Commit 46e9a32f authored by Mike Hibler's avatar Mike Hibler
Browse files

First cut at firewall documentation.

parent 311ededb
Copyright (c) 2004 University of Utah and the Flux Group.
All rights reserved.
<h1>Simple Firewalled Experiment Support in Emulab</h1>
<li> <a href="#Overview">Overview</a>
<li> <a href="#Use">Use</a>
<li> <a href="#Limitations">Current Limitations</a>
<li> <a href="#KnownBugs">Known Bugs</a>
<li> <a href="#Example">A Complete Example</a>
<a NAME="Overview"></a><h2>Overview</h2>
Emulab allows the ability to setup a
<a href="docwrapper.php3?docname=tutorial.html#ControlNet">control-network</a>
firewall between an experiment and the outside world.
This is <b>not</b> a firewall
between nodes within an experiment, that is, the firewall is not part of
your NS-specified network topology.
The purpose of an Emulab firewall is to prevent experimental traffic from
escaping, via the control network, to the internet due to a mis-configured
application or router within an experiment. In its initial
<a href="#Limitations">limited</a>
form, the Emulab firewall is intended to prevent accidental misuse and not
malicious abuse.
The firewall is implemented by allocating an additional node to the experiment.
This node is set as the default route for all other nodes in the experiment.
Thus, all outgoing, non-experimental traffic will pass through the node.
This firewall node, running IPFW on FreeBSD, can be configured in a number
of ways to block or allow certain types of traffic. Inbound traffic directed
to the nodes does not pass through the firewall.
<a NAME="Use"></a><h2>Use</h2>
To add a firewall to an Emulab experiment, you specify a Firewall object
in your NS file (there is currently no way to add a firewall via the Emulab
client or experiment creation GUI):
set fw [new Firewall $ns]
$fw set-style &lt;style&gt;
This tells Emulab to add a node called "fw" to your experiment, and provides
a handle for adding rules to the firewall. The "style" of the firewall is
one of:
<li><i>closed</i>: A closed firewall allowing no outgoing traffic.
This translates to the IPFW rule "deny ip from any to any".
<li><i>open</i>: A completely open firewall allowing all traffic.
This translates to the IPFW rule "allow ip from any to any".
<li><i>basic</i>: A mostly closed firewall allowing only <code>ssh</code>
and ICMP traffic.
The firewall can be augmented with user-specified rules as well:
$fw add-rule &lt;IPFW format rule string&gt;
In the basic form, rules are numbered starting at 1. Thus, rules are
interpreted in the firewall in the order in which they appear in the NS file.
The default rules for a firewall style are numbered starting at 50000, so
user-specified rules are also interpreted before the default rule set.
There is also a method allowing explicit numbering of rules:
$fw add-numbered-rule &lt;ruleno&gt; &lt;IPFW format rule string&gt;
allowing more precise control over their interpretation and allowing for
linked rules. A <a href="#Example">complete example</a> is shown below.
<a NAME="Limitations"></a><h2>Limitations</h2>
One should note carefully the following issues, some of which limit an
Emulab firewall from being safely used as a containment environment for
Really Bad Stuff.
Our ultimate goal is to produce a highly secure containment environment,
this implementation is only the first step.
<li><b>The firewall is implemented using OS-provided routing</b>.
Specifically, every node has its default route changed to point to the
firewall node. Sufficiently powerful applications could accidentally
or intentionally change the default route back to the Emulab router
thus circumventing all protection. In the near future, nodes in an
experiment will have their control net interfaces in a private LAN
not including the Emulab router. Thus they will be able to change their
default route, but they won't be able to reach the outside world.
<li><b>The firewall is just another node in the topology</b>.
It is setup just like all other nodes, this includes enabling accounts
and NFS access. Thus anyone who can login to a node, can login to the
firewall. If they are root on a node, they are root on the firewall.
In the future, firewalls will have much more constrained access.
<li><b>Intra-Emulab traffic is not firewalled</b>.
As mentioned, this is not a firewall in the experiment topology so traffic
over topology links (10.x.x.x address) is not firewalled.
Additionally, traffic between nodes over the control net
(155.98.36.x addresses) is not filtered since the control net is a LAN
and all other nodes are directly reachable. This will probably not
change in the future. Finally, traffic between the nodes and the Emulab
infrastructure (boss and ops) does not pass through the firewall. Host
routes are explicitly setup to avoid the firewall. This is a simplification
for node setup, but will be fixed in the future.
<li><b>The firewall must run Freebsd with IPFW</b>.
However, this is not fundamental to the design and we hope that the
firewall syntax will be general enough to support at least Linux with
ipchains as well.
<li><b>The firewall rule syntax is fixed</b>.
In particular, it allows only fixed strings with no form of variable
substitution. This is evident in the <a href="#Example">example</a>
where there are a number of unsightly hardwired IP addresses for Emulab
<li><b>Firewall setup is static</b>.
Rules are specified in the NS file at experiment creation time.
Rules can be changed using experiment modify, or by logging into the firewall
and changing rules by hand with IPFW.
A more dynamic interface may be desirable,
in particular integration with the event system.
On the other hand, it is not necessarily
a good thing to have a firewall that is too easy to (mis)configure.
<li><b>There is always a firewall for experiments</b>.
Keep in mind that Emulab has an external firewall already which imposes
some site-wide restrictions (yes, there should be a link here...)
So just allowing some ports through your experimental firewall doesn't
guarantee that the affected traffic will make it to the outside world.
<a NAME="Known Bugs"></a><h2>Known Bugs</h2>
There is a mighty fine line between a "limitation" and a "bug".
For now, we are viewing things as the <a href="#Limitations">former</a>.
<a NAME="Example"></a><h2>A Complete Example</h2>
Here is the NS code for a simple two node experiment with a firewall:
source tb_compat.tcl
set ns [new Simulator]
set n1 [$ns node]
tb-set-node-os $n1 FBSD-STD
set n2 [$ns node]
tb-set-node-os $n2 RHL-STD
set link [$ns duplex-link $n1 $n2 100Mb 0ms DropTail]
# create a firewall node
set fw [new Firewall $ns]
$fw set-style basic
# allow tracroute through
$fw add-rule "allow udp from to any 33434-33524"
$fw add-rule "allow udp from any 33434-33524 to"
$ns run
Notice that the second added rule filters incoming traffic and
is not strictly necessary right now since inbound traffic does not pass
through the firewall. This NS specification yields a topology that looks like:
<img src="firewall-experiment.png" alt="&lt;Three nodes, two connected&gt;">
No surprises. Two nodes connected by a link, with a "disconnected" firewall
off to the side.
What isn't shown is that all three are connected via the control net.
The firewall is accessible via a DNS name of
similarly to
other nodes. You can login to the firewall, reboot it, etc. just as any
other node.
The <code>Experiment Details</code> web page for the experiment lists
all the rules for the firewall:
Firewall information:
ID Type Style Rule# Rule
--------------- -------- -------- ----- -----------------------------------
fw ipfw basic 1 allow udp from to any 33434-33524
2 allow udp from any 33434-33524 to
55000 allow ip from me to me
55100 allow ip from me to
55101 allow ip from to me
55200 allow icmp from any to
55201 allow icmp from to any
55300 allow tcp from any to 22
55301 allow tcp from 22 to any
55400 allow tcp from me 16534 to
55401 allow tcp from to me 16534
55500 allow udp from 2917 to 2917
65534 deny ip from any to any
Notice the rules involving multicast (224) and 155.98.32/23. These allow all
nodes to talk to Emulab infrastructure.
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment