Building a Gateway Server

Building a Gateway Server
Dashamir Hoxha <dashohoxha@gmail.com>

Copyright (C) 2009 Dashamir Hoxha. Permission is granted to copy, distribute and/or modify this
document under the terms of the GNU Free Documentation License, Version 1.1 or any later version
published by the Free Software Foundation; with no Invariant Sections, with no Front-Cover Texts, and
with no Back-Cover Texts. A copy of the license is included in the section entitled "GNU Free
Documentation License."

1. Abstract
A gateway server is a server through which the computers in a LAN access the Internet. This is
usually done through NAT. It should also provide firewall protection for the LAN and it can also serve
as a DNS and DHCPD server for the LAN.
Besides these, the gateway server can also provide the following services:
● access control for the LAN (which user can access the Internet and which not)
● accounting and statistics (how much Internet resources each user has consumed)
● management (cut the line of the users that have consumed above the allowed limit)
● traffic control (offer different quality of service to different groups of users)
● load balancing and backup link (connecting to the Internet through more than one line and
using them efficiently)
● etc.
Some years ago I have been involved in a project for building gateway servers like this, using
slackware on old PCs. In this article I will try to explain the things that I have done on this project and
how I did them. Later I discovered that in fact I was re-inventing the wheel, because most of these
requirements can be fulfilled quite well by using MikroTik, HotSpot, ChilliSpot, freeRadius, etc.
Anyway, the shell scripts and web interfaces that I built for this project can still have any educational
values, or they can serve as inspiration for somebody else who would like to build such things himself.
So, if you are a results oriented people, maybe the article is not so interesting for you. However, if you
are a hands-on, try-for-yourself kind of people, who would like to experiment with such things, then
this article can be just right for you.

2. The Requirements (Features)
The gateway server that we built was named SCARLET. Don't ask me why, it was the choice of my
partner and apparently there was no particular reason for naming it like that.
The aim of scarlet was to manage the access to the Internet of the clients of a small ISP. However, it
can be used as well to manage the connections of the employees of a company or organization. It has
two network interfaces, one that is connected on the client side (LAN), and one that is connected on
the Internet side (WAN). It can also have more than two network interfaces. For example, there can be
more than one LAN of clients. This is the case when we have more than 250 clients, or when we have
different classes of clients, having different limitations and quality of services. It is also possible to
have another network interface on the WAN side. In this case scarlet will have two different gateways
to the Internet and it can do load balancing between them and backup connection. Load balancing
means that part of the network traffic will go through one connection and the rest of it through the
other. Backup connection means that when one of the connections fails, the traffic is automatically

routed through the other one.

Illustration 1: Diagram of how SCARLET is used.
Scarlet gives also DHCP and DNS services to the computers of the LAN(s). It also has a firewall in
order to protect itself and the LAN(s) and has a web proxy (squid) in order to improve the quality and
efficiency of web access. All the clients that access the Internet through scarlet are registered. Scarlet
knows their MAC addresses and allows only the registered clients to access the Internet. When the
time of registration expires, the connection of the client is interrupted automatically. Scarlet can also
limit the download and upload speed of the Internet connection of the LAN. It also counts the network
traffic of each client (MAC) and builds statistics about the network usage.

3. Installing and Accessing Scarlet
In order to develop and test a gateway server we usually need a small lab with at least a few
machines and network connections. However, it is possible to simulate such a testing environment
using virtual machines and virtual network connections. For this purpose I have used QEMU. Other
options like VirtualBox, VMWare, etc. could have been used as well.

3.1. Installing QEMU
From the download page: http://fabrice.bellard.free.fr/qemu/download.html download and install
QEMU. The installation is very easy:

bash# tar xfz qemu-0.10.2.tar.gz
bash# cd qemu-0.10.2/
bash# ./configure && make && make install

The documentation of QEMU can be found at /usr/local/share/doc/qemu/, or at the web site, or using
`man qemu`.
The command `qemu` usually takes as parameter the image of the disk of the computer that is being

emulated. This is usually a file in the host system that is treated by qemu as the disk of the guest
computer. This image file can be created by the command `qemuimg` like this:

bash$ qemu-img create images/scarlet.img 3500MB

After this, we will have the file scarlet.img which can be used by qemu as the hard disk of the
guest computer. The size of this virtual hard disk is 3.5GB (sufficient for installing and testing the
gateway server). If you try now the command:

bash$ qemu images/scarlet.img

you will see the emulation window, and then an error message, saying that the disk is not bootable
(there is no operating system in it). This is right, because we have installed nothing in it yet.

3.2. Installing SCARLET
First of all we should download scarlet.iso from internet [...].
If we already have a scarlet CD, we can copy it like this:

bash# dd if=/dev/cdrom of=images/scarlet.iso

To install scarlet in the virtual disk scarlet.img use this command:

bash$ qemu -boot d -cdrom images/scarlet.iso images/scarlet.img

The option boot d tells qemu to boot from CDROM and the option cdrom tells it to use the given
ISO image as a CD. Press Enter several times, until you get to the prompt. Then type :

# install-scarlet.sh

This is the installation script. I swear it is not user friendly at all. It makes a professional installation for
sure; fast, efficient and straight to the point. It will partition the disk and then format the parts with
reiserfs, and then restore there the scarlet system. The partitions are restored from the backup files
scarlet.hda2.tar.gz and scarlet.hda5.tar.gz (we will see later how to make these backups). Let's have a
look at the script:

#!/bin/sh
### install scarlet

### create the partition table of /dev/hda
echo "Erasing the current partition table of /dev/hda"
dd if=/dev/zero of=/dev/hda count=63b
echo "Creating a new partition table for /dev/hda"
fdisk_commands="
nn pn 1n n +32n
nn pn 2n n +64n
nn en 3n n n
nn ln n +128n
nn ln n +128n
nn ln n n
tn 1n 82n

wn"
#echo -e $fdisk_commands ##debug
echo -e $fdisk_commands | fdisk /dev/hda
fdisk -l /dev/hda

### mount cdrom
mkdir /mnt/cdrom
mount /dev/hdc /mnt/cdrom

### install the maintenance system in /dev/hda2
echo "Formating /dev/hda2:"
mkreiserfs /dev/hda2
echo "Installing the maintenance system:"
mkdir /hda2
mount /dev/hda2 /hda2
cd /hda2
tar --extract --gunzip --preserve
--file=/mnt/cdrom/scarlet.hda2.tar.gz -v
cd /
echo "Installing lilo from the maintenance system:"
chroot /hda2 lilo
### install the scarlet system in /dev/hda5
echo "Formating /dev/hda5:"
mkreiserfs /dev/hda5
echo "Installing the scarlet-1 system:"
mkdir /hda5
mount /dev/hda5 /hda5
cd /hda5
tar --extract --gunzip --preserve
--file=/mnt/cdrom/scarlet.hda5.tar.gz -v
cd /

### format the swap partition
echo "Formating the swap partition:"
mkswap /dev/hda1

### start the maintenance script from the maintenance system
echo "Starting the maintenance script from the maintenance system:"
chroot hda2 /usr/local/maintenance/maintenance.sh

A usual installation of scarlet uses the disk as shown in the following diagram:
hda1 hda2 /dev/hda5 /dev/hda6 /dev/hda7
+---+--------+================+================+===========+
|swp| maintn | scarlet-1 | scarlet-2 |squid-cache|
|256| 512 MB | 1 GB | 1 GB | ... |
+---+--------+================+================+===========+

Scarlet itself is installed in two different partitions, which are labeled scarlet-1 and scarlet-2. There is
also a maintenance system, a swap partition and a partition used for the cache of squid. By default,
scarlet-1 is used, and scarlet-2 is a backup copy of it. The backup can be done from the maintenance
system. In case that something happens to the first scarlet, we can quickly boot to the other one, until
we find and fix the problem. The maintenance system is used to make backup/restore etc.
After installation, scarlet can be booted with the command:

bash$ qemu images/scarlet.img

(the option '-boot c' which says to boot from the HDD image is the default).

3.3. Using a Better Network Configuration
After booting scarlet like that, you will notice some error messages about device eth1 which is missing.
The lack of network interface eth1 will also cause some other problems, because the scarlet system
assumes that there are two network interfaces, one external (eth0) and one internal (eth1).
This problem is created because qemu by default creates only one network interface for the emulated
system. We can correct this problem by using advanced qemu options about network configuration.
Let's create some scripts for starting Scarlet with qemu.
start-scarlet.sh:

#!/bin/bash
### Start scarlet with two network cards.
### Note: it should be run by root, so that tapX interfaces
### can be created.

### make sure that the script is called by root
if [ $(whoami) != "root" ]
then
echo "Should be called by root"
exit
fi

### emulate a testbox with two network cards
qemu -m 64 -serial /dev/tty8 images/scarlet.img
-net nic,vlan=0 -net nic,vlan=1
-net tap,vlan=0,ifname=tap0,script=./tap0.sh
-net tap,vlan=1,ifname=tap1,script=./tap1.sh &

tap0.sh:

#!/bin/bash
/sbin/ip link set tap0 up
/sbin/ip address flush dev tap0
/sbin/ip address add 192.168.100.1/24 dev tap0

tap1.sh:

#!/bin/bash

There are a few qemu options to be explained here. First of all, m 64 tells qemu to use up to 64MB
RAM for the emulated scarlet. It really does not need more that this for testing. The option serial /
dev/tty8 tells qemu to connect the serial port of the emulated computer to the device /dev/tty8.

Press now Ctrl+Alt+F8, and you will access the serial console of scarlet (press Ctrl+Alt+F7 to get
back). You already know that the images/scarlet.img is the HDD of the emulated system.
The option net nic tells qemu to create a virtual network interface for the emulated (guest) system.
Since it is used twice, the emulated scarlet will have two network interfaces (eth0 and eth1). The
option net tap tells qemu to create a virtual network interface on the host system. Two tap
interfaces will be created, tap0 and tap1. The parameter vlan is used to create virtual LANs. The
parameter vlan=0 means that the network card should be connected to the virtual hub (or switch)
with number 0. So, all the network cards that have the same vlan number are connected to the same
hub/switch, being thus on the same LAN. We also tell qemu to use the script tap0.sh in order to
configure the virtual interface tap0 of the host, and the script tap1.sh for the interface tap1.
The logical diagram of the virtual network that is created by qemu is like this:

If we run `/sbin/ip addr` on the host, we see that the interface tap1 has address
192.168.10.101/24 and interface tap0 has two addresses: 192.168.100.1/24 and 10.0.0.10/24. If we
run `ip addr` inside scarlet , we will see something like this:

root@scarlet:/# ip addr
1: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
2: eth0: <BROADCAST,MULTICAST,UP> mtu 1500 qdisc htb qlen 1000
link/ether 52:54:00:12:34:56 brd ff:ff:ff:ff:ff:ff
inet 192.168.0.250/24 scope global eth0
3: eth1: <BROADCAST,MULTICAST,PROMISC,UP> mtu 1500 qdisc htb qlen
link/ether 52:54:00:12:34:57 brd ff:ff:ff:ff:ff:ff
root@scarlet:/#

In order to check the "physical" connections of the network that is built by qemu, press Ctrl+Alt+2 to
switch to the qemu monitor, then, in the (qemu) prompt, give the command info network. Finally
press Ctrl+Alt+1 to get back to the Linux console. This command should output something like this:

(qemu) info network
VLAN 0 devices:
tap: ifname=tap0 setup_script=./tap0.sh
ne2000 pci macaddr=52:54:00:12:34:56
VLAN 1 devices:
tap: ifname=tap1 setup_script=./tap1.sh
ne2000 pci macaddr=52:54:00:12:34:57
(qemu)

3.4. Accessing Scarlet From the Web Interface

Try to open http://192.168.10.1/ in a browser, and you will see this interesting page:

This is the screen that gets an internet client when he tries to access a web page but he is not
registered yet (or his registration has expired).
Try also http://192.168.10.1/admin/ and login with username 'superuser' and password 'admin'. This is
a screenshot from the administrative interface of Scarlet:

For more screenshots from Scarlet check the appendix.

