How to hook up PPP

in Linux

W.G. Unruh

You can download this file as a text file (no HTML formatting) at www.theory.physics.ubc.ca/ppp-linux.txt
or as an HTML version at http://www.theory.physics.ubc.ca/ppp-linux.html

This document is copyright W. Unruh. It may be copied freely as a whole for non-commercial purposes, but must retain the indication of authorship. If you want to include it in a collection for sale, please contact the author. Furthermore, any alterations must have the permission of the author.

Reason for Document

The key problem in hooking up a PPP link to an Internet Service Provider (ISP) is that the ISP's seem to compete with each other as to who can find another obscure way of authenticating users. Thus we have login, PAP, CHAP 05, CHAP80, CHAP80-lanman, CHAP81, ... and combinations of these. The chief difficulty of connecting to an ISP is discovering which technique is actually being used by the ISP in an orderly way. Since few of them know anything about Linux, and since few of them even understand what technique they actually use, this procedure should allow you to set up without their help, or to understand what their help means if it is given.

The following has a number of steps to connecting to your ISP. The temptation is to skip steps. This is a bad idea, as it will almost certainly lead to grief. Do each step and you will succeed, and you will also learn something about your system in the process.

These instructions were developed with the 2.0.x versions of the Linux kernels, and also work with the 2.2.x or 2.4.x series of kernels as well. Similarly, they work with versions of pppd at least from version 2.3.3 on (up to 2.4.1 by now).(For the 2.4.x kernels you must use pppd version 2.4.1 or later.) In fact some of the features below only work with later versions of pppd and I would urge you to upgrade to the latest version. You can get the latest pppd from ftp://ftp.samba.org/pub/ppp. You may be able to get versions from ftp://cs.anu.edu.au/pub/software/ppp/ but these sites may not be up to date.

Unfortunately the man page for pppd has gotten out of date, and a number of options have been added/subtracted without proper documentation. For a latest man page for pppd 2.4.1 written by James Carlson ( who ported pppd to Solaris and is the Solaris maintainer), see pppd-jc.8

INDEX

Initial Setup

For much of this document, the steps are essentially independent of your ISP. Many ISPs do not know their own systems, and their advice must often be taken with a grain of salt. Thus, the document below shows you how to find out what your ISP wants, independent of what your ISP says he wants. However, for the first step you cannot do without your ISP. You must get a username and a password from your ISP-- which usually means that you must open an account with them and pay your money. It is worth asking them to make sure that they tell you exactly what form that user name must take when you sign on. For example some of them demand that you use your full email address with them as your username, rather than simply the name itself. Others demand additional additions to the bare username. Try to get them to be as specific as possible about exactly what you need to use.
You also need to get the ISP's Domain Name Server (DNS-IP) address. (This is an address in Internet Protocol (IP) format, which is four numbers each less than 256 separated by dots (eg 142.103.234.29).) Often they will give you more than one. Occasionally an ISP will refuse. In that case they may supply the DNS dynamically instead. Using ppp version greater than 2.3.7 will allow you still to sign on with them as described below.
  Having that information, log on as root (use root as the user name when you log on), or, after you have logged on as a user, run
su
and enter your root password at the prompt.)

Modem

Make sure that you know which port your modem is on. The two programs at modem-chk.html can help you to determine which port your modem is on. Alternatively, running
setserial -bg /dev/ttyS* can help find serial devices with something attached.
Remember that COM1 (serial port 1) is /dev/ttyS0. COM2 is /dev/ttyS1, etc. In the following I will assume /dev/ttyS1. Change for your situation.
If you have an internal PCI card modem (which is not a winmodem-- see below), looking in /proc/pci can give you a clue as to where the PCI bus placed the modem (ie port and irq). You can use the setserial command to set up one of the ttyS devices to use that port and irq.
It is preferable to use the ttyS device rather than the cua device or the modem device. The cua devices will (have) disappear in future releases of Linux, and the /dev/modem device hides what you are actually doing and can lead to conflicts with other programs which use the serial ports.
[Note that if your modem is a winmodem, it will not work as is under Linux. Some winmodems now have drivers for Linux. See www.linmodems.org--- It is probably a winmodem if it is a PCI modem, less so if it is an ISA and improbable if it is an external modem, and also improbably if it is not a 56K modem.
If you have the latest kernels with USB support (Kernels 2.2.17 with the USB patches or 2.4.x) you may also be able to use USB modems. (They must comply with the ACM standard.) See The file Documentation/usb/acm.txt included with the Linux source (/usr/src/linux* if included on your system, or here is a copy from kernel 2.2.17 ) Note that for USB modems, the correct port to use is not the serial port ttyS? series but probably the /dev/usb/acm/? series of ports, where ? is a usb number. I have never used a usb modem, so can provide no guidance on setting them up correctly.

Message Logging

In order to figure out what is happening while you are trying to log on to your ISP, you must turn on debugging and log the debug messages to some file. The key programs are pppd, and a program it uses, called chat. We will store those debugging messages in a file called /var/log/ppp. (You can use any file you wish, including /var/log/messages. Some distributions are now installing a directory called /var/log/ppp (Eg, The Corel distribution does this). If yours does, you must use a different name, eg, /var/log/ppplog. Use that name instead everywhere this document talks about /var/log/ppp.)

