Docs: add note on spf lookup type, in experimental-spec
[exim.git] / doc / doc-txt / experimental-spec.txt
1 From time to time, experimental features may be added to Exim.
2 While a feature is experimental, there will be a build-time
3 option whose name starts "EXPERIMENTAL_" that must be set in
4 order to include the feature. This file contains information
5 about experimental features, all of which are unstable and
6 liable to incompatible change.
9 Brightmail AntiSpam (BMI) suppport
10 --------------------------------------------------------------
12 Brightmail AntiSpam is a commercial package. Please see
13 for more information on
14 the product. For the sake of clarity, we'll refer to it as
15 "BMI" from now on.
18 0) BMI concept and implementation overview
20 In contrast to how spam-scanning with SpamAssassin is
21 implemented in exiscan-acl, BMI is more suited for per
22 -recipient scanning of messages. However, each messages is
23 scanned only once, but multiple "verdicts" for multiple
24 recipients can be returned from the BMI server. The exiscan
25 implementation passes the message to the BMI server just
26 before accepting it. It then adds the retrieved verdicts to
27 the messages header file in the spool. These verdicts can then
28 be queried in routers, where operation is per-recipient
29 instead of per-message. To use BMI, you need to take the
30 following steps:
32 1) Compile Exim with BMI support
33 2) Set up main BMI options (top section of Exim config file)
34 3) Set up ACL control statement (ACL section of the config
35 file)
36 4) Set up your routers to use BMI verdicts (routers section
37 of the config file).
38 5) (Optional) Set up per-recipient opt-in information.
40 These four steps are explained in more details below.
42 1) Adding support for BMI at compile time
44 To compile with BMI support, you need to link Exim against
45 the Brighmail client SDK, consisting of a library
46 ( and a header file (bmi_api.h).
47 You'll also need to explicitly set a flag in the Makefile to
48 include BMI support in the Exim binary. Both can be achieved
49 with these lines in Local/Makefile:
52 CFLAGS=-I/path/to/the/dir/with/the/includefile
53 EXTRALIBS_EXIM=-L/path/to/the/dir/with/the/library -lbmiclient_single
55 If you use other CFLAGS or EXTRALIBS_EXIM settings then
56 merge the content of these lines with them.
58 Note for BMI6.x users: You'll also have to add -lxml2_single
59 to the EXTRALIBS_EXIM line. Users of 5.5x do not need to do
60 this.
62 You should also include the location of
63 in your dynamic linker configuration
64 file (usually /etc/ and run "ldconfig"
65 afterwards, or else the produced Exim binary will not be
66 able to find the library file.
69 2) Setting up BMI support in the Exim main configuration
71 To enable BMI support in the main Exim configuration, you
72 should set the path to the main BMI configuration file with
73 the "bmi_config_file" option, like this:
75 bmi_config_file = /opt/brightmail/etc/brightmail.cfg
77 This must go into section 1 of Exim's configuration file (You
78 can put it right on top). If you omit this option, it
79 defaults to /opt/brightmail/etc/brightmail.cfg.
81 Note for BMI6.x users: This file is in XML format in V6.xx
82 and its name is /opt/brightmail/etc/bmiconfig.xml. So BMI
83 6.x users MUST set the bmi_config_file option.
86 3) Set up ACL control statement
88 To optimize performance, it makes sense only to process
89 messages coming from remote, untrusted sources with the BMI
90 server. To set up a messages for processing by the BMI
91 server, you MUST set the "bmi_run" control statement in any
92 ACL for an incoming message. You will typically do this in
93 an "accept" block in the "acl_check_rcpt" ACL. You should
94 use the "accept" block(s) that accept messages from remote
95 servers for your own domain(s). Here is an example that uses
96 the "accept" blocks from Exim's default configuration file:
99 accept domains = +local_domains
100 endpass
101 verify = recipient
102 control = bmi_run
104 accept domains = +relay_to_domains
105 endpass
106 verify = recipient
107 control = bmi_run
109 If bmi_run is not set in any ACL during reception of the
110 message, it will NOT be passed to the BMI server.
113 4) Setting up routers to use BMI verdicts
115 When a message has been run through the BMI server, one or
116 more "verdicts" are present. Different recipients can have
117 different verdicts. Each recipient is treated individually
118 during routing, so you can query the verdicts by recipient
119 at that stage. From Exim's view, a verdict can have the
120 following outcomes:
122 o deliver the message normally
123 o deliver the message to an alternate location
124 o do not deliver the message
126 To query the verdict for a recipient, the implementation
127 offers the following tools:
130 - Boolean router preconditions. These can be used in any
131 router. For a simple implementation of BMI, these may be
132 all that you need. The following preconditions are
133 available:
135 o bmi_deliver_default
137 This precondition is TRUE if the verdict for the
138 recipient is to deliver the message normally. If the
139 message has not been processed by the BMI server, this
140 variable defaults to TRUE.
142 o bmi_deliver_alternate
144 This precondition is TRUE if the verdict for the
145 recipient is to deliver the message to an alternate
146 location. You can get the location string from the
147 $bmi_alt_location expansion variable if you need it. See
148 further below. If the message has not been processed by
149 the BMI server, this variable defaults to FALSE.
151 o bmi_dont_deliver
153 This precondition is TRUE if the verdict for the
154 recipient is NOT to deliver the message to the
155 recipient. You will typically use this precondition in a
156 top-level blackhole router, like this:
158 # don't deliver messages handled by the BMI server
159 bmi_blackhole:
160 driver = redirect
161 bmi_dont_deliver
162 data = :blackhole:
164 This router should be on top of all others, so messages
165 that should not be delivered do not reach other routers
166 at all. If the message has not been processed by
167 the BMI server, this variable defaults to FALSE.
