CD Actual 13/CDA13.ISO/DOC/HOWTO/OTHER_FO/SGML/IPX_HOWT.GZ/IPX-HOWTO.sgml

SGML Document | 1996-06-02 | 64.4 KB | 1,683 lines

<!doctype linuxdoc system>

<!-- This is the Linux IPX-HOWTO, please forward any comments or suggestions
     to the author: [e-mail]
 -->

<article>

<!-- Title information -->

<title>Linux IPX-HOWTO
<author>Terry Dawson, <tt>[e-mail]</tt>
<date>v2.0, 1 June 1996

<abstract>
This document aims to describe how to obtain, install and configure various
tools available for the the Linux operating system that use the Linux kernel
IPX protocol support.
</abstract>

<!-- Table of contents -->
<toc>

<!-- Begin the document -->

<sect><heading>Introduction.

<p>
This is the Linux IPX-HOWTO. You should read the Linux NET-2-HOWTO in
conjunction with this document.

<sect1><heading>Changes from the previous release.

<p>
<verb>
Additions:
    FAQ section - thanks to Gregory Steuck, Volker Lendecke, Alan Cox
    PPP section - thanks Gilbert, David, and Chad.
    details of 'linware' list.
    ipx_internal_net to router section - comments ?

Corrections/Updates:
    Copyright is now GNU GPL version 2.0
    Updated location of Greg's IPX tools.
    Suggest Volkers ipxd in place of old daemons.
    Updated locations of Volkers ftp site location.
    Updated versions of ncpfs and mars_nwe.
</verb>

<sect1><heading>Introduction.

<p>
The Linux Kernel has a completely new network implementation as compared
to other Unix like operating systems. The ability to take a fresh approach
to developing the kernel networking software has led to the Linux kernel
having support for a range of non tcp/ip protocols being built. The IPX
protocol is one of those that have been included.

The Linux kernel supports the IPX protocol only. It does not yet support
protocols such as IPX/RIP, SAP or NCP, these are supported by other software
such as that documented elsewhere in this document.

The IPX support was originally developed by Alan Cox
<tt><[e-mail]></tt> and has been significantly enhanced by
Greg Page <tt><[e-mail]></tt>.

<sect><heading>Disclaimer.

<p>
I do not and cannot know everything there is to know about the Linux network
software. Please accept and be warned that this document probably does contain
errors. Please read any README files that are included with any of the various
pieces of software described in this document for more detailed and accurate
information. I will attempt to keep this document as error-free and up-to-date
as possible. Versions of software are current as at time of writing.
<p>
In no way do I or the authors of the software in this document offer protection
against your own actions. If you configure this software, even as described in
this document and it causes problems on your network then you alone must
carry the responsibility. I include this warning because IPX network design
and configuration is not always a simple matter and sometimes undesirable
interaction with other routers and fileservers can result if you do not design
or configure your network carefully. I also include this warning because I
was asked to by someone unfortunate enough to have discovered this lesson the
hard way.

<sect><heading>Related Documentation.

<p>
This document presumes you understand how to build a Linux kernel with
the appropriate networking options selected and that you understand how
to use the basic network tools such as <em>ifconfig</em> and <em>route</em>.
If you do not, then you should read the <url url="http://sunsite.unc.edu/mdw/HOWTO/NET-2-HOWTO.html" name="NET-2-HOWTO">
in conjunction with this document as it describes these.
<p>
Other Linux HOWTO documents that might be useful are:
<p>
The <url url="http://sunsite.unc.edu/mdw/HOWTO/Ethernet-HOWTO.html"
name="Ethernet-HOWTO">, which describes the details of configuring an
Ethernet device for Linux.

The <url url="http://sunsite.unc.edu/mdw/HOWTO/PPP-HOWTO.html"
name="PPP-HOWTO"> as IPX support is available for version 2.2.0d and later of
the Linux PPP implementation.

<sect1><heading>New versions of this document.

<p>
If your copy of this document is more than two months old then I strongly
recommend you obtain a newer version. The networking support for Linux is
changing very rapidly with new enhancements and features, so this document
also changes fairly frequently. The latest released version of this document
can always be retrieved by anonymous ftp from:

<bf>sunsite.unc.edu</bf>
<tscreen><verb>
/pub/Linux/docs/HOWTO/IPX-HOWTO
</verb></tscreen>
or:
<tscreen><verb>
/pub/Linux/docs/HOWTO/other-formats/IPX-HOWTO{-html.tar,ps,dvi}.gz
</verb></tscreen>

via the World Wide Web from the
<url url="http://sunsite.unc.edu/mdw/linux.html"
name="Linux Documentation Project Web Server">, at page:
<url url="http://sunsite.unc.edu/mdw/HOWTO/IPX-HOWTO.html" name="IPX-HOWTO">
or directly from me, <tt><[e-mail]></tt>. It may
also be posted to the newsgroups: <tt>comp.os.linux.networking</tt>,
<tt>comp.os.linux.answers</tt> and <tt>news.answers</tt> from time to time.

<sect1><heading>Feedback.

<p>
Please send any comments, updates, or suggestions to me,
<tt><[e-mail]></tt>. The sooner I get feedback, the
sooner I can update and correct this document. If you find any problems
with it, please mail me directly as I now very rarely read the newsgroups.
You might also catch me as <em>terryd</em> on the <tt>#linpeople</tt> IRC
channel on the <tt>undernet</tt> IRC network.

<sect1><heading>Mailing list support.

<p>
There is a mailing list established for discussion of the various Linux
IPX software packages described in this document. You can subscribe to it
by sending a mail message to <tt>[e-mail]</tt> with
<tt>add linware</tt> in the body of the message. To post to the list your
send your mail to <tt>[e-mail]</tt>.
<p>
The mailing list is archived at
<url url="http://www.kin.vslib.cz/hypermail/linware/" name="www.kin.vslib.cz">.

<sect><heading>Some of the terms used in this document.

<p>
You will often see the terms <tt>client</tt> and <tt>server</tt> used in this
document. They are normally fairly specific terms but in this document I have
generalized their definitions a little so that they mean the following:
<descrip>
<tag>client</tag>The machine or program that initiates an action or a
connection for the purpose of gaining use of some service or data.
<tag>server</tag>The machine or program that accepts incoming connections from
multiple remote machines and provides a service or data to those.
</descrip>

These definitions are not very reliable either, but they provide a means of
distinguishing the ends of peer to peer systems such as <em>SLIP</em> or
<em>PPP</em> which truly do not actually have clients and servers.
<p>
Other terms you will see are:
<descrip>
<tag>IPX</tag>Internet Packet eXchange is a protocol used by the Novell
corporation to provide internetworking support for their NetWare(tm) product.
IPX is similar in functionality to the IP protocol used by the tcp/ip community.
<tag>IPX network address</tag>This is a number which uniquely identifies
a particular IPX network. The usual notation for this address is in
hexadecimal. An example might look like: 0x23a91002
<tag>IPX Internal network</tag>This is a virtual IPX network. It is virtual
because it does not correspond to a physical network. This is used to provide
a means of uniquely identifying and addressing a particular IPX host. This
is generally only useful to IPX hosts that exist on more than one physical
IPX network such as fileservers. The address is coded in the same form as
for a physical IPX network.
<tag>RIP</tag>Routing Information Protocol is a protocol used to automatically
propagate network routes in an IPX network. It is functionally similar to the
RIP used within the tcp/ip community.
<tag>NCP</tag>NetWare Core Protocol is a networked filesystem protocol designed
by the Novell Corporation for their NetWare(tm) product. NCP is functionally
similar to the NFS used in the tcp/ip community.
<tag>SAP</tag>Service Advertisement Protocol is a protocol designed by the
Novell Corporation that is used to advertise network services in a NetWare(tm)
environment.
<tag>Hardware address</tag>This is a number that uniquely identifies a host
in a physical network at the media access layer. Examples of this are
<em>Ethernet Addresses</em>. An Ethernet address is generally coded as
six hexadecimal values separated by colon characters eg. 00:60:8C:C3:3C:0F
<tag>route</tag>The <em>route</em> is the path that your packets take
through the network to reach their destination.
</descrip>

<sect><heading>The IPX related files in the <tt>/proc</tt> filesystem.

