CD Actual 13/CDA13.ISO/DOC/HOWTO/OTHER_FO/SGML/ETHERNET.GZ/Ethernet-HOWTO.sgml

SGML Document | 1996-02-11 | 174.9 KB | 4,812 lines

<!--

    The SGML source of the Ethernet-Howto
    =====================================

    Maintained by Paul Gortmaker.

    Modification Date: Nov 18th, 1995

    (Don't forget to update the reference to the current kernel version
    in the Introduction section, and the date/version in the abstract!)

    The "==" are because two dashes together inside an SGML comment
    will crash it.  :-(

    History:

    1) About July '93 == I was looking for info for e'net support for
       a notebook, and was browsing the files on Don's ftp site
       (was ftp.super.org then) == Don had a bunch of informational
       files scattered about, and while talking to him, he mentioned
       that someday it would be good if all that info was rolled
       into one document. The Howto project was just started, with
       the NET-Howto being the first howto, (grown from all the
       confusion of the big net-switch at 0.99pl10). I rolled
       all Don's info into one, added a bunch of info that I had
       gleaned from postings, copied the general style of the
       NET-Howto, and the first Ethernet-Howto hit the internet
       sometime in August 93.

    2) About December 93, Don hacked together a HTML version of the
       Ethernet-Howto, but it didn't get much distribution, and
       I only kept the ASCII version up to date.

    3) Spring 94, and the Ethernet-Howto gets published in the
       Linux Bible  == 30 pages of a ~750 page book, published
       by Linux System Labs.

    4) Don moves jobs, and sets up a WWW home page with up-to-date
       information on new linux drivers, and other ethernet news.
       Look for the url contained in this document.

    5) A year (July 94) after its beginning, and the ugly tab+space
       formatting is dead.  The last ASCII version that was maintained
       by hand was called version 1.03, and dated June 22/94. Any ASCII
       versions newer than that were generated from this SGML source.

    6) I sat on the converted doc until Oct 94 because of a bug in the
       cross-referencing with HTML. A fix was to be forthcoming, but
       didn't eventuate. I eventually found a work-around.

    7) Nov 95 sees a major reorganization, which puts the FAQ section
       and all the interesting bits first. People never read more
       than the first few pages anyway... :-)

TODO:

    Add more URLs to this and that?


-->

<!doctype linuxdoc system>

<article>

<title>Linux Ethernet-Howto
<author>Paul Gortmaker, Editor.
<date>v2.5, 18 November 1995

<abstract>
    This is the Ethernet-Howto, which is a compilation of information
    about which ethernet devices can be used for Linux, and how to
    set them up. It hopefully answers all the frequently asked questions
    about using ethernet cards with Linux. Note that this Howto is focused
    on the hardware and low level driver aspect of the ethernet cards,
    and does not cover the software end of things. See the NET2-Howto
    for that stuff.
</abstract>

<toc>

<sect>Introduction<label id="main-intro">
<p>

    The Ethernet-Howto covers what cards you should and
    shouldn't buy; how to set
    them up, how to run more than one, and other common problems and
    questions. It contains detailed information on the current level
    of support for <em/all of the most common ethernet cards available./
    It does <em/not/ cover the software end of things, as that
    is covered in the NET-2 Howto. Also note that general non-Linux
    specific questions about Ethernet are not (or at least they should
    not be) answered here. For those types of questions, see the
    excellent amount of information in the <em/comp.dcom.lans.ethernet/
    FAQ. You can FTP it from <tt/dorm.rutgers.edu/ in the directory
    <tt>/pub/novell/info_and_docs/</tt>

    This present revision covers distribution kernels up to and
    including v1.2.13. Information pertaining to development
    kernels up to version 1.3.42 is also documented.

    The Ethernet-Howto is edited and maintained by:
<quote>
    Paul Gortmaker, <tt/[e-mail]/
</quote>

    The primary source of the information for the initial
    Ethernet-Howto was from:
<quote>
    Donald J. Becker, <tt/[e-mail]/
</quote>

    who we have to thank for writing the vast majority of ethernet
    card drivers that are presently available for Linux. He also
    is the original author of the NFS server too. Thanks Donald! We
    owe ya one!  :-)

    Net-surfers may wish to check out the following URL:

    <url url="http://cesdis.gsfc.nasa.gov/pub/people/becker/whoiam.html"
        name="Donald Becker">

<sect1>New Versions of this Document<label id="new-doc">
<p>

    New versions of this document can be retrieved via anonymous
    FTP from:

    <url url="ftp://sunsite.unc.edu/pub/Linux/docs/HOWTO/"
        name="Sunsite HOWTO Archive">

    and various Linux ftp mirror sites. Updates will be made
    as new information and/or drivers becomes available. If this copy
    that you are reading is more than 4 months old, it is either out of
    date, or it means that I have been lazy and haven't updated it.
    This document was produced by using the SGML system that was
    specifically set up for the Linux Howto project, and there are
    various output formats available, including, postscript, dvi,
    ascii, html, and soon TeXinfo.

    I would recommend viewing it in the html (via a WWW browser) or the
    Postscript/dvi format. Both of these contain cross-references
    that are lost in the ascii translation.

    If you want to get the official copy off sunsite, here is URL.

    <url url="http://sunsite.unc.edu/mdw/HOWTO/Ethernet-HOWTO.html"
        name="Ethernet-HOWTO">

    If minor additions and changes have been made, you can view the
    latest working copy from this URL.

    <url url="http://rsphy1.anu.edu.au/˜gpg109/Ethernet-HOWTO.html"
        name="Working Copy">

<sect1>Using the Ethernet-Howto<label id="using">
<p>

    As this guide is getting bigger and bigger, you probably don't want
    to spend the rest of your afternoon reading the whole thing. And you
    don't <em/have/ to read it all. If you haven't got an ethernet card, then
    you will want to start with
    <ref id="what-card" name="What card should I buy...">
    to see what you should buy,
    and what you should avoid. If you have already got an ethernet card,
    but are not sure if you can use it with Linux, then you will want to
    read <ref id="card-intro" name="Vendor Specific...">
    which contains specific information on each
    manufacturer, and their cards. If you are having trouble with your
    card, then you will want to read the specific information about
    your card mentioned above, and the troubleshooting information in
    <ref id="faq" name="the FAQ section">.
    If you are interested in some of the technical aspects
    of the device drivers, then you can find that information in
    <ref id="tech-intro" name="Technical Information">

<sect1>Disclaimer and Copyright<label id="copyright">
<p>

    This document is <em/not/ gospel. However, it is probably the most
    up to date info that you will be able to find. Nobody is responsible
    for what happens to your hardware but yourself. If your ethercard
    or any other hardware goes up in smoke (...nearly impossible!)
    we take no responsibility. ie. THE AUTHORS ARE NOT RESPONSIBLE
    FOR ANY DAMAGES INCURRED DUE TO ACTIONS TAKEN BASED ON THE
    INFORMATION INCLUDED IN THIS DOCUMENT.