3.5. Accessing Scarlet Through SSH

Besides accessing scarlet through the QEMU console and the serial port console (Ctrl+Alt+F8) we can
access it through ssh as well. However, for security reasons a gateway server has to be restrictive on
giving access, so the ssh can be accessed only on the port 2222 of the IP 10.0.0.1 on the interface
eth0. Try to login using the command:

bash$ ssh -p 2222 root@10.0.0.1

The password of root is 'admin'.
Accessing the emulated system through ssh is more convenient than accessing it through the QEMU
console and the serial console because it is possible to copy/paste (commands, configurations, etc.)

between the host system and the emulated system. However, if something goes wrong with the
network configuration or the firewall of the emulated system and we cannot login anymore, then we
can still access it through the serial console and the QEMU console.
While you are logged in to scarlet, go to /var/www/htdocs/scarlet

cd /var/www/htdocs/scarlet

This directory contains almost all the customizations that are done to a common Slackware system.
Most of the content of the directory itself is made up from the web application that is used to configure
and manage the gateway remotely. However, there is a subdirectory in it, called server-config, which
is used to configure the gateway server itself.
In this directory you will find the configuration file gateway.cfg which contains the configuration
parameters of the gateway server. Notice that they are the same configuration parameters that are
managed through the web interface. This configuration file is included in all the configuration scripts, in
order to get the parameters.
The configuration script update-config-files.sh updates several configuration files of the server:
/etc/resolv.conf, /etc/resolv.dnsmasq.conf, /etc/hosts, /etc/dnsmasq.conf, /etc/apache/httpd.conf and
/etc/squid/squid.conf. These configuration files are updated by replacing variables with configuration
parameters in template files (which can be found in the subdirectory templates/ ). But what are the
modifications that are made in these template files? How they are different from the original
configuration files? This can be seen in the .diff files in the templates directory.
The directory firewall/ contains the scripts that are used to build the firewall of the gateway server. This
firewall works also as a gateway keeper, because it allows only certain clients to have access to the
Internet.
The directory trafic/ contains scripts that are used to do traffic control. Traffic control makes sure that
the clients do not have download and upload speeds that are higher than the assigned limits, that
interactive traffic gets priority over the bulk traffic, that the connection resources are distributed fairly
between the clients, etc.
The directory traffic-counter/ contains scripts that are used to count the traffic (download and upload
traffic) of each client and to generate usage statistics for each of them.
Have a look at all these configuration files and scripts and try to understand how they work. If you
have difficulties understanding them, don't worry. They are a bit difficult and do contain some
advanced stuff. We will try to discuss them later in more details.

4. Preparing a Scarlet Installation CD
We used the CD image scarlet.iso in order to install Scarlet. But how is prepared this image? Suppose
that we have made some modifications/improvements to the Scarlet system. How can we prepare a
new version of scarlet.iso CD image?

4.1. Making a Scarlet ISO
The scripts that are used to prepare the Scarlet ISO can be checked out from the subversion
repository on SourceForge:

bash$ svn checkout
https://netaccess.svn.sourceforge.net/svnroot/netaccess/scarlet/install

The image scarlet.iso is prepared by the script make-scarlet-iso.sh . It is actually a modified Slackware
bootable CD (the first installation CD of Slackware 11). The modification is that all the Slackware
packages are removed from the CD and instead of them the backup files scarlet.hda2.tar.gz,
scarlet.hda5.tar.gz and wan.hda2.tar.gz are copied into it. Also initrd is modified so that the scripts
install-scarlet.sh and install-wan.sh are copied inside it.
make-scarlet-iso.sh:

#!/bin/sh
### Make the ISO image of the installation CD. This CD contains
### the backups scarlet.hda2.tar.gz, scarlet.hda5.tar.gz and
### wan.hda2.tar.gz. It also contains the installation scripts
### install-scarlet.sh and install-wan.sh .
### This CD is a Slackware bootable CD, without any slackware packages.
### After it is booted, the script install-scarlet.sh or the script
### install-wan.sh can be used to install the systems scarlet or wan.

if [ $(whoami) != "root" ]; then echo "Should be called by root"; exit; fi

### go to this directory
cd $(dirname $0)

### make initrd.img
./make-initrd.sh

### make backups of scarlet.hda2, scarlet.hda5 and wan.hda2
./make-system-backups.sh

### create the ISO image
mkisofs -R -J -v -d -N -hide-rr-moved -graft-points
-no-emul-boot -boot-load-size 4 -boot-info-table
-sort slack-cd1/isolinux/iso.sort
-b isolinux/isolinux.bin
-c isolinux/isolinux.boot
-V "Scarlet Install"
-A "Scarlet Install CD"
-o images/scarlet.iso
-x slack-cd1/isolinux/initrd.img
-x slack-cd1/slackware
slack-cd1/
scarlet-cd/
/=images/scarlet.hda2.tar.gz
/=images/scarlet.hda5.tar.gz
/=images/wan.hda2.tar.gz
/=install-scarlet.sh
/=install-wan.sh

### clean up
rm -rf scarlet-cd/

make-initrd.sh :

#!/bin/bash
### modify initrd by adding the scripts install-scarlet.sh
### and install-wan.sh


cd $(dirname $0)
path=$(pwd)

### copy the original initrd.img, unzip and mount it
cp slack-cd1/isolinux/initrd.img initrd-img.gz
gunzip initrd-img.gz
rm -rf initrd/
mkdir initrd/
mount -o loop initrd-img initrd/

### modify initrd by adding install-scarlet.sh and install-wan.sh
cp install-scarlet.sh initrd/
cp install-wan.sh initrd/
chown root.root initrd/install-scarlet.sh
chown root.root initrd/install-wan.sh
### copy initrd to scarlet-cd
umount initrd/
gzip initrd-img
rm -rf scarlet-cd/
mkdir -p scarlet-cd/isolinux/
mv initrd-img.gz scarlet-cd/isolinux/initrd.img

### clean
rmdir initrd

make-system-backups.sh :

#!/bin/bash
### make backups of the partitions hda2 of images/wan.img,
### hda2 of images/scarlet.img and hda5 of images/scarlet.img


### go to the image/ directory
cd $(dirname $0)
cd images/

### mount the second partition of wan.img to the directory wan.hda2/
/sbin/losetup -o 271434240 /dev/loop0 wan.img
mkdir -p wan.hda2/
mount /dev/loop0 wan.hda2/

### mount the second partition of scarlet.img to the directory scarlet.hda2/
/sbin/losetup -o 271434240 /dev/loop1 scarlet.img
mkdir -p scarlet.hda2/
mount /dev/loop1 scarlet.hda2/

### mount the partition 5 of scarlet.img to the directory scarlet.hda5/
/sbin/losetup -o 806109696 /dev/loop2 scarlet.img
mkdir -p scarlet.hda5/
mount /dev/loop2 scarlet.hda5/

### make backups of wan.hda2/, scarlet.hda2/ and scarlet.hda5/
tar --create --gzip --preserve --one-file-system --verbose
--file=wan.hda2.tar.gz --directory=wan.hda2/ .
--file=scarlet.hda2.tar.gz --directory=scarlet.hda2/ .
--file=scarlet.hda5.tar.gz --directory=scarlet.hda5/ .

### clean up
umount wan.hda2/
umount scarlet.hda2/
umount scarlet.hda5/
/sbin/losetup -d /dev/loop0
rmdir wan.hda2/
rmdir scarlet.hda2/
rmdir scarlet.hda5/

The way that we make the backups of the partitions hda2, hda5 etc. is quite interesting. We create
some loop devices, then we mount them and suddenly we are able to access the file systems in the
partitions that are inside the virtual disk. Then we just make a backup (archive) of the whole file
system there.
The way that the loop devices are created seems like magic, using some strange numbers. However
these numbers do have sense, they are the number of bytes (offset) where the partition that we want
to access starts. But how do we know what is the offset of a partition on the disk? We can find it out
from the partition table. Login to Scarlet, give the command `fdisk /dev/hda`, press 'u' to change
the units to sectors then press 'p' to print the partition table. You will see that the unit is a sector of 512
bytes, /dev/hda2 starts on unit 530145 and /dev/hda5 starts on unit 1574433. Then the partition /dev/
hda2 starts on byte 512*530145=271434240 of the disk, and the offset of /dev/hda5 is
512*1574433=806109696.

4.2. What Was Created First, Egg or Chicken?
At the beginning of the article we installed a scarlet system from scarlet.iso . On the previous section
we said that we make scarlet.iso from a scarlet system. Then, what was created first, a scarlet system
or scarlet.iso?
It is like the problem of egg and chicken. Eggs come out from chickens. Chickens come out from eggs.
Then, what was created first, eggs or chickens?
Another similar problem is the problem of compilers. All the executable programs have to be compiled
by a compiler. A compiler is an executable program. Then, what compiles the compiler? Maybe
another compiler? Then, what compiles the other compiler?
I am not sure about eggs-or-chickens and about compilers. However, in the case of Scarlet, I initially
installed a plain, simple and minimal Slackware system, using the first Slackware installation CD
(Slack11). Then I installed additional required packages and modified the system until it was working
as I wished. Then I created scarlet.iso in order to install this system easily to other machines.

The file dependencies.txt contains a list of packages that are required by scarlet, and the file
packages.txt contains a list of all the slackware packages that are installed in Scarlet (as listed in
/var/log/packages/). Both of these files are appended in the appendix.
Besides the usual Slackware packages, I have also installed the scarlet web interface and the
configuration/management scripts, like this:

root@scarlet:/# cd /var/www/htdocs/
root@scarlet:/var/www/htdocs# svn ls
https://netaccess.svn.sourceforge.net/svnroot/netaccess/scarlet/branches/
root@scarlet:/var/www/htdocs# svn checkout
https://netaccess.svn.sourceforge.net/svnroot/netaccess/scarlet/branches/release
-5 scarlet
root@scarlet:/var/www/htdocs# cd scarlet/
root@scarlet:/var/www/htdocs# vim sysconfig
root@scarlet:/var/www/htdocs# ./install.sh

The script install.sh itself uses other scripts that are located on install/, cron/, etc. Some of the
configuration files and scripts that are modified or created by it are these:
● /etc/ssh/sshd_config
● /etc/rc.d/rc.local
● /etc/inittab
● /etc/securetty
● /etc/mail/aliases
● /etc/cron.hourly/netaccess.check-time-limits.sh
● /etc/cron.daily/netaccess.check-traffic.sh
● /etc/cron.netaccess.weekly/maintain.sh
● /var/spool/cron/crontabs/root
● /etc/apache/php.ini
● /etc/apache/mod_php.conf
● /etc/HOSTNAME
● etc.

5. Building a Network Testing Virtual Environment
In order to test the gateway router we should install it in an old PC with at least two network cards,
using at least another PC as client, ensuring that it has two different gateways, making the necessary
network connections (with cables and hubs/switches), etc. This requires some resources. It also takes
some time to modify the environment for testing purposes, like adding another network card,
decreasing the RAM of the gateway PC, etc. So, it is more practical and efficient to build a virtual
testing environment. We have already emulated a virtual network with QEMU in the previous
sections. Here we are going to extend it further so that it can match better the network conditions in
which Scarlet is supposed to work. For this purpose we are going to use some shell scripts and
configurations. Using these scripts, all that is needed for testing is just a computer with sufficient RAM,

and optionally a network card for having access to the Internet.
Let us call the real computer the HOST system. Inside it we are going to emulate two guest systems.
On one of them, called TESTBOX, we will install Scarlet, and on the other one, called TESTBOX1, we
will install WAN. We can install both TESTBOX and TESTBOX1 using the scarlet.iso CD image, as
described previously. For installing TESTBOX we should run the script install-scarlet.sh, and for
installing TESTBOX1 we should run the script install-wan.sh . TESTBOX1 (WAN) will be used as a
second gateway for TESTBOX (SCARLET).
The figure below gives an idea about this situation: we have two guest systems which are
emulated and run simultaneously within the host system.

As we have seen already, QEMU gives also the possibility to emulate network interfaces of the
TESTBOX and TEXTBOX1 and to connect them to some virtual interfaces of the HOST and to
each-other. This figure shows the network connections that should be set up between TESTBOX,
TESTBOX1 and HOST:

TESTBOX has 3 network interfaces (eth0, eth1, eth2) which are "physically" connected to the virtual
interfaces tap01, tap1 and tap2 of the HOST. TESTBOX1 has the network interface eth0 which is
connected to tap02 of the HOST and there is also a connection between the interface eth0 of
TESTBOX and the interface eth1 of TESTBOX1. The arrows point in the direction of the network
gateways.
By properly configuring these "virtually-physical" network interfaces and connections, we can manage
to build a logical network like that which is shown in the following figure, which is sufficient for testing
the functionality of Scarlet.

