This manual page documents briefly the compile definition files that are used
by the
netscript-compile(8)
command from the netscript router/firewall network configuration
package. This compiler creates a compiled iptables rules file in
/etc/netscript/ipfilter-defs.conf
(it is a shell script portion) that is sourced by the netscript
netscript(8)
command to configure the
iptables(8)
firewall rules in the kernel.
STARTUP COMPILATION
The rules can be compiled and automatically loaded on boot by setting the
IPV4_CONFIGURE_SWITCH switch in
network.conf(5)
to the value of the function used to configure the kernel.
Net-compile(8)
creates this function as
Configure.
If this switch is set, the netscript startup will run
netscript-compile(8)
to make sure everything is up to date and load the rules from
/etc/netscript/ipfilter-defs.conf,
and the relevant settings in
network.conf(5)
which are used to establish packet grooming and configure the built in
kernel netfilter INPUT and FORWARD chains in the filter table. If
compilation fails, the previous rule set is not replaced and it is used
instead. See the
netscript(8)
manpage to see how to load and use backup copies of the rule set.
CHAIN STRUCTURE
Each chain in the
iptables(8)
filter table is set up by a corresponding coonstruction function of the same
name as the chain. The chains are laced into the
iplcl
(which is laced in to the INPUT chain) and
ipfwd
(laced into FORWARD) chains respectively, and the forwarding control chains
are set up to take traffic in both directions, with the destination
network/interface and source network/interface being used in the lacing chain,
and network protocol and port being tied down in each specific chain.
For the new in kernel Linux IPSEC, traffic to and from the VPN can be
controlled via the iptables policy match module, if you have it patched and
compiled into your kernel and iptables. Future versions of the kernel and
iptables should have this included in the distributed source.
FILE STRUCTURE
All the files defining the rules set are in the
/etc/netscript/ipfilter-defs
directory. The
network-defs
file is used to define the regions and network blocks used in the rest of the
rules. The
prototypes-defs
file is used to define protoype rules that can be referenced elsewhere in
the rule set. The
prototypes.sh
file is used to construct shell functions for the
netscript-compile(8)
command that can be used in the definitions files. DNAT and SNAT are set up
in the
dnat-defs
and
masq-defs
files respectively. Any file ending in .def is taken as general rule set
input for
netscript-compile(8).
The files generally take the form of tables, with the columns tab or space
seperated. The '#' character is supported for commenting, and comments
can be on a line by themselves, or at the end of a configuration line.
Everything after the '#' is treated as a comment by the
netscript-compile(8)
compiler.
RULE STRUCTURE
The structure of the rule sets is thus. Each chain is started by calling
a shell compilation function, (generally ipv4_compile_chain) to create the
chain, with the chain name and source/destination regions as arguments, and
each rule in the chain by starting a fresh line with the chain name in the
first column.
Regions are defined as network interface tuples, and are set up in
network-defs.
They are syntactically the same as shell script variables, and are used
the same way in the
.def
rule set files. Technically this magic is achieved by using
eval
within the
netscript-compile(8)
shell script.
Any interface name can have either of the keywords
=clear
or
=ipsec
tied to them by using the `=' character on the end of the interface name.
This is used to specifically match IPSEC traffic, or non-IPSEC traffic
going over the interface. Typically you would use this when defining a region,
though the syntax is valid elsewhere as well. It is recommended that you use
this feature to prevent packet injection from adjacent external sources when
setting up iptables rules for VPN tunnel traffic.
The regions are given as arguments to the compilation function, with the region
always being 2 arguments in network/interface order to the function.
Each
chain rule
in the chain is defined by giving first of all the
chain name,
then the
rule type,
and its
direction.
All columns after the 3rd one are specific to and are defined by the
rule type.
The
direction
may have a '-' in it.
The rules produced by the compiler use the iptables connection based state
tracking. Packet by packet rules will be added later.
EXAMPLE
Here is an example of part of a .def file:
# Access from Office to internet
# - only allow outgoing tcp and UDP
# and ping traffic - anything else is most
# like a tunneling protocol.
# We have VPNs for tunneling
ipv4_compile_chain -p 90 offcInet droplog $OFFICE_REGN $INTERNET_REGN
offcInet ACCEPT_EST BOTH
offcInet ACCEPT_PING L2R
offcInet ACCEPT_TCP L2R 1:65535
offcInet ACCEPT_UDP L2R 1:65535
The ACCEPT_EST line accepts packets for ESTABLISHED and RELATED connections
to the new ones already accepted. New connections are accepted by the
ACCEPT_PING, ACCEPT_TCP, and ACCEPT_UDP rules. Please see the
iptables(8)
manpage for the details on stateful filtering.