    This document is Copyright (c) 1993, 1994, 1995 by Donald Becker and
    Paul Gortmaker. Permission is granted to make and distribute
    verbatim copies of this manual provided the copyright notice
    and this permission notice are preserved on all copies.

    Permission is granted to copy and distribute modified versions
    of this document under the conditions for verbatim copying,
    provided that this copyright notice is included exactly as in
    the original, and that the entire resulting derived work is
    distributed under the terms of a permission notice identical
    to this one.

    Permission is granted to copy and distribute translations
    of this document into another language, under the above
    conditions for modified versions.

    If you are intending to incorporate this document into a
    published work, please contact me, and I will make an effort
    to ensure that you have the most up to date information
    available. In the past, out of date versions of the Linux
    howto documents have been published, which caused the developers
    undue grief from being plagued with questions that were already
    answered in the up to date versions.

<sect>What card should I buy for Linux?<label id="what-card">
<p>

    The answer to this question depends heavily on exactly what
    you intend on doing with your net connection, and how much
    traffic it will see.

    If you only expect a single user to be doing the occasional
    ftp session or WWW connection, then an old 8 bit wd8003 card
    will keep you happy.

    If you intend to set up a server, and you require the CPU
    overhead of Rx'ing and Tx'ing ether packets to be kept
    at a minimum, you probably want to look at one of the
    newer PCI cards with the DEC 21040 chip, or the
    AMD PCnet-PCI chip.

    If you fall somewhere in the middle of the above, then any
    one of the 16 bit ISA cards with stable drivers will do the
    job for you.

<sect1>So What Drivers are Stable?
<p>

    Of the 16 bit ISA cards, the following drivers are very
    mature, and you shouldn't have any problems if you buy
    a card that uses these drivers.

    SMC-Ultra/EtherEZ, WD80x3, 3c509, 3c503/16, Lance, NE2000.

    This is not to say that all the other drivers are unstable.
    It just happens that the above are the oldest and most
    used of all the linux drivers, making them the safest
    choice.

    Note that some el-cheapo motherboards can have trouble
    with the bus-mastering that the lance cards do, and
    some el-cheapo NE2000 clones can have trouble getting
    detected at boot.

    As for PCI cards, the PCnet-PCI cards that use the
    lance driver are a safe choice (except for the Boca
    cards as they have hardware flaws). The Allied Telsyn AT2450
    is a PCnet-PCI implementation that is known to work well.

    The DEC 21040 `tulip' driver and the 3c59x `vortex' driver
    are relatively new drivers, but have proven themselves to be quite
    stable already. (The newer 21041 based cards may be slightly
    problematic at the moment, but this will no doubt be
    resolved quickly.)

<sect1>Eight bit vs 16 bit Cards<label id="8-vs-16">
<p>

    You probably can't buy a new 8 bit ISA ethercard anymore,
    but you will find lots of them turning up at computer swap
    meets and the like for the next few years, at very low prices.
    This will make them popular for ``home-ethernet'' systems.

    Some 8 bit cards that will provide adequate performance for
    light to average use are the wd8003, the 3c503 and the ne1000.
    The 3c501 provides poor performance, and these poor 10 year
    old relics of the XT days should be avoided.

    The 8 bit data path doesn't hurt performance that much, as you
    can still expect to get about 500 to 800kB/s ftp download speed
    to an 8 bit wd8003 card (on a fast ISA bus) from a fast host.
    And if most of your net-traffic is going to remote sites, then
    the bottleneck in the path will be elsewhere, and the only speed
    difference you will notice is during net activity on your local
    subnet.

<sect1>32 Bit / VLB / PCI Ethernet Cards
<p>

    There aren't many 32 bit ethercard device drivers because there
    aren't that many 32 bit ethercards.
    There aren't many 32 bit ethercards out there because a 10Mbs
    network doesn't justify spending the 5x price increment for
    the 32 bit interface. (For a 100Mbs network, it is a different
    story though.)

    See <ref id="data-xfer" name="Programmed I/O vs. ..."> as to why
    having an ethercard on an 8MHz ISA bus is really not a
    bottleneck. Even though having the ethercard on a fast bus won't
    necessarily mean faster transfers, it will usually mean reduced
    CPU overhead, which is good for multi-user systems.

    AMD has the 32 bit PCnet-VLB and PCnet-PCI chips.
    See <ref id="pcnet-32" name="AMD PCnet-32"> for info on the
    32 bit versions of the LANCE / PCnet-ISA chip.

    The DEC 21040 PCI chip is another option
    (see <ref id="dec-21040" name="DEC 21040">) for power-users.
    Many manufacturers produce cards that use this chip.

    3Com's `Vortex' PCI cards are also another option, and the price
    is quite cheap if you can get one under their evaluation
    deal while it lasts.  (see <ref id="vortex" name="3c590/3c595">)

<sect1>Available 100Mbs Cards and Drivers
<p>

    The present list of supported 100Mbs hardware is as follows:
     cards with the DEC 21140 chip; the 3c595 Vortex card;
    and the HP 100VG ANY-LAN. The drivers for the first two are
    quite stable, but feedback on the HP driver has been low
    so far as it has only been around since early 1.3.x kernels.

    The EtherExpressPro10/100 is not presently supported.
    Getting documentation out of intel is difficult to
    impossible, which is the main reason.

    The 21140 100Base-? chip is supported with the same driver as
    its 10Mbs counterpart, the 21040. SMC's 100Mbs EtherPower PCI
    card uses this chip. As with the 21040, you have a choice of
    two drivers to pick from.

    Also have a look at the information on Donald's WWW site,
    at the following URL:

    <url url="http://cesdis.gsfc.nasa.gov/linux/misc/100mbs.html"
        name="100Mbs Ethernet">

    Donald has been doing most of his work with the
    SMC EtherPower-10/100 cards, and reports getting about 4.6MB/s
    application to application with TCP on P5-100 Triton machines.