The TESTBOX has two gateways to the Internet, the primary one (TESTBOX_GW1) is through
the interface eth1 of TESTBOX1 (which in the logical diagram is labeled WAN), and the secondary
one (TESTBOX_GW2) is through the interface tap01 of the HOST. The gateway of TESTBOX1
(WAN) is through the interface tap02 of the HOST. CLIENT1 is a client in the LAN1 (which is
connected to eth1 of TESTBOX) and CLIENT2 is a client in LAN2. Both of them use TESTBOX as
gateway. 'Admin' is a box that can be used to access TEXTBOX via ssh. CLIENT1, CLIENT2 and
Admin can be performed by HOST, which can be set in several roles.
So, we have built a network where the TESTBOX is a gateway server for two LANs, each of them
connected to one of its LAN interfaces. The HOST system can be configured to be on the WAN side of
the TESTBOX or on each of the LAN networks, in order to test it from each side. It can also access it
by ssh, using the Admin IP (because TESTBOX should deny shell access from outside, for security
reasons, except from the access IP). TESTBOX has also two gateways to the Internet, the primary
one through TESTBOX1 and the secondary one through HOST. TESTBOX1 itself is connected to the
Internet through HOST.

The scripts that are used to build a network testing virtual environment can be checked out from the
subversion repository at sourceforge:

bash$ svn checkout
https://netaccess.svn.sourceforge.net/svnroot/netaccess/scarlet/testing

The parameters that are used for the configuration of this virtual testing network are in the file
network.cfg. They correspond to the labels on the last diagram.

### the parameters that are used
### for the configuration of the test network

### gateways
HOST_GW=10.10.10.1 # gateway of the HOST to internet
TESTBOX_GW1=192.168.0.1 # eth1 of TESTBOX1
TESTBOX_GW2=192.168.100.1 # tap01 of HOST
TESTBOX1_GW=192.168.1.1 # tap02 of HOST
LAN1_GW=192.168.10.1 # eth1 of TESTBOX
LAN2_GW=192.168.11.1 # eth2 of TESTBOX

### IP-s
HOST_WAN_IP=${HOST_GW/%1/111}/24 # eth0 of HOST
HOST_LAN_IP1=${TESTBOX_GW2}/24 # tap01 of HOST
HOST_LAN_IP2=${TESTBOX1_GW}/24 # tap02 of HOST

TESTBOX_WAN_IP1=${TESTBOX_GW1/%1/250}/24 # eth0 of TESTBOX
TESTBOX_WAN_IP2=${TESTBOX_GW2/%1/250}/24 # eth0 of TESTBOX
TESTBOX_LAN1_IP=${LAN1_GW}/24 # eth1 of TESTBOX
TESTBOX_LAN2_IP=${LAN2_GW}/24 # eth2 of TESTBOX

CLIENT1_IP=${LAN1_GW/%1/101}/24 # tap1 of HOST (as eth0 of CLIENT1)
CLIENT2_IP=${LAN2_GW/%1/101}/24 # tap2 of HOST (as eth0 of CLIENT2)

ACCESS_IP=10.0.0.1/24 # eth0 of TESTBOX, used to access (login to) it
ADMIN_IP=10.0.0.10/24 # tap01 of HOST (as eth0 of ADMIN)

The gateways of the boxes are taken as primary parameters in the configuration of the network:
● HOST_GW -- gateway of HOST to the Internet.
● TESTBOX_GW1 -- primary gateway of TESTBOX (interface eth1 of TESTBOX1)
● TESTBOX_GW2 -- secondary gateway of TESTBOX (interface tap01 of HOST)
● TESTBOX1_GW -- gateway of TESTBOX1 (interface tap02 of HOST)
● LAN1_GW -- gateway of LAN1 (interface eth1 of TESTBOX)
● LAN2_GW -- gateway of LAN2 (interface eth2 of TESTBOX)
The other parameters are derived from them. For example, HOST_WAN_IP is derived from
HOST_GW by replacing 1 with 111 in the last part of the IP, and also appending a netmask '/24'. The
other IP-s are derived similarly.
In order to access the shell of TESTBOX, another network is created by adding another IP to the
interfaces tap01 of HOST and eth0 of TESTBOX. These IPs are kept in the parameters ADMIN_IP
and ACCESS_IP. This additional network is needed because, for security reasons, shell access to
TESTBOX is forbidden for the normal networks (WAN or LAN).
The configuration file network.cfg is usually included in all the scripts that deal with the configuration

of the network: setting IP-s, adding routes, default routes (gateways), etc. These scripts are
network.sh, tap01-setup.sh, tap02-setup.sh, tap1-setup.sh, tap2-setup.sh, etc.
The script network.sh is used to configure the interface eth0 of the HOST and the routes in the HOST.

#!/bin/bash

### include the config file
cd $(dirname $0)
. network.cfg

### setup eth0
/sbin/ip link set dev eth0 up
/sbin/ip address flush dev eth0
/sbin/ip address add $HOST_WAN_IP dev eth0

### add a default route
/sbin/ip route add to default via $HOST_GW

### setup different routing (gateway) for packets
### coming from clients (out of interfaces tap1 and tap2);
### these packets are marked with the mark 10 by the firewall
sed -e "/201 clients/d" -i /etc/iproute2/rt_tables
echo "201 clients" >> /etc/iproute2/rt_tables

/sbin/ip rule del fwmark 10 table clients 2> /dev/null
/sbin/ip rule add fwmark 10 table clients
/sbin/ip route del default via $HOST_GW dev eth0
table clients 2> /dev/null
/sbin/ip route add default via $HOST_GW dev eth0 table clients

The scripts tapX-setup.sh are used to configure the interfaces tap01, tap02, tap1 and tap2 of the
HOST. These interfaces are created by the emulation program (qemu), and the configuration scripts
are called by it as well.
tap01-setup.sh:

#!/bin/bash

cd $(dirname $0)
. network.cfg

### setup the interface
/sbin/ip address add $HOST_LAN_IP1 dev tap01
/sbin/ip address add $ADMIN_IP dev tap01

tap02-setup.sh:

#!/bin/bash


cd $(dirname $0)
. network.cfg

/sbin/ip address add $HOST_LAN_IP2 dev tap02

tap1-setup.sh:

#!/bin/bash

cd $(dirname $0)
. network.cfg

/sbin/ip address add $CLIENT1_IP dev tap1

tap2-setup.sh:

#!/bin/bash

cd $(dirname $0)
. network.cfg

/sbin/ip address add $CLIENT2_IP dev tap2

The emulation of TESTBOX, TESTBOX1 and the testing network is started by the script start-
testing.sh. It basically calls the program `qemu` with the appropriate parameters (for creating virtual
network cards etc.). It also modifies appropriately the existing firewall of HOST.
start-testing.sh:

#!/bin/bash
### Start the emulation of the gateway/router testing environment.
### Note: it should be run by root, so that the firewall can be modified
### and so that tapX interfaces can be created.


### get the system to be tested
testbox=${1:-images/netaccess.img}


cd $(dirname $0)

### accept in firewall tapX and enable source NAT
./firewall.sh enable

### emulate a testbox with three network cards
qemu -kernel-kqemu -m 64 -serial /dev/tty8 $testbox
-net nic,vlan=0 -net nic,vlan=1 -net nic,vlan=2
-net socket,vlan=0,listen=:1234
-net tap,vlan=0,ifname=tap01,script=./tap01-setup.sh
-net tap,vlan=2,ifname=tap2,script=./tap2-setup.sh &

### configure a route to ip 10.0.0.1 of tap01
sleep 10
./network.sh

### wait until the first testbox boots up
sleep 30

### emulate a second testbox with two network cards,
### which will be gateway for the first testbox
qemu -kernel-kqemu -m 64 images/wan.img
-net nic,vlan=0,macaddr=52:54:00:12:34:60
-net nic,vlan=1,macaddr=52:54:00:12:34:61
-net socket,vlan=1,connect=127.0.0.1:1234

### close the firewall for tapX and disable source NAT
./firewall.sh disable

The modification of the firewall is done/undone (enabled/disabled) by calling the script firewall.sh.
This script inserts in firewall rules for accepting input from the interfaces tap01, tap02, tap1 and
tap2. It also enables forwarding for the packets coming from the interface tap01 and tap02 and
enables source NAT (masquerading) for the packets that are forwarded through eth0, since it should
serve as a gateway for the TESTBOX and TESTBOX1.

firewall.sh:

#!/bin/bash
### accept/not-accept input from tapX and
### enable or disable source NAT (masquerading) for tap0

function usage
{
echo "Usage: ${0} [enable | disable | list]"
exit
}

### check that there is one parameter and get it
if [ -z $1 ]; then usage; fi
ACTION=$1

cd $(dirname $0)
. network.cfg

### variables
WAN_TAPS="tap01 tap02"
LAN_TAPS="tap1 tap2"
ALL_TAPS="$WAN_TAPS $LAN_TAPS"
IPT=/sbin/iptables

function insert_rules
{
### mark packets comming from wan tap interfaces
for tap in $WAN_TAPS
do
$IPT --table mangle --insert PREROUTING
--in-interface $tap --jump MARK --set-mark 10
done

### accept input from the tap interfaces
for tap in $ALL_TAPS
do
$IPT --table filter --insert INPUT --in-interface $tap
--jump ACCEPT
done

### accept forward from the wan tap interfaces
do
$IPT --table filter --insert FORWARD --in-interface $tap
--jump ACCEPT
done

### enable forwarding in the kernel
echo 1 > /proc/sys/net/ipv4/ip_forward

### enable source NAT
/sbin/iptables --table nat --insert POSTROUTING
--out-interface eth0 --jump MASQUERADE
}

function delete_rules
{
### del marking rules for packets coming from wan tap interfaces
do
$IPT --table mangle --delete PREROUTING
--in-interface $tap --jump MARK --set-mark 10
done

### delete input rules for tap interfaces
for tap in $ALL_TAPS
do
$IPT --table filter --delete INPUT --in-interface $tap
--jump ACCEPT
done

### delete forward rules for wan tap interfaces
do
$IPT --table filter --delete FORWARD --in-interface $tap
--jump ACCEPT
done

### disable forwarding in the kernel

### disable source NAT
/sbin/iptables --table nat --delete POSTROUTING
--out-interface eth0 --jump MASQUERADE
}

### list the existing rules
function list-rules
{
/sbin/iptables-save | grep -E 'MASQUERADE|tap'
}

case $ACTION in
enable ) insert_rules ;;
disable ) delete_rules ;;
* ) list-rules ;;
esac

After the emulation of the testboxes and the testing network has been started, the HOST system can
be placed in several roles inside this network (as it has been discussed above), in order to be able to
test better the TESTBOX. Switching of the roles is done by calling the script set-role.sh.
set-role.sh:

#!/bin/bash
### Sets the role of the main computer
### (by changing the default route etc.).


cd $(dirname $0)
. network.cfg

function usage()
{
echo "Usage: $0 [ls | normal | client1 | client2]"
echo "Sets the role of the HOST (by changing the default route)."
exit 0
}

function show_current_role
{
gateway=$(/sbin/ip route | grep default | cut -d' ' -f3)

case $gateway in
$HOST_GW ) echo "normal" ;;
$LAN1_GW ) echo "client1" ;;
$LAN2_GW ) echo "client2" ;;
* ) echo "unknown, gateway=$gateway" ;;
esac
}

function set_normal
{
### delete the table clients
/sbin/ip route flush table clients

### change the default route
/sbin/ip route del default 2> /dev/null
/sbin/ip route add default via $HOST_GW

### change resolv.conf as well
echo "nameserver $HOST_GW" > /etc/resolv.conf
}

function set_client1
{
### copy all routes of eth0 to table 'clients'
/sbin/ip route show | grep eth0
| while read ROUTE ; do /sbin/ip route add $ROUTE table clients ; done

### change the default routes
/sbin/ip route add default via $LAN1_GW
/sbin/ip route del default table clients 2> /dev/null
/sbin/ip route add default via $HOST_GW table clients

echo "nameserver $LAN1_GW" > /etc/resolv.conf
}

function set_client2
{
### copy all routes of eth0 to table 'clients'
/sbin/ip route show | grep eth0
| while read ROUTE ; do /sbin/ip route add $ROUTE table clients ; done

### change the default routes
/sbin/ip route add default via $LAN2_GW
/sbin/ip route del default table clients 2> /dev/null
/sbin/ip route add default via $HOST_GW table clients

echo "nameserver $LAN2_GW" > /etc/resolv.conf
}

if [ "$1" = "" ]; then usage; fi
role=$1

case $role in
ls ) show_current_role ;;
normal ) set_normal ;;
client1 ) set_client1 ;;
client2 ) set_client2 ;;
* ) usage ;;
esac

When HOST is set in the role of, say, CLIENT1, then its default gateway is changed to
LAN1_GW (through tap1) and its nameserver is changed to TESTBOX. So, HOST now, being in
stead of CLIENT1, accesses the Internet using TESTBOX as a gateway. On the other hand,
TESTBOX itself accesses the Internet using HOST as a gateway. Doesn't it sound like an infinite
loop?
In order to avoid this infinite loop, some special configurations are done. We want to make sure that
all the packets that come through the interfaces tap01/tap02 of HOST are always sent out to the
Internet through the interface eth0 of HOST (using HOST_GW as gateway), no matter what is their
source (TESTBOX or CLIENT1). This would break the network loop.
This is implemented by marking (in the firewall) all the packets that come through the interfaces
tap01/tap02, and then using a different routing table for the marked packets, which always has
HOST_GW as gateway no matter what is the role of HOST. See the scripts firewall.sh and
network.sh for the details of how this is done.