170 - A list router precondition to query if rules "fired" on
171 the message for the recipient. Its name is "bmi_rule". You
172 use it by passing it a colon-separated list of rule
173 numbers. You can use this condition to route messages that
174 matched specific rules. Here is an example:
176 # special router for BMI rule #5, #8 and #11
177 bmi_rule_redirect:
178 driver = redirect
179 bmi_rule = 5:8:11
180 data =
183 - Expansion variables. Several expansion variables are set
184 during routing. You can use them in custom router
185 conditions, for example. The following variables are
186 available:
188 o $bmi_base64_verdict
190 This variable will contain the BASE64 encoded verdict
191 for the recipient being routed. You can use it to add a
192 header to messages for tracking purposes, for example:
194 localuser:
195 driver = accept
196 check_local_user
197 headers_add = X-Brightmail-Verdict: $bmi_base64_verdict
198 transport = local_delivery
200 If there is no verdict available for the recipient being
201 routed, this variable contains the empty string.
203 o $bmi_base64_tracker_verdict
205 This variable will contain a BASE64 encoded subset of
206 the verdict information concerning the "rules" that
207 fired on the message. You can add this string to a
208 header, commonly named "X-Brightmail-Tracker". Example:
210 localuser:
211 driver = accept
212 check_local_user
213 headers_add = X-Brightmail-Tracker: $bmi_base64_tracker_verdict
214 transport = local_delivery
216 If there is no verdict available for the recipient being
217 routed, this variable contains the empty string.
219 o $bmi_alt_location
221 If the verdict is to redirect the message to an
222 alternate location, this variable will contain the
223 alternate location string returned by the BMI server. In
224 its default configuration, this is a header-like string
225 that can be added to the message with "headers_add". If
226 there is no verdict available for the recipient being
227 routed, or if the message is to be delivered normally,
228 this variable contains the empty string.
230 o $bmi_deliver
232 This is an additional integer variable that can be used
233 to query if the message should be delivered at all. You
234 should use router preconditions instead if possible.
236 $bmi_deliver is '0': the message should NOT be delivered.
237 $bmi_deliver is '1': the message should be delivered.
240 IMPORTANT NOTE: Verdict inheritance.
241 The message is passed to the BMI server during message
242 reception, using the target addresses from the RCPT TO:
243 commands in the SMTP transaction. If recipients get expanded
244 or re-written (for example by aliasing), the new address(es)
245 inherit the verdict from the original address. This means
246 that verdicts also apply to all "child" addresses generated
247 from top-level addresses that were sent to the BMI server.
250 5) Using per-recipient opt-in information (Optional)
252 The BMI server features multiple scanning "profiles" for
253 individual recipients. These are usually stored in a LDAP
254 server and are queried by the BMI server itself. However,
255 you can also pass opt-in data for each recipient from the
256 MTA to the BMI server. This is particularly useful if you
257 already look up recipient data in Exim anyway (which can
258 also be stored in a SQL database or other source). This
259 implementation enables you to pass opt-in data to the BMI
260 server in the RCPT ACL. This works by setting the
261 'bmi_optin' modifier in a block of that ACL. If should be
262 set to a list of comma-separated strings that identify the
263 features which the BMI server should use for that particular
264 recipient. Ideally, you would use the 'bmi_optin' modifier
265 in the same ACL block where you set the 'bmi_run' control
266 flag. Here is an example that will pull opt-in data for each
267 recipient from a flat file called
268 '/etc/exim/bmi_optin_data'.
270 The file format:
276 The example:
278 accept domains = +relay_to_domains
279 endpass
280 verify = recipient
281 bmi_optin = ${lookup{$local_part@$domain}lsearch{/etc/exim/bmi_optin_data}}
282 control = bmi_run
284 Of course, you can also use any other lookup method that
285 Exim supports, including LDAP, Postgres, MySQL, Oracle etc.,
286 as long as the result is a list of colon-separated opt-in
287 strings.
289 For a list of available opt-in strings, please contact your
290 Brightmail representative.
295 Sender Policy Framework (SPF) support
296 --------------------------------------------------------------
298 To learn more about SPF, visit This
299 document does not explain the SPF fundamentals, you should
300 read and understand the implications of deploying SPF on your
301 system before doing so.
303 SPF support is added via the libspf2 library. Visit
307 to obtain a copy, then compile and install it. By default,
308 this will put headers in /usr/local/include and the static
309 library in /usr/local/lib.
311 To compile Exim with SPF support, set these additional flags in
312 Local/Makefile:
315 CFLAGS=-DSPF -I/usr/local/include
316 EXTRALIBS_EXIM=-L/usr/local/lib -lspf2
318 This assumes that the libspf2 files are installed in
319 their default locations.
321 You can now run SPF checks in incoming SMTP by using the "spf"
322 ACL condition in either the MAIL, RCPT or DATA ACLs. When
323 using it in the RCPT ACL, you can make the checks dependent on
324 the RCPT address (or domain), so you can check SPF records
325 only for certain target domains. This gives you the
326 possibility to opt-out certain customers that do not want
327 their mail to be subject to SPF checking.
329 The spf condition takes a list of strings on its right-hand
330 side. These strings describe the outcome of the SPF check for
331 which the spf condition should succeed. Valid strings are:
333 o pass The SPF check passed, the sending host
334 is positively verified by SPF.
335 o fail The SPF check failed, the sending host
336 is NOT allowed to send mail for the domain
337 in the envelope-from address.
338 o softfail The SPF check failed, but the queried
339 domain can't absolutely confirm that this
340 is a forgery.
341 o none The queried domain does not publish SPF
342 records.
343 o neutral The SPF check returned a "neutral" state.
344 This means the queried domain has published
345 a SPF record, but wants to allow outside
346 servers to send mail under its domain as well.
347 This should be treated like "none".
348 o permerror This indicates a syntax error in the SPF
349 record of the queried domain. You may deny
350 messages when this occurs. (Changed in 4.83)
351 o temperror This indicates a temporary error during all
352 processing, including Exim's SPF processing.
353 You may defer messages when this occurs.
354 (Changed in 4.83)
355 o err_temp Same as permerror, deprecated in 4.83, will be
356 removed in a future release.
357 o err_perm Same as temperror, deprecated in 4.83, will be
358 removed in a future release.
360 You can prefix each string with an exclamation mark to invert
361 its meaning, for example "!fail" will match all results but
362 "fail". The string list is evaluated left-to-right, in a
363 short-circuit fashion. When a string matches the outcome of
364 the SPF check, the condition succeeds. If none of the listed
365 strings matches the outcome of the SPF check, the condition
366 fails.
368 Here is an example to fail forgery attempts from domains that
369 publish SPF records:
371 /* -----------------
372 deny message = $sender_host_address is not allowed to send mail from ${if def:sender_address_domain {$sender_address_domain}{$sender_helo_name}}. \
373 Please see${if def:sender_address_domain {mfrom}{helo}};identity=${if def:sender_address_domain {$sender_address}{$sender_helo_name}};ip=$sender_host_address
374 spf = fail
375 --------------------- */
377 You can also give special treatment to specific domains:
379 /* -----------------
380 deny message = AOL sender, but not from AOL-approved relay.
381 sender_domains =
382 spf = fail:neutral
383 --------------------- */
385 Explanation: AOL publishes SPF records, but is liberal and
386 still allows non-approved relays to send mail from
387 This will result in a "neutral" state, while mail from genuine
388 AOL servers will result in "pass". The example above takes
389 this into account and treats "neutral" like "fail", but only
390 for Please note that this violates the SPF draft.
392 When the spf condition has run, it sets up several expansion
393 variables.
395 $spf_header_comment
396 This contains a human-readable string describing the outcome
397 of the SPF check. You can add it to a custom header or use
398 it for logging purposes.
400 $spf_received
401 This contains a complete Received-SPF: header that can be
402 added to the message. Please note that according to the SPF
403 draft, this header must be added at the top of the header
404 list. Please see section 10 on how you can do this.
406 Note: in case of "Best-guess" (see below), the convention is
407 to put this string in a header called X-SPF-Guess: instead.
409 $spf_result
410 This contains the outcome of the SPF check in string form,
411 one of pass, fail, softfail, none, neutral, permerror or
412 temperror.
414 $spf_smtp_comment
415 This contains a string that can be used in a SMTP response
416 to the calling party. Useful for "fail".
418 In addition to SPF, you can also perform checks for so-called
419 "Best-guess". Strictly speaking, "Best-guess" is not standard
420 SPF, but it is supported by the same framework that enables SPF
421 capability. Refer to
422 for a description of what it means.
424 To access this feature, simply use the spf_guess condition in place
425 of the spf one. For example:
427 /* -----------------
428 deny message = $sender_host_address doesn't look trustworthy to me
429 spf_guess = fail
430 --------------------- */
432 In case you decide to reject messages based on this check, you
433 should note that although it uses the same framework, "Best-guess"
434 is NOT SPF, and therefore you should not mention SPF at all in your
435 reject message.
437 When the spf_guess condition has run, it sets up the same expansion
438 variables as when spf condition is run, described above.
440 Additionally, since Best-guess is not standardized, you may redefine
441 what "Best-guess" means to you by redefining spf_guess variable in
442 global config. For example, the following:
444 /* -----------------
445 spf_guess = v=spf1 a/16 mx/16 ptr ?all
446 --------------------- */
448 would relax host matching rules to a broader network range.
451 A lookup expansion is also available. It takes an email
452 address as the key and an IP address as the database:
454 $lookup (username@domain} spf {ip.ip.ip.ip}}
456 The lookup will return the same result strings as they can appear in
457 $spf_result (pass,fail,softfail,neutral,none,err_perm,err_temp).
458 Currently, only IPv4 addresses are supported.
462 SRS (Sender Rewriting Scheme) Support
463 --------------------------------------------------------------
465 Exiscan currently includes SRS support via Miles Wilton's
466 libsrs_alt library. The current version of the supported
467 library is 0.5.
469 In order to use SRS, you must get a copy of libsrs_alt from
473 Unpack the tarball, then refer to MTAs/README.EXIM
474 to proceed. You need to set
478 in your Local/Makefile.
481 DCC Support
482 --------------------------------------------------------------
484 *) Building exim
486 In order to build exim with DCC support add
490 to your Makefile. (Re-)build/install exim. exim -d should show
491 EXPERIMENTAL_DCC under "Support for".
494 *) Configuration
496 In the main section of add at least
497 dccifd_address = /usr/local/dcc/var/dccifd
498 or
499 dccifd_address = <ip> <port>
501 In the DATA ACL you can use the new condition
502 dcc = *
504 After that "$dcc_header" contains the X-DCC-Header.
506 Return values are:
507 fail for overall "R", "G" from dccifd
508 defer for overall "T" from dccifd
509 accept for overall "A", "S" from dccifd
511 dcc = */defer_ok works as for spamd.
513 The "$dcc_result" variable contains the overall result from DCC
514 answer. There will an X-DCC: header added to the mail.
516 Usually you'll use
517 defer !dcc = *
518 to greylist with DCC.
520 If you set, in the main section,
521 dcc_direct_add_header = true
522 then the dcc header will be added "in deep" and if the spool
523 file was already written it gets removed. This forces Exim to
524 write it again if needed. This helps to get the DCC Header
525 through to eg. SpamAssassin.
527 If you want to pass even more headers in the middle of the
528 DATA stage you can set
529 $acl_m_dcc_add_header
530 to tell the DCC routines to add more information; eg, you might set
531 this to some results from ClamAV. Be careful. Header syntax is
532 not checked and is added "as is".
534 In case you've troubles with sites sending the same queue items from several
535 hosts and fail to get through greylisting you can use
536 $acl_m_dcc_override_client_ip
538 Setting $acl_m_dcc_override_client_ip to an IP address overrides the default
539 of $sender_host_address. eg. use the following ACL in DATA stage:
541 warn set acl_m_dcc_override_client_ip = \
542 ${lookup{$sender_helo_name}nwildlsearch{/etc/mail/multipleip_sites}{$value}{}}
543 condition = ${if def:acl_m_dcc_override_client_ip}
544 log_message = dbg: acl_m_dcc_override_client_ip set to \
545 $acl_m_dcc_override_client_ip
547 Then set something like
548 # cat /etc/mail/multipleip_sites
552 Use a reasonable IP. eg. one the sending cluster acutally uses.
554 DMARC Support
555 --------------------------------------------------------------
557 DMARC combines feedback from SPF, DKIM, and header From: in order
558 to attempt to provide better indicators of the authenticity of an
559 email. This document does not explain the fundamentals, you
560 should read and understand how it works by visiting the website at
563 DMARC support is added via the libopendmarc library. Visit:
567 to obtain a copy, or find it in your favorite rpm package
568 repository. If building from source, this description assumes
569 that headers will be in /usr/local/include, and that the libraries
570 are in /usr/local/lib.
