Static IP Addresses and Accounting

Over the last few posts I’ve talked about how to set up the basic parts needed to run a small ISP.

In this post I’m going to cover adding a few extra features such as static IP addresses, Bandwidth accounting and Bandwidth limiting/shaping.

Static IP Addresses

We can add a static IP address by adding a field to the users LDAP entry. To do this first we need to add the Freeradius schema to the list of fields that the LDAP server understands. The Freeradius schema files can be found in the /usr/share/doc/freeradius/schemas/ldap/openldap/ and have been gzipped. I unzipped them and copied them to /etc/ldap/schema then imported it with

$ sudo ldapadd -Y EXTERNAL -H ldapi:/// -f /etc/ldap/schema/freeradius.ldif

Now we have the schema imported we can now add the radiusprofile objectClass to the user along with a radiusFramedIPAddress entry with the following ldif file.

dn: uid=isp1,ou=users,dc=hardill,dc=me,dc=uk
changetype: modify
add: objectClass
objectClass: radiusprofile
add: radiusFramedIPAddress

We then use ldapmodify to update the isp1 users record

$ ldapmodify -f addIPAddress.ldif -D cn=admin,dc=hardill,dc=me,dc=uk -w password

Now we have the static IP address stored against the user, we have to get the RADIUS server to pass that information back to the PPPoE server after it has authenticated the user. To do this we need to edit the /etc/freeradius/3.0/mods-enabled/ldap file. Look for the `update` section and add the following

update {
  reply:Framed-IP-Address     := 'radiusFramedIPAddress'

Running radtest will now show Framed-IP-Address in the response message and when pppoe-server receives the authentication response it will use this as the IP address for the client end of the connection.


Out of the box pppoe-server will send accounting messages to the RADIUS server at the start and end of the session.

Sat Aug 24 21:35:17 2019
	Acct-Session-Id = "5D619F853DBB00"
	User-Name = "isp1"
	Acct-Status-Type = Start
	Service-Type = Framed-User
	Framed-Protocol = PPP
	Acct-Authentic = RADIUS
	NAS-Port-Type = Virtual
	Framed-IP-Address =
	NAS-IP-Address =
	NAS-Port = 0
	Acct-Delay-Time = 0
	Event-Timestamp = "Aug 24 2019 21:35:17 BST"
	Tmp-String-9 = "ai:"
	Acct-Unique-Session-Id = "290b459406a25d454fcfdf3088a2211c"
	Timestamp = 1566678917

Sat Aug 24 23:08:53 2019
	Acct-Session-Id = "5D619F853DBB00"
	User-Name = "isp1"
	Acct-Status-Type = Stop
	Service-Type = Framed-User
	Framed-Protocol = PPP
	Acct-Authentic = RADIUS
	Acct-Session-Time = 5616
	Acct-Output-Octets = 2328
	Acct-Input-Octets = 18228
	Acct-Output-Packets = 32
	Acct-Input-Packets = 297
	NAS-Port-Type = Virtual
	Acct-Terminate-Cause = User-Request
	Framed-IP-Address =
	NAS-IP-Address =
	NAS-Port = 0
	Acct-Delay-Time = 0
	Event-Timestamp = "Aug 24 2019 23:08:53 BST"
	Tmp-String-9 = "ai:"
	Acct-Unique-Session-Id = "290b459406a25d454fcfdf3088a2211c"
	Timestamp = 1566684533

The Stop message includes the session length (Acct-Session-Time) in seconds and the number of bytes downloaded (Acct-Output-Octets) and uploaded (Acct-Input-Octets).

Historically in the days of dial up that probably would have been sufficient as sessions would probably only last for hours at a time, not weeks/months for a DSL connection. pppoe-server can be told to send updates at regular intervals, this setting is also controlled by a field in the RADIUS authentication response. While we could add this to each user, it can be added to all users with a simple update to the /etc/freeradius/3.0/sites-enabled/default file in the post-auth section.

post-auth {
   update reply {
      Acct-Interim-Interval = 300

This sets the update interval to 5mins and the log now also contains entries like this.

