fixed/improved logging

[rxpd] / README

RXPD Documentation

Overview

The RegexPolicyDaemon (rxpd) can be used to efficiently check data against
different lists of regular expressions. This can be used to build whitelists/
blacklists to protect many kinds of Internet services. It uses a simple textual
protocol that is easily implementable in scripting languages. Example usages
are access and content control (spam filtering) for CGI scripts, wikis, email,
revision control systems, IRC servers and clients, and so on.

Rxpd encourages users to distribute their lists in a friend2friend network. It
has features to fetch, updaten, filter and merge lists. The idea is that users/
administrators maintain manageable lists which cover single topics and then
merge them together.

Building and Installing

Release Tarballs

Release tarballs are attached to the wiki at:

  * http://www.pipapo.org/pipawiki/RegexPolicyDaemon?action=AttachFile

I am using gpg signed tarballs for distribution. As first step one has to check
the signature

$ gpg rxpd-X.Y.tar.gz.gpg

This will produce a rxpd-X.Y.tar.gz and report if the signature could be
validated.

Since the package is built with gnu autotools, the usual build and install
procedure works:

$ tar xzvf rxpd-X.Y.tar.gz
$ cd rxpd-X.Y
$ mkdir build     # using a build directory is optional
& cd build
$ ../configure
$ make
$ make install

Development Version via git

The development version is available via git from 'git://git.pipapo.org/rxpd'
or mirrored at repo.or.cz 'git://repo.or.cz/rxpd.git'.

After you cloned the repository you need to bootstrap the autotools first

$ autoreconf -i

Then the usual configure / make will work.

There is a special makefile target make meta to bring several files (README,
AUTHORS, NEWS, TODO) in sync with the Rxpd Documentation wiki and update the
ChangeLog.

Dependencies

Rxpd requires gnu-pth and its development headers.

What gets installed

A single executable called 'rxpd' will be installed in $prefix/bin.

Concepts

Rxpd targets to be simple and efficently validating data against Regular
expressions. It has (yet) no configuration file for the daemon itself and is
controlled by commandline options. Most management of regular expression lists
can be done remotely over a simple protocol. By itself it has has no
authentication but there is a 'policy' check which validates incoming requests
against an special regex list which then defines if the client is allowed to do
a certain task. Any further management like distributing the lists,
authenticate sessions more strongly and so on should be done by other means and
are not planned to be included in rxpd.

The goal it to create a common place which applications can use to validate any
kind of data. This works efficently because short lived programs like CGI
scripts take the advantage of regular expressions which are precompiled in core
and generally such lists might be shared between different applications.

Commandline Options and Starting

<!> WIP

       Increase verbosity level. Verbosity levels correspondend to syslog
-v     levels, where LOG_WARNING is the default. Thus -v can be given up to
       three times to get LOG_DEBUG.
-V     Just shows version information and then exit.
-d     Detach from terminal and run in background.
-D     Run in debug mode, that is, increasing verbosity to at lest LOG_INFO and
       don't detach. An additional -v can be used for LOG_DEBUG.
-b dir Give the basedir for rules. Rules have to be in a single directory, rxpd
       will never access any data outside of this directory.
-q     Turn rxpd quiet, only LOG_ALERT or worse will be logged.
-t     Give a port number for a tcp port to listen on. This option can be
port   appear multiple times. Only root can listen on ports less than 1024.
-u     Path for a unix local socket to listen on. This option can appear
name   multiple times.
-p     Define which rules-list will be be used to limit access to the rxpd
policy itself. If not given, no access restrictions apply (everything
       allowed!). Policy matching will be descriped in detail later.
-i     Turn all regular expressions case insensitive.
-4     Use only IPv4, default is IPv4 and IPv6 when compiled in.
-6     Use only IPv6.
-h     Give a short usage notice and then exit.
-U     When started as root, switch to 'user', if not given, 'nobody' is used.
user

List Syntax

There are only 2 things allowed in a list file:

  * Comments
      + Begining with a '#' at the first column followed by arbitary text.
        Comments are preserved and have semantic meaning as they can be used to
        organize the data. Comments in the form '#UPPERCASE: ' are special/
        reserved, the engine uses them to disable rules when they expire or
        flag errorneous rules, '#lowercase: ' can be used for custom enabling/
        disableing of rules, see the FILTER command.
  * Rules
      + Starting with an optional accesstime entry, then a name, followed by a
        regex. This three parts are delimited by colons.
      + 'atime' will be maintained by the daemon to reflect the last time the
        rule matched some data. This is time in seconds since epoch in UTC.
      + 'name' is an arbitary string which has not special meaning for the rxpd
        but will send back to the calling applications and be used there to
        classify results.
      + the regex is a POSIX extended regular expression, regex are currently
        case-insensitive this will become configureable later.

Lines in can be at most 4095 bytes long.

Example list file:

# Free things are good!
:accept:GNU|Linux
0:accept:FreeBSD
# Bad things
0:reject:M.*soft

Matches will later report the line matched, without the atime and first colon
part. "Macrosoft" matches "M.*soft" thus "reject:M.*soft" will be returned.

Note that the first 'accept' rule has no atime, to initiate atimes they can be
initalized with '0' the daemon will update them on access and rewrite the List
files with the 'SAVE' command or when it recieves a SIGTERM.

When there is an error in a regular expression, it will be replaced with #
ERROR:, followed by the cause of the error, followed by the rules string in
quote.

Protocol

Rxpd uses a simple line based text protocol. The first line is always the
command and list which will be used on the following data, it is not possible
to change the command throughout a session. Each session will generate at least
one line of response. When no other output is available '#OK:' is send, in case
of an error a line starting with '#ERROR:' is send.

Lines end with any combination of the 'newline' and/or 'carriage return'
character.

The protocol is line based where lines which are longer than 4095 characters
are broken (may be word-wraped on the last whitespace character in the line in
future).

Commands:

  * 'CHECK:list\n..data..'
      + check all following data against the list. Returns the first matching
        rule (excluding the 'atime' field), if any. When a empty line is send,
        the daemon answers with "#OK:". This can be used to syncronize the
        queries before sending new data.
  * 'APPEND:list\n..rules..'
      + append the following lines to list.
  * 'PREPEND:list\n..rules..'
      + prepend the following lines to list.
  * 'REMOVE:list\n..rules..'
      + remove all matching lines from list.
  * 'REPLACE:list\nrule\n..replacements..'
      + find the position matching the first line, which can be a rule or a
        comment and replaces it with the following rules. Updates are atomic
        and done when either an empty line is send or when the connection gets
        closed.
  * 'LOAD:list\n'
      + reload list from disk, this resets the 'atime' to the values stored on
        disk.
  * 'SAVE:list\n'
      + save list to disk, saves new atime records.
  * 'EXPIRE:list\nseconds'
      + marks all rules from list which are subject of atime updates and where
        not touched for some (much) seconds with a '#EXPIRED:' comment,
        effectively disabling them.
  * 'FILTER:list\nfilters...'
      + runs 'filters..' (also rules lists) over list and takes action for
        matches. There are 2 Actions defined:
          o 'DELETE' removes the matching rule completely from the list
          o 'ACTIVATE' tries to reactivate a previously commented out rule (#
            something: comment)
          o any other name will just comment out the matching rule using the
            name itself
  * 'FETCH:list\nremote'
      + fetches a list from remote storing it in local 'list'. 'remote' has the
        form address/listname where address is either 'ip:port' or a path to a
        unix domain socket.

        Idea: do we want 'FETCH:list\
        nremote:policylist' which gives a local list filtering remote first?
  * 'UPDATE:list\nsource...'
      + updates atimes in 'list' from sources, uses an efficent forward looking
        algorithm, rule reordering in sources is not supported (adding/removing
        rules works). sources have to be other local lists.
  * 'MERGE:list\nsources..'
      + Adds new lines from other 'sources..' to 'list', order will be
        preserved. Does NOT delete removed lines.
  * 'DUMP:list\n'
      + dump the content of list.
  * 'LIST:\n'
      + list all loaded lists.
  * 'SHUTDOWN:\n'
      + exits the daemon gracefully, pending connections will still be served
        but no new connections are accepted.
  * 'VERSION:\n'
      + prints package and version information.
  * 'HELP:\n'
      + gives a short list of available commands.

