man FBB::Milter (3): Interface to the sendmail mail filter facilities

SYNOPSIS

#include <bobcat/milter>
Linking option: -lmilter -lbobcat

DESCRIPTION

Milter defines an abtract base class interfacing to the sendmail mail filter (milter) facilities. It defines a C++ interface, based on the assumption that a single mail filter program does not implement multiple mail filters. The traditional sendmail C-based Milter API uses a (SMFICTX) pointer representing a mail connection, and a pointer to connection-specific `private' data, requiring the Milter constructor to perform quite a few administrative tasks. While acceptable in a C environment these administratve tasks distract from the main task: the Milter's mail filtering functionality. The FBB::Milter class hides these administrative tasks from the programmer, who is then able to concentrate on filtering mail. The main benefits of Milter are therefore

Basic administration is performed by the Milter class
The class' interface is more C++ like than the raw C interface offered by the milter API.
Administration, allocation and communicating of connection specific data is no longer required
It is not normally necessary to use connection-specific data, like a pointer identifying the connection, anymore when implementing the Milter.
Milter uses current-day design patterns enforcing principles of reuable software, thus simplifying the construction of the actual Milter. To activate a milter from the sendmail.mc configuration file, use, e.g., INPUT_MAIL_FILTER(`name', `S=socket'), where name is the milter's name, and socket is the name of the socket. See also the setConnection member below.

NAMESPACE

FBB
All constructors, members, operators and manipulators, mentioned in this man-page, are defined in the namespace FBB.

INHERITS FROM

ENUMERATIONS

The class defines four enumerations. One enumeration is used to indicate the callback-functions that need to be called, the second one renames the flags that can be passed to sendmail to indicate which actions the milter is allowed to perform. The third one defines status values that may be used to inform sendmail how to further process a message. The fourth one defines return values. The enumerations are:

enum CallBack
This enumeration holds the following values:

CONNECT:
Indicates that the milter defines (overrides) the connection-functionality. This is the first callback function that can be called by sendmail.
HELO:
Indicates that the milter defines (overrides) the helo-functionality. This indicates that the helo function should be called by sendmail, providing the milter with information about the connecting client.
SENDER:
Indicates that the milter defines (overrides) the sender-functionality. This indicates that the sender function should be called by sendmail, providing the milter with the sender's envelope information.
RECIPIENT:
Indicates that the milter defines (overrides) the recipient-functionality. This indicates that the recipient function should be called by sendmail, providing the milter with the recipient's envelope information.
HEADER:
Indicates that the milter defines (overrides) the header-functionality. This indicates that the header function should be called by sendmail for each mail header that is used in the current mail message.
EOH:
Indicates that the milter defines (overrides) the end-of-header-functionality. This indicates that the eoh function should be called by sendmail once all header lines have been processed.
BODY:
Indicates that the milter defines (overrides) the body-functionality. This indicates that the body function should be called by sendmail, offering the mail-body to the milter.
EOM:
Indicates that the milter defines (overrides) the end-of-message-functionality. This indicates that the eom function should be called by sendmail, once all elements of the e-mail message have been processed.
ABORT:
Indicates that the milter defines (overrides) the abort-functionality. The abort function may be called by sendmail before eom is called. It should reclaim all resources used by the message, but not delete any memory allocated by the milter, as this is close's job.
CLOSE:
Indicates that the milter defines (overrides) the close-functionality. The close function should delete all (connection specific) memory allocated by the milter. It may be called `out-of-order', i.e. even before connect is called and developers should anticipate this possibility when crafting their close code. In particular, it is incorrect to assume the private context pointer will be something other than 0 in this callback.
UNKNOWN:
Currently not used. Reserved for versions exceeding version 2 of the sendmail milter API
DATA:
Currently not used. Reserved for versions exceeding version 3 of the sendmail milter API
ALL_CALLBACKS:
Shortcut to indicate all callback facilities. The CallBack values are bit-flags. The bit_or operator may be used to combine them, and the bit_not operator may be used to remove a flag from ALL_CALLBACKS (e.g., ALL_CALLBACKS && ~ABORT).