<p>
There are a number of files related to the Linux IPX support that are located
within the <tt>/proc</tt> filesystem. They are:

<descrip>
<tag>/proc/net/ipx_interface</tag>This file contains information about the IPX
interfaces configured on your machine. These may have been configured manually
by command or automatically detected and configured.
<tag>/proc/net/ipx_route</tag>This file contains a list of the routes that
exist in the IPX routing table. These routes may have been added manually by
command or automatically by an IPX routing daemon.
<tag>/proc/net/ipx</tag>This file is a list of the IPX sockets that are
currently open for use on the machine.
</descrip>

<sect><heading>Greg Pages IPX tools.

<p>
Greg Page <tt><[e-mail]></tt> of Caldera Incorporated has written
a suite of IPX configuration tools and enhanced the Linux IPX kernel support.
<p>
The kernel enhancements allow linux to be configured as a fully featured
IPX bridge or router. The enhanced IPX support has already been fed back into
the mainstream kernel distribution so you will probably already have it.
<p>
The network configuration tools provide you with the capability to configure
your network devices to support IPX and allow you to configure IPX routing
and other facilities under Linux. The Linux IPX network tools are available
from:
<url url="ftp://sunsite.unc.edu/pub/Linux/system/Filesystems/ncpfs/ipx.tgz"
    name="sunsite.unc.edu">.

<sect1><heading>The IPX tools in more detail.

<p>
<descrip>
<tag>ipx_interface</tag>This command is used to manually add, delete or check
ipx capability to an existing network device. Normally the network device
would be an Ethernet device such at <tt>eth0</tt>. At least one IPX interface
must be designated the <em>primary</em> interface and the <em>-p</em> flag
to this command does this. For example to enable Ethernet device <tt>eth0</tt>
for IPX capability as the primary IPX interface using the IEEE 802.2 frame type
and IPX network address <tt>39ab0222</tt> you would use:
<tscreen><verb>
# ipx_interface add -p eth0 802.2 0x39ab0222
</verb></tscreen>

<tag>ipx_configure</tag>This command enables or disables the automatic setting
of the interface configuration and primary interface settings. The
<tt>--auto_interface</tt> argument allows you to select whether new network
devices should be automatically configured as IPX devices or not. The
<tt>--auto_primary</tt> argument allows you to select whether the IPX software
should automatically select a primary interface or not. A typical example
of this commands use would be to enable both automatic interface configuration
and automatic primary interface setting:
<tscreen><verb>
# ipx_configure --auto_interface=on --auto_primary=on
</verb></tscreen>

<tag>ipx_internal_net</tag>This command allows you to configure or deconfigure
an internal network address. An internal network address is optional, but when
it is configured it will always be the primary interface. To configure an
IPX network address of <tt>ab000000</tt> on IPX node <tt>1</tt>
you would use:
<tscreen><verb>
# ipx_internal_net add 0xab000000 1
</verb></tscreen>

<tag>ipx_route</tag>The command allows you to manually modify the IPX routing
table. For example to add a route to IPX network <tt>39ab0222</tt> via
a router with node number <tt>00608CC33C0F</tt> on IPX network
<tt>39ab0108</tt>:
<tscreen><verb>
# ipx_route add 0x39ab0222 0x39ab0108 0x00608CC33C0F
</verb></tscreen>
</descrip>

<sect><heading>Configuring your Linux machine as an IPX router.

<p>
If you have a number of IPX segments that you wish to internetwork you
need the services of a router. In the Novell environment there are two
pieces of information which are necessary to be propagated around the network.
They are the network routing information propagated using Novell RIP, and
the service advertisement information propagated using Novell SAP. Any
router must support both of these protocols to be useful in most situations.
<p>
Linux has support for both of these protocols and can be fairly easily
made to function as a fully Novell compliant router.
<p>
The Linux kernel IPX support actually manages the IPX packet forwarding
across interfaces, but it does this according to the rules coded into the
IPX routing table. Linux needs a program to implement the Novell RIP and SAP
to ensure that the IPX routing table is built correctly and updated periodically
to reflect changes in the network status.
<p>
Volker Lendecke <tt><[e-mail]></tt> has developed a routing
daemon that will do this for you.
<p>
You can find <em>ipxripd</em> at:

<url url="ftp://sunsite.unc.edu/pub/Linux/system/Filesystems/ipxripd-0.7.tgz"
    name="sunsite.unc.edu">

or at Volkers home site at:

<url url="ftp://ftp.gwdg.de/pub/linux/misc/ncpfs/ipxripd-0.7.tgz"
    name="ftp.gwdg.de">
<p>
Configuring your Linux machine to act as a router is very straightforward.
The steps you must take are:
<enum>
<item>Build your kernel with IPX, Ethernet and <tt>/proc</tt> support.
<item>Obtain, compile and install the <em>ipxd</em> daemon program.
<item>Boot the new kernel and ensure that each of the Ethernet cards has
been properly detected and there are no hardware conflicts.
<item>Enable the IPX protocol on each of the interfaces using the
<em>ipx_interface</em> command described above.
<item>Start the <em>ipxd</em> daemon program.
</enum>
<p>
Consider the following simple network:

<verb>
 IPX Addr: 0x01000000  802.2
|--------------------------|
              |
              \_________________________
                                        \          Linux Router
 IPX Addr: 0x02000000  802.2             \
|--------------------------|              \   eth0/-----------\
              |                            \--====|           |
              \_________________________          | IPX route |
                                        \     eth1|  Table    |
 IPX Addr: 0x03000000  etherII           \----====|    ^      |
|--------------------------|                      |    |      |
              |                               eth2|  IPXd     |
              \______________________________/====|           |
                                                  |  SAPd     |
 IPX Addr: 0x04000000  etherII                eth3|           |
|--------------------------|                 /====|           |
              |                              |    \___________/
              \______________________________/
</verb>
<p>
The configuration for the above network would look like:
<tscreen><verb>
# ipx_interface add eth0 802.2 0x0100000000
# ipx_interface add eth1 802.2 0x0200000000
# ipx_interface add eth2 etherii 0x0300000000
# ipx_interface add eth3 etherii 0x0400000000
# ipxd
</verb></tscreen>
<p>
You should then wait a moment or two and check your
<tt>/proc/net/ipx_route</tt> file and you should see it populated with
the IPX routes relevant to your configuration and any learned from any other
routers in the network.

<sect1><heading>Do I need to configure an internal network ?

<p>
Novell has a feature called an internal network, which it uses to simplify
routing in situations where a host has more than one network device connected.
This is useful in the case of a fileserver connected to multiple networks
as it means that only one route needs to be advertised to reach the server
regardless of which network you are attempting from.
<p>
In the case of a configuration where you are not running a fileserver and
your machine acting only as an IPX router the question is not as simple to
answer. It has been reported that configuring for IPX/PPP works `better' if
you also configure an internal network.
<p>
In any case it is easy to do, but may require a rebuild of your kernel.
When you are working through the kernel <tt>make config</tt> you must answer
<tt>Y</tt> when asked <tt>Full internal IPX network</tt> as illustrated:

<tscreen><verb>
 ...
 ...
Full internal IPX network (CONFIG_IPX_INTERN) [N/y/?] y
 ...
 ...
</verb></tscreen>
<p>
To configure the internal network interface, use the <em>ipx_internal_net</em>
command described earlier in the IPX tools section. The main precaution to
take is to ensure that they IPX network address you assign is unique on your
network and that no other machine or network is using it.

<sect><heading>Configuring your Linux machine as an NCP client.

