safer coding of router toplevel

[exim.git] / doc / doc-txt / experimental-spec.txt

From time to time, experimental features may be added to Exim.
While a feature  is experimental, there  will be a  build-time
option whose name starts  "EXPERIMENTAL_" that must be  set in
order to include the  feature. This file contains  information
about experimental  features, all  of which  are unstable and
liable to incompatible change.


Brightmail AntiSpam (BMI) support
--------------------------------------------------------------

Brightmail  AntiSpam  is  a  commercial  package.  Please  see
http://www.brightmail.com    for    more    information     on
the product. For  the sake of  clarity, we'll refer  to it  as
"BMI" from now on.


0) BMI concept and implementation overview

In  contrast  to   how  spam-scanning  with   SpamAssassin  is
implemented  in  exiscan-acl,  BMI  is  more  suited  for  per
-recipient  scanning of  messages. However,  each messages  is
scanned  only  once,  but  multiple  "verdicts"  for  multiple
recipients can be  returned from the  BMI server. The  exiscan
implementation  passes  the  message to  the  BMI  server just
before accepting it.  It then adds  the retrieved verdicts  to
the messages header file in the spool. These verdicts can then
be  queried  in  routers,  where  operation  is  per-recipient
instead  of per-message.  To use  BMI, you  need to  take the
following steps:

  1) Compile Exim with BMI support
  2) Set up main BMI options (top section of Exim config file)
  3) Set up ACL control statement (ACL section of the config
     file)
  4) Set up your routers to use BMI verdicts (routers section
     of the config file).
  5) (Optional) Set up per-recipient opt-in information.

These four steps are explained in more details below.

1) Adding support for BMI at compile time

  To compile with BMI support,  you need to link Exim  against
  the  Brightmail  client   SDK,  consisting   of  a   library
  (libbmiclient_single.so)  and  a  header  file  (bmi_api.h).
  You'll also need to explicitly set a flag in the Makefile to
  include BMI support in the Exim binary. Both can be achieved
  with  these lines in Local/Makefile:

  EXPERIMENTAL_BRIGHTMAIL=yes
  CFLAGS=-I/path/to/the/dir/with/the/includefile
  EXTRALIBS_EXIM=-L/path/to/the/dir/with/the/library -lbmiclient_single

  If  you use  other CFLAGS  or EXTRALIBS_EXIM  settings then
  merge the content of these lines with them.

  Note for BMI6.x users: You'll also have to add -lxml2_single
  to the EXTRALIBS_EXIM line. Users of 5.5x do not need to  do
  this.

  You    should     also    include     the    location     of
  libbmiclient_single.so in your dynamic linker  configuration
  file   (usually   /etc/ld.so.conf)   and   run    "ldconfig"
  afterwards, or  else the  produced Exim  binary will  not be
  able to find the library file.


2) Setting up BMI support in the Exim main configuration

  To enable BMI  support in the  main Exim configuration,  you
  should set the path to the main BMI configuration file  with
  the "bmi_config_file" option, like this:

  bmi_config_file = /opt/brightmail/etc/brightmail.cfg

  This must go into section 1 of Exim's configuration file (You
  can  put it  right on  top). If  you omit  this option,  it
  defaults to /opt/brightmail/etc/brightmail.cfg.

  Note for BMI6.x users: This  file is in XML format  in V6.xx
  and its  name is  /opt/brightmail/etc/bmiconfig.xml. So  BMI
  6.x users MUST set the bmi_config_file option.


3) Set up ACL control statement

  To  optimize performance,  it makes  sense only  to process
  messages coming from remote, untrusted sources with the  BMI
  server.  To set  up a  messages for  processing by  the BMI
  server, you MUST set the "bmi_run" control statement in  any
  ACL for an incoming message.  You will typically do this  in
  an "accept"  block in  the "acl_check_rcpt"  ACL. You should
  use the "accept" block(s)  that accept messages from  remote
  servers for your own domain(s). Here is an example that uses
  the "accept" blocks from Exim's default configuration file:


  accept  domains       = +local_domains
          endpass
          verify        = recipient
          control       = bmi_run

  accept  domains       = +relay_to_domains
          endpass
          verify        = recipient
          control       = bmi_run

  If bmi_run  is not  set in  any ACL  during reception of the
  message, it will NOT be passed to the BMI server.