    (See <ref id="vortex" name="3c595"> and
    <ref id="dec-21040" name="DEC 21140"> for more details.)

    For 100VG information, see
    <url url="http://cesdis.gsfc.nasa.gov/linux/drivers/100vg.html"
        name="Donald's 100VG Page">

    You may also be interested in looking at:

    <url url="http://alumni.caltech.edu/~dank/fe/"
        name="Dan Kegel's Fast Ethernet Page">

<sect1>Programmed I/O vs. Shared Memory vs. DMA<label id="data-xfer">
<p>

    Ethernet is 10Mbs. (Don't be pedantic, 3Mbs and 100Mbs don't count.)
    If you can already send and receive back-to-back packets, you just
    can't put more bits over the wire. Every modern ethercard can receive
    back-to-back packets. The Linux DP8390 drivers come pretty close to
    sending back-to-back packets (depending on the current interrupt
    latency) and the 3c509 and AT1500 hardware has no problem at all
    automatically sending back-to-back packets.

    The ISA bus can do 5.3MB/sec (42Mb/sec), which sounds like more than
    enough. You can use that bandwidth in several ways:

<sect2>Programmed I/O (e.g. NE2000, HP)
<p>

    Pro: Doesn't use any constrained system resources,
    just a few I/O registers, and has no 16M limit.

    Con: Usually the slowest transfer rate, the CPU is waiting
    the whole time, and interleaved packet access is usually
    difficult to impossible.

<sect2>Shared memory (e.g. WD80x3, SMC-Ultra, 3c503)
<p>

    Pro: Simple, faster than programmed I/O, and allows random
    access to packets. The linux drivers compute the checksum of
    incoming IP packets as they are copied off the card, resulting in
    a further reduction of CPU usage vs. an equivalent PIO card.

    Con: Uses up memory space (a big one for DOS users, only a minor
    issue under Linux), and it still ties up the CPU.

<sect2>Slave (normal) Direct Memory Access (e.g. none for Linux!)
<p>

    Pro: Frees up the CPU during the actual data transfer.

    Con: Checking boundary conditions, allocating contiguous buffers,
    and programming the DMA registers makes it the slowest
    of all techniques.  It also uses up a scarce DMA
    channel, and requires aligned low memory buffers.

<sect2>Bus Master Direct Memory Access (e.g. LANCE, DEC 21040) <label id="master">
<p>

    Pro: Frees up the CPU during the data transfer, can string
    together buffers, can require little or no CPU time lost on
    the ISA bus.

    Con: Requires low-memory buffers and a DMA channel. Any
    bus-master will have problems with other bus-masters that
    are bus-hogs, such as some primitive SCSI adaptors. A few
    badly-designed motherboard chipsets have problems with
    bus-masters. And a reason for not using <em/any/ type of
    DMA device is using a 486 processor designed for
    plug-in replacement of a 386: these processors must
    flush their cache with each DMA cycle. (This includes
    the Cx486DLC, Ti486DLC, Cx486SLC, Ti486SLC, etc.)

<sect1>Type of cable that your card should support<label id="cable-intro">
<p>

    If you are setting up a small ``personal'' network, you will probably
    want to use thinnet or thin ethernet cable. This is the style with the
    standard BNC connectors. See <ref id="cable" name="Cables, Coax...">
    for other concerns with different types of ethernet cable.

    Most ethercards also come in a `Combo' version for only
    $10-$20 more.
    These have both twisted pair and thinnet transceiver built-in,
    allowing you to change your mind later.

    The twisted pair cables, with the RJ-45 (giant phone jack)
    connectors is technically called 10BaseT. You may also
    hear it called UTP (Unsheilded Twisted Pair).

    The thinnet, or thin ethernet cabling, (RG-58 coaxial cable)
    with the BNC (metal push and turn-to-lock) connectors is
    technically called 10Base2.

    The older thick ethernet (10mm coaxial cable) which is only
    found in older installations is called 10Base5.

    Large corporate installations will most likely use 10BaseT
    instead of 10Base2. 10Base2 does not offer an easy upgrade
    path to the new upcoming 100Base-whatever.

<sect>Frequently Asked Questions<label id="faq">
<p>

    Here are some of the more frequently asked questions about using
    Linux with an Ethernet connection. Some of the more specific
    questions are sorted on a `per manufacturer basis'.
    However, since this
    document is basically `old' by the time you get it, any `new' problems
    will not appear here instantly. For these, I suggest that you make
    efficient use of your newsreader. For example, nn users would type

<tscreen>
        nn -xX -s'3c'
</tscreen>

    to get all the news articles in your subscribed list that have
    `3c' in the subject. (ie. 3com, 3c509, 3c503, etc.)
    The moral: Read the man page for your newsreader.

<sect1>Alpha Drivers -- Getting and Using them<label id="alfa">
<p>

    I heard that there is an updated or alpha driver available
    for my card. Where can I get it?

    The newest of the `new' drivers can be found on Donald's new
    ftp site: <tt/cesdis.gsfc.nasa.gov/ in the
    <tt>/pub/linux/</tt> area. Things
    change here quite frequently, so just look around for it.
    There is still all the stuff on the old ftp site <tt/ftp.super.org/
    in <tt>/pub/linux</tt>, but this is not being actively maintained,
    and hence will be of limited value to most people.

    Note that most of the `useable' alpha drivers have been
    included in the standard kernel source tree. When running
    <tt/make config/ you will be asked if you want to be offered
    ALPHA test drivers.

    Now, if it really is an alpha, or pre-alpha driver, then please
    treat it as such. In other words, don't complain because you
    can't figure out what to do with it. If you can't figure out
    how to install it, then you probably shouldn't be testing it.
    Also, if it brings your machine down, don't complain. Instead,
    send us a well documented bug report, or even better, a patch!

    People reading this while <em/net-surfing/ may want to check out:

    <url url="http://cesdis.gsfc.nasa.gov/pub/linux/linux.html"
        name="Don's Linux Home Page">

    for the latest dirt on what is new and upcoming.

<sect1>Using More than one Ethernet Card per Machine<label id="two-card">
<p>

    What needs to be done so that Linux can run two ethernet cards?