<p>
If you are a user of a mixed technology network that comprises both IP and IPX
protocols it is likely that at some time or another you have wanted to have
your Linux machine access data stored on a Novell fileserver on your
network. Novell have long offered an NFS server package for their
fileservers that would allow this, but if you are a small site or have only
a small number of people interested in doing this it is difficult to justify
the cost of the commercial package.

Volker Lendecke <tt><[e-mail]></tt> has written a Linux
filesystem kernel module that supports a limited subset of the Novell NCP
that will allow you to mount Novell volumes into your Linux filesystem
without requiring any additional products for your fileserver. The software
causes Linux to emulate a normal Novell workstation for file services. It also
includes a small print utility that allows you to print to Novell print
queues. Because this is a limited subset it does not work very well with any
of the Novell fileserver emulators described later in this document.

Volker has called the package <em>ncpfs</em> and derived the necessary
information mainly from the book "Netzwerkprogrammierung in C" by
Manfred Hill and Ralf Zessin (further details of the book are contained within
the README file in the <em>ncpfs</em> package). Volker also used the IPX tools
written by Greg Page <tt><[e-mail]></tt>.

<sect1><heading>Obtaining <bf><em>ncpfs</em></bf>.

<p>
The latest <em>ncpfs</em> package was designed to be built against the version
<tt>1.2.13</tt> kernel or kernels later than <tt>1.3.71</tt> so you if you are
not using a kernel in either of these categories it then you should obtain
one and use it.
<p>
You can obtain the <em>ncpfs</em> package by anonymous ftp from
Volker's home site at:
<url url="ftp://ftp.gwdg.de/pub/linux/misc/ncpfs/" name="ftp.gwdg.de">
or
<url url="ftp://sunsite.unc.edu/pub/Linux/system/Filesystems/ncpfs"
    name="sunsite.unc.edu">
or mirror sites. The current version at the time of writing was:
<tt>ncpfs-0.24.tgz</tt>

<sect1><heading>Building <bf><em>ncpfs</em></bf> for kernels 1.2.*.

<p>
<descrip>
<tag>Build a kernel with Ethernet and IPX support</tag>The first thing you
need to do is ensure that your kernel has been built with IPX support enabled.
In the <tt>1.2.13</tt> version kernel you need only ensure that you have
answered <tt>Y</tt> to the question: '<tt>The IPX protocol</tt>' as
illustrated:
<verb>
 ...
 ...
Assume subnets are local (CONFIG_INET_SNARL) [y]
Disable NAGLE algorithm (normally enabled) (CONFIG_TCP_NAGLE_OFF) [n]
The IPX protocol (CONFIG_IPX) [n] y
*
* SCSI support
 ...
 ...
</verb>
You will also need to ensure that you include an appropriate driver for your
Ethernet card. If you do not know how to do this then you should read the
<url url="http://sunsite.unc.edu/mdw/HOWTO/Ethernet-HOWTO.html"
name="Ethernet-HOWTO">.
<p>
You can then proceed to build your kernel. Make sure you remember to run
<em>lilo</em> to install it when you have finished.


<tag>Untar the <em>ncpfs</em> software</tag>
<verb>
# cd /usr/src
# tar xvfz ncpfs-0.24.tgz
# cd ncpfs
</verb>

<tag>Make the <em>ncpfs</em> software</tag>The software should compile cleanly
with no configuration necessary:
<verb>
# make
</verb>

<tag>Copy the IPX tools somewhere useful if you don't already have them.
</tag>After the <em>make</em> has completed you should find all of the tools
you need in the <tt>ncpfs/bin</tt> directory. You can use <tt>make install</tt>
to install the tools in Volkers choice of directories, but I prefer to place
them in different directories. I use the following commands to copy them
to my preferred directories:
<verb>
# cd bin
# cp ncp* pserver /usr/local/sbin
# cp nwmsg /sbin
# cp ../ipxdump/ipxdump ../ipxdump/ipxparse /usr/local/sbin
# cp ncopy nprint nsend nwfsinfo pqlist slist /usr/local/bin
</verb>
<bf>note:</bf> the <em>nwmsg</em> command must  be installed in the
<tt>/sbin</tt> because it is invoked by the <em>kerneld</em> daemon if you
are running it. If you want to change the location of it then you need to
edit the <tt>/usr/src/linux/include/linux/ncp_fs.h</tt> file and change
the <tt>NCP_MSG_COMMAND</tt> definition.
<p>
If you do not already have Greg Pages IPX tools installed then you should
also do the following:
<verb>
# cp ipx* /usr/local/sbin
</verb>

<tag>Copy the <em>ncpfs.o</em> module somewhere useful if necessary.</tag>
If you are compiling for a <tt>1.2.*</tt> kernel then you will find a file
called <tt>ncpfs.o</tt> in the <tt>ncpfs/bin</tt> directory after the
<em>make</em> has completed. This is the <em>ncpfs</em> kernel module.
You should copy this somewhere useful. On my <em>debian</em> system I have
copied it to the <tt>/lib/modules/1.2.13/fs</tt> directory and added
<tt>ncpfs</tt> to the <tt>/etc/modules</tt> file so that it will be
automatically started at boot time. If you are using some other distribution
you should find where it keeps its modules and copy it there, or just copy it
to your <tt>/etc</tt> directory. To load the modules manually you need to use
the command:
<verb>
# insmod ncpfs.o
</verb>

</descrip>

<sect1><heading>Building <bf><em>ncpfs</em></bf> for kernels 1.3.71++.

<p>
For the latest version of <em>ncpfs</em> you must use kernel <tt>1.3.71</tt>
or newer. If you want to use older <tt>1.3.*</tt> kernels then you must locate
a version of <em>ncpfs</em> like <em>ncpfs-0.12.tgz</em>.
<p>
If you intend using a kernel that is version <tt>1.3.71</tt> or newer then the
<em>ncpfs</em> kernel code has been included in the standard kernel
distribution. You need only answer <tt>Y</tt> to:
<tscreen><verb>
 ...
 ...
The IPX protocol ?
 ...
 ...
NCP filesystem support (to mount NetWare volumes) ?
 ...
 ...
</verb></tscreen>
<p>
You will still need to follow the instructions for building for kernels
<tt>1.2.*</tt> so that you can build the tools but there will not be a module
file for you to install.

<sect1><heading>Configuring and using <bf><em>ncpfs</em></bf>.

<p>

<descrip>
<tag>Configure the IPX network software</tag>There are two ways of configuring
the IPX network software. You can manually configure all of your IPX network
information or you can choose to let the software determine for itself some
reasonable settings. In most installations the automatic method will work
ok. If it doesn't work for you then read the 'IPX tools' section below to
configure your software manually:
<verb>
# ipx_configure --auto_interface=on --auto_primary=on
</verb>

<tag>Test the configuration</tag>After your IPX network is configured you
should be able to use the <em>slist</em> command to see a list of all of
the Novell fileserver on your network:
<verb>
# slist
</verb>
If the slist command displays a message like:
<tt>ncp_connect: Invalid argument</tt> then your kernel probably does not
support IPX. Check that you have actually booted off the appropriate kernel.
If the <em>slist</em> command does not list all of your fileservers then
you may need to use the manual network configuration method.

<tag>Mount a Novell(tm) volume.</tag>If your IPX network software is working
ok you should now be able to mount a Novell fileserver volume into your
Linux filesystem. The <em>ncpmount</em> command is used for this purpose
and requires that you specify at least the following information:
<enum>
<item>The fileserver name
<item>The fileserver login id. If it has a password you will also need that.
<item>The mount point ie. where you want the mount to go.
</enum>

An example command to mount fileserver <tt>ACCT_FS01</tt>, with a login id of
<tt>guest</tt> with no password, under the <tt>/mnt/Accounts</tt> directory
might look like the following:
<verb>
# ncpmount -S ACCT_FS01 /mnt/Accounts -U guest -n
</verb>
Note the use of the <tt>-n</tt> option to indicate that no password is
required for the login. The same login specifying a password of <tt>secret</tt>
would look like:
<verb>
# ncpmount -S ACCT_FS01 /mnt/Accounts -U guest -P secret
</verb>
If you don't specify either the <tt>-n</tt> or the <tt>-P</tt> options you
will be prompted for a password.

