postfwd2 - postfix firewall daemon


postfwd2 [OPTIONS] [SOURCE1, SOURCE2, ...]

        Ruleset: (at least one, multiple use is allowed):
        -f, --file <file>               reads rules from <file>
        -r, --rule <rule>               adds <rule> to config
        -s, --scores <v>=<r>            returns <r> when score exceeds <v>
        -i, --interface <dev>           listen on interface <dev>
        -p, --port <port>               listen on port <port>
            --proto <proto>             socket type (tcp or unix)
            --server_socket <sock>      e.g. tcp:
        -u, --user <name>               set uid to user <name>
        -g, --group <name>              set gid to group <name>
            --pidfile <path>            create pidfile under <path>
            --min_servers <i>           spawn at least <i> children
            --max_servers <i>           do not spawn more than <i> children
            --min_spare_servers <i>     minimum idle children
            --max_spare_servers <i>     maximum idle children
        -c, --cache <int>               sets the request-cache timeout to <int> seconds
            --cleanup-requests <int>    cleanup interval in seconds for request cache
            --cache_interface <dev>     listen on interface <dev>
            --cache_port <port>         listen on port <port>
            --cache_proto <proto>       socket type (tcp or unix)
            --cache_socket <sock>       e.g. tcp:
            --cacheid <list>            list of request items for cache-id
            --cache-rdomain-only        skip recipient localpart for cache-id
            --cache-no-sender           skip sender address for cache-id
            --cache-no-size             skip size for cache-id
            --no_parent_request_cache   disable parent request cache
            --no_parent_rate_cache      disable parent rate cache
            --no_parent_dns_cache       disable parent dns cache
            --no_parent_cache           disable all parent caches
            --cleanup-rates <int>       cleanup interval in seconds for rate cache
        -k, --kill, --stop              terminate postfwd2
            --reload, --hup             reload postfwd2
            --watchdog <w>              watchdog timer in seconds
            --respawn <r>               respawn delay in seconds
            --failures <f>              max respawn failure counter
            --daemons <list>            list of daemons to start
            --dumpcache                 show cache contents
            --dumpstats                 show statistics
        -n, --nodns                     skip any dns based test
            --dns_timeout <i>           dns query timeout in seconds
            --dns_timeout_max <i>       disable dnsbl after <i> timeouts
            --dns_timeout_interval <i>  reenable dnsbl after <i> seconds
            --cache-rbl-timeout <i>     default dns ttl if not specified in ruleset
            --cache-rbl-default <s>     default dns pattern if not specified in ruleset
            --cleanup-rbls <i>          cleanup old dns cache items every <i> seconds
            --dns_async_txt             perform dnsbl A and TXT lookups simultaneously
            --dns_max_ns_lookups        max names to look up with sender_ns_addrs
            --dns_max_mx_lookups        max names to look up with sender_mx_addrs
        -t, --test                      testing, always returns "dunno"
        -S, --summary <i>               show stats every <i> seconds
            --noidlestats               disables statistics when idle
            --norulestats               disables per rule statistics
        -I, --instantcfg                reloads ruleset on every new request
            --config_timeout <i>        parser timeout in seconds
        -l, --logname <label>           label for syslog messages
            --facility <s>              use syslog facility <s>
            --nodnslog                  do not log dns results
            --anydnslog                 log any dns (even cached) results
            --norulelog                 do not log rule actions
            --nolog|--perfmon           no logging at all
        -v, --verbose                   verbose logging, use twice to increase
            --debug <s>                 list of debugging classes
        Information (use only at command-line!):
        -h, --help                      display this help and exit
        -m, --manual                    shows program manual
        -V, --version                   output version information and exit
        -D, --defaults                  show postfwd2 settings and exit
        -C, --showconfig                show postfwd2 ruleset and exit (-v allowed)
        -L, --stdout                    redirect syslog messages to stdout
        -q, --quiet                     no syslogging, no stdout (-P works for compatibility)
        Obsolete (only for compatibility with postfwd v1):
        -d|--daemon, --shortlog, --dns_queuesize, --dns_retries



postfwd2 is written to combine complex postfix restrictions in a ruleset similar to those of the most firewalls. The program uses the postfix policy delegation protocol to control access to the mail system before a message has been accepted (please visit for more information).

postfwd2 allows you to choose an action (e.g. reject, dunno) for a combination of several smtp parameters (like sender and recipient address, size or the client's TLS fingerprint). Also it offers simple macros/acls which should allow straightforward and easy-to-read configurations.


* Complex combinations of smtp parameters

* Combined RBL/RHSBL lookups with arbitrary actions depending on results

* Scoring system

* Date/time based rules

* Macros/ACLs, Groups, Negation

* Compare request attributes (e.g. client_name and helo_name)

* Internal caching for requests and dns lookups

* Built in statistics for rule efficiency analysis


A configuration line consists of optional item=value pairs, separated by semicolons (`;`) and the appropriate desired action:

        [ <item1>[=><~]=<value>; <item2>[=><~]=<value>; ... ] action=<result>


        client_address= ; sender==no@bad.local ; action=REJECT

This will deny all mail from with envelope sender no@bad.local. The order of the elements is not important. So the following would lead to the same result as the previous example:

        action=REJECT ; client_address= ; sender==no@bad.local

The way how request items are compared to the ruleset can be influenced in the following way:

         ITEM == VALUE                true if ITEM equals VALUE
         ITEM => VALUE                true if ITEM >= VALUE
         ITEM =< VALUE                true if ITEM <= VALUE
         ITEM =~ VALUE                true if ITEM ~= /^VALUE$/i
         ITEM != VALUE                false if ITEM equals VALUE
         ITEM !> VALUE                false if ITEM >= VALUE
         ITEM !< VALUE                false if ITEM <= VALUE
         ITEM !~ VALUE                false if ITEM ~= /^VALUE$/i
         ITEM =  VALUE                default behaviour (see ITEMS section)

To identify single rules in your log files, you may add an unique identifier for each of it:

        id=R_001 ; action=REJECT ; client_address= ; sender==no@bad.local

You may use these identifiers as target for the `jump()` command (see ACTIONS section below). Leading or trailing whitespace characters will be ignored. Use '#' to comment your configuration. Others will appreciate.