In order that your system will record the debugging information from both these programs, do the following:
cp /etc/syslog.conf /etc/syslog.conf.orig
to make sure you have a copy of a good version in case you mess up the next command. If you have trouble, copy back the original version.
Edit /etc/syslog.conf and add the lines

local2.*                                /var/log/ppp 
daemon.*                                /var/log/ppp
(Some versions of syslogd (eg on SunOS) demand that those be tabs not spaces between the two parts on each line. Some editors refuse to insert tabs and convert them to spaces. The syslogd shipped with most recent distributions of Linux do not seem to care, but Be warned.)

Then, to tell syslogd to actually log the information, do
killall -1 syslogd

SUSE has now gone over to using syslog-ng instead of syslogd. While it does give much more room for fine tuning exactly what and how you want things logged, it comes with a much more complex setup. As far as I can tell from the docs for syslog-ng, you need to add the following lines to the /etc/syslog-ng.config or /etc/syslog-ng/syslog-ng.conf file

destination ppp { file("/var/log/ppp"); };

filter f_ppp {facility(daemon) and facility(local2);};

log { source(src); filter(f_ppp); destination(ppp); };
and then restart the syslog-ng daemon. (killall -1 syslog-ng)
I do not have such a syslog-ng system, so am unable to test this. Perhaps someone who does could drop me a note letting me know if it works or to correct my mistakes.

Resolver Information

The computers on the internet are referenced by numbers called IP numbers. However, people, being bad at remembering numbers, use names instead (eg, www.theory.physics.ubc.ca instead of 142.103.234.29). The name resolver is a program that asks another computer, the nameserver, to translate internet names to IP numbers which your computer can actually use. In order that, once you are connected, you are able to actually resolve host names,
edit /etc/resolv.conf and add a line (or lines) like
nameserver 111.222.333.444
Instead of 111.222.333.444 use the Domain Name Server IP address (DNS-IPaddress) which your ISP gave you. If they gave you two numbers enter another nameserver line with the second address.
It can also be useful to include a line like
domain ISP.domain.name
where ISP.domain.name is the domain name of your ISP. For example, my /etc/resolv.conf file looks like
domain  physics.ubc.ca
nameserver 137.82.43.9
nameserver 137.82.1.1
[Note: Do NOT put leading zeros on any of those numbers. They will not work. 137.082.43.9 is a different number from 137.82.43.9]

Edit /etc/host.conf and make sure the line
order hosts, bind
is in there.
Also check /etc/nsswitch.conf and make sure a line like
hosts: files dns
is there. (that line may also contain other options like nis, which will do no harm)

Also make sure that all of these files are readable by everyone.
chmod a+r /etc/resolv.conf /etc/host.conf /etc/nsswitch.conf

No Default Route

Your computer passes messages for addresses it does not know to the default route. ppp assumes that there can only be one default route on any system. This is almost always the route to ISP's computer. However, if pppd finds that a default route already exists, it will not set up one to your ISP, causing problems (nothing gets through). In order that you not have any conflicts between the routing across ppp and onto some network you have, run
/sbin/route -n
(or you can also run
/bin/netstat -nr
which does the same thing)
while not connected via ppp, and look for a default entry. (First entry in line is 0.0.0.0). This almost certainly should not be there. (Unless you already have a hookup to the internet via an Ethernet connection, you do not want such a default route.) Unless you know that your computer has an Ethernet connection to other computers already, the only thing there should be a route to 127.0.0.1 or 127.0.0.0 If there is such a default route (line starting with 0.0.0.0), and you want the ppp interface to be your default route, (the usual situation), run
/sbin/route del default
Also find out what was setting that wrong default route, and stop it from doing so. (If you cannot figure out what is doing so, put that line
/sbin/route del default
at the end of whichever of /etc/rc.local or /etc/rc.d/rc.local exists on your system.)

PPPD module support

Now check to see if the module support for ppp is set up properly. Run
/usr/sbin/pppd
You should get a line of garbage which repeats about 10 times and then the command exits. This is good. Skip to the PPPD Options section.
If you get nothing and the command return to the prompt after about 10 seconds, this is probably OK as well, but try
/usr/sbin/pppd notty
This should give you the lines of garbage. If you do get this line of garbage, skip the next section. If you do not get that line of garbage, it is probable that the module support is not set up properly.

Setting up module support

If you did not get that line of garbage, but instead get "permission denied", you are not running as root. Log on as root and try again.
If you get a message saying that your kernel does not support ppp, do not recompile the kernel. This error message is a grab bag error message which the pppd program produces almost anytime something goes wrong. The response depends on the kernel version you are running.

PPPoE

Many cable companies are offering ppp connections via the TV cable connection. This connection is an Ethernet connection. Unfortunately many of the cable companies have bought onto a bastardisation called PPP over Ethernet (PPPoE). Why one should want to layer a ppp connection over something which already provides a layer which can handle any of the higher layers (Ethernet) is mysterious, but as consumers with monopoly cable suppliers we often have no say in the matter. Thus, if this is the bind you are caught in, go to Roaring Penguin to get their implementation. All reports I have read state that it is vastly superiour to anything your cable company may try to supply you with.
In the 2.4.x kernels, there is also experimental support for kernel mode pppoe, which should in theory be faster. It seems that some distributions supply it already compiled in as a module (Mandrake 8.0 and later), but otherwise you will have to recompile your kernel. Then got to Mike Mostrows' page to get a patched version of ppp 2.4.1. Also go to the Roaring Penguin page to get the other programs needed to run pppoe. Then be prepared to solve any problems on your own since neither I nor Roaring Penguin will or can help you.

