SPAMCONTROL is an extension for qmail. It provides the following features:
Enhancements for qmail-smtpd:
Enhancements for qmail-remote:
Enhancements for qmail-pop3d:
Enhancements for qmail-queue:
Enhancements for qmail-send:
With SPAMCONTROL, qmail-smtpd can stand the two most common threats:
Additionally, qmail-smtpd allows
Throughout this document, I assume that qmail-smtpd is under control of supervise (out of the Daemontools package) and served by tcpserver (part of the UCSPI package) or a patched version of sslserver (part of the UCSPI-SSL package).
A typical - minimal - so called run script looks like follows:
Qmail - and SPAMCONTROL - relies on the concept of environment variables which are available for a task (sharing the same environment). qmail-smtpd may be fed by environment variables in three different fashions:
While the first three cases define static and "global" environments variables, the last case makes the environment variables client-dependent and - by means of tcprules - dynamically changeable. Any mixture is possible, though only the "last" setting of an environment variable is effective!
As a convention, I will call the tcperver's cdb, which rules the behavior of qmail-smtpd, tcp.smtp. A typical tcp.smtp would look like
The cdb is constructed on the fly:
Caution: For use with tcpserver, the value of the environment variable has to be included in quotes.
Though qmail can live happily without the knowledge of domains to be responsible for as provided by rcpthosts/morercpthosts.cdb, it is highly adviceable to include all domains to receive emails for (as per DNS MX Records) into those control files. Otherwise, qmail-smtpd may act as an Open Relay. Further, some LOCALMFCHECKs will fail, as discussed below.
qmail-smtpd's "Mail From:" parameter parser is used to detect and evaluate the SIZE parameter and to eventually reject messages which initially exceed the databytes limit.
Nevertheless, qmail-smtpd checks the size of the incoming message anyway.
For incoming E-Mails which exceed the message size values (in Bytes) defined in
or via the $DATABYTES environment variable.
SMTP Authentication requires a Client to authenticate and a Server to honor the authentication procedure. In this version of SPAMCONTROL, Qmail acts as an Authentication Server for qmail-smtpd and as an Authentication Client for qmail-remote.
Usually, a MTA (such as Qmail) will accept transmissions of E-Mails anyway as long as the "Rcpt To: <forwarding-path>" is targeted to a local Recipient (according to control/rcpthosts). However, with SMTP Authentication you may allow an authenticated User to relay E-Mails. In this respect, SMTP Authentication copes with the deficiencies of the POP3/IMAP4 protocol and is applied as an alternative to SMTP-after-POP, which is ugly as well.
I have taken the SMTP-Auth Patch from Krzysztof Dabrowski and included this into SPAMCONTROL. However, SPAMCONTROL's implementation is compliant with the checkpassword API designed by Dan Bernstein (the Plugable Authentication Module PAM).
SPAMCONTROL provides the following features:
While SASL is a generic concept, the information flow for authentication between e.g. qmail-smtpd and the PAM is defined by Dan Bernstein's checkpassword API. SPAMCONTROL provides the PAM on file descriptor 3 as an informational string composed of:
You are free to choose or even write your own PAM program, but in any case, the SASL Procedure of the client and the server has to match and the procedure has to be advertised. Compliant PAMs:
qmail-smtpd including SMTP Authentication may be called by tcpserver/sslserver in a supervise run script. Here is an example (with some more features):
Beware! Unlike the original implementation, I omitted the inclusion of the Hostname as argument for qmail-smtpd.
Unlike the standard qmail-smtpd, now you have
The environment variable SMTPAUTH may be left blank to allow Authentication types "PLAIN" and "LOGIN" or may be currently set to "cram..." (lower case) to enable CRAM-MD5 authentication in addition.
For SMTP Authentication, a User Database has to be generated and maintained. The SMTP Authentication User may exist independently of any System Users, Qmail Users, or E-Mail Accounts. In case of the modified cmd5checkpw I decided to keep the User in the Qmail directory as
There exist other flavors, in particular the saslpasswd scheme or the Cyrus SASL Library you may want to use. Further, for users with POP3/IMAP4 Accounts on the system it is advisable to use a common User Database. For Vpopmail you may use vchkpw.
However, since you are free to use any other checkpassword compliant PAM, it's up to you whatever you apply. Please remember: In order to access the Unix /etc/passwd the respective program has to run as root.