<tag>Check the mount</tag>If the mount is successful you will find the
volumes accessible to the userid used for login listed as directories under
the mount point. You should then also be able to traverse the directory
structure to find other files. Because NCP does not provide uid or gid
ownership of files, all of the files will have the permission and ownership
assigned to the mount point directory, keep this in mind when sharing
mounts between Linux users.

<tag>Test printing</tag>You can test to see if printing works using the
<em>nprint</em> command. The <em>nprint</em> command allows you to print
to a file to a NetWare print queue. The <em>pqlist</em> command allows you
the list the available print queues on a NetWare server. Both commands
require that you supply username and password so you might normally
consider building some shell scripts to make the task of printing easier.
An example might look like:
<tscreen><verb>
# pqlist -S ACCT_FS01 -U guest -n
# nprint -S ACCT_FS01 -q LASER -U guest -n filename.txt
</verb></tscreen>
The login syntax is similar to the <em>ncpmount</em> command. The examples
above assume that fileserver <tt>ACCT_FS01</tt> has a <tt>guest</tt> account
with no password, that a print queue called <tt>LASER</tt> exists and that
<tt>guest</tt> is allowed to print to it.

<tag>Configure mounts to be automatically performed.</tag>If you have some
need to permanently have an ncp mount then you will want to configure the
commands above into your <em>rc</em> files so that they occur automatically
at boot time. I recommend you place them in your <tt>/etc/rc.local</tt> file
if you have one. I have placed them in a file called <tt>NetWare</tt> in
my <tt>/etc/init.d</tt> directory and created appropriate symbolic links
into my <tt>/etc/rc2.d</tt> directory to cause it to be started. You might
use something like:
<tscreen><verb>
#
# Start the ncp filesystem
/sbin/insmod /lib/modules/1.2.13/fs/ncpfs.o

# configure the IPX network
ipx_configure --auto_interface=on --auto_primary=on

# guest login to the Accounting fileserver
ncpmount -S ACCT_FS01 /mnt/Accounts -U guest -n

#
</verb></tscreen>
There is another means of configuring NCP mounts and that is by building
a <tt>$HOME/.nwclient</tt> file. This file contains details of temporary
or user specific NCP mounts that would be performed regularly. It allows
you to store the details of mounts so that you can recreate them without
having to specify all of the detail each time.
<p>
Its format is quite straightforward:
<tscreen><verb>
# The first entry is the 'preferred server' entry and is
# used whenever you do not specify a server explicitly.
#
# User TERRY login to DOCS_FS01 fileserver with password 'password'
DOCS_FS01/TERRY password
#
# Guest login to the ACCT_FS01 fileserver with no password.
ACCT_FS01/GUEST -
</verb></tscreen>
To activate these mounts you could use:
<tscreen><verb>
$ ncpmount /home/terry/docs
</verb></tscreen>
to mount: DOCS_FS01 with a login of TERRY under the /home/terry/docs
directory. Note that this entry was chosen because no fileserver was
specified in the mount command. If the following command were used:
<tscreen><verb>
$ ncpmount -S ACCT_FS01 /home/terry/docs
</verb></tscreen>
then a GUEST login to ACCT_FS01 would be mounted there instead.
<p>
<bf>Note:</bf> for this mechanism to work the permissions of the
<tt>$HOME/.nwclient</tt> file must be <tt>0600</tt> so you would
need to use the command:
<tscreen><verb>
$ chmod 0600 $HOME/.nwclient
</verb></tscreen>
If non-root users are to be allowed to use this mechanism then
the <em>ncpmount</em> command must be Set Userid Root, so you
would need to give it permissions:
<tscreen><verb>
# chmod 4755 ncpmount
</verb></tscreen>

<tag>Try out the <em>nsend</em> utility</tag>a utility to send messages
to Novell users is also included in the package, it is called <em>nsend</em>
and is used as follows:
<verb>
# nsend rod hello there
</verb>
would send the message "hello there" to a logged in user "rod" on your
"primary" fileserver (the first one appearing in your <tt>.nwclient</tt>
file. You can specify another fileserver with the same syntax as for the
<em>ncpmount</em> command.

</descrip>

<sect><heading>Configuring your Linux machine as an NCP server.

<p>
There are two packages available that allow Linux to provide the functions of
a Novell Fileserver. They both allow you to share files on your linux
machine with users using Novell NetWare client software. Users can attach and
map filesystems to appear as local drives on their machines just as they would
to a real Novell fileserver. You may want to try both to see which best
serves your intended purpose.

<sect1><heading>The <bf><em>mars_nwe</em></bf> package.

<p>
Martin Stover <tt><[e-mail]></tt> developed <em>mars_nwe</em>
to enable linux to provide both file and print services for NetWare clients.
<p>
In case you are wondering about the name: <em>mars_nwe</em> is Martin Stovers
Netware Emulator.

<sect2><heading>Capability of <bf><em>mars_nwe</em></bf>.

<p>
<em>mars_nwe</em> implements a subset of the full Novell NCP for file services,
disk based bindery and also print services. It is likely to contain bugs but
there are many people using it now and the number of bugs is steadily
decreasing as new versions are released.

<sect2><heading>Obtaining <bf><em>mars_nwe</em></bf>.

<p>
You can obtain <em>mars_nwe</em> from
<url url="ftp://ftp.gwdg.de/pub/linux/misc/ncpfs/" name="ftp.gwdg.de">. The
current version at the time of writing was: <tt>mars_nwe-0.97.pl6.tgz</tt>.

<sect2><heading>Building the <em>mars_nwe</em> package.

<p>
<descrip>
<tag>Build a kernel with Ethernet and IPX Support</tag>In the <tt>1.2.13</tt>
version kernel you need only ensure that you have answered <tt>Y</tt> to the
question: '<tt>The IPX protocol</tt>' and <tt>N</tt> to the question:
'<tt>Full internal IPX network</tt>' as illustrated:
<verb>
 ...
 ...
The IPX protocol (CONFIG_IPX) [n] y
 ...
 ...
Full internal IPX network (CONFIG_IPX_INTERN) [N/y/?] n
 ...
 ...
</verb>
In newer kernels a similar process is adopted but the actual text of the prompt
may have changed slightly.
<p>
You will also need to ensure that you include an appropriate driver for your
Ethernet card. If you do not know how to do this then you should read the
<url url="http://sunsite.unc.edu/mdw/HOWTO/Ethernet-HOWTO.html"
name="Ethernet-HOWTO">.
<p>
You can then proceed to build your kernel. Make sure you remember to run
<em>lilo</em> to install it when you have finished.

<tag>Untar the <em>mars_nwe</em> package.</tag>
<tscreen><verb>
# cd /usr/src
# tar xvfz mars_nwe-0.97.pl6.tgz
</verb></tscreen>

<tag>Make <em>mars_nwe</em>.</tag>To make the package is very simple. First
you should look at and edit the <tt>config.h</tt> file if necessary. It
allows you to configure items such as the installation directories that will
be used and the maximum number of sessions and volumes that the server will
support. The really important entries to look at are:
<verb>
FILENAME_NW_INI       the location of the initialisation file
PATHNAME_PROGS        where the executable support programs will be found.
PATHNAME_BINDERY      where the 'bindery' files will go.
PATHNAME_PIDFILES     the directory for the 'pid' files to be written.
MAX_CONNECTIONS       the maximum number of simultaneous connections allowed.
MAX_NW_VOLS           the maximum number of volumes mars_nwe will support.
MAX_FILE_HANDLES_CONN the maximum number of open files per connection.
WITH_NAME_SPACE_CALLS if you want to support ncpfs clients.
INTERNAL_RIP_SAP      whether you want mars_nwe to provide rip/sap routing.
</verb>

When this is done:
<tscreen><verb>
# make
# make install
</verb></tscreen>

will build the servers and install them in the appropriate directory. The
installation script also installs the configuration file
<tt>/etc/nwserv.conf</tt>.