Hostname

One final item, especially for Redhat systems, make sure that your give your machine a name (any single word name you want). You can set it with
hostname <yourhostname>
But you also want to change it permanently. In Redhat/Mandrake, put the line
HOSTNAME=<yourhostname>
into /etc/sysconfig/network. On other systems I am not sure, but you could always put the hostname command at the end of your rc.local file. This is not essential to running ppp, but if you do not do it, you will find yourself unable to run any X application while running ppp.

PPP Options

You have now set up the critical files on your system and you know that you have an operational pppd. You now need to set up the ppp options so that you can get debugging information when you try to connect, and so that it is minimally usable.
Edit /etc/ppp/options so it contains
lock
crtscts
defaultroute
noauth
[note that if you are in the unlikely situation that you already have a working default route to an Ethernet, omit the defaultroute option. If you are not sure, put that option in.]
[Also other options can be used. In general do not do so unless advised to do so. More harm is done by putting in too many obscure options than by putting in too few. If you have a local Ethernet, and if your ISP demands that they assign your IP address for the ppp connection, then also use
noipdefault
You could also put in a line containing debug here but we will use it explicitly on the command line.]


Figuring out how your ISP wants to log you on

Unfortunately there is no one procedure for logging on to an ISP. Each one uses some slightly different procedure. The following steps through the various possibilities to see what the ISP wants. Note that your ISP may have been creative and found some other way to confuse the users. This technique should still help you figure out what your ISP really wants.

In a separate window set up a running tally of the entries to the log file with
tail -f /var/log/ppp
This will report in the window what is logged to the /var/log/ppp file as it is logged. It sometimes scrolls by too fast to see-- then use the window bar to scroll back in the output. We are now going to systematically go through the options to find out what your ISP wants. During these experiments you may occasionally find that your pppd has not died (or you want to kill it). Just do
killall pppd
to kill it. Occasionally pppd will leave a lock file behind if it has been killed. If this is the case, do
rm /var/lock/LCK..ttyS1
where you replace ttyS1 by whichever serial port your modem uses.

Immediate PPP:

First run (all on one line)
/usr/sbin/pppd /dev/ttyS1 57600 debug connect "/usr/sbin/chat -v   ''   AT  OK  ATD5555555  CONNECT  '\d\c'"
(Note that that is doubled apostrophe ' after the chat -v, not a single double quote mark ".)
This command accomplishes two things-- it tests your modem and that you are able to call and have the modems connect, and it begins the process of figuring out how your ISP wants you to connect. This will not produce a ppp connection, but it should dial your phone, (where 5555555 is to be replaced by your ISP's phone number). (Note that I use 57600 as a conservative option for modem speed. If you are on a sufficiently fast system, and you have a new 56K or 33.6K modem, this should almost certainly be changed to 115200. However, I will stay conservative here to make sure it is not modem speed problems which are causing grief.)

For those with an interest, This command starts pppd on the terminal /dev/ttyS1 and sets the computer to modem speed at 57600 bps, tells pppd to log debugging messages to the file listed under daemon.* in /etc/syslog.conf, and says to use the program chat to connect. The chat program is told (-v) to log its debugging messages to the file listed under local2.* in /etc/syslog.conf, wait for nothing, send the command AT to the modem, wait for an return answer of OK, send the dialing command ATD5555555 to the modem, wait for CONNECT, then wait one second (\d) and then do NOT send anything-- especially not any carriage return back to the modem (\c). This last bit is important as for many ISPs if you send a carriage return, then assume you want to log on and give you a logon script, which can totally confuse everything.

[ If it does not dial your phone, then you will have to figure out on which port your modem is on, and perhaps send your modem some init string. For example to tell most modems to reset themselves to factory default, do the following (again all on one line)
/usr/sbin/pppd /dev/ttyS1 57600 debug connect "/usr/sbin/chat -v   ''   'AT&F0'   OK   ATD5555555   CONNECT   '\d\c' "
instead. You can add anything else you need to send to the modem either instead of the &F0 or after it.]
http://www.56k.com/inits/ contains modem initialisation strings for a large variety of modems.

[If there is a significant length of time between the log entry for the "send AT" "expect OK" lines and the resulting "got it" in the /var/log/ppp file, the modem is using a different interrupt line (IRQ) than the kernel is expecting. Try using the setserial command as follows:
setserial /dev/ttyS1 autoconfig auto_irq
If this corrects the problem put that line into /etc/rc.local or /etc/rc.d/rc.local or...]
(Although the man page for setserial and the setserial usage blurb state that the parameter is autoconfigure, the program actually uses autoconfig instead-- one of Linux's charms.--thanks to M. Cook for pointing this out.)

I will assume that this dialed your phone. This will not connect you to your ISP via ppp unless your ISP is incompetent (there is as yet no authentication) but it may have begun the ppp negotiations. We can use the debugging output of this command to determine what kind of authentication your ISP wants.

Which Authentication Scheme?

Look at the end of /var/log/ppp (eg do
less /var/log/ppp
To page back, use ^B (control B) The space bar will page forward. Or if you ran the tail -f command above look at the end of its output.) You should see a bunch of messages from chat, telling you what it sends and what it receives from the far end. (In this case it will end when chat receives the CONNECT string
got it (CONNECT)
from your modem)

Then pppd will start reporting, and will probably give some error message. One possibility is the message containing the line
Problem: all had bit 7 set to 0
This means that your ISP was not expecting you to negotiate ppp at this point. It almost certainly means that your ISP wants you to log on first.