A ruleset consists of one or multiple rules, which can be loaded from files or passed as command line arguments. Please see the COMMAND LINE section below for more information on this topic.

Rules can span multiple lines by adding a trailing backslash ``\'' character:

        id=R_001 ;  client_address=; sender==no@bad.local; \
                    action=REJECT please use your relay from there


        id                      - a unique rule id, which can be used for log analysis
                                  ids also serve as targets for the "jump" command.
        date, time              - a time or date range within the specified rule shall hit
                                  # FORMAT:
                                  # Feb, 29th
                                  # Dec, 24th - 26th
                                  # from today until Nov, 23rd
                                  # from April, 1st until today
        days, months            - a range of weekdays (Sun-Sat) or months (Jan-Dec)
                                  within the specified rule shall hit
        score                   - when the specified score is hit (see ACTIONS section)
                                  the specified action will be returned to postfix
                                  scores are set global until redefined!
        request_score           - this value allows to access a request's score. it
                                  may be used as variable ($$request_score).
        rbl, rhsbl,             - query the specified RBLs/RHSBLs, possible values are:
        rhsbl_client,             <name>[/<reply>/<maxcache>, <name>/<reply>/<maxcache>]
        rhsbl_sender,             (defaults: reply=^127\.0\.0\.\d+$ maxcache=3600)
        rhsbl_reverse_client      the results of all rhsbl_* queries will be combined
                                  in rhsbl_count (see below).
        rblcount, rhsblcount    - minimum RBL/RHSBL hitcounts to match. if not specified
                                  a single RBL/RHSBL hit will match the rbl/rhsbl items.
                                  you may specify 'all' to evaluate all items, and use
                                  it as variable in an action (see ACTIONS section)
                                  (default: 1)
        sender_localpart,       - the local-/domainpart of the sender address
        recipient_localpart,    - the local-/domainpart of the recipient address
        helo_address            - postfwd2 tries to look up the helo_name. use
                                  helo_address=!!( to check for unknown.
                                  Please do not use this for positive access control
                                  (whitelisting), as it might be forged.
        sender_ns_names,        - postfwd2 tries to look up the names/ip addresses
        sender_ns_addrs           of the nameservers for the sender domain part.
                                  Please do not use this for positive access control
                                  (whitelisting), as it might be forged.
        sender_mx_names,        - postfwd2 tries to look up the names/ip addresses
        sender_mx_addrs           of the mx records for the sender domain part.
                                  Please do not use this for positive access control
                                  (whitelisting), as it might be forged.
        version                 - postfwd2 version, contains "postfwd2 n.nn"
                                  this enables version based checks in your rulesets
                                  (e.g. for migration). works with old versions too,
                                  because a non-existing item always returns false:
                                  id=R01; version~=1.10; \
                                        ; action=REJECT sorry no access

Besides these you can specify any attribute of the postfix policy delegation protocol. Feel free to combine them the way you need it (have a look at the EXAMPLES section below).

Most values can be specified as regular expressions (PCRE). Please see the table below for details:

        # ==========================================================
        # ITEM=VALUE                            TYPE
        # ==========================================================
        id=something                            mask = string
        date=01.04.2007-22.04.2007              mask = date (DD.MM.YYYY-DD.MM.YYYY)
        time=08:30:00-17:00:00                  mask = time (HH:MM:SS-HH:MM:SS)
        days=Mon-Wed                            mask = weekdays (Mon-Wed) or numeric (1-3)
        months=Feb-Apr                          mask = months (Feb-Apr) or numeric (1-3)
        score=5.0                               mask = maximum floating point value                    mask = <name>/<reply>/<maxcache>[,...]
        rblcount=2                              mask = numeric, will match if rbl hits >= 2
        helo_address=<a.b.c.d/nn>               mask = CIDR[,CIDR,...]
        sender_ns_names=some.domain.tld         mask = PCRE
        sender_mx_names=some.domain.tld         mask = PCRE
        sender_ns_addrs=<a.b.c.d/nn>            mask = CIDR[,CIDR,...]
        sender_mx_addrs=<a.b.c.d/nn>            mask = CIDR[,CIDR,...]
        # ------------------------------
        # Postfix version 2.1 and later:
        # ------------------------------
        client_address=<a.b.c.d/nn>             mask = CIDR[,CIDR,...]
        client_name=another.domain.tld          mask = PCRE
        reverse_client_name=another.domain.tld  mask = PCRE
        helo_name=some.domain.tld               mask = PCRE
        sender=foo@bar.tld                      mask = PCRE
        recipient=bar@foo.tld                   mask = PCRE
        recipient_count=5                       mask = numeric, will match if recipients >= 5
        # ------------------------------
        # Postfix version 2.2 and later:
        # ------------------------------
        sasl_method=plain                       mask = PCRE
        sasl_username=you                       mask = PCRE
        sasl_sender=                            mask = PCRE
        size=12345                              mask = numeric, will match if size >= 12345
        ccert_subject=blackhole.nowhere.local   mask = PCRE (only if tls verified)
        ccert_issuer=John+20Doe                 mask = PCRE (only if tls verified)
        ccert_fingerprint=AA:BB:CC:DD:EE:...    mask = PCRE (do NOT use "..." here)
        # ------------------------------
        # Postfix version 2.3 and later:
        # ------------------------------
        encryption_protocol=TLSv1/SSLv3         mask = PCRE
        encryption_cipher=DHE-RSA-AES256-SHA    mask = PCRE
        encryption_keysize=256                  mask = numeric, will match if keysize >= 256

the current list can be found at Please read carefully about which attribute can be used at which level of the smtp transaction (e.g. size will only work reliably at END_OF_DATA level). Pattern matching is performed case insensitive.