6. Configuration of Services
Let us discuss some details about the configuration of network and services in Scarlet (sshd, dns,
dhcpd, httpd, mysql, squid, etc). We can find out these configuration details by looking in the directory
/var/www/htdocs/scarlet/server-config/ and in its subdirectory templates/ .
All the configuration parameters are saved in a single file, gateway.cfg :

DOWNLOAD_LIMIT=750
UPLOAD_LIMIT=300
LAN_IP1=192.168.10.1/24
LAN_IP2=
LAN_IP3=
ENABLE_DHCPD=false
ENABLE_SQUID=true
OPEN_PORTS=true
PORT_LIST="21 25 53 80 110 119 222 443 1506 1507 1863 1972 2222 4899 5061 5050
5100 6667 9000 9010"
DNS1=192.168.0.1
DNS2=192.168.100.1
DNS3=
G2_PORTS=""
GATEWAY1=192.168.0.1
GATEWAY2=192.168.100.1
WAN_IP1=192.168.0.250/24
WAN_IP2=192.168.100.250/24

DOMAIN=localnet.net
ACCESS_IP=10.0.0.1/24

This file can be modified manually, if we login with ssh. It can also be modified from the admin web
interface of Scarlet. This configuration file is included in all the configuration scripts that use
configuration parameters.
After something is modified in gateway.cfg the script reconfig.sh must be called, in order to apply the
new settings:

#!/bin/bash
### reconfigure the network after changing
### any configuration variables

cd $(dirname $0)

### update the configuration files
./update-config-files.sh

### configure the server
./configure.sh

The script update-config-files.sh modifies the configuration files of the server, so that the new
configuration settings are used:

#!/bin/bash
### modify the configuration files after changing
### any configuration variables

cd $(dirname $0)

### include the configuration file
. gateway.cfg

### change the resolv.conf files
echo "nameserver $DNS1" > /etc/resolv.conf
echo "nameserver $DNS1" > /etc/resolv.dnsmasq.conf
if [ -n "$DNS2" ]
then
echo "nameserver $DNS2" >> /etc/resolv.conf
echo "nameserver $DNS2" >> /etc/resolv.dnsmasq.conf
fi
if [ -n "$DNS3" ]
then
echo "nameserver $DNS3" >> /etc/resolv.conf
echo "nameserver $DNS3" >> /etc/resolv.dnsmasq.conf
fi

### update /etc/hosts
LOCAL_IP=${LAN_IP1%/*}
sed -e "s/|DOMAIN|/$DOMAIN/g"

-e "s/|LOCAL_IP|/$LOCAL_IP/g"
templates/hosts > /etc/hosts

### change the configuration of dnsmasq
LAN_NETWORK=${LAN_IP1%.*/*}
sed -e "s/|DOMAIN|/$DOMAIN/g"
-e "s/|LAN_NETWORK|/$LAN_NETWORK/g"
templates/dnsmasq.conf > /etc/dnsmasq.conf
if [ "$ENABLE_DHCPD" = "false" ]
then
### comment any DHCP settings
sed -e "s+^dhcp+#dhcp+" -i /etc/dnsmasq.conf
fi

### change the configuration of HTTPD
APP_PATH=$(dirname $(pwd))
sed -e "s#|APP_PATH|#$APP_PATH#g"
templates/httpd.conf > /etc/apache/httpd.conf

### change the configuration of proxy (squid)
LAN1=${LAN_IP1%.*/*}.0/24
OUR_NETWORKS="$LAN1"
sed -e "s+|OUR_NETWORKS|+$OUR_NETWORKS+g"
templates/squid.conf > /etc/squid/squid.conf

The configuration files are modified by using the templates in the subdirectory templates/. Using sed
we replace any template variables with the values that are taken from gateway.cfg. The template
variables are denoted by starting and ending them with vertical bars, like this: |VAR_NAME| . After the
variables are replaced, the configuration file is written to the proper place.
The configuration files that are modified are these:
● /etc/resolv.conf
● /etc/resolv.dnsmasq.conf
● /etc/hosts
● /etc/dnsmasq.conf
● /etc/apache/httpd.conf
● /etc/squid/squid.conf
What modifications are made in these files and how they are different from the original (Slackware)
configuration files can be checked on templates/*.diff, which are also appended in the appendix.

7. Network Configuration
Scarlet can have two network gateways (network connections). The configuration of these two
connections is done using the parameters GATEWAY1, GATEWAY2, WAN_IP1, WAN_IP2 of the
configuration file gateway.cfg . The parameter G2_PORTS has a list of ports that are routed through
GATEWAY2. The rest of the network traffic is done through GATEWAY1. This is some kind of load
balancing between the network connections. In case that one of the gateways fails, then the other one
is used automatically. When the failed connection is back again, then the network traffic is adjusted

again so that both of the connections are used again. This is automatic backup connection (when a
connection fails, the network traffic is switched to the other one automatically).
The network configuration of scarlet is done by the script network.sh (which is called by configure.sh):

#!/bin/bash
### reconfigure the network after changing any parameters

### go to this dir
cd $(dirname $0)

### variables
IP=/sbin/ip

### include the configuration file
. ./gateway.cfg

### set up the links, in case that they are not up
$IP link set eth0 up
$IP link set eth1 up

### flush any existing ip addresses
$IP address flush eth0
$IP address flush eth1

### flush any existing routes of table main
$IP route flush table main
### add new addresses
$IP address add $WAN_IP1 dev eth0
$IP address add $WAN_IP2 dev eth0
$IP address add $LAN_IP1 dev eth1

### copy the routes of the main table to another routing table
$IP route flush table 2
$IP route show table main | grep -Ev ^default
| while read ROUTE ; do $IP route add table 2 $ROUTE ; done

### add the default routes
./set-default-routes.sh $GATEWAY1 $GATEWAY2

### use the second routing table for the packets marked with 2 by the firewall
$IP rule del fwmark 2 table 2 2>/dev/null
$IP rule add fwmark 2 table 2

### add the access IP to the external (WAN, eth0) interface
$IP address add $ACCESS_IP dev eth0

$IP route flush cache

Notice that the script creates two routing tables: the table main and the table 2. When the addresses
are added to the interfaces some routes are added implicitly to the main routing table. The script
copies all these routes (except the default route) to the second table. Then it sets a different gateway
(default route) for each routing table. Finally, a rule is added which says that we should use the
second routing table for all the packets that are marked by 2 from the firewall (and the main table for
the rest of packets). Since the second table has a different gateway than the first one, then these
packets go to the Internet passing through the second gateway.

The script firewall/mark-rules.sh builds the iptables' rules that decide which packets are marked by 2:

#!/bin/bash
### Mark with 2 the packets that should be routed through the gateway2.

### get variable $G2_PORTS from the configuration file
. ../gateway.cfg

IPT='/usr/sbin/iptables --table mangle'

$IPT --new-chain MARK-RULES
$IPT --append PREROUTING --jump MARK-RULES
$IPT --append OUTPUT --jump MARK-RULES

for PORT in $G2_PORTS
do
$IPT --append MARK-RULES --match tcp --protocol tcp
--destination-port $PORT --jump MARK --set-mark 0x2
done

It creates the chain of rules MARK-RULES in the table mangle, which is used in both PREROUTING
and OUTPUT. Then, for each port in G2_PORTS a new rule is added in the chain MARK-RULES,
which sets the mark 2 to the tcp packets that have such a port. The configuration parameter
G2_PORTS is defined in gateway.cfg and it is a list of space separated port numbers.
The default routes are set by the script set-default-routes.sh (which is called by network.sh):

#!/bin/bash
### (re)configure the default routes
### if only one parameter is given, then it is used for both gateways
### (gateway1 and gateway2 in this case are the same)

if [ "$1" = "" ]
then
echo "Usage: $0 gateway1 [gateway2]"
echo "If only gateway1 is given, then it is used as the second one as well"
exit 1
fi

GATEWAY1=$1
GATEWAY2=${2:-$GATEWAY1}

### change the default route of the main table
/sbin/ip route add default via $GATEWAY1

### change the default route of the second routing table
/sbin/ip route del table 2 default 2> /dev/null
/sbin/ip route add table 2 default via $GATEWAY2

### delete the cache of routes
/sbin/ip route flush cache

In order to check for the availability of the gateways, the script watch.sh is used:

#!/bin/bash
### Check periodicly the status of the gateways (whether they
### are dead or alive) and modify the network configuration
### accordingly, so that if one line goes down it is backup-ed
### by the other one.

cd $(dirname $0)

### read the config file
. gateway.cfg

### ping each gateway and find out whether it is on or off
function check-status
{
if ping -r -I eth0 -c 2 -W 1 $GATEWAY1 > /dev/null 2> /dev/null
then G1=on ; else G1=off
fi

if ping -r -I eth0 -c 2 -W 1 $GATEWAY2 > /dev/null 2> /dev/null
then G2=on ; else G2=off
fi
}

### compare the status with the status of the previous check,
### return true if the status of the gateways has changed
function status-changed
{
test "$G1" != "$G_1" -o "$G2" != "$G_2"
}

### change the default routes and the SNAT rules of the firewall
function change-network-config
{
#echo $G1 -- $G2 # debug

IP1=${WAN_IP1%/*}
IP2=${WAN_IP2%/*}

if [ "$G1" = "on" ]
then
gateway1=$GATEWAY1
ip1=$IP1
else
gateway1=''
ip1=''
fi

if [ "$G2" = "on" ]
then
gateway2=$GATEWAY2
ip2=$IP2
else
gateway2=''
ip2=''
fi

### if at least one of the gateways is on
### then change the config of the default routes and SNAT
if [ "$G1" = "on" -o "$G2" = "on" ]
then
./set-default-routes.sh $gateway1 $gateway2
firewall/source-nat.sh enable eth0 $ip1 $ip2
fi
}

### save the current status of the gateways
### in order to compare them in the next check
function save-current-status
{
G_1=$G1
G_2=$G2
}

### start an endless loop that periodicly checks the status of the interfaces,
### and if it is changed, updates the network config accordingly
while true
do
check-status

if status-changed
then
change-network-config
save-current-status
fi

sleep 6 # wait for a while
done

This script checks periodically the status of the gateways, and if there is any change (gateway online
or offline), it changes the configuration of the default routes (by calling set-default-routes.sh).
Note: Actually, just pinging the IP of the gateway is not enough for deciding whether the network
connection through this gateway is OK or not. The network connection may be broken somewhere
behind the gateway. This remains to be fixed and improved in the future.
The LAN computers are connected to the network using SNAT (Source Network Address Translation).
The script that manages SNAT rules is in the file firewall/source-nat.sh :

#!/bin/bash
### Enable or disable source NAT (masquerading).
### If two IPs are given, then the second one is used for
### the packets marked with 2, and the first one for the rest.
### If only one IP is given, then the second one is taken to be
### equal to the first one. If no IPs at all are given, then
### MASQUERADING is used instead.