You may at this point get no response from the far system at all-- ie, your system sends out LCP Config Requests and gets no response. Try replacing the '\d\c' in the above line with the word CLIENT, and try again. If this starts ppp negotiations, it indicates that you have a Win NT RAS server as your ISP. If so, in all of the discussions below, continue to replace `\d\c' with CLIENT.

Alternatively, One of the lines at the end of /var/log/ppp reported to be from pppd had something that started out with rcvd and then had something in it that looked like
<auth pap>
or
<auth chap ...>
(as an example here is one of mine
Jan 15 23:10:28 wormhole pppd[511]: rcvd [LCP ConfReq id=0x1 <mru 1524> <asyncmap 0xa0000> <auth pap> <pcomp> <accomp> < 13 09 03 00 c0 7b 63 d6 e6>] )
This means that they are ready to negotiate ppp and want to use pap (chap) authorization, not login authorization. If so, go down below to the PAP/CHAP authorization section.

Login Authorization?

Let us assume that they want login authorization (you got the "bit 7" error message). Try (on one line)
/usr/sbin/pppd /dev/ttyS1 57600 debug connect "/usr/sbin/chat -v   ''  ATD5555555  CONNECT   ''  ogin:  <yourusername>  assword:  <yourpassword>"
where <yourusername>; is your user name on the remote machine , and similarly for <yourpassword>. (Note that the <> surrounding the works yourusername and yourpassword are not to be included in your script. Note again that those are doubled apostrophes ' not single double quotes " inside the chat -v script, but are double quotes " surrounding it.)

Again look in /var/log/ppp.
You should see chat logging you in (sending your remote name and your password). If not, look at what chat received from the far end to get a clue as to what they expected. For example on some machines the login request comes via a Name: or Username: request instead of Login:. In that case, change the "ogin:" to "ame:" instead in the above command line.
If both your username and your password got sent (both show up on the lines in /var/log/ppp) but you got a login rejection, check to make sure that you have the right password and username for the remote system.

If it logged you in but again you got a message saying the 7 bit is all zero, your ISP is expecting something else from you after you logged in. This is most likely a, ppp or a pppd command. Insert a
"" ppp
or
"" pppd
at the end of the chat string. In one case they put in a request "Do you want PPP? y/n" In that case put in
"PPP? y/n" "\dy"
at the end of the chat script instead. (The \d tells chat to wait one second, to make sure that the remote computer is ready to receive your "y". (Try one of these. If it does not work, the lines in /var/log/ppp from chat will give you a clue as to what was expected).

Occasionally, your ISP will want both login authorization and PAP or CHAP authorization. You will see this by the <auth pap> or <auth chap ...> in the pppd lines in /var/log/ppp file after you have logged in. In this case go to the PAP/CHAP section and follow those directions as well.

If, in the var/log/ppp file you see a line giving your local and the remote IP address, you are connected and skip the next section.

PAP/CHAP

Let me assume that in one of the various lines in /var/log/ppp, there was the phrase <auth pap> (<auth chap ...>). This means that the remote system wants to use PAP (CHAP) authentication. Let me first make a digression on the types of CHAP authentications.

Types of CHAP

In the case of CHAP, there is an extra number after the <auth chap ..>, indicated by the dots, which indicates which type of CHAP authentication they are using (Yes, there are different types). The 05 one (or md5) is standard, and your system will have no problem with it. The types 80 (also called m$oft) and 81 are special Microsoft types. Your pppd will state in /var/log/ppp if it does not support them with error messages like-- unknown digest type or Unknown CHAP code 80 received..
Your pppd (certainly in the 2.3.x series) can and may already support type 80 (m$oft). In this case you are OK. The only thing to beware of is that the username in chap-secrets file and in the user option to pppd may need the special domain addition.

Similarly if you see something like
.... < auth 0xc027 01 ....> ...
in one of the lines from the far end, they want a patented version of PAP called Shiva PAP (or SPAP). Because of those patents, Linux does not support it. This is probably an NT server, and should also accept other versions of authentications if properly set up (a big if).
If your version of pppd does not support type 80 (m$oft) (the one supplied by all distributions do, and the source code pppd2.4.1 enables this automatically), it may be possible to recompile your pppd from source to support the type 80 chap. I leave that exercise to you.
Read the file
README.MSCHAP80 from the pppd source for hints. Also see the file http://www.van-dijk.net/PPP-NT-HOWTO/PPP-NT-HOWTO.html
Often a server will first try to see if you are willing to use the chap 80 but if your system does not agree, they will fall back to asking for either chap 05(md5) or pap.
Finally note that there are two separate type 80 (m$oft) chap implementations. The older, obsolete standard is Microsoft's lanman standard. Microsoft's newer is the default NT standard. If your ISP uses the older standard, (and you can only find that out from them) and your pppd has been compiled to support type 80 and this MSLANMAN option, then you can persuade pppd to use the older one by use of the ms-lanman option to the pppd command.]
Note that the Windows 98 server only uses the ms-lansman version of the chap 80 protocol(See Andrew Sun's comments on this). Many versions of pppd have not been compiled to support this version. Thus, you must either download the source for pppd (see the above sources) and compile pppd with
make -DMSLANMAN
try to get the system to use the pap authentication, or (preferably) flee from that ISP as quickly as you can.
If your ISP uses type 81 and refuses to use anything else, yell at them for using this new Microsoft non-standard. If they refuse to use anything else (eg chap 05 or chap md5) then note that efforts are being made to also support MSChap 81 in Linux.
See the PPPTP- client Project which includes patches for chap 81 and MPPE support and docs for setting up a PPTP client .