Wed Aug 28 08:38:56 2019
	Acct-Session-Id = "5D62ACB7070100"
	User-Name = "isp1"
	Acct-Status-Type = Interim-Update
	Service-Type = Framed-User
	Framed-Protocol = PPP
	Acct-Authentic = RADIUS
	Acct-Session-Time = 230105
	Acct-Output-Octets = 10915239
	Acct-Input-Octets = 17625977
	Acct-Output-Packets = 25918
	Acct-Input-Packets = 31438
	NAS-Port-Type = Virtual
	Framed-IP-Address =
	NAS-IP-Address =
	NAS-Port = 0
	Acct-Delay-Time = 0
	Event-Timestamp = "Aug 28 2019 08:38:56 BST"
	Tmp-String-9 = "ai:"
	Acct-Unique-Session-Id = "f36693e4792eafa961a477492ad83f8c"
	Timestamp = 1566977936

Having this data written to a log file is useful, but if you want to trigger events based on it (e.g. create a rolling usage graph or restrict speed once a certain allowance has been passed) then something a little more dynamic is useful. Freeradius has a native plugin interface, but it also has plugins that let you write Perl and Python functions that are triggered at particular points. I’m going to use the Python plugin to publish the data to a MQTT broker.

To enable the Python plugin you need to install the freeradius-python package

$ sudo apt-get install freeradius-python

And then we need to symlink the mods-available/python to mods-enabled and then edit the file. First we need to set the path that the plugin will use to file Python modules and files. And then enable the events we want to pass to the module.

python {
    python_path = "/etc/freeradius/3.0/mods-config/python:/usr/lib/python2.7:/usr/local/lib/python/2.7/dist-packages"
    module = example

    mod_instantiate = ${.module}
    func_instantiate = instantiate

    mod_accounting = ${.module}
    func_accounting = accounting

The actual code follows, it publishes the number of bytes used in the session to the topic isp/[username]/usage. Each callback gets pass a tuple containing all the values available.

import radiusd
import paho.mqtt.publish as publish

def instantiate(p):
  print "*** instantiate ***"
  print p
  # return 0 for success or -1 for failure

def accounting(p):
  print "*** accounting ***"
  radiusd.radlog(radiusd.L_INFO, '*** radlog call in accounting (0) ***')
  print p
  d = dict(p)
  if d['Acct-Status-Type'] == 'Interim-Update':
      topic = "isp/" + d['User-Name'] + "/usage"
      usage = d['Acct-Output-Octets']
      print "publishing data to " + topic
      publish.single(topic, usage, hostname="", retain=True)
      print "published"
  return radiusd.RLM_MODULE_OK

def detach():
  print "*** goodbye from ***"
  return radiusd.RLM_MODULE_OK

I was going to talk about traffic shaping next, but that turns out to be real deep magic and I need to spend some more time playing before I have something to share.


As mentioned in the last post, I’m building a PoC ISP and to do this I need to set both an LDAP and RADIUS servers.

I’m going to run all of this on the latest version of Raspbian Buster.


Lets start by installing the LDAP server.

$ sudo apt-get install ldap-server

This will install OpenLDAP. The first thing to do is to set the admin password and configure the base dn. To do this we first create a hashed version of the password with slappasswd

$ slappasswd
New password:
Re-enter new password: 

We can then create the following ldif file called config.ldif. This sets the admin password and updates the base dn

dn: olcDatabase={1}mdb,cn=config
changetype: modify
add: olcRootPW
olcRootPW: {SSHA}TXcmvaldskl312012cKsPK1cY2321+aj
replace: olcRootDN
olcRootDN: cn=admin,dc=hardill,dc=me,dc=uk
replace: olcSuffix
olcSuffix: dc=hardill,dc=me,dc=uk

We the apply these changed with the ldapmodify command

$ ldapmodify -a -Q -Y EXTERNAL -H ldapi:/// -f config.ldif

Now we have the admin user set up we can start to add the normal users. First we need to create the structure to hold them.