You can use the environment variable REQUIREAUTH to enforce authentication for particular clients. A typical run script to require SMTP Authentication for particular SMTP clients looks like:
SMTP Authentication works well with vpopmail, however, you have to use a checkpassword compatible PAM. Older versions of vchkpw have to be patched accordingly (see http://www.fehcom.de/qmail/smtpauth.html).
vchpkw offers a lot of authentication capabilities; it supports login, plain, and CRAM-MD5 and may authenticate the user against a mysql database and others. In start-up script for qmail-smtpd you have to make sure to access the user database with the correct user access rights:
If you use Sqwebmail in addition, the user is free to set his/her own password.
SPAMCONROL's STARTTLS support for qmail-smtpd is aligned with Scott Gifford's approach and depends on the following:
My STARTTLS implemention conforms with RFC 3207.
Most current STARTTLS/TLS solutions depend on the existance and availability of the OpenSSL libraries -- so does SPAMCONROL. However, unlike other implementations, qmail-smtpd is insulated against OpenSSL by means of sslserver. In fact, all encryption and certificate verification is facilitated by sslserver. In this respect, Scott's and my STARTTLS implementation is very much OSI-like. The communication and presentation happens at a well defined environment, typically assigned to the user and group ssl. Any potential attacks or bugs are kept away from the application and don't harm.
The reading and response to client cerificates and the actual encryption happens in the assigned user spaces; which should never be root.
Install Superscript's ucspi-ssl (version 0.70). Apply the attached patch "ucspi-ssl-0.70-ucspitls-0.4.patch_" against the source, typically found at /package/host/superscript.com/net/ucspi-ssl-0.70/src and execute package/install base. This patch includes (delayed) STARTTLS support to sslserver and will allow to substitude tcpserver completely, even if no SSL/TLS communication is required.
Further, it is helpful to create a low privileged user and group ssl, which will be used by sslserver for SSL/TLS communication purposes. Please follow Scott Giffords' advices.
Substiute tcpserver with sslserver in the run script for qmail-smtpd. If you use softlimits, it might be necessary to raise those settings significantly due to the increased memory requirements. Here is my run script:
It is absolutely necessary to use the "-n" flag for sslserver, since this will trigger the availability of encrypted communications channels between sslserver and qmail-smtpd.
The environment variables needed to feed sslserver are included in the "profile" /var/qmail/ssl/env. In my system, they have the following settings:
Of course it is required, to have raised the directory /var/qmail/ssl before and generate via openssl the appropriate keys before.
Note: These settings are "global"; however, by means of sslserver and the settings in your tcp.smtpd file it is possible to use different certificates per connection.
Comment: Please read the documentation of UCSPI-SSL carefully w.r.t. the "mod-ssl" variables. It might in addition be necessary to define CAFILE, CADIR and other SSL options to your needs.
After you verified your stettings, restart qmail-smtpd. Whether qmail-smtpd will present "STARTTLS" in the EHLO dialogue, depends on the presence of the UCSPITLS environment variable. These can be set i.e. per IP in the tcp.smtpd control file.
In order to allow a selecitive use of STARTTLS and certificate verification you can use the following environment variables:
In case qmail-smtpd is instructed to use STARTTLS and SMTPAUTH, SMTP Authentication always takes place after the TLS session is active, but never reverse. Thus, all SMTP parameters like username and password are already encrypted. Of course, SMTP Authentication is still available for unencrypted SMTP connections, and STARTTLS does not require per se SMTP Authentication. However, STARTTLS and SMTP Authentication is a strong and powerful couple to secure the SMTP communication
SPAMCONTROL displays the use of STARTTLS in the Received header (according to RFC 3207). The following information is added:
Here's a sample using Thunderbird als email client:
Mail From: Adress Verification (MAV) is a mean to enforce the use of the SMTP "Mail From:" address for particular Relayclients. Former versions of SPAMCONTROL used a "LOCALMF" check which allowed only a very limited granularity. However, with MAV you can control/enforce
MAV is in particluar very useful, if emails from your domains have to be undoubtly "officially" send.
Mail From: Address Verification is only be done if the flag 'relayclient' is set. This flag is set if
For these circumstances MAV can be enforced by means of the environment variable LOCALMFCHECK:
The file control/mfrules follows roughly the same syntax as the common file tcp.smtpd used for tcpserver/sslserver. It assigns either a complete SMTP address, a FQDN, an IP adress or a domain to a set of allowed Originator addresses. In practice control/mfrules allows
Once you have populated control/mfrules run qmail-mfrules to derive control/mfrules.cdb from the input file. Additionally, define LOCALMFCHECK="!" either gobally or in the tcp.smtpd file.
Mail Clients may be setup wrongly or a user may want to use the relaying MTA to send emails for a different name. In case, MAV is in place and well configured, the particular user will not be allowed to send the mail over the gateway receiving the following SMTP reply:
Since this might not be helpful for the (innocent) sender, you might use the environment variable REPLYMAV to add a qualification to that message.