572 1. To compile Exim with DMARC support, you must first enable SPF.
573 Please read the above section on enabling the EXPERIMENTAL_SPF
574 feature. You must also have DKIM support, so you cannot set the
575 DISABLE_DKIM feature. Once both of those conditions have been met
576 you can enable DMARC in Local/Makefile:
579 LDFLAGS += -lopendmarc
580 # CFLAGS += -I/usr/local/include
581 # LDFLAGS += -L/usr/local/lib
583 The first line sets the feature to include the correct code, and
584 the second line says to link the libopendmarc libraries into the
585 exim binary. The commented out lines should be uncommented if you
586 built opendmarc from source and installed in the default location.
587 Adjust the paths if you installed them elsewhere, but you do not
588 need to uncomment them if an rpm (or you) installed them in the
589 package controlled locations (/usr/include and /usr/lib).
592 2. Use the following global settings to configure DMARC:
594 Required:
595 dmarc_tld_file Defines the location of a text file of valid
596 top level domains the opendmarc library uses
597 during domain parsing. Maintained by Mozilla,
598 the most current version can be downloaded
599 from a link at
601 Optional:
602 dmarc_history_file Defines the location of a file to log results
603 of dmarc verification on inbound emails. The
604 contents are importable by the opendmarc tools
605 which will manage the data, send out DMARC
606 reports, and expire the data. Make sure the
607 directory of this file is writable by the user
608 exim runs as.
610 dmarc_forensic_sender The email address to use when sending a
611 forensic report detailing alignment failures
612 if a sender domain's dmarc record specifies it
613 and you have configured Exim to send them.
614 Default: do-not-reply@$default_hostname
617 3. By default, the DMARC processing will run for any remote,
618 non-authenticated user. It makes sense to only verify DMARC
619 status of messages coming from remote, untrusted sources. You can
620 use standard conditions such as hosts, senders, etc, to decide that
621 DMARC verification should *not* be performed for them and disable
622 DMARC with a control setting:
624 control = dmarc_disable_verify
626 A DMARC record can also specify a "forensic address", which gives
627 exim an email address to submit reports about failed alignment.
628 Exim does not do this by default because in certain conditions it
629 results in unintended information leakage (what lists a user might
630 be subscribed to, etc). You must configure exim to submit forensic
631 reports to the owner of the domain. If the DMARC record contains a
632 forensic address and you specify the control statement below, then
633 exim will send these forensic emails. It's also advised that you
634 configure a dmarc_forensic_sender because the default sender address
635 construction might be inadequate.
637 control = dmarc_enable_forensic
639 (AGAIN: You can choose not to send these forensic reports by simply
640 not putting the dmarc_enable_forensic control line at any point in
641 your exim config. If you don't tell it to send them, it will not
642 send them.)
644 There are no options to either control. Both must appear before
645 the DATA acl.
648 4. You can now run DMARC checks in incoming SMTP by using the
649 "dmarc_status" ACL condition in the DATA ACL. You are required to
650 call the spf condition first in the ACLs, then the "dmarc_status"
651 condition. Putting this condition in the ACLs is required in order
652 for a DMARC check to actually occur. All of the variables are set
653 up before the DATA ACL, but there is no actual DMARC check that
654 occurs until a "dmarc_status" condition is encountered in the ACLs.
656 The dmarc_status condition takes a list of strings on its
657 right-hand side. These strings describe recommended action based
658 on the DMARC check. To understand what the policy recommendations
659 mean, refer to the DMARC website above. Valid strings are:
661 o accept The DMARC check passed and the library recommends
662 accepting the email.
663 o reject The DMARC check failed and the library recommends
664 rejecting the email.
665 o quarantine The DMARC check failed and the library recommends
666 keeping it for further inspection.
667 o none The DMARC check passed and the library recommends
668 no specific action, neutral.
669 o norecord No policy section in the DMARC record for this
670 sender domain.
671 o nofrom Unable to determine the domain of the sender.
672 o temperror Library error or dns error.
673 o off The DMARC check was disabled for this email.
675 You can prefix each string with an exclamation mark to invert its
676 meaning, for example "!accept" will match all results but
677 "accept". The string list is evaluated left-to-right in a
678 short-circuit fashion. When a string matches the outcome of the
679 DMARC check, the condition succeeds. If none of the listed
680 strings matches the outcome of the DMARC check, the condition
681 fails.
683 Of course, you can also use any other lookup method that Exim
684 supports, including LDAP, Postgres, MySQL, etc, as long as the
685 result is a list of colon-separated strings.
687 Several expansion variables are set before the DATA ACL is
688 processed, and you can use them in this ACL. The following
689 expansion variables are available:
691 o $dmarc_status
692 This is a one word status indicating what the DMARC library
693 thinks of the email. It is a combination of the results of
694 DMARC record lookup and the SPF/DKIM/DMARC processing results
695 (if a DMARC record was found). The actual policy declared
696 in the DMARC record is in a separate expansion variable.
698 o $dmarc_status_text
699 This is a slightly longer, human readable status.
701 o $dmarc_used_domain
702 This is the domain which DMARC used to look up the DMARC
703 policy record.
705 o $dmarc_domain_policy
706 This is the policy declared in the DMARC record. Valid values
707 are "none", "reject" and "quarantine". It is blank when there
708 is any error, including no DMARC record.
710 o $dmarc_ar_header
711 This is the entire Authentication-Results header which you can
712 add using an add_header modifier.
715 5. How to enable DMARC advanced operation:
716 By default, Exim's DMARC configuration is intended to be
717 non-intrusive and conservative. To facilitate this, Exim will not
718 create any type of logging files without explicit configuration by
719 you, the admin. Nor will Exim send out any emails/reports about
720 DMARC issues without explicit configuration by you, the admin (other
721 than typical bounce messages that may come about due to ACL
722 processing or failure delivery issues).
