chiark - git - mdw - exim-config/blob - spam.m4

   1 ### -*-m4-*-
   2 ###
   3 ### Spam filtering for distorted.org.uk Exim configuration
   4 ###
   5 ### (c) 2012 Mark Wooding
   6 ###
   7
   8 ###----- Licensing notice ---------------------------------------------------
   9 ###
  10 ### This program is free software; you can redistribute it and/or modify
  11 ### it under the terms of the GNU General Public License as published by
  12 ### the Free Software Foundation; either version 2 of the License, or
  13 ### (at your option) any later version.
  14 ###
  15 ### This program is distributed in the hope that it will be useful,
  16 ### but WITHOUT ANY WARRANTY; without even the implied warranty of
  17 ### MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  18 ### GNU General Public License for more details.
  19 ###
  20 ### You should have received a copy of the GNU General Public License
  21 ### along with this program; if not, write to the Free Software Foundation,
  22 ### Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  23
  24 DIVERT(null)
  25 ###--------------------------------------------------------------------------
  26 ### Spam filtering.
  27
  28 SECTION(global, policy)m4_dnl
  29 spamd_address = CONF_spamd_address CONF_spamd_port
  30
  31 SECTION(routers, allspam)m4_dnl
  32 ## If we're verifying an address and the recipient has a `~/.mail/spam-limit'
  33 ## file, then look up the recipient and sender addresses to find a plausible
  34 ## limit and insert it into the `address_data' where the RCPT ACL can find
  35 ## it.  This router always declines, so it doesn't affect the overall outcome
  36 ## of the verification.
  37 fetch_spam_limit:
  38         driver = redirect
  39         data = :unknown:
  40         verify_only = true
  41         local_part_suffix = CONF_user_suffix_list
  42         local_part_suffix_optional = true
  43         check_local_user
  44         address_data = \
  45                 ${if def:address_data {$address_data}{}} \
  46                 ${if exists {CONF_userconf_dir/spam-limit} \
  47                      {${lookup {$local_part_prefix\
  48                                         $local_part\
  49                                         $local_part_suffix\
  50                                         @$domain/\
  51                                         $sender_address} \
  52                                nwildlsearch {CONF_userconf_dir/spam-limit} \
  53                                {${if match{$value}{\N^[0-9]+$\N} \
  54                                      {spam_limit=$value} \
  55                                      {}}} \
  56                                {}}} \
  57                      {}} \
  58                 ${if exists {CONF_userconf_dir/spam-limit.userv} \
  59                      {${run {timeout 5s -- \
  60                                 userv $local_part exim-spam-limit \
  61                                         $sender_address \
  62                                         $local_part_prefix \
  63                                         $local_part \
  64                                         $local_part_suffix \
  65                                         @$domain} \
  66                             {${if match{$value}{\N^[0-9]+$\N} \
  67                                   {spam_limit=$value} \
  68                                   {}}} \
  69                             {}}} \
  70                      {}}
  71
  72 SECTION(acl, rcpt-hooks)m4_dnl
  73         ## Do per-recipient spam-filter processing.
  74         require  acl = rcpt_spam
  75
  76 SECTION(acl, misc)m4_dnl
  77 skip_spam_check:
  78
  79         ## If the client is trusted, or this is a new submission, don't
  80         ## bother with any of this.  We will have verified the sender
  81         ## fairly aggressively before granting this level of trust.
  82         accept   hosts = +trusted
  83         accept   condition = ${if eq{$acl_c_mode}{submission}}
  84
  85         ## Otherwise we should check.
  86         deny
  87
  88 rcpt_spam:
  89
  90         ## See if we should do this check.
  91         accept   acl = skip_spam_check
  92
  93         ## Always accept mail to `postmaster'.  Currently this is not
  94         ## negotiable; maybe a tweak can be added to `domains.conf' if
  95         ## necessary.
  96         accept   local_parts = postmaster
  97
  98         ## Collect the user's spam threshold from the `address_data'
  99         ## variable, where it was left by the `fetch_spam_limit' router
 100         ## during recipient verification.  (This just saves duplicating this
 101         ## enormous expression.)
 102         warn     set acl_m_this_spam_limit = \
 103                         ${sg {${extract {spam_limit} \
 104                                         {${if def:address_data \
 105                                               {$address_data}{}}} \
 106                                         {$value}{nil}}} \
 107                              {^(|.*\\D.*)\$}{CONF_spam_max}}
 108
 109         ## If there's a spam limit already established, and it's different
 110         ## from this user's limit, then the sender will have to try this user
 111         ## again later.