IPT="/usr/sbin/iptables --table nat"

function usage
{

echo "Usage: $0 [enable|disable|list] [interface] [ip1] [ip2]"
echo " ip1 is used for the default traffic"
echo " ip2 is used for the packets that have the mark 2"
echo " If ip2 is missing, then ip1 is used as ip2."
echo " If both ip-s are missing, then masquerading is used instead."
exit
}

### check that there is one parameter
if [ -z $1 ]; then usage; fi

ACTION=$1
OUT_IF=${2:-eth0}
IP1=$3
IP2=${4:-$IP1}


### append snat rules
function append-snat-rules
{
delete-snat-rules 2>/dev/null

$IPT --new-chain SNAT-RULES
$IPT --append POSTROUTING --out-interface $OUT_IF
--jump SNAT-RULES

if [ -z $IP1 ]
then
$IPT --append SNAT-RULES --jump MASQUERADE
else
$IPT --append SNAT-RULES
--match mark --mark 0x2 --jump SNAT --to-source $IP2
$IPT --append SNAT-RULES --jump SNAT --to-source $IP1
fi
}

### delete snat rules
function delete-snat-rules
{
$IPT --delete POSTROUTING --out-interface $OUT_IF
--jump SNAT-RULES
$IPT --flush SNAT-RULES
$IPT --delete-chain SNAT-RULES
}

### list the existing SNAT rules
function list-snat-rules
{
/usr/sbin/iptables-save --table nat | grep 'SNAT-RULES'
}

case $ACTION in
enable ) append-snat-rules ;;
disable ) delete-snat-rules ;;

list ) list-snat-rules ;;
* ) usage ;;
esac

A new iptables chain, SNAT-RULES is created in the table nat. Then, if the packet is marked by 2,
then its source address is set to IP2, otherwise it is set to IP1. If IP2 is not given, then IP1 is used for
both marked and unmarked packets. If no IPs are given at all, then MASQURADING is used instead.

8. Managing Ports
In the configuration file gateway.cfg there are two variables about managing ports:

OPEN_PORTS=true
PORT_LIST="21 25 53 80 110 119 222 443 1506 1507 1863 1972 2222 4899 5061 5050
5100 6667 9000 9010"

These correspond to the section Ports of Settings in the web interface (see the screenshots in the
appendix).
If OPEN_PORTS is true, then all the ports are closed, except for the ones that are listed in the
PORT_LIST. If OPEN_PORTS is closed, then all the ports are closed, except for the ones that are
listed in the PORT_LIST. This logic is implemented by the script manage-ports.sh :

#!/bin/bash
### Manage ports, either allow all except the listed ones
### or forbid all except the listed ones.

IPT='/usr/sbin/iptables --table filter'

### get variables $PORT_LIST, $OPEN_PORTS
. ../gateway.cfg

function allow-ports
{
for PORT in $PORT_LIST
do
$IPT --append CHECK-PORTS --match tcp --protocol tcp
--destination-port $PORT --jump RETURN
$IPT --append CHECK-PORTS --match udp --protocol udp
--destination-port $PORT --jump RETURN
done
$IPT --append CHECK-PORTS --jump REJECT
}

function forbid-ports
{
for PORT in $PORT_LIST
do
$IPT --append CHECK-PORTS --match tcp --protocol tcp
--destination-port $PORT --jump REJECT
$IPT --append CHECK-PORTS --match udp --protocol udp
--destination-port $PORT --jump REJECT
done

$IPT --append CHECK-PORTS --jump RETURN
}

### create the chain CHECK-PORTS and jump to it
### for each packet that is forwarded out through eth0
$IPT --new-chain CHECK-PORTS
$IPT --insert FORWARD --out-interface eth0
--match tcp --protocol tcp --jump CHECK-PORTS
$IPT --insert FORWARD --out-interface eth0
--match udp --protocol udp --jump CHECK-PORTS

### call one of the functions allow-ports or forbid-ports,
### depending on the value of $OPEN_PORTS
if [ "$OPEN_PORTS" = "true" ] ; then allow-ports ; else forbid-ports ; fi

In order to manage the ports, the chain CHECK-PORTS is created in the table filter and is used for all
the packets that are sent out from the LAN through the WAN interface (eth0). When PORT_LIST
contains the ports that should be allowed, then the chain CHECK-PORTS contains a RETURN for
each port in the list (which means that the packet will continue to be processed by the firewall), and at
the end of the chain there is a REJECT for all the ports that are not in the list. When PORT_LIST
contains the ports that should be blocked, then the chain contains a REJECT for each port in the list,
and at the end of the chain there is a RETURN for all the ports that are not in the list, which means
that they will continue to be processed by the firewall.

9. Protecting the Gateway Server
The directory firewall/ contains the scripts that are related to iptables. Actually not all of the iptables
scripts are used to build a firewall, most of them are used to manage connections.
The rules of iptables are constructed by the script iptables.sh :

#!/bin/bash
### this script configures the firewall

cd $(dirname $0)

### include network configuration variables
. ../gateway.cfg

### update the list of allowed macs from DB
./get-allowed-macs.sh

IPT=/usr/sbin/iptables

### clear all the tables
$IPT --table filter --flush
$IPT --table filter --delete-chain
$IPT --table nat --flush
$IPT --table nat --delete-chain
$IPT --table mangle --flush
$IPT --table mangle --delete-chain

### accept by default anything in the OUTPUT chain
$IPT --table filter --policy OUTPUT ACCEPT

### set a mark to each packet that goes to internet
./mark-rules.sh

### forward ports
./forward-ports.sh

### redirect some ports (like 53, 80, etc.) to localhost
./redirect.sh

### enable SNAT-ing
IP1=${WAN_IP1%/*}
IP2=${WAN_IP2%/*}
./source-nat.sh enable eth0 $IP1 $IP2
#./source-nat.sh enable eth0

### the rules of the INPUT chain
./input-rules.sh

### the rules of the FORWARD chain
./forward-rules.sh

### if there is any argument, display the rules
if [ $# != 0 ]
then
/usr/sbin/iptables-save
fi

First of all, it cleans any existing rules, before constructing the new ones.
The script iptables-stop.sh can be used as well to clear all the iptables rules:

#!/bin/bash
### clear all the iptables rules

path=$(dirname $0)

### clear all the tables
$IPT --table filter --flush
$IPT --table filter --delete-chain
$IPT --table nat --flush
$IPT --table mangle --flush

### set default policies to ACCEPT
$IPT --table filter --policy INPUT ACCEPT
$IPT --table filter --policy OUTPUT ACCEPT
$IPT --table filter --policy FORWARD ACCEPT

Since the clients on the LAN use SNAT to access the Internet, they are automatically protected from
external access, because no body can access them directly. However the gateway server itself needs

to be protected both from external (WAN) and internal (LAN) sides. These protecting rules are placed
on the chain INPUT of iptables, which checks all the connections that come in to the server.
The rules of the INPUT chain are constructed by the script input-rules.sh :

#!/bin/bash
### the rules of the INPUT chain

. ../gateway.cfg

APPEND="$IPT --append INPUT"

### drop anything that does not match
$IPT --policy INPUT DROP

### accept anything from localhost
$APPEND --in-interface lo --jump ACCEPT

### accept any type of icmp
$APPEND --protocol icmp --icmp-type any --jump ACCEPT

### accept already established connections
$APPEND --match state --state ESTABLISHED,RELATED --jump ACCEPT

### create the chains WAN and LOCALNET for the connections
### comming from the WAN interface and from the LAN interfaces
$IPT --new-chain WAN
$IPT --new-chain LOCALNET
$APPEND --in-interface eth0 --jump WAN
$APPEND --in-interface ! eth0 --jump LOCALNET

### add the rules for these chains
./wan-rules.sh
./localnet-rules.sh

### reject anything else
$APPEND --jump REJECT --reject-with icmp-host-prohibited

A new chain, named WAN, is created and used for checking the connections that come from the
external interface (eth0), and another one, named LOCALNET, for the connections that come from the
LANs (internal interfaces, eth1 and eth2). The rules of these chains are constructed by the scripts
wan-rules.sh and localnet-rules.sh.
wan-rules.sh :

#!/bin/bash
### The rules of the chain WAN, which are applied to input from eth0,
### which is connected to the WAN network.

APPEND="$IPT --append WAN"

### accept httpd (80, 8080)
./port.sh 80 accept WAN

./port.sh 8080 accept WAN

### accept ssh (port 2222)
. ../gateway.cfg
ACCESS_IP=${ACCESS_IP%/*}
$APPEND --match state --state NEW --destination $ACCESS_IP
--match tcp --protocol tcp --destination-port 2222
--jump ACCEPT

### return to the previous chain
#$APPEND --jump RETURN
#
# This is commented because the default is to return to the previous chain.
# Also, if it is commented, the script 'port.sh' can be used to
# accept or reject any other ports (by appending or deleting rules).
# If it is uncommented, then any other rules that are added later
# will come after this one, and since this one matches anything, they
# will not be reached. The order of the rules does matter!

localnet-rules.sh :

#!/bin/bash
### The rules of the chain LOCALNET.

APPEND="$IPT --append LOCALNET"

### accept port 67 (dhcp)
./port.sh 67 accept LOCALNET tcp-udp

### accept port 53 (DNS)
./port.sh 53 accept LOCALNET tcp-udp

### accept port 80 (http)
./port.sh 80 accept LOCALNET

### accept port 3128 (proxy/squid)
./port.sh 3128 accept LOCALNET

### accept samba ports
#$IPT --new-chain SAMBA
#$APPEND --jump SAMBA
#./samba-rules.sh

#$APPEND --jump RETURN
#
# This is commented because the default is to return to the previous chain.
# Also, if it is commented, the script 'port.sh' can be used to
# accept or reject any other ports (by appending or deleting rules).
# If it is uncommented, then any other rules that are added later
# will come after this one, and since this one matches anything, they
# will not be reached. The order of the rules does matter!

In order to accept or deny certain ports, the script port.sh is used:

root@scarlet# ./port.sh
Usage: ./port.sh [port] [accept|deny|status] [chain] [proto]
where port is a number (like 80) or a service name (like http)
chain is: INPUT, FORWARD, LOCALNET, WAN, etc.
and proto is: tcp or udp or tcp-udp (default is tcp)

port.sh :

#!/bin/bash
### Accept (or deny, or show status of) tcp connections in the given port
### by appending or deleting a rule in the given chain.

function usage
{
echo "Usage: $0 [port] [accept|deny|status] [chain] [proto]
where port is a number (like 80) or a service name (like http)
chain is: INPUT, FORWARD, LOCALNET, WAN, etc.
and proto is: tcp or udp or tcp-udp (default is tcp)"
exit
}

if [ $# = 0 ]; then usage; fi

PORT=$1
ACTION=$2
CHAIN=${3:-INPUT}
PROTO=${4:-tcp}

## appends or deletes a rule, according to the parameter
function iptables-rule
{
command=$1

if [ $PROTO = "tcp" -o $PROTO = "tcp-udp" ]
then
$IPT --table filter --$command $CHAIN
--match state --state NEW
--match tcp --protocol tcp --destination-port $PORT
--jump ACCEPT
fi

if [ $PROTO = "udp" -o $PROTO = "tcp-udp" ]
then
$IPT --table filter --$command $CHAIN
--match udp --protocol udp --destination-port $PORT
--jump ACCEPT
fi
}

case $ACTION in

accept ) iptables-rule append ;;
deny ) iptables-rule delete ;;
* ) /usr/sbin/iptables-save | grep $CHAIN | grep "dport $PORT" ;;
esac

This script adds an ACCEPT rule for the port that should be accepted. If the port should be denied,
then it is sufficient to delete the ACCEPT rule that was added previously, since the default policy is to
deny everything that is not explicitly accepted.

There is also a script for accepting samba ports (in case that we want to share some folders on the
server with the local network):
samba-rules.sh :

#!/bin/bash
### The rules of the chain SAMBA.
### This chain is used to accept the samba ports.