724 In order to log statistics suitable to be imported by the opendmarc
725 tools, you need to:
726 a. Configure the global setting dmarc_history_file.
727 b. Configure cron jobs to call the appropriate opendmarc history
728 import scripts and truncating the dmarc_history_file.
730 In order to send forensic reports, you need to:
731 a. Configure the global setting dmarc_forensic_sender.
732 b. Configure, somewhere before the DATA ACL, the control option to
733 enable sending DMARC forensic reports.
736 6. Example usage:
737 (RCPT ACL)
738 warn domains = +local_domains
739 hosts = +local_hosts
740 control = dmarc_disable_verify
742 warn !domains = +screwed_up_dmarc_records
743 control = dmarc_enable_forensic
745 warn condition = (lookup if destined to mailing list)
746 set acl_m_mailing_list = 1
748 (DATA ACL)
749 warn dmarc_status = accept : none : off
750 !authenticated = *
751 log_message = DMARC DEBUG: $dmarc_status $dmarc_used_domain
752 add_header = $dmarc_ar_header
754 warn dmarc_status = !accept
755 !authenticated = *
756 log_message = DMARC DEBUG: '$dmarc_status' for $dmarc_used_domain
758 warn dmarc_status = quarantine
759 !authenticated = *
760 set $acl_m_quarantine = 1
761 # Do something in a transport with this flag variable
763 deny condition = ${if eq{$dmarc_domain_policy}{reject}}
764 condition = ${if eq{$acl_m_mailing_list}{1}}
765 message = Messages from $dmarc_used_domain break mailing lists
767 deny dmarc_status = reject
768 !authenticated = *
769 message = Message from $dmarc_used_domain failed sender's DMARC policy, REJECT
773 Event Actions
774 --------------------------------------------------------------
776 (Renamed from TPDA, Transport post-delivery actions)
778 An arbitrary per-transport string can be expanded upon various transport events.
779 Additionally a main-section configuration option can be expanded on some
780 per-message events.
781 This feature may be used, for example, to write exim internal log information
782 (not available otherwise) into a database.
784 In order to use the feature, you must compile with
788 in your Local/Makefile
790 and define one or both of
791 - the event_action option in the transport
792 - the event_action main option
793 to be expanded when the event fires.
795 A new variable, $event_name, is set to the event type when the
796 expansion is done. The current list of events is:
798 msg:complete after main per message
799 msg:delivery after transport per recipient
800 msg:host:defer after transport per attempt
801 msg:fail:delivery after main per recipient
802 msg:fail:internal after main per recipient
803 tcp:connect before transport per connection
804 tcp:close after transport per connection
805 tls:cert before both per certificate in verification chain
806 smtp:connect after transport per connection
808 The expansion is called for all event types, and should use the $event_name
809 variable to decide when to act. The value of the variable is a colon-separated
810 list, defining a position in the tree of possible events; it may be used as
811 a list or just matched on as a whole. There will be no whitespace.
814 There is an auxilary variable, $event_data, for which the
815 content is event_dependent:
817 msg:delivery smtp confirmation mssage
818 msg:host:defer error string
819 tls:cert verification chain depth
820 smtp:connect smtp banner
822 The msg:host:defer event populates one extra variable, $event_defer_errno.
824 The following variables are likely to be useful depending on the event type:
826 router_name, transport_name
827 local_part, domain
828 host, host_address, host_port
829 tls_out_peercert
830 lookup_dnssec_authenticated, tls_out_dane
831 sending_ip_address, sending_port
832 message_exim_id, verify_mode
835 An example might look like:
837 event_action = ${if eq {msg:delivery}{$event_name} \
838 {${lookup pgsql {SELECT * FROM record_Delivery( \
839 '${quote_pgsql:$sender_address_domain}',\
840 '${quote_pgsql:${lc:$sender_address_local_part}}', \
841 '${quote_pgsql:$domain}', \
842 '${quote_pgsql:${lc:$local_part}}', \
843 '${quote_pgsql:$host_address}', \
844 '${quote_pgsql:${lc:$host}}', \
845 '${quote_pgsql:$message_exim_id}')}} \
846 } {}}
848 The string is expanded when each of the supported events occur
849 and any side-effects of the expansion will happen.
850 Note that for complex operations an ACL expansion can be used.
853 The expansion of the event_action option should normally
854 return an empty string. Should it return anything else the
855 following will be forced:
857 msg:delivery (ignored)
858 msg:host:defer (ignored)
859 msg:fail:delivery (ignored)
860 tcp:connect do not connect
861 tcp:close (ignored)
862 tls:cert refuse verification
863 smtp:connect close connection
865 No other use is made of the result string.
868 Known issues:
869 - the tls:cert event is only called for the cert chain elements
870 received over the wire, with GnuTLS. OpenSSL gives the entire
871 chain including those loaded locally.
874 Redis Lookup
875 --------------------------------------------------------------
877 Redis is open source advanced key-value data store. This document
878 does not explain the fundamentals, you should read and understand how
879 it works by visiting the website at
881 Redis lookup support is added via the hiredis library. Visit:
885 to obtain a copy, or find it in your operating systems package repository.
886 If building from source, this description assumes that headers will be in
887 /usr/local/include, and that the libraries are in /usr/local/lib.
889 1. In order to build exim with Redis lookup support add
893 to your Local/Makefile. (Re-)build/install exim. exim -d should show
894 Experimental_Redis in the line "Support for:".
897 LDFLAGS += -lhiredis
898 # CFLAGS += -I/usr/local/include
899 # LDFLAGS += -L/usr/local/lib
901 The first line sets the feature to include the correct code, and
902 the second line says to link the hiredis libraries into the
903 exim binary. The commented out lines should be uncommented if you
904 built hiredis from source and installed in the default location.
905 Adjust the paths if you installed them elsewhere, but you do not
906 need to uncomment them if an rpm (or you) installed them in the
907 package controlled locations (/usr/include and /usr/lib).
910 2. Use the following global settings to configure Redis lookup support:
912 Required:
913 redis_servers This option provides a list of Redis servers
914 and associated connection data, to be used in
915 conjunction with redis lookups. The option is
916 only available if Exim is configured with Redis
917 support.