MAV puts the burden of SMTP Originator address verification ot the relaying MTA; that is the reverse scheme compared to SPF and others. Emails qualified through MAV are labeled with "ESMTPM" in the Received header, which is generated by qmail-smtpd.
The SMTP Envelope consists of three parts:
SPAMCONTROL allows to filter E-Mails according to the bad* criteria with a so-called wildmat search, which is a subset of the known Regualar Expressions (RegEx). The wildmat search works in order least significant to most significant and includes
The following sets of wildmat control characters can be used:
Any E-Mail address, lets say <user@host.domain.com> consists of a
E-Mail addresses for local accounts are considered case-insensitive and delivered irrespective of their case.
Lets say - if the local account is "admin" and the RCPT to: tells <AdMin> or <adMIN> the delivery will be successful.
RFC 2821 says: "These commands (HELO/EHLO) are used to identify the SMTP client to the SMTP server. The argument field contains the fully-qualified domain name of the SMTP client if one is available. In situations in which the SMTP client system does not have a meaningful domain name (e.g., when its address is dynamically allocated and no reverse mapping record is available), the client SHOULD send an address literal (see section 4.1.3), optionally followed by information that will help to identify the client system. y The SMTP server identifies itself to the SMTP client in the connection greeting reply and in the response to this command."
Qmail records the HELO/EHLO greeting string for every received message in the E-Mail "Received:" header in case the provided HELO/EHLO string is different from the connecting hosts FQDN:
The HELO/EHLO string is included as "(HELO foo)". The HELO/EHLO string is usually generated by the sending MTA without much control (MUAs often use their generic hostname).
SPAMCONTROL allows a flexible filtering of the clients HELO/EHLO greeting string, which depends on the setting of the environment variable HELOCHECK:
The HELOCHECKs are only done, in case RELAYCLIENT is not set (split-horizon fashion). In my current setup, a useful setting is HELOCHECK="." and with the following input in control/badhelo
These settings exclude the spoofing of the MTA's own address, which is typical for spam senders, since they determine the EHLO/EHLO greeting from the initial IP/SMTP session parameter.
SPAMCONTROL allows four types of checks against the provided "Mail From:" SMTP envelope address (which I often call the "Originator"):
Invoking the environment variable MFDNSCHECK in the qmail-smtpd startup script enables globally the DNS check for the envelope's Sender.
Example:
Additionally, the environment variable may be defined individually within a cdb of tcpserver/sslserver. Typically, this is done for "non-trusted" hosts within a tcpservers cdb:
If environment variable MFDNSCHECK is not set, qmail-smtpd does not perform this DNS MX check.
Note: All DNS checks are either done by means of the libresolv library which comes with BIND, or my means of DJBDNS's routines, which can be included installing DJBDNS and using Nikola Vladov's enhancements for DJBDNS in addition with the modified Makefile.djbdns.
control/badmailfrom was the only SMTP envelope filter Dan Bernstein originally implemented for qmail-smtpd. Here, only particular names or perhaps domains were listeted to be rejected in the SMTP dialogue. Since then, various flavours of badmailfrom have been brought out. However, the approach to reduce spam emails feeding control/badmailfrom with known spammer addresses is comparable trying to hit a moving target. Almost all Originator addresses spammer use today are fake and in this sense are meaningless.
There exist a special case, where you expect an email with a specific Originator address to be send via particular MTAs. For instance, if you see an email with Originator address "support@microsoft.com", it has to be send from a Microsoft MTA. qmail-smtpd has the knowledge of the sender's IP and FQDN (by means of the environment variables TCPREMOTEIP and TCPREMOTEHOST) in case you use tcp-env, tcpserver, or sslserver with the appropriate argument, i.e. tcpserver -h.
MTAs for which the FQDN can't be resolved are unqualified. In particular, emails from the large webmail providers (aol, hotmail, yahoo, gmx, t-online ...) have always to be send from qualified MTAs. Reversely, you can safely reject emails with those Originator hostparts, which can not be resolved tcpserver/sslserver records them as "unkonwn".
With SPAMCONTROL's badmailfrom implementation, you simply include the Originator addresses for which you enforce a qualified TCPREMOTEINFO into control/badmailfrom in the following way appending a dash ("-"):
Note 1: Since tcp-env/tcpserver/sslserver
relies on a qualified DNS lookup, it is certainly helpful to use
DJBDNS' dnscache as frontend.
Note 2: Wildmat support is not provided; thus an
entry "@*.yahoo.com-" won't work.
