Gerard Beekmans gerard at
Sun Sep 23 12:28:56 PDT 2001

Hey Ian,

It's about time that I send in something again for our hints database, it's
been a while.

Gerard Beekmans

-*- If Linux doesn't have the solution, you have the wrong problem -*-
-------------- next part --------------
TITLE:		scp-before-relay
AUTHOR:		Gerard Beekmans <gerard at>

	An alternative to the pop-before-relay scheme using postfix


	This document is, at the moment, mostly a quick hack. I had to
	implement this in between releasing the LFS-3.0 book and some
	other pending LFS work, because my ISP's SMTP servers went offline.
	The scripts are anything but effecient, but it gets the job done, 
	which is good enough for me at the moment. There's nothing elegant 
	about it, so be warned ;)

	Having that said, let's get the show on the road.

what is pop-before-relay

A lot of ISP's give their users dynamically assigned IP addresses. One of
the side-effects of this is that remote servers never know who is behind a 
particular IP address. You cannot simply tell the SMTP server you wish to relay 
with to allow it from one or a block of IP addresses, because this would allow
anybody from your ISP to exploit the relay.

So, people came up with the pop-before-relay scheme. When you login to a
server to check for new mail, your current IP address is logged and added
to the SMTP's server relay database. A pop-before-relay (helper) daemon
will check that database and expire IP addresses after they have been
in the database for a certain amount of time, because you could have gone offline
after you checked for new email, so the IP address should be flushed from
that database as soon as possible to lower the chance of somebody else obtaining 
the IP address and exploit the relay, unless you check for new mail again within 
the expire time.


What is this scp-before-relay and why not use pop-before-relay if software
has already been written to support it?

The problem I encountered is that I use ssh tunneling to login to my pop3
server at With pop3 (imap and others probably too) your 
password is transmitted as clear text, much like happens with telnet. I don't 
quite like that, so I use port forwarding with openssh. The downside is that
the pop3 daemon now logs my login attempt with the IP address of the server
itself (since I'm ssh'ed into the server and connect to localhost there, so
the originating IP address is that of localhost as well).

So I had to come up with something different, and I call it

How it works

My workstation, called gwaihir, runs the following script every 4

	# Begin send-ip


	ifconfig ppp0 > /home/gerard/tmp/ppp.tmp || exit 1

	IP=$(cat /home/gerard/tmp/ppp.tmp|grep "inet addr"| \
		cut -f 2 -d ":"|cut -f 1 -d " ")

	echo "$IP 0" > /home/gerard/tmp/gwaihir-ip

	/usr/bin/scp /home/gerard/tmp/gwaihir-ip \

	# End send-ip

The above script will obtain gwaihir's current IP address on the ppp0 link, 
if it's up at the moment. Then send the file, gwaihir-ip, to the server in my ~/tmp directory over there.

The format of the file is:
		IP 0

The 0 can be any value you want, it's not used in the current scheme. It's
just to satisfy the hash database format used by postfix.

The server, shadowfax, runs the following script every 10 minutes:

	# Start update-postfix-relay


	if [ -f /home/gerard/tmp/gwaihir-ip ]
	        mv /home/gerard/tmp/gwaihir-ip /etc/postfix
	        chown root.root /etc/postfix/gwaihir-ip
	        postmap /etc/postfix/gwaihir-ip
	        date +%s > /etc/postfix/gwaihir-ip.log
	        LAST=$(cat /etc/postfix/gwaihir-ip.log)
	        if [ $(expr $(date +%s) - $LAST) -lt 1200 ]
        	        echo "0 0" > /etc/postfix/gwaihir-ip
	                postmap /etc/postfix/gwaihir-ip
	# End update-postfix-relay

This script checks for a new gwaihir-ip file. If found, move the file to
/etc/postfix and run postmap to update the database (else postfix will
complain that the database file is older than it's source file). Also log
the time when this update took place. The +%s option will log the amount of
seconds since January 1970. This makes it easier to calculate when it's
supposed to expire.

Now, if there is no file gwaihir-ip present on shadowfax, it usually is
caused by one of the following:

1) both scripts were run at the same time. There will always be a point in
time where both systems will be running the scripts at the exact same
time, so gwaihir could be in the middle of sending the new file while
shadowfax is checking for it to exist.

2) gwaihir is still online, but there's a network problem between gwaihir's
ISP and shadowfax, which could have caused shadowfax to be unreachable from
gwaihir for a short period of time.

3) gwaihir went offline

There is no way for shadowfax to know which of the 3 reasons applies at the
moment. It's most likely reason 1 or 2, so we don't want to turn off the 
relay authorization right away.

So instead, shadowfax checks gwaihir-ip.log's contents against the current
time. If the difference is less than 1200 seconds (20 minutes), then
it won't do anything. Since shadowfax checks every 10 minutes, gwaihir will have
10 minutes to send the file. Because gwaihir tries to send it every 4
minutes, it gives gwaihir 2 more chances to get the file delivered before
shadowfax checks again, when the 1200 seconds are up.

If during the next check (after 1200 seconds) gwaihir still hasn't sent a file, 
we'll have to disable gwaihir's ability to relay. At that time "0 0" is put in 
postfix's database. There is no IP address that matches '0', so this effectively 
disables the relay.


There is much room for improvement.

Shortly after I activated this scheme I got the idea to put the date +%s
output in the gwaihir-ip file as the second value, which would elimate the 
need for the gwaihir-ip.log file. gwaihir can't put the date +%s in it's
file, because the systems are in different timezones. So shadowfax will
have to do this when it found the new file.

The send-ip script could be changed to write the gwaihir-ip file as:

	echo "$IP stub" > /home/gerard/tmp/gwaihir-ip

Then the modified update-postfix-relay script could look like:
	# Start update-postfix-relay


	if [ -f /home/gerard/tmp/gwaihir-ip ]
		sed s/stub/$(date +%s)/ /home/gerard/tmp/gwaihir-ip \
			> /etc/postfix/gwaihir-ip
		rm /home/gerard/tmp/gwaihir-ip
		postmap /etc/postfix/gwaihir-ip
		LAST=$(cat /etc/postfix/gwaihir-ip|cut -f 2 -d " ")
		if [ $(expr $(date +%s) - $LAST) -lt 1200 ]
			echo "0 0" > /etc/postfix/gwaihir-ip
			postmap /etc/postfix/gwaihir-ip
	# End update-postfix-relay

I'm pretty sure this works, I just haven't tried it yet.

Another obvious improvent would be using variables in the scripts for the
gwaihir-ip file locations, like:


Other than that, this is designed to work with one user. If you have more
people you want to be able to use the server as a relay, you need to make some 
changes. You could use seperate files for every user, and put the script in a loop 
to parse the files and disable or refresh somebody's relay abilities
accordingly. I haven't bothered giving multi-user functionality any thought, 
so you're on your own for now. Feel free to get back to me if you have any ideas 
on the matter. I'm not going to pursue it right now because I don't have the 
need for it. The need may arise in the future, so who knows.

More information about the hints mailing list