WebSVN - Diff - Rev 67 and 73 - /tags/4.2.1-14/README


COPYRIGHTS
----------
SA-Exim was written by Marc MERLIN <marc_soft@merlins.org>
You can find the latest version here: 
    http://sa-exim.sf.net/
or here:
    http://marc.merlins.org/linux/exim/sa.html

greylisting was written by and is copyright Mark Lawrence <nomad@null.net>


INSTALL
-------
See the file named INSTALL for installations instructions (either compiled
in exim, or as a stand-alone shared library)

If you got sa-exim prepackaged (like on debian), you have to make sure that
your exim supports a dynamically loadable local_scan (which is true on debian
and probably on other distros too if they shipped sa-exim as a package), and
that your exim4.conf file contains the following:
local_scan_path = /usr/lib/exim4/local_scan/sa-exim.so
If you are using the split configuration file on debian with the sa-exim deb
package, you'll be fine. If you're using the monolithic file, you are on your
own until/unless the sa-exim packages try to do an in place edit (i.e. you have
to add the above configuration line yourself)


UPGRADING
---------
Deleting greylisting tuplets pre-4.2.1:
If you are installing this package yourself, and ever installed the old
greylistclean.cron which contained the complicated shell commands to clean
old tuplets, you should stop using those commands and upgrade to greylistclean.
Upgrading Greylisting.pm should also create safer tuplets without whitespace,
but it's better to get rid of the old shell cron jobs either way


PRIVACY WARNING
---------------
SA-Exim can add a header with the list of recipients in an Email (including
Bcced folks).
X-SA-Exim-Rcpt-To is used to allow you to see who a spam went to easily (i.e.
without scanning the exim logs), and to write SpamAssassin rules on the envelope
To (like adding a score if there were too many recipients or a recipient who you
know only receives spam)
X-SA-Exim-Rcpt-To is not added anymore by default, you need to enable it by
setting SAmaxrcptlistlength to a value up to 8000, but if you do add it,
you should consider removing it in exim's system_filter or in a transport. 
If SARewriteBody is true you should also consider setting
SAaddSAEheaderBeforeSA to false (see the config) as all the recipients
will be visible in the attached spam, note that this disables the
ability to write SpamAssassin rules based on X-SA-Exim-Rcpt-From/To.
In real life, who a spam was sent to isn't really a problem, but it could be if
a private message is mis-categorized as spam
Note however that if you disable X-SA-Exim-Rcpt-To by setting 
SAmaxrcptlistlength to 0, you will not be able to use greylisting, which
depends on this header (however you'd still be welcome to remove the header in
system_filter)


CONFIGURATION
-------------
You should read sa-exim.conf, all the options there should be well
documented.

Note that the code will not act on any mail before it is flagged as SPAM by SA.

Having SA flag the mail however doesn't mean the code rejects it or throws
the alleged spam away, you control what you want to do depending on the score.
The only restriction is that things happen in this order (for increasing SA
scores)

    - Save in SAnotspamsave if enabled
    - Save in SAspamacceptsave if enabled
    - Temporarily reject and optionally save if enabled
    - Permanently reject and optionally save if enabled
    - Accept, drop the mail, and optionally save if enabled
    - Teergrube (i.e. stall) the sender to waste his resources (and yours)

Note that you cannot set a teergrube threshold of 12, and a permreject
threshold of 20 (not that it would make much sense anyway).
Threshold scores should decrease as you apply the highest to the lowest penalty
(i.e. the rules are run in this order: teergrube, devnull, permreject,
tempreject)

Now, as of SA-Exim 4.2, things get slightly more complicated as scores are 
actually full exim conditions, and therefore you could have:
SAteergrube: ${if and { {!eq {$sender_host_address}{127.0.0.1}} {!eq {$sender_host_address}{127.0.0.2}} } {25}{1048576}}
This means that if your condition succeeds, the teergrube score is set to 25,
and if the condition fails, the teergrube score is set to 2^20, which for all
intents and purposes, disables teergrubing.
Regardless of what your scores end up being after the conditions are evaluated,
sa-exim still tests them in this order: teergrube, devnull, permreject,
tempreject)




CONFIGURING SPAMASSASSIN
------------------------
A good example of spamassassin configuration would be:

    report_safe            0
    use_terse_report       1	# for SA < 3.x

This will put a non-verbose SPAM-report in the headers, but leave the
message itself intact for easy analyzing and for easy feeding to
sa-learn when mis-flagged as spam or ham. The only way to see the
message is spam, is by looking in the headers. 

If you have an older version of SpamAssassin (<= 2.50), you'd probably
want to add 'report_header 1' to that list. But this is default and
un-needed in new versions of SA)

If you set 'report_safe' to a true value, you might also want to set
use_terse_report to a false value, in case you'll get the long header
which might be friendlier to your users.

