Table of Contents
If you're reading this, you have found the README.Debian file. This is good, thanks! Please continue reading this file in its entirety. It is full of important information and has been written with the questions in mind that keep popping up on the mailing lists.
Exim comes with very extensive documentation. Here is how to find it.
The very extensive Upstream documentation is shipped
/usr/share/doc/exim4-base/spec.txt.gz
)
with the binary packages.
exim4-doc-html
exim4-doc-info
Please note that documentation found on the web or in other parts of the Debian system (such as the Debian Reference) might be outdated and thus give wrong advice. In doubt, the documentation listed above should take precedence.
For your questions and comments, there is a Debian-specific mailing list. Please ask Debian-specific questions there, and only write to the upstream exim-users mailing list if you are sure that your question is not Debian-specific. Debian-specific questions are more likely to find answers on our pkg-exim4-users mailing list, while complex custom configuration issues might be more easily solved on the upstream exim-users mailing list because of the broader and more experienced audience there. You can subscribe to pkg-exim4-users via the subscription web page; you need to be subscribed to post.
If you think that your question might be more easily answered if one knows a bit about your configuration, you might want to execute reportbug --subject="none" --offline --quiet --severity=wishlist --body="none" --output=exim4.reportbug exim4-config on the system in question, answer yes to both "include [extended] configuration" questions and include the contents of the exim4.reportbug file generated by this command with your question. Please check whether the file contains any confidential information before sending.
Similar to the Apache2 package, Exim 4 is an entirely different package that does not currently offer a smooth upgrade path from Debian's Exim 3 packages.
It is the first Exim package in Debian that can be configured using debconf. However, the entire configuration framework is extremely flexible, allowing you to get exactly the amount of control you need for the job at hand.
To use Exim 4, you need at least the following packages:
Just apting the metapackage exim4 will pull in the other packages per dependency. You'll get an exim daemon with minimal feature set (no external lookups).
If you need more advanced features like LDAP, sqlite, PostgreSQL and MySQL data lookups, SASL and SPA SMTP authentication, embedded Perl interpreter, and exiscan-acl for integration of virus-scanners and SpamAssassin, you can replace exim4-daemon-heavy instead of exim4-daemon-light. Additionally, the source package offers infrastructure to build your own custom-tailored exim4-daemon-custom which exactly fits your special local needs. The infrastructure to do so is already in place, see debian/rules for instructions.
Generally, the Debian Exim 4 packages are configured through debconf. You have been asked some questions on package installation, and your initial Exim configuration has been created from your answers. You can repeat the configuration process any time by invoking dpkg-reconfigure exim4-config. If you are an experienced Exim administrator and prefer to have your own, hand-crafted, non-automatic Exim configuration, you will find information about how to do so in Section 2.1.6, “Using a completely different configuration scheme”.
The debconf-driven configuration is mainly geared for a one-domain shell account machine/workstation with local delivery as suggested by the original upstream default configuration. If you configure the packages to handle more than one local domain, all local domains are treated identically. The domain part is not used for routing and filtering decisions.
Despite the default configuration being extended somewhat from the original upstream, chances are that you'll need to manually change the Exim configuration with an editor if you intend to do something that is not covered by the debconf-driven configuration. It has never been the packages' intention to offer all possible configuration methods through debconf. The configuration files are there to be changed, feel free to do so if you see fit. The Debian Exim 4 maintainers have tried to make the configuration as flexible as possible so that manual intervention can be minimized.
If you need to make manual changes to the Exim configuration,
please be familiar with how Exim works. At minimum, have read this
README file and the manpages delivered with the Debian Exim 4
packages, and /usr/share/doc/exim4-base/spec.txt.gz
chapters "How Exim receives and delivers mail" and
"The Exim run time configuration file".
spec.txt.gz
is an excellent reference.
Please note that while most free-form fields in the debconf-driven configuration have the entered string end up verbatim in Exim's configuration file (and thus using more advanced features like host, address and domain lists is possible and will probably work), this is not officially supported. Only plain lists are supported in the debconf dialogs. You may use more advanced features, but they may stop working any time during upgrades.
In this section, we try to document and explain the debconf
questions, which are themselves limited to a small screen of
information and might leave questions unanswered. Since you
can usually read this file only after having answered the
questions, the process can always be repeated by invoking
dpkg-reconfigure exim4-config.
/etc/exim4/update-exim4.conf.conf
,
documented in the update-exim4.conf
manual page, is
a simple shell-script snippet used to store the answers
that you passed to debconf when initially configuring Exim.
You may also modify this file with an editor of your choice.
The package maintainer scripts can handle this and will
preserve your changes.
This is the main configuration question which will control which of the remaining questions are presented to you. It also controls things like daemon invocation and delivery of outgoing mail.
This option is suitable for a standalone system with full internet connectivity.
The Exim SMTP daemon will accept messages to local domains, and deliver them locally.
Outgoing mail will be delivered directly to the mail exchange servers of the recipient domain
This option is suitable for a standalone client system which has restricted internet connectivity, for example on a residential connection where an SMTP smarthost is used. Some ISPs block outgoing SMTP connections to combat the spam problem, thus requiring the use of their smarthosts. It is generally a good idea to use the ISPs smart host if one is connected with a dynamic IP address since quite a few sites do not accept mail directly delivered from a dial-in pool.
fetchmail can be used to retrieve incoming mail from the ISP's POP3 or IMAP mail server and deliver it to Exim via SMTP.
