Table of Contents
- Primary Commands
- interface, control what the firewall host can do
- router, control what traffic can pass through the firewall host
- Sub-Commands
- Helper commands
- action, setup custom filtering actions
- blacklist, setup a unidirectional or bidirectional blacklist
- classify, classify traffic for traffic shapping tools
- connmark, set a statefull mark on a connection
- dnat, setup a Destination NAT rule for routed traffic
- dscp, set the DSCP field in the packet header, to a raw value or a DiffServ class
- ecn_shame, disable ECN for all hosts in the ECN Shame list
- iptables, add some custom iptables commands to the firewall
- mac, setup a source mac address with IP match
- mark, mark traffic for traffic shapping tools
- masquerade, setup masquerading (NAT) to the output of a network interface
- nat, setup a NAT rule for routed traffic
- redirect, setup a port redirection rule
- snat, setup a Source NAT rule for routed traffic
- tcpmss, set the MSS of TCP SYN packets for routers
- tos, set the Type of Service (TOS) field in the packet header
- tosfix, fix the TOS value of packets
- transparent_proxy, setup a transparent TCP proxy running on the firewall host
- transparent_squid, setup a transparent web proxy running on the firewall host
- version, require a specific version of FireHOL
- Actions
- accept, allow the traffic to reach its destination
- reject, don't allow the traffic to pass, but send a rejection message to the sender
- drop, discard the traffic without sending anything back to the sender
- deny, an alias for drop
- tarpit, tarpit TCP traffic making the remote client to timeout after several minutes
- return, return to processing the parent flow of rules
- mirror, send the traffic back to the sender at the port of the destination
- redirect, redirect the traffic to another port on the local host
- Optional Rule Parameters
- src, match the source of traffic
- dst, match the destination of traffic
- srctype, match the source address type of traffic
- dsttype, match the destination address type of traffic
- inface, match the network interface traffic is received via
- outface, match the network interface traffic is send via
- physin, match the physical network interface (for bridges) traffic is received via
- physout, match the physical network interface (for bridges) traffic is send via
- custom, pass a few custom parameters to the generated iptables statements
- log, write something to the syslog when traffic is matched
- loglimit, write (limited) something to the syslog when traffic is matched
- proto, match a specific protocol
- limit, limit the frequency traffic is matched
- sport, match the source ports
- dport, match the destination ports
- uid, user, match the users sending this traffic
- gid, group, match the user groups sending this traffic
- pid, process, match the process IDs sending this traffic
- sid, session, match the process session IDs sending this traffic
- cmd, command, match the command name sending this traffic
- mac, match the source MAC address of packets.
- mark, match the MARK ID of packets.
- tos, match the Type of Service (TOS) of packets.
- dscp, match the DSCP raw value or DiffServ class value of packets.
- Variables that control FireHOL
- Variables FireHOL offers
Primary commands form groups of rules within a FireHOL firewall.
The optional rule parameters given to the primary commands are indirectly applied to all sub-commands given within this primary command.
interface <real interface> <name> [optional rule parameters]
Description
The interface command creates a firewall for protecting the host the firewall is running, from the given interface.
The default interface policy is drop, so that if no subcommands are given, the
firewall will just drop all incoming and outgoing traffic using this interface.
INPORTANT
Note that unlike ipchains, in iptables traffic passing through the firewall
host (FORWARDed traffic) does not pass through the INPUT/OUTPUT chains of the firewall and
therefore the interface rules in FireHOL never match it.
To match the traffic passing through (even if DNATed) you have to place
the filtering rules in router statements. Interface
statements filter traffic only from/to the firewall host. Nothing else.
Parameters
- real interface is the interface name as shown by ip link show.
Generally anything iptables accepts, including the pattern character + (the plus sign), is valid.
The plus sign after some text will match all interfaces that start with this text.
It is allowed to use more than one interfaces separated by spaces, but all of them should be
given within one quoted argument:
Example: interface "eth0 eth1 ppp0" myname
- name is a name for this interface.
Generally you should use short names (10 characters max) without spaces or other symbols.
You should not use the same name more than once in FireHOL primary commands.
- optional rule parameters is a set of rules that allow further restriction of the traffic that
gets matched for this interface. See Optional Rules Parameters for more information.
Example 1: interface eth0 intranet src 10.0.0.0/16
Example 2: interface eth0 internet src not "$UNROUTABLE_IPS"
(note: UNROUTABLE_IPS is a variable defined by FireHOL that includes all IPs that should not be routable
by the Internet).
router <name> [optional rule parameters]
Description
The router command creates a firewall for the traffic passing through the host running the firewall.
The default policy on router commands is return. This means that by default no packets are dropped in a router.
Packets not matched by any router command will be dropped at the end of the firewall.
Change the default policy of a router only if you understand clearly what gets matched by the router statement. It is very common
to have overlapping definitions of routers and changing this policy to anything other than the default may have strange results
for your configuration. (Changing the policy on routers appeared in v1.248).
INPORTANT
Note that unlike ipchains, in iptables traffic passing through the firewall
host (FORWARDed traffic) does not pass through the INPUT/OUTPUT chains of the firewall and
therefore the interface rules in FireHOL never match it.
To match the traffic passing through (even if DNATed) you have to place
the filtering rules in router statements. Interface
statements filter traffic only from/to the firewall host.
Router statements filter traffic that passes through the
firewall host.
Parameters
- name is a name for this router.
The same restrictions of interface names apply here too.
- optional rule parameters is a set of rules that allow further restriction of the traffic that
gets matched for this router. See Optional Rules Parameters for more information.
Router statements produce similar iptables commands the interface statements produce.
For each router statement an in_<name> and an
out_<name> chain are produced to match the traffic
in both directions of the router.
To match some client or server traffic the administrator has to specify the input/output
interface or the source/destination of the request.
All inface/outface, src/dst optional rule parameters can be given either on the router
statement in which case will be applied to all subcommands for this router,
or on each subcommand within a router.
Both are valid.
For example:
router mylan inface ppp+ outface eth0
server http accept
client smtp accept
The above says: Define a router that matches all requests that originate from some PPP interface and go out to eth0.
There is an HTTP server in eth0 that client from the PPP interfaces are allowed to reach.
Clients on eth0 are allowed to get SMTP traffic from the PPP interfaces.
While:
router mylan
server http accept inface ppp+ outface eth0
server smtp accept inface eth0 outface ppp+
The above says: Define a router that matches any kind of forwarded traffic.
For HTTP traffic the clients are on a PPP interface and the servers on eth0.
For SMTP traffic the clients are on a eth0 interface and the servers o a PPP interface.
Please note that in the second example the SMTP traffic is matched again with a server subcommand,
not a client (as in the first example).
The client subcommand reverses all the optional rules that are applied indirectly to it.
Indirect rule parameters are those that are inherited from the parent command (router in this case).
To make it simple, for FireHOL a client is: "a server with all the implicit optional rule parameters reversed".
So, in the first example, the client simply flipped the inface and outface rules defined at the router
and became an SMTP server.
In the second example there is nothing to be flipped, so server and client
are exactly the same.
I suggest to use client subcommands in routers only if you have inface/outface or src/dst in the router statements.
If you are building routers like the second example, don't use client, it is confusing.
Older versions of FireHOL did not allow server and client subcommands in routers.
Only the route subcommand was allowed. Today, route is just an alias for server and can be
used only in routers, not interfaces.
Any number of router statements can exist. Since the policy is RETURN on all of them, any traffic
not matched by a router will continue to be checked against the second.
Subcommands must be given within Primary commands.
policy <action>
Description
The policy subcommand defines the default policy for an interface.
This directive accepts all the actions specified in Actions.
Please note that FireHOL is only useful for "DROP/REJECT EVERYTHING, ALLOW EXPLICITLY" type of firewalls.
By changing the policy to accept, you CANNOT create valid "ACCEPT EVERYTHING, DROP EXPLICITLY" firewalls.
The policy should be set to accept only if you really trust the hosts that can reach the machine via this interface.
For information about this, see: Bug 927532.
protection [reverse] <type> [requests/sec [burst]]
Description
The protection subcommand sets a number of protection rules on an interface.
In router configurations, protections are setup on inface. The reverse keyword will make the
protections setup on outface. Please note that the reverse keyword must be the first given.
Otherwise it will not work.
type can be:
strong, or full, or all
turns on all the possible protections.
Currently it includes: invalid fragments new-tcp-w/o-syn icmp-floods syn-floods malformed-xmas malformed-null malformed-bad.
bad-packets
turns on all the possible protections that detect bad packets based one the packet contents themselves, not the packets transfer rate.
Currently it includes: invalid fragments new-tcp-w/o-syn malformed-xmas malformed-null malformed-bad.
invalid
Drops all incoming INVALID packets. Invalid packets are those that are matched by "-m state --state INVALID", so they are all those
packets that the iptables connection tracker believes are INVALID.
There is also the FIREHOL_DROP_INVALID variable which controls the same function globally, for the
whole firewall.
fragments
Drops all packet fragments. Please note that most probably this rule will never match anything since
iptables reconstructs all packets automatically, before the iptables firewall rules are processed, when its connection tracker is running.
I left this here for psychological reasons and in case some one finds some use for it.
new-tcp-w/o-syn
Drops all TCP packets that initiate a socket but have no the SYN bit set.
syn-floods
Allows only a certain amount of new TCP connections per second. The optional two arguments [requests/sec] and [burst]
are used by this rule in order to provide control on the number of connections to be allowed.
The default is 100 connections per second that can match 50 (it was 4 in v1.38 and before) packets initially
(this is implemented using the limit module of iptables: see man iptables for more).
Note that this rule applies to all connections attempted regardless of their final result (rejected, dropped, established, etc).
Therefore it might not be a good idea to set it too low.
icmp-floods
Allows only a certain amount of ICMP echo requests per second. The optional two arguments [requests/sec] and [burst]
are used by this rule in order to provide control on the number of connections to be allowed.
The default is 100 connections per second that can match 50 (it was 4 in v1.38 and before) packets initially
(this is implemented using the limit module of iptables: see man iptables for more).
all-floods
Allows only a certain amount of new connections per second. The optional two arguments [requests/sec] and [burst]
are used by this rule in order to provide control on the number of connections to be allowed.
The default is 100 connections per second that can match 50 (it was 4 in v1.38 and before) packets initially
(this is implemented using the limit module of iptables: see man iptables for more).
Note that this rule applies to all connections attempted regardless of their final result (rejected, dropped, established, etc).
Therefore it might not be a good idea to set it too low.
malformed-xmas
Drops all TCP packets that have all the TCP flags set.
malformed-null
Drops all TCP packets that have all the TCP flags unset.
malformed-bad
Drops all TCP packets that have illegal combinations of TCP flags set.
server <service> <action> [optional rule parameters]
Description
The server subcommand defines a server of a service. For FireHOL a server is the destination of a request,
and even if this is more complex for multi-socket services, for FireHOL a server always accepts requests.
The optional rule parameters given to the parent primary command (interface or router) are inherited by the server as they have been given.
This subcommand can be used on both interfaces and routers.
Parameters
client <service> <action> [optional rule parameters]
The client subcommand defines a client of a service. For FireHOL a client is the source of a request.
FireHOL follows this simple rule even on multi-socket complex protocols, so that for FireHOL a client always sends requests.
The parameters are exactly the same with the server subcommand.
The optional rule parameters given to the parent primary command (interface or router) are inherited by the client, but they are reversed.
For an explanation of this please refer to the documentation of the router primary command.
This subcommand can be used on both interfaces and routers.
route <service> <action> [optional rule parameters]
The route subcommand is an alias for the server command that can be used only on routers, not interfaces.
group with [optional rule parameters]
group end
The group subcommand allows grouping for server and client
subcommands that share common optional rule parameters.
There is only one reason for using groups: Optimization of the generated firewall rules.
With groups, all the optional rule parameters are checked only once, while entering the
group, and then each service is applied without extra processing.
For example:
interface any world
server http accept
# trusted services
group with src "1.2.3.4 5.6.7.8"
server ssh accept
server telnet accept
group end
client all accept
|
The above will be optimal compared with server "ssh telnet" accept src "1.2.3.4 5.6.7.8".
FireHOL supports nested groups.
Helper commands provide shortcuts for common functions not handled by the core of FireHOL's logic.
action [chain <name> <action>
The action helper creates an iptables chain which can be used to control the action of other firewall rules during runtime.
For example, you can setup the custom action ACT1, which by default is ACCEPT, but under certain cases it can be changed to DROP, REJECT or RETURN without
restarting the firewall.
The first argument must always be the word chain, for the moment.
- name can be any chain name accepted by iptables. It is suggested to keep it between 5 to 10 letters.
- action can be any action supported by FireHOL, although only ACCEPT, REJECT, DROP, RETURN may
have any meaning under this use.
Example 1:
At the top of firehol.conf, create the action ACT1:
action chain ACT1 accept
later, in interfaces and routers, create rules that use the ACT1 action:
server smtp ACT1
client imap ACT1
Please note that actions created this way are case sensitive.
At some point, and while the firewall is running, the action ACT1 can be changed to DROP, with this linux command (this is not FireHOL specific):
iptables -t filter -I ACT1 -j DROP
The above command inserts (-I) the new action DROP above the default action ACCEPT, and therefore all the traffic matching the FireHOL rules that have
the action ACT1 will now be dropped.
To return to the default action (ACCEPT), run the following linux command:
iptables -t filter -D ACT1 -j DROP
This command deletes (-D) the DROP action that was inserted above the default action. If you delete all actions in the chain ACT1, the default action will be RETURN,
in which case all rules with action ACT1 will be neutralized (it will be the same as they were not specified at all in firehol.conf).
Example 2: action chain "ACT1 ACT2 ACT3" accept chain "ACT4 ACT5 ACT6" drop # will create 6 actions, ACT1, ACT2, ACT3 with ACCEPT and ACT4, ACT5, ACT6 with DROP
blacklist [option] <IPs>
The blacklist helper creates a blacklist for the IP addresses given. It supports two modes of operation
based on the option given (or the absence of it).
The option can be:
- one of the words them, him, her, it, this, these, input in which case it will generate a unidirectional
statefull blacklist, meaning that you will be able to ask (initiate connections) anything from them,
but they will not be able to ask (initiate connections) anything from you or the remote hosts you protect (routing).
- one of the words all, full or ommitted (no option given), in which case FireHOL will create bidirectional
stateless rules that will DROP all traffic comming in from these IPs and will REJECT all traffic going to them.
The blacklist helper affects both interfaces and routers and can be used as many times as needed, but before the first interface statement.
The blacklist helper accepts multiple IPs both as one quoted and space separated list and as separate arguments.
Example 1: blacklist this "195.97.5.202 10.1.1.1" # please note the quotes
Example 2: blacklist full 195.97.5.202 10.1.1.1 # please note the absence of quotes
classify <CLASS> [optional rule parameters]
The classify helper classifies the traffic into a specific traffic shapping class.
Parameters
- CLASS is a class, as accepted by iptables (i.e. MAJOR:MINOR).
- optional rule parameters is a set of rules that allow further restriction of the traffic that
gets matched by this rule. See Optional Rules Parameters for more information.
Example 1: classify 1:1 outface eth1 proto tcp dport 25, will send all smtp traffic going out via eth1 to class 1:1.
connmark <VALUE> <WHERE> [optional rule parameters]
The connmark helper is used to set a mark on a whole connection in both directions of the traffic automatically. The difference with mark is that the later just marks the packets matched by the optional rule parameters, while connmark marks the whole statefull connection.
Parameters
- VALUE can be a mark value to set, or the words save to convert the existing mark of the packets to connmark, or restore to convert the existing connmark of the connection to a mark on the packets.
- WHERE tells FireHOL where to search for the specific traffic to be marked.
Currently, WHERE can be one of the build-in iptables chains attached to table mangle.
(for example: INPUT, FORWARD, OUTPUT, PREROUTING, POSTROUTING - case does matter here).