Multiple use of the same item is allowed and will compared as logical OR, which means that this will work as expected:

        id=TRUST001; action=OK; encryption_keysize=64;          \
                ccert_fingerprint=11:22:33:44:55:66:77:88:99;   \
                ccert_fingerprint=22:33:44:55:66:77:88:99:00;   \
                ccert_fingerprint=33:44:55:66:77:88:99:00:11;   \

client_address, rbl and rhsbl items may also be specified as whitespace-or-comma-separated values:

        id=SKIP01; action=dunno; \
        id=SKIP02; action=dunno; \
                client_address=       \

The following items must be unique:

        id, minimum and maximum values, rblcount and rhsblcount

Any item can be negated by preceeding '!!' to it, e.g.:

        id=TLS001 ;  hostname=!!^secure\.trust\.local$ ;  action=REJECT only please

or using the right compare operator:

        id=USER01 ;  sasl_username !~ /^(bob|alice)$/ ;  action=REJECT who is that?

To avoid confusion with regexps or simply for better visibility you can use '!!(...)':

        id=USER01 ;  sasl_username=!!( /^(bob|alice)$/ )  ;  action=REJECT who is that?

Request attributes can be compared by preceeding '$$' characters, e.g.:

        id=R-003 ;  client_name = !! $$helo_name      ;  action=WARN helo does not match DNS
        # or
        id=R-003 ;  client_name = !!($$(helo_name))   ;  action=WARN helo does not match DNS

This is only valid for PCRE values (see list above). The comparison will be performed as case insensitive exact match. Use the '-vv' option to debug.


Since postfwd1 v1.15 and postfwd2 v0.18 long item lists can be stored in separate files:

        id=R001 ;  ccert_fingerprint==file:/etc/postfwd/wl_ccerts ;  action=DUNNO

postfwd2 will read a list of items (one item per line) from /etc/postfwd/wl_ccerts. comments are allowed:

        # client1
        # client2
        # client3

To use existing tables in key=value format, you can use:

        id=R001 ;  ccert_fingerprint==table:/etc/postfwd/wl_ccerts ;  action=DUNNO

This will ignore the right-hand value. Items can be mixed:

        id=R002 ;  action=REJECT \
                client_name==unknown; \

and for non pcre (comma separated) items:

        id=R003 ;  action=REJECT \
                client_address==, file:/etc/postfwd/blacklisted
        id=R004 ;  action=REJECT \
                rbl=myrbl.home.local,, file:/etc/postfwd/rbls_changing

You can check your configuration with the --show_config option at the command line:

        # postfwd2 --showconfig --rule='action=DUNNO; client_address=, file:/etc/postfwd/wl_clients,'

should give something like:

        Rule   0: id->"R-0"; action->"DUNNO"; client_address->"=;, =;, =;, =;"

If a file can not be read, it will be ignored:

        # postfwd2 --showconfig --rule='action=DUNNO; client_address=, file:/etc/postfwd/wl_clients,'
        [LOG warning]: error: file /etc/postfwd/wl_clients not found - file will be ignored ?
        Rule   0: id->"R-0"; action->"DUNNO"; client_address->"=;, =;"

File items are evaluated at configuration stage. Therefore postfwd2 needs to be reloaded if a file has changed

If you want to specify a file, that will be reloaded for each request, you can use lfile: and ltable:

        id=R001; client_address=lfile:/etc/postfwd/client_whitelist; action=dunno

This will check the modification time of /etc/postfwd/client_whitelist every time the rule is evaluated and reload it as necessary. Of course this might increase the system load, so please use it with care.

The --showconfig option illustrates the difference:

        ## evaluated at configuration stage
        # postfwd2 --nodaemon -L --rule='client_address=table:/etc/postfwd/clients; action=dunno' -C
        Rule   0: id->"R-0"; action->"dunno"; client_address->"=;, =;, =;"
        ## evaluated for any rulehit
        # postfwd2 --nodaemon -L --rule='client_address=ltable:/etc/postfwd/clients; action=dunno' -C
        Rule   0: id->"R-0"; action->"dunno"; client_address->"=;ltable:/etc/postfwd/clients"

Files can refer to other files. The following is valid.

        -- FILE /etc/postfwd/ --
        id=R01; client_address=file:/etc/postfwd/; action=DUNNO
        -- FILE /etc/postfwd/ --
        -- FILE /etc/postfwd/ --
        -- FILE /etc/postfwd/ --

Remind that there is currently no loop detection (/a/file calls /a/file) and that this feature is only available with postfwd1 v1.15 and postfwd2 v0.18 and higher.



Actions will be executed, when all rule items have matched a request (or at least one of any item list). You can refer to request attributes by preceeding $$ characters, like:

        id=R-003; client_name = !!$$helo_name; action=WARN helo '$$helo_name' does not match DNS '$$client_name'
        # or
        id=R-003; client_name = !!$$helo_name; action=WARN helo '$$(helo_name)' does not match DNS '$$(client_name)'

postfix actions

Actions will be replied to postfix as result to policy delegation requests. Any action that postfix understands is allowed - see ``man 5 access'' or for a description. If no action is specified, the postfix WARN action which simply logs the event will be used for the corresponding rule.