The Exim SMTP daemon will accept messages to local domains, and deliver them locally.
Outgoing mail will always be delivered to the smarthost configured in exim4.
This option is suitable for a client system in a computer pool which is not responsible for a local e-mail domain. All locally generated e-mail is sent to the smarthost without any local domains.
This option is suitable for a standalone system with no networking at all. Only messages for configured local domains are accepted and delivered locally; messages for all other domains are rejected: ``Mailing to remote domains not supported''.
This option disables most of Debian's automatisms
and leaves exim in an unconfigured state.
update-exim4.conf will still copy
/etc/exim4/exim4.conf.template
or concatenate the files from
/etc/exim4/conf.d,
and will
not generate any configuration control macros.
Unless you manually edit the configuration source,
this will leave Exim with a syntactically invalid
configuration file, thus in a state where the
daemon won't even start.
Only choose this option if you know what you're doing and are prepared to create your own Exim configuration.
dpkg-conffile handling is still in place, and you will be offered updates for configuration snippets, as soon as they become available.
The "mail name" is the domain name used to "qualify" mail addresses without a domain name.
This name will also be used by other programs. It should be the single, full domain name (FQDN).
For example, if a mail address on the local host is foo@example.org, then the correct value for this option would be example.org.
Exim, as a rule, handles only fully qualified mail addresses, that is, addresses with a local part, an @ sign and a domain. If confronted with an unqualified address, that is, one without @ sign and without domain, first thing exim does is qualify the address by adding the @ sign and a domain.
This qualification happens for all addresses exim encounters, be it sender, recipient or else.
The domain name used to qualify unqualified mail addresses
is called ``mail name'' on Debian systems and entered
in this debconf dialog. What you enter here will end
up in /etc/mailname,
which is a
file that might be used by other programs as well.
In some configuration types, the package configuration will offer you, at a later step, to hide this name from outgoing messages by rewriting the headers.
Please enter a semicolon-separated list of IP addresses. The Exim SMTP listener daemon will listen on all IP addresses listed here.
An empty value will cause Exim to listen for connections on all available network interfaces.
If this system does only receive e-mail directly from local services (and not from other hosts), it is suggested to prohibit external connections to the local Exim daemon. Such services include e-mail programs (MUSs) which talk to localhost only as well as fetchmail. External connections are impossible when 127.0.0.1 is entered here, as this will disable listening on public network interfaces.
Do not change this unless you know what you are doing. Altering this value could post a security risk to your system. For most users, the default value is sufficient.
Please enter a semicolon-separated list of recipient domains for which this machine should consider itself the final destination. These domains are commonly called 'local domains'. The local hostname and 'localhost' are always added to the list given here.
By default all local domains will be treated identically. If both a.example and b.example are local domains, acc@a.example and acc@b.example will be delivered to the same final destination. If different domain names should be treated differently, it is necessary to edit the config files afterwards.
The answer to this question ends up in the list of domains that Exim will consider local domains. Mail for recipients in one of these domains will be subject to local alias expansion and then delivered locally in the appropriate configuration types.
Please enter a semicolon-separated list of recipient domains for which this system will relay mail, for example as a fallback MX or mail gateway. This means that this system will accept mail for these domains from anywhere on the Internet and deliver them according to local delivery rules.
Do not mention local domains here. Wildcards may be used.
The answer to this question is a list of the domains for which Exim will relay messages coming in from anywhere on the Internet.
Please enter a semicolon-separated list of IP address ranges for which this system will unconditionally relay mail, functioning as a smarthost.
You should use the standard address/prefix format (e.g. 194.222.242.0/24 or 5f03:1200:836f::/48).
If this system should not be a smarthost for any other host, leave this list blank.
Please note that systems not listed here can still use SMTP AUTH to relay through this system. If this system only has clients on dynamic IP addresses that use SMTP AUTH, leave this list blank as well. Do NOT list 0.0.0.0/0!
Warning: While it is possible to use hostnames instead of IP addresses in this list extra care needs to be taken in this case. Unresolvable names in the host list will break relaying. See Exim specification chapter "Domain, host, address, and local part lists" and the exim4-config_files man page.
Please enter the IP address or the host name of a mail server that this system should use as outgoing smarthost. If the smarthost only accepts your mail on a port different from TCP/25, append two colons and the port number (for example smarthost.example::587 or 192.168.254.254::2525). Colons in IPv6 addresses need to be doubled.
If the smarthost requires authentication, please refer to Section 2.3, “SMTP-AUTH” for notes about setting up SMTP authentication.
Multiple smarthost entries are permitted, semicolon separated. Each of the hosts is tried, in the order specified (See Exim specification, chapter "The manualroute router", section "How the list of hosts is used".)
The headers of outgoing mail can be rewritten to make it appear to have been generated on a different system, replacing the local host name in From, Reply-To, Sender and Return-Path.
If you ask Exim to hide the local mail name in outgoing mail, it will next ask you for the domain name that should be visible for your local users. These information is then used to establish the appropriate rewriting rules.
In normal mode of operation Exim does DNS lookups at startup, and when receiving or delivering messages. This is for logging purposes and allows keeping down the number of hard-coded values in the configuration.
If this system does not have a DNS full service resolver available at all times (for example if its Internet access is a dial-up line using dial-on-demand), this might have unwanted consequences. For example, starting up Exim or running the queue (even with no messages waiting) might trigger a costly dial-up-event.