Signals

Rxpd acts on following signals:

  * SIGHUP:
      + Reload all files from disk
  * SIGTERM:
      + Save all files and exit
  * SIGINT:
      + Exit immediately without saving
  * SIGALRM:
      + Save all files while continue running

Using the rxpd

<!> WIP

Namespaces

Rxpd is intended to be used as distributed system, this makes clear that data
put into it has to follow some rules to avoid clashing with foreign data. While
rxpd doesn't enforce any naming rules I write some suggestions/proposals here.

Rule lists in rxpd are files, with 0.3 they are organized as directory hierachy
under the rules-basedir.

Most importantly we need to distinguish the authority for a rules list, this
can be either a specifc user or some server. For users we use the users email
as unique identifier, for servers their hostname will suffice. Further we need
to address the purpose or protocol for which a list is primary used and finally
the list shall have a descriptive name.

This triple forms a uniqe identifier for any list.

Access Policies

One list of rules can be used to define access policies for the rxpd itself (-p
option). Each command will be extended with access protocol (one of tcp4, tcp6
or unix) and the peer address and then checked against this policy list. When
this check yields in an 'ACCEPT:..' rule, the command is allowed, for
everything else will result in an error and drop the connection.

For example if '-p policy' is used:

# Syntax:
# [atime]:rulename:command:list:proto:address
#
# Allow dumping of the 'policy' list itself
:ACCEPT:DUMP:policy
# Clients from local network are allowed to do anything
:ACCEPT:.*:tcp.:10\..*$
# Forbid all others to do anything else with the policy
:REJECT:.*:policy
# Finally allow anything else
:ACCEPT:.*

Example

We want to protect a wiki or such against vandalism: blacklists are in
$blacklists.d/ lets say /etc/blacklists.d/

The wiki engine builds a tuple hostname;ip which is checked against a blacklist
which classify the 'hosts'

this is /etc/blacklist.d/hosts

# local access are always trusted, thats localhost any my local network
:allow:localhost;127.0.0.1
:allow:mydomain.org;10.10.
# some really bad guys are put on a blacklist which never ever shall get access
:deny:.*aol.com;
# everyone else shall just get the content checked
:check:

so printf("CHECK:hosts\n%s;%s\n", hostname, ipaddr) send to the blacklist
daemon will result in either 'allow', 'deny' or 'check' send back. The first
both (allow/deny) results are handled obliviously. With the 'check' result the
edited content will be filtered against another list '/etc/blacklists.d/
content'

#example .. see BadContent on this wiki
:deny:sex.com
:deny:warez

Demonstation

<!> WIP <!> Not always running

Note: there is an almost unrestricted rxpd running here as demo:

$ echo -e 'LIST:' | nc www.pipapo.org 2345
example
policy

$ echo -e 'DUMP:policy' | nc www.pipapo.org 2345
# syntax:
# rule:command:list:proto:address
#
# Allow dumping of the 'policy' list itself
:ACCEPT:DUMP:policy
# Clients from local network are allowed to do anything
:ACCEPT:.*:tcp.:10\..*$
# Forbid all others to do anything else with the policy
:REJECT:.*:policy
# Hey and no one shall remotely shutdown this daemon!
:REJECT:SHUTDOWN
# Finally allow anything else
:ACCEPT:.*

Also checkout the 'HELP:' and 'VERSION:' command. See this documentation for
more details.

This Documentation is maintained at:

  * http://www.pipapo.org/pipawiki/RegexPolicyDaemon/Documentation

RegexPolicyDaemon/Documentation (last edited 2007-11-02 02:28:10 by ct)