COURIERTCPDSection: (1)Updated: 07 February 2004 |
COURIERTCPDSection: (1)Updated: 07 February 2004 |
couriertcpd -pid=pidfile [ option ... ] list program arg ...
couriertcpd -pid=pidfile -stop
couriertcpd -pid=pidfile -restart
couriertcpd accepts incoming network connections, and runs program after establishing each network connection. The program's standard input and output are set to the network connection.
list is a comma-separated list of TCP port numbers where incoming connections are created. program is the program to run. If program requires any arguments, they are specified on the command line, after program itself.
Before running program, couriertcpd initializes several environment variables that describe the network connection. The environment inherited by program will be the environment inherited by couriertcpd, plus any additional environment variables initialized by couriertcpd. It is also possible to reject certain network connections. Several options are available to specify which network connections will be rejected.
The list argument can be a comma-separated list of multiple port numbers. couriertcpd will create network connections on any listed port. Each port number can be optionally specified as "address.port", for example:
couriertcpd -pid=/var/run/smtp.pid 127.0.0.1.25,999 program
This instance accepts network connections to either port 25 or port 999, however connections on port 25 are created only on the IP address 127.0.0.1, the loopback interface.
Whenever an IP address is not specified, network connections are accepted to any IP address (called "wildcarding"). On IPv6-capable systems, couriertcpd will attempt to create two incoming network connection ports, if an IP address is not specified. After creating the first port as an IPv6 wildcard port, couriertcpd will then attept to create an IPv4 wildcard port, with the same port number. Some BSD-derived systems must use separate IPv6 and IPv4 wildcard ports to create incoming network connections. Most other systems only need an IPv6 port to create both IPv6 and IPv4 incoming network connections. couriertcpd quietly ignores a failure to create an IPv4 wildcard port, as long as an IPv6 wildcard was succesfully created.
The -address option can be used to default a specific IP address for every listed port number. For example:
couriertcpd -pid=/var/run/smtp.pid 127.0.0.1.25,127.0.0.1.999 program
and
couriertcpd -pid=/var/run/smtp.pid -address=127.0.0.1 25,999 program
will create network connections on ports 25 and 999 of the IP address 127.0.0.1.
The access file lists IP addresses that couriertcpd will accept or reject connections from. An access file is optional. Without an access file couriertcpd accepts a connection from any IP address.
Both IPv4 and IPv6 addresses can be specified, if IPv6 support is available. A non-standard syntax is currently used to specify IPv6 addresses. This is subject to change in the near future. IPv6 support is currently considered to be experimental.
The access file is a binary database file that's usually created by a script, such as makesmtpaccess(8), from one or more plain text files. Blank lines in the text file are ignored. Lines that start with the # character are also ignored.
The following line instructs couriertcpd to reject all connections from an IP address range:
netblock<tab>deny
netblock is an IP address, such as 192.68.0.2. <tab> is the ASCII tab character. There MUST be exactly one tab character after the IP address and the word "deny".
You can also block connections from an entire network C block:
192.68.0<tab>deny
This blocks connections from IP addresses 192.68.0.0 through 192.68.0.255. Blocking connections from an entire B or A network block works the same way.
Use the word "allow" instead of "deny" to explicitly allow connections from that IP address or netblock. For example:
192.68.0<tab>deny 192.68.0.10<tab>allow
This blocks all connections from 192.68.0.0 to 192.68.0.255 except for 192.68.0.10. These two lines can occur in any order. couriertcpd always uses the line with the most specific IP address.
If the IP address of the connection is not found in the access file the connection is accepted by default. The following line causes unlisted connections to be rejected:
*<tab>deny
The access file can also specify IPv6 addresses, if IPv6 support is available. The existing IPv4 address format is used for IPv6-mapped IPv4 addresses, and no changes are required. For all other IPv6 addresses use the following format:
:hhhh:hhhh:hhhh:hhhh:hhhh:hhhh:hhhh:hhhh<tab>action
The IPv6 address must begin with :. The initial : character is not really a part of the IPv6 address, it is only used to designate this record as an IPv6 address, allowing an access file to contain a mixture of IPv4 and IPv6 addresses. The IPv6 address follows the initial : character, and it must be spelled out using zero-padded lowercase hexadecimal digits. For example:
:0000:0000:0000:0000:0000:f643:00a2:9354<tab>deny
Netblocks must be specified using even-word boundaries only:
:0000:0000:0000:e240<tab>deny
This puts the entire ::e240/32 IPv6 CIDR into the access file.
allow can be optionally followed by a list of environment variable assignments, separated by commas. The environment variables are set before executing program or checking access lists (see below). For example:
192.68.0<tab>allow,RELAYCLIENT 192.68.0.10<tab>allow,RELAYCLIENT,SIZELIMIT=1000000
This sets RELAYCLIENT environment variable for connections from the 192.68.0 block. In addition to that, the SIZELIMIT environment variable is set to 1000000 if the connection comes from the IP address 192.68.0.10.
Note that RELAYCLIENT must be explicitly specified for the IP address 192.68.0.10. The first line is NOT used for connections from this IP address. couriertcpd only reads one entry from the access file, the entry for the most specific IP address.
An alternative to listing banned IP addresses is to use an external DNS-based IP access list.
There is no provision to support IPv6-based lists, because none yet exist. IPv6-based access list support will be added in the future.
couriertcpd does not automatically reject connections from any IP address listed on a DNS-based list. If the connecting IP address is listed couriertcpd simply sets an environment variable. It's up to the program, run by couriertcpd, to read the environment variable and choose what to do if the environment variable is set.
Please note that if the environment variable is already set, couriertcpd will NOT search the access list. This can be used to override the access list where program only recognizes the access list if the environment variable is not empty. By setting the environment variable to an empty string in the access file (see above), you can override access lists for selected IP addresses.
The -block option queries a DNS list for each connecting IP address. The only required argument to -block is the DNS zone that is used to publish thelist. The name of the zone can optionally be followed by a comma and the name of the environment variable to set if the DNS list includes the IP address. couriertcpd sets the environment variable BLOCK if you do not specify the name yourself.
The name of the environment variable can be optionally followed by a slash and an IP address. Normally couriertcpd sets the environment variable if the access list includes any A record entry for the specified IP address. Some access lists may offer additional information by returning one of several possible A records. If the name of the environment variable is followed by a slash and an IP address, the environment variable will be initialized only if the access list includes an A record containing the indicated IP address.
The contents of the environment variable will be the contents of any TXT record for the listed IP address. var[/n.n.n.n] can be optionally followed by a comma and a text message, which will be used instead of the TXT record. The text message may include a single @ character somewhere in it, which will be replaced by the listed IP address.
Multiple -block options can be used. The connecting IP address will be looked up in multiple access lists. This is implemented as follows.
couriertcpd processes all -block options one at a time. If the indicated environment variable is already set, couriertcpd skips the DNS list lookup (this is also true if only one -block option is specified). Therefore, if multiple -block options are used, and an IP address is found in the first access list, the remaininglists that use the same environment variable will not be checked. But other lists that use a different environment variable WILL be checked.
The same zone can be specified more than once, with different environment variables and different IP addresses. For example:
couriertcpd -block=block.example.org,BLOCK1/127.0.0.2 \
-block=block.example.org,BLOCK2/127.0.0.3
If the specified access list contains an A record for the listed address, and the A record contains the IP address 127.0.0.2, couriertcpd initializes the BLOCK1 environment variable. If the A record contains the IP address 127.0.0.3, couriertcpd initializes BLOCK2. If both records are present, both variables are initialized.
couriertcpd uses the following logic to determine what kind of DNS query to issue:
If neither the IP address, nor msg is specified, couriertcpd will query for existence of TXT records, for the IP address.
If only msg is specified, couriertcpd looks up the existence of A records, for the IP address.
If /n.n.n.n is used, and msg is not specified for at least one -block option for this same zone, couriertcpd will query for existence of ANY records, which should return both TXT and all the A records for this IP address.
If /n.n.n.n is used, and msg is specified for every -block option for this same zone, couriertcpd will query for existence of A records only.
couriertcpd also initializes the following environment variables prior to running program:
courier(8).