This option should be selected if this system is using Dial-on-Demand. If it has always-on Internet access, this option should be disabled.
Exim is able to store locally delivered mail in different formats. The most commonly used ones are mbox and Maildir. mbox uses a single file for the complete mail folder stored in /var/mail/. With Maildir format every single message is stored in a separate file in ~/Maildir/.
Please note that most mail tools in Debian expect the local delivery method to be mbox in their default.
Our packages offer two (actually three, see Section 2.1.6, “Using a completely different configuration scheme”) possibilities:
/etc/exim4/exim4.conf.template,
which is basically a normal Exim run-time
configuration file which will be supplemented
with some macros generated from Debconf in a
post-processing step before it is passed to exim.
/etc/exim4/conf.d/
. The
directories in
/etc/exim4/conf.d/
correspond to the sections of the Exim
run-time configuration file, so you should
easily find your way around there.
Splitting the configuration across multiple files
means that you have the actual configuration file
automatically generated from the files below
/etc/exim4/conf.d/
by invoking
update-exim4.conf. Each section
of Exim's configuration has its own subdirectory and
the files in there are supposed to be read in
alphanumeric order.
router/00_exim4-config_header
is followed by
router/100_exim4-config_domain_literal
,
...
If you chose unsplit configuration,
update-exim4.conf builds the
configuration from
/etc/exim4/exim4.conf.template
,
which is basically the files from
/etc/exim4/conf.d/
concatenated
together at package build time, and thus guarantees
consistency on the target system.
In both cases, update-exim4.conf generates exim configuration macros from the debconf configuration values and puts them into the actual configuration file, which is then used by the Exim daemon. See the update-exim4.conf manual page for more in-depth information about this mechanism.
Benefits of the split configuration approach:
/etc/exim4/conf.d
.
This needs, however quite exact syncing
between the exim4 packages and the other,
cooperating package.
Drawbacks of the split configuration approach:
Benefits of the unsplit configuration approach:
exim4.conf.template
basically is a complete Exim configuration
file which will only undergo some basic
string replacement before is it passed to
exim.
Drawbacks of the unsplit configuration approach:
If in doubt go for the unsplit config, because it is easier to roll back to Debian's default configuration in one step. If you intend to do many changes to the Debian setup, you might want to use the split config at the price of having to more closely examine the config file after an update.
We'd appreciate a patch that uses ucf and the 3-way-merge mechanism offered by that package. It might be the best way to handle the big configuration file.
If you are using unsplit configuration, have local
changes to /etc/exim4/conf.d/
(either made by yourself or by other packages dropping
their own routers or transports in) and want to
re-generate
/etc/exim4/exim4.conf.template
to
activate these changes, you can do so by using
update-exim4.conf.template.
The Debian exim 4 packages come with a default configuration that allows flexible access control and blacklisting of sites and hosts. The acls involved can be found in /etc/exim4/conf.d/acl, or in /etc/exim4/exim4.conf.template, depending on which configuration scheme you use. Most rejections of messages due to this mechanism happen at RCPT time. Local configuration of the mechanisms happens through data files in /etc/exim4 or via Exim macros that you can set in /etc/exim4/conf.d/main, so there is normally no need to change the files in the acl subdirectory in a split-config setup. If you use the non-split config, you need to edit /etc/exim4/exim4.conf.template, which, as a big dpkg-conffile, won't give you any advantage of the .ifdef scheme.
The data files are documented in the exim4-config_files man page.
The access lists delivered with the exim4 packages also contain quite a few configuration options that are too restrictive to be active by default on a real-life site. These are masked by .ifdef statements, can be activated by setting the appropriate macros, and are documented in the ACL files itself.
Our configuration can be controlled in a limited way by setting macros. That way, you can switch on and off certain parts of the default configuration and/or override values set in Debconf without having to touch the dpkg-conffiles. While touching dpkg-conffiles itself is explicitly allowed and wanted, it can be quite a nuisance to be asked on package upgrade whether one wants to use the locally changed file or the file changed by the package maintainer.
Whenever you see an .ifdef or
.ifndef clause in the configuration file,
you can control the appropriate clause by setting the macro in
a local configuration file. .ifndef checks
whether a specific macro is set (to a nonempty value), the
actual value does not matter. (Both
“EXIM4_EXAMPLE = true” and
“EXIM4_EXAMPLE = false” pass this test.)
For split configuration, you can
drop the local configuration file anywhere in
/etc/exim4/conf.d/main
. Just make sure it
gets read before the macro is first used.
000_localmacros
is a possible name,
guaranteeing first order. For a non-split configuration,
/etc/exim4/exim4.conf.localmacros
gets
read before
/etc/exim4/exim4.conf.template
. To
actually set the macro EXIM4_EXAMPLE
to the
value "this is a sample", write the following line
EXIM4_EXAMPLE = this is a sample
into the appropriate file. For more detailed discussion of the general macro mechanism, see the Exim specification, chapter "The Exim run time configuration file", for details how macro expansion works.
The script update-exim4.conf parses the
/etc/exim4/update-exim4.conf.conf
file
and provides the configuration for the exim daemon.
Depending on the value of
dc_use_split_config
, it either
/etc/exim4/conf.d/
and
concatenates them together or
exim4.conf.template
as
input.
The debconf-managed information from
/etc/exim4/update-exim4.conf.conf
is
merged into the generated configuration file by generating a
number of Exim configuration macros.