postfwd2 will return dunno if it has reached the end of the ruleset and no rule has matched. This can be changed by placing a last rule containing only an action statement:

        action=dunno ; sender=@domain.local     # sender is ok
        action=reject                           # default deny

postfwd2 actions

postfwd2 actions control the behaviour of the program. Currently you can specify the following:

        jump (<id>)
        jumps to rule with id <id>, use this to skip certain rules.
        you can jump backwards - but remember that there is no loop
        detection at the moment! jumps to non-existing ids will be skipped.
        score (<score>)
        the request's score will be modified by the specified <score>,
        which must be a floating point value. the modificator can be either
                +n.nn   adds n.nn to current score
                -n.nn   sustracts n.nn from the current score
                *n.nn   multiplies the current score by n.nn
                /n.nn   divides the current score through n.nn
                =n.nn   sets the current score to n.nn
        if the score exceeds the maximum set by `--scores` option (see
        COMMAND LINE) or the score item (see ITEMS section), the action
        defined for this case will be returned (default: 5.0=>"REJECT postfwd2 score exceeded").
        set (<item>=<value>,<item>=<value>,...)
        this command allows you to insert or override request attributes, which then may be
        compared to your further ruleset. use this to speed up repeated comparisons to large item lists.
        please see the EXAMPLES section for more information. you may separate multiple key=value pairs
        by "," characters.
        rate (<item>/<max>/<time>/<action>)
        this command creates a counter for the given <item>, which will be increased any time a request
        containing it arrives. if it exceeds <max> within <time> seconds it will return <action> to postfix.
        rate counters are very fast as they are executed before the ruleset is parsed.
            # no more than 3 requests per 5 minutes
            # from the same "unknown" client
            id=RATE01 ;  client_name==unknown ; \
               action==rate($$client_address/3/300/450 4.7.1 sorry, max 3 requests per 5 minutes)
        size (<item>/<max>/<time>/<action>)
        this command works similar to the rate() command with the difference, that the rate counter is
        increased by the request's size attribute. to do this reliably you should call postfwd2 from
        smtpd_end_of_data_restrictions. if you want to be sure, you could check it within the ruleset:
           # size limit 1.5mb per hour per client
           id=SIZE01 ;  state==END_OF_DATA ;  client_address==!!(; \
              action==size($$client_address/1572864/3600/450 4.7.1 sorry, max 1.5mb per hour)
        ask (<addr>:<port>[:<ignore>])
        allows to delegate the policy decision to another policy service (e.g. postgrey). the first
        and the second argument (address and port) are mandatory. a third optional argument may be
        specified to tell postfwd2 to ignore certain answers and go on parsing the ruleset:
           # example1: query postgrey and return it's answer to postfix
           id=GREY; client_address==; action=ask(
           # example2: query postgrey but ignore it's answer, if it matches 'DUNNO'
           # and continue parsing postfwd's ruleset
           id=GREY; client_address==; action=ask(^dunno$)
        wait (<delay>)
        pauses the program execution for <delay> seconds. use this for
        delaying or throtteling connections.
        note (<string>)
        just logs the given string and continues parsing the ruleset.
        if the string is empty, nothing will be logged.
        quit (<code>)
        terminates the program with the given exit-code. postfix doesn`t
        like that too much, so use it with care.

You can reference to request attributes, like

        id=R-HELO ;  helo_name=^[^\.]+$ ;  action=REJECT invalid helo '$$helo_name'

These special attributes will be reset for any new rule:

        rblcount        - contains the number of RBL answers
        rhsblcount      - contains the number of RHSBL answers
        matches         - contains the number of matched items
        dnsbltext       - contains the dns TXT part of all RBL and RHSBL replies in the form
                          rbltype:rblname:<txt>; rbltype:rblname:<txt>; ...

These special attributes will be changed for any matching rule:

        request_hits    - contains ids of all matching rules

This means that it might be necessary to save them, if you plan to use these values in later rules:

        # set vals
        id=RBL01 ; rhsblcount=all ; rblcount=all ; \
      ,,, ; \
      ,, ; \
      ,, ; \
        # compare
        id=RBL02 ; HIT_rhls>=1 ; HIT_rbls>=1 ; action=554 5.7.1 blocked using $$HIT_rhls RHSBLs and $$HIT_rbls RBLs [INFO: $$HIT_txt]
        id=RBL03 ; HIT_rhls>=2               ; action=554 5.7.1 blocked using $$HIT_rhls RHSBLs [INFO: $$HIT_txt]
        id=RBL04 ; HIT_rbls>=2               ; action=554 5.7.1 blocked using $$HIT_rbls RBLs [INFO: $$HIT_txt]


Multiple use of long items or combinations of them may be abbreviated by macros. Those must be prefixed by '&&' (two '&' characters). First the macros have to be defined as follows:

        &&RBLS {,,,,; };

Then these may be used in your rules, like:

        &&RBLS ;  client_name=^unknown$                         ; action=REJECT
        &&RBLS ;  client_name=(\d+[\.-_]){4}                    ; action=REJECT
        &&RBLS ;  client_name=[\.-_](adsl|dynamic|ppp|)[\.-_]   ; action=REJECT

Macros can contain actions, too:

        # definition
        &&GONOW { action=REJECT your request caused our spam detection policy to reject this message. More info at http://www.domain.local; };
        # rules
        &&GONOW ;  &&RBLS ;  client_name=^unknown$
        &&GONOW ;  &&RBLS ;  client_name=(\d+[\.-_]){4}
        &&GONOW ;  &&RBLS ;  client_name=[\.-_](adsl|dynamic|ppp|)[\.-_]

Macros can contain macros, too:

        # definition (note the trailing "\" characters)
        &&RBLS {                                                \
       ;                          \
       ;                             \
       ;                            \
       ;                           \
       ;                       \
        &&DYNAMIC {                                             \
                client_name=^unknown$ ;                         \
                client_name=(\d+[\.-_]){4} ;                    \
                client_name=[\.-_](adsl|dynamic|ppp|)[\.-_] ;   \
        &&GOAWAY { &&RBLS; &&DYNAMIC; };
        # rules
        &&GOAWAY ; action=REJECT dynamic client and listed on RBL

Basically macros are simple text substitutions - see the PARSER section for more information.


Please visit



The following arguments are used to specify the source of the postfwd2 ruleset. This means that at least one of the following is required for postfwd2 to work.