enum Flags
This enumeration holds the following values:

NO_FLAGS:
No flags are defined.
ADD_HEADERS:
This flag indicates that the milter is allowed to add headers to the current e-mail message.
ADD_RECIPIENTS:
This flag indicates that the milter is allowed to add recipients to the current e-mail message.
CHANGE_BODY:
This flag indicates that the milter is allowed to modify the message's body contents.
CHANGE_HEADERS:
This flag indicates that the milter is allowed to change headers of the current e-mail message.
DELETE_RECIPIENTS:
This flag indicates that the milter is allowed to remove recipients from the current e-mail message.
QUARANTINE:
This flag indicates that the milter is allowed to request sendmail to quarantine the current e-mail message.
ALL_FLAGS:
Shortcut to indicate all callback facilities. The Flags values are bit-flags. The bit_or operator may be used to combine them, and the bit_not operator may be used to remove a flag from ALL_FLAGS (e.g., ALL_FLAGS && ~QUARANTINE).

Status
This enumeration simplifies the extended SMFIS_ values used by the C API. These values may be used to return sfsistat values:

ACCEPT:
This value is equal to SMFIS_ACCEPT, indicating that Sendmail should accept the message. For a connection-oriented callback (see below at PROTECTED VIRTUAL MEMBER FUNCTIONS), accept this connection without further filter processing, call close (see below). For other callbacks: accept this message without further filtering.
CONTINUE:
This value is equal to SMFIS_CONTINUE, indicating that Sendmail should continue processing. This is the default for all callback functions which are not overridden by the class derived from Milter.
DISCARD:
This value is equal to SMFIS_DISCARD, indicating that Sendmail should discard the mail message. It should not be returned by a connection-oriented callback. For other callbacks: the message is accepted, but silently discarded.
REJECT:
This value is equal to SMFIS_REJECT, indicating that Sendmail should reject the mail message. For a connection-oriented callback, reject this connection; call close. For a message-oriented callback (except for eom or abort, see below), reject this message. For a recipient-oriented callback, reject the current recipient (but continue processing the current message).
TEMPFAIL:
This value is equal to SMFIS_TEMPFAIL, indicating that Sendmail should return a `temporary unavailable' message to the sender of the mail message. For a message-oriented callback (except sender, see below), fail for this message. For a connection-oriented callback, fail for this connection and call close. For a recipient-oriented callback, only fail for the current recipient and continue message processing.

Return
This enumeration simplifies the extended MI_ values used by the C API. Most return values used by the Milter class, however, are bool values. The Return values are:

FAILURE:
This value is equal to MI_FAILURE, indicating that a C-api function failed to perform its task.
SUCCESS:
This value is equal to MI_SUCCESS, indicating that a C-api function succeeded in performing its task.

CONSTRUCTORS

The default- and copy constructors are available for derived classes. They perform no actions.

PUBLIC STATIC MEMBER FUNCTIONS

These functions form the heart of the Milter base-class. They can be called to initialize, start and stop the Milter.

void initialize(std::string const &name, Milter &milter, callback_set callbacks = CONNECT, flag_set flags = NO_FLAGS):
This function initializes the Milter's administration. It expects the name of the mailfilter as its first argument. Its second argument is a reference to a Milter object. Since Milter is an abstract base class the actual object is always an object of a class derived from Milter. Its third argument specifies the callbacks to call for this milter. By default the connect callback will be called. Note that the close callback is not called by default. This is ok, since the Milter object doesn't have to cleanup `private' data, as is normal with the C API. The last argument defines flags, specifying the Milter's capabilities.
std::string const &name():
This function returns the milter's name.
bool start():
This member function calls smfi_main, controlling the milter's event loop. It returns true if the event-loop is successfully terminated.
void stop():
This member function terminates the milter's event loop, after finishing all threads. Following this call start may be called again to continue the milter.

PROTECTED VIRTUAL MEMBER FUNCTIONS

The remaining functionality of the class Milter is useful only for Milter-implementations in classes derived from Milter. The following members can be overridden by derived classes. Note that clone must be overridden. Except for clone, all the members in this sections are callback functions. I.e., the MTA will call them to process parts of the mail message. Recipient-, message-, and connection-oriented callbacks are distinguished.

The recipient-oriented callback (recipient, see below) may affect the processing of a single message to a single recipient. Connection-oriented callbacks (connect, helo and close) affect the processing of the entire connection (during which multiple messages may be delivered to multiple sets of recipients). The remaining callbacks are message-oriented, affecting the processing of a single message to all its recipients.

virtual sfsistat abort():
This message-oriented member may be called at any time during message processing (i.e. between some message-oriented routine and eom). abort reclaim any resources allocated on a per-message basis (which are not the connection specific data, which should be handled by the derived class' destructor), and must be tolerant of being called between any two message-oriented callbacks. abort is only called if the message is aborted outside the filter's control and the filter has not completed its message-oriented processing. For example, if a filter has already returned ACCEPT, REJECT, or DISCARD from a message-oriented routine, abort will not be called even if the message is later aborted outside its control.
virtual sfsistat body(unsigned char *text, size_t length):
This message-oriented member is called zero or more times between eoh and eom. text points to a sequence of bytes. It is not necessarily a 0-terminated. Moreover, the sequence may contain 0-characters. Since message bodies can be very large, defining body can significantly impact filter performance. End-of-lines are represented as received from SMTP (normally CR/LF). Later filters will see body changes made by earlier ones, and message bodies may be sent in multiple chunks, with one call to body per chunk.
virtual Milter *clone() const = 0:
This pure virtual function must be implemented by derived classes to return a newly allocated copy of the derived object passed to the initialize static member. It is used by the standard `virtual constructor' design pattern. The destruction of the allocated object is the responsibility of clone's caller.
virtual sfsistat close():
This connection-oriented member is always called once at the end of each connection. It may be called "out-of-order", i.e. before even the connect is called. After a connection is established by the MTA to the filter, if the MTA decides this connection's traffic will be discarded (e.g. via an access_db result), no data will be passed to the filter from the MTA until the client closes down. At that time, close is called. It can therefore be the only callback ever used for a given connection, and developers should anticipate this possibility when crafting their close code. The member close is called on close even if the previous mail transaction was aborted.
virtual sfsistat connect(char *hostname, _SOCK_ADDR *hostaddr):
This connection-oriented member may be called once, at the start of each SMTP connection. The parameter hostname is he host name of the message sender, as determined by a reverse lookup on the host address. If the reverse lookup fails, hostname will contain the message sender's IP address enclosed in square brackets (e.g. [a.b.c.d]). The parameter hostaddr is the host address, as determined by a getpeername(2) call on the SMTP socket. It is 0 if the type is not supported in the current version or if the SMTP connection is made via stdin.
virtual sfsistat data():
Not yet supported. Will be available with libmilter versions beyond 3.
virtual sfsistat eoh():
This message-oriented member is called once after all headers have been processed.
virtual sfsistat eom():
This message-oriented member is called once after all calls to body for a given message have been completed. Note that only in this function modifications to the message headers, body, and envelope can be made (see the add-, change- and delete- members listed below).
virtual sfsistat header(char *headerf, char *headerv):
This message-oriented member is called zero or more times between recipient and eoh, once per message header. The headerf parameter contains the text of the header, the headerv parameter contains its value. E.g., if an e-mail message contains the following headers:
```
    From: sender <[email protected]>
    Subject:no
        
```
then header will be called twice with the following values for, respectively headerf and headerv:
```
    First header:   "From",     " sender <[email protected]>"
    Second header:  "Subject",  "no"
        
```
Further details about header information is given in RFC 882.
virtual sfsistat helo(char *helohost):
This connection-oriented member is called whenever the client sends a HELO/EHLO command. It may therefore be called between zero and three times. The helohost parameter should be the domain name of the sending host (but is, in practice, anything the sending host wants to send).
virtual sfsistat recipient(char **argv):
This recipient-oriented member is called once per recipient, hence one or more times per message, immediately after sender. The parameter argv is a 0-terminated array of pointers to SMTP command arguments; argv[0] is guaranteed to be the recipient address. Later arguments are the ESMTP arguments. TEMPFAIL may be returned indicate that sendmail should return a temporary failure for this particular recipient; further recipients may still be sent, abort is not called. REJECT will reject this particular recipient; further recipients may still be sent, abort is not called. DISCARD will accept (but discard) the message, abort will be called. ACCEPT will accept recipient, abort will not be called. More details on ESTMP responses, are described in RFC 1869.
virtual sfsistat sender(char **argv):
This message-oriented member is called once at the beginning of each message, before recipient. argv[0] is guaranteed to be the sender's envelope address. Later arguments are the ESMTP arguments. TEMPFAIL may be returned, indicating that sendmail should return a temporary failure for this particular message, abort is not called. REJECT will reject this message, abort is not called. DISCARD will accept (but discard) the message, abort will be called. ACCEPT will accept recipient, abort will not be called. More details on ESTMP responses, are described in RFC 1869.
virtual sfsistat unknown(char *ptr):
Not yet supported. Will be available with libmilter versions beyond 2.