### accept the udp ports 137 and 138
APPEND='/usr/sbin/iptables --table filter --append SAMBA'
$APPEND --match udp --protocol udp --dport 137 --jump ACCEPT
$APPEND --match udp --protocol udp --dport 138 --jump ACCEPT

### accept the tcp ports 139 and 445
./port.sh 139 accept SAMBA
./port.sh 445 accept SAMBA

$APPEND --jump RETURN

10. Managing Clients
The PCs on the LAN (clients) can connect to the Internet only if their MAC address is registered with
Scarlet and is in the list of the allowed MACs. The list of the allowed MACs is in the file scarlet/server-
config/allowed-macs , one MAC per line, like this:

52:54:00:12:34:56
00:0a:e4:c5:56:28
00-16-6f-81-5c-9c
00:FF:A0:4C:51:31

This list of macs is retrieved from the database of the web application by the script get-allowed-
macs.sh :

#!/bin/bash
### get a list of the macs that are allowed to connect to internet
### and save them in the file 'allowed-macs'

cd $(dirname $0)

### include the DB connection params
. ../../db/vars

### get them and save them in allowed-macs
mysql --host=$host --user=$user --password=$passwd
--database $database
--execute="select mac from macs where connected='true'"
--vertical
| grep '^mac:' | cut -d' ' -f2
> allowed-macs

The file allowed-macs is used in the script forward-rules.sh (which is called by iptables.sh) :

#!/bin/bash
### the rules of the FORWARD chain

. ../gateway.cfg

APPEND="$IPT --append FORWARD"


### the default is to drop anything that does not match
$IPT --policy FORWARD DROP

### accept any type of icmp
$APPEND --protocol icmp --icmp-type any --jump ACCEPT

### manage ports (allow or forbid them)
./manage-ports.sh

### accept already established connections
$APPEND --match state --state ESTABLISHED,RELATED --jump ACCEPT

### create the chain ALLOWED-MACS
$IPT --new-chain ALLOWED-MACS
for MAC in $(< allowed-macs)
do
$APPEND --match mac --mac-source $MAC --jump ALLOWED-MACS
done
### add rules for the chain ALLOWED-MACS
$IPT --append ALLOWED-MACS --jump ACCEPT

### reject anything else
$APPEND --jump REJECT --reject-with icmp-net-unreachable

The right place for checking the traffic between the clients (PCs in LAN) and the Internet is the chain
FORWARD. The connections that are already established are accepted. Then, for each MAC in the
list of allowed MACs, we add a rule that accepts the packets originating from that MAC. Everything

else is rejected.
However, the DNS and HTTP requests are handled differently. Scarlet serves as a transparent DNS
and HTTP proxy for the LAN, so all the DNS and HTTP packets are redirected to the localhost. This is
done by the script redirect.sh (which is called from iptables.sh) :

#!/bin/bash
### redirect some ports (like 53, 80, etc.) to localhost

IPT='/usr/sbin/iptables --table nat'

### get the configuration variables
. ../gateway.cfg

### redirect any DNS requests (port 53) to the localhost
$IPT --append PREROUTING --protocol udp --destination-port 53
--jump REDIRECT
$IPT --append PREROUTING --protocol tcp --destination-port 53
--jump REDIRECT

### handle the HTTP requests (port 80)
$IPT --new-chain HTTP
$IPT --append PREROUTING --protocol tcp --destination-port 80
--jump HTTP

### handle HTTP requests of the allowed macs
### in the chain HTTP-ALLOWED-MACS
$IPT --new-chain HTTP-ALLOWED-MACS
for MAC in $(< allowed-macs)
do
$IPT --append HTTP --match mac --mac-source $MAC
--jump HTTP-ALLOWED-MACS
done

### the rules of the chain HTTP-ALLOWED-MACS
if [ "$ENABLE_SQUID" = "false" ]
then
### direct connection to internet
$IPT --append HTTP-ALLOWED-MACS --jump ACCEPT
else
### don't use proxy for the connections to the local server
local_ips="$WAN_IP1 $WAN_IP2 $LAN_IP1 $LAN_IP2 $LAN_IP3"
for ip in $local_ips
do
local_ip=${ip%/*} ### remove /24 from the ip
$IPT --append HTTP-ALLOWED-MACS --destination $local_ip
--jump ACCEPT
done
### connection through transparent proxy (squid)
$IPT --append HTTP-ALLOWED-MACS --protocol tcp
--jump REDIRECT --to-port 3128
fi

### redirect HTTP requests of the others (not in HTTP-ALLOWED-MACS)
### to the localhost
$IPT --append HTTP --jump REDIRECT

All the packets with destination port 53 (DNS requests) are jumped to REDIRECT, which means that
their destination address is changed so that they end up to localhost. This is done in the
PREROUTING chain, which is checked and applied before routing is done.
Then a new chain named HTTP is created and all the http packets are send to this chain. In this chain
are added rules that match each allowed mac and send such packets to the chain HTTP-ALLOWED-
MACS. The HTTP packets that do not originate from the allowed macs will match the last rule of the
HTTP chain and will be redirected to the localhost. In the chain HTTP-ALLOWED-MACS we either
accept just the packets, if the transparent proxy is not enabled, or we redirect them to the port 3128 (of
squid), if the transparent proxy is enabled. However we are careful not to send to squid the requests to
the local webserver.

11. Advanced Connection Management
Actually these things are not more complicated that what we have already seen, however they are
called “advanced” because they have to be done manually, by calling the appropriate scripts (while the
other things are managed either automatically or from the web interface). The reason for this is that
they may be needed only in rare cases.

11.1. Port Forwarding
We need to do port forwarding when we have some servers in the LAN (for example a web server),
behind the gateway server. Since the gateway server uses SNAT, then the web server cannot be
accessed from the Internet. However, if we forward the port 80 to the web server, then all the http
requests that go to the gateway server will be forwarded back to the web server. The browser will think
that it gets the answer from the gateway server, although actually it is the web server that sends
replies to it.
Port forwarding in scarlet can be done by calling the script port-forward.sh :

root@scarlet # ./port-forward.sh
Usage: ./port-forward.sh PORT EXT_IP INT_IP SERVER_IP [enable|disable|status]
PORT is the port that will be forwarded (like 80 or http)
EXT_IP is the external IP of the server
INT_IP is the IP of through which it will be forwarded
SERVER_IP is the IP of the server to which it will be forwarded

#!/bin/bash
### enables or disables port forwardings

function usage
{
echo "Usage: $0 PORT EXT_IP INT_IP SERVER_IP [enable|disable|status]
PORT is the port that will be forwarded (like 80 or http)
EXT_IP is the external IP of the server
INT_IP is the IP of through which it will be forwarded
SERVER_IP is the IP of the server to which it will be forwarded"
exit
}

### check that there are at least 4 parameters
if [ $# -lt 4 ]; then usage; fi

### get the parameters and remove any netmask from IP-s
PORT=$1
EXT_IP=${2%/*}
INT_IP=${3%/*}
SERVER_IP=${4%/*}
ACTION=$5

### append or delete the forward rules according to the parameter
function iptables-fwd-rules
{
command=$1

# accept the port for forwarding (both tcp and udp)
$IPT --table filter --$command FORWARD
--match state --state NEW
--match tcp --protocol tcp --dport $PORT
--jump ACCEPT
$IPT --table filter --$command FORWARD
--match udp --protocol udp --dport $PORT
--jump ACCEPT

### forward the port
$IPT --table nat --$command PREROUTING
--protocol tcp --destination $EXT_IP --dport $PORT
--jump DNAT --to-destination $SERVER_IP
$IPT --table nat --$command PREROUTING
--protocol udp --destination $EXT_IP --dport $PORT
--jump DNAT --to-destination $SERVER_IP

### in case that $INT_IP is not gateway for $SERVER_IP
$IPT --table nat --$command POSTROUTING
--protocol tcp --destination $SERVER_IP --dport $PORT
--jump SNAT --to-source $INT_IP
}

case $ACTION in
enable ) iptables-fwd-rules append ;;
disable ) iptables-fwd-rules delete ;;
* ) /usr/sbin/iptables-save | grep $SERVER_IP | grep "dport $PORT" ;;
esac

This script can be called from another script, like this:
forward-ports.sh :

#!/bin/bash
### ports that are forwarded to local servers


. ../gateway.cfg

### forward port 443 to 192.168.255.2
./port-forward.sh 443 $WAN_IP1 $LAN_IP1 192.168.255.2 enable

11.2. IP Forwarding
In this case we assign an additional IP to the WAN interface of the gateway server. However, all the
traffic that comes to this IP is forwarded to another server inside the LAN. So, this server is logically on
the WAN network, although physically it is on the LAN network.

root@scarlet:# ./ip-forward.sh
Usage: ./ip-forward.sh EXTERNAL_IP INTERNAL_IP SERVER_IP
EXTERNAL_IP is the IP that will be forwarded
INTERNAL_IP is the IP through which it is forwarded
SERVER_IP is the IP to which it is forwarded

ip-forward.sh :

#!/bin/bash
### forward an IP to another server

function usage
{
echo "Usage: $0 EXTERNAL_IP INTERNAL_IP SERVER_IP
EXTERNAL_IP is the IP that will be forwarded
INTERNAL_IP is the IP through which it is forwarded
SERVER_IP is the IP to which it is forwarded"
exit
}


### get the parameters and remove any netmask from IP-s
EXTERNAL_IP=${1%/*} # IP that will be forwarded
INTERNAL_IP=${2%/*} # IP through which it is forwarded
SERVER_IP=${3%/*} # IP to which it is forwarded

IPT='/usr/sbin/iptables'

### accept for forward everything related with $SERVER_IP
$IPT --table filter --append FORWARD
--destination $SERVER_IP --jump ACCEPT
--source $SERVER_IP --jump ACCEPT

### forward everything with destination $EXTERNAL_IP to $SERVER_IP
$IPT --table nat --append PREROUTING

--destination $EXTERNAL_IP --jump DNAT --to-destination $SERVER_IP

### in case that $INTERNAL_IP is not gateway for $SERVER_IP
### if this is uncomented, then the internal server does not see
### anything else, except $INTERNAL_IP
#$IPT --table nat --append POSTROUTING
# --destination $SERVER_IP --jump SNAT --to-source $INTERNAL_IP

### everything comming from $SERVER_IP should be
### sent out to internet with source $EXTERNAL_IP
#$IPT --table nat --append POSTROUTING
# --out-interface eth0 --source $SERVER_IP
# --jump SNAT --to-source $EXTERNAL_IP

11.3. Source Forwarding
All the traffic that comes from a certain network can be forwarded to a certain internal server. There
may be cases when such a thing is needed.

root@scarlet:l# ./source-forward.sh
Usage: ./source-forward.sh SOURCE_NET DESTINATION_IP
SOURCE_NET the source network (e.g. 192.168.1.0/25)
DESTINATION_IP the destination server

source-forward.sh :

#!/bin/bash
### forward all the packets coming from a given source network
### to another server

function usage
{
echo "Usage: $0 SOURCE_NET DESTINATION_IP
SOURCE_NET the source network (e.g. 192.168.1.0/25)
DESTINATION_IP the destination server"
exit
}


SOURCE_NET=$1 # the source network
DESTINATION_IP=${2%/*} # the destination server

IPT='/usr/sbin/iptables'

### accept for forward everything comming from the $SOURCE_NET
--source $SOURCE_NET --jump ACCEPT

--destination $SOURCE_NET --jump ACCEPT

### forward everything with source $SOURCE_NET to $DESTINATION_IP
$IPT --table nat --append PREROUTING
--source $SOURCE_NET --jump DNAT --to-destination $DESTINATION_IP

12. Controlling the Traffic (tc)
The scripts that are used for traffic control are located in the directory trafic/ . They basically use the
utility /sbin/tc . The script that is used to set up the traffic rules is set-rate-limits.sh :

#!/bin/bash
### update the upload and download limits

### go to this dir
cd $(dirname $0)

### get the upload and download rates
. ../gateway.cfg

### set upload limit
if [ "$UPLOAD_LIMIT" = "" ]
then
./rate-limit.sh off eth0
else
./rate-limit.sh on eth0 $UPLOAD_LIMIT
fi

### set download limit
if [ "$DOWNLOAD_LIMIT" = "" ]
then
./rate-limit.sh off eth1
else
./rate-limit.sh on eth1 $DOWNLOAD_LIMIT
fi

It gets the variables $UPLOAD_LIMIT and $DOWNLOAD_LIMIT from the configuration file
../gateway.cfg and calls the script rate-limit.sh to set these rates.

root@scarlet: # ./rate-limit.sh
Usage: ./rate-limit.sh [on | off | ls] [dev] [rate]
dev can be eth0, eth1, etc.
rate can be 48, 96, etc.

This script sets a rate limit in one of the network interfaces
(eth0, eth1, etc.). If the the limit is set in the external interface
then it is an upload limit; if it is set on the LAN interface, then

it is a download limit.

rate-limit.sh :

#!/bin/bash
### This script sets a rate limit in one of the network interfaces
### (eth0, eth1, etc.). If the the limit is set in the external interface
### then it is an upload limit; if it is set on the LAN interface, then
### it is a download limit.