- optional rule parameters is a set of rules that allow further restriction of the traffic that
gets matched by this rule. See Optional Rules Parameters for more information.
Example 1: Let's say that you have 3 ethernet interfaces: eth0, eth1 and eth2. eth0 connects you to the LAN, while eth1 and eth2 connect you to the internet via 2 different service providers.
You want to guarrantee that when a connection is coming from provider 1 (eth1) will go back through this provider and the same for provider 2 (eth2).
Step 1: You mark the connections when they enter the firewall host from the provider interfaces:
connmark 1 PREROUTING inface eth1
connmark 2 PREROUTING inface eth2
Step 2: You restore the mark (from the connmark) when the packets come in from the LAN:
connmark restore OUTPUT
connmark restore PREROUTING inface eth0
Now you can use iproute2 to pick the right routing table (eth1 or eth2 and the default gateway) based on the MARK of the packets.
Example 2: connmark 1 PREROUTING inface eth1, Set the CONNMARK 1 to all packets coming in from eth1.
Example 3: connmark save PREROUTING inface eth1, Save the MARK value to the CONNMARK value on all packets coming in from eth1.
dnat [to] <target> [optional rule parameters]
The dnat helper sets up a Destination NAT rule for routed traffic, by calling
nat to-destination <target> [optional rule parameters]
See the nat helper.
Example: dnat to 1.1.1.1 inface eth0 src 2.2.2.2 dst 3.3.3.3
dscp <NUMBER> <WHERE> [optional rule parameters]
dscp class <CLASSID> <WHERE> [optional rule parameters]
The dscp helper sets the DSCP field in the header of the packets matching the optional rule parameters.
Parameters
- NUMBER is a decimal or hex (0xNN) number to set the DSCP field to.
- CLASSID is any of the iptables supported DiffServ class values (EF, BE, CSxx, AFxx - check iptables -j DSCP --help for more information).
- WHERE tells FireHOL where to search for the specific traffic to be marked.
Currently, WHERE can be one of the build-in iptables chains attached to table mangle.
(for example: INPUT, FORWARD, OUTPUT, PREROUTING, POSTROUTING - case does matter here).
- optional rule parameters is a set of rules that allow further restriction of the traffic that
gets matched by this rule. See Optional Rules Parameters for more information.
Example 1: dscp 32 OUTPUT, will set the DSCP field to 32 of all packets sent by the local machine.
Example 2: dscp 0x20 FORWARD, will set DCSP to 0x20 (32) of all packets passing through the local machine.
Example 3: dscp class EF FORWARD proto tcp dport 25 dst 1.1.1.1 src 2.2.2.2, will set DSCP to DeffServ class EF
for all packets sent by 2.2.2.2, passing through the local machine and targeting port TCP/25 of host 1.1.1.1.
ecn_shame
The ecn_shame helper checks if Explicit Congestion Notification (ECN) is enabled in the kernel, and if it is, it fetches
the ECN Hall of Shame list and disables ECN for all hosts present in the list.
iptables <arguments>
Description
The iptables command passes all its arguments to the real iptables command, during run-time.
You should not use in FireHOL configurations /sbin/iptables or other means to alter a FireHOL firewall.
If you do, your commands will be run before FireHOL activates its firewall and while the previous
firewall is still running. Also, since FireHOL will delete all previous firewall rules in order to
activate the new firewall, any changes you will make, will be deleted too.
Always use the iptables directive to hook iptables commands in a FireHOL firewall. Nothing else.
mac <IP> <MAC>
The mac helper verifies that all traffic comming in with source the IP address, comes from the
MAC address. The helper applies its rules to filter/INPUT and filter/FORWARD and checks the source IP address
in combination with the source MAC address.
The same MAC address is allowed to use other IPs; only the specific IP is required to be used with the specified MAC
address.
Packets with the given IP address but with wrong MAC address will be DROPped and a log (as in loglimit)
with label "MAC MISSMATCH" will appear in the system logs.
This helper has to be used before all interface or router statements. Of course, you can use as many mac statements as you wish.
Example: mac 195.97.5.202 00:02:8a:21:a9:d8
mark <NUMBER> <WHERE> [optional rule parameters]
The mark helper marks the traffic with a specific mark NUMBER that can be matched by traffic shapping tools
for controlling the traffic.
Parameters
- NUMBER is a number to mark the packets with.
- WHERE tells FireHOL where to search for the specific traffic to be marked.
Currently, WHERE can be one of the build-in iptables chains attached to table mangle.
(for example: INPUT, FORWARD, OUTPUT, PREROUTING, POSTROUTING - case does matter here).
- optional rule parameters is a set of rules that allow further restriction of the traffic that
gets matched by this rule. See Optional Rules Parameters for more information.
Keep in mind that if you need to do policy based routing based on iptables marks, you should disable the Root Path Filtering on the interfaces involved (rp_filter in sysctl).
Example 1: mark 1 OUTPUT, will mark with 1 all packets send by the local machine.
Example 2: mark 2 FORWARD, will mark with 2 all packets passing through the local machine.
Example 3: mark 3 FORWARD proto tcp dport 25 dst 1.1.1.1 src 2.2.2.2, will match with 3
all packets sent by 2.2.2.2, passing through the local machine and targeting port TCP/25 of host 1.1.1.1.
masquerade [reverse | interface] [optional rule parameters]
Masquerading is a special from of SNAT (Source NAT) that changes the source of requests when they go out and replaces their original source when replies come in.
This way a Linux box can become an internet router for a LAN of clients having unroutable IP addresses. Masquerading takes care to re-map IP addresses and ports as
required.
Masquerading is "expensive" compared to SNAT because it checks the IP address of the ougoing interface every time for every packet, and therefore it is suggested that
if you connect to the internet with a static IP address, to prefer SNAT.
The masquerade helper sets up masquerading on the output of a network interface (not the interface command, but a real network interface).
If the masquerade command is placed within an interface command, its network interface(s) will be used.
If the masquerade command is placed within a router command that has an outface defined, then the outface network interface(s) will be used.
If placed within a router command but the keyword reverse is specified and the router command has an inface defined, then the inface network interface(s) will be used.
If placed outside and before all primary commands, an interface (or list of space separated interfaces, within double quotes) can be specified on the masquerade command.
In all cases, masquerade will setup itself on the output of the given interface(s).
Please note that if masquerade is used within some interface or router, it does not respect the optional rule parameters
given to this interface or router command. Masquerade uses only its own optional rule parameters.
inface and outface should not be given as parameters to masquerade (inface because iptables does not support this
in the POSTROUTING chain, and outface because it will be overwritten by the interface(s) mentioned above).
Finally, the masquerade helper will turn on FIREHOL_NAT and instruct the kernel to do packet forwarding (like the router
commands do).
Example 1: before the first interface or router: masquerade eth0 src 10.0.0.0/8 dst not 10.0.0.0/8
Example 2: within an interface: masquerade, to masquerade on the output of this interface
Example 3: within a router: masquerade reverse, to masquerade on the output of the router's inface.
nat <type> <target> [optional rule parameters]
The nat helper sets up a NAT rule for routed traffic.
The type parameter can be:
- to-source, to define a Source NAT (created in NAT/POSTROUTING).
The target in this case is the source address to be set in packets matching the
optional rule parameters (if no optional rule parameters, all forwarded traffic will be matched).
target accepts all --to-source values iptables accepts (see iptables -j SNAT --help).
Multiple --to-source values can be given, if separated by space and quoted as a single argument.
inface should not be used in SNAT, because iptables does not provide this information
at this point.
YOU HAVE TO MAKE SURE THAT THE SOURCE ADDRESS YOU SPECIFY IS SUCH THAT THE REPLIES WILL BE SEND BACK
TO THE HOST DOING THE NAT. If the traffic does not get back to this host, then SNAT will simply not work.
THE PACKET FILTERING MECHANISM WILL NOT KNOW THE CHANGE OF THE SOURCE ADDRESS. For packet filtering
(client, server, route statements in interfaces and routers), the
traffic will flow in both directions with its real source address (i.e. real = as it is, before SNATed).
For the host doing the NAT, real is "where do I have to really send it?"
- to-destination, to define a Destination NAT (created in NAT/PREROUTING).
The target in this case is the destination address to be set in packets matching the
optional rule parameters (if no optional rule parameters, all forwarded traffic will be matched).
target accepts all --to-destination values iptables accepts (see iptables -j DNAT --help).
Multiple --to-destination values can be given, if separated by space and quoted as a single argument (in
which case DNAT becomes a load-balancer).
outface should not be used in DNAT, because iptables does not provide this information
at this point.
THE PACKET FILTERING MECHANISM WILL KNOW THE CHANGE OF THE DESTINATION ADDRESS. For packet filtering
(client, server, route statements in interfaces and routers), the
traffic will flow in both directions with its real destination address (i.e. real = as it is, after DNATed).
For the host doing the NAT, real is "where do I have to really send it?"
- redirect-to, to catch traffic comming in and send it to the local machine (created in NAT/PREROUTING).
The target in this case is a port or a range of ports (XXX-YYY) that packets matching the rule
will be redirected to (if no optional rule parameters are given, all incomming traffic will be matched).
target accepts all --to-ports values iptables accepts (see iptables -j REDIRECT --help).
outface should not be used in REDIRECT, because iptables does not provide this information
at this point.
THE PACKET FILTERING MECHANISM WILL KNOW THE CHANGE OF THE DESTINATION PORT. For packet filtering
(client, server statements in interfaces), the
traffic will flow in both directions with its real destination port (i.e. real = as it is, after REDIRECTed).
For the host doing the NAT, real is "where do I have to really send it?"
Please understand that the optional rule parameters are used only to limit the traffic to be matched.
Consider these examples:
Command | Description |
nat to-destination 1.1.1.1 |
Sends to 1.1.1.1 all traffic comming in or passing trhough the firewall host. |
nat to-destination 1.1.1.1 dst 2.2.2.2 |
Redirects to 1.1.1.1 all traffic comming in or passing through, and going to 2.2.2.2. |
nat to-destination 1.1.1.1 proto tcp dst 2.2.2.2 |
Redirects to 1.1.1.1 all TCP traffic comming in or passing through and going to 2.2.2.2. |
nat to-destination 1.1.1.1 proto tcp dport 25 dst 2.2.2.2 |
Redirects to 1.1.1.1 all traffic comming in or passing through and going to 2.2.2.2 to port tcp/25. |
nat to-destination 1.1.1.1 proto tcp dport 25 dst 2.2.2.2 src 3.3.3.3 |
Redirects to 1.1.1.1 all traffic comming in or passing through from 3.3.3.3 and going to 2.2.2.2 to port tcp/25. |
Example 1: nat to-source 1.1.1.1 outface eth0 src 2.2.2.2 dst 3.3.3.3
Example 2: nat to-destination 4.4.4.4 inface eth0 src 5.5.5.5 dst 6.6.6.6
Example 3: nat redirect-to 8080 inface eth0 src 2.2.2.0/24 proto tcp dport 80
redirect [to] <target> [optional rule parameters]
The redirect helper catches all incomming traffic matching the optional rule parameters given and
redirects it to ports on the local host, by calling
nat redirect-to <target> [optional rule parameters]
See the nat helper.
Example: redirect to 8080 inface eth0 src 2.2.2.0/24 proto tcp dport 80
snat [to] <target> [optional rule parameters]
The snat helper sets up a Source NAT rule for routed traffic, by calling
nat to-source <target> [optional rule parameters]
See the nat helper.
Example: snat to 1.1.1.1 outface eth0 src 2.2.2.2 dst 3.3.3.3
tcpmss <what>
The tcpmss helper sets the MSS (Maximum Segment Size) of TCP SYN packets routed through the firewall.
Its purpose is to overcome situations where Path MTU Discovery is not working and packet
fragmentation is not possible.
Within FireHOL, it can be defined either before any primary command, in which case it is applied to all
traffic passing through the firewall, or to any router, in which case it is applied to traffic going out on the outfaces of the router.
The argument can either be the word auto or a number:
- The word auto will make the TCP connections have MSS equal to the MTU of the
outgoing intefrace minus 40 (clamp-mss-to-pmtu).
- A numeric argument will make the TCP connections have MSS equal to the number given.
See also TCPMSS target in iptables tutorial.
Example 1: tcpmss auto
Example 2: tcpmss 500
tos <NUMBER> <WHERE> [optional rule parameters]
The tos helper sets the Type of Service (TOS) in packets.
Parameters
- NUMBER is a number to set TOS to. FireHOL supports decimal numbers, hex numbers and the
descriptive values iptables supports. For more information see iptables -j TOS --help.
- WHERE tells FireHOL where to search for the specific traffic to be marked.
Currently, WHERE can be one of the build-in iptables chains attached to table mangle.
(for example: INPUT, FORWARD, OUTPUT, PREROUTING, POSTROUTING - case does matter here).
- optional rule parameters is a set of rules that allow further restriction of the traffic that
gets matched by this rule. See Optional Rules Parameters for more information.
Example 1: tos 16 OUTPUT, will set TOS to 16 for all packets sent by the local machine.
Example 2: tos 0x10 FORWARD, will set TOS to 0x10 (16) for all packets passing through the local machine.
Example 3: tos Maximize-Throughput FORWARD proto tcp dport 25 dst 1.1.1.1 src 2.2.2.2, will set TOS to Maximize-Throughput (8)
for all packets sent by 2.2.2.2, passing through the local machine and targeting port TCP/25 of host 1.1.1.1.
tosfix
The tosfix helper is based on the suggestions given by Erik Hensema for iptables and tc shapping tricks (check the source here).
It does:
- All TCP ACK packets with length less than 128 bytes are assigned the Minimize-Delay TOS, while bigger ones are assigned the Maximize-Throughput.
- All packets with TOS Minimize-Delay, that are bigger than 512 bytes are set to Maximize-Throughput, except for short bursts of 2 packets per second.
transparent_proxy <service> <port> <user> [optional rule parameters]
The transparent_proxy helper sets up trasparent proxy server for TCP traffic.
The proxy is assumed to be running on the firewall host at port port, with the credentials of the local user user serving TCP port's service requests.
The transparent_proxy helper can be used for two kinds of traffic:
- Incomming TCP traffic, either targeted to the firewall host or passing through the firewall host.
The optional rule parameters can be used to specify which kind of incomming traffic to be catched
(by using inface, src, dst, etc --
outface should not be used here, because the rules generated are placed before
the routing decision and therefore the outgoing interface is not yet known).
If no optional rule parameters are given, then the transparent
proxy will be setup on all network interfaces for all TCP traffic (use this with care since you are
risking to serve requests from the internet using your proxy).
- Locally generated outgoing TCP traffic, except TCP traffic generated by processes running as the user argument.
The optional rule parameters inface, outface and src are
ignored for this type of traffic.
This kind of matching makes it possible to support transparent proxying for clients running on the firewall
host, as far as they do not run as the user excluded. More than one users can be specified by space-separating
and enclosing them in double quotes.
This rule can be disabled by specifing as user the empty string: ""
Of course, make sure that your firewall allows requests to reach your proxy server.
In kernel versions prior to 2.6 you need to enable CONFIG_IP_NF_NAT_LOCAL for locally generated outgoing traffic to be redirected correctly.
Example 1: transparent_proxy 80 3128 squid inface eth0 src 10.0.0.0/8 to redirect all tcp/80 requests coming in from eth0 (10.0.0.0/8) to your local web caching proxy server listening at port 3128 and running as user squid.
Example 2: transparent_proxy "80 3128 8080" 8080 "squid privoxy root bin" inface not "ppp+ ipsec+" dst not "a.not.proxied.server"
to redirect all web and proxy requests (tcp: 80, 3128, 8080) coming in from all interfaces except ppp+ and ipsec+ to your local squid server listening at tcp/8080 except those requests generated by the local users squid, privoxy, root and bin, and all matching traffic that is targeting host 'a.not.proxied.server'.
transparent_squid <port> <user> [optional rule parameters]
The transparent_squid helper sets up trasparent caching for HTTP traffic.