    The hooks for multiple ethercards are all there.
    However, note that only <em/one/ ethercard is
    auto-probed for by default.  This avoids a lot of possible
    boot time hangs caused by probing sensitive cards.

    There are two ways that you can enable auto-probing for
    the second (and third, and...) card. The easiest
    method is to pass boot-time arguments to the kernel,
    which is usually done by LILO. Probing for the
    second card can be achieved by using a boot-time argument
    as simple as <tt/ether=0,0,eth1/. In this
    case <tt/eth0/ and <tt/eth1/ will be assigned in the order
    that the cards are found at boot.  Say if you want
    the card at <tt/0x300/ to be <tt/eth0/ and
    the card at <tt/0x280/ to be <tt/eth1/ then you could use

<tscreen>
    LILO: linux ether=5,0x300,eth0 ether=15,0x280,eth1
</tscreen>

    The <tt/ether=/ command accepts more than the IRQ + i/o
    + name shown above. Please have a look at
    <ref id="lilo" name="Passing Ethernet Arguments...">
    for the full syntax, card specific parameters, and LILO tips.

    These boot time arguments can be made permanent so that you
    don't have to re-enter them every time. See the LILO
    configuration option `<tt/append/' in the LILO manual.

    The second way (not recommended) is to edit the file
    <tt/Space.c/ and replace the <tt/0xffe0/ entry for the
    i/o address with a zero. The <tt/0xffe0/ entry tells it
    not to probe for that device -- replacing it with a zero
    will enable autoprobing for that device.
    If you <em/really/ need more than four ethernet
    cards in one machine, then you can add an <tt/eth4/ entry
    and an <tt/eth5/ entry and so on.

    Note that if you are intending to use Linux as a gateway between
    two networks, you will have to re-compile a kernel with IP
    forwarding enabled. Usually using an old AT/286 with something
    like the `kbridge' software is a better solution.

    If you are viewing this while <em/net-surfing/, you may wish
    to look at a mini-howto Donald has on his WWW site. Check out
    <url url="http://cesdis.gsfc.nasa.gov/linux/misc/multicard.html"
        name="Multiple Ethercards">.

    For module users with 8390 based cards, you can have a single
    module control multiple cards as well. Please see
    <ref id="multmod8390" name="8390 Based Cards as Modules"> for
    module specific information about using multiple cards.



<sect1>Poor NE2000 Clones<label id="ne2k-clones">
<p>

    Here is a list of some of the NE-2000 clones that are known to
    have various problems. Most of them aren't fatal. In the case
    of the ones listed as `bad clones' -- this usually indicates that
    the cards don't have the two NE2000 identifier bytes. NEx000-clones
    have a Station Address PROM (SAPROM) in the packet buffer memory space.
    NE2000 clones have <tt/0x57,0x57/ in bytes <tt/0x0e,0x0f/
    of the SAPROM, while other supposed NE2000 clones must be
    detected by their SA prefix.

    This is not a comprehensive list of all the NE2000 clones that
    don't have the <tt/0x57,0x57/ in bytes <tt/0x0e,0x0f/
    of the SAPROM. There are probably hundreds of them. If you
    get a card that causes the driver to report an `invalid
    signature' then you will have to add your cards signature
    to the driver. The process for doing this is described
    below.

    <bf/Accton NE2000/
    -- might not get detected at boot, see below.

    <bf/Aritsoft LANtastic AE-2/
    -- OK, but has flawed error-reporting registers.

    <bf/AT-LAN-TEC NE2000/
    -- clone uses Winbond chip that traps SCSI drivers

    <bf/ShineNet LCS-8634/
    -- clone uses Winbond chip that traps SCSI drivers

    <bf/Cabletron E10**, E20**, E10**-x, E20**-x/
    -- bad clones, but the driver checks for them. See
    <ref id="e10xx" name="E10**">.

    <bf/D-Link Ethernet II/
    -- bad clones, but the driver checks for them. See
    <ref id="de-100" name="DE-100 / DE-200">.

    <bf/DFI DFINET-300, DFINET-400/
    -- bad clones, but the driver checks for them. See
    <ref id="dfi-300" name="DFI-300 / DFI-400">

    <bf/EtherNext UTP8, EtherNext UTP16/
    -- bad clones, but the driver checks for them.

<sect1>Problems with NE1000 / NE2000 cards (and clones)<label id="ne2k-probs">
<p>

    <bf/Problem:/
    NE*000 card hangs machine, sometimes with a `DMA conflict'
    message, sometimes completely silently.

    <bf/Reason:/
    There were some bugs in the driver and the upper networking
    layers that caused this. They have been fixed in kernels
    v1.2.9 and above. Upgrade your kernel.

    <bf/Problem:/
    NE*000 card hangs machine during NE probe, or can not
    read station address properly.

    <bf/Reason:/
    Kernels previous to v1.3.7 did not fully reset the card
    after finding it at boot. Some cheap cards are not left in
    a reasonable state after power-up and need to be fully
    reset before any attempt is made to use them. Also, a
    previous probe may have upset the NE card prior to the
    NE probe taking place. In that case, look in to using
    the  ``reserve='' boot keyword to protect the card from other
    probes.

    <bf/Problem:/
    NE*000 driver reports `not found (no reset ack)' during boot
    probe.

    <bf/Reason:/
    This is related to the above change. After the initial
    verification that an 8390 is at the probed i/o address, the
    reset is performed. When the card has completed the reset,
    it is supposed to acknowedge that the reset has completed.
    Your card doesn't, and so the driver assumes that no NE card
    is present.

    <bf/Solution:/

    Change the two lines as shown below in <tt>drivers/net/ne.c</tt>
<code>
-            printk(" not found (no reset ack).\n");
-            return ENODEV;
+            printk(" (warning: no reset ack)");
+            break;
</code>

    This will allow the card detection to continue, even if your
    card doesn't ACK the reset.

    <bf/Problem:/
    NE*000 card hangs machine at first network access.

    <bf/Reason:/
    This problem has been reported for kernels as old as 1.1.57
    to the present. It appears confined to the software configurable
    clone cards. It appears that they expect to be initialized in
    some special way.

    <bf/Solution:/
    Several people have reported that running the supplied DOS
    software config program and/or the supplied DOS driver prior
    to warm booting (i.e. loadlin or the `three-finger-salute')
    into linux allowed the card to work. This would indicate
    that these cards need to be initialized in a particular
    fashion, slightly different than what the present Linux
    driver does.