dn: dc=hardill,dc=me,dc=uk
objectClass: dcObject
objectClass: organization
dc: hardill
o: hardill

dn: ou=users,dc=hardill,dc=me,dc=uk
objectClass: organizationalUnit
ou: users

And since we’ve set the admin password we need to change the ldapadd command as well

 ldapadd -f domain.ldif -D cn=admin,dc=hardill,dc=me,dc=uk -w password

Again we need to use the slappasswd command to create password that we can use in the user.ldif file. I’ve added the inetOrgPerson in to the user entry so I can also include the mail item.

dn: uid=isp1,ou=users,dc=hardill,dc=me,dc=uk
objectClass: top
objectClass: person
objectClass: inetOrgPerson
displayName: Joe Blogs
cn: Joe
sn: Blogs
uid: isp1
userPassword: {SSHA}rozJD+T37NqRQp36myXf1KJ35+7tf2LN

Added to the ldap with

$ ldapadd -f user.ldif -D cn=admin,dc=hardill,dc=me,dc=uk -w password 


Next we need to install the RADIUS

$ sudo apt-get install freeradius

Once installed we need to enable the LDAP module and configure it to use the server we have just setup. To do this we need to symlink the ldap file from /etc/freeradius/3.0/mods-available to /etc/freeradius/3.0/mods-enabled. Next edit the identity, password and base_dn in the ldap config file to match the settings in config.ldif.

	#  additional schemes:
	#  - ldaps:// (LDAP over SSL)
	#  - ldapi:// (LDAP over Unix socket)
	#  - ldapc:// (Connectionless LDAP)
	server = 'localhost'
#	server = ''
#	server = ''

	#  Port to connect on, defaults to 389, will be ignored for LDAP URIs.
#	port = 389

	#  Administrator account for searching and possibly modifying.
	#  If using SASL + KRB5 these should be commented out.
	identity = 'cn=admin,dc=hardill,dc=me,dc=uk'
	password = password

	#  Unless overridden in another section, the dn from which all
	#  searches will start from.
	base_dn = 'ou=users,dc=hardill,dc=me,dc=uk'

	#  SASL parameters to use for admin binds

Once we’ve restarted freeradius we can test if we can authenticate the isp1 user with the radtest command.

$ radtest isp1 secret testing123
Sent Access-Request Id 159 from to length 78
	User-Name = "isp1"
	User-Password = "secret"
	NAS-IP-Address =
	NAS-Port = 0
	Message-Authenticator = 0x00
	Cleartext-Password = "secret"
Received Access-Accept Id 159 from to length 51

testing123 is the default password for a RADIUS client connecting from, you can change this and add more clients in the /etc/freeradius/3.0/clients.conf file.

In the next post I’ll talk about setting up PPPoE

Building an ISP

I’ve had this idea in the back of my head for ages, it’s centred round buying a building (something like an old Yorkshire mill, or better yet a private island) and dividing it up into a number of homes/offices/co-working paces.

To go with this fantasy I’ve been working out how to build a small scale boutique ISP (most of this would probably work for a small town community fibre or wireless mesh system) to share the hugely expensive high bandwidth symmetric dedicated fibre .

Over the next few posts I’m going to walk through building the PoC for this (which is likely to be where it stays unless I win the lottery)

To work out what I’d need lets first look roughly how home internet connections works.

At the advent of Home Internet there were two methods of delivering IP packets over a telephone/serial line, SLIP and PPP protocol. PPP became the dominant player and was extended to encapsulate PPP packets carried over both ATM (PPPoA) and Ethernet (PPPoE) frames in order to facilitate the move to DSL Home Broadband connections. PPPoE became the standard for the next evolution, FTTX (Where X can be B for building, P for premisses, or H for Home ). Modern home routers include a modem that converts DSL signal back to Ethernet frames and a PPPoE client to unpack the PPP connection back into IP packets to forward on to the network.

This means we need a PPPoE server for the users router to connect to, Linux has PPPoE support both as a client and as a server. I’ve already used the PPPoE client when the router for my FTTC Broadband service was late arriving.