DCsmarthost
, for example, is set to the
value of $dc_smarthost
in /etc/exim4/update-exim4.conf.conf
which holds the answer to "Which machine will act as the
smarthost and handle outgoing mail?"
The result of these operations is saved as
/var/lib/exim4/config.autogenerated
,
which is not a dpkg-conffile! Manual
changes to this file will be overwritten by
update-exim4.conf.
Please consult update-exim4.conf manpage for more detailed information.
update-exim4.conf is invoked by the init script prior to any operation that may invoke an exim process, and gives an error message if the generated config file is syntactically invalid. If you want to activate your changes to files in conf.d/ just execute invoke-rc.d exim4 restart.
Some times, you want to do minor adjustments to the Exim configuration to make Exim behave exactly like you want it to behave. There are the following possibilities to modify Exim's behavior.
If you want to modify parameters that are supported by the
debconf configuration, things are easy. Just invoke
dpkg-reconfigure exim4-config or hand-edit
/etc/exim4/update-exim4.conf.conf
to your
liking and restart Exim.
You can find explanation of the debconf questions in Section 2.1.1, “The Debconf questions”.
Additionally,
/etc/exim4/update-exim4.conf.conf
is documented in the update-exim4.conf
man page.
Some aspects of the Debian Exim configuration can be controlled by Exim macros. To find out about these, you need basic understanding of Exim configuration. Just look in our Exim configuration and see which macro needs to be set to a different value to alter Exim's behavior.
Section 2.1.3, “Using Exim Macros to control the configuration” gives a closer explanation about how to do this.
You can, of course, make direct change to the configuration. All configuration files in /etc/exim4 are dpkg-conffiles, and you can thus edit them any time. Your changes will be preserved through updates. You need to know about how to configure Exim to be successful.
If you use unsplit configuration, edit
/etc/exim4/exim4.conf.template
. If you use
split configuration, edit the Exim configuration snippets in
/etc/exim4/conf.d
.
More information about how the Exim configuration is built can be found in this document and in the update-exim4.conf manual page.
If you are an experienced Exim administrator, you might feel working with our pre-fabricated configuration cumbersome and complex. You might feel right if you need to make more complex changes and do not need to receive updates from us. This section is going to tell about how to use your own configuration.
But, you might profit from keeping the Debian magic. Most files that come with Debian exim4 are conffiles. Debian is going to care about your changes and keeps them around. Additionally, a lot of configuration options can be overridden with a macro, which does not require you to actually change our configuration file. A lot of people are using our configuration scheme, and maybe it is going to save you a lot of time if you decide to spend some time familiarizing yourself with our scheme.
If you are only running a small number of systems and
want to completely disable Debian's magic, just take
your monolithic configuration file and install it as
/etc/exim4/exim4.conf
. Exim will
use that file verbatim. To have something to start,
you can either take
/etc/exim4/exim4.conf.template
,
run update-exim4.conf --keepcomments --output
/etc/exim4/exim4.conf, or use upstream's
default configuration file that is installed as
/usr/share/doc/exim4-base/examples/example.conf.gz
.
You are going to lose all magic you get from packaging
though, so you need to be familiar with Exim to build
an actually working config.
/var/lib/exim4/config.autogenerated
,
the file generated by
update-exim4.conf, is ignored as soon
as /etc/exim4/exim4.conf
is found.
You should not edit
/etc/exim4/exim4.conf
directly when
Exim is running, because the forked processes Exim starts
for SMTP receiving or queue running would use the new
configuration file, while the original main exim-daemon
would still use the old configuration file.
Some third-party HOWTOs that reference Debian and
claim to make things easy suggest dumping a
pre-fabricated, static config file to
/etc/exim4/exim4.conf
. This is
considered bad advice by the Debian maintainers since
you are going to disable all updates and service magic
that Debian might deliver in the future this way. If
you do not know exactly what you're doing here, this
is a bad choice. We try to comment on external HOWTOs
found on the web in the Debian
Exim4 User FAQ to help you find out which
advice to follow.
We split off Exim's configuration system (debconf,
update-exim4.conf, and the files in
/etc/exim4/conf.d)
to a separate
package, exim4-config. If you want to, you can replace
exim4-config by something entirely different. The other
packages don't care. Your package needs to:
/var/lib/exim4/config.autogenerated
or into /etc/exim4/exim4.conf
.
Your package must provide an executable update-exim4.conf
that must be in root's path (/usr/sbin
recommended). The init
script will invoke that executable prior to invoking the
actual exim daemon. If you do not need that script, have it exit 0.
If you want to create your own configuration packages, there is a number of helpers available.
debian/config-custom/create-custom-config-package
which will create a new source package
"exim4-config-custom" with the debconf-driven config
scheme of exim4-config for your local modification.
Please note that exim4-config-simple and exim4-config-medium are only targeted to be used as a template. The configurations contained are not suitable for productive use. Of course, the Debian maintainers appreciate any patches you might find suitable. The scripts in exim4-config-simple and exim4-config-medium may not work at all in your environment. Unfortunately, they have not been updated in a long time as well. We are willing to accept patches.
See the development web page for links to the subversion repository.
Exchanging the entire exim4-config package with something custom comes particularly handy for sites that have more than a few machines that are similarly configured, but do not want to use the original exim4-config package. Build your own exim4-config-custom or exim4-config-foo, and simply apt that package to the machines that need to have that configuration. Future updates can then be handled via the dpkg-conffile mechanism, properly detecting local modifications.
In the future, it might be possible that Debian will contain multiple flavours of Exim4 configuration. However, these packages would have to be maintained by someone else because the exim4 package maintainers think that the scheme delivered with exim4-config is the least of all evils and would rather not spend the time to maintain multiple configuration schemes while only actually using one. It would be nice to have a configuration scheme using a monolithic config file, managed by ucf in three-way-merge mode. If anybody feels ready to maintain it, please go ahead.
Both exim4-daemon-heavy and exim4-daemon-light support TLS/SSL using the GnuTLS library.
Exim will use TLS via STARTTLS automatically as client if the server Exim connects to offers it.
This means that you will not need any special configuration if you want to use opportunistic TLS for outgoing mail. However, to enforce TLS and successful certificate verification, a few things need to be configured.
To enforce TLS and prevent fallback to unencrypted connections, ensure that hosts_require_tls = * is in effect on the respective transport. For the remote_smtp_smarthost transport, this setting can be controlled via the REMOTE_SMTP_SMARTHOST_HOSTS_REQUIRE_TLS macro.
The certificate presented by the remote host is checked
against the system CA certificate store
(/etc/ssl/certs/
) and the verification
result is logged (CV=...).
For the remote_smtp_smarthost transport successful
certificate verification against the system trust store is
enforced by default on encrypted connections.
(“REMOTE_SMTP_SMARTHOST_TLS_VERIFY_HOSTS = *”
is set by default). Set this macro to an empty value to
disable this. To check against a certificate not present in
the system trust store point
REMOTE_SMTP_SMARTHOST_TLS_VERIFY_CERTIFICATES (which sets
tls_verify_certificates) to a file containing this (set of)
trusted certificates.
Successful certificate verification is not enforced by default for other transports.
Another possibility would be to use DANE for certificate verification. This requires support on the server side and a resolver with DNSSEC support on the client side.
If your server setup mandates the use of client certificates, you need to amend your remote_smtp and/or remote_smtp_smarthost transports with a tls_certificate option. This is not commonly needed.
To make exim send a TLS certificate to the remote host set REMOTE_SMTP_TLS_CERTIFICATE/REMOTE_SMTP_PRIVATEKEY or for the remote_smtp_smarthost transport REMOTE_SMTP_SMARTHOST_TLS_CERTIFICATE/REMOTE_SMTP_SMARTHOST_PRIVATEKEY respectively.
To use TLS on connect set “protocol = smtps” on the respective transport. (For the remote_smtp_smarthost transport the macro REMOTE_SMTP_SMARTHOST_PROTOCOL can be used.
Exim supports incoming opportunistic TLS by using on-connect autogenerated self-signed certificates. This is not optimal both for performance reasons and because these certificates cannot be verified by connecting clients/servers.
Each time an on-demand cert is created exim will log a info message in the mainlog that looks like this:
Warning: No server certificate defined; will use a selfsigned one. Suggested action: either install a certificate or change tls_advertise_hosts option
This informative message can be ignored.
To avoid the (small) performance issue and the log message one can
locally create certificates. The exim-gencert script (which requires
openssl) can be helpful for this purpose. It is shipped in
/usr/share/doc/exim4-base/examples/
and
takes care of proper access privileges on the private key
file when installing key/certificate in
/etc/exim4/
.
One can also get a certificate from a CA and install the key in
/etc/exim4/exim.key
and the certificate
in /etc/exim4/exim.crt
.
To enable use of the installed certificates set the macro MAIN_TLS_ENABLE in a local configuration file as described in Section 2.1.3, “Using Exim Macros to control the configuration”.
After this configuration, Exim will advertise STARTTLS when
connected to on the normal SMTP ports. Some broken clients
(most prominent example being nearly all versions of Microsoft
Outlook and Outlook Express, and Incredimail) insist on doing
TLS on connect on Port 465. If you need to support these, set
SMTPLISTENEROPTIONS='-oX 465:25 -oP /run/exim4/exim.pid'
in /etc/default/exim4
and
"tls_on_connect_ports=465" in the main configuration section.
The -oP is needed because Exim does not write an implicit pid file if -oX is given. Without pid file, init script and cron job will malfunction.
It might be appropriate to add "+tls_cipher" to any log_selector statement you might already have, or to add a log_selector statement setting these two options in a local configuration file. (For Debian's configuration simply define the MAIN_LOG_SELECTOR macro.) This option makes Exim log what cipher your Exim and the peer's mailer have negotiated to use to encrypt the transaction.
Exim can be configured to ask a client for a certificate and to try to verify it. Debian's exim configuration used to enable this by default, but stopped doing so since it caused TLS errors with a couple of popular clients (Outlook, Incredimail, etc.). To enable this again set the macro MAIN_TLS_TRY_VERIFY_HOSTS to the lists hosts whose certificates you want to check. (Use * to try checking all hosts. The value of the macro is used to populate exim's main option tls_try_verify_hosts.) You should also point MAIN_TLS_VERIFY_CERTIFICATES to a file containing the accepted certificates, since its default setting (/etc/ssl/certs/ca-certificates.crt) can contain a large list of certificates which causes the interoperabilty problems with Outlook et.al. noted above.
The server certificate is only used for incoming connections, please consult Section 2.2.1, “Exim 4 as TLS/SSL client” for the corresponding outgoing conncection options.