<tag>Configure the server.</tag>Configuration is fairly simple. You need to
edit the <tt>/etc/nwserv.conf</tt> file. The format of this file may at first
look a little cryptic, but it is fairly straightforward. The file contains a
number of single line configuration items. Each line is whitespace delimited
and begins with a number that indicates the contents of the line. All
characters following a '<tt>#</tt>' character are considered a comment and
ignored. Martin supplies an example configuration file in the package,
but I'll present what I consider to be a simplified example to offer an
alternative for you.

<tscreen><verb>
# VOLUMES (max. 5)
# Only the SYS volume is compulsory. The directory containing the SYS
# volume must contain the directories: LOGIN, PUBLIC, SYSTEM, MAIL.
# The 'k' option converts all filenames in NCP requests to lowercase.
# The 'm' option marks the volume as removable (useful for cdroms etc.)
# The 'r' option set the volume to read-only.
# The 'o' option indicates the volume is a single mounted filesystem.
# The 'P' option allows commands to be used as files.
# The default is upper case.
# Syntax:
#    1 <Volumename> <Volumepath>   <Options>

1   SYS         /home/netware/SYS/               # SYS
1   DATA        /home/netware/DATA/      k       # DATA
1   CDROM       /cdrom                   kmr     # CDROM

# SERVER NAME
# If not set then the linux hostname will be converted to upper case
# and used. This is optional, the hostname will be used if this is not
# configured.
# Syntax:
#    2 <Servername>

2   LINUX_FS01

# INTERNAL NETWORK ADDRESS
# The Internal IPX Network Address is a feature that simplifies IPX routing
# for multihomed hosts (hosts that have ports on more than one IPX network).
# Syntax:
#    3 <Internal Network Address> [<Node Number>]
# or:
#    3 auto
#
# If you use 'auto' then your host IP address will be used. NOTE: this may
# be dangerous, please be sure you pick a number unique to your network.

3   0x49a01010  1


# NETWORK DEVICE(S)
# This entry configures your IPX network. If you already have your
# IPX network configured then you do not need this. This is the same as
# using ipx_configure/ipx_interface before you start the server.
# Syntax:
#    4 <IPX Network Number> <device_name> <frametype> [<ticks>]
#                         Frame types: ethernet_ii, 802.2, 802.3, SNAP

4  0x39a01010  eth0  802.3  1

# SAVE IPX ROUTES AFTER SERVER IS DOWNED
# Syntax:
#    5 <flag>
#       0 = don't save routes, 1 = do save routes

5 0

# NETWARE VERSION
# Syntax:
#    6 <version>
#      0 = 2.15, 1 = 3.11

6 1

# PASSWORD HANDLING
# Real Novell DOS clients support a feature which encypts your
# password when changing it. You can select whether you want your
# mars server to support this feature or not.
# Syntax
#    7 <flag>
#    <flag> is:
#        0 to force password encryption. (Clients can't change password)
#        1 force password encryption, allow unencrypted password change.
#        7 allow non-encrypted password but no empty passwords.
#        8 allow non-encrypted password including empty passwords.
#        9 completely unencrypted passwords (doesn't work with OS/2)

7 1

# MINIMAL GID UID rights
# permissions used for attachments with no login. These permissions
# will be used for the files in your primary server attachment.
# Syntax:
#    10 <gid>
#    11 <uid>
#    <gid> <uid> are from /etc/passwd, /etc/groups

10  200
11  201

# SUPERVISOR password
# May be removed after the server is started once. The server will
# encrypt this information into the bindery file after it is run.
# You should avoid using the 'root' user and instead use another
# account to administer the mars fileserver.
#
# This entry is read and encrypted into the server bindery files, so
# it only needs to exist the first time you start the server to ensure
# that the password isn't stolen.
#
# Syntax:
#    12 <Supervisor-Login> <Unix username> [<password>]

12  SUPERVISOR  terry  secret

# USER ACCOUNTS
# This associates NetWare logins with unix accounts. Password are
# optional.
# Syntax:
     13 <User Login> <Unix Username> [<password>]

13  TERRY  terry
13  MAR    mar

# LAZY SYSTEM ADMIN CONFIGURATION
# If you have a large numbers of users and could not be bothered using
# type 13 individual user mappings, you can automatically map mars_nwe
# logins to linux user names. BUT, there is currently no means of making
# use of the linux login password so all users configured this way are
# will use the single password supplied here. My recommendation is not
# to do this unless security is absolutely no concern to you.
# Syntax:
#    15 <flag> <common-password>
#    <flag> is: 0  - don't automatically map users.
#               1  - do automatically map users not configured above.
#               99 - automatically map every user in this way.

15  0  duzzenmatta

# SANITY CHECKING
# mars_nwe will automatically ensure that certain directories exist if
# you set this flag.
# Syntax:
#    16 <flag>
#    <flag> is 0 for no, don't, or 1 for yes, do.

16  0

# PRINT QUEUES
# This associates NetWare printers with unix printers. The queue
# directories must be created manually before printing is attempted.
# The queue directories are NOT lpd queues.
# Syntax:
#    21 <queue_name> <queue_directory> <unix_print_cmd>

21  EPSON  SYS:/PRINT/EPSON lpr -h
21  LASER  SYS:/PRINT/LASER lpr -Plaser

# DEBUG FLAGS
# Syntax:
#    <debug_item> <debug_flag>
#
#    100 = IPX KERNEL
#    101 = NWSERV
#    102 = NCPSERV
#    103 = NWCONN
#    104 = start NWCLIENT
#    105 = NWBIND
#    106 = NWROUTED
#                 0 = disable debug, 1 = enable debug

100 0
101 0
102 0
103 0
104 0
105 0
106 0

# RUN NWSERV IN BACKGROUND AND USE LOGFILE
# Syntax:
#    200 <flag>
#        0 = run NWSERV in foreground and don't use logfile
#        1 = run NWSERV in background and use logfile

200  1

# LOGFILE NAME
# Syntax:
#    201 <logfile>

201  /tmp/nw.log

# APPEND LOG OR OVERWRITE
# Syntax:
#    202 <flag>
#        0 = append to existing logfile
#        1 = overwrite existing logfile

202  1

# SERVER DOWN TIME
# This item sets the time after a SERVER DOWN is issued that the
# server really goes down.
# Syntax:
#    210 <time>
#        in seconds. (defaults 10)

210  10

# ROUTING BROADCAST INTERVAL
# The time is seconds between server broadcasts
# Syntax:
#    211 <time>
#        in seconds. (defaults 60)

211  60

# ROUTING LOGGING INTERVAL
# Set how many broadcasts take place before logging of routing
# information occurs.
# Syntax:
#    300  <number>

300  5

# ROUTING LOGFILE
# Set the name of the routing logfile
# Syntax:
#    301 <filename>

301  /tmp/nw.routes

# ROUTING APPEND/OVERWRITE
# Set whether you want to append to an existing log file or
# overwrite it.
# Syntax:
#    302 <flag>
#        <flag> is 0 for append, 1 for create/overwrite

302  1

# WATCHDOG TIMING
# Set the timing for watchdog messages that ensure the network is
# still alive.
# Syntax:
#    310 <value>
#        <value> =   0 - always send watchdogs
#                  < 0 - (-ve) for disable watchdogs
#                  > 0 - send watchdogs when network traffic
#                        drops below 'n' ticks

310  7

# STATION FILE
# Set the filename for the stations file which determine which
# machines this fileserver will act as the primary fileserver for.
# The syntax of this file is described in the 'examples' directory
# of the source code.
# Syntax:
#    400 <filename>

400  /etc/nwserv.stations

# GET NEAREST FILESERVER HANDLING
# Set how SAP Get Nearest Fileserver Requests are handled.
# Syntax:
#    401 <flag>
#        <flag> is: 0 - disable 'Get Nearest Fileserver' requests.
#                   1 - The 'stations' file lists stations to be excluded.
#                   2 - The 'stations' file lists stations to be included.