Setting up PAP/CHAP

You now need to make sure that the remote system gets the proper PAP/CHAP authentication. There are two steps here.

a) Edit the file /etc/ppp/pap-secrets (/etc/ppp/chap-secrets).
You will now add a line to this file. The first entry in the line is your user name on the remote system. The second is a *. The third is your password and the fourth can also be a *. Thus there will be a line like

<yourusername>     *     <yourpassword>        *

eg.
unruh                *       dontyouwish             *

(This means that this line is the pap (chap) secret for user <yourusername> on any remote system (*) and <yourpassword> is that secret. Also the connection can use any IP address-- the second *.)
The second entry (first star) may have to be replaced by the name of the remote system if your ISP told you to do so or you have accounts on many ISPs. The last star may have to be removed. But this line as written should work for a single ISP.

If you have accounts on multiple systems, then you will have to replace the second item (first *) with the name for the remote system so your system knows which secret to use for that particular remote system. There are three options for that name.
(i)You can ask your ISP for a name.

(ii)You can look in /var/log/ppp for a line like
pppd[1155]: rcvd [CHAP Challenge id=0x1 <...>, name = "www.theory"]
where ... is a long string of random numbers and letters. That name (www.theory in this case) is the name the remote system thinks of itself as. chap 80 (Microsoft chap) does not report the remote system name.

(iii)Or you can assign the remote system an arbitrary name, put the option
remotename <TheNameYOuDecidedOn>
after the pppd command.
Whichever option you choose, use that name as the second field in the line in the chap-secrets file (or pap-secrets).

If your ISP uses NT, you may have to add a domain name to your username. Thus, in both the chap secrets file and in the "user" option to pppd, instead of <username>, use <domainname/username> instead. (or in some cases it has been reported as <domainname\\username>) You will have to get the domainname from your ISP.

[Note that PAP or CHAP also has the option for symmetric authentication, where your machine also requires authentication of the remote system. For most home user systems, this will not be used--your ISP will refuse to agree to authenticate themselves-- and the above is sufficient. If in your /var/log/ppp file you see your system asking the remote system for authentication-- ie a line like
Jan 15 23:10:28 wormhole pppd[511]: sent [LCP ConfReq ... <auth pap> ...
ie your system sent (not received) a request (ConfReq) for pap authentication, then your system wants the other side to authenticate themselves, which they will almost certainly refuse to do. Put the line
noauth
into your /etc/ppp/options file and remove any options like +chap, +pap, require-pap, require chap, auth from the /etc/ppp/options file or from the pppd command line. Some new versions of pppd may, by default, require the remote system to authenticate themselves, and will thus need the noauth option.]

b) In both the case of PAP and CHAP you also have to use the "user" option to pppd, to let the remote machine and your machine know what your user name is for PAP/CHAP authentication when looking up secrets in the pap-secrets or chap-secrets files. By default pppd uses the name of your local machine, which is probably not your user name on the remote machine. So now try
/usr/sbin/pppd /dev/ttyS1 57600 debug user <yourusername> connect "/usr/sbin/chat -v <chatstring>"
where <chatstring> is whichever chat string successfully got you to this point. (Remember that the < and > are not to be included in strings.)
Eg., here would be a line for my system
/usr/sbin/pppd /dev/ttyS1 57600 debug user unruh connect "/usr/sbin/chat -v '' ATDT5551234 CONNECT '\d\c' "

Occasionally the remote system is broken, and after they have asked for PAP authentication, they seem to refuse to listen to your systems sending them the requested information. The symptom will be your system sending a whole string of PAP authentication attempts
.... sent [PAP AuthReq id=0x1 user="username" password="password"]
.... sent [PAP AuthReq id=0x2 user="username" password="password"]
....
with no response from the other side. Try putting the line
asyncmap 0xa0000
or even
default-asyncmap
into your /etc/ppp/options file.

Occasionally you may find, in reading /var/log/ppp, that the remote end cannot seem to hear you. Your side sends requests to the far side, and the far side keeps sending back the same requests to you. The session will terminate after a while. You may have a misconfigured serial port. Run
setserial /dev/ttyS1
and make sure that the UART listed is actually the same as the one on your serial or modem card. Almost all newer computers ( since the mid 90s) use the 16550A UART. This is different from the 16550 UART or 16450.

Are you connected?

You are now, I hope, connected via ppp. The /var/log/ppp file should have a line like
Connect: ppp0 <--> /dev/ttyS1
1 Now, run
/sbin/route -n
and look for a default (0.0.0.0) option on the ppp0 link (ppp0 is the last item in the line, and 0.0.0.0 is the first). If so, success, you are connected. If not, you now at least have the far end negotiating for a ppp connection. Your /var/log/ppp file should now have lines which read
sent [LCP ConfReq ... rcvd [LCP ConfAck ...
pppd will also report in the log file your local and the remote IP addresses. If so, you are connected.

Connected!!:

At this point you should be connected. You should see lines like
Jan 16 14:54:50 wormhole pppd[905]: local IP address 137.82.66.22
Jan 16 14:54:50 wormhole pppd[905]: remote IP address 137.82.47.122
in /var/log/ppp. (The above two lines are from my own system. The addresses, names, dates and times will vary for yours, but the form should be the same.)

ppp is now connected.