        -f, --file <file>
        Reads rules from <file>. Please see the CONFIGURATION section
        below for more information.
        -r, --rule <rule>
        Adds <rule> to ruleset. Remember that you might have to quote
        strings that contain whitespaces or shell characters.


        A file containing plugin routines for postfwd. Please see the
        PLUGINS section for more information.


        -s, --scores <val>=<action>
        Returns <action> to postfix, when the request's score exceeds <val>

Multiple usage is allowed. Just chain your arguments, like:

        postfwd2 -r "<item>=<value>;action=<result>" -f <file> -f <file> --plugins <file> ...
        postfwd2 --scores 4.5="WARN high score" --scores 5.0="REJECT postfwd2 score too high" ...

In case of multiple scores, the highest match will count. The order of the arguments will be reflected in the postfwd2 ruleset.


postfwd2 can be run as daemon so that it listens on the network for incoming requests. The following arguments will control it's behaviour in this case.

        -d, --daemon
        postfwd2 will run as daemon and listen on the network for incoming
        queries (default
        -i, --interface <dev>
        Bind postfwd2 to the specified interface (default
        -p, --port <port>
        postfwd2 listens on the specified port (default tcp/10040).
        --proto <type>
        The protocol type for postfwd's socket. Currently you may use 'tcp' or 'unix' here.
        To use postfwd2 with a unix domain socket, run it as follows:
            postfwd2 --proto=unix --port=/somewhere/postfwd.socket
        -u, --user <name>
        Changes real and effective user to <name>.
        -g, --group <name>
        Changes real and effective group to <name>.
        -R, --chroot <path>
        Chroot the process to the specified path.
        Test this before using - you might need some libs there.
        --pidfile <path>
        The process id will be saved in the specified file.
        -l, --logname <label>
        Labels the syslog messages. Useful when running multiple
        instances of postfwd.
        --loglen <int>
        Truncates any syslog message after <int> characters.

Optional arguments

These parameters influence the way postfwd2 is working. Any of them can be combined.

        -v, --verbose
        Verbose logging displays a lot of useful information but can cause
        your logfiles to grow noticeably. So use it with caution. Set the option
        twice (-vv) to get more information (logs all request attributes).
        -c, --cache <int>    (default=600)
        Timeout for request cache, results for identical requests will be
        cached until config is reloaded or this time (in seconds) expired.
        A setting of 0 disables this feature.
        Ignores size attribute for cache comparisons which will lead to better
        cache-hit rates. You should set this option, if you don't use the size
        item in your ruleset.
        Ignores sender address for cache comparisons which will lead to better
        cache-hit rates. You should set this option, if you don't use the sender
        item in your ruleset.
        This will strip the localpart of the recipient's address before filling the
        cache. This may considerably increase cache-hit rates.
        --cache-rbl-timeout <timeout>     (default=3600)
        This default value will be used as timeout in seconds for rbl cache items,
        if not specified in the ruleset.
        --cache-rbl-default <pattern>    (default=^127\.0\.0\.\d+$)
        Matches <pattern> to rbl/rhsbl answers (regexp) if not specified in the ruleset.
        --cacheid <item>, <item>, ...
        This csv-separated list of request attributes will be used to construct
        the request cache identifier. Use this only, if you know exactly what you
        are doing. If you, for example, use postfwd2 only for RBL/RHSBL control,
        you may set this to
                postfwd2 --cache=3600 --cacheid=client_name,client_address
        This increases efficiency of caching and improves postfwd's performance.
        Warning: You should list all items here, which are used in your ruleset!
        --cleanup-requests <interval>    (default=600)
        The request cache will be searched for timed out items after this <interval> in
        seconds. It is a minimum value. The cleanup process will only take place, when
        a new request arrives.
        --cleanup-rbls <interval>    (default=600)
        The rbl cache will be searched for timed out items after this <interval> in
        seconds. It is a minimum value. The cleanup process will only take place, when
        a new request arrives.
        --cleanup-rates <interval>    (default=600)
        The rate cache will be searched for timed out items after this <interval> in
        seconds. It is a minimum value. The cleanup process will only take place, when
        a new request arrives.
        -S, --summary <int>    (default=600)
        Shows some usage statistics (program uptime, request counter, matching rules)
        every <int> seconds. This option is included by the -v switch.
        This feature uses the alarm signal, so you can force postfwd2 to dump the stats
        using `kill -ALRM <pid>` (where <pid> is the process id of postfwd).
        Aug 19 12:39:45 mail1 postfwd[666]: [STATS] Counters: 213000 seconds uptime, 39 rules
        Aug 19 12:39:45 mail1 postfwd[666]: [STATS] Requests: 71643 overall, 49 last interval, 62.88% cache hits
        Aug 19 12:39:45 mail1 postfwd[666]: [STATS] Averages: 20.18 overall, 4.90 last interval, 557.30 top
        Aug 19 12:39:45 mail1 postfwd[666]: [STATS] Contents: 44 cached requests, 239 cached dnsbl results
        Aug 19 12:39:45 mail1 postfwd[666]: [STATS] Rule ID: R-001   matched: 2704 times
        Aug 19 12:39:45 mail1 postfwd[666]: [STATS] Rule ID: R-002   matched: 9351 times
        Aug 19 12:39:45 mail1 postfwd[666]: [STATS] Rule ID: R-003   matched: 3116 times
        Disables per rule statistics. Keeps your log clean, if you do not use them.
        This option has no effect without --summary or --verbose set.
        -L, --stdout
        Redirects all syslog messages to stdout for debugging. Never use this with postfix!
        -t, --test
        In test mode postfwd2 always returns "dunno", but logs according
        to it`s ruleset. -v will be set automatically with this option.
        -n, --nodns
        Disables all DNS based checks like RBL checks. Rules containing
        such elements will be ignored.
        -n, --nodnslog
        Disables logging of dns events.
        --dns_timeout     (default: 14)
        Sets the timeout for asynchonous dns queries in seconds. This value will apply to
        all dns items in a rule.
        --dns_timeout_max    (default: 10)
        Sets the maximum timeout counter for dnsbl lookups. If the timeouts exceed this value
        the corresponding dnsbl will be deactivated for a while (see --dns_timeout_interval).
        --dns_timeout_interval    (default=1200)
        The dnsbl timeout counter will be cleaned after this interval in seconds. Use this
        in conjunction with the --dns_timeout_max parameter.
        Perform dnsbl A and TXT lookups simultaneously (otherwise only for listings with at
        least one A record). This needs more network bandwidth due to increased queries but
        might increase throughput because the lookups can be parallelized.
        --dns_max_ns_lookups     (default=0)
        maximum ns names to lookup up with sender_ns_addrs item. use 0 for no maximum.
        --dns_max_mx_lookups     (default=0)
        maximum mx names to lookup up with sender_mx_addrs item. use 0 for no maximum.
        -I, --instantcfg
        The config files, specified by -f will be re-read for every request
        postfwd2 receives. This enables on-the-fly configuration changes
        without restarting. Though files will be read only if necessary
        (which means their access times changed since last read) this might
        significantly increase system load.
        --config_timeout    (default=3)
        timeout in seconds to parse a single configuration line. if exceeded, the rule will
        be skipped. this is used to prevent problems due to large files or loops.
I<Informational arguments>

These arguments are for command line usage only. Never ever use them with postfix!

        -C, --showconfig
        Displays the current ruleset. Use -v for verbose output.
        -V, --version
        Displays the program version.
        -h, --help
        Shows program usage.
        -m, --manual
        Displays the program manual.
        -D, --defaults
        displays complete postfwd2 settings.
        -P, --perfmon
        This option turns of any syslogging and output. It is included
        for performance testing.


In daemon mode postfwd2 reloads it's ruleset after receiving a HUP signal. Please see the description of the '-I' switch to have your configuration refreshed for every request postfwd2 receives.


        ## whitelisting
        # 1. networks,
        # 2. client_names * and *
        # 3. sender *@someshop.tld from
        id=WL001; action=dunno ; client_address=,
        id=WL002; action=dunno ; client_name=\.gmx\.(net|de)$
        id=WL003; action=dunno ; sender=@someshop\.tld$ ; client_address=
        ## TLS control
        # 1. *@authority.tld only with correct TLS fingerprint
        # 2. *@secret.tld only with keysizes >=64
        id=TL001; action=dunno                          ; sender=@authority\.tld$ ; ccert_fingerprint=AA:BB:CC..
        id=TL002; action=REJECT wrong TLS fingerprint   ; sender=@authority\.tld$
        id=TL003; action=REJECT tls keylength < 64      ; sender=@secret\.tld$ ; encryption_keysize=64
        ## Combined RBL checks
        # This will reject mail if
        # 1. listed on
        # 2. listed on (sbl and xbl, dns cache timeout 1200s instead of 3600s)
        # 3. listed on min 2 of,,
        # 4. listed on and one of,
        id=RBL01 ; action=REJECT listed on  ;
        id=RBL02 ; action=REJECT listed on     ;[2-8]/1200
        id=RBL03 ; action=REJECT listed on too many RBLs        ; rblcount=2 ;,,
        id=RBL04 ; action=REJECT combined RBL+RHSBL check       ; ;,
        ## Message size (requires message_size_limit to be set to 30000000)
        # 1. 30MB for systems in *.customer1.tld
        # 2. 20MB for SASL user joejob
        # 3. 10MB default
        id=SZ001; state==END-OF-MESSAGE; action=REJECT message too large; size=30000000 ; client_name=\.customer1.tld$
        id=SZ002; state==END-OF-MESSAGE; action=REJECT message too large; size=20000000 ; sasl_username==joejob
        id=SZ003; state==END-OF-MESSAGE; action=REJECT message too large; size=10000000
        ## Selective Greylisting
        # 1. if listed on with results or .11, dns cache timeout 1200s
        # 2. Client has no rDNS
        # 3. Client comes from several dialin domains
        id=GR001; action=greylisting ;,[01]/1200
        id=GR002; action=greylisting ; client_name=^unknown$
        id=GR003; action=greylisting ; client_name=\.(t-ipconnect|alicedsl|ish)\.de$
        ## Date Time
        date=24.12.2007-26.12.2007          ;  action=450 4.7.1 office closed during christmas
        time=04:00:00-05:00:00              ;  action=450 4.7.1 maintenance ongoing, try again later
        time=-07:00:00 ;  sasl_username=jim ;  action=450 4.7.1 to early for you, jim
        time=22:00:00- ;  sasl_username=jim ;  action=450 4.7.1 to late now, jim
        months=-Apr                         ;  action=450 4.7.1 see you in may
        days=!!Mon-Fri                      ;  action=greylist
        ## Usage of jump
        # The following allows a message size of 30MB for different
        # users/clients while others will only have 10MB.
        id=R001 ; action=jump(R100) ; sasl_username=^(Alice|Bob|Jane)$
        id=R002 ; action=jump(R100) ; client_address=
        id=R003 ; action=jump(R100) ; ccert_fingerprint=AA:BB:CC:DD:...
        id=R004 ; action=jump(R100) ; ccert_fingerprint=AF:BE:CD:DC:...
        id=R005 ; action=jump(R100) ; ccert_fingerprint=DD:CC:BB:DD:...
        id=R099 ; state==END-OF-MESSAGE; action=REJECT message too big (max. 10MB); size=10000000
        id=R100 ; state==END-OF-MESSAGE; action=REJECT message too big (max. 30MB); size=30000000
        ## Usage of score
        # The following rejects a mail, if the client
        # - is listed on 1 RBL and 1 RHSBL
        # - is listed in 1 RBL or 1 RHSBL and has no correct rDNS
        # - other clients without correct rDNS will be greylist-checked
        # - some whitelists are used to lower the score
        id=S01 ; score=2.6              ; action=greylisting
        id=S02 ; score=5.0              ; action=REJECT postfwd2 score too high
        id=R00 ; action=score(-1.0)     ;,,,
        id=R01 ; action=score(2.5)      ;,,
        id=R02 ; action=score(2.5)      ;,
        id=N01 ; action=score(-0.2)     ; client_name==$$helo_name
        id=N02 ; action=score(2.7)      ; client_name=^unknown$
        ## Usage of rate and size
        # The following temporary rejects requests from "unknown" clients, if they
        # 1. exceeded 30 requests per hour or
        # 2. tried to send more than 1.5mb within 10 minutes
        id=RATE01 ;  client_name==unknown ;  state==RCPT ; \
                action==rate($$client_address/30/3600/450 4.7.1 sorry, max 30 requests per hour)
        id=SIZE01 ;  client_name==unknown ;  state==END_OF_DATA ; \
                action==size($$client_address/1572864/600/450 4.7.1 sorry, max 1.5mb per 10 minutes)
        ## Macros
        # definition
        &&RBLS {,,,,; };
        &&GONOW { action=REJECT your request caused our spam detection policy to reject this message. More info at http://www.domain.local; };
        # rules
        &&GONOW ;  &&RBLS ;  client_name=^unknown$
        &&GONOW ;  &&RBLS ;  client_name=(\d+[\.-_]){4}
        &&GONOW ;  &&RBLS ;  client_name=[\.-_](adsl|dynamic|ppp|)[\.-_]
        ## Groups
        # definition
        &&RBLS { \
       ;          \
       ;             \
       ;            \
       ;           \
       ;       \
        &&RHSBLS { \
        &&DYNAMIC { \
                client_name==unknown ;                          \
                client_name~=(\d+[\.-_]){4} ;                   \
                client_name~=[\.-_](adsl|dynamic|ppp|)[\.-_] ;  \
        &&BAD_HELO { \
      ;         \
                helo_name~=^([^\.]+)$;          \
                helo_name~=\.(local|lan)$;      \
        &&MAINTENANCE { \
                date=15.01.2007 ; \
                date=15.04.2007 ; \
                date=15.07.2007 ; \
                date=15.10.2007 ; \
                time=03:00:00 - 04:00:00 ; \
        # rules
        id=COMBINED    ;  &&RBLS ;  &&DYNAMIC ;  action=REJECT dynamic client and listed on RBL
        id=MAINTENANCE ;  &&MAINTENANCE       ;  action=DEFER maintenance time - please try again later
        # now with the set() command, note that long item
        # lists don't have to be compared twice
        id=RBL01    ;  &&RBLS      ;  action=set(HIT_rbls=1)
        id=HELO01   ;  &&BAD_HELO  ;  action=set(HIT_helo=1)
        id=DYNA01   ;  &&DYNAMIC   ;  action=set(HIT_dyna=1)
        id=REJECT01 ;  HIT_rbls==1 ;  HIT_helo==1  ; action=REJECT please see for more info
        id=REJECT02 ;  HIT_rbls==1 ;  HIT_dyna==1  ; action=REJECT please see for more info
        id=REJECT03 ;  HIT_helo==1 ;  HIT_dyna==1  ; action=REJECT please see for more info
        ## combined with enhanced rbl features
        id=RBL01 ; rhsblcount=all ; rblcount=all ; &&RBLS ; &&RHSBLS ; \
        id=RBL02 ; HIT_dnsbls>=2  ; action=554 5.7.1 blocked using $$HIT_dnsbls DNSBLs [INFO: $$HIT_dnstxt]