Another special case is given, rejecting none-Relayclient emails with Originator addreses spoofing your domain name or email addresses. Email can be rejected if the "responsible domains" are included with a trailing plus ("+") in the following way into control/badmailfrom:
Note: Wildcard support is not provided in this case.
Apart for the RECIPIENTS mechanism, which is detailed later, you can reject SMTP Recipient addresses (Rcpt To: <Recipient>) by means of control/badrcptto. However, qmail-smtpd lets you effectively
Note: The provided Rcpt To: <Recipient> information by the SMTP client is (apart from it's IP/FQDN) the only information which can not be faked, though these addresses are today often randomly generated by means of lexical/dictionary attacks by spammers or gathered by address harvesting. Standard qmail will accept any addresses which matches an entry in control/rcpthosts or control/morercpthosts.cdb and in case the Recipient does not exist tires to bounce the email to the Originator after control/qeueulifetime has exceeded (default one week).
By populating control/badrcptto you reject emails to Recipients listed there in already in the SMTP session. Wildcards are allowed. If you don't wont to receive emails for root (from the Internet) include in control/badrcptto:
Alternatively to the Recipients mechanism, as a side-effect of the wildmat filtering, you can use the control/badrcptto file as an effective whitelisting mechanism. The trick is, to initially reject everything while later to allow specific Recipients:
*
!*@otherdomain.com
!user1@maydomain.com
!user2@mydoamain.con
Note: The evaluation of control/badrcptto is done independent from the setting of the RELAYCLIENT environment variable.
The environment variable
can be used to restrict the number of counted "Rcpt To: "s in the SMTP session. By default, no restriction is facilitated.
I have included Chris Johnson's TARPITTING patch into SPAMCONTROL:
"What is tarpitting? It's the practice of inserting a small sleep in a SMTP session for each "Rcpt To:" after some set number of "Rcpt To:"s. The idea is to that spammers who would hand your SMTP server a single message with a long list of RCPT TOs. If a spammer were to attempt to use your server to relay a message, say, 10,000 Recipients, and you inserted a five-second delay for each Recipient after the fiftieth, the spammer would be 'tarpitted', and would most likely assume that the connection had stalled and give up."
Two additional control files can be employed:
is the number of "Rcpt To:"s to accept before starting tarpitting and defaults to 0 (no tarpitting).
is the number of seconds of delay to introduce after each subsequent "Rcpt To:". Default is 5 seconds.
Instead, the environment variables TARPITCOUNT and TARPITDELAY can be used.
Note: In combination with the Recipients extension, the TARPITCOUNT is used to terminate the SMTP session if the number of invalid Recipients ("Rcpt to:") exceeds the TARPITCOUNT. Unlike the typical tarpitting mechanism, this is a hard limit (Smart Rejection).
qmail-smtpd accepts messages if the SMTP domain part of Recipient address ("Rcpt to: <recip@domain>") matches an entry in control/rcpthosts or control/morercpthosts.cdb. The existence of a mailbox/maildir for the corresponding SMTP Recipient is checked later in the delivery chain. In case no Mailbox/Maildir exists, the message is bounced back to the SMTP Sender ("Mail From: <send@example.com>").
For normal SMTP mail traffic that's fine as long as the rate of undeliverable messages don't exceed 10% and the Sender is 'legitmate'; ie. exists. Today's situation is different: Spam and Virus attacks with forged/faked Sender addresses to a bunch of random Recipient addresses yield an undeliverable rate up to 90%. Worse, the generated bounces will never reach the Sender and a double-bounce is eventually send to the postmaster.
The Recipients extension makes qmail-smtpd aware of acceptable Recipients and is employed in a none-RELAYCLIENT case only. The Recipients are kept in 'fastforward' compatible cdbs for a case-insensitive quick lookup during the SMTP session.
The Recipients mechanism supports Qmail's address extensions. If a Recipient address like 'foo@mydomain.com' is included in a cdb, all VERP addresses like 'foo-bar@mydomain.com' are accepted for SMTP reception.
Within the Recipients mechanism you can define domain-wide wildlisting. Simply include '!@domain' into one of the cdbs to allow all addresses for domain "domain" to be accepted. For performance reasons it is advisable to put those "wilddomains" at the beginning of the first cdb.
Unqualified Recipient addresses are always translated to full qualified, appending the domain part '@localhost' (eg. 'foo' -> 'foo@localhost').
The qmail-smtpd Recipients extension is available by means of the control file
Here, you include the path to fastforward compatible cdbs in which you keep the Recipient addresses (in lower case).
- control/wilddomains.cdb
- users/recipients.cdb
etc/fastforward.cdb
The Recipients extension needs no customization except for the following circumstances:
With the Recipients extension qmail-smtpd will act for none-RELAYCLIENTs like follows
In any other case, a SMTP temporary failure protocol error is issued to the client saying:
Based on the "qmail-smtp-viruscan-1.1.patch" by Russell Nelson (and Charles Cazabon), SPAMCONTROL includes my WARLORD extension, which is a much robuster and efficient filter for BASE64 encoded MIME attachments and bundled with the Qmail High Performance Scanner Interface (QHPSI):
In case a badmimetype or badloadertype filter condition is met or a virus is detected, qmail-smtpd sends a SMTP 554 reply to the sender "554 sorry, invalid message content (#5.3.2)". Populating the REPLY554 environment variable, allows to include additional information (typically an URL), which can be used to deal with potential false-positives.
The badmimetype filter becomes active if
The control file control/badmimetypes.cdb is populated by the additional program qmail-badmimetypes which takes the input of control/badmimetypes. New MIME signatures can be added/removed on-the-fly. Bad MIME Type signatures have to have the length of at least 9 significant characters.
The currently included MIME signatures are:
Adding new badmimetypes is simple:
Comments (starting with "#") are allowed in badmimetypes; the length of the signature will be truncated to nine characters.
The badloadertype filter becomes active if
The BADLOADERTYPE mechanism deals in particular with "transport stealth" worms, ie. UPX encoded Windows executables.
badloadertypes.cdb is populated by the additional program qmail-badloadertypes which takes the input of control/badloadertypes The badloadertype mechanism looks for five significant strings in the BASE64 encoded data-stream which is matched against an entry in control/badloadertypes.cdb. badloadertype signatures can be added/removed on-the-fly.
The currently included Windows OS badloadertype signatures are:
Comments (starting with "#") are allowed in badloadertypes; the length of the signature will be truncated to five characters.
Caution: Unlike the badmimetype, the badloadertype signatures are placed anywhere in the BASE64 encoded datastream and are difficult to find out. In order to make the search efficient, a common character has to be providen in the environment variable BADLOADERTYPE. The provided pattern look basically for a string like "32.dll" as a subpart of "Kernel32.dll" which is an indication for an executable for the Windows OS. However, there is a small chance for false positives. Some - lets say - Word document attached as BASE64 MIME part in the message containing the buzz words "kernel32.dll" might become flagged and finally rejected as well.
Unlike all other AV Scanners currently in use for Qmail, with Qmail High Performance Scanner Interface (QHPSI) there is no need for any other umbrella program, neither qmail-scanner, AMAViS, qscanq or whatsoever. Further, no additional MIME analyzing program like reformime, metamail, or ripmime is required. Even better, no "staging" area for temporary files are needed, except the one, the AV Scanners requires for itself.
Today's AV Scanners - and in particular Clam AV - are able to read the BASE64 encoded message and eventually dig out the files in archives, ie. in zip format. In order to use an AV Scanner with QHPSI, the AV Scanner has to have the following qualifications:
The QHPSI allows to use the following environment variables:
The AV Scanner is directly called in the start scripts of Qmail (i.e. the run script for qmail-smtpd) or by means of tcpserver's capabilities. Here is a typical example, how to customize QHPSI together with Clam AV (clamd/clamdscan) for a tcpserver tcp.smtpd file:
Comments:
Note:
Results:
Here is a sample of Clam AV without and with the argument "--disable-summary":
Note: Even in case no virus is detected, the "SCAN SUMMARY" is provided.
Attention:
Note: As with this writing, clamav 0.8x is broken, since it writes all logs to STDOUT instead of STDERR; thus no scanning messages will apear in the qmail-smtpd log.
The badmimtypes and badloadertypes mechanism provides a wire-speed filtering of incoming emails. However, typically all not-filtered emails are subject of the AV Scannner as defined via the QHPSI. Almost all worms and virii are transported as BASE64 encoded attachments (except some trojans, encapsulated as HTML files). By means of the environment variable
one can advice QHPSI to scan only those emails which contain a BASE64 encoded attachment.
qmail-smtpd will send a SMTP 554 Error Reply under the following conditions:
The SMTP Reply code for the first three conditions is always "554 sorry, invalid message content (#5.3.2)". The rejection of email because of the message content is due to some internal policy. For those users, which a subject of this policy innocently (and did not send ie. a virus mail on purpose), it might be advisable to explain the company's email policy.
The environment variable
allows to include a particular SMTP 554 Reply. Typically, an URL is referenced: REPLY554="[ see: http://www.fehcom.de/emailpolicy.html ]" which allows to detail possible circumventions.