4) Setting up routers to use BMI verdicts

  When a message has been  run through the BMI server,  one or
  more "verdicts" are  present. Different recipients  can have
  different verdicts. Each  recipient is treated  individually
  during routing, so you  can query the verdicts  by recipient
  at  that stage.  From Exim's  view, a  verdict can  have the
  following outcomes:

  o deliver the message normally
  o deliver the message to an alternate location
  o do not deliver the message

  To query  the verdict  for a  recipient, the  implementation
  offers the following tools:


  - Boolean router  preconditions. These  can be  used in  any
    router. For a simple  implementation of BMI, these  may be
    all  that  you  need.  The  following  preconditions   are
    available:

    o bmi_deliver_default

      This  precondition  is  TRUE  if  the  verdict  for  the
      recipient is  to deliver  the message  normally. If  the
      message has not been  processed by the BMI  server, this
      variable defaults to TRUE.

    o bmi_deliver_alternate

      This  precondition  is  TRUE  if  the  verdict  for  the
      recipient  is to  deliver the  message to  an alternate
      location.  You  can  get the  location  string  from the
      $bmi_alt_location expansion variable if you need it. See
      further below. If the message has not been processed  by
      the BMI server, this variable defaults to FALSE.

    o bmi_dont_deliver

      This  precondition  is  TRUE  if  the  verdict  for  the
      recipient  is  NOT  to   deliver  the  message  to   the
      recipient. You will typically use this precondition in a
      top-level blackhole router, like this:

        # don't deliver messages handled by the BMI server
        bmi_blackhole:
          driver = redirect
          bmi_dont_deliver
          data = :blackhole:

      This router should be on top of all others, so  messages
      that should not be delivered do not reach other  routers
      at all. If   the  message  has  not  been  processed  by
      the  BMI server, this variable defaults to FALSE.


  - A list router  precondition to query  if rules "fired"  on
    the message for the recipient. Its name is "bmi_rule". You
    use  it  by  passing it  a  colon-separated  list of  rule
    numbers. You can use this condition to route messages that
    matched specific rules. Here is an example:

      # special router for BMI rule #5, #8 and #11
      bmi_rule_redirect:
        driver = redirect
        bmi_rule = 5:8:11
        data = postmaster@mydomain.com


  - Expansion variables. Several  expansion variables are  set
    during  routing.  You  can  use  them  in  custom   router
    conditions,  for  example.  The  following  variables  are
    available:

    o $bmi_base64_verdict

      This variable  will contain  the BASE64  encoded verdict
      for the recipient being routed. You can use it to add  a
      header to messages for tracking purposes, for example:

      localuser:
        driver = accept
        check_local_user
        headers_add = X-Brightmail-Verdict: $bmi_base64_verdict
        transport = local_delivery

      If there is no verdict available for the recipient being
      routed, this variable contains the empty string.

    o $bmi_base64_tracker_verdict

      This variable  will contain  a BASE64  encoded subset of
      the  verdict  information  concerning  the  "rules" that
      fired  on the  message. You  can add  this string  to a
      header, commonly named "X-Brightmail-Tracker". Example:

      localuser:
        driver = accept
        check_local_user
        headers_add = X-Brightmail-Tracker: $bmi_base64_tracker_verdict
        transport = local_delivery

      If there is no verdict available for the recipient being
      routed, this variable contains the empty string.

    o $bmi_alt_location

      If  the  verdict  is  to  redirect  the  message  to  an
      alternate  location,  this  variable  will  contain  the
      alternate location string returned by the BMI server. In
      its default configuration, this is a header-like  string
      that can be added to the message with "headers_add".  If
      there is  no verdict  available for  the recipient being
      routed, or if the  message is to be  delivered normally,
      this variable contains the empty string.

    o $bmi_deliver

      This is an additional integer variable that can be  used
      to query if the message should be delivered at all.  You
      should use router preconditions instead if possible.

      $bmi_deliver is '0': the message should NOT be delivered.
      $bmi_deliver is '1': the message should be delivered.


  IMPORTANT NOTE: Verdict inheritance.
  The  message  is passed  to  the BMI  server  during message
  reception,  using the  target addresses  from the  RCPT TO:
  commands in the SMTP transaction. If recipients get expanded
  or re-written (for example by aliasing), the new address(es)
  inherit the  verdict from  the original  address. This means
  that verdicts also apply to all "child" addresses  generated
  from top-level addresses that were sent to the BMI server.