For SA before 3.x, add 'always_add_report 1' to always have a spamcheck report
put in the message. This might be useful to test rules.
For SA 3.x onward, the syntax you'd want, is:
add_header                      all Report _REPORT_

Since SA is usually configured to pass messages on that are beyond the SA 
spam threshold, it can make sense to rewrite the subject line. 
To achieve this, you would use this for SA 2.x:
    rewrite_subject        1
    subject_tag            SPAM: _HITS_:

For SA 3.x, the syntax is:
    rewrite_header Subject SPAM: _HITS_:
    

If you are using SA 2.50 or better, by default, you should probably set:
    report_safe            0

Now, if you are willing to take a small speed and I/O hit, you can have
sa-exim read the body back from SA, and replace the original mail with
the new body.

You would use this if you want to set SA's report_safe to 1 or 2 (in
which case you also have to set SARewriteBody: 1 in SA-Exim's config)

Note that if you do so, unfortunately archived messages will have the
body modified by SA. This is not very trivial to fix, so if you archive
anything, you may not want to use SARewriteBody


Important:

You want to run spamd as such (you have to create the spamd user
yourself):

/usr/sbin/spamd -d -u spamd -H /var/spool/spamassassin/

It may not work if you run spamd with -c (debian default), 
(you shouldn't run spamassassin as root for this purpose anyway (there
is no reason to, so why take the risk)

You can edit this in /etc/default/spamassassin (debian) and probably
/etc/sysconfig/spamassassin (redhat)

With SA 3.x is better, the updated syntax would look like this:
/usr/sbin/spamd --max-children 50 --daemonize --username=spamd --nouser-config --helper-home-dir=/var/spool/spamassassin/



CONFIGURING EXIM4.CONF
----------------------
This code works without anything in the exim conf, but you probably want to use
some knobs to disable scanning for some users (like setting X-SA-Do-Not-Rej
or X-SA-Do-Not-Run in the rcpt ACL and removing those headers in the right 
places)

See http://marc.merlins.org/linux/exim/#conf and more specifically
http://marc.merlins.org/linux/exim/exim4-conf/exim4.conf

Note that obviously if you set those headers, spammers can set them too, so
if you are concerned about this, you can either change the header name, or set 
it to something else than 'Yes' and check for that value in sa-exim.conf
(or as a 3rd option, you can use exim ACL variables to pass values to SA-Exim
without generating headers; see the section contributed by Chirik, lower in
this file)



EXIM4 INTEGRATION / NOT SCANNING YOUR OWN MAILS
-----------------------------------------------
For a very complete exim4 config, including settings for SA, you should
look at sa-exim.conf and play with:

SAEximRunCond: ${if and{ \
                            {def:sender_host_address} \
                            {!eq {$sender_host_address}{127.0.0.1}} \
                            {!eq {$h_X-SA-Do-Not-Run:}{Yes}} \
                        } \
                    {1}{0} \
                }

PLEASE NOTE: This conditional statement must be on one line. SA-Exim's
configfile parser does not support \-lineconitunation!! 

You may also want to look at my exim4.conf config if you haven't done so yet:
http://marc.merlins.org/linux/exim/#conf

The check_rcpt ACL has:
  warn     message       = X-SA-Do-Not-Rej: Yes
           local_parts   = +nosarej:postmaster:abuse

  warn     message       = X-SA-Do-Not-Run: Yes
           hosts         = +relay_from_hosts

  warn     message       = X-SA-Do-Not-Run: Yes
           authenticated = *

Then, you'll want to strip SA headers for messages that aren't local
This means you should strip them at least in the remote_smtp transport
with this configuration snippet:

  # This is generally set on messages originating from local users and it tells
  # SA-Exim not to scan the message or that the message was scanned.
  # Let's remove these headers if the message is sent remotely
  headers_remove = "X-SA-Do-Not-Run:X-SA-Exim-Scanned:X-SA-Exim-Mail-From:X-SA-Exim-Rcpt-To:X-SA-Exim-Connect-IP"


You can also use another option, which can't be spoofed by a spammer, but
won't show you why a mail didn't get scanned if it was sent to multiple
people (which is why I personally prefer the above, even if it's spoofable)

Contributed by Chirik <chirik@castlefur.com>:
----------------------------------------------------------------------------
I have the following:

SAEximRunCond: ${if !eq {$acl_m0}{do-not-scan} {1} {0}}
SAEximRejCond: ${if !eq {$acl_m0}{do-not-reject} {1} {0}}

Then, in my recipient ACL, I have:

  ##### Checks for postmaster or abuse - we'll scan, still, but not reject
  ##### Don't reject for certain users
  warn     local_parts   = postmaster : abuse
           set acl_m0    = do-not-reject

  ##### Check for situations we don't even scan (local mail)
  ##### Don't scan if hosts we relay for (probably dumb MUAs),
  warn     hosts         = +relay_from_hosts:127.0.0.1/8
           set acl_m0    = do-not-scan

  ##### Don't scan non-smtp connections (empty host list)
  warn     hosts         = :
           set acl_m0    = do-not-scan

  ##### Don't scan if authenticated
  warn     authenticated = *
           set acl_m0    = do-not-scan
----------------------------------------------------------------------------



TEERGRUBING: SAteergrube
------------------------
The idea is for mail that you know for sure is spam (I use a threshold of 25),
you can stall the spammer for as long as possible by sending a continuation
line every 10 seconds:
451- wait for more output
451- wait for more output
451- wait for more output
(...)

You can go there for details:
http://www.iks-jena.de/mitarb/lutz/usenet/teergrube.en.html

What should you know?
1) This is obviously going to use up some of your resources
2) You should not teergrube SMTP servers that relay mail for you, be
   courteous (set a condition in SAteergrube like in the example
   provided). Besides they are real mail relays, so they will diligently
   try to send you the spam over and over for days)
   (note that you should probably not teergrube mailling lists you subscribed
   to either, or you risk getting unsubscribed)
   See a sample in sa-exim.conf for example syntax.
3) Because of limitations in the current exim code, teergrubing will not work
   over TLS.
   This shouldn't be a problem since real spammers should not be using TLS,
   and you shouldn't teergrube relays that do TLS with you.
   If you do teergrube a TLS connection, it will break the connection and you
   will see this in your logs:
18640m-0000Vb-00 SSL_write error 5
TLS error (SSL_write): error:1408F10B:SSL routines:SSL3_GET_RECORD:wrong version number
   This is not ideal, but in real life, that's ok.



GREYLISTING
-----------
See README.greylisting



READING ARCHIVED SPAMS
----------------------
Spams are optionally saved in individual files in a 'new' subdirectory
of some place like /var/spool/sa-exim/SAteergrube.

There are two ways to read them:
1) cat new/*  > /tmp/mailbox, and use  the resulting file as  a standard
   mbox file with any mail client (if SAPrependArchiveWithFrom is true)
2) Use a maildir capable mail client, like mutt, and run something like
   'mutt -f /var/spool/sa-exim/SAteergrube'. This will read the messages in
   place, since what sa-exim creates looks like a valid Maildir spool.

If you configured SA-Exim to set X-SA-Exim-Rcpt-To, you can even resend
archived refused messages to the users they were meant for

Note that sa-exim runs with the same uid/gid than the exim daemon (something
like mail, exim, or Debian-Exim), so /var/spool/sa-exim/SAteergrube must exist
and be writeable by exim.
SA-Exim will then create (sub-)directories with the permissions 0770 as
needed (those permissions aren't a configuration option, but you can change
them after the fact or pre-create the directories with the permissions of your
choice)
Files are created with 0664 permissions so that anyone who has directory access
can read (and maybe write) the files. 
If you chgrp the parent 'new' directory to a group of your choice, and give it
permissions 2770 or 2775, the files will be created with that group instead of
the default exim group



LOG AND SMTP OUTPUT
-------------------
As of SA-Exim 3.0, SMTP output does not contain the spam score anymore,
and you can change the messages or re-add the score by changing the
runtime SAmsg* variables

All SA-Exim log now looks like this:
- "SA: PANIC: "		-> severe errors
- "SA: Warning: "	-> config file parsing errors
- "SA: Notice: "	-> misc info on what SA-Exim is doing or not doing
- "SA: Action: "	-> what action SA-Exim took on a mail after scanning
- "SA: Debug[X]: "	-> misc debug info if enabled

Marin Balvers has written a nice log parser here:
http://nossie.addicts.nl/projects/sa-exim-stats/



FAQ
---
Why do I get this in my exim logs?

2004-05-15 12:43:57 1BP54T-0002gV-Nu TLS send error on connection from internalmx1.company.tld (internalmx.company.tld) [192.168.1.1]:51552: Error in the push function.                                           
2004-05-15 12:43:57 TLS recv error on connection from internalmx1.company.tld (internalmx.company.tld)              
[192.168.1.1]:51552: The specified session has been invalidated for some reason.

This is because you are teergrubing a host that is doing TLS. Teergrubing does
not work with TLS, and people doing TLS with you are probably known relays which
you should exclude from your teergrube list (SAteergrubecond)
Subversion Repositories

(root)/tags/4.2.1-14/README @ 86 - Rev 67 → 73