    <bf/Problem:/
    NE*000 ethercard at <tt/0x360/ doesn't get detected anymore.

    <bf/Reason:/
    Recent kernels ( > 1.1.7X) have more sanity checks with respect
    to overlapping i/o regions. Your NE2000 card is <tt/0x20/ wide in
    i/o space, which makes it hit the parallel port at <tt/0x378/.
    Other devices that could be there are the second floppy
    controller (if equipped) at <tt/0x370/ and the secondary
    IDE controller at <tt/0x376--0x377/.
    If the port(s) are already registered by another driver, the
    kernel will not let the probe happen.

    <bf/Solution:/
    Either move your card to an address like <tt/0x280, 0x340, 0x320/
    or compile without parallel printer support.

    <bf/Problem:/
    Network `goes away' every time I print something (NE2000)

    <bf/Reason:/
    Same problem as above, but you have an older kernel that
    doesn't check for overlapping i/o regions. Use the
    same fix as above, and get a new kernel while you are at it.

    <bf/Problem:/
    NE*000 ethercard probe at 0xNNN: 00 00 C5 ... not found.
    (invalid signature yy zz)

    <bf/Reason:/
    First off, do you have a NE1000 or NE2000 card at the addr. 0xNNN?
    And if so, does the hardware address reported look like a valid
    one? If so, then you have a poor NE*000 clone. All NE*000 clones
    are supposed to have the value <tt/0x57/ in bytes 14 and 15 of the
    SA PROM on the card. Yours doesn't -- it has `yy zz' instead.

    <bf/Solution:/
    The driver (/usr/src/linux/drivers/net/ne.c) has a "Hall of Shame"
    list at about line 42. This list is used to detect poor clones.
    For example, the DFI cards use `DFI' in the first 3 bytes of the
    prom, instead of using 0x57 in bytes 14 and 15, like they are
    supposed to.

    You can determine what the first 3 bytes of your card PROM are
    by adding a line like:

<verb>
    printk("PROM prefix: %2.2x %2.2x %2.2x\n",SA_prom[0],SA_prom[1],SA_prom[2]);
</verb>

    into the driver, right after the error message you got above, and
    just before the "return ENXIO" at line 227.

    Reboot with this change in place, and after the detection fails, you
    will get the three bytes from the PROM like the DFI example above.
    Then you can add your card to the bad_clone_list&lsqb&rsqb at about
    line 43. Say the above line printed out:

<tscreen>
    PROM prefix: 0x3F 0x2D 0x1C
</tscreen>

    after you rebooted. And say that the 8 bit version of your card was
    called the "FOO-1k" and the 16 bit version the "FOO-2k". Then you would
    add the following line to the bad_clone_list&lsqb&rsqb:

<tscreen>
    {"FOO-1k", "FOO-2k", {0x3F, 0x2D, 0x1C,}},
</tscreen>

    Note that the 2 name strings you add can be anything -- they are just
    printed at boot, and not matched against anything on the card.
    You can also take out the "printk()" that you added above, if you want.
    It shouldn't hit that line anymore anyway. Then recompile once more,
    and your card should be detected.

    <bf/Problem:/
    Errors like <tt/DMA address mismatch/

    Is the chip a real NatSemi 8390? (DP8390, DP83901, DP83902 or DP83905)?
    If not, some clone chips don't correctly implement the transfer
    verification register. MS-DOS drivers never do error checking,
    so it doesn't matter to them. (Note: The DMA address check is not done
    by default as of v1.2.4 for performance reasons. Enable it with
    the `NE_SANITY' define in <tt/ne.c/ if you want the check done.)

    Are most of the messages off by a factor of 2?
    If so:  Are you using the NE2000 in a 16 bit slot?
    Is it jumpered to use only 8 bit transfers?

    The Linux driver expects a NE2000 to be in a 16 bit slot. A NE1000 can
    be in either size slot. This problem can also occur with some clones,
    notably D-Link 16 bit cards, that don't have the correct ID bytes
    in the station address PROM.

    Are you running the bus faster than 8Mhz?
    If you can change the speed (faster or slower), see if that
    makes a difference. Most NE2000 clones will run at 16MHz, but
    some may not. Changing speed can also mask a noisy bus.

    What other devices are on the bus?
    If moving the devices around changes the reliability, then you
    have a bus noise problem -- just what that error message was
    designed to detect. Congratulations, you've probably found the
    source of other problems as well.

    <bf/Problem:/
    The machine hangs during boot right after the `8390...'  or
    `WD....' message. Removing the NE2000 fixes the problem.

    <bf/Solution:/
    Change your NE2000 base address to <tt/0x340/. Alternatively, you
    can use the device registrar implemented in 0.99pl13 and later
    kernels.

    <bf/Reason:/
    Your NE2000 clone isn't a good enough clone. An active
    NE2000 is a bottomless pit that will trap any driver
    autoprobing in its space. The other ethercard drivers take
    great pain to reset the NE2000 so that it's safe, but some
    clones cannot be reset. Clone chips to watch out for:
    Winbond 83C901. Changing the NE2000 to a less-popular
    address will move it out of the way of other autoprobes,
    allowing your machine to boot.


    <bf/Problem:/
    The machine hangs during the SCSI probe at boot.

    <bf/Reason:/
    It's the same problem as above, change the
    ethercard's address, or use the device registrar.

    <bf/Problem:/
    The machine hangs during the soundcard probe at boot.

    <bf/Reason:/
    No, that's really during the silent SCSI probe, and it's
    the same problem as above.

    <bf/Problem:/
    NE2000 not detected at boot - no boot messages at all

    Donald writes:
    `A few people have reported a problem with detecting the Accton NE2000.
    This problem occurs only at boot-time, and the card is later detected
    at run-time by the identical code my (alpha-test) ne2k diagnostic
    program. Accton has been very responsive, but I still haven't tracked
    down what is going on. I've been unable to reproduce this problem
    with the Accton cards we purchased. If you are having this problem,
    please send me an immediate bug report. For that matter, if you have
    an Accton card send me a success report, including the type of the
    motherboard. I'm especially interested in finding out if this problem
    moves with the particular ethercard, or stays with the motherboard.'

    Here are some things to try, as they have fixed it for some people:

<itemize>

<item>    Change the bus speed, or just move the card to a different slot.

<item>    Change the `I/O recovery time' parameter in the BIOS
    chipset configuration.