COMPILE FUNCTIONS
Unless a function is defined in
prototypes.sh,
there is only one function provided. However this is not limiting as there is
a facility for rule macros, as well as the ability to tell the function to
use one of the default base rule sets.
If you do define a function in
prototypes.sh,
be careful to handle all errors to function and command calls as otherwise
netscript-compile(8)
will break, as it runs with set -e set.
You can see the source region and destination region on the end of it. The
default-target
is one of RETURN, DROP, droplog, or log.
The options to this function are as follows:
-i
Create an input chain for attaching to
iplcl
instead of the default forward chain for attaching to
ipfwd.
-n
Don't lace the chain into
iplcl
or
ipfwd.
-b base-chain
Specify an alternate ruleset chain to use.
-s slave-chain
Configure/deconfigure this chain as well as the one specified. Useful for
adjusting input rule set when manipulating the access chain for an IPsec
VPN.
-p priority
Specify the priority of the chain in the lacing rule set. Priority is between
00 and 99, with 00 at the top of the lacing chain, and 99 at the bottom. This
is useful for making sure that host specific rule sets occur before more
general network related ones, and for putting Internet related ones at the
bottom of the lacing chain.
DIRECTION STATEMENTS
The direction is as per FreeS/WAN - it uses left and right terminology.
Here are the valid chain rules, and the arguments they expect.
COMMENT [word1] [word2] ...
Insert a comment into the compile shell script. Fill the 3rd column direction
in with '-'.
MACRO <macro-name>
Specify a macro rule set. Rule set must name start with `MACRO_'. Direction
again should be `-'.
LOG [word1] [word2] ...
Insert a logging rule using the given log meesage, or if none given, using
the curretlog message for the chain.
LOG_MSG [word1] [word2] ...
Set the log message for the chain away from the default
of `Chain: <chain-name>' or from previous LOG_MSG setting. Up to 26 letters
can be used until truncation limit is reached.
RESET_LOG_MSG
Reset log message to the default of `Chain: <chain-name>'.
REJECT_SMB
Jump to smb control chain. Creates smb chain if it does not already exist.
DROP_MARTIANS
Jump to martian source address control chain. Creates chain if it does not
already exist.
LOG_PORTSCAN
Use the psd module to detect and log portscans. Creates portscan log chain
(if not already there) which puts `PORTSCAN DETECTED - ' in the log.
DROP_BROADCAST
Drop ethernet broadcast packets.
LOG_BROADCAST
Log ethernet broadcast packets with the current log messages for the chain.
ACCEPT_EST
Accept ESTABLISH,RELATED packets via the
iptables(8)
state module.
ACCEPT_RELATED
Accept RELATED packets via the
iptables(8)
state module. Useful for ICMP type 3 packets used for maximum MTU detection.
ACCEPT_PROTO <protocol>
Accept NEW connections for a protocol. Accepts one argument in the 4th column which is the protocol name from
/etc/protocols
or the protocol number between 0 and 255.
REJECT_PROTO <protocol>
Reject NEW connections for a protocol with ICMP reject packets. Accepts one
argument in the 4th column which is the protocol name from
/etc/protocols
or the protocol number between 0 and 255.
DROP_PROTO <protocol>
Drop all packets for a protocol with nothing in reply. Accepts one
argument in the 4th column which is the protocol name from
/etc/protocols
or the protocol number between 0 and 255.
LOG_PROTO <protocol>
Log NEW connections for a protocol with the current log message for the chain.
Accepts one argument in the 4th column which is the protocol name from
/etc/protocols
or the protocol number between 0 and 255.
ACCEPT_TCP [src-port-range] <dst-port-range>
Accept NEW TCP connections. If one argument given, it is the destinaion port
(range). If 2 arguments, the first is the source port (range), and second the
destination port (range). Port ranges are specified by separating them with
a `:' character, and ports must be in the
/etc/services
file, or a number between 0 and 65535.
REJECT_TCP [src-port-range] <dst-port-range>
Reject NEW TCP connections with an ICMP REJECT packet. If one argument given,
it is the destination port(range). If 2 arguments, the first is the source
port (range), and second the destination port (range). Port ranges are
specified by separating them with a `:' character, and ports must be in
the
/etc/services
file, or a number between 0 and 65535.