It is equivalent to:
transparent_proxy 80 <port> <user> [optional rule parameters]
Example 1: transparent_squid 3128 squid inface eth0 src 10.0.0.0/8
Example 2: transparent_squid 8080 "squid privoxy root bin" inface not "ppp+ ipsec+" dst not "a.not.proxied.server"
version <number>
Description
The version command states the FireHOL release the configuration file was
created for. In case the configuration file is newer than FireHOL, FireHOL will deny
to run it.
This command is here to allow you or anyone else design and distribute FireHOL configuration files, while
ensuring that the correct FireHOL version is going to run them.
Since FireHOL configurations are BASH script, it is relatively easy to use FireHOL configurations as small
scripts that dynamically process rules stored in a database, in a file system as separate files, or elsewhere.
This directive will help the developers of FireHOL configurations to control the required version of
FireHOL.
The FireHOL release is increased every time the format of the configuration file and the internals of FireHOL are changed.
Since FireHOL v1.67 version is not required to be present in every configuration file.
Actions are the actions to be taken on services and traffic described by other commands and functions.
Please note that normally, FireHOL will pass-through to the generated iptables statements all the possible actions iptables accepts, but only the ones defined here can be used
with lower case letters and currently it will be impossible to pass arguments to some unknown action. Also, keep in mind that the iptables action LOG is a FireHOL optional rule
parameter (see log and loglimit) that can be defined together with one of the following actions and FireHOL will actually produce
multiple iptables statements to achieve both the logging and the action.
accept [with limit <frequency> <burst> [overflow <action>]]
accept [with knock <name>]
accept [with recent <name> <seconds> <hits>]
accept allows the traffic matching the rules to reach its destination.
with limit
The optional parameter with limit offers control over the allowed frequency of NEW connections. frequency and burst have the same syntax of the limit
optional rule parameter.
The overflow action offers control over the overflowed NEW connections. The default is to REJECT overflowed connections (not DROP them, since DROP produces timeouts
on the otherwise valid service clients). Also, the REJECT overflow action, will reject TCP connections with tcp-reset and all others with icmp-host-unreachable.
The overflowed NEW connection attempts will be logged with a OVERFLOW message, with the options the loglimit parameter works.
with knock
The optional parameter with knock allows easy integration with knockd, a server that allows you to control access to services, by sending certain packets to "knock" the
door, before the door is open for service.
This parameter accepts just a name. This name is used to build a special chain knock_<name> which will contain just the rules to allow established connections to work,
but it prevent any new connections from taking place. In case, knockd has not allowed new connections, this traffic entering this chain will just return back
and continue to match against the other rules until the end of the firewall.
How to use it: lets say that you want to allow https traffic based on a knock. In FireHOL you write:
server https accept with knock hidden
and you configure knockd so that it runs:
iptables -A knock_hidden -s %IP% -j ACCEPT
to enable the https service (notice that there is no need to match anything else than the IP. FireHOL already matches everything needed for its rules to work), and:
iptables -D knock_hidden -s %IP% -j ACCEPT
to disable this service for the given IP.
You can use the same knock name in more than one FireHOL services to enable/disable all the services together, without any extra effort in knockd configuration (just one ACCEPT row like above
- i.e. the same knock as far as knockd is concerned - will allow all the FireHOL services with the same knock name to serve requests for the given IP).
Check also this forum post.
with recent
The optional parameter with recent allows new connections to be limited per remote IP for a number of seconds and/or hits.
The name parameter allows multiple statements to share the same recent list of IPs.
If the time or the hits are not required, set them to empty.
Keep in mind that when a new connection is not allowed, the traffic will continue to be matched by the rest of the firewall. In other words, if the traffic is not allowed due to the limitations set here, it is not dropped. It is just not matched by this rule.
Example 1: server smtp accept, to allow SMTP requests and their replies to flow.
Example 2: server smtp accept with limit 10/s 100, to allow SMTP requests to come at 10/second max, with a burst of 100, and their replies to flow back.
Example 3: server smtp accept with limit 10/s 100 overflow drop, to allow SMTP requests to come at 10/second max, with a burst of 100, and their replies to flow back. The overflow requests will be dropped.
Example 4: server smtp accept with limit 10/s 100 src "1.2.3.4 5.6.7.8", same as example 2, but the only valid clients are given as argument to src
Example 4: server smtp accept with recent mail 60 2, to allow only 2 connections every 60 seconds per remote IP, to the smtp server.
reject [with <message>]
reject discards the matching traffic but sends a rejecting message back to the sender.
with is used to offer control on the message to be returned to the sender.
with accepts all the arguments the --reject-with iptables expression accepts. For an updated list
of these messages type iptables -j REJECT --help. Bellow you can find the list my system accepts.
By default (no with argument given), reject will send an icmp-port-unreachable on all protocols
except TCP, for which it will send a tcp-reset.
Message | Alias | Description | Reference |
icmp-net-unreachable |
net-unreach |
ICMP network unreachable
From RFC 1812 section 5.2.7.1
Generated by a router if a forwarding path (route) to the destination network is not available.
Notes
Use this with care. The sender and the routers between you and the sender may conclude that
the whole network your host resides in, is unreachable and prevent other traffic from
reaching you.
|
RFC 1812
RFC 792 |
icmp-host-unreachable |
host-unreach |
ICMP host unreachable
From RFC 1812 section 5.2.7.1
Generated by a router if a forwarding path (route) to the destination host on a directly connected network
is not available (does not respond to ARP).
Notes
Use this with care. The sender and the routers between you and the sender may conclude that
your server is unreachable and prevent the transmission of other traffic to you.
|
RFC 1812
RFC 792 |
icmp-proto-unreachable |
proto-unreach |
ICMP protocol unreachable
From RFC 1812 section 5.2.7.1
Generated if the transport protocol designated in a datagram is not supported in the transport layer
of the final destination.
|
RFC 1812
RFC 792 |
icmp-port-unreachable |
port-unreach |
ICMP port unreachable (default)
From RFC 1812 section 5.2.7.1
Generated if the designated transport protocol (e.g. TCP, UDP, etc) is unable to demultiplex the
datagram in the transport layer of the final destination but has no protocol mechanism to inform
the sender.
In other words
Generated by hosts to indicate that the required port is not active. |
RFC 1812
RFC 792 |
icmp-net-prohibited |
net-prohib |
ICMP communication with destination network administratively prohibited
From RFC 1812 section 5.2.7.1
This code was intended for use by end-to-end encryption devices used by U.S military agencies.
Routers SHOULD use the newly defined Code 13 (Communication Administratively Prohibited) if they
administratively filter packets.
In other words
Use it, but don't expect that this will be widely understood.
|
RFC 1812
RFC 1122 |
icmp-host-prohibited |
host-prohib |
ICMP communication with destination host administratively prohibited
From RFC 1812 section 5.2.7.1
This code was intended for use by end-to-end encryption devices used by U.S military agencies.
Routers SHOULD use the newly defined Code 13 (Communication Administratively Prohibited) if they
administratively filter packets.
In other words
Use it, but don't expect that this will be widely understood.
|
RFC 1812
RFC 1122 |
tcp-reset |
tcp-reset |
TCP RST
This is the port unreachable message of the TCP stack. It is useful when you want to prevent
timeouts on rejected TCP services when the client is badly written to ignore ICMP port unreachable
messages.
I suggest to use this for rejecting idents: server ident reject with tcp-reset.
|
RFC 1122 |
Example 1: policy reject with host-unreach
Example 2: server ident reject with tcp-reset
Example 3: UNMATCHED_INPUT_POLICY="reject with host-prohib"
drop
drop silently discards the matching traffic.
The fact that the traffic is silently discarded makes the sender timeout in order to conclude that it is not
possible to use the wanted service.
Example: server smtp drop, to silently discard SMTP requests and their replies.
deny
deny is just an alias for drop, made for those who are used to ipchains terminology.