</itemize>

<sect1>Problems with WD80*3 cards<label id="8013-probs">
<p>

    <bf/Problem:/
    A WD80*3 is falsely detected. Removing the sound or
    MIDI card eliminates the `detected' message.

    <bf/Reason:/
    Some MIDI ports happen to produce the same checksum as a
    WD ethercard.

    <bf/Solution:/
    Update your ethercard driver: new versions include an
    additional sanity check. If it is the midi chip at 0x388
    that is getting detected as a WD living at 0x380, then
    you could also use:
<verb>
    LILO: linux reserve=0x380,8
</verb>

    <bf/Problem:/
    You get messages such as the following with your 80*3:

<verb>
    eth0: bogus packet size, status = NNNN
    kmalloc called with impossibly large argument (65400)
    eth0: Couldn't allocate sk_buff of size 65400
    eth0: receiver overrun
</verb>

    <bf/Reason:/
    There is a shared memory problem.

    <bf/Solution:/
    If the problem is sporadic, you have hardware problems.
    Typical problems that are easy to fix are board conflicts,
    having cache or `shadow ROM' enabled for that region, or
    running your bus faster than 8Mhz. There are also a
    surprising number of memory failures on ethernet cards,
    so run a diagnostic program if you have one for your
    ethercard.

    If the problem is continual, and you have have to reboot
    to fix the problem, record the boot-time probe message
    and mail it to [e-mail] - Take
    particular note of the shared memory location.

    <bf/Problem:/
    WD80*3 will not get detected at boot.

    <bf/Reason:/
    Earlier versions of the Mitsumi CD-ROM (mcd) driver probe
    at 0x300 will succeed if just about <em/anything/ is that I/O location.
    This is bad news and needs to be a bit more robust.
    Once another driver registers that it `owns' an I/O
    location, other drivers (incl. the wd80x3) are `locked
    out' and can not probe that addr for a card.

    <bf/Solution:/
    Recompile a new kernel without any excess drivers that
    you aren't using, including the above mcd driver.
    Or try moving your ethercard to a new I/O addr. Valid
    I/O addr. for all the cards are listed in
    <ref id="probe" name="Probed Addresses">
    You can also point the mcd driver off in another direction
    by a boot-time parameter (via LILO) such as:
<tscreen>
    mcd=0x200,12
</tscreen>

    <bf/Problem:/
    Old wd8003 and/or jumper-settable wd8013 always get the IRQ wrong.

    <bf/Reason:/
    The old wd8003 cards and jumper-settable wd8013 clones don't
    have the EEPROM that the driver can read the IRQ setting from.
    If the driver can't read the IRQ, then it tries to auto-IRQ
    to find out what it is. And if auto-IRQ returns zero, then
    the driver just assigns IRQ 5 for an 8 bit card or IRQ 10 for
    a 16 bit card.

    <bf/Solution:/
    Avoid the auto-IRQ code, and tell the kernel what the IRQ
    that you have jumpered the card to is via a boot time argument.
    For example, if you are using IRQ 9, using the following
    should work.

<tscreen>
    LILO: linux ether=9,0,eth0
</tscreen>

<sect1>Problems with 3Com cards<label id="3com-probs">
<p>
    <bf/Problem:/
    The 3c503 picks IRQ N, but this is needed for some
    other device which needs IRQ N. (eg. CD ROM driver, modem, etc.)
    Can this be fixed without compiling this into the kernel?

    <bf/Solution:/
    The 3c503 driver probes for a free IRQ line in the order
    {5, 9/2, 3, 4}, and it should pick a line which isn't being
    used. Very old drivers used to pick the IRQ line
    at boot-time, and the current driver (0.99pl12 and newer) chooses when
    the card is open()/ifconfig'ed.

    Alternately, you can fix the IRQ at boot by passing
    parameters via LILO. The following selects IRQ9, base
    location 0x300, <ignored value>, and if_port #1 (the
    external transceiver).

<tscreen>
      LILO: linux ether=9,0x300,0,1,eth0
</tscreen>

    The following selects IRQ3, probes for the base location,
    <ignored value>, and the default if_port #0 (the internal
    transceiver)

<tscreen>
      LILO: linux ether=3,0,0,0,eth0
</tscreen>

    <bf/Problem:/
    3c503: configured interrupt X invalid, will use autoIRQ.

    <bf/Reason:/
    The 3c503 card can only use one of IRQ{5, 2/9, 3, 4}
    (These are the only lines that are connected to the card.)
    If you pass in an IRQ value that is not in the above
    set, you will get the above message.
    Usually, specifying an interrupt value for the 3c503 is
    not necessary. The 3c503 will autoIRQ when it gets
    ifconfig'ed, and pick one of IRQ{5, 2/9, 3, 4}.

    <bf/Solution:/
    Use one of the valid IRQs listed above, or enable
    autoIRQ by not specifying the IRQ line at all.

    <bf/Problem:/
    The supplied 3c503 drivers don't use the AUI (thicknet) port.
    How does one choose it over the default thinnet port?

    <bf/Solution:/
    The 3c503 AUI port can be selected at boot-time with 0.99pl12
    and later. The selection is overloaded onto the low bit of
    the currently-unused dev->rmem_start variable, so a boot-time
    parameter of:

<tscreen>
    LILO: linux ether=0,0,0,1,eth0
</tscreen>

    should work. A boot line to force IRQ 5, port base 0x300,
    and use an external transceiver is:

<tscreen>
    LILO: linux ether=5,0x300,0,1,eth0
</tscreen>

    With kernels 1.3.42 and newer, you can specify the
    AUI port when loading as a module as well. Just append
    <tt/xcvr=1/ to the <tt/insmod/ command line along with
    your i/o and irq values.

    Also note that kernel revisions 1.00 to 1.03 had an
    interesting `feature'. They would switch to the AUI port
    when the internal transceiver failed. This is a problem,
    as it will <em/never/ switch back if for example you
    momentarily disconnect the cable. Kernel versions 1.04
    and newer only switch if the very first Tx attempt fails.

<sect1>Problems with Hewlett Packard Cards<label id="hp-probs">
<p>

    <bf/Problem:/
    HP Vectra using built in AMD LANCE chip gets IRQ and DMA wrong.