Bruce Guenter's Qmail QUEUE_EXTRA patch has almost the rank of a recommended patch, because it's used by many Qmail extensions like the Qmail-Scanner and qmail-qfilter.
The actual use is controlled via the content of environment variable "QMAILQUEUE", which usually set in a tcpserver's cdb ie. tcp.smtp or globally defined in the qmail-smtpd's run script. A typical use is:
which advices qmail-smtpd to use the executable qmail-qfilter as first stage queueing program instead of qmail-queue itself.
Note: The QUEUE_EXTRA patch is not applied against qmail-smtpd but rather against the module qmail.c itself, since it is just an extension to the general queue call-mechanism.
.
SPAMCONTROL modifies and extends the behavior of qmail-remote in the following ways:
The qmail-remote (qmail-smtp-auth-send) authentication from Bjoern Kalkbrenner has been included in a modular and RFC-complient version. qmail-remote sessions can be SMTP authenticated with the types PLAIN and LOGIN on a per-sender (Reverse-Path) base. Thus for each sender you can advice qmail-remote to use SMTP authentication with a particular username and password connecting to relay at port.
The qmail-remote authentication follows in this respect the smtproutes mechanism. Authentication for outgoing SMTP sessions is faciliated, if the control file
is populated accordingly. Sample:
Typically sites/domains on the Internet are reachable over serveral MTA listed and deployed in the DNS MX records (o.k. qmail.org is an exception). By theory, the MX with the smallest weight is the primary MX for that domain;though often sites have redundant MTA with equal weights:
In order to deliver emails, qmail-remote follows two strategies:
In case this MTA is exhausted and rejects the connection during the EHELO/HELO greeting, qmail-remote exists and retries the very same MTA again with it's quadratic queue schedule mechanism.
Running EZMLM with many messages to the vary same domain but different Recipients, email delivery may become throttled, which particularly happens for t-online.de sites which don't allow too many connections from the same client MTA (a policy which is actually not covered by any SMTP RFC).
Back 10 years ago, when Dan was designing qmail he already was aware of that problem:
and included already the code base into qmail-remote which I simply activated. Thus, in case qmail-remote receives a rejection during the EHLO/HELO greeting it will simply try the next MTA for the DNS MX list.
qmail-remote incorporates two performance critical steps for the delivery:
Bruce Guenter recognized the last fact and patched qmail-remote accordingly ("fastremote"), thus qmail-remotes processes the input data in chunks of 4 Kbyte. This patch has been included into SPAMCONTROL.
The STLS (Start TLS support) for qmail-popup follows the same scheme as for qmail-smtpd
Requirements:
Actually, you use the same "env" file as for qmail-smtpd. In this case qmail-popup announces in the capability list "STLS" and the following POP3 dialogue is encrypted as is the transmission of the received emails as well.
In order to use SSL encryption for a POP3 connection, the following run script for qmail-pop3d is appropriate:
Note: In this run script I use Bruce Guenter's checkvpw as PAM (for vmailmgr), which requires the additional presence of the "Maildir" argument after the call of qmail-pop3d.
The profile /var/qmail/ssl/env is the same as for qmail-smtpd. Defining the environment variable UCSPITLS directly in the run script instead of the profiles, allows a flexible use of the STARTTLS/STLS option for qmail-pop3d and qmail-smtpd without modifying the common profile.
Bounces have generally a Null-Sender address (Mail From: <>) and are out-of-band error-messages to indicate a failure in the delivery process. In fact, RFC 2821/821 requires that all notification E-Mails have to have a Null-Sender address!
For every undeliverable message, qmail-send generates a bounce to the Sender.While this is legitimate and necessary for normal operation, in case of SPAM attacks the bounces are meaningless:
Unless you use a 'whitelisting' of Recipient E-Mail addresses, there is not much to do about. However, SPAMCONTORL helps you in three cases:
By definition, a bounce is a SMTP notification for a failure situation. It is common practice, to include the original message in the bounce. Qmail uses a specific format, introduced by Dan Bernstein and called "QSBMF" (qmail-send Bounce Message Format); other MTA encapsulate the original message as MIME attachment in rfc822/message format.
Anyway, for a legitimate bounce reaching the Sender the original message is usually of no interest, except for identification purposes. In order to save bandwidth, you can limit the size of bounces using the control file
Unlike the original patch (from Frank DENIS aka Jedi/Sector One <j@4u.net>), the default value is '0' byte, meaning no limits. A useful limit would by 2000 (byte), which covers the header and some body part information. The average size of a SPAM E-Mail is 5 Kbyte.
The original message included in the bounce will be limited to the defined bouncemaxbytes and truncated, which is displayed in the bounce with "--- Rest of message truncated." at the end of the bounce.