Example: server smtp deny, to silently discard SMTP requests and their replies.
tarpit
Captures and holds incoming TCP connections using no local per-connection resources. Connections are accepted, but immediately switched to the persist state (0 byte window),
in which the remote side stops sending data and asks to continue every 60-240 seconds. Attempts to close the connection are ignored, forcing the remote side to
time out the connection in 12-24 minutes.
Example: server smtp tarpit, to force remote clients to timeout.
return
return will return the flow of processing to the parent of the current command.
Currently, it has meaning to specify the action return only as a policy to some interface.
Example: policy return, to have traffic not matched by any rule within an interface to continue traveling through the
firewall and possibly matched by other interfaces bellow.
mirror
mirror will return the traffic to the wanted port, back to the sending host.
Use this with care, and only if you understand what you doing.
Keep also in mind that FireHOL will apply this action to both requests and replies comming in or passing through,
and will replace it with REJECT for traffic generated by the local host.
redirect [to-port port]
redirect is used internally by FireHOL Helper Commands to redirect traffic to ports on the local host.
Unless you are a developer, you will never need to use this directly.
Optional rule parameters are accepted by many commands to narrow the match they do by default.
The parameters described bellow are all that FireHOL supports. You should check the documentation of each command to find which parameters should not be used with it.
Normally, all FireHOL commands are designed so that if you specify a parameters that is also used internally, the internal one will overwrite the one given in the configuration file. In such a case,
FireHOL will present you a warning with the old and the new value.
Not all parameters should be used in all cases. For example sport and dport should not be used in normal server and
client commands since such ports are internally defined by the services themselves. In any case, FireHOL will complain
about optional rule parameters that should not be used in certain commands.
src [not] <host>
Description
src defines the source IP address of the REQUEST. If src is defined on a server statement it matches the
source of the request which is the remote host, while if it is defined on a client statement it matches again
the source of the request, but this time it is the local host. Focus on the REQUEST!!! Forget the reply.
Parameters
- not is an optional argument that reverses the match. When defined, the rule will match all hosts
except the ones defined.
Example: server smtp accept src not 1.2.3.4
- host can be an IP address, a hostname, or a subnet.
Multiple hosts/networks can be defined if separated by space and quoted as a single argument.
Example 1: server smtp accept src 1.2.3.4
Example 2: server smtp accept src not "1.2.3.0/24 5.6.7.8 badhost.example.com"
dst [not] <host>
Description
dst defines the destination of the REQUEST. If dst is defined on a server statement it matches the
destination of the request which is the local host, while if it is defined on a client statement it matches again
the destination of the request, but this time it is the remote host. Focus on the REQUEST!!! Forget the reply.
dst accepts the same parameters as src
srctype [not] <type>
Description
srctype defines the source IP address type. The following types are supported:
- UNSPEC an unspecified address (i.e. 0.0.0.0)
- UNICAST an unicast address
- LOCAL a local address
- BROADCAST a broadcast address
- ANYCAST an anycast packet
- MULTICAST a multicast address
- BLACKHOLE a blackhole address
- UNREACHABLE an unreachable address
- PROHIBIT a prohibited address
- THROW
- NAT
- XRESOLVE
More that one types can be specified if given as a space separated list enclosed in quotes.
FireHOL will convert to upper case any strings given as types.
For more information check: iptables -m addrtype --help (or man iptables).
dsttype [not] <type>
Description
dsttype defines the destination IP address type. This parameter has the same rules with srctype.
inface [not] <interface>
Description
inface defines the interface the REQUEST is received via.
inface cannot be used in interface commands.
Parameters
- not is an optional argument that reverses the match. When defined, the rule will match all interfaces
except the ones defined.
Example: server smtp accept inface not eth0
- interface if an interface name in the same format the interface command accepts.
Multiple interfaces can be defined if separated by space and quoted as a single argument.
Example 1: server smtp accept inface not eth0
Example 2: server smtp accept inface not "eth0 eth1"
outface [not] <interface>
Description
outface defines the interface the REQUEST is send via.
outface cannot be used in interface commands.
outface accepts the same parameters as inface.
physin [not] <interface>
Description
physin defines the physical interface the REQUEST is received via and is used in cases where inface is defined as
a virtual interface (such as a bridge interface). In such cases, the physdev iptables module (which is used by physin)
can be used to match the physical network interface the REQUEST is recieved via.
physin accepts the same parameters as inface.
physout [not] <interface>
Description
physout defines the physical interface the REQUEST is send via and is used in cases where outface is defined as
a virtual interface (such as a bridge interface). In such cases, the physdev iptables module (which is used by physout)
can be used to match the physical network interface the REQUEST is send via.
physout accepts the same parameters as outface.
custom "parameters"
Description
custom passes its arguments to the generated iptables commands.
It is required to quote all the parameters given to custom. If the parameters include a space character
between some text that is required to be given to iptables as one argument, it is required to escape another set of
quotes in order. Another way is to use double quotes externally and single quotes internally.
Example 1: server smtp accept custom "--some-iptables-option and_its_value"
Example 2: server smtp accept custom "--some-iptables-option 'one_value another_value'"
log "<some text>" [level a_level]
Description
log will log the matching packets to syslog. Note that this is not an action (in iptables it is).
FireHOL will actually produce multiple iptables commands to accomplish both the action for the rule and the logging.
You can control how logging works, by altering the variables FIREHOL_LOG_MODE, FIREHOL_LOG_OPTIONS and
FIREHOL_LOG_LEVEL. You can also change the level of just one rule by using the
level argument of the log parameter (only when FIREHOL_LOG_MODE=LOG).
FireHOL logs traffic, exactly the same way iptables does. Many users have complained about packet logs appearing at
their console. To avoid this you will have to:
- setup klogd to log only more important traffic
- change FIREHOL_LOG_LEVEL to log at a not so important log-level
Actually klogd's -c option and iptables' --log-level option are the same thing (iptables accepts also
the numeric values klogd accepts). If iptables logs at a
higher priority than klogd is configured to use, then your packets will appear in the console too.
In most kernels klogd is by default configured to log everything, so if you don't also change klogd's -c option,
the --log-level setting of iptables has no effect. Use the table bellow to match klogd options with --log-level options:
Priority | klogd | iptables | Description |
0 |
0 |
emerg |
system is unusable |
1 |
1 |
alert |
action must be taken immediately |
2 |
2 |
crit |
critical conditions |
3 |
3 |
error |
error conditions |
4 |
4 |
warning (default) |
warning conditions |
5 |
5 |
notice |
normal but significant condition |
6 |
6 |
info |
informational |
7 |
7 (default) |
debug |
debug-level messages |
To prevent packet logs appearing to the console, klogd option must be LOWER than the one iptables uses.
On RedHat systems, you can configure klogd by changing
/etc/sysconfig/syslog and adding to KLOGD_OPTIONS the required -c level.
loglimit "<some text>"
Description
loglimit is the same with log but limits the frequency of logging according to the setting of
FIREHOL_LOG_FREQUENCY and FIREHOL_LOG_BURST.
proto [not] <protocol>
Description
proto sets the required protocol for the traffic. This command accepts anything iptables accepts as
protocols.
limit <frequency> <burst>
Description
limit will limit the match in both directions of the traffic (request and reply). This is used internally by FireHOL and its effects has not been tested in the high level
configuration file directives.
sport <port>
Description
sport defines the source port of a request. It accepts port names, port numbers, port ranges (FROM:TO) and multiple ports (or ranges) seperated by spaces and quoted as a single argument.
This parameter should not be used in normal services definitions (client and server commands) or
interface and router definitions, unless you really understand what you are doing.
dport <port>
Description
dport defines the destination port of a request. It accepts port names, port numbers, port ranges (FROM:TO) and multiple ports (or ranges) seperated by spaces and quoted as a single argument.
This parameter should not be used in normal services definitions (client and server commands) or
interface and router definitions, unless you really understand what you are doing.
uid [not] <user>
user [not] <user>
Description
uid or user define the operating system user sending this traffic. The parameter can be a username, a user number or a list of these two, seperated by spaces and quoted as
a single argument.