    <bf/Solution:/
    The HP Vectra uses a different implementation to the
    standard HP-J2405A. The `lance.c' driver used to
    <em/always/ use the value in the setup register of an HP Lance
    implementation.  In the Vectra case it's reading an invalid
    0xff value. Kernel versions newer than about 1.1.50
    now handle the Vectra in an appropriate fashion.

    <bf/Problem:/
    HP Card is not detected at boot, even though kernel was
    compiled with `HP PCLAN support'.

    <bf/Solution:/
    You probably have a HP PCLAN+  -- note the `plus'. Support
    for the PCLAN+ was added to mid/late versions of 1.1.
    Recompile a (possibly newer) kernel with support for the
    HP PCLAN+ and you should be in business.


<sect1>FAQs Not Specific to Any Card.
<p>

<sect2>Ethercard is Not Detected at Boot.
<p>

    The usual reason for this is that people are using a kernel
    that does not have support for their particular card built
    in. If you are using a pre-compiled kernel that is part of
    a distribution set, then check the documentation to see which
    kernel you installed, and if it was built with support
    for your particular card. If it wasn't, then your
    options are to try and get one that has support
    for your card, or build your own.

    It is usually wise to build your own kernel with only
    the drivers you need, as this cuts down on the kernel
    size (saving your precious RAM for applications!) and
    reduces the number of device probes that can upset
    sensitive hardware. Building a kernel is not as complicated
    as it sounds. You just have to answer yes or no to
    a bunch of questions about what drivers you want, and
    it does the rest.

    The next main cause is having another device using part
    of the i/o space that your card needs. Most cards are
    16 or 32 bytes wide in i/o space. If your card is set
    at 0x300 and 32 bytes wide, then the driver will ask
    for 0x300-0x31f. If any other device driver has registered
    even one port anywhere in that range, the probe will not
    take place at that address and  the driver will silently
    continue to the next of the probed addresses. So, after
    booting, do a <tt>cat /proc/ioports</tt> and verify that the
    full iospace that the card will require is vacant.

    Another problem is having your card jumpered to an i/o
    address that isn't probed by default. There is a list
    <ref id="probe" name="probed addresses"> for each card
    in this document. Even if the i/o setting of your card is
    not in the list of porbed addresses, you can supply
    it at boot with the <tt/ether=/ command as described in
    <ref id="lilo" name="Passing Ethernet Arguments...">


<sect2><tt/ifconfig/ reports the wrong i/o address for the card.
<p>

    No it doesn't. You are just interpreting it incorrectly.
    This is <em/not/ a bug, and the numbers reported are correct. It just
    happens that some 8390 based cards (wd80x3, smc-ultra, etc) have the
    actual 8390 chip living at an offset from the first assigned i/o port.
    Try <tt>cd /usr/src/linux/drivers/net;grep NIC_OFFSET *.c|more</tt>
    to see what is going on. This is the value stored in
    <tt/dev->base_addr/, and is what <tt/ifconfig/ reports. If you
    want to see the full range of ports that your card uses, then try
    <tt>cat /proc/ioports</tt> which will give the numbers you expect.

<sect2>Shared Memory ISA cards in PCI Machine
<p>

    This will usually show up as reads of lots of <tt/0xffff/ values.
    No shared memory cards of any type will work in a PCI machine
    unless you have the PCI ROM BIOS/CMOS SETUP configuration set
    properly. You have to set it to allow shared memory access
    from the ISA bus for the memory region that your card is trying
    to use. If you can't figure out which settings are applicable
    then ask your supplier or local computer guru.

<sect2>Asynchronous Transfer Mode (ATM) Support
<p>

    Werner Almesberger has been playing around with ATM support
    for linux.
    He has been working with the Efficient Networks ENI155p
    board (<url url="http://www.efficient.com/"
        name="Efficient Networks">)
    and the Zeitnet ZN1221 board
    (<url url="http://www.zeitnet.com/" name="Zeitnet">).

    Werner says that the driver for the ENI155p is rather
    stable, while the driver for the ZN1221 is presently
    unfinished.

    Check the latest/updated status at the following URL:

    <url url="http://lrcwww.epfl.ch/linux-atm/"
        name="Linux ATM Support">

<sect2>FDDI Support
<p>
    Is there FDDI support for Linux?

    Donald writes: `No, there is no Linux driver for any FDDI boards.
    I come from a place with supercomputers, so an external
    observer might think
    FDDI would be high on my list. But FDDI never delivered end-to-end
    throughput that would justify its cost, and it seems to be a nearly
    abandoned technology now that 100base{X,Anynet} seems imminent.
    (And yes, I know you can now get FDDI boards for <$1K. That
    seems to be a last-ditch effort to get some return on the
    development investment. Where is the next generation of FDDI
    going to come from?)'

<sect2>Ethernet Cards for Linux on Alpha/AXP PCI Boards
<p>

    At present only the depca, de4x5 and all the 8390 drivers
    (wd, smc-ultra, ne, 3c503, e2100, ac3200, hp, hp-plus) have
    been made `architecture independent' so as to work on the
    DEC Alpha CPU based systems.

    Note that the changes that are required aren't that complicated.
    You only need to do the following:

    -multiply all <tt/jiffies/ related values by HZ/100 to account
    for the different HZ value that the Alpha uses.
    (i.e <tt/timeout=2;/ becomes <tt>timeout=2*HZ/100;</tt>)

    -replace any i/o memory (640k to 1MB) pointer dereferences with
    the appropriate readb() writeb() readl() writel() calls, as
    shown in this example.

<code>
-    int *mem_base = (int *)dev->mem_start;
-    mem_base[0] = 0xba5eba5e;
+    unsigned long mem_base = dev->mem_start;
+    writel(0xba5eba5e, mem_base);
</code>

    -replace all memcpy() calls that have i/o memory as source or
    target destinations with the appropriate one of
    <tt/memcpy_fromio()/ or <tt/memcpy_toio()/.

<sect2>Linking 10BaseT without a Hub
<p>

     Can I link 10BaseT (RJ45) based systems together without a hub?

     You can link 2 machines easily, but no more than that, without
     extra devices/gizmos. See <ref id="utp" name="Twisted Pair">
     -- it explains
     how to do it. And no, you can't hack together a hub just by
     crossing a few wires and stuff. It's pretty much impossible
     to do the collision signal right without duplicating a hub.

<sect2>SIOCSIFxxx: No such device
<p>

    I get a bunch of `SIOCSIFxxx: No such device' messages at
    boot, followed by a `SIOCADDRT: Network is unreachable'
    What is wrong?