Testing Connection

First try to ping the address of the remote host. In /var/log/ppp there will be two lines that look like
Feb 7 12:52:21 ppptest pppd[27378]: local IP address 137.82.43.5
Feb 7 12:52:21 ppptest pppd[27378]: remote IP address 142.103.234.16
(Note that these are for a particular connection on my system. The numbers will differ on yours) These two lines give your and the remote computer's IP numbers for the duration of the connection. (They may well change the next time you connect.)

Again run
/sbin/route -n
and look for an entry in which the first entry is that remote IP address, and the last entry is ppp0. It must be there for the connection to work.

Then do
ping <remoteIPnumber>
eg in the above case this would be
ping 142.103.234.16
You should get back a sequence of lines about one every second. This shows that your connection to that remote host is OK, and that you have a connection to it. Type ^C (control-C) to stop ping. (Sometimes the remoteIPnumber does not respond to pings because of the way the ISP has set it up. Do not worry yet unless you get a Network or Host unreachable message.)

Now, look at those lines that you added to /etc/resolv.conf back at the beginning of this script. Do
ping <nameserverIP>
where the entry in /etc/resolv.conf was
nameserver <nameserverIP>
Again you should get a series of responses. If you do not, they may not support ping on those machines. However, if you get a response saying that the host or network is unreachable, there is definitely something wrong, either with your setup or with the nameserver address they gave you.

Now try
ping 142.103.234.29
(Using that actual number)
which should work (unless of course there is something wrong with my machine. In which case try other IP numbers.) If none of these work, then there is some problem with your connection (eg default route problems).

Now try pinging some name, eg.,
ping www.theory.physics.ubc.ca
(which is the location where this page is kept). The key thing is that ping reports back on its first line with the IP address of www.theory. If it did so, it means your nameserver is working. If the previous one worked, but this one does not, then you either forgot to follow the initial instructions on this page about setting up the file /etc/resolv.conf, or your ISP has nameserver problems.
Try a few more names of machines that you know. If all of these tests have passed, try running Netscape and connecting to some page. Everything, I hope, now works.

If not, collect all of the details you can (eg the output in /var/log/ppp-- making sure that you remove your passwords from that file) and ask for help in comp.os.linux.networking, alt.os.linux.dial-up, or alt.comp.linux.isp.
As politeness, look in the archives on www.dejanews.com since your problem may already have been answered ( many times). When you ask for help or search, make your subject and description as specific as possible. "Nothing works" as a symptom is very very hard for anyone to suggest fixes for. 



 

Various Possible Problems

The PPP setup returns the error "Could not determine remote IP address". This means that your ISP has a damaged implementation of ppp, which does not know (or refuses to report) who it is. So you need to assign your ISP an IP number. Put the lines
:192.168.255.1
ipcp-accept-remote

into the /etc/ppp/options file. These assign your ISP the IP number 192.168.255.1 (a "reserved" IP number) since they refuse to tell you theirs, and also tells pppd to accept their version if at this point they wake up and finally send you an IP address for themselves.
You might also notify your ISP that maybe they should know their own IP number.

ping works --ie the ping tests above all worked--, but telnet, or surfing does not seem to work.
Again your ISP seems to have a defective ppp. Try putting the line
novj
into /etc/ppp/options. This disables compression of the headers on the packets which seems to cause problems for some ISPs..

Miscellaneous Error messages

There are a few error messages which crop up with the newer versions of ppp and the newer (2.2.x) kernels.
On the 2.3.9 and 2.3.10 versions of pppd, the message
modprobe: can't locate module char-major-108
is displayed in the log file. Ignore this. These versions of pppd use some new features of the ppp kernel drivers in the development kernels on linux if they are available. This message says ppp cannot find these features. This is fine as these versions of ppp also work perfectly well with the older (2.0.x and 2.2.x) kernels.

You may see a series of lines like
modprobe: Can't locate ppp_compress-21
or 24 or 26. Put the following three lines into /etc/conf.modules (/etc/modules.conf on the 2.4.x series of kernels)
alias ppp-compress-21 bsd_comp
alias ppp-compress-24 ppp_deflate # From original RFC draft
alias ppp-compress-26 ppp_deflate # Final standard per ppp-2.3.4

Again these error messages cause no harm, except, in the unlikely event that the remote side supports these compression modes, you will not be able to use these compressions. (Modems nowadays do compression, so the advantages of software compression are often minimal).

 

Automation and Cleanup

Typing the above command line every time you want to connect can get tiring and be prone to errors. It also suffers from the problem that it must be run by root, not an ordinary user, and is also insecure in that details like passwords are left lying around on the command line for anyone who can run the command ps to see. Thus we want to make the programs runnable by any user, make them easy to use, and we want to make sure that it is at least somewhat secure.

SUID

To allow ordinary users (or usually you as an ordinary user) to use ppp, some changes need to be made. Just make sure that the pppd file is rx permission for users
chmod a+rx /usr/sbin/pppd
and that /usr/sbin/pppd is setuid for root
chmod +s /usr/sbin/pppd In addition, you may have to make the serial port accessible by your users. Assuming that your port is /dev/ttyS1 (COM2) do
chmod a+rw /dev/ttyS1
Note that some programs (like linuxconf or mgetty) like to change the permissions of the serial port back to a more protected form. So, don't run linuxconf or mgetty, or get used to having to keep resetting the permissions of the serial port. Alternatively, get one of the more recent versions of pppd (eg 2.3.9). If you put the port (eg., /dev/ttyS1) into /etc/ppp/options, instead of on the pppd command line as I do in the scripts below, pppd will connect to the port as root, rather than as user. Then you do not have to worry about the permissions on the port.
(If you have users who you do not want using your modems, edit /etc/group and insert a line like
ppp::25:<Comma separated List of names of users>
where the List of users is a comma separated list of the users you want to allow use of ppp. The group number -- 25 in the above example-- is arbitrary, but should not be the same as any other group in /etc/group. Then do
chown root.ppp /usr/sbin/pppd
chmod o-rx /usr/sbin/pppd
in addition to the above changes.)
Alternatively if you are running mgetty put your modem users into the group uucp, which should already be defined in /etc/group. (mgetty keeps resetting the ownership and permissions of the device, eg, /dev/ttyS1, to have group uucp, and no rw permissions for ordinary users)