 112         defer   !hosts = +trusted
 113                  message = "You'd better try this one later"
 114                  condition = ${if def:acl_m_spam_limit {true}{false}}
 115                  condition = ${if ={$acl_m_spam_limit} \
 116                                    {$acl_m_this_spam_limit} \
 117                                   {false}{true}}
 118
 119         ## There's no limit set yet, or the user's limit is the same as the
 120         ## existing one, or the client's local and we're not checking for
 121         ## spam anyway.  Whichever way, it's safe to set it now.
 122         warn     set acl_m_spam_limit = $acl_m_this_spam_limit
 123
 124         ## All done.
 125         accept
 126
 127 SECTION(acl, data-spam)m4_dnl
 128         ## Do spam checking.
 129         require  acl = data_spam
 130
 131 SECTION(acl, misc)m4_dnl
 132 data_spam:
 133
 134         ## See if we should do this check.
 135         accept   acl = skip_spam_check
 136
 137         ## Check header validity.
 138         require  verify = header_syntax
 139
 140         ## Check the message for spam, comparing to the configured limit.
 141         deny     spam = exim:true
 142                  message = Tinned meat product detected ($spam_score)
 143                  condition = ${if >{$spam_score_int}{$acl_m_spam_limit} \
 144                                   {true}{false}}
 145
 146         ## Insert headers from the spam check now that we've decided to
 147         ## accept the message.
 148         warn
 149
 150                  ## Convert the limit (currently 10x fixed point) into a
 151                  ## decimal for presentation.
 152                  set acl_m_spam_limit_presentation = \
 153                         ${sg{$acl_m_spam_limit}{\N(\d)$\N}{.\$1}}
 154
 155                  ## Convert the report into something less obnoxious.  Plain
 156                  ## old SpamAssassin has an `X-Spam-Status' header which
 157                  ## lists the matched rules and provides some other basic
 158                  ## information.  Try to extract something similar from the
 159                  ## report.
 160                  ##
 161                  ## This is rather fiddly.
 162
 163                  ## Firstly, escape angle brackets, because we'll be using
 164                  ## them for our own purposes.
 165                  set acl_m_spam_tests = ${sg{$spam_report}{([!<>])}{!\$1}}
 166
 167                  ## Trim off the blurb paragraph and the preview.  The rest
 168                  ## should be fairly well behaved.  Wrap double angle-
 169                  ## brackets around the remainder; these can't appear in the
 170                  ## body because we escaped them all earlier.
 171                  set acl_m_spam_tests = \
 172                         ${sg{$acl_m_spam_tests} \
 173                             {\N^(?s).*\n Content analysis details:(.*)$\N} \
 174                             {<<\$1>>}}
 175
 176                  ## Extract the information about the matching rules and
 177                  ## their scores.  Leave `<<...>>' around everything else.
 178                  set acl_m_spam_tests = \
 179                         ${sg{$acl_m_spam_tests} \
 180                             {\N(?s)\n\s*(-?[\d.]+)\s+([-\w]+)\s\N} \
 181                             {>>\$2:\$1,<<}}
 182
 183                  ## Strip everything still in `<<...>>' pairs, including any
 184                  ## escaped characters inside.
 185                  set acl_m_spam_tests = \
 186                         ${sg{$acl_m_spam_tests}{\N(?s)<<([^!>]+|!.)*>>\N}{}}
 187
 188                  ## Trim off a trailing comma.
 189                  set acl_m_spam_tests = ${sg{$acl_m_spam_tests}{,\s*\$}{}}
 190
 191                  ## Undo the escaping.
 192                  set acl_m_spam_tests = ${sg{$acl_m_spam_tests}{!(.)}{\$1}}
 193
 194                  ## Insert the headers.
 195                  add_header = X-SpamAssassin-Score: \
 196                         $spam_score/$acl_m_spam_limit_presentation \
 197                         ($spam_bar)
 198                  add_header = X-SpamAssassin-Status: \
 199                         score=$spam_score, \
 200                         limit=$acl_m_spam_limit_presentation, \n\t\
 201                         tests=$acl_m_spam_tests
 202
 203         ## We're good.
 204         accept
 205
 206 DIVERT(null)
 207 ###----- That's all, folks --------------------------------------------------