5) Using per-recipient opt-in information (Optional)

  The  BMI server  features multiple  scanning "profiles"  for
  individual recipients.  These are  usually stored  in a LDAP
  server and are  queried by the  BMI server itself.  However,
  you can also  pass opt-in data  for each recipient  from the
  MTA to the  BMI server. This  is particularly useful  if you
  already look  up recipient  data in  Exim anyway  (which can
  also be  stored in  a SQL  database or  other source).  This
  implementation enables you  to pass opt-in  data to the  BMI
  server  in  the  RCPT   ACL.  This  works  by   setting  the
  'bmi_optin' modifier in  a block of  that ACL. If  should be
  set to a list  of comma-separated strings that  identify the
  features which the BMI server should use for that particular
  recipient. Ideally, you  would use the  'bmi_optin' modifier
  in the same  ACL block where  you set the  'bmi_run' control
  flag. Here is an example that will pull opt-in data for each
  recipient      from       a      flat       file      called
  '/etc/exim/bmi_optin_data'.

  The file format:

    user1@mydomain.com: <OPTIN STRING1>:<OPTIN STRING2>
    user2@thatdomain.com: <OPTIN STRING3>


  The example:

    accept  domains       = +relay_to_domains
            endpass
            verify        = recipient
            bmi_optin     = ${lookup{$local_part@$domain}lsearch{/etc/exim/bmi_optin_data}}
            control       = bmi_run

  Of course,  you can  also use  any other  lookup method that
  Exim supports, including LDAP, Postgres, MySQL, Oracle etc.,
  as long as  the result is  a list of  colon-separated opt-in
  strings.

  For a list of available opt-in strings, please contact  your
  Brightmail representative.




DCC Support
--------------------------------------------------------------
Distributed Checksum Clearinghouse; http://www.rhyolite.com/dcc/

*) Building exim

In order to build exim with DCC support add

EXPERIMENTAL_DCC=yes

to your Makefile. (Re-)build/install exim. exim -d should show
EXPERIMENTAL_DCC under "Support for".


*) Configuration

In the main section of exim.cf add at least
  dccifd_address = /usr/local/dcc/var/dccifd
or
  dccifd_address = <ip> <port>

In the DATA ACL you can use the new condition
        dcc = *

After that "$dcc_header" contains the X-DCC-Header.

Return values are:
  fail    for overall "R", "G" from dccifd
  defer   for overall "T" from dccifd
  accept  for overall "A", "S" from dccifd

dcc = */defer_ok works as for spamd.

The "$dcc_result" variable contains the overall result from DCC
answer.  There will an X-DCC: header added to the mail.

Usually you'll use
  defer   !dcc = *
to greylist with DCC.

If you set, in the main section,
  dcc_direct_add_header = true
then the dcc header will be added "in deep" and if the spool
file was already written it gets removed. This forces Exim to
write it again if needed.  This helps to get the DCC Header
through to eg. SpamAssassin.

If you want to pass even more headers in the middle of the
DATA stage you can set
  $acl_m_dcc_add_header
to tell the DCC routines to add more information; eg, you might set
this to some results from ClamAV.  Be careful.  Header syntax is
not checked and is added "as is".

In case you've troubles with sites sending the same queue items from several
hosts and fail to get through greylisting you can use
$acl_m_dcc_override_client_ip

Setting $acl_m_dcc_override_client_ip to an IP address overrides the default
of $sender_host_address. eg. use the following ACL in DATA stage:

  warn    set acl_m_dcc_override_client_ip = \
    ${lookup{$sender_helo_name}nwildlsearch{/etc/mail/multipleip_sites}{$value}{}}
          condition = ${if def:acl_m_dcc_override_client_ip}
          log_message = dbg: acl_m_dcc_override_client_ip set to \
                        $acl_m_dcc_override_client_ip

Then set something like
# cat /etc/mail/multipleip_sites
mout-xforward.gmx.net           82.165.159.12
mout.gmx.net                    212.227.15.16

Use a reasonable IP. eg. one the sending cluster actually uses.



DSN extra information
---------------------
If compiled with EXPERIMENTAL_DSN_INFO extra information will be added
to DSN fail messages ("bounces"), when available.  The intent is to aid
tracing of specific failing messages, when presented with a "bounce"
complaint and needing to search logs.


The remote MTA IP address, with port number if nonstandard.
Example:
  Remote-MTA: X-ip; [127.0.0.1]:587
Rationale:
  Several addresses may correspond to the (already available)
  dns name for the remote MTA.