    Your ethernet device was not detected at boot, and when
    <tt/ifconfig/ and <tt/route/ are run, they have no device
    to work with. Use <tt>dmesg | more</tt> to review the
    boot messages and see if there are any messages about
    detecting an ethernet card.

<sect2>SIOCSFFLAGS: Try again
<p>

    I get `SIOCSFFLAGS: Try again' when I run `ifconfig' -- Huh?

    Some other device has taken the IRQ that your ethercard
    is trying to use, and so the ethercard can't use the IRQ.
    You don't necessairly need to reboot to resolve this, as
    some devices only grab the IRQs when they need them and
    then release them when they are done. Examples are some
    sound cards, serial ports, floppy disk driver, etc. You
    can type <tt>cat /proc/interrupts</tt> to see which interrupts
    are presently <em/in use/. Most of the
    Linux ethercard drivers only grab the IRQ when they are
    opened for use via `ifconfig'. If you can get the other
    device to `let go' of the required IRQ line, then you
    should be able to `Try again' with ifconfig.

<sect2>Link UNSPEC and HW-addr of 00:00:00:00:00:00
<p>

    When I run ifconfig with no arguments, it reports that
    LINK is UNSPEC (instead of 10Mbs Ethernet) and it
    also says that my hardware address is all zeros.

    This is because people are running a newer version of
    the `ifconfig' program than their kernel version. This
    new version of ifconfig is not able to report these properties
    when used in conjunction with an older kernel. You can either
    upgrade your kernel, `downgrade' ifconfig, or simply ignore
    it. The kernel knows your hardware address, so it really
    doesn't matter if ifconfig can't read it.

<sect2>Huge Number of RX and TX Errors
<p>

    When I run ifconfig with no arguments, it reports that I
    have a huge error count in both rec'd and transmitted
    packets. It all seems to work ok -- What is wrong?

    Look again. It says <tt/RX packets/ <em/big number/ <bf/PAUSE/
    <tt/errors 0/ <bf/PAUSE/ <tt/dropped 0/ <bf/PAUSE/ <tt/overrun 0/.
    And the same for the <tt/TX/ column.
    Hence the big numbers you are seeing are the total number of
    packets that your machine has rec'd and transmitted.
    If you still find it confusing, try typing
    <tt>cat /proc/net/dev</tt> instead.

<sect2>Entries in <tt>/dev/</tt> for Ethercards
<p>

     I have /dev/eth0 as a link to /dev/xxx. Is this right?

     Contrary to what you have heard, the files in /dev/* are not used.
     You can delete any <tt>/dev/wd0, /dev/ne0</tt> and similar entries.

<sect2>Linux and ``trailers''
<p>

    Should I disable trailers when I `ifconfig' my ethercard?

    You can't disable trailers, and you shouldn't want
    to. `Trailers' are a hack to avoid data copying in the
    networking layers. The idea was to use a trivial
    fixed-size header of size `H', put the variable-size header
    info at the end of the packet, and allocate all
    packets `H' bytes before the start of a page. While it was a
    good idea, it turned out to not work well in practice.
    If someone suggests the use of `-trailers', note that it
    is the equivalent of sacrificial goats blood. It won't do
    anything to solve the problem, but if problem fixes itself then
    someone can claim deep magical knowledge.


<sect2>Access to the raw Ethernet Device
<p>

    How do I get access to the raw ethernet device in linux,
    without going through TCP/IP and friends?

<code>
    int s=socket(AF_INET,SOCK_PACKET,htons(ETH_P_ALL));
</code>

    This gives you a socket receiving every protocol type.
    Do <tt/recvfrom()/ calls to it and it will fill the sockaddr
    with device type in sa_family and the device name in the
    sa_data array. I don't know who originally invented
    SOCK_PACKET for Linux (its been in for ages) but its superb stuff.
    You can use it to send stuff raw too via <tt/sendto()/ calls.
    You have to have root access to do either of course.

<sect>Performance Tips<label id="perf">
<p>

    Here are some tips that you can use if you are suffering
    from low ethernet throughput, or to gain a bit more
    speed on those ftp transfers.

    The <tt/ttcp.c/ program is a good test for measuring
    raw throughput speed. Another common trick is to do
    a <tt>ftp> get large_file /dev/null</tt> where
    <tt/large_file/ is > 1MB and residing in the buffer
    cache on the Tx'ing machine. (Do the `get' at least
    twice, as the first time will be priming the buffer
    cache on the Tx'ing machine.) You want the file in
    the buffer cache because you are not interested in
    combining the file access speed from the disk into
    your measurement. Which is also why you send the
    incoming data to <tt>/dev/null</tt> instead of onto
    the disk.

<sect1>General Concepts
<p>

    Even an 8 bit card is able to receive back-to-back packets
    without any problems. The difficulty arises when the computer
    doesn't get the Rx'd packets off the card quick enough to
    make room for more incoming packets. If the computer does not
    quickly clear the card's memory of the packets already received,
    the card will have no place to put the new packet.

    In this case
    the card either drops the new packet, or writes over top of
    a previously received packet. Either one seriously interrupts
    the smooth flow of traffic by causing/requesting re-transmissions
    and can seriously degrade performance by up to a factor of 5!

    Cards with more onboard memory are able to ``buffer'' more
    packets, and thus can handle larger bursts of
    back-to-back packets without dropping packets.
    This in turn means that the card does not require as low
    a latency from the the host computer with respect to pulling
    the packets out of the buffer to avoid dropping packets.

    Most 8 bit cards have an 8kB buffer, and most 16 bit cards have
    a 16kB buffer. Most Linux drivers will reserve 3kB of that
    buffer (for two Tx ping-pong buffers), leaving only 5kB of
    receive space for an 8 bit card. This is room enough for
    only three full sized (1500 bytes) ethernet packets.

<sect1>ISA Bus Speed
<p>
    As mentioned above, if the packets are removed from the card
    fast enough, then a drop/overrun condition won't occur even
    when the amount of Rx packet buffer memory is small. The
    factor that sets the rate at which packets are removed from
    the card to the computer's memory is the speed of the data path
    that joins the two -- that being the ISA bus speed. (If the
    CPU is a dog-slow 386sx-16, then this will also play a role.)

    The recommended ISA bus clock is about 8MHz, but many