If Exim complains in an SMTP session that TLS is unavailable, the Exim mainlog or paniclog frequently has exact information about what might be wrong. Fo example, you might see
2003-01-27 19:06:45 TLS error on connection from localhost [127.0.0.1] (cert/key setup): Error while reading file)
showing that there has been an error while accessing the certificate or the private key file.
Insuffient entropy available is a frequent cause of TLS
failures in Exim context. If Exim logs "not enough random bytes
available", or simply hangs silently when an encrypted
connection should be established, then Exim was
unable to read enough random data from
/dev/random
to do whatever cryptographic
operation is requested. Please check that your
/dev/random
device is setup properly.
You might also find "TLS error on connection to [...] (gnutls_handshake): The Diffie-Hellman prime sent by the server is not acceptable (not long enough)." given as reason. Exim by default requires a DH prime length of 1024 bits. This requirement can be downgraded by setting the tls_dh_min_bits option on the SMTP transport. The setting is accessible in the Debian configuration by setting the macro TLS_DH_MIN_BITS. (e.g. "TLS_DH_MIN_BITS = 768").
Exim can do SMTP AUTH both as a client and as a server.
AUTH PLAIN and AUTH LOGIN are disabled for connections which are not protected by SSL/TLS per default. These authentication methods use cleartext passwords, and allowing the transmission of cleartext passwords on unencrypted connections is a security risk. Therefore, the default configuration configures Exim not to use and/or allow AUTH PLAIN and AUTH LOGIN over unencrypted connections.
It is thus recommended to set up Exim to use TLS to encrypt the connections. Please refer to Section 2.2, “Using TLS” for documentation about this. Note that most Microsoft clients need special handling for TLS.
If you want to set up Exim as SMTP AUTH client for delivery
to your internet access provider's smarthost put the name of
the server, your login and password in
/etc/exim4/passwd.client
. See the man
page for exim4-config_files(5) for more information about the
required format.
If you need to enable AUTH PLAIN or AUTH LOGIN for unencrypted connections because your service provider does support neither TLS encryption nor the CRAM MD5 authentication method, you can do so by setting the AUTH_CLIENT_ALLOW_NOTLS_PASSWORDS macro. Please refer to Section 2.1.3, “Using Exim Macros to control the configuration” for an explanation of how best to do this.
/etc/exim4/passwd.client
needs to be
readable for the exim user (user Debian-exim, group
Debian-exim). It is suggested that you keep the default
permissions root:Debian-exim 0640.
The configuration files include many, verbosely commented, examples for server-side smtp-authentication which just need to be uncommented.
If you need to enable AUTH PLAIN or AUTH LOGIN for unencrypted connections because your clients neither support TLS encryption nor the CRAM MD5 authentication method, you can do so by setting the AUTH_SERVER_ALLOW_NOTLS_PASSWORDS macro. Please refer to Section 2.1.3, “Using Exim Macros to control the configuration” for an explanation of how best to do this.
If you want to authenticate against system passwords (e.g.
/etc/shadow
) the easiest way is to use
saslauthd in the Debian package sasl2-bin. You have to add the
exim-user (currently Debian-exim) to the sasl group, to give
exim permission to use the saslauthd service.
The Debian exim4 maintainers consider using system login passwords a bad idea for the following reasons:
The Debian Exim 4 packages' init script is located in
/etc/init.d/exim4
. Apart from the
functions that are required by Debian policy and the LSB, it
supports the commands what, which executes
exiwhat to show what your Exim processes
are doing, and force_stop which
unconditionally kills all Exim processes.
The init script can be configured to start listening and/or
queue running daemons. This configuration can be found in
/etc/default/exim4
. This file is
extensively documented.
Exim4's daily cron job
(/etc/cron.daily/exim4-base
)
does basic housekeeping tasks:
/etc/default/exim4
, so you
can use this file to change any of the variables used in
the cron job.
A non-empty paniclog is a nearly sure sign of bad things going on. Thus, the cron job will send out warning messages to the syslog and root if it finds the panic log non-empty. Please note that the paniclog is not rotated daily, so existing issues will be reported daily until either the paniclog is rotated due to its sheer size, or you manually move it away, for example by calling logrotate -f /etc/logrotate.d/exim4-paniclog from a shell.
Just in case your system logs transient error situations to the panic log as well (see, for example, Exim Bug 92), you can configure $E4BCD_PANICLOG_NOISE to a regular expression. If the paniclog contains only lines that match that regular expression, no warning messages are generated.
If you want to disable paniclog monitoring completely, set $E4BCD_WATCH_PANICLOG to no. E4BCD_WATCH_PANICLOG=once will rotate a non-empty paniclog automatically after sending out the warning e-mail.
The E4BCD_PANICLOG_LINES setting can be used to limit the number of lines of paniclog quoted in warning email. It is set to 10 by default.
Exim4 is run as a separate daemon instead of inetd/xinetd for two reasons:
If you introduce bugs on your systems by running from (x)inetd you are on your own! If you want to run exim from xinetd, follow these steps:
Create /etc/xinetd.d/exim4
service smtp { disable = no flags = NAMEINARGS socket_type = stream protocol = tcp wait = no user = Debian-exim group = Debian-exim server = /usr/sbin/exim4 server_args = exim4 -bs }
If you want to use plain inetd, insert following line into
/etc/inetd.conf
:
smtp stream tcp nowait Debian-exim /usr/sbin/exim4 exim4 -bs
Since system accounts (mail, uucp, lp etc) are usually aliased to root, and root's mailbox is usually read by a human, these account names have started to be a common target for spammers. The Debian Exim 4 packages have a mechanism to deal with this situation. However, since this derives rather far from normal behavior, it is disabled by default.