The remote MTA connect-time greeting.
Example:
  X-Remote-MTA-smtp-greeting: X-str; 220 the.local.host.name ESMTP Exim x.yz Tue, 2 Mar 1999 09:44:33 +0000
Rationale:
  This string sometimes presents the remote MTA's idea of its
  own name, and sometimes identifies the MTA software.

The remote MTA response to HELO or EHLO.
Example:
  X-Remote-MTA-helo-response: X-str; 250-the.local.host.name Hello localhost [127.0.0.1]
Limitations:
  Only the first line of a multiline response is recorded.
Rationale:
  This string sometimes presents the remote MTA's view of
  the peer IP connecting to it.

The reporting MTA detailed diagnostic.
Example:
  X-Exim-Diagnostic: X-str; SMTP error from remote mail server after RCPT TO:<d3@myhost.test.ex>: 550 hard error
Rationale:
  This string sometimes give extra information over the
  existing (already available) Diagnostic-Code field.


Note that non-RFC-documented field names and data types are used.


Queuefile transport
-------------------
Queuefile is a pseudo transport which does not perform final delivery.
It simply copies the exim spool files out of the spool directory into
an external directory retaining the exim spool format.

The spool files can then be processed by external processes and then
requeued into exim spool directories for final delivery.
However, note carefully the warnings in the main documentation on
qpool file formats.

The motivation/inspiration for the transport is to allow external
processes to access email queued by exim and have access to all the
information which would not be available if the messages were delivered
to the process in the standard email formats.

The mailscanner package is one of the processes that can take advantage
of this transport to filter email.

The transport can be used in the same way as the other existing transports,
i.e by configuring a router to route mail to a transport configured with
the queuefile driver.

The transport only takes one option:

* directory - This is used to specify the directory messages should be
copied to.  Expanded.

The generic transport options (body_only, current_directory, disable_logging,
debug_print, delivery_date_add, envelope_to_add, event_action, group,
headers_add, headers_only, headers_remove, headers_rewrite, home_directory,
initgroups, max_parallel, message_size_limit, rcpt_include_affixes,
retry_use_local_part, return_path, return_path_add, shadow_condition,
shadow_transport, transport_filter, transport_filter_timeout, user) are
ignored.

Sample configuration:

(Router)

scan:
   driver = accept
   transport = scan

(Transport)

scan:
  driver = queuefile
  directory = /var/spool/baruwa-scanner/input


In order to build exim with Queuefile transport support add or uncomment

EXPERIMENTAL_QUEUEFILE=yes

to your Local/Makefile. (Re-)build/install exim. exim -d should show
Experimental_QUEUEFILE in the line "Support for:".


ARC support
-----------
Specification: https://tools.ietf.org/html/draft-ietf-dmarc-arc-protocol-11
Note that this is not an RFC yet, so may change.

[RFC 8617 was published 2019/06.  Draft 11 was 2018/01.  A review of the
changes has not yet been done]

ARC is intended to support the utility of SPF and DKIM in the presence of
intermediaries in the transmission path - forwarders and mailinglists -
by establishing a cryptographically-signed chain in headers.

Normally one would only bother doing ARC-signing when functioning as
an intermediary.  One might do verify for local destinations.

ARC uses the notion of a "ADministrative Management Domain" (ADMD).
Described in RFC 5598 (section 2.3), this is essentially a set of
mail-handling systems that mail transits that are all under the control
of one organisation.  A label should be chosen to identify the ADMD.
Messages should be ARC-verified on entry to the ADMD, and ARC-signed on exit
from it.


Building with ARC Support
--
Enable using EXPERIMENTAL_ARC=yes in your Local/Makefile.
You must also have DKIM present (not disabled), and you very likely
want to have SPF enabled.


Verification
--
An ACL condition is provided to perform the "verifier actions" detailed
in section 6 of the above specification.  It may be called from the DATA ACL
and succeeds if the result matches any of a given list.
It also records the highest ARC instance number (the chain size)
and verification result for later use in creating an Authentication-Results:
standard header.

  verify = arc/<acceptable_list>   none:fail:pass

  add_header = :at_start:${authresults {<admd-identifier>}}

        Note that it would be wise to strip incoming messages of A-R headers
        that claim to be from our own <admd-identifier>.  Eg:

  remove_header = \N^(?i)Authentication-Results\s*::\s*example.org;\N

There are four new variables:

  $arc_state            One of pass, fail, none
  $arc_state_reason     (if fail, why)
  $arc_domains          colon-sep list of ARC chain domains, in chain order.
                        problematic elements may have empty list elements
  $arc_oldest_pass      lowest passing instance number of chain

Example:
  logwrite = oldest-p-ams: <${reduce {$lh_ARC-Authentication-Results:} \
                                {} \
                                {${if = {$arc_oldest_pass} \
                                        {${extract {i}{${extract {1}{;}{$item}}}}} \
                                        {$item} {$value}}} \
                            }>

Receive log lines for an ARC pass will be tagged "ARC".


Signing
--
arc_sign = <admd-identifier> : <selector> : <privkey> [ : <options> ]
An option on the smtp transport, which constructs and prepends to the message
an ARC set of headers.  The textually-first Authentication-Results: header
is used as a basis (you must have added one on entry to the ADMD).
Expanded as a whole; if unset, empty or forced-failure then no signing is done.
If it is set, all of the first three elements must be non-empty.

The fourth element is optional, and if present consists of a comma-separated list
of options.  The options implemented are

  timestamps            Add a t= tag to the generated AMS and AS headers, with the
                        current time.
  expire[=<val>]        Add an x= tag to the generated AMS header, with an expiry time.
                        If the value <val> is an plain number it is used unchanged.
                        If it starts with a '+' then the following number is added
                        to the current time, as an offset in seconds.
                        If a value is not given it defaults to a one month offset.

[As of writing, gmail insist that a t= tag on the AS is mandatory]

Caveats:
 * There must be an Authentication-Results header, presumably added by an ACL
   while receiving the message, for the same ADMD, for arc_sign to succeed.
   This requires careful coordination between inbound and outbound logic.

   Only one A-R header is taken account of.  This is a limitation versus
   the ARC spec (which says that all A-R headers from within the ADMD must
   be used).

 * If passing a message to another system, such as a mailing-list manager
   (MLM), between receipt and sending, be wary of manipulations to headers made
   by the MLM.
   + For instance, Mailman with REMOVE_DKIM_HEADERS==3 might improve
     deliverability in a pre-ARC world, but that option also renames the
     Authentication-Results header, which breaks signing.

 * Even if you use multiple DKIM keys for different domains, the ARC concept
   should try to stick to one ADMD, so pick a primary domain and use that for
   AR headers and outbound signing.

Signing is not compatible with cutthrough delivery; any (before expansion)
value set for the option will result in cutthrough delivery not being
used via the transport in question.



Dovecot authenticator via inet socket
--------------------------------------------------------------
If Dovecot is configured similar to :-

service auth {
...
#SASL
  inet_listener {
    name = exim
    port = 12345
  }
...
}

then an Exim authenticator can be configured :-

  dovecot-plain:
    driver =            dovecot
    public_name =       PLAIN
    server_socket =     dovecot_server_name 12345
    server_tls =        true
    server_set_id =     $auth1

If the server_socket does not start with a / it is taken as a hostname (or IP);
and a whitespace-separated port number must be given.




Logging protocol unusual states
---------------------------------------------------------------
An extra log_selector, "protocol_detail" has been added in the default build.
The name may change in future, hence the Experimental status.

Currrently the only effect is to enable logging, under TLS,
of a TCP RST received directly after a QUIT (in server mode).

Outlook is consistently doing this; not waiting for the SMTP response
to its QUIT, not properly closing the TLS session and not properly closing
the TCP connection.  Previously this resulted is an error from SSL_write
being logged.



XCLIENT proxy support
---------------------------------------------------------------
Per https://www.postfix.org/XCLIENT_README.html

XCLIENT is an ESMTP extension supporting an inbound proxy.
The only client immplementation known is in Nginx
(https://nginx.org/en/docs/mail/ngx_mail_proxy_module.html)

If compiled with EXPERIMENTAL_XCLIENT=yes :-

As a server, Exim will advertise XCLIENT support (conditional on a new option
"hosts_xclient") and service XCLIENT commands with parameters
  ADDR
  NAME
  PORT
  LOGIN
  DESTADDR
  DESTPORT
A fresh HELO/EHLO is required after a succesful XCLIENT, and the usual
values are derived from that (making the HELO and PROTO paramemters redundant).

An XCLIENT command must give both ADDR and PORT parameters if no previous
XCLIENT has succeeded in the SMTP session.

After a success:
  $proxy_session variable becomes "yes"
  $proxy_local_address, $proxy_local_port have the proxy "inside" values
  $proxy_external_address, $proxy_external_port have the proxy "outside" values
  $sender_host_address, $sender_host_port have the remot client values




--------------------------------------------------------------
End of file
--------------------------------------------------------------