401  2
</verb></tscreen>

<tag>Start the server</tag>If you've configured the server to expect external
programs to configure your network and/or provide the routing function then
you should start those before starting the server. Presuming you have
configured the server so that itwill configure your interfaces for you and
provide the routing services you need only issue the command:
<tscreen><verb>
# nwserv
</verb></tscreen>

<tag>Test the server</tag>To test the server you should first try to attach
and login from a NetWare client on your network. You then set a
<tt>CAPTURE</tt> from the client and attempt a print. If both of these are
successful then the server is working.
</descrip>

<sect1><heading>The <bf><em>lwared</em></bf> package.

<p>
Ales Dryak <tt><[e-mail]></tt> developed <em>lwared</em> to allow
Linux to function as an NCP based fileserver.
<p>
Ales has called the package <em>lwared</em>, an abbreviation for <em>LinWare
Daemon</em>.

<sect2><heading>Capability of <bf><em>lwared</em></bf>.

<p>
The <em>lwared</em> server is capable of providing a subset of the full
function of the Novell NCP. It incorporates messaging but it does not provide
any printing facilities at all. It does not currently work very well with
either Windows95 or Windows NT clients. The <em>lwared</em> server relies
on external programs to build and update the IPX routing and SAP tables.
Misbehaving clients can cause the server to crash. Importantly, filename
translation facilities have not been included.
<p>
The server does work for NETX and VLM NetWare shells.

<sect2><heading>Obtaining <bf><em>lwared</em></bf>

<p>
The <em>lwared</em> package can be built for any kernel newer than
<tt>1.2.0</tt>, I recommend you use version <tt>1.2.13</tt> as no kernel
patches are required if you do. Some of the IPX functionality has changed
with the version <tt>1.3.*</tt> kernels and this means that patches are
now required to make it work properly. Appropriate patches are included for
the new kernels, so if you must use an alpha kernel you should still be able
to get <em>lwared</em> to work properly for you.
<p>
You can obtain the <em>lwared</em> package by anonymous ftp from:
<url url="ftp://klokan.sh.cvut.cz/pub/linux/linware/"
    name="klokan.sh.cvut.cz">

or from:

<url url="ftp://sunsite.unc.edu/pub/Linux/system/Network/daemons"
    name="sunsite.unc.edu">
or mirror sites. The current version at the time of writing was:
<tt>lwared-0.95.tar.gz</tt>

<sect2><heading>Building <bf><em>lwared</em></bf>

<p>
<descrip>
<tag>Untar the <em>lwared</em>package</tag>Something like:
<verb>
# cd /usr/src
# tar xvpfz lwared-0.95.tar.gz
</verb>

<tag>Build a kernel with Ethernet and IPX support</tag>
If you are using an alpha <tt>1.3.*</tt> kernel then you should try and use
kernel version <tt>1.3.17</tt> or newer because the supplied patches were built
against it. <tt>1.3.*</tt> kernels older than <tt>1.3.17</tt> will require
hand patching to install. (<em>some information on how to do this is included
in the <tt>INSTALL</tt> file in the package.</em>). To install the patches
against a <tt>1.3.17</tt> kernel or newer you should try:
<verb>
# make patch
</verb>
<p>
After applying the patches if necessary, the next thing you need to do is
ensure that your kernel has been built with IPX support enabled. In the
<tt>1.2.13</tt> version kernel you need only ensure that you have answered
<tt>Y</tt> to the question: '<tt>The IPX protocol</tt>' as
illustrated:
<verb>
 ...
 ...
Assume subnets are local (CONFIG_INET_SNARL) [y]
Disable NAGLE algorithm (normally enabled) (CONFIG_TCP_NAGLE_OFF) [n]
The IPX protocol (CONFIG_IPX) [n] y
*
* SCSI support
 ...
 ...
</verb>
In newer kernels a similar process is adopted by the actual text of the prompt
may have changed slightly.
<p>
You will also need to ensure that you include an appropriate driver for your
Ethernet card. If you do not know how to do this then you should read the
<url url="http://sunsite.unc.edu/mdw/HOWTO/Ethernet-HOWTO.html"
name="Ethernet-HOWTO">.
<p>
You can then proceed to build your kernel. Make sure you remember to run
<em>lilo</em> to install it when you have finished.

<tag>Compile and install <em>lwared</em>.</tag>To compile <em>lwared</em> you
should first check, edit if necessary, the <tt>server/config.h</tt> file. This
file contains various settings that will govern the way your server will
behave when it is running. The defaults are reasonable, though you might want
to check that the directories specified for the log files and configuration
files suit your system.
<verb>
# make depend
# make
# make install
</verb>
I found that the '<tt>make depend</tt>' complained about not finding the
<tt>float.h</tt> file on my system but appeared to work anyway.
I also found that when I tried compiling with gcc <tt>2.6.3</tt> I found
I had to change the line:
<verb>
#include <net/route.h>
</verb>
to
<verb>
#include <net/if_route.h>
</verb>
in <tt>lib/ipxkern.c</tt> as this file changed name sometime.
<p>
The '<tt>make install</tt>' will attempt to install the server and routing
daemon programs into your <tt>/usr/sbin</tt> directory, the <em>lwpasswd</em>
program into your <tt>/usr/bin</tt> directory, the IPX utility programs will
be installed into your <tt>/sbin</tt> directory and last but not least the
manual pages will go into the <tt>/usr/man</tt> directory structure. If any
of these locations are not suitable for your system then you should edit
the relevant <tt>Makefile</tt> and change the target directories to suit.

</descrip>

<sect2><heading>Configuring and using <bf><em>lwared</em></bf>

<p>
Now the fun bit!

<descrip>
<tag>Configuring the IPX network</tag>The first thing you must do is configure
your Ethernet interfaces to support the IPX networks your server will support.
To do this you will need to know the IPX network addresses for each of your
LAN segments, which Ethernet device (<tt>eth0</tt>, <tt>eth1</tt> etc.) is
on which segment, what frame type (<tt>802.3</tt>, <tt>EtherII</tt> etc.) each
LAN segment uses and what Internal Network address your server should use
(this is really needed if your server will service more than one LAN segment).
A configuration for a server that is on two dis-similar segments with IPX
network addresses <tt>23a91300</tt> and <tt>23a91301</tt> and internal network
address <tt>bdefaced</tt> might look like:
<verb>
# ipx_internal_net add BDEFACED 1
# ipx_interface add eth0 802.3 23a91300
# ipx_interface add eth1 etherii 23a91301
</verb>

<tag>Start the routing daemons</tag>The kernel software itself actually does
the IPX packet forwarding as it does for IP, but the kernel requires additional
programs to manage the routing table updates. In the case of IPX two daemons
are needed and both are supplied with <em>lwared</em>: <em>ipxripd</em> manages
the IPX routing information and <em>ipxsapd</em> manages the SAP information.
To start the daemons you need only specify the location of where they should
write their log messages:
<verb>
# ipxripd /var/adm/ipxrip
# ipxsapd /var/adm/ipxsap
</verb>

<tag>Configure the <em>lwared</em> server</tag>There are two files that you
must manually configure to allow user login to your <em>lwared</em> server.
They are:

<p>
<descrip>
<tag><tt>/etc/lwpasswd</tt></tag>This is where LinWare user account
information is kept. The <em>lwpasswd</em> program is to keep it up to date.
In its simplest form the <tt>/etc/lwpasswd</tt> file looks like:
<verb>
ales:
terryd:
guest:
</verb>
Its format is a simple list of login id followed by a ':' character and then
the encrypted version of the login passwd. A couple of important caveats here:
No encrypted password means no password, LinWare users must have Linux
accounts, that is any user you place in <tt>/etc/lwpasswd</tt> must also
appear in <tt>/etc/passwd</tt> and <tt>root</tt> is the only account that can
change the password of another LinWare user. If you are logged in as
<tt>root</tt> you can change the password of a LinWare user as this transcript
demonstrates:
<verb>
# lwpasswd rodg
Changing password for RODG
Enter new password:
Re-type new password:
Password changed.
</verb>
<tag><tt>/etc/lwvtab</tt></tag>This is the LinWare volume tables and it stores
information about what directories should be made available to LinWare users
(this file is similar in nature to the NFS <tt>/etc/exports</tt> file). A
simple example of its format is as follows:
<verb>
SYS             /lwfs/sys
DATA            /lwfs/data
HOME            /home
</verb>
The format is simple: Volume name followed by whitespace followed by Linux
directory to export. You must have at <bf>least</bf> an entry for the
<tt>SYS</tt> volume for the server to start. If you intend your DOS based
users to be able use your LinWare server as their primary server then you must
install a standard <tt>SYS</tt> volume directory structure underneath the
directory you export as your <tt>SYS</tt> volume. Since these files are
proprietary and copyright to the Novell corporation you should have a license
for these. If you users will be using a Novell fileserver as their primary
server then this will not be necessary.
</descrip>

<tag>Start the <em>lwared</em> server.</tag>tada!
<verb>
# lwared
</verb>
It is almost an anticlimax isn't it ? Ok so you've got a question, right?
What is the fileserver name that is being advertised ? If you started the
server as shown then the LinWare server name being advertised will be
based on what is returned by the Linux <em>hostname</em>. If you'd like it
to be something else then you can give the server the name when you start
it, for example:
<verb>
# lwared -nlinux00
</verb>
would start the server with the name <tt>linux00</tt>.

<tag>Test the <em>lwared</em> server.</tag>The very first thing to test is
that your LinWare server appears in an <em>slist</em> from a DOS client
on your network. The <em>slist</em> program is stored on the <tt>SYS</tt>
volume of a Novell fileserver so you must do this from a machine that is
already logged in somewhere. If this is not successful then check that
<em>ipxsapd</em> and <em>lwared</em> are both running. If the <em>slist</em>
is successful then you should try attaching to the server and mapping
a volume:
<verb>
C:> attach linux00/ales
 ...
 ...
C:> map l:=linux00/data:
C:> l:
</verb>
You should then be able to treat the new map just like any other map. The
file permissions you will have will be based on those allowed to the
<em>linux</em> account that parallels your LinWare login.

</descrip>

<sect><heading>Configuring PPP for IPX support.

<p>
New versions of the <em>pppd</em> PPP daemon for Linux have support that
allows you to carry IPX packets across a PPP serial link. You need at least
version <tt>ppp-2.2.0d</tt> of the daemon. See the
<url url="http://sunsite.unc.edu/mdw/HOWTO/PPP-HOWTO.html" name="PPP-HOWTO">
for details on where to find it. When you compile <em>pppd</em> you must
ensure you enable the IPX support by adding the following two lines:

<tscreen><verb>
IPX_CHANGE = 1
USE_MS_DNS = 1
</verb></tscreen>

to: <tt>/usr/src/linux/pppd-2.2.0f/pppd/Makefile.linux</tt>.
<p>
The <tt>IPX_CHANGE</tt> is what configures the IPX support into PPP.
The <tt>USE_MS_DNS</tt> define allows Microsoft Windows95 machines to do
Name Lookups.
<p>
The real trick to getting it to work in knowing how to configure it.
<p>
There are many ways of doing this, but I'm only going to describe the two that
I've received any information on. I've tried neither yet, so consider this
section experimental, and if you get something to work, please let me know.

<sect1><heading>Configuring an IPX/PPP server.

<p>
The first thing you need to do is configure your Linux machine as an IP/PPP
server. When you have this done there are a couple of simple  modifications
you need to make to get IPX working over the same configuration.

<sect2><heading>First steps.

<p>
Unless all you want to do is connect to your linux machine supporting the
ppp dialin, and even then perhaps, you must configure your linux machine
as an IPX router as described earlier. You won't need to use the
<em>ipx_route</em> command because <em>pppd</em> will configure these for you
as it does for IP. When you have the <em>ipxd</em> daemon running it will
automatically detect any new IPX interfaces and propogates routes for them.
This is what you will need so that other machines will know about the client
machines that connect.
<sect2><heading>Design.

<p>
When you are running as a server it will normally be your responsibility
to assign network address to each of the PPP links when they are established.
This is an important point, each PPP link will be an IPX network and will have
a unique IPX network address. This means that you must decide how you will
allocate addresses and what what they will be. A simple convention is to
allocate one IPX network address to each serial device that will support
IPX/PPP. You could allocate IPX network addresses based on the login id
of the connecting user, but I don't see any particularly good reason to do
so.
<p>
I will assume that this is what you have done, and that there are two serial
devices (modems) that we will use. The addresses I've assigned in this
contrived example are:
<tscreen><verb>
device IPX Network Address
------ -------------------
ttyS0  0xabcdef00
ttyS1  0xabcdef01
</verb></tscreen>

<sect2><heading>Configure <em>pppd</em>.
<p>
Configure your <tt>/etc/ppp/options.ttyS0</tt> file as follows:

<tscreen><verb>
ipx-network abcdef00
ipx-node 2:0
ipxcp-accept-remote
</verb></tscreen>

and your <tt>/etc/ppp/options.ttyS1</tt> file as:

<tscreen><verb>
ipx-network abcdef01
ipx-node 3:0
ipxcp-accept-remote
</verb></tscreen>

These will ask <em>pppd</em> to allocate the appropriate IPX network addresses
to the link when the link is established, set the local node number to
<tt>2</tt> or <tt>3</tt> and will let the remote node overwrite what the
remote node number with what it thinks it is. Note that each of the addresses
are hexadecimal numbers but that no <tt>0x</tt> is required.
<p>
There are other places this information could be configured. If you have only
one dialin modem then an entry could go into the <tt>/etc/ppp/options</tt>
file. Alternatively this information can be passed on the command line to
<em>pppd</em>.

<sect2><heading>Test the server configuration.

<p>
To test the configuration you will need to have a client configuration that
is known to work. When the caller dials in, logs in and <em>pppd</em> starts
it will assign the network address, advise the client of the servers node
number and negotiate the clients node number. When this has completed, and
after <em>ipxd</em> has detected the new interface the client should be able
to establish IPX connections to remote hosts.

<sect1><heading>Configuring an IPX/PPP client.

<p>
In a client configuration, whether or not you configure your Linux machine
as an IPX router depends on whether you have a local LAN that you wish to
act as an IPX router for. If you are a standalone machine connecting to an
IPX/PPP dialin server then you won't need to run <em>ipxd</em>, but if you
have a LAN and wish all of the machines on the LAN to make use of the
IPX/PPP route then you must configure and run <em>ipxd</em> as described.
This configuration is much simpler because you do not have multiple serial
devices to configure.

<sect2><heading>Configuring <em>pppd</em>

<p>
The simplest configuration is one that allows the server to supply all of
the IPX network configuration information. This configuration would be
compatible with the server configuration described above.
<p>
Again you need to add options to your <tt>/etc/ppp/options</tt> file,
they are:
<tscreen><verb>
ipxcp-accept-network
ipxcp-accept-remote
ipxcp-accept-local
</verb></tscreen>

These options tell <em>pppd</em> to act completely passively and accept
all of the configuration details from the server. You could supply default
values here for servers that don't supply details by adding
<tt>ipx-network</tt> and <tt>ipx-node</tt> entries similar to the server
configuration.

<sect2><heading>Testing the IPX/PPP client.