919 For example:
921 redis_servers = - using database 10 with no password
922 redis_servers = - to make use of the default database of 0 with a password
923 redis_servers = - for default database of 0 with no password
925 3. Once you have the Redis servers defined you can then make use of the
926 experimental Redis lookup by specifying ${lookup redis{}} in a lookup query.
928 4. Example usage:
930 (Host List)
931 hostlist relay_from_ips = <\n ${lookup redis{SMEMBERS relay_from_ips}}
933 Where relay_from_ips is a Redis set which contains entries such as "" "" and so on.
934 The result set is returned as
937 ..
938 .
940 (Domain list)
941 domainlist virtual_domains = ${lookup redis {HGET $domain domain}}
943 Where $domain is a hash which includes the key 'domain' and the value '$domain'.
945 (Adding or updating an existing key)
946 set acl_c_spammer = ${if eq{${lookup redis{SPAMMER_SET}}}{OK}}
948 Where SPAMMER_SET is a macro and it is defined as
950 "SET SPAMMER <some_value>"
952 (Getting a value from Redis)
954 set acl_c_spam_host = ${lookup redis{GET...}}
957 Proxy Protocol Support
958 --------------------------------------------------------------
960 Exim now has Experimental "Proxy Protocol" support. It was built on
961 specifications from:
963 Above URL revised May 2014 to change version 2 spec:
966 The purpose of this function is so that an application load balancer,
967 such as HAProxy, can sit in front of several Exim servers and Exim
968 will log the IP that is connecting to the proxy server instead of
969 the IP of the proxy server when it connects to Exim. It resets the
970 $sender_address_host and $sender_address_port to the IP:port of the
971 connection to the proxy. It also re-queries the DNS information for
972 this new IP address so that the original sender's hostname and IP
973 get logged in the Exim logfile. There is no logging if a host passes or
974 fails Proxy Protocol negotiation, but it can easily be determined and
975 recorded in an ACL (example is below).
977 1. To compile Exim with Proxy Protocol support, put this in
978 Local/Makefile:
982 2. Global configuration settings:
984 proxy_required_hosts = HOSTLIST
986 The proxy_required_hosts option will require any IP in that hostlist
987 to use Proxy Protocol. The specification of Proxy Protocol is very
988 strict, and if proxy negotiation fails, Exim will not allow any SMTP
989 command other than QUIT. (See end of this section for an example.)
990 The option is expanded when used, so it can be a hostlist as well as
991 string of IP addresses. Since it is expanded, specifying an alternate
992 separator is supported for ease of use with IPv6 addresses.
994 To log the IP of the proxy in the incoming logline, add:
995 log_selector = +proxy
997 A default incoming logline (wrapped for appearance) will look like this:
999 2013-11-04 09:25:06 1VdNti-0001OY-1V <=
1000 [] P=esmtp S=433
1002 With the log selector enabled, an email that was proxied through a
1003 Proxy Protocol server at will look like this:
1005 2013-11-04 09:25:06 1VdNti-0001OY-1V <=
1006 [] P=esmtp PRX= S=433
1008 3. In the ACL's the following expansion variables are available.
1010 proxy_host_address The (internal) src IP of the proxy server
1011 making the connection to the Exim server.
1012 proxy_host_port The (internal) src port the proxy server is
1013 using to connect to the Exim server.
1014 proxy_target_address The dest (public) IP of the remote host to
1015 the proxy server.
1016 proxy_target_port The dest port the remote host is using to
1017 connect to the proxy server.
1018 proxy_session Boolean, yes/no, the connected host is required
1019 to use Proxy Protocol.
1021 There is no expansion for a failed proxy session, however you can detect
1022 it by checking if $proxy_session is true but $proxy_host is empty. As
1023 an example, in my connect ACL, I have:
1025 warn condition = ${if and{ {bool{$proxy_session}} \
1026 {eq{$proxy_host_address}{}} } }
1027 log_message = Failed required proxy protocol negotiation \
1028 from $sender_host_name [$sender_host_address]
1030 warn condition = ${if and{ {bool{$proxy_session}} \
1031 {!eq{$proxy_host_address}{}} } }
1032 # But don't log health probes from the proxy itself
1033 condition = ${if eq{$proxy_host_address}{$sender_host_address} \
1034 {false}{true}}
1035 log_message = Successfully proxied from $sender_host_name \
1036 [$sender_host_address] through proxy protocol \
1037 host $proxy_host_address
1039 # Possibly more clear
1040 warn logwrite = Remote Source Address: $sender_host_address:$sender_host_port
1041 logwrite = Proxy Target Address: $proxy_target_address:$proxy_target_port
1042 logwrite = Proxy Internal Address: $proxy_host_address:$proxy_host_port
1043 logwrite = Internal Server Address: $received_ip_address:$received_port
1046 4. Recommended ACL additions:
1047 - Since the real connections are all coming from your proxy, and the
1048 per host connection tracking is done before Proxy Protocol is
1049 evaluated, smtp_accept_max_per_host must be set high enough to
1050 handle all of the parallel volume you expect per inbound proxy.
1051 - With the smtp_accept_max_per_host set so high, you lose the ability
1052 to protect your server from massive numbers of inbound connections
1053 from one IP. In order to prevent your server from being DOS'd, you
1054 need to add a per connection ratelimit to your connect ACL. I
1055 suggest something like this:
1057 # Set max number of connections per host
1058 LIMIT = 5
1059 # Or do some kind of IP lookup in a flat file or database
1060 # LIMIT = ${lookup{$sender_host_address}iplsearch{/etc/exim/proxy_limits}}
1062 defer message = Too many connections from this IP right now
1063 ratelimit = LIMIT / 5s / per_conn / strict
1066 5. Runtime issues to be aware of:
1067 - The proxy has 3 seconds (hard-coded in the source code) to send the
1068 required Proxy Protocol header after it connects. If it does not,
1069 the response to any commands will be:
1070 "503 Command refused, required Proxy negotiation failed"
1071 - If the incoming connection is configured in Exim to be a Proxy
1072 Protocol host, but the proxy is not sending the header, the banner
1073 does not get sent until the timeout occurs. If the sending host
1074 sent any input (before the banner), this causes a standard Exim
1075 synchronization error (i.e. trying to pipeline before PIPELINING
1076 was advertised).