The postfwd2 ruleset can be specified at the commandline (-r option) or be read from files (-f). The order of your arguments will be kept. You should check the parser with the -C | --showconfig switch at the command line before applying a new config. The following call:

        postfwd2 --showconfig \
                -r "id=TEST; recipient_count=100; action=WARN mail with 100+ recipients" \
                -f /etc/ \
                -r "id=DEFAULT; action=dunno";

will produce the following output:

        Rule   0: id->"TEST" action->"WARN mail with 100+ recipients"; recipient_count->"100"
        ... <content of /etc/> ...
        Rule <n>: id->"DEFAULT" action->"dunno"

Multiple items of the same type will be added to lists (see the ITEMS section for more info):

        postfwd2 --showconfig \
                -r "client_address=; client_address=; action=dunno"

will result in:

        Rule   0: id->"R-0"; action->"dunno"; client_address->","

Macros are evaluated at configuration stage, which means that

        postfwd2 --showconfig \
                -r "&&RBLS {; client_name=^unknown$; };" \
                -r "id=RBL001; &&RBLS; action=REJECT listed on spamcop and bad rdns";

will result in:

        Rule   0: id->"RBL001"; action->"REJECT listed on spamcop and bad rdns"; rbl->""; client_name->"^unknown$"

Request processing

When a policy delegation request arrives it will be compared against postfwd`s ruleset. To inspect the processing in detail you should increase verbority using use the ``-v'' or ``-vv'' switch. ``-L'' redirects log messages to stdout.