Now we have the basic connection between the users equipment and the ISPs network we need to be able to authenticate each user so we know who is actually trying to connect. You can hard code credentials and details into the PPPoE configuration files, but this doesn’t scale and means you need to restart everything when ever something changes.

The better solution is something called a RADIUS server. RADIUS is a AAA service that can be used to not only authenticate users, but also supply information to the PPPoE server about that user, e.g. a static IP address allocation. RADIUS can also be used for accounting to record how much bandwidth each user has consumed.

A rasperry Pi and a Acer Revo hooked up to a ethernet switch
Initial testing

RADIUS servers can be backed by a number of different databases but the usual approach is to use LDAP.

In the next post I’ll cover installing the LDAP and RADIUS servers, then configuring them.

Update to node-red-node-ldap

I’ve updated the node-red-node-ldap node to use the ldapjs node rather than the LDAP node.

Node-RED LDAP node

The LDAP node was a NodeJS wrapper round the OpenLDAP libraries which meant there was an external native dependancy and limited the number of platforms that node could be deployed on.

The ldapjs node is all pure Javascript so should run everywhere.

Everything should be backward compatible but raise issues on github if you find problems.

Version 0.0.2 can be found on npmjs here and on github here

Securing Node-RED

Node-RED added some new authentication/authorisation code in the 0.10 release that allows for a plugable scheme. In this post I’m going to talk about how to use this to make Node-RED use a LDAP server to look up users.


First of all, to do this properly we will need to enable HTTPS to ensure the communication channel between the browser and Node-RED is properly protected. This is done by adding a https value to the settings.js file like this:

https: {
  key: fs.readFileSync('privkey.pem'),
  cert: fs.readFileSync('cert.pem')

You also need to un-comment the var fs = require(‘fs’); line at the top of settings.js.

You can generate the privkey.pem and cert.pem with the following commands in your node-red directory:

pi@raspberrypi ~/node-red $ openssl genrsa -out privkey.pem
Generating RSA private key, 1024 bit long modulus
e is 65537 (0x10001)
pi@raspberrypi ~/node-red $ openssl req -new -x509 -key privkey.pem -out cert.pem -days 1095
You are about to be asked to enter information that will be incorporated
into your certificate request.
What you are about to enter is what is called a Distinguished Name or a DN.
There are quite a few fields but you can leave some blank
For some fields there will be a default value,
If you enter '.', the field will be left blank.
Country Name (2 letter code) [AU]:GB
State or Province Name (full name) [Some-State]:Hampshire
Locality Name (eg, city) []:Eastleigh
Organization Name (eg, company) [Internet Widgits Pty Ltd]:Node-RED
Organizational Unit Name (eg, section) []:
Common Name (e.g. server FQDN or YOUR name) []:raspberrypi.local
Email Address []:

The important bit is the Common Name value, this needs to match either the name or IP address that you will use to access your Node-RED console. In my case I have avahi enabled so I can access my pi using it’s host name raspberrypi with .local as the domain, but you may be more used to using an IP address like

Since this is a self signed certificate your browser will reject it the first time you try to connect with a warning like this:

Chrome warning about unsafe cert.
Chrome warning about unsafe cert.

This is because your certificate is not signed by one of the trusted certificate authorities, you can get past this error by clicking on Advanced then Proceed to raspberrypi.local (unsafe). With Chrome this error will be shown once every time you access the page, you can avoid this by copying the cert.pem file to your client machine and import it into Chrome:

  1. Open Chrome settings page chrome://settings
  2. Scroll to the bottom of page and click on the “+Show advanced settings” link
  3. Scroll to the HTTPS/SSL and click on “Manage certificates…”
  4. Select the Servers tab and select import
  5. Select the cert.pem you copied from your Raspberry Pi

Usernames and Passwords

In previous Node-RED releases you could set a single username and password for the admin interface, any static content or one that covered bh. This was done by adding a object to the settings.js file containing the user name and password. This was useful but could be a little limited. Since the 0.10 release there is now a pluggable authentication interface that also includes support for things like read only access to the admin interface. Details of these updates can be found here.