1077 - This is not advised, but is mentioned for completeness if you have
1078 a specific internal configuration that you want this: If the Exim
1079 server only has an internal IP address and no other machines in your
1080 organization will connect to it to try to send email, you may
1081 simply set the hostlist to "*", however, this will prevent local
1082 mail programs from working because that would require mail from
1083 localhost to use Proxy Protocol. Again, not advised!
1085 6. Example of a refused connection because the Proxy Protocol header was
1086 not sent from a host configured to use Proxy Protocol. In the example,
1087 the 3 second timeout occurred (when a Proxy Protocol banner should have
1088 been sent), the banner was displayed to the user, but all commands are
1089 rejected except for QUIT:
1091 # nc 25
1092, ESMTP Exim 4.82+proxy, Mon, 04 Nov 2013 10:45:59
1093 220 -0800 RFC's enforced
1094 EHLO localhost
1095 503 Command refused, required Proxy negotiation failed
1096 QUIT
1097 221 closing connection
1102 SOCKS
1103 ------------------------------------------------------------
1104 Support for proxying outbound SMTP via a Socks 5 proxy
1105 (RFC 1928) is included if Exim is compiled with
1108 If an smtp transport has a nonempty socks_proxy option
1109 defined, this is active. The option is expanded and
1110 should be a list (colon-separated by default) of
1111 proxy specifiers. Each proxy specifier is a list
1112 (space-separated by default) where the initial element
1113 is an IP address and any subsequent elements are options.
1115 Options are a string <name>=<value>.
1116 These options are currently defined:
1117 - "auth", with possible values "none" and "name".
1118 Using "name" selects username/password authentication
1119 per RFC 1929. Default is "none".
1120 - "name" sets the authentication username. Default is empty.
1121 - "pass" sets the authentication password. Default is empty.
1122 - "port" sets the tcp port number for the proxy. Default is 1080.
1123 - "tmo" sets a connection timeout in seconds for this proxy. Default is 5.
1125 Proxies from the list are tried in order until
1126 one responds. The timeout for the overall connection
1127 applies to the set of proxied attempts.
1129 If events are used, the remote IP/port during a
1130 tcp:connect event will be that of the proxy.
1135 DANE
1136 ------------------------------------------------------------
1137 DNS-based Authentication of Named Entities, as applied
1138 to SMTP over TLS, provides assurance to a client that
1139 it is actually talking to the server it wants to rather
1140 than some attacker operating a Man In The Middle (MITM)
1141 operation. The latter can terminate the TLS connection
1142 you make, and make another one to the server (so both
1143 you and the server still think you have an encrypted
1144 connection) and, if one of the "well known" set of
1145 Certificate Authorities has been suborned - something
1146 which *has* been seen already (2014), a verifiable
1147 certificate (if you're using normal root CAs, eg. the
1148 Mozilla set, as your trust anchors).
1150 What DANE does is replace the CAs with the DNS as the
1151 trust anchor. The assurance is limited to a) the possibility
1152 that the DNS has been suborned, b) mistakes made by the
1153 admins of the target server. The attack surface presented
1154 by (a) is thought to be smaller than that of the set
1155 of root CAs.
1157 It also allows the server to declare (implicitly) that
1158 connections to it should use TLS. An MITM could simply
1159 fail to pass on a server's STARTTLS.
1161 DANE scales better than having to maintain (and
1162 side-channel communicate) copies of server certificates
1163 for every possible target server. It also scales
1164 (slightly) better than having to maintain on an SMTP
1165 client a copy of the standard CAs bundle. It also
1166 means not having to pay a CA for certificates.
1168 DANE requires a server operator to do three things:
1169 1) run DNSSEC. This provides assurance to clients
1170 that DNS lookups they do for the server have not
1171 been tampered with. The domain MX record applying
1172 to this server, its A record, its TLSA record and
1173 any associated CNAME records must all be covered by
1174 DNSSEC.
1175 2) add TLSA DNS records. These say what the server
1176 certificate for a TLS connection should be.
1177 3) offer a server certificate, or certificate chain,
1178 in TLS connections which is traceable to the one
1179 defined by (one of?) the TSLA records
1181 There are no changes to Exim specific to server-side
1182 operation of DANE.
1184 The TLSA record for the server may have "certificate
1185 usage" of DANE-TA(2) or DANE-EE(3). The latter specifies
1186 the End Entity directly, i.e. the certificate involved
1187 is that of the server (and should be the sole one transmitted
1188 during the TLS handshake); this is appropriate for a
1189 single system, using a self-signed certificate.
1190 DANE-TA usage is effectively declaring a specific CA
1191 to be used; this might be a private CA or a public,
1192 well-known one. A private CA at simplest is just
1193 a self-signed certificate which is used to sign
1194 cerver certificates, but running one securely does
1195 require careful arrangement. If a private CA is used
1196 then either all clients must be primed with it, or
1197 (probably simpler) the server TLS handshake must transmit
1198 the entire certificate chain from CA to server-certificate.
1199 If a public CA is used then all clients must be primed with it
1200 (losing one advantage of DANE) - but the attack surface is
1201 reduced from all public CAs to that single CA.
1202 DANE-TA is commonly used for several services and/or
1203 servers, each having a TLSA query-domain CNAME record,
1204 all of which point to a single TLSA record.
1206 The TLSA record should have a Selector field of SPKI(1)
1207 and a Matching Type field of SHA2-512(2).
1209 At the time of writing,
1210 is useful for quickly generating TLSA records; and commands like
1212 openssl x509 -in -pubkey -noout <certificate.pem \
1213 | openssl rsa -outform der -pubin 2>/dev/null \
1214 | openssl sha512 \
1215 | awk '{print $2}'
1217 are workable for 4th-field hashes.
1219 For use with the DANE-TA model, server certificates
1220 must have a correct name (SubjectName or SubjectAltName).