Keeping the order of the ruleset in general, items will be compared in random order, which basically means that

        id=R001; action=dunno; client_address=; sender=bob@alice.local

equals to

        id=R001; sender=bob@alice.local; client_address=; action=dunno

Lists will be evaluated in the specified order. This allows to place faster expressions at first:

        postfwd2 --nodaemon -vv -L -r "id=RBL001; rbl=localrbl.local; action=REJECT" /some/where/request.sample

produces the following

        [LOGS info]: compare rbl: "[]"  ->  "localrbl.local"
        [LOGS info]: count1 rbl:  "2"  ->  "0"
        [LOGS info]: query rbl:   localrbl.local (
        [LOGS info]: count2 rbl:  "2"  ->  "0"
        [LOGS info]: match rbl:   FALSE
        [LOGS info]: compare rbl: "[]"  ->  ""
        [LOGS info]: count1 rbl:  "2"  ->  "0"
        [LOGS info]: query rbl: (
        [LOGS info]: count2 rbl:  "2"  ->  "0"
        [LOGS info]: match rbl:   FALSE
        [LOGS info]: Action: dunno

The negation operator !!(<value>) has the highest priority and therefore will be evaluated first. Then variable substitutions are performed:

        postfwd2 --nodaemon -vv -L -r "id=TEST; action=REJECT; client_name=!!($$heloname)" /some/where/request.sample

will give

        [LOGS info]: compare client_name:     "unknown"  ->  "!!($$helo_name)"
        [LOGS info]: negate client_name:      "unknown"  ->  "$$helo_name"
        [LOGS info]: substitute client_name:  "unknown"  ->  ""
        [LOGS info]: match client_name:  TRUE
        [LOGS info]: Action: REJECT

Ruleset evaluation

A rule hits when all items (or at least one element of a list for each item) have matched. As soon as one item (or all elements of a list) fails to compare against the request attribute the parser will jump to the next rule in the postfwd2 ruleset.

If a rule matches, there are two options:

* Rule returns postfix action (dunno, reject, ...) The parser stops rule processing and returns the action to postfix. Other rules will not be evaluated.

* Rule returns postfwd2 action (jump(), note(), ...) The parser evaluates the given action and continues with the next rule (except for the jump() or quit() actions - please see the ACTIONS section for more information). Nothing will be sent to postfix.

If no rule has matched and the end of the ruleset is reached postfwd2 will return dunno without logging anything unless in verbose mode. You may place a last catch-all rule to change that behaviour:

        ... <your rules> ...
        id=DEFAULT ;  action=dunno

will log any request that passes the ruleset without having hit a prior rule.


To debug special steps of the parser the '--debug' switch takes a list of debug classes. Currently the following classes are defined:

        all devel
        verbose request config
        cache getcache setcache
        dns getdns setdns getdnspacket
        parent_cache parent_dns_cache parent_rate_cache parent_request_cache
        child_cache  child_dns_cache  child_rate_cache  child_request_cache


Integration via daemon mode

The common way to use postfwd2 is to start it as daemon, listening at a specified tcp port. postfwd2 will spawn multiple child processes which communicate with a parent cache. This is the prefered way to use postfwd2 in high volume environments. Start postfwd2 with the following parameters:

        postfwd2 -d -f /etc/ -i -p 10040 -u nobody -g nobody -S

For efficient caching you should check if you can use the options --cacheid, --cache-rdomain-only, --cache-no-sender and --cache-no-size.

Now check your syslogs (default facility ``mail'') for a line like:

        Aug  9 23:00:24 mail postfwd[5158]: postfwd2 n.nn ready for input

and use `netstat -an|grep 10040` to check for something like

        tcp  0  0*  LISTEN

If everything works, open your postfix and insert the following      = 3600                                          <--- integration
        smtpd_recipient_restrictions    = permit_mynetworks                             <--- recommended
                                          reject_unauth_destination                     <--- recommended
                                          check_policy_service inet:     <--- integration

Reload your configuration with `postfix reload` and watch your logs. In it works you should see lines like the following in your mail log:

        Aug  9 23:01:24 mail postfwd[5158]: rule=22, id=ML_POSTFIX,[], sender=owner-postfix-users@postfix.tld, recipient=someone@domain.local,, proto=ESMTP, state=RCPT, action=dunno

If you want to check for size or rcpt_count items you must integrate postfwd2 in smtp_data_restrictions or smtpd_end_of_data_restrictions. Of course you can also specify a restriction class and use it in your access tables. First create a file /etc/postfix/policy containing:

        domain1.local           postfwdcheck
        domain2.local           postfwdcheck

Then postmap that file (`postmap hash:/etc/postfix/policy`), open your and enter

        # Restriction Classes
        smtpd_restriction_classes       = postfwdcheck, <some more>...                          <--- integration
        postfwdcheck                    = check_policy_service inet:             <--- integration      = 3600                                                  <--- integration
        smtpd_recipient_restrictions    = permit_mynetworks,                                    <--- recommended
                                          reject_unauth_destination,                            <--- recommended
                                          ...                                                   <--- optional
                                          check_recipient_access hash:/etc/postfix/policy,      <--- integration
                                          ...                                                   <--- optional

Reload postfix and watch your logs.


First you have to create a ruleset (see Configuration section). Check it with

        postfwd2 -f /etc/ -C

There is an example policy request distributed with postfwd, called 'request.sample'. Simply change it to meet your requirements and use

        postfwd2 -f /etc/ <request.sample

You should get an answer like


For network tests I use netcat:

        nc 10040 <request.sample

to send a request to postfwd. If you receive nothing, make sure that postfwd2 is running and listening on the specified network settings.


Some of these proposals might not match your environment. Please check your requirements and test new options carefully!

        - use caching options
        - use the correct match operator ==, <=, >=
        - use ^ and/or $ in regular expressions
        - use item lists (faster than single rules)
        - use set() action on repeated item lists
        - use jumps and rate limits
        - use a pre-lookup rule for rbl/rhsbls with empty note() action


See for a description of how Postfix policy servers work.


postfwd2 is free software and released under BSD license, which basically means that you can do what you want as long as you keep the copyright notice:

Copyright (c) 2009, Jan Peter Kessler All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

 * Redistributions of source code must retain the above copyright
   notice, this list of conditions and the following disclaimer.
 * Redistributions in binary form must reproduce the above copyright
   notice, this list of conditions and the following disclaimer in
   the documentation and/or other materials provided with the
 * Neither the name of the authors nor the names of his contributors
   may be used to endorse or promote products derived from this
   software without specific prior written permission.



Jan Peter Kessler <info (AT) postfwd (DOT) org>. Let me know, if you have any suggestions.