Script

Let us now automate the running of pppd by writing two scripts. The first script file will contain the commands, while the second will contain the expect/send pairs that you have found that chat needs to log you on. The first file I will call pppon (to distinguish it from the ppp-on which comes with your distribution), while the second will be chatscript. The pppon script is simple, but comes in two versions. The one for the case in which you discovered that your ISP expected you to log on
pppon
===============================================
#!/bin/sh
#This script is for the case where you log on to your ISP
/usr/sbin/pppd /dev/ttyS1 57600 connect "/usr/sbin/chat -f $HOME/chatscript"
----------------------------------------------
In this case I have assumed that each user will keep their chatscript in their home directory. Replace $HOME with the path if you want it kept elsewhere.
It is important to note that I have removed the -v from the chat command and the debug option from the pppd command. This is important as otherwise your password will be recorded in the /var/log/ppp file-- a bad idea.

In this case, where the remote ISP wants you to log in, the chat script is most complex. I will also add some extra features to the chat script to make it more robust than the minimal commands I used above. (The lines starting with # are comments which may safely be removed or left in.)
chatscript

============================================
                      ABORT "NO CARRIER"
                      ABORT "NO DIALTONE"
                      ABORT "ERROR"
                      ABORT "NO ANSWER"
                      ABORT "BUSY"
#The above lines indicate conditions for chat quitting
                      ""    AT
                      'OK-\d\d+++\d\d\c-OK'   ATH0
# The A-B-C form of the expect sequence means that it is to expect A,
# if A does not arrive, send B and then expect C.
# So this line means-- expect OK, if it does not arrive, wait two
# seconds,(each \d is a one second wait) send +++
# wait for two more seconds and then do not send a carriage return.
# Finally again wait for OK again. If it arrives, send ATH0
# The +++ is in case your modem has not hung up. You may want to know if your 
# modem is not hanging up, if so, remove this whole line.
              OK    'AT&F\\Q3\\V1&D3M1#CC1'
# These are various controls sent to my particular modem. CHANGE FOR YOURS!
#The &F,  &D3,  and M1 seem to be quite generic for setting the modem
# with factory defaults,  allowing the DTR line to reset the modem
#(Used when pppd shuts itself down), and  M1 allows the speaker on while the
#call is being established, but off after the remote system has answered.
#USR/3COM Sportster modems require &F1 instead of &F.
# \Q3, \V1, and #CC1 are almost certainly peculiar to my particular
# modem.
# Note that a backslash must be doubled to get through. This sequence 
# should also always be enclosed in single  quotes
# because of the special characters like &.\,#.
              OK     ATD8765432
              CONNECT ""
              ogin:--ogin:    unruh
# Again the A-B-C, except the middle B to be sent if ogin: is not
# received is nothing except a  carriage return
              assword:      "dontyouwish"

---------------------------------------------------------------\
If in your determination of what your ISP wanted, you found you had to expect other text (eg 'PPP? y/n') and send a response, insert those onto the end of this file. Also, since your username, password and phone numbers are highly unlikely to be as in this script, change those.

The other possibility is that the ISP uses PAP/CHAP authorization. In this case the pppon script needs the "user" option. You can put it into the file permanently (eg, if you will ever only use a single ISP) or, as in the following, you can allow it to be given as an option to pppd.
pppon

==============================================================
#!/bin/sh
if [ "$1" = "" ] ;then
   echo "Usage: pppon username"
   exit 1
fi
/usr/sbin/pppd /dev/ttyS1 57600 user $1 connect "/usr/sbin/chat -f $HOME/chatscript"
----------------------------------------------------------------
This script asks for your username on the remote system as part of the options to the pppon. You could replace the lines between the if and fi and just put the username in instead of the $1 in the pppd command line if you wished.
It is important to note that I have removed the debug option to pppd. Under especially PAP, using debug will store your PAP password in the /var/log/ppp file, which is not what you want. Additionally, erase your /var/log/ppp file at this point to remove the possibility of anyone reading the password in that file.
The chatscript here is simpler, since you do not need to log on (for an explanation of the various lines in this script, look at the above example where I set up a sample chatscript for the case the ISP expects you to log on first).
chatscript
===============================================================
              ABORT "NO CARRIER"
              ABORT "NO DIALTONE"
              ABORT "ERROR"
              ABORT "NO ANSWER"
              ABORT "BUSY"
              ""    AT
              OK-\d\d+++\d\d\c-OK   ATH0
              OK    'AT&F\\Q3\\V1&D3M1#CC1'
              OK     ATD8765432
              CONNECT '\d\c'
# The \d means wait a second, the \c means do not send a carriage
# return (sending a carriage return at this point almost always confuses
# the ISP)
--------------------------------------------------------------------

When you have created the above scripts, (perhaps with the debug option for pppd in pppon and adding a -v after the chat command so you can see what goes wrong if something is wrong with the script--- just remember to remove both of these once you have finished testing), copy them to their final home.

mv pppon /usr/bin
chmod a+rx /usr/bin/pppon
chown <user> chatscript
su <user>
mv chatscript $HOME
chmod go-rwx $HOME/chatscript

Note that we have made the chatscript readable only by the user. This is important since it may contain the user's password (in the logon case).