<p>
To test the client you will need a known working server to dial into. After
you have dialled in and pppd has run you should see the IPX details configured
on your <tt>ppp0</tt> device when you run the <em>ifconfig</em> command and
you should be able to use <em>ncpmount</em>.
<p>
I'm not sure whether you will have to manually add IPX routes so that you
can reach distant fileserver or not. This seems likely. If anyone running
this configuration could tell me I'd be grateful.

<sect><heading>IPX tunnel over IP

<p>
Many of you will be in a situation where you have two Novell Local Area Netorks
with only an IP connection between them. How do you play multiplayer deathmatch
DOOM for DOS via this arrangement you might ask ? Andreas Godzina
<tt><[e-mail]></tt> has an answer for you in the form of
<em>ipxtunnel</em>.
<p>
<em>ipxtunnel</em> provides a bridge-like facility for IPX by allowing
IPX packets to be encapsulated with tcp/ip datagrams so that they can
be carried by a tcp/ip connection. It listens for IPX packets and when it
hears one it wraps it within a tcp/ip datagram and routes it to a remote
IP address that you specify. For this to work of course the machine that
you route the encapsulated IPX must also be running a copy of the same
version of <em>ipxtunnel</em> as you.

<sect1><heading>Obtaining <em>ipxtunnel</em>

<p>
You can obtain <em>ipxtunnel</em> from
<url url="ftp://sunsite.unc.edu/pub/Linux/system/Network/daemons"
        name="sunsite.unc.edu">
or mirror sites.

<sect1><heading>Building <em>ipxtunnel</em>

<p>
<em>ipxtunnel</em> built cleanly for me using the following commands:
<tscreen><verb>
# cd /usr/src
# tar xvfz .../ipxtunnel.tgz
# cd ipxtunnel
# make
</verb></tscreen>

<sect1><heading>Configuring <em>ipxtunnel</em>

<p>
Configuration for <em>ipxtunnel</em> is easy. Lets say that your friends
machine is <tt>gau.somewhere.com</tt> and your machine is called
<tt>gim.sw.edu</tt>. <em>ipxtunnel</em> uses a configuration file called
<tt>/etc/ipxtunnel.conf</tt>. This file allows you to specify the default UDP
port to use for the tcp/ip connection, where to send the encapsulated data
and which of your local interfaces <em>ipxtunnel</em> should listen on
and deliver IPX packets to.
<p>
A simple configuration file would look like the following:
<tscreen><verb>
#
# /etc/ipxtunnel.conf for gim.sw.edu
#
# The UDP port to use:                       (default 7666)
port 7777
#
# The remote machine to send IPX packets to: (no default)
remote gau.somewhere.com
#
# The local interfaces to listen for IPX on: (default eth0)
interface eth0
interface eth1
</verb></tscreen>
Obviously the other machine would have a similar configuration file specifying
this machine as a <tt>remote</tt> host.

<sect1><heading>Testing and using <em>ipxtunnel</em>

<p>
<em>ipxtunnel</em> acts <bf>like</bf> an IPX bridge, so the IPX networks
at either end of the link should probably be the same. Andreas has never
tested the <em>ipxtunnel</em> in an environment that actually supports
Novell file servers so if you do try this in a real environment let Andreas
know if it works or not.
<p>
If the <em>ipxtunnel</em> is working you should be able to start your
DOOM machines up at each end of the link running IPX mode and they should
see each other.
<p>
Andreas has only used this code over good high speed lines and he makes no
claim as to its performance when your link is low speed. Again, let him
know what works for you and what doesn't.


<sect><heading>Commercial IPX support for Linux.

<p>
Caldera Inc., produce a Linux distribution that features a range of
commercially supported enhancements including fully functional Novell NetWare
client support. The base distribution is the well respected Red Hat Linux
Distribution and Caldera have added their "Network Desktop" products to this.
The NetWare support provides a fully featured Novell NetWare client built on
technology licensed from Novell Corporation. The client provides full client
access to Novell 3.x and 4.x fileservers and includes features such as NetWare
Directory Service (NDS) and RSA encryption.
<p>
You can obtain much more information and ordering details from the:
<url url="http://www.caldera.com/" name="Caldera Inc Web Server">.
<p>
If you work within a Netware 4.x and/or NDS environment then the Caldera
Netware Client is the only solution available.
<p>
If you have a business critical application for Novell support for Linux
then the Caldera product should be something you take a close look at.

<sect><heading>Some Frequently Asked Questions

<p>
<descrip>
<tag>Where can I find commercially supported IPX software for Linux ?
</tag>The Caldera Corporation offers a fully licensed and fully supported
Netware 3.x and 4.x client. You can obtain information about it from the
<url url="http://www.caldera.com/" name="Caldera Inc Web Server">.

<tag>Does the IPX software work with Arcnet/Token Ring/etc. ?
</tag>The Linux IPX software does work with ArcNet and Token Ring interfaces.
I haven't heard of anyone trying it with AX.25 yet. Configuration is the same
as for configuring for ethernet except you will have to substitute appropriate
device names in place of 'eth0' and appopriate hardware addresses where
necessary.

<tag>How do I configure more than one IPX interface ?
</tag>If you have more than one interface in your machine you should use the
<em>ipx_interface</em> command to manually configure each one, you should not
use the `plug n play' configuration.

<tag>Why do I get the message `invalid argument' when I configure IPX ?
</tag>You are probably not running a kernel that supports IPX, either recompile
your kernel so it does, or double check that you have actually used lilo to
install and run the new kernel.

<tag>Why do I get the message `package not installed' when I configure IPX ?
</tag>You are probably not running a kernel that supports IPX, either recompile
your kernel so it does, or double check that you have actually used lilo to
install and run the new kernel.

<tag>How do I NFS export a mounted NCP filesystem ?
</tag>To use NFS to export an NCP filesystem you must mount it using the
<em>ncpmount</em> <tt>-V</tt> option. This option allows you to mount only
one volume of a fileserver instead of the usual mounting of all of them.
When you do this your NFS daemon will allow you to export that filesystem in
the usual way.

<tag>Does ncpfs package work with mars_nwe ?
</tag>Martin and Volker's code is slowly beginning to converge. Recent versions
of <em>mars_nwe</em> have an option to enable it to work with <em>ncpfs</em>.
You must enable the <tt>WITH_NAME_SPACE_CALLS</tt> in the <em>mars_nwe</em>
<tt>config.h</tt> file.

<tag>Is there any free DOS software to work with mars_nwe ?
</tag>A contrived question deserves a contrived answer. I'm glad you asked,
Martin has a package that he distributes alongside his <em>mars_nwe</em>
package that offers free DOS client support for the <em>mars_nwe</em> server.
You can find it at the same sites as the server, and it will be called
<tt>mars_dosutils-0.01.tgz</tt>. It includes C source code for programs such
as <em>slist.exe</em>, <em>login.exe</em>, <em>map.exe</em> etc. The source
is compilable with Borland(tm) C.

</descrip>

<sect><heading>Copyright Message.

<p>
The IPX-HOWTO, a guide to software supporting the IPX protocol for Linux.
Copyright (c) 1995 Terry Dawson.

This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; either version 2 of the License, or (at your option) any later
version.

This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with
this program; if not, write to the:

Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.

<sect><heading>Miscellaneous and Acknowledgements.

<p>
David E. Storey <tt><[e-mail]></tt> and
Volker Lendecke <tt><[e-mail]></tt>
both assisted greatly by supplying me with information for this document.
Gilbert Callaghan <tt><[e-mail]></tt>,
David Higgins <tt><[e-mail]></tt> and
Chad Robinson <tt><[e-mail]></tt> each contributed
information on configuring IPX/PPP.
<p>
"NetWare" is a registered trademark of the
<url url="http://www.novell.com/" name="Novell Corporation">.
"Caldera" is a registered trademark of the
<url url="http://www.caldera.com/" name="Caldera Corporation">.


<p>regards
<tt>Terry Dawson, vk2ktj.</tt>
<p>
<tt><[e-mail]></tt>

</article>