To enable it, set the macro FIRST_USER_ACCOUNT_UID to a numeric, non-zero value. Incoming mail for local users that have a UID lower than FIRST_USER_ACCOUNT_UID is rejected with the message "no mail to system accounts". Incoming mail for local users that have a UID greater or equal FIRST_USER_ACCOUNT_UID are processed as usual. Therefore, the default value of 0 ensures that the mechanism is disabled. On Debian systems, setting FIRST_USER_ACCOUNT_UID to 500 or 1000 (depending on your local policy) will disable incoming mail for system accounts.
Just in case that you need exceptions to the rule,
/etc/exim4/lowuid-aliases
is an alias
file that is only honored for local accounts with UID lower
than FIRST_USER_ACCOUNT_UID. If you define an alias for such an
account here, incoming mail is processed according to the
alias. If you alias the account to itself, messages are
delivered to the account itself, which is an exception to the
rule that messages for low-UID accounts are rejected. The
format of /etc/exim4/lowuid-aliases
is
just another alias file.
Sometimes, it might be desirable to be able to bypass local routing specialities like the alias file or a user-forward file. This is possible in the Debian Exim4 packages by prefixing the account name with "real-". For a local account name "foo", "real-foo@hostname.example" will result in direct delivery to foo's local Mailbox.
This feature is by default only available for locally
generated messages. If you want it to be accessible for
messages delivered from remote as well, set the Exim macro
COND_LOCAL_SUBMITTER to true. If you do not want this at all,
set the macro to false. Please note that the userforward
router uses this feature to get error messages delivered, i.e.
notifying the user of a syntax error in her
.forward
file.
Delivery to arbitrary files, directory or to pipes in the
/etc/aliases
file is disabled by default
in the Debian Exim 4 packages. The delivery process including the
program being piped to would run as the exim admin-user
Debian-exim, which might open up security holes.
Invoking pipes from /etc/aliases
file is
widely considered obsolete and deprecated. The Debian Exim
package maintainers would like to suggest using a dedicated
router/transport pair to invoke local processes for mail
processing. For example, the Debian mailman package contains a
/usr/share/doc/mailman/README.Exim4.Debian
file
that gives a good example how to implement this. Using a
dedicated router/transport pair have the following advantages:
The router/transport pair can be put in place by another package, giving a well-defined transaction point between Exim 4 and $PACKAGE.
Not allowing pipe deliveries from alias files makes it harder to accidentally run programs with wrong privileges.
It is possible to run different pipe processes under different accounts.
Even if only invoking a single local program, it is easier to do with your dedicated router/transport since you won't need to change this file, making automatic updates of this file possible for future versions of the Exim 4 packages. If you do local changes here, dpkg conffile handling will bother you on future updates.
If you insist on using /etc/aliases
in
the traditional way, you will need to activate the
respective functions by setting the transport options on the
system_aliases router appropriately. Macros are defined to make
this easier. See
/etc/exim4/conf.d/router/400_exim4-config_system_aliases
for information about which macros are available. You might
find the address_file, address_pipe and/or address_directory
transports that are used for the userforward router helpful in
writing your own transports for use in the system_aliases router.
If any of your aliases expand to pipes or files or directories you should set up a user and a group for these deliveries to run under. You can do this by setting the "user" and - if necessary - a "group" option and adding a "group" option if necessary. Alternatively, you can specify "user" and/or "group" on the transports that are used.
UUCP is a traditional way to execute remote jobs (e.g. spool mails), and as a lot of old things there are much more than one way to do it. However, today, the ways to handle it have boiled down to more or less two different ways.
Our recommendation is to use bsmtp/rsmtp wherever possible, because it supports all kinds of mail addresses (also the empty ones in bounces), and is also better from the security point of view.
rmail is the oldest way to transfer mail to a remote system. However, today it is normally required to use addresses with full domains for that (Well, they look like any normal address for you, and we do not tell about the other way to not confuse you ;). If you want this, you can use this transport:
rmail: debug_print = "T: rmail for $pipe_addresses" driver=pipe command = uux - -r -a$sender_address -gC $domain_data!rmail $pipe_addresses return_fail_output user=uucp batch_max = 20
However, all recipients are handled via the command line, so you are discouraged to use it.
This is a more efficient way to transfer mails. It works like sending SMTP via a pipe, but instead of waiting for an answer, the SMTP is just batched; from this is also the name batched SMTP or short bsmtp.
Furthermore, this way won't fail on addresses like " "@do.main. If you want this, please use this, if the remote site uses rsmtp (e.g. is Exim 4):
rsmtp: debug_print = "T: rsmtp for $pipe_addresses" driver=pipe command = /usr/bin/uux - -r -a$sender_address -gC $domain_data!rsmtp use_bsmtp return_fail_output user=uucp batch_max = 100
and this if it wants bsmtp as the command:
bsmtp: debug_print = "T: bsmtp for $pipe_addresses" driver=pipe command = /usr/bin/uux - -r -a$sender_address -gC $domain_data!bsmtp use_bsmtp return_fail_output user=uucp batch_max = 100
Of course, these examples can be extended for e.g. compression (but you can also use ssh for compression, if you want).