This parameter can be used only in services (client and server commands) defined within interfaces, not routers.
FireHOL is "smart" enough to apply this parameter only to traffic send by the localhost, i.e. the replies of servers and requests of clients.
It is not possible, and FireHOL will simply ignore this parameter, on traffic coming in or passign through the firewall host.
Example 1: client "pop3 imap" accept user not "user1 user2 user3" dst mymailer.example.com
The above will allow local users except user1, user2 and user3 to use POP3 and IMAP services on mymailer.example.com.
You can use this, for example, to allow only a few of the local users use the fetchmail program to fetch their mail from the mail server.
Example 2: server http accept user apache
The above will allow all HTTP to reach the local http server, but only if the web server is running as user apache the replies will be send back to the HTTP client.
gid <group>
group <group>
Description
gid or group define the operating system user group sending this traffic. The parameter can be a group name, a group number or a list of these two, seperated by spaces and quoted as
a single argument.
This parameter can be used only in services (client and server commands) defined within interfaces, not routers.
FireHOL is "smart" enough to apply this parameter only to traffic send by the localhost, i.e. the replies of servers and requests of clients.
It is not possible, and FireHOL will simply ignore this parameter, on traffic coming in or passign through the firewall host.
pid <process>
process <process>
Description
pid or process define the operating system process ID (or PID) sending this traffic. The parameter can be a PID or a list of PIDs, seperated by spaces and quoted as
a single argument.
This parameter can be used only in services (client and server commands) defined within interfaces, not routers.
FireHOL is "smart" enough to apply this parameter only to traffic send by the localhost, i.e. the replies of servers and requests of clients.
It is not possible, and FireHOL will simply ignore this parameter, on traffic coming in or passign through the firewall host.
sid <session>
session <session>
Description
sid or session define the operating system session ID of the process sending this traffic (The session ID of a process is the process group ID of the session leader).
The parameter can be a list of such IDs, seperated by spaces and quoted as a single argument.
This parameter can be used only in services (client and server commands) defined within interfaces, not routers.
FireHOL is "smart" enough to apply this parameter only to traffic send by the localhost, i.e. the replies of servers and requests of clients.
It is not possible, and FireHOL will simply ignore this parameter, on traffic coming in or passign through the firewall host.
cmd <name>
command <name>
Description
cmd or command define the operating system command name of the process sending this traffic.
The parameter can be a list of such command names, seperated by spaces and quoted as a single argument.
This parameter can be used only in services (client and server commands) defined within interfaces, not routers.
FireHOL is "smart" enough to apply this parameter only to traffic send by the localhost, i.e. the replies of servers and requests of clients.
It is not possible, and FireHOL will simply ignore this parameter, on traffic coming in or passign through the firewall host.
mac [not] <address>
Description
mac matches the source MAC address of packets comming into the firewall. The mac parameter does nothing for outgoing traffic.
For interfaces, the mac parameter matches against all traffic that comes into the firewall, whether it is server
or client traffic.
For routers, the mac parameter matches also against all traffic comming into the firewall, but firehol considers the router
input differently based on the command given. For server or route statements,
the mac parameter matches the MAC address of the client (the host sending the request),
while for client statements it matches the source MAC address of the server (the host accepting requests).
In principle, the mac parameter behaves the same for both interfaces and routers and this is why:
The mac parameter matches the source MAC address of what FireHOL considers the "remote" host, not the one that FireHOL considers the "protected" one.
For interfaces, this is simple, because always the "remote" host is a remote host and the "protected" host is the one running the
firewall. For routers though, the command chosen (client or server) defines what the firewall protects. Therefore, a client statement
protects the client making the "remote" host the server, while a server statement protects the server and therefore the "remote"
host is the client.
More than one MAC addresses can be given if separated by spaces and enclosed in quotes as a single argument to the mac parameter.
The not argument will reverse the match. In case there are many MAC addresses defined, positive expressions are ORed
(either address should be matched), while negative expressions are ANDed (none of the addresses should be matched).
mark [not] <ID>
Description
mark matches the traffic against the given IDs. This command accepts anything iptables accepts as MARKs (see iptables -m mark --help).
More than one MARK IDs can be given if separated by spaces and enclosed in quotes as a single argument to the mark parameter.
tos [not] <ID>
Description
tos matches the traffic against the given IDs. This command accepts anything iptables accepts as TOS (see iptables -m tos --help).
More than one IDs can be given if separated by spaces and enclosed in quotes as a single argument to the tos parameter.
dscp [not] <ID>
dscp [not] class <ID>
Description
dscp matches the traffic against the given DSCP IDs. This command accepts anything iptables accepts as DSCP (see iptables -m dscp --help).
More than one IDs can be given if separated by spaces and enclosed in quotes as a single argument.
DEFAULT_INTERFACE_POLICY
Description
DEFAULT_INTERFACE_POLICY controls the default action to be taken on traffic not matched by any rule within an interface.
Actually, this is a global setting for what policy does for an interface.
All packets that reach the end of an interface are logged only if the action is not return or accept.
You can control the frequency of this logging by altering the frequency loglimit uses.
Default: DEFAULT_INTERFACE_POLICY="DROP"
Example: DEFAULT_INTERFACE_POLICY="REJECT"
DEFAULT_ROUTER_POLICY
Description
DEFAULT_ROUTER_POLICY controls the default action to be taken on traffic not matched by any rule within a router.
Actually, this is a global setting for what policy does for a router.
All packets that reach the end of a router are logged only if the action is not return or accept.
You can control the frequency of this logging by altering the frequency loglimit uses.
Default: DEFAULT_ROUTER_POLICY="RETURN"
Example: DEFAULT_ROUTER_POLICY="REJECT"
UNMATCHED_INPUT_POLICY
UNMATCHED_OUTPUT_POLICY
UNMATCHED_ROUTER_POLICY
Description
UNMATCHED_INPUT_POLICY controls the default action to be taken for incoming traffic not matched by any interface command.
UNMATCHED_OUTPUT_POLICY controls the default action to be taken for outgoing traffic not matched by any interface command.
UNMATCHED_ROUTER_POLICY controls the default action to be taken for forwarded traffic not matched by any router command.
All variables accept all the Actions FireHOL supports.
All packets that reach the end of firewall in all three chains are logged (always, regardless of these settings).
You can control the frequency of this logging by altering the frequency loglimit uses.
Default: UNMATCHED_INPUT_POLICY="DROP"
Default: UNMATCHED_OUTPUT_POLICY="DROP"
Default: UNMATCHED_ROUTER_POLICY="DROP"
Example: UNMATCHED_INPUT_POLICY="REJECT"
Example: UNMATCHED_OUTPUT_POLICY="REJECT"
Example: UNMATCHED_ROUTER_POLICY="REJECT"
FIREHOL_INPUT_ACTIVATION_POLICY
FIREHOL_OUTPUT_ACTIVATION_POLICY
FIREHOL_FORWARD_ACTIVATION_POLICY
Description
All these variables have been added in v1.133
FIREHOL_INPUT_ACTIVATION_POLICY controls the default action to be taken for incoming traffic during firewall activation.
FIREHOL_OUTPUT_ACTIVATION_POLICY controls the default action to be taken for outgoing traffic during firewall activation.
FIREHOL_FORWARD_ACTIVATION_POLICY controls the default action to be taken for forwarded traffic during firewall activation.
All variables accept either ACCEPT, DROP or REJECT.
The default is ACCEPT in order to prevent a denial of service during a firewall restart. This is by design correct,
since FireHOL will block all invalid connections after the firewall is fully activated (remember that FireHOL allows
specific traffic to pass in both directions of the firewall).
Default: FIREHOL_INPUT_ACTIVATION_POLICY="ACCEPT"
Default: FIREHOL_OUTPUT_ACTIVATION_POLICY="ACCEPT"
Default: FIREHOL_FORWARD_ACTIVATION_POLICY="ACCEPT"
Example: FIREHOL_INPUT_ACTIVATION_POLICY="REJECT"
Example: FIREHOL_OUTPUT_ACTIVATION_POLICY="REJECT"
Example: FIREHOL_FORWARD_ACTIVATION_POLICY="REJECT"
FIREHOL_LOG_MODE
FIREHOL_LOG_LEVEL
FIREHOL_LOG_OPTIONS
FIREHOL_LOG_FREQUENCY
FIREHOL_LOG_BURST
FIREHOL_LOG_PREFIX
Description
FIREHOL_LOG_MODE controls the method of logging used by FireHOL. Currently,
two modes are supported: LOG and ULOG.
FIREHOL_LOG_LEVEL controls the level at which iptables will log things to the syslog.
For a description of the possible values supported and for per-rule control of log level,
see the log optional rule parameter. FIREHOL_LOG_LEVEL is ignored when FIREHOL_LOG_MODE=ULOG.
FIREHOL_LOG_OPTIONS controls the way iptables will log things to the syslog.
The value of this variable is passed as is to iptables, so use exact iptables parameters.
This variable can have special arguments for the LOG or ULOG actions of iptables.
FIREHOL_LOG_FREQUENCY and FIREHOL_LOG_BURST (added in v1.39 of FireHOL) control the frequency at each each logging
rule will write packets to the syslog. FIREHOL_LOG_FREQUENCY is set to the maximum average
frequency and FIREHOL_LOG_BURST specifies the maximum initial number of packets to match.
FIREHOL_LOG_PREFIX adds its contents to every log line, so that FireHOL logs can easily be detected in the
system log. By default it is empty.
Default: FIREHOL_LOG_MODE="LOG"
Default: FIREHOL_LOG_OPTIONS="--log-level warning"
Default: FIREHOL_LOG_FREQUENCY="1/second"
Default: FIREHOL_LOG_BURST="5"
Default: FIREHOL_LOG_PREFIX=""
Example: FIREHOL_LOG_OPTIONS="--log-level info --log-tcp-options --log-ip-options"
Example: FIREHOL_LOG_FREQUENCY="30/minute"
Example: FIREHOL_LOG_BURST="2"
Example: FIREHOL_LOG_PREFIX="FIREHOL: "
To see the available iptables log options, run /sbin/iptables -j LOG --help or /sbin/iptables -j ULOG --help (depending on FIREHOL_LOG_MODE)
To see what iptables accepts as frequencies and bursts run, /sbin/iptables -m limit --help
You can also check man iptables.
FIREHOL_DROP_INVALID
Description
If FIREHOL_DROP_INVALID is set to 1, FireHOL will drop all packets that are matched by the INVALID state of the
iptables connection tracker. The default in versions prior to v1.183 was to drop all those packers. The new default
is not to drop those packets globally, but only as part of the protection statement.
See also Bug 927509 that discusses the effects of globally dropping invalid packets.
Default: FIREHOL_DROP_INVALID="0"
Example: FIREHOL_DROP_INVALID="1"
DEFAULT_CLIENT_PORTS
Description
DEFAULT_CLIENT_PORTS controls the port range to be used when a remote client is specified.
For localhost clients, FireHOL finds the exact client ports by querying the kernel options.
Default: DEFAULT_CLIENT_PORTS="1024:65535"
Example: DEFAULT_CLIENT_PORTS="0:65535"
FIREHOL_NAT
Description
If FIREHOL_NAT is set to 1, FireHOL will load NAT kernel modules for those services that they are require such.
FireHOL sets this to 1 automatically if you use the Helper Commands that do NAT.
Default: FIREHOL_NAT="0"
Example: FIREHOL_NAT="1"
FIREHOL_AUTOSAVE
Description
FIREHOL_AUTOSAVE controls the file that will be created when FireHOL is called with the save command line argument.
If this variable is empty (the default), FireHOL will try to detect where to save the file. Currently, the RedHat way
(/etc/sysconfig/iptables) and the Debian way (/var/lib/iptables/autosave) are automatically detected (in the order given here)
based on the existance of the directory this file should be created in.
Default: FIREHOL_AUTOSAVE=""
Example: FIREHOL_AUTOSAVE="/tmp/firehol-saved.txt"
FIREHOL_LOAD_KERNEL_MODULES
Description
FIREHOL_LOAD_KERNEL_MODULES controls the way FireHOL handles required kernel modules.
If set to 0 (zero), FireHOL will not attempt to load kernel modules at all. Set this to 0 (zero)
if you have compiled a kernel that has all the modules build into it.
Since v1.165 of the FireHOL this feature is almost useless.
FireHOL now tries to find the .config file of your kernel
(in /proc/config or /lib/modules/`uname -r`/build/.config
or /usr/src/linux/.config, in this order) and figures out which kernel modules are compiled
build-in and which are modules.
So, this variable is only used in cases where FireHOL has no access at all to kernel configuration.
Keep in mind, that FireHOL is able to detect the ip_tables and ip_conntrack modules
by examining the relative entries in /proc, so it can detect whether these are loaded without the
need for the kernel configuration.
Default: FIREHOL_LOAD_KERNEL_MODULES=1
Example: FIREHOL_LOAD_KERNEL_MODULES=0
FIREHOL_TRUST_LOOPBACK
Description
FIREHOL_TRUST_LOOPBACK allows you to choose if FireHOL will trust the lo interface and
accept all traffic to or from it (although not FORWARD, it matches only INPUT and OUTPUT).
The default is to accept all traffic from/to lo. If however you need to control what
the local system users can use on the local machine, you can set this to anything other than 1
allowing you to setup the firewall as you will.
IMPORTANT: If you choose not to trust interface lo and you don't specifically setup the lo
interface in firehol.conf, no local process will be able to communicate to any other local process,
which most likely will break several things. Be sure you know what you are doing.
This option appeared in FireHOL version 1.195.
Default: FIREHOL_TRUST_LOOPBACK=1
Example: FIREHOL_TRUST_LOOPBACK=0
FIREHOL_DROP_ORPHAN_TCP_ACK_FIN
Description
FIREHOL_DROP_ORPHAN_TCP_ACK_FIN allows you to choose if FireHOL will drop all TCP connections
with ACK FIN set without logging them.
In busy environments the iptables connection tracker removes from the connection tracking list
entries when it receives a FIN. This makes the ACK FIN appear as an invalid packet (it does not
match anything on the connection tracking list, and it is not a new connection).
Normally FireHOL logs this traffic. If FIREHOL_DROP_ORPHAN_TCP_ACK_FIN is set to 1, FireHOL
will silentrly drop such packets.
Default: FIREHOL_DROP_ORPHAN_TCP_ACK_FIN=0
Example: FIREHOL_DROP_ORPHAN_TCP_ACK_FIN=1
RESERVED_IPS
Description
This variable includes all the IP addresses defined as IANA - Reserved by IANA.
You can overwrite the internal default definion of this variable by creating a file with same name in /etc/firehol.
This file accepts an IP definition per line.
The supplied get-iana.sh script creates this file by directly fetching this document.
Example: interface eth0 internet src not "${RESERVED_IPS}"
PRIVATE_IPS
Description
This variable includes all the IP addresses defined as Private or Test by RFC 3330.
You can overwrite the internal default definion of this variable by creating a file with same name in /etc/firehol.
This file accepts an IP definition per line.
Example: interface eth0 internet src not "${PRIVATE_IPS}"
UNROUTABLE_IPS
Description
This variable is both RESERVED_IPS and PRIVATE_IPS together. I suggest to use this variable on interfaces and routers
accepting Internet traffic.
You can overwrite the internal default definion of this variable by creating a file with same name in /etc/firehol.
This file accepts an IP definition per line.
Example: interface eth0 internet src not "${UNROUTABLE_IPS}"
MULTICAST_IPS
Description
You can overwrite the internal default definion of this variable by creating a file with same name in /etc/firehol.
This file accepts an IP definition per line.
Example: interface eth0 internet src not "${MULTICAST_IPS}"
|
$Id: commands.html,v 1.75 2010/06/07 15:44:09 ktsaou Exp $
FireHOL, a firewall for humans...
© Copyright 2004
Costa Tsaousis <costa@tsaousis.gr>
|