableimap=1
disablewebmail=1
wbnochangingfrom=1wbnochangepass=1wbusexsender=1wbnoimages=1wbnodsn=1In the simple examples above, we have been using the system password file to authenticate users. When creating new "E-mail only" accounts on your system, you probably don't want your users to be able to login to Unix using ssh or telnet. To disable this, you can simply create their accounts with a nonexistent shell.
# pw useradd username -m -s /nonexistent
To improve security and scalability further, you may wish to keep all your mail accounts in a completely separate password file or database; these users won't be known to Unix at all. The mail directories and messages have to be owned by some Unix user, so we can choose to make them all owned by the 'mailnull' user.
You have many choices of authentication module: for example an LDAP database, a mysql or postgresql database, or a local dbm file (courier supports a format called 'userdb'). These databases will contain the mail login usernames and passwords, and the directories where the mail will be stored. You'll need to configure Courier to use this new database as a login source, and also configure Exim to read this source to determine whether a user exists and where to deliver mail to.
The following example shows how to do this with authuserdb, which
is described in "man makeuserdb" and "man userdb".
We'll create two users in the table, and two empty maildirs. We'll
make them under directory /var/vmail01,
which we'll assume is a fast SCSI hard drive. To support many domains
we'll make the POP3 login be user@domain instead of just a
username. We'll have a separate directory for each domain, and also
use the first two characters of the username as subdirectories, so
that if we have ten thousand users for one domain we don't end up
with ten thousand accounts within the same directory.
First we make an empty userdb, and make sure it's not world readable
# touch /usr/local/etc/userdb # chmod 600 /usr/local/etc/userdb
Next we create some accounts and empty maildirs:
# userdb fred@flintstone.org set uid=26 gid=6 home=/var/vmail01/flintstone.org/f/r/fred # userdbpw -md5 | userdb fred@flintstone.org set systempw Password: wibble Reenter password: wibble # mkdir -p /var/vmail01/flintstone.org/f/r/fred # maildirmake /var/vmail01/flintstone.org/f/r/fred/Maildir # chown -R mailnull:mail /var/vmail01/flintstone.org/f/r/fred # userdb wilma@flintstone.org set uid=26 gid=6 home=/var/vmail01/flintstone.org/w/i/wilma # userdbpw -md5 | userdb wilma@flintstone.org set systempw Password: boing Reenter password: boing # mkdir -p /var/vmail01/flintstone.org/w/i/wilma # maildirmake /var/vmail01/flintstone.org/w/i/wilma/Maildir # chown -R mailnull:mail /var/vmail01/flintstone.org/w/i/wilma
(In real life you'd write a script to automate this process for creating new accounts). The compulsory fields we need to provide are "home" and numeric "uid" and "gid"; these are documented in "man makeuserdb".
Now we check the userdb contents - it's just a plain text file - then convert it into userdb.dat which is the fast indexed version that authuserdb reads. Note that authuserdb requires encrypted passwords.
# cat /usr/local/etc/userdb fred@flintstone.org home=/var/vmail01/f/flintstone.org/fred|systempw=$1$96YgsKCe$0oey3dzw0mztdby6ICFxR0|gid=6|uid=26 wilma@flintstone.org home=/var/vmail01/f/flintstone.org/wilma|systempw=$1$nXNJyXcB$1mItZjaFmOV/3YHby8SGu0|gid=6|uid=26 # makeuserdb # ls -l /usr/local/etc/userdb.dat -rw-r--r-- 1 root wheel 65536 May 14 16:23 /usr/local/etc/userdb.dat
Now you can configure courier-imap and sqwebmail to login using these new accounts. If you leave authpam in the configuration then you can login with both the userdb accounts and the system accounts. If you removed authuserdb from the list of authentication modules earlier, then put it back now:
# vi /usr/local/etc/authlib/authdaemonrc ... authmodulelist="authuserdb authpam" ... # /usr/local/etc/rc.d/courier-authdaemond stop # /usr/local/etc/rc.d/courier-authdaemond start
At this point you should be able to login using one of the new accounts, and see the (empty) mail directory.
# telnet localhost 110 Trying 127.0.0.1... Connected to localhost Escape character is '^]'. +OK Hello there. user fred@flintstone.org +OK Password required. pass wibble +OK logged in. stat +OK 0 0 quit +OK Bye-bye. Connection closed by foreign host.
If this doesn't work, follow the instructions for debugging
authentication problems given earlier. Remember the authtest and
authenumerate commands, and look in /var/log/maillog
and /var/log/debug.log
Now all that is necessary is for Exim to know how to deliver
messages to these users. There are a couple of ways this can be done;
Exim can be configured to read /usr/local/etc/userdb.dat
directly, or it can be configured to talk to courier's authdaemond
process. Both can be set up using Exim's general-purpose
configuration language.
It's a tribute to the flexibility of Exim that this can be done even though Exim does not have any specific features for using userdb or authdaemond, although it does mean that the configuration looks complicated at first glance. If you want to find out in detail how these configurations work, you will need to read the Exim documentation carefully.
Firstly, we need a separate list of which domains which need lookups in the userdb database; for simplicitly we will make this a plain text file which is searched linearly. If this gets large it can be converted into an indexed database.
# vi /usr/exim/vdomains flintstone.org # vi /usr/local/etc/exim/configure Change the local domains setting to say: domainlist local_domains = @ : lsearch;/usr/exim/vdomains
To read /usr/local/etc/userdb.dat
directly, we use the "dbmnz" lookup type. It's made a bit
awkward because courier uses a vertical bar to separate fields,
whereas Exim's "extract" operator expects fields separated
by spaces, but we can use the regular expression substitution
operator (sg) to convert this into the form we want.
[put this immediately after 'begin routers']
userdb:
driver = accept
transport = local_delivery_userdb
domains = lsearch;/usr/exim/vdomains
address_data = ${lookup{$local_part@$domain}dbmnz{/usr/local/etc/userdb.dat}\
{${sg{$value}{([^=]+)=([^|]+)\\|?}{\$1="\$2" }}}fail}
# note the space between "\$2" and }}}
# If the address_data lookup succeeds, then we'll go to the transport.
# But if the address_data lookup fails, then we fall through to here; all
# remaining addresses in vdomains need to be bounced.
userdb_bounce:
driver = redirect
domains = lsearch;/usr/exim/vdomains
data = :fail:unknown user
allow_fail
fail_verify
[put this anywhere after 'begin transports']
local_delivery_userdb:
driver = appendfile
directory = ${extract{home}{$address_data}}/${extract{mail}{$address_data}{$value}{Maildir}}/
maildir_format
maildir_use_size_file
delivery_date_add
envelope_to_add
return_path_add
user = ${extract{uid}{$address_data}}
group = ${extract{gid}{$address_data}}
maildir_tag = ,S=$message_size
quota_size_regex = ,S=(\d+)
quota = ${if match{${extract{quota}{$address_data}}}{([0-9]+)S}{$1}{}}
quota_filecount = ${if match{${extract{quota}{$address_data}}}{([0-9]+)C}{$1}{}}
quota_warn_threshold = 90%
Once you've done this, test using
# /usr/exim/bin/exim -bt brian@flintstone.org brian@flintstone.org is undeliverable: unknown user # /usr/exim/bin/exim -bt fred@flintstone.org fred@flintstone.org router = userdb, transport = local_delivery_userdb # /usr/exim/bin/exim -v fred@flintstone.org Subject: test hello . LOG: MAIN <= root@noc.presanog.org.bt U=root P=local S=302 LOG: MAIN => fred <fred@flintstone.org> R=userdb T=local_delivery_userdb LOG: MAIN Completed
It's even possible for Exim to send a request to courier's authdaemond process to perform the lookup, which has the advantage that it will work for any courier authentication module or combination of modules. However there are similar difficulties with parsing the response properly, and the debug output you get from courier is not as good as Exim produces.
[put this immediately after 'begin routers']
# We use manualroute with empty route_data as a dummy router, just to
# set address_data to the value read from the socket.
authdaemon_lookup:
driver = manualroute
route_data =
domains = lsearch;/usr/exim/vdomains
address_data = ${sg{${readsocket{/usr/local/var/spool/authdaemon/socket}\
{PRE . exim $local_part@$domain\n}}}{([^=]+)=([^\n]+)\n}{\$1="\$2" }}
# note the space between "\$2" and }}
# Next, if the response contains HOME= then we know this address is valid
# and we can send it to the local delivery transport
courier:
driver = accept
transport = local_delivery_courier
domains = lsearch;/usr/exim/vdomains
condition = ${extract{HOME}{$address_data}{1}{0}}
# Otherwise, the address was bad. If it contains FAIL then it's a permanant
# failure, otherwise it's a temporary failure
courier_bounce:
driver = redirect
domains = lsearch;/usr/exim/vdomains
data = ${if match{$address_data}{FAIL}{:fail:unknown user}fail}
allow_fail
fail_verify
[transports]
local_delivery_courier:
driver = appendfile
directory = ${extract{HOME}{$address_data}}/${extract{MAILDIR}{$address_data}{$value}{Maildir}}/
maildir_format
maildir_use_size_file
delivery_date_add
envelope_to_add
return_path_add
user = ${extract{UID}{$address_data}}
group = ${extract{GID}{$address_data}}
maildir_tag = ,S=$message_size
quota_size_regex = ,S=(\d+)
quota = ${if match{${extract{QUOTA}{$address_data}}}{([0-9]+)S}{$1}{}}
quota_filecount = ${if match{${extract{QUOTA}{$address_data}}}{([0-9]+)C}{$1}{}}
quota_warn_threshold = 90%
Finally, if you are building a box where you know that all the mail will be owned by the 'exim' or 'mailnull' user, then in fact the daemons no longer need to run with root privileges - they can run as the exim user only, as they never have to change to any other userid. This in theory should make your system more secure.
There are some steps you can take to do this for each daemon:
/usr/exim/configure.
You'll probably also have to adjust permissions so that the
exim user has access to the authdaemon socket:
# chown mailnull:mail /usr/local/var/spool/authdaemon
... TCPDOPTS="-nodnslookup -noidentlookup -user=mailnull" ...
# su mailnull -c "/usr/local/etc/rc.d/sqwebmail-sqwebmaild.sh start"However, for this to work, you first need to sort out some permissions issues: you need to recompile and reinstall sqwebmail using
# make WITH_CACHEOWNER=mailnulland change the entry in /etc/crontab to run cleancache.pl as user 'mailnull' not user 'bin'. You also need to change ownerships of the configuration files and temporary directory:
# cd /usr/local/etc # chown mailnull:mail sqwebmail/sqwebmaild authlib/authdaemonrc # chown -R mailnull:mail /usr/local/var/sqwebmail