DROP_TCP [src-port-range] <dst-port-range>
Drop all tcp packets, returning nothing at all. If one argument given,
it is the destinaion port (range). If 2 arguments, the first is the source
port (range), and second the destination port (range). Port ranges are
specified by separating them with a `:' character, and ports must be in
the
/etc/services
file, or a number between 0 and 65535.
LOG_TCP [src-port-range] <dst-port-range>
Log NEW TCP connections with the current log text for the chain.
If one argument given, it is the destination port(range). If 2 arguments,
the first is the source port (range), and second the destination port (range). Port ranges are specified by separating them with a `:' character,
and ports must be in the
/etc/services
file, or a number between 0 and 65535.
ACCEPT_UDP [src-port-range] <dst-port-range>
Accept NEW UDP connections. If one argument given, it is the destinaion port
(range). If 2 arguments, the first is the source port (range), and second the
destination port (range). Port ranges are specified by separating them with
a `:' character, and ports must be in the
/etc/services
file, or a number between 0 and 65535.
REJECT_UDP [src-port-range] <dst-port-range>
Reject NEW UDP connections with an ICMP REJECT packet. If one argument given,
it is the destination port(range). If 2 arguments, the first is the source
port (range), and second the destination port (range). Port ranges are
specified by separating them with a `:' character, and ports must be in
the
/etc/services
file, or a number between 0 and 65535.
DROP_UDP [src-port-range] <dst-port-range>
DROP all UDP packets, returning nothing at all. If one argument given,
it is the destinaion port (range). If 2 arguments, the first is the source
port (range), and second the destination port (range). Port ranges are
specified by separating them with a `:' character, and ports must be in
the
/etc/services
file, or a number between 0 and 65535.
LOG_UDP [src-port-range] <dst-port-range>
Log NEW UDP connections with the current log message for the chain. If one
argument given, it is the destination port(range). If 2 arguments, the first
is the source port (range), and second the destination port (range). Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
ACCEPT_PING
Accept ICMP type 8 echo request packets for network diagnosis.
DROP_PING
Drop ICMP type 8 packets with no reply.
LOG_PING
Log an ICMP echo request with the current log message for the chain.
Accept NEW TCP connections from given source (optional) to destination.
Network
is given in IPv4 address/netmask or address/masklen format. Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
Reject NEW TCP conections with an ICMP reject packet which come from a given
source (optional), going to given destination.
Network
is given in IPv4 address/netmask or address/masklen format. Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
Drop all TCP packets which come from a given source (optional), going to given
destination.
Network
is given in IPv4 address/netmask or address/masklen format. Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
Log all NEW TCP connections from given source (optional) to destination, with
the current log message for the chain.
Network
is given in IPv4 address/netmask or address/masklen format. Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
Accept NEW UDP connections from given source (optional) to destination.
Network
is given in IPv4 address/netmask or address/masklen format. Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
Reject NEW UDP conections with an ICMP reject packet which come from a given
source (optional), going to given destination.
Network
is given in IPv4 address/netmask or address/masklen format. Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
Drop all UDP packets which come from a given source (optional), going to given
destination.
Network
is given in IPv4 address/netmask or address/masklen format. Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
Log all NEW UDP connections from given source (optional) to destination, with
the current log message for the chain.
Network
is given in IPv4 address/netmask or address/masklen format. Port
ranges are specified by separating them with a `:' character, and ports
must be in the
/etc/services
file, or a number between 0 and 65535.
ACCEPT_IFACE <interface>
Accept all incoming NEW connections from an incoming interface.
REJECT_IFACE <interface>
Reject all incoming NEW conections with an ICMP reject packet, from an
interface.
DROP_IFACE <interface>
Drop all incoming packets from an interface.
LOG_IFACE <interface>
Log all incoming NEW conections from an interface.
ACCEPT_NET <network>
Accept all NEW connections from
network.Network
is given in IPv4 address/netmask or address/masklen format.
REJECT_NET <network>
Reject all NEW conections from
network
with an ICMP reject packet.
Network
is given in IPv4 address/netmask or address/masklen format.
DROP_NET <network>
Drop all packets from
network.Network
is given in IPv4 address/netmask or address/masklen format.
LOG_NET <network>
Log all NEW conections from
network.Network
is given in IPv4 address/netmask or address/masklen format.