1222 The use of OCSP-stapling should be considered, allowing
1223 for fast revocation of certificates (which would otherwise
1224 be limited by the DNS TTL on the TLSA records). However,
1225 this is likely to only be usable with DANE-TA. NOTE: the
1226 default of requesting OCSP for all hosts is modified iff
1227 DANE is in use, to:
1229 hosts_request_ocsp = ${if or { {= {0}{$tls_out_tlsa_usage}} \
1230 {= {4}{$tls_out_tlsa_usage}} } \
1231 {*}{}}
1233 The (new) variable $tls_out_tlsa_usage is a bitfield with
1234 numbered bits set for TLSA record usage codes.
1235 The zero above means DANE was not in use,
1236 the four means that only DANE-TA usage TLSA records were
1237 found. If the definition of hosts_request_ocsp includes the
1238 string "tls_out_tlsa_usage", they are re-expanded in time to
1239 control the OCSP request.
1241 This modification of hosts_request_ocsp is only done if
1242 it has the default value of "*". Admins who change it, and
1243 those who use hosts_require_ocsp, should consider the interaction
1244 with DANE in their OCSP settings.
1247 For client-side DANE there are two new smtp transport options,
1248 hosts_try_dane and hosts_require_dane. They do the obvious thing.
1249 [ should they be domain-based rather than host-based? ]
1251 DANE will only be usable if the target host has DNSSEC-secured
1252 MX, A and TLSA records.
1254 A TLSA lookup will be done if either of the above options match
1255 and the host-lookup succeded using dnssec.
1256 If a TLSA lookup is done and succeeds, a DANE-verified TLS connection
1257 will be required for the host.
1259 (TODO: specify when fallback happens vs. when the host is not used)
1261 If DANE is requested and useable (see above) the following transport
1262 options are ignored:
1263 hosts_require_tls
1264 tls_verify_hosts
1265 tls_try_verify_hosts
1266 tls_verify_certificates
1267 tls_crl
1268 tls_verify_cert_hostnames
1270 If DANE is not usable, whether requested or not, and CA-anchored
1271 verification evaluation is wanted, the above variables should be set
1272 appropriately.
1274 Currently dnssec_request_domains must be active (need to think about that)
1275 and dnssec_require_domains is ignored.
1277 If verification was successful using DANE then the "CV" item
1278 in the delivery log line will show as "CV=dane".
1280 There is a new variable $tls_out_dane which will have "yes" if
1281 verification succeeded using DANE and "no" otherwise (only useful
1282 in combination with EXPERIMENTAL_EVENT), and a new variable
1283 $tls_out_tlsa_usage (detailed above).
1288 ------------------------------------------------------------
1290 Internationalised mail name handling.
1291 RFCs 6530, 6533, 5890
1293 Compile with EXPERIMENTAL_INTERNATIONAL and libidn.
1295 New main config option smtputf8_advertise_hosts, default '*',
1296 a host list. If this matches the sending host and
1297 accept_8bitmime is true (the default) then the ESMTP option
1298 SMTPUTF8 will be advertised.
1300 If the sender specifies the SMTPUTF8 option on a MAIL command
1301 international handling for the message is enabled and
1302 the expansion variable $message_smtputf8 will have value TRUE.
1304 The option allow_utf8_domains is set to true for this
1305 message. All DNS lookups are converted to a-label form
1306 whatever the setting of allow_utf8_domains.
1308 Both localparts and domain are maintained as the original
1309 utf8 form internally; any matching or regex use will
1310 require appropriate care. Filenames created, eg. by
1311 the appendfile transport, will have utf8 name.
1313 Helo names sent by the smtp transport will have any utf8
1314 components expanded to a-label form.
1316 Any certificate name checks will be done using the a-label
1317 form of the name.
1319 Log lines and Received-by: header lines will aquire a "utf8"
1320 prefix on the protocol element, eg. utf8esmtp.
1322 New expansion operators:
1323 ${utf8_domain_to_alabel:str}
1324 ${utf8_domain_from_alabel:str}
1325 ${utf8_localpart_to_alabel:str}
1326 ${utf8_localpart_from_alabel:str}
1328 New "control = utf8_downconvert" ACL modifier,
1329 sets a flag requiring that addresses are converted to
1330 a-label form before smtp delivery, for use in a
1331 Message Submission Agent context. Can also be
1332 phrased as "control = utf8_downconvert/1" and is
1333 mandatory. The flag defaults to zero and can be cleared
1334 by "control = utf8_downconvert/0". The value "-1"
1335 may also be used, to use a-label for only if the
1336 destination host does not support SMTPUTF8.
1338 If mua_wrapper is set, the utf8_downconvert control
1339 defaults to -1 (convert if needed).
1342 There is no explicit support for VRFY and EXPN.
1343 Configurations supporting these should inspect
1344 $smtp_command_argument for an SMTPUTF8 argument.
1346 There is no support for LMTP on Unix sockets.
1347 Using the "lmtp" protocol option on an smtp transport,
1348 for LMTP over TCP, should work as expected.
1350 Known issues:
1351 - DSN unitext handling is not present
1352 - no provision for converting logging from or to UTF-8
1354 ----
1355 IMAP folder names
1357 New expansion operator:
1359 ${imapfolder {<string>} {<sep>} {<specials>}}
1361 The string is converted from the charset specified by the headers charset
1362 command (in a filter file) or headers_charset global option, to the
1363 modified UTF-7 encoding specified by RFC 2060, with the following
1364 exception: All occurences of <sep> (which has to be a single character)
1365 are replaced with periods ("."), and all periods and slashes that aren't
1366 <sep> and are not in the <specials> string are BASE64 encoded.
1368 The third argument can be omitted, defaulting to an empty string.
1369 The second argument can be omitted, defaulting to "/".
1371 This is the encoding used by Courier for Maildir names on disk, and followed
1372 by many other IMAP servers.
1374 Example 1: ${imapfolder {Foo/Bar}} yields "Foo.Bar".
1375 Example 2: ${imapfolder {Foo/Bar}{.}{/}} yields "Foo&AC8-Bar".
1376 Example 3: ${imapfolder {Räksmörgås}} yields "R&AOQ-ksm&APY-rg&AOU-s".
1378 Note that the source charset setting is vital, and also that characters
1379 must be representable in UTF-16.
1384 --------------------------------------------------------------
1385 End of file
1386 --------------------------------------------------------------