### Set the rate limit to your uplink's *actual* speed, minus a few percent
### If you have a really fast modem, raise 'burst' a bit.

function usage
{
echo "Usage: $0 [on | off | ls] [dev] [rate]"
echo " dev can be eth0, eth1, etc."
echo " rate can be 48, 96, etc."
echo
echo "This script sets a rate limit in one of the network interfaces"
echo "(eth0, eth1, etc.). If the the limit is set in the external interface"
echo "then it is an upload limit; if it is set on the LAN interface, then"
echo "it is a download limit."
echo
exit
}

function set_limit
{
### check that the rate (third parameter) is supplied
if [ "$RATE" = "" ]
then
echo
echo "Please supply a rate as well..."
echo
usage
exit 1
fi

### delete any existing settings
/sbin/tc qdisc del dev $DEV root 2>/dev/null >/dev/null

### add the HTB qdisc with just one class (used for limiting the rate)
/sbin/tc qdisc add dev $DEV root handle 1: htb default 1
/sbin/tc class add dev $DEV parent 1: classid 1:1 htb rate ${RATE}kbit burst
5k

### add a PRIO qdisc below the HTB (in order to
### prioritize interactive traffic)
/sbin/tc qdisc add dev $DEV parent 1:1 handle 10: prio
# automatically creates classes 10:1,, 10:2, 10:3

### add a SFQ qdisc below each priority class (to handle connections fairly)
/sbin/tc qdisc add dev $DEV parent 10:1 handle 11: sfq perturb 10


### Note: with these classes and queues as above, there should be
### no need for filters.

### apply the new rate limit
#/sbin/tc qdisc add dev $DEV root tbf rate ${RATE}kbit latency 50ms burst 3000
}

### get the arguments
ACTION=$1
DEV=$2
RATE=$3

TC="/sbin/tc qdisc"
case $ACTION in
on ) set_limit ;;
off ) /sbin/tc qdisc del dev $DEV root 2>/dev/null >/dev/null;;
ls ) /sbin/tc -s qdisc ls dev $DEV ;;
* ) usage ;;
esac

This script limits the rate on the given interface. It also creates three priority queues on the interface,
where the one with the highest priority is for the interactive traffic, and the one with the most low
priority for the bulk traffic.

13. Counting the Traffic
The traffic is counted for each MAC in the LAN and statistics are generated (using RRD tools). The
scripts that count the traffic and generate the statistics are located in the directory server-config/traffic-
counter/ . Initially an RRD (Round Robin Database) file is created for each MAC that is allowed to
access the internet. This is accomplished by the scripts init.sh and rrdinit.sh :
init.sh :

#!/bin/bash
### initialize the databases for each allowed mac

rm -rf rrd/
mkdir -p rrd
for mac in $(< ../firewall/allowed-macs)
do
./rrdinit.sh $mac
done

rrdinit.sh :

#!/bin/bash
### Create a RRD for the MAC that is given as argument.
### This database keeps a DS for download and one for upload.

### The interval in which the data is fed into the database is 10 sec.
### and the type of the data is ABSOLUTE (after feeding the data into
### the database the counters that count upload and download traffic
### are reseted to 0).
###
### The round robin archives that are kept in the database are these:
###
### - One that keeps each data that is inputed and has 60 items
### (time interval 10s, time span 10m.)
###
### - One that consolidates 30 primary values for each item and has 24 items
### (time interval 300s = 5m, time span 120m = 2h)
###
### - One that consolidates 360 values for each item and has 48 items
### (time interval 3600s = 1h, time span 48h = 2 days)
###
### - One that averages 360*24=8640 values and has 240 items
### (time interval 1h*24 = 1day, time span 1day * 240 = 8 months)

if [ "$1" = "" ]
then
echo "Usage: $0 mac"
exit 1
fi

mac=$1
dbname="mac_${mac//[:-]/}"

cd $(dirname $0)

### don't overwrite an existing database
if [ -f "rrd/$dbname" ] ; then exit ; fi

### calculate start time to be a multiple of 86400 (one day)
now=$(date +%s)
day=86400
start=$(( $now / $day * $day ))

### create a new database
rrdtool create rrd/$dbname
--start $start --step 10
DS:download:ABSOLUTE:30:0:U
DS:upload:ABSOLUTE:30:0:U
RRA:AVERAGE:0.5:1:60

To learn more about RRD and its tools, have a look at its documentation on /usr/share/doc/rrdtool-
1.2.8/ . You can also read the man pages ( man rrdtool , man rrdtutorial , etc.)

After initializing the RRD files, we can start counting the traffic for each MAC. Counting is started by

the script start.sh :

#!/bin/bash

cd $(dirname $0)

### clean up any pid files
rm -f *.pid

### start tcpdump
#./tcpdump.sh | ./counter.sh & # debug
./tcpdump.sh | ./counter.sh | rrdtool - >/dev/null &

### start timer
./timer.sh &

Here we see some scripts and commands processing data continuously in a pipeline.
The script tcpdump.sh listens the network traffic on the interface eth1 and outputs it on a short
(concise) format, one line for each packet:

#!/bin/bash

### run tcpdump on the LAN interface
/usr/sbin/tcpdump -i eth1 -e -n -nn -N -tt -q &

### save the process id to a file
echo $! > tcpdump.pid

###test
#echo $$ > tcpdump.pid
#while true
#do
# cat test/tcpdump.output
# sleep 2
#done

The output that is produced by this script looks like this:

1245873613.101635 52:54:00:12:34:56 > 00:ff:70:3a:98:18, IPv4, length 226:
10.0.0.1.2222 > 10.0.0.10.39877: tcp 160
1245873613.102139 52:54:00:12:34:56 > 00:ff:70:3a:98:18, IPv4, length 226:
10.0.0.1.2222 > 10.0.0.10.39877: tcp 160
1245873613.102781 00:ff:70:3a:98:18 > 52:54:00:12:34:56, IPv4, length 66:
10.0.0.10.39877 > 10.0.0.1.2222: tcp 0
1245873613.103568 00:ff:70:3a:98:18 > 52:54:00:12:34:56, IPv4, length 66:
10.0.0.10.39877 > 10.0.0.1.2222: tcp 0

This output is fed by the pipeline to the script count.sh , which counts the upload/download bytes for
each MAC:

#!/bin/bash
### This script watches the network traffic by reading the output
### of tcpdump through a pipe. It processes this output and accumulates
### the bytes used for upload/download by each MAC.
### It also receives periodically a sigalarm (14) from the script
### 'timer.sh'. When this signal is received, the number of bytes that
### is counted for each MAC is saved to a file or to a database.

### create upload/download counters (variables) with value 0
### for each database in the directory 'rrd/'
function init_counters()
{
db_list=$(ls rrd/)
for dbname in $db_list
do
eval ${dbname}_upload=0
eval ${dbname}_download=0
done
}

### save the values of the counters in the databases
function dump_counters()
{
now=$(date +%s)
now=$(($now / 10 * 10))
echo
for dbname in $db_list
do
eval download=$${dbname}_download
eval upload=$${dbname}_upload
echo update rrd/$dbname $now:$download:$upload
done

init_counters
}

### ----------------------------------------------------------

echo $$ > counter.pid

### init counters
init_counters

### call the function dump_counters when the sigalarm is received
trap dump_counters 14

### get the mac of the LAN interface (eth1)
MY_MAC=$(/sbin/ip addr | sed -n -e '/eth1:/,+1 p' | gawk '/link/ {print $2}')

### read package logs from standard input and process them
while read line
do
### get from the line $mac1, $mac2 and $bytes
mac1=${line:18:17}
mac2=${line:38:17}

bytes=${line:70:10}
bytes=${bytes%%:*}

### increment the counters
if [ "$mac1" = "$MY_MAC" ]
then
dbname="mac_${mac2//[:-]/}"
eval let ${dbname}_download+=$bytes
elif [ "$mac2" = "$MY_MAC" ]
then
dbname="mac_${mac1//[:-]/}"
eval let ${dbname}_upload+=$bytes
fi

### debug
#echo "$mac1 > $mac2 $bytes"
done

This script keeps two counters for each MAC, one for upload and one for download. Each line that
comes from tcpdump.sh is processed and the counters are incremented accordingly. Whenever the
script gets the signal 14 (sigalarm), it calls the function dump_counters which writes the current values
of the counters to RRD files and initializes them again (resets the counters). The signal 14 is sent
periodically each 10 seconds by the script timer.sh :

#!/bin/bash
### send periodically a sigalarm to the counter

period=10

echo $$ > timer.pid

### get the process id of the counter
while [ "$pid" = "" ]
do
pid=$(cat counter.pid 2> /dev/null)
done

### send a sigalarm (14) to the counter periodically
while true
do
sleep $period
kill -s 14 $pid
done

The files counter.pid, tcpdump.pid and timer.pid keep the process id of the scripts. They are used to
send signals to each other and also to stop them. Counting the traffic can be used by the script
stop.sh :

#!/bin/bash

### stop timer
pid=$(cat timer.pid)

kill -9 $pid

### stop tcpdump
pid=$(cat tcpdump.pid)
kill -9 $pid

### clean up any pid files
rm -f *.pid

The script counter.sh dumps the counters to the database by outputting update command to the
pipeline, which are then read and executed by the rrdtool . If we called rrdtool directly from the script
counter.sh this would be much less efficient, because each time that it is called, it has to be loaded
and initialized in RAM, before being executed.
After the statistics have been collected, we can generate graphs using the script rrdgraph.sh :

root@scarlet:# ./rrdgraph.sh
Usage: ./rrdgraph.sh fname interval [ mac]+
Generates the traffic usage of one or more macs in an interval.
'fname' is the path and name of the image that will be created
'interval' is: 10minutes | 2hours | 2days | 1month | 8months
One or more MACs should be supplied, separated by space.

rrdgraph.sh :

#!/bin/bash

function usage
{
echo "Usage: $0 fname interval [ mac]+"
echo " Generates the traffic usage of one or more macs in an interval."
echo " 'fname' is the path and name of the image that will be created"
echo " 'interval' is: 10minutes | 2hours | 2days | 1month | 8months"
echo " One or more MACs should be supplied, separated by space."
exit 1
}

if [[ $# -lt 3 ]] ; then usage ; fi

cd $(dirname $0)

### get file name and the time interval to be graphed
fname=$1 ; shift
interval=$1 ; shift

### get the resolution
case $interval in
10minutes ) resolution=10 ;; # 10 sec
2hours ) resolution=300 ;; # 5 min

2days ) resolution=3600 ;; # 1 hour
1month ) resolution=86400 ;; # 1 day
8months ) resolution=86400 ;; # 1 day
* ) usage ;;
esac

### get start and end times
now=$(date +%s)
end=$(($now/$resolution*$resolution))
start="end-$interval"

### process the MACs
nr_macs=0
DEFs=''
upload='0'
download='0'

while [ "$1" != "" ]
do
let nr_macs++
mac=$1 ; shift
mac=${mac//[:-]/}

DEFs="$DEFs DEF:upload_$mac=rrd/mac_$mac:upload:AVERAGE"
DEFs="$DEFs DEF:download_$mac=rrd/mac_$mac:download:AVERAGE"

upload="upload_$mac,$upload,+"
download="download_$mac,$download,+"
done

### get the average
#upload="$upload,$nr_macs,/"
#download="$download,$nr_macs,/"

### convert from Bytes/s to Bits/s
upload="$upload,8,*"
download="$download,8,*"

### generate the graphic image
rm -f $fname
#echo
rrdtool graph $fname
--imgformat PNG --interlaced --end $end --start $start
--width 550 --vertical-label Bits/s
$DEFs
CDEF:upload=$upload
CDEF:download=$download
AREA:upload#00EEEE:'Upload'
LINE1:upload#00CCAA
LINE1:download#FF0000:'Download'

14. Summary and Future Work

15. Appendix

15.1. Screenshots From SCARLET
Here are listed some screenshots from the Scarlet web interface.

This is the screen that gets an internet client when he tries to access a web page but he is not
registered yet (or his registration is expired).

A client can have many connections. He registers them using his username and access code. His
connections have a time expiration and an upload/download limit.

Scarlet generates upload/download traffic statistics for each connection of the client and for all of them
in total. It also generates usage graphs for different periods of time (hourly, daily, monthly, etc).

The configuration of the WAN side of Scarlet. It can use two different gateways to the Internet.

Scarlet can block certain network ports.

Scarlet can have more than one LAN interface.

Scarlet can limit the bandwidth of the network traffic.

Scarlet can also use transparent proxy, can give DHCP service, etc.

Scarlet has some interesting administrative features, like backup, update, upgrade, etc.

Every important action is logged and the logs can be filtered (searched) and printed.

Scarlet also has a flexible statistics module.