PROTECTED MEMBER FUNCTIONS

The following members are non-virtual. They can be called by members of classes derived from Milter:

bool addHeader(std::string const &hdrName, std::string const &hdrValue):
This member may only be called from eom, and the flag ADD_HEADERS must have been specified or it will fail. The hdrName and hdrValue must be non-empty strings. Each line of the header must be under 2048 characters and should be under 998 characters. If longer headers are needed, make them multi-line. To make a multi-line header, insert a line feed (\n) followed by at least one whitespace character such as a space or tab (\t). The line feed should not be preceded by a carriage return. It is the filter writer's responsibility to ensure that no standards are violated.
bool addRecipient(std::string const &rcptName):
This member may only be called from eom, and the flag ADD_RECIPIENTS must have been specified or it will fail.
bool changeHeader(std::string const &hdrName, size_t headerNr, std::string const &hdrValue):
This member may only be called from eom, and the flag CHANGE_HEADERS must have been specified or it will fail. See addHeader for the header-requirements. The headerNr parameter is a 1-based header index value. A headerNr value of 1 will modify the first occurrence of a header named hdrValue. If headerNr is greater than the number of times hdrName appears, a new hdrName-header will be added. If hdrValue is empty, the header is deleted.
bool deleteRecipient(std::string const &rcptName):
This member may only be called from eom, and the flag DELETE_RECIPIENTS must have been specified or it will fail. This member removes the named recipient from the current message's envelope.
SMFICTX *id() const:
This member may be called by the Milter object to obtain a pointer identifying its sendmail-connection. Normally it should not be necessary to call this member.
bool insertHeader(size_t hdrIdx, std::string const &hdrName, std::string const &hdrValue):
This member may only be called from eom, and the flag ADD_HEADERS must have been specified or it will fail. See addHeader for the header-requirements. The headerNr parameter is a header index value. A headerNr value of 0 will insert this header as the first of the hdrName headers. If headerNr is greater than the number of times hdrName appears, a new hdrName-header will be added.
bool openSocket(bool removeIfTrue = true):
This member should be called before start is called. This member attempts to create the socket specified by setConnection (see below). This allows the calling application to ensure that the socket can be created, possibly changing its protection (access rights) before the milter starts its work. If this member is not called, it will be called implicitly when run is started. It returns true if the socket could be created.
bool quarantine(std::string const &reason):
This member may only be called from eom and causes the MTA to quarantines the message using the given reason.
bool replaceBody(std::string const &body):
This member may only be called from eom, and the flag CHANGE_BODY must have been specified or it will fail. It may be called multiple times in which case the various body contents are concatenated in the final message. Newlines should be coded as CRLF.
bool setBacklog(size_t backlog = 5):
This member should be called before start is called. Sets the incoming socket backlog used by listen(2). If setBacklog is not called, the operating system default is used. The function returns false if the backlog could not be set as requested. It is the responsibility of the programmer not to call this function with a 0 argument.
bool setConnection(std::string const &name):
This member should be called before start is called. Sets the socket through which the filter communicates with sendmail. The socket may be specified using one of the following variants:
{unix|local}:/path/to/file - A named pipe;
net:port@{hostname|ip-address} - An IPV4 socket;
inet6:port@{hostname|ip-address} - An IPV6 socket.
If possible, filters should not run as root when communicating over unix/local domain sockets.
Unix/local sockets should have their permissions set to 0600 (read/write permission only for the socket's owner) or 0660 (read/write permission for the socket's owner and group) which is useful if the sendmail RunAsUser option is used. The permissions for a unix/local domain socket are determined as usual by umask, which should be set to 007 or 077.
Possible failure of this function cannot be determined from its return value. Rather, run will fail.
setReply(std::string const &rcode, std::string const &xcode = "", std::string const &msg = ""):
This member sets the default SMTP error reply code. It may be called from any callback member, except connect. The parameter rcode should be a he three-digit (RFC 821/2821) SMTP reply code and it must be a valid 4XX or 5XX reply code. The parameter xcode, when specified, must be a extended (RFC 1893/2034) reply code. The parameter msg may be an additional textual message. The Milter class has no member comparable to the libmilter API function smfi_setmlreply. the Milter class.
void setTimeout(size_t seconds = 7210):
This member should be called before start is called. Sets the number of seconds libmilter will wait for an MTA connection before timing out a socket. If setTimeout is not called, a default timeout of 7210 seconds is used. It is the responsibility of the programmer not to call this function with an argument equal to 0.
char const *symval(std::string const &name) const:
This member returns the value of a specific sendmail macro. The name parameter should be set to he name of a sendmail macro . Single letter macros can optionally be enclosed in braces ({ and }), longer macro names must be enclosed in braces, just as in a sendmail.cf file.0 is returned if the macro is not defined. By default, the following macros are valid in the given contexts:
for connect: daemon_name, if_name, if_addr, j, _;
for helo: tls_version, cipher, cipher_bits, cert_subject, cert_issuer;
for sender: i, auth_type, auth_authen, auth_ssf, auth_author, mail_mailer, mail_host, mail_addr;
for recipient: rcpt_mailer, rcpt_host, rcpt_addr.
All macros stay in effect from the point they are received until the end of the connection for the first two sets, the end of the message for the third (sender), and just for each recipient for the final set (recipient).
The following macros may be specified in the sendmail.mc configuration file:
define(`confMILTER_MACROS_CONNECT', `m1', ...), define(`confMILTER_MACROS_HELO', ...), define(`confMILTER_MACROS_ENVFROM', ...), define(`confMILTER_MACROS_ENVRCPT', ...), where `m1', ... represents a comma separated list of returnable macros. Single letter macros can optionally be enclosed in braces ({ and }), longer macro names must be enclosed in braces.
bool wait():
This member may only be called from eom and tells the MTA that the filter is still working on a message, causing the MTA to re-start its timeouts.

EXAMPLE

To do

FILES

bobcat/milter - defines the class interface

BUGS

-lmilter must be specified before -lbobcat.

DISTRIBUTION FILES

bobcat_4.02.00-x.dsc: detached signature;
bobcat_4.02.00-x.tar.gz: source archive;
bobcat_4.02.00-x_i386.changes: change log;
libbobcat1_4.02.00-x_*.deb: debian package holding the libraries;
libbobcat1-dev_4.02.00-x_*.deb: debian package holding the libraries, headers and manual pages;
http://sourceforge.net/projects/bobcat: public archive location;

BOBCAT

Bobcat is an acronym of `Brokken's Own Base Classes And Templates'.

COPYRIGHT

This is free software, distributed under the terms of the GNU General Public License (GPL).

AUTHOR

Frank B. Brokken ([email protected]).