Double bounces are generated, if the bounce can not be delivered to the Sender.
Double bounces are usually delivered to the 'Postmaster' account. It is convenient that this account is local and eventual double bounces are stored in a mbox/Maildir for later inspection. However, Qmail allows you to forward double bounces to some other account defined in
However, due to the forged Sender address in SPAM E-Mails, practically all bounces become double bounces eventually. In this case any storage and inspection is fruitless. Taken from Russell Nelson and Charles Cazabon, you can optionally dump all double bounces immediately. This is facilitated if doublebounceto contains a '@' in the first line.
Those dumped double bounces show up in the qmail-send log as: "double bounce: discarding".
Scott Gifford's "ipme.c" patch (or qmail-0.0.0.0-patch) is included by default. According to RFC 1122, Sec. 3.2.1.3 the IP address "0.0.0.0" is a special address which always refers to "this host, this network". You may wish to tell Qmail about arbitrary IP addresses employing the moreipme patch and include the following control files:
See the enhanced man page qmail-smtpd and or consult the README.moreipme.
As a further gadget, the qmail-send control files
are re-read by means of a HUP signal (eg. svc -h /service/qmail-send).
The queue directories ./intd and ./todo may be splitted (as per conf-split) into subdirectories to allow a more efficient treatment of many incoming messages.
Caution: Make sure, that the directories ./queue/todo and ./queue/intd are empty before applying the patch; otherwise qmail-send will not be able to process those messages anymore!
Note: The shell script qmail-qstat and in addition some qmail-mrtg analyses are affected by this patch.
Hint: Consider raising the value in conf-split in the first place !
The following fixes for Qmail's sendmail wrapper have been included for compatibility reasons:
SMTP allows to reject Sessions based on some technical and/or political criteria, which are not well expressed in the RFCs (2821, 2554, 2505, 1122).
The SMTP protocol mechanism between the client and the server are defined as Commands and Replies. SMTP uses a three-letter Reply Code. The first digit tells whether a command was accepted and completed (2), transaction begin (3), or whether there was as transient (4) or permanent failure (5). In addition, an explanatory description may be given.
RFC 1893 introduces a concept of "Enhanced Mail System Status Codes" (EMSSC) which should provide easily parseable SMTP server conditions and transaction statuses, usually at the end of the SMTP reply and included in parenthesis, eg. (#5.5.1).
The SMTP Reply Codes and the EMSSC are detailed in the corresponding RFCs, but don't fit well to each other, thus either providing redundant information or almost no additional information at all. In short, the EMSSC is nowadays almost meaningless.
Here's a breakdown of SPAMCONTROL's SMTP Reply Codes, informational texts, and the used EMSSC.
Reply |
Informational text |
EMSSC |
| 450 | sorry, mailbox currently unavailable | (#4.2.1) |
| 451 | DNS temporary failure | (#4.3.0) |
| 454 | TLS not available due to temporary reason | (#5.7.3) |
| 501 | auth exchange canceled | (#5.0.0) |
| 501 | malformed auth input | (#5.5.4) |
| 503 | you're already authenticated | (#5.5.0) |
| 503 | no auth during mail transaction | (#5.5.0) |
| 503 | sorry, SMTP Authentication not available | (#5.7.3) |
| 504 | auth type unimplemented | (#5.5.1) |
| 535 | authorization failed | (#5.7.1) |
| 535 | authentication required | (#5.7.1) |
| 535 | STARTTLS required | (#5.7.1) |
| 550 | sorry, invalid HELO/EHLO greeting | (#5.7.1) |
| 550 | sorry, your envelope recipient is in my badrcptto list | (#5.7.1) |
| 550 | sorry, invalid sender address specified | (#5.7.1) |
| 550 | sorry, too many recipients | (#5.5.3) |
| 550 | sorry, bounce messages should have a single envelope recipient | (#5.7.1) |
| 552 | sorry, that message size exceeds my databytes limit | (#5.3.4) |
| 553 | sorry, your envelope sender is in my badmailfrom list | (#5.7.1) |
| 550 | sorry, that domain isn't in my list of allowed rcpthosts | (#5.7.1) |
| 553 | sorry, your envelope sender domain must exist | (#5.7.1) |
| 554 | too many hops, this message is looping | (#5.4.6) |
| 554 | sorry, invalid message content (optional text) | (#5.3.2) |
Normally, qmail-smtpd doesn't log anything.With SPAMCONTROL qmail-smtpd logs accepted and some (important) rejected SMTP session attempts. The logging is done at the end of the "Rcpt To:" and eventually at the end of the "Data" phase.
Action |
Type |
Condition |
Explanation |
| Reject | SMTP | Toomany_Hops | Message Hop count exceeded |
| Reject | SMTP | Syntax_Error | Malformed SMTP address (e.g. missing brackets) |
| Reject | DATA | Invalid_Size | DATA exceeds sizelimit |
| Reject | DATA | Bad_MIME | DATA includes BASE 64 MIME type listed in badmimetypes |
| Reject | DATA | Bad_Loader | DATA includes BASE64 loader type listed in badmimetypes |
| Reject | DATA | Virus_Infected | DATA includes virus infected message (QHPSI) |
| Reject | SNDR | Bad_Helo | SNDR's HELO is in the badhelo |
| Reject | SNDR | DNS_HELO | SNDR's HELO has no DNS A/MX RR |
| Reject | SNDR | Invalid_Relay | SNDR's tries relaying; but not allowd |
| Reject | SNDR | Missing_TLS | STARTTLS was required but not granted by client |
| Accept | Accept | Relay_Client | SNDR was identfied as relay client |
| Accept | Accept | Start_TLS | SNDR succesfully started TLS |
| Reject | ORIG | Bad_Mailfrom | ORIG is in badmailfrom |
| Reject | ORIG | DNS_MF | Domain part of ORIG has no DNS MX RR |
| Reject | ORIG | Failed_Auth | ORIG tried SMTP Authentication; but failed |
| Reject | ORIG | Invalid_Sender | ORIG not allowed to send |
| Reject | ORIG | Missing_Auth | SMTP Authentication required, but not granted |
| Accept | ORIG | Valid_Auth | ORIG was successful authenticated |
| Accept | ORIG | Local_Sender | ORIG was identified as local sender address |
| Accept | ORIG | Relay_Mailfrom | ORIG was accepted als Relaymailfrom |
| Reject | RCPT | Bad_Rcptto | RCPT is in badrcptto |
| Reject | RCPT | Toomany_Rcptto | Too many RCPTs |
| Reject | RCPT | Failed_Rcptto | RCPT could not acceptd as per recipients/cdb. |
| Accept | RCPT | Recipients_Rcptto | RCPT was accepted as per recipients/cdb. |
| Accept | RCPT | Recpients_VERP | RCPT was accepted per VERP address in recipients/cdb. |
| Accept | RCPT | Recpients_Domain | RCPT was accepted per Domain wildlisting in recipients/cdb. |
| Accept | RCPT | Rcpthosts_Rcptto | RCPT was accepted as per rcpthosts/morercpthosts |
The Information includes typically the following
This scheme is easy extensible to other successful/deferred SMTP sessions. Sample:
Accept::SNDR::Relay_Client: P:orion.fehnet.de S:81.173.229.48:xdsl-81-173-229-48.netcologne.de H:mail.fehcom.de F:feh@fehcom.de T:erwin@example.com
A typical tcpserver start script applying standard splogger:
Since splogger is now facilitated, ACCUSTAMP time information is included.
A better choice would be multilog. multilog allows you to write separate filtered logs; to individual directories, and/or files, STDERR respectively.A typical Daemontools qmail-smtpd run script would look like:
Note: tcpserver's logging via the '-v' flag can be omitted to get mostly a full comprehensive and terse one-line logging of the SMTP session.
The corresponding multilog run script allows not only to filter the log information and write them to the file "current" in a specific directory but in addition to feed a file with specific information; here's a sample:
In this case, multilog adds at first a TAI64 time stamp.
In this version of SPAMCONTROL I have substantially reduced the number of compile-time options:
In order to consistently change all relevant binaries, use the file conf-spamcontrol which is evaluated by the installation routine install_spamcontrol.sh and passes the changes to the Qmail c-files:
In case your E-Mail environment complies to the assumption in PURPOSE do the following:
to your needs.
See above samples and check the included samples for ./badmailfrom and ./badrcptto.
*) Not useful, if tcpserver in employed.
Good luck!
Attention: For 64 bit OS'es you may need in addition the qmail-isoc patch from James Craig Burley (http://www.jcb-sc.com/qmail/patches/qmail-isoc.patch).
The SPAMCONTROL patch is incompatible with the Qmail LDAP patch. It should be applied against qmail-1.03 and not against netqmail-1.0x.
Actually, the Qmail LDAP enhancement is a super set of SPAMCONTROL. If you need to incorporate parts of SPAMCONTROL into the LDAP patch, look at the different pieces and pick 'em up from the sources.
Thanks to the discussion in the Qmail Mailing List (qmail@list.cr.yp.to) in particular:
Erwin Hoffmann (feh@fehcom.de)
Cologne, 2006-06-15