15.2. Slackware Packages Installed in Scarlet

dependencies.txt :

apache
cyrus-sasl
diffutils
dnsmasq
gawk
gettext-tools
glibc-i18n
iproute2
iptables
isapnptools
lftp
libart_lgpl
libidn
libxml2
mhash
mysql
openssh
openssl
php
procmail
sendmail
stunnel
subversion
wget
wireless-tools

packages.txt :

aaa_base-11.0.0-noarch-2
aaa_elflibs-11.0.0-i486-9
acl-2.2.39_1-i486-1
acpid-1.0.4-i486-2
apache-1.3.37-i486-2
apmd-3.0.2-i386-1
apr-1.2.7-i486-1
apr-util-1.2.7-i486-1
at-3.1.10-i486-1
attr-2.4.32_1-i486-1
bash-3.1.017-i486-1
bin-11.0-i486-3
bind-9.3.2_P1-i486-1
bzip2-1.0.3-i486-3
coreutils-5.97-i486-1
curl-7.15.5-i486-1
cxxlibs-6.0.3-i486-1
cyrus-sasl-2.1.22-i486-1

dcron-2.3.3-i486-5
devs-2.3.1-noarch-25
dhcp-3.0.4-i486-2
dhcpcd-2.0.4-i486-2
diffutils-2.8.1-i486-3
dnsmasq-2.33-i486-1
e2fsprogs-1.38-i486-2
elvis-2.2_0-i486-2
etc-11.0-noarch-2
findutils-4.2.28-i486-1
gawk-3.1.5-i486-3
genpower-1.0.5-i486-1
gettext-0.15-i486-1
gettext-tools-0.15-i486-1
glibc-i18n-2.3.6-noarch-6
glibc-solibs-2.3.6-i486-6
glibc-zoneinfo-2.3.6-noarch-6
gnupg-1.4.5-i486-1
grep-2.5-i486-3
groff-1.19.2-i486-1
gzip-1.3.5-i486-1
hdparm-6.6-i486-1
hotplug-2004_09_23-noarch-11
inetd-1.79s-i486-7
iproute2-2.6.16_060323-i486-1
iptables-1.3.5-i486-2
iptraf-2.7.0-i386-1
isapnptools-1.26-i386-1
joe-3.5-i486-1
jove-4.16.0.61-i386-1
kernel-ide-2.4.33.3-i486-1
kernel-modules-2.4.33.3-i486-1
less-394-i486-1
lftp-3.5.4-i486-1
libart_lgpl-2.3.17-i486-1
libidn-0.6.5-i486-1
libxml2-2.6.26-i486-1
lilo-22.7.1-i486-2
logrotate-3.7.4-i486-1
mailx-12.1-i486-1
man-1.6c-i486-2
man-pages-2.39-noarch-1
mc-4.6.1-i486-2
mhash-0.9.7-i486-1
minicom-2.1-i486-2
mkinitrd-1.0.1-i486-3
module-init-tools-3.2.2-i486-2
mutt-1.4.2.2i-i486-1
mysql-5.0.24a-i486-1
nc-1.10-i386-1
neon-0.25.5-i486-2
netwatch-1.0a-i386-1
nmap-4.11-i486-1
openldap-client-2.3.27-i486-1
openssh-4.4p1-i486-1
openssl-0.9.8d-i486-1

openssl-solibs-0.9.8d-i486-1
pciutils-2.2.3-i486-2
pcmcia-cs-3.2.8-i486-3
php-4.4.4-i486-3
pidentd-3.0.19-i486-1
pkgtools-11.0.0-i486-4
portmap-5.0-i486-3
procmail-3.22-i486-2
procps-3.2.7-i486-1
reiserfsprogs-3.6.19-i486-1
rrdtool-1.2.8-i486-1jm
scarlet-2.0
screen-4.0.2-i486-1
sed-4.1.5-i486-1
sendmail-8.13.8-i486-4
shadow-4.0.3-i486-13
sharutils-4.6.3-i486-1
slocate-3.1-i486-1
smartmontools-5.36-i486-1
squid-2.6.STABLE1-i386-1
stunnel-4.17-i486-1
subversion-1.4.0-i486-1
sudo-1.6.8p12-i486-1
sysfsutils-2.0.0-i486-2
sysklogd-1.4.1-i486-9
syslinux-2.13-i486-1
sysvinit-2.84-i486-69
tar-1.15.1-i486-2
tcpdump-3.9.4-i486-2
tcpip-0.17-i486-39
tcsh-6.14.00-i486-2
texinfo-4.8-i486-1
traceroute-1.4a12-i386-2
udev-097-i486-10
umsdos-progs-1.13-i386-1
usbutils-0.72-i486-1
utempter-1.1.3-i486-1
util-linux-2.12r-i486-5
vim-7.0.109-i486-1
wget-1.10.2-i486-2
whois-4.7.15-i486-1
wireless-tools-28-i486-3

15.3. Modifications of the Configuration Files
sshd_config.diff :

--- /etc/ssh/sshd_config.bak 2006-04-25 16:20:16.000000000 +0200
+++ templates/sshd_config 2006-04-25 16:17:32.000000000 +0200
@@ -11,6 +11,8 @@
# default value.

#Port 22
+### dasho

+Port 2222
#Protocol 2,1
#AddressFamily any
#ListenAddress 0.0.0.0

rc.local.diff :

--- rc.local.1 2007-01-11 01:39:51.000000000 +0100
+++ rc.local 2007-01-04 15:06:21.000000000 +0100
@@ -4,3 +4,20 @@
#
# Put any local setup commands in here:

+### application path
+app_path=|APP_PATH|
+
+### record the startup event
+$app_path/scripts/record_startup.sh
+
+### configure the server
+$app_path/server-config/reconfig.sh
+
+### check periodicly the status of the gateways
+$app_path/server-config/watch.sh &
+
+### start the traffic counter
+$app_path/server-config/traffic-counter/start.sh
+
+### start 'cron/check-traffic-limits.php'
+$app_path/cron/check-traffic-limits.php &

hosts.diff :

--- /etc/hosts.bak 2006-04-25 15:18:00.000000000 +0200
+++ templates/hosts 2006-04-25 14:33:39.000000000 +0200
@@ -13,7 +13,7 @@

# For loopbacking.
127.0.0.1 localhost
+|LOCAL_IP| scarlet.|DOMAIN| scarlet router.|DOMAIN| router

# End of hosts.

dnsmasq.diff :

--- dnsmasq.conf.1 2007-01-11 01:35:46.000000000 +0100
+++ dnsmasq.conf 2007-01-11 01:32:05.000000000 +0100
@@ -30,6 +30,7 @@
# Change this line if you want dns to get its upstream servers from
# somewhere other that /etc/resolv.conf
#resolv-file=
+resolv-file=/etc/resolv.dnsmasq.conf

# By default, dnsmasq will send queries to any of the upstream
# servers it knows about and tries to favour servers to are known
@@ -54,6 +55,7 @@
# Add local-only domains here, queries in these domains are answered
# from /etc/hosts or DHCP only.
#local=/localnet/
+local=/|DOMAIN|/

# Add domains which you want to force to an IP address here.
# The example below send any host in doubleclick.net to a local
@@ -72,6 +74,7 @@
#interface=
# Or you can specify which interface _not_ to listen on
#except-interface=
+except-interface=eth0
# Or which to listen on by address (remember to include 127.0.0.1 if
# you use this.)
#listen-address=
@@ -97,6 +100,7 @@
# Set this (and domain: see below) if you want to have a domain
# automatically added to simple names in a hosts-file.
#expand-hosts
+expand-hosts

# Set the domain for dnsmasq. this is optional, but if it is set, it
# does the following things.
@@ -106,6 +110,7 @@
# domain of all systems configured by DHCP
# 3) Provides the domain part for "expand-hosts"
#domain=thekelleys.org.uk
+domain=|DOMAIN|

# Uncomment this to enable the integrated DHCP server, you need
# to supply the range of addresses available for lease and optionally
@@ -113,6 +118,7 @@
# repeat this for each network on which you want to supply DHCP
# service.
#dhcp-range=192.168.0.50,192.168.0.150,12h
+dhcp-range=|LAN_NETWORK|.50,|LAN_NETWORK|.250,12h

# This is an example of a DHCP range where the netmask is given. This
# is needed for networks we reach the dnsmasq DHCP server via a relay

httpd.conf.diff :

--- /etc/apache/httpd.conf.default 2006-08-10 02:50:08.000000000 +0200
+++ httpd.conf 2007-01-11 01:29:28.000000000 +0100
@@ -136,14 +136,17 @@
# a new spare. If there are more than MaxSpareServers, some of the
# spares die off. The default values are probably OK for most sites.
#
-MinSpareServers 5
-MaxSpareServers 10
+#MinSpareServers 5

+#MaxSpareServers 10
+MinSpareServers 2
+MaxSpareServers 5

#
# Number of servers to start initially --- should be a reasonable ballpark
# figure.
#
-StartServers 5
+#StartServers 5
+StartServers 2
#
# Limit on total number of servers running, i.e., limit on the number
@@ -152,7 +155,8 @@
# It is intended mainly as a brake to keep a runaway server from taking
# the system with it as it spirals down...
#
-MaxClients 150
+#MaxClients 150
+MaxClients 50

#
# MaxRequestsPerChild: the number of requests each child process is
@@ -168,7 +172,8 @@
# an initial request and 10 subsequent "keptalive" requests, it
# would only count as 1 request towards this limit.
#
-MaxRequestsPerChild 0
+#MaxRequestsPerChild 0
+MaxRequestsPerChild 1000

#
# Listen: Allows you to bind Apache to specific IP addresses and/or
@@ -327,7 +332,7 @@
# e-mailed. This address appears on some server-generated pages, such
# as error documents.
#
-ServerAdmin root@tree.slackware.lan
+ServerAdmin root@scarlet.localnet.net

#
# ServerName allows you to set a host name which is sent back to clients for
@@ -346,13 +351,15 @@
# local testing and development, you may use 127.0.0.1 as the server name.
#
#ServerName www.example.com
+ServerName scarlet.localnet.net

#
# DocumentRoot: The directory out of which you will serve your
# documents. By default, all requests are taken from this directory, but
# symbolic links and aliases may be used to point to other locations.
#
-DocumentRoot "/var/www/htdocs"
+#DocumentRoot "/var/www/htdocs"
+DocumentRoot "|APP_PATH|"

#
# Each directory to which Apache has access, can be configured with respect
@@ -377,7 +384,8 @@
#
# This should be changed to whatever you set DocumentRoot to.
#
-<Directory "/var/www/htdocs">
+#<Directory "/var/www/htdocs">
+<Directory "|APP_PATH|">

#
# This may also be "None", "All", or any combination of "Indexes",
@@ -386,7 +394,8 @@
# Note that "MultiViews" must be named *explicitly* --- "Options All"
# doesn't give it to you.
#
- Options Indexes FollowSymLinks MultiViews
+ #Options Indexes FollowSymLinks MultiViews
+ Options FollowSymLinks MultiViews

#
# This controls which options the .htaccess files in directories can
@@ -911,6 +920,8 @@
#
# 2) local redirects
#ErrorDocument 404 /missing.html
+ErrorDocument 404 /connect.php
+
# to redirect to local URL /missing.html
#ErrorDocument 404 /cgi-bin/missing_handler.pl
# N.B.: You can redirect to a script or a document using server-side-includes.
@@ -1030,7 +1041,7 @@
# mysql (AP series), gmp (L series), mhash (L series),
# and apache (N series)
#
-#Include /etc/apache/mod_php.conf
+Include /etc/apache/mod_php.conf

# ==> mod_ssl configuration settings <==
#

mod_php.conf.diff :

--- /etc/apache/mod_php.conf.bak 2006-04-25 15:16:47.000000000 +0200
+++ templates/mod_php.conf 2006-04-25 14:33:39.000000000 +0200
@@ -10,3 +10,10 @@

# This will display PHP files in colored syntax form. Use with caution.
#AddType application/x-httpd-php-source .phps
+
+#
+# Add index.php to the list of files that will be served as directory
+# indexes.
+#

+DirectoryIndex index.php
+

php.ini.diff :

--- /etc/apache/php.ini.bak 2006-04-26 14:33:06.000000000 +0200
+++ /etc/apache/php.ini 2006-04-26 14:27:13.000000000 +0200
@@ -282,14 +282,17 @@
;
; - Show all errors
;
-error_reporting = E_ALL
+;error_reporting = E_ALL
+error_reporting = E_ALL & ~E_NOTICE

; Print out errors (as a part of the output). For production web sites,
; you're strongly encouraged to turn this feature off, and use error logging
; instead (see below). Keeping display_errors enabled on a production web site
; may reveal security information to end users, such as file paths on your Web
; server, your database schema or other information.
-display_errors = Off
+display_errors = On

; Even when display_errors is on, errors that occur during PHP's startup
; sequence are not displayed. It's strongly recommended to keep

squid.conf :

squid.logrotate :

Building a Gateway Server

More Related Content

What's hot (20)

Viewers also liked (8)

Similar to Building a Gateway Server (20)

More from Dashamir Hoxha (15)

Building a Gateway Server