Extra Security

When you are connected to the internet via ppp, others are also connected to you, and this could open you to break-ins while you are connected. To help alleviate this, put the line
ALL: ALL
into the file /etc/hosts.deny and
ALL: 127.0.0.1 127.0.0.0
into /etc/hosts.allow. This will disallow all external hookups by anything to your machine. If you want more detailed control, read
man 5 hosts_access
For example, if you have an internal Ethernet network you obviously want to allow them, but not others to log on to your system.

Stopping PPP

To stop the connection, use the command
killall pppd
This sends a signal to the pppd program telling it to shut down. pppd shuts itself down by dropping the DTR line to the modem, and by running the commands in /etc/ppp/ip-down and /etc/ppp/ip-down.local. The biggest problem with the above is that the command may not work unless you are logged in as root. (I have had this trouble for a while, others, including me now, have not.) If the above does not work (ie your ppp connection stays up and your modem does not hang up after a minute or so) you will either have to log on as root and then run the above command, or you can use a "wrapper" so that any user can run the above command as root. This wrapper is a C program which makes running the command as suid root safe. The details are given on the web page
http://www.theory.physics.ubc.ca/ppp-kill-wrapper.html

Since pppd tries to hang up the modem by switching the DTR line to the modem, the modem must be set up to reset the modem when this occurs. The modem will eventually hang itself up if pppd stops, but it takes a while (a minute or so at times). thus it is useful (but not crucial) that you set up your modem to use the DTR line to reset the modem. You must look up the command in your documentation for your modem. For the three different modems I have, the command is &D3 (Reset modem on ON to OFF transition of DTR). So you would add &D3 to the modem initialisation string in the chat script files. Ie, the line
OK 'AT&D3'
should be added just before the line where you dial up your service provider.

One useful safety net (against running up huge bills with your ISP and making your family angry at your hogging the phone) is that pppd has an option to hang up the connection if no traffic has gone across the ppp network in a certain amount of time. Enter the line
idle 1800 into /etc/ppp/options file. This will hang up the phone if there has been no traffic for 1800 seconds (30 min). Change that number (1800) if you want this idle option to work more quickly (or slowly).
Note that pppd will interpret anything on the link as resetting the idle timer. Thus, and LcpEcho request (sent out by some ISPs to see if you are still alive) will reset the timer and stop the idle timeout from ever operating. However as of the 2.4.x series of kernels, the ppp kernel has the option of active filtering which will disregard certain items coming down the ppp line for purposes of the idle timer. The kernel must have been compiled with the
PPP_FILTER
option. Whether yours has been depends on the distribution.


You might want your system to do demand dialing-- ie, to have your system phone your ISP on its own every time you (or your computer) wants to connect. (This is especially handy if you have a number of computers connected to yours and yours acts as the gateway via PPP to the outside world). To do so use the program diald from http://diald.sourceforge.net/.
An alternative to diald is masqdialer which some have found easier to set up than diald. It allows users on other machines in your network to bring up and down the connection on your linux server.
Finally, the newer versions of pppd and the kernel (2.4.x series) have the demand option, which will also bring up the connection when something tries to send out a request to some machine on the net. diald has better filtering abilities to filter out things which you do not want to bring up the interface (eg some of the Windows internet noise).
To have others connect to your machine, (ie, to have yourself act as an ISP), use mgetty from http://mgetty.greenie.net/. Also get the guide http://www.swcp.com/~jgentry/dialin2.html to setting up dial in access or see ./mgetty.html.
Here, http://www.theory.physics.ubc.ca/mgetty-point.html are also a number of links to documentation to getting mgetty up and running (including with multiple incoming telephone lines).

For a variety of scripts, which can be of help in a some situations, see C Kite's site http://users3.ev1.net/~ckite/public_html/download/index.html
In particular
For dialing a number of ISPs in a row if they are busy who all use chap or pap authentication: http://users3.ev1.net/ckite/public_html/download/chap-many2one.gz
For a patch to chat (although this is by now an old version of chat) to allow use of SecureID see: http://users3.ev1.net/ckite/public_html/download/SecurID.gz
For another "SecurID" type one time password patch see http://users3.ev1.net/ckite/public_html/download/chatbypipe.tar.gz
For a discussion on RAS callback for Linux ppp, see http://users3.ev1.net/ckite/public_html/download/RAScb.gz
These scripts have not been tested by me, so use at your own risk.


Finally, if it is really kppp which you want to run, see http://www.beginnerslinux.org -> Connecting to the Web


A really big helping of thanks to staehle@nevalue.net for a lot of help in whipping this page into shape. Thanks also to pacman@cqc.com for helpful improvements to this document. Finally, but not least, I would thank James Carlson and Clifford Kite from whose posting on comp.protocols.ppp and elsewhere I have learned a lot of what I know about ppp.


This document was written by W. G. Unruh, . I would appreciate any feedback you can give, especially if your ISP has some other strange way of authenticating you. Also if anything in this document is ambiguous, hard to follow, or wrong, please let me know.

Good Luck in your connection.

Last modified Mar 23, 2002