To implement a authentication plugin you need to create a NodeJS module based on this skeleton:

var when = require("when");

module.exports = {
   type: "credentials",
   users: function(username){
      //returns a promise that checks the authorisation for a given user
      return when.promise(function(resolve) {
         if (username == 'foo') {
            resolve({username: username, permissions: "*"});
         } else {
   authenticate: function(username, password) {
      //returns a promise that completes when the user has been authenticated
      return when.promise(function(resolve) {
         if (username == 'foo' && password == 'bar' ) {
            resolve({username: username, permissions: "*"});
         } else {
   default: function() {
      // Resolve with the user object for the default user.
      // If no default user exists, resolve with null.
      return when.promise(function(resolve) {

This comprises on 3 functions, one to authenticate a user against the backend, one to check the level of authorisation (used by Node-REDs built in oAuth mechanism once a user has been authenticated), and finally default which matches unauthenticated users.

For my LDAP example I’m not going to implement different read only/write levels of authorisation to make things a little easier.

The source for the module can be found here or on npmjs here. The easiest way to install it is with:

npm install -g node-red-contrib-ldap-auth

Then edit the Node-RED settings.js to include the following:

adminAuth: require('node-red-contrib-ldap-auth').setup({
    base: 'ou=group,', 
    filterTemplate: 'mail={{username}}'

The filterTemplate is a mustache template to use to search the LDAP for the username, in this case the username is mail address.

Once it’s all up and running Node-RED should present you with something that looks like this:

Node-RED asking for credentials
Node-RED asking for credentials

LDAP and NFC Node-RED Nodes

About a week ago a colleague asked me to help resurrect some code I had written to use our work ID badges to look up information on the card owner in order to log into a system for a demonstration.

The ID badges are basically mifare cards so can be read by a NFC reader. The content of the cards is encrypted, but each card has a unique ID. Unfortunately the security team will not share the mapping of these IDs to actual people, but since this is for a demonstration that will only be given by a relatively small number of people it’s not a problem to set up a list of mappings our selves.

The original version of this code used nfc-eventd and some java code to the IDs then do a lookup in a little database to convert these to email addresses. It worked but was a bit of a pig to setup and move between machines as it required a number of different apps and config files so I decided to have a go at rewriting it all in Node-RED.


To do this I was going to need 2 new nodes, one to read the NFC card and one to look up details in the LDAP. Both of these actually proved reasonable easy and quick to write as there are existing Node NPM modules that do most of the heavy lifting. The flow has a couple of extra bit, it uses a mongodb to store the id to email address mappings and if there is no match it uses websockets to populate a field in a separate web page to enter a email address to update the database.


I did a first pass using the nfc npm and it worked but there was no way to shut the connection to the NFC reader down in the code which meant I couldn’t clean up properly when Node-RED shut down or when the node needed to be restarted.

The nfc on is actually a bit out of date compared to the git repository it’s hosted in. So I moved over to using the upstream version of the code. This changed the API a little and still didn’t have a mechanism to allow the interface to be stopped. I forked the project and after a little bit of playing I ended up with some working shutdown code.

The only call back is for when at NFC tag is detected and it polls in tight loop so the stream of data from the node is way too high to feed into a Node-RED flow really. The Node-RED wrapper rate limits to only reporting the same tag once every 10 seconds. This is good enough for the original problem I was looking to solve but I still think it can be done better. I’m planning on adding call backs for tag seen and when it is removed, this is similar to how nfc-eventd works. I also want to look at doing NDEF decoding.

You can install the current version of the node with:

npm install

It depends on libnfc which should work on the Linux and OSx and I’ve even seen instructions to build it for Windows.
Once I’ve got a bit further I’ll add it to


This one was even simpler. The LDAP npm modules links to the openldap libraries and does all the hard work.

It just needed a config dialog creating to take a base DN and a filter and a connection setup that takes a server, port and if needed a bind DN and password. The filter is a mustache template so values can be passed in.

This node is pretty much done, you can find the code on github here and the node can be installed with the following:

npm install node-red-node-ldap

Like with the NFC node, openldap should be available for Linux and OSx and there looks to be a Windows port.