You need a router to tell Exim 4 which mails to forward to UUCP. You can use this one; please adopt the last line. Of course, it is also possible to send mail via more than one way.
uucp_router: debug_print = "R: uucp_router for $local_part@$domain" driver=accept require_files = +/usr/bin/uux domains = wildlsearch;/etc/exim4/uucp transport = rsmtp
The file /etc/exim4/uucp
looks like:
*.do.main uucp.name.of.remote.side
If you have a leaf system (i.e. all your mail not for your
local system goes to a single remote system), you can just
forward all non-local mail to the remote UUCP system. In
this case, you can replace "domains = ..." with "domains = !
+local_domains", but then you need also to replace
$domain_data in the transport by the UUCP-name of your
smarthost. The file /etc/exim4/uucp
is
not needed in this case.
Depending how much you trust your local users, you might use trusted_users and add uucp to it or use local_sender_retain=true and local_from_check=false.
Exim can run SpamAssassin while receiving a message by SMTP which allows one to avoid acceptance of spam messages. The Debian configuration contains some example code for running SpamAssassin, but like all filtering this needs to be handled carefully.
SpamAssassin's default report should not be used in a add_header statement since it contains empty lines. (This triggers e.g. Amavis' warning "BAD HEADER SECTION, Improper folded header field made up entirely of whitespace".) This is a safe, terse alternative:
clear_report_template report (_SCORE_ / _REQD_ requ) _TESTSSCORES(,)_ autolearn=_AUTOLEARN_
Rejecting spam messages: Do not reject spam-messages received on (non-spam) mailing lists, this can/will cause auto-unsubscription. This also applies to messages received via forwarding services (e.g. @debian.org addresses). If theses messages are rejected the forwarding services will need to send a bounce address to the spammer and will probably disable the forwarding if it happens all the time. You will need to have some kind of whitelist to exclude these hosts.
Security considerations: By default spamd runs as root and changes uid/gid to the requested user to run SpamAssassin. The example uses SpamAssassin default non-privileged user (nobody) which prevents use of Bayesian filtering since this requires persistent storage. You might want to setup a dedicated user for exim spam scanning and use that one, either for a separate SpamAssassin user profile or to run SpamAssassin as non-privileged user.
If you use exim4-config from Debian, you will get the debconf based configuration scheme that is intended to cover the majority of cases.
If exim4-config is installed while an Exim 3
package is present on the system,
exim4-config tries to parse the Exim 3 config
file to determine the answers that were given to
eximconfig on Exim 3 installation. These
answers are then taken as default values for the debconf based
configuration process. Be warned! eximconfig
from the Exim 3 packages does not record the explicit answers
given on Exim 3 configuration. So we have to guess the answers
from the Exim 3 configuration file
/etc/exim/exim.conf
, which is bound to fail
if the config file has been modified after using
eximconfig.
This is the reason why we refrained from doing a "silent update", but only use the guessed answers to get reasonable defaults for our debconf based configuration process.
Please note that we do not use the exim_convert4r4 script, but try to configure the Exim 4 package in the same way Exim 3 was. This will hopefully aid future updates.
If you have used a customized Exim 3 configuration, you can of
course use exim_convert4r4, and install the
resulting file as /etc/exim4/exim4.conf
after careful inspection. Exim 4 will then use that file and
ignore the file that it generated from the debconf
configuration. To aid future updates, we do, however, encourage
you not to use the
exim_convert4r4-generated
file verbatim but
instead drop appropriate configuration snippets in their
appropriate place in /etc/exim4/conf.d
.
On Debian systems the PAM modules run as the same user
as the calling program, so they cannot do anything you
could not do yourself, and in particular cannot access
/etc/shadow
unless the user is in group
shadow. - If you want to use
/etc/shadow
for Exim's SMTP AUTH you
will need to run exim as group shadow. Only
exim4-daemon-heavy is linked against libpam. We suggest using
saslauthd instead.
In the default configuration, Exim cannot locally deliver mail to accounts which have capitals in their name. This is caused by the fact that Exim converts the local part of incoming mail to lower case before the comparison done by the check_local_user directive in routers is done.
The router option caseful_local_part can be used to control this, and we decided not to set this option in the Debian configuration since it would be a rather big change to Exim's default behavior.
No Exim 4 version released with any Debian OS can run
deliveries as root. If you don't redirect mail for root via
/etc/aliases
to a nonprivileged
account, the mail will be delivered to
/var/mail/mail
with permissions 0600 and
owner mail:mail.
This redirection is done by the mail4root router which is last in the list and will thus catch mail for root that has not been taken care of earlier.
Most of the scripts that come with this Debian package do a set -x if invoked with the environment variable EX4DEBUG defined and non-zero. This is particularly handy if you need to debug the maintainer scripts that are invoked during package installation. Since dpkg redirects stdout of maintainer scripts, calling dpkg with EX4DEBUG set might yield interesting results. If in doubt, invoke the maintainer scripts with EX4DEBUG set manually directly from the command line.
There is no SELinux policy for Exim4 available so far. Until this is resolved, users should use postfix or sendmail if they intend to run SELinux.
The Debian Exim4 maintainers would appreciate if somebody could write an SELinux policy. We will gladly use them in the Debian packages as long as there is somebody available to test, debug and support.
/usr/sbin/exim_convert4r4.
Documentation updates
/path/to/sharedobject
.