Barricade MX
smtpf/2.6
|
route:local | FORWARD: localhost:26 |
route:127.0.0.1 | FORWARD: 127.0.0.1:26 |
route:::1 | FORWARD: [::1]:26 |
The local route is very important in smtpf. It is used for queuing outbound
messages from a RELAY
, queuing messages for SMTP authenticated senders, and
for processing messages for unqualified recipients, ie. those that have no domain name part
in the address. The local route can be a list of other hosts machines as shown below.
Note that the hosts specified in the local route must not use the smtpf
server as a smart-host, otherwise you can end up with a mail loop.
This example shows all the possible variants for the local route,
though the preferred form is route:local
. The other variants are
still supported for backwards compatibility with previous releases.
route:local FORWARD: [2001:db8::1]:26, other.host route:127.0.0.1 FORWARD: 192.0.2.1 route:::1 FORWARD: 2001:db8::1
After the local route, the next most important records that must be added to the route-map are the recipient domains we will accept mail for. For example:
route:example.com FORWARD: 127.0.0.1:26, 192.168.0.9 route:snertsoft.com FORWARD: 192.0.2.1 route:info@snertsoft.com FORWARD: 192.0.2.2
The FORWARD:
parameter tells smtpf where to send a message for a recipient
of that domain. When you specify more than one FORWARD:
host, as was done for
route:example.com above, then they are tried in left to right order until one of them
answers (see also route-forward-selection). If no host answers, then that recipient will be
temporarily rejected.
Note that it is also possible to route individual mail addresses as shown by route:info@snertsoft.com to a different host, instead of the default for the domain, route:snertsoft.com.
When creating a route-map, it is usually a good idea to specify which hosts or subnets from your LAN are allowed to relay. For example:
route:10.0.0.1 RELAY route:192.168.0 RELAY route:FE80:0000 RELAY route:example.com FORWARD: 192.168.1.2; RELAY
The RELAY
tag should be used sparingly, typically only for your
LAN and trusted hosts. Any client connection that is marked as a RELAY, will
have their mail sent to the local route for queuing and white listed
through many, but not all, of the tests (anti-virus scanning will always be done
when enabled and if required, EMEW transformation of the Message-Id header).
When a RELAY
connects, all mail is sent to the local route
for queuing. It is imperative that the SMTP servers handling the local route
know how to relay their domains correctly, otherwise mail may fail to be delivered.
The route:example.com shows an instance
where inbound mail destined for example.com is forwarded to a specific host
on the LAN and outbound mail from any host within the example.com domain is
relayed. The point to note here is that inbound mail route selection is taken based on the domain
name of the recipient address, while relay selection is based on the
DNS PTR record found based on the connected SMTP
client's IP address. If you do not have reverse DNS entries for your internal hosts, then
you must specify RELAY
entries by IP address or family as shown above.
The RCPT:
parameter tells smtpf to perform a call-ahead
to another set of hosts in order to verify if the recipient is valid. This parameter is optional
and is only recommended when the FORWARD:
host is some intermediate MTA,
such as an anti-virus appliance or gateway machine that has no knowledge of recipients
for that domain. RCPT:
allows you to jump over one or more intermediate
servers to talk directly with the final mail store.
Accept and reject results from a call-ahead are cached for future use, while
temporary failure results are not cached at all. Only the expire time of an accept result
will be touched (see cache-accept-ttl).
route:example.net FORWARD: scan.our.domain; RCPT: 192.168.1.2 route:example.org FORWARD: scan.our.domain; RCPT: exchange.our.domain
Note that some MTA by default use an accept-then-bounce model and so do not work with call-ahead very well, such as default configurations for Microsoft Exchange and Domino servers. An accept-then-bounce server will always accept any SMTP RCPT TO: command whether it be valid or not. Typically such servers can be configured to verify the RCPT TO: argument the moment the command is received (instead of delaying until after the message body) so as to provide a more useful and immediate response to the call-ahead.
The call-ahead test will perform a false-RCPT test to detect accept-then-bounce
servers and cache the result. If the server does accept-then-bounce, then future
call-ahead attempts will be skipped. Alternatively, if an
accept-then-bounce server cannot be configured otherwise, then do not use the
RCPT:
parameter.
smtpf has support for SMTP AUTH PLAIN and AUTH LOGIN methods. If you need SMTP AUTH support, then be sure to enable both smtp-enable-esmtp, auth-delay-checks, which will postpone some connection and HELO/EHLO tests until a MAIL FROM: is received.
If the authentication-Id given to the AUTH command is fully qualified, ie. it has the form
userid@sender.domain
, then AUTH command is first proxied to the RCPT:
or
FORWARD:
host of the route:sender.domain; if that fails,
then the unqualified version of the authentication-Id is tried.
If the authentication-Id is not qualified, ie. is just a simple userid
, then the
AUTH command is proxied to the RCPT:
or FORWARD:
host of the local route.
All mail sent over an authenticated session is forwarded (queued) to the SMTP host that authenticated the credentials. SMTP AUTH support in smtpf is limited and it is recommended that a proper MSA be employed.
smtpf has support for SMTP ETRN, where ETRN commands are simply relayed to the local route, which is responsible for queuing. If the local route refers to more than one forward host, then those hosts must all share the same mail queue, otherwise the behaviour is undefined.
- TOP -