blocklistd is a daemon similar to
syslogd(8) that listens to sockets at paths specified in the
sockpathsfile for notifications from other daemons
about successful or failed connection attempts. If no such file is
specified, then it only listens to the socket path specified by
sockpath or if that is not specified to
/var/run/blocklistd.sock. Each notification contains
an (action, port, protocol, address, owner) tuple that identifies the remote
connection and the action. This tuple is consulted against entries from the
configfile, with the syntax specified in
blocklistd.conf(5). If an entry is matched, a state entry is
created for that tuple. Each entry contains a number of tries limit and a
duration.
If configfile is a directory, or a directory
exists with the same name as configfile with
".d" appended to it, each file in the directory will be read as
configuration file. If configfile exists as a file it
will be processed before the contents of the
configfile.d directory if that also exists.
The way blocklistd does configuration
entry matching is by having the client side pass the file descriptor
associated with the connection the client wants to blocklist as well as
passing socket credentials.
The file descriptor is used to retrieve information (address and
port) about the remote side with
getpeername(2) and the local side with
getsockname(2).
By examining the port of the local side,
blocklistd can determine if the client program
“owns” the port. By examining the optional address portion on
the local side, it can match interfaces. By examining the remote address, it
can match specific allow or deny rules.
Finally blocklistd can examine the socket
credentials to match the user in the configuration file.
While this works well for TCP sockets, it cannot be relied on for
unbound UDP sockets. It is also less meaningful when it comes to connections
using non-privileged ports. On the other hand, if we receive a request that
has a local endpoint indicating a UDP privileged port, we can presume that
the client was privileged to be able to acquire that port.
Once an entry is matched blocklistd can
perform various actions. If the action is “add” and the number
of tries limit is reached, then a control script
controlprog is invoked with arguments:
control add <rulename> <proto> <address> <mask> <port>
and should invoke a packet filter command to block the connection
specified by the arguments. The rulename argument can
be set from the command line (default blocklistd).
The script could print a numerical id to stdout as a handle for the rule
that can be used later to remove that connection, but that is not required
as all information to remove the rule is kept.
If the action is “rem” Then the same control script
is invoked as:
control rem <rulename> <proto> <address> <mask> <port> <id>
where id is the number returned from the
“add” action.
blocklistd maintains a database of known
connections in dbfile. On startup it reads entries
from that file, and updates its internal state.
blocklistd checks the list of active
entries every timeout seconds (default
15) and removes entries and block rules using the
control program as necessary.
The following options are available:
-C
controlprog
- Use controlprog to communicate with the packet
filter, instead of the default, which is
/usr/libexec/blocklistd-helper. The following
arguments are passed to the control program:
- action
- The action to perform:
add,
rem, or flush; to add,
remove or flush a firewall rule.
- name
- The rule name.
- protocol
- The optional protocol name (can be empty):
tcp, tcp6,
udp, udp6.
- address
- The IPv4 or IPv6 numeric address to be blocked or released.
- mask
- The numeric mask to be applied to the blocked or released address
- port
- The optional numeric port to be blocked (can be empty).
- id
- For packet filters that support removal of rules by rule identifier,
the identifier of the rule to be removed. The add command is expected
to return the rule identifier string to stdout.
-c
configuration
- The name of the configuration file to read. The default when
-c is not given is
/etc/blocklistd.conf.
-D
dbfile
- The Berkeley DB file where
blocklistd stores its
state. It defaults to /var/db/blocklistd.db.
-d
- Normally,
blocklistd disassociates itself from the
terminal unless the -d flag is specified, in which
case it stays in the foreground.
-f
- Truncate the state database and flush all the rules named
rulename are deleted by invoking the control script
as:
-P
sockpathsfile
- A file containing a list of pathnames, one per line that
blocklistd will create sockets to listen to. This
is useful for chrooted environments.
-R
rulename
- Specify the default rule name for the packet filter rules, usually
blocklistd.
-r
- Re-read the firewall rules from the internal database, then remove and
re-add them. This helps for packet filters that do not retain state across
reboots.
-s
sockpath
- Add sockpath to the list of Unix sockets
blocklistd listens to.
-t
timeout
- The interval in seconds
blocklistd polls the state
file to update the rules.
-v
- Cause
blocklistd to print diagnostic messages to
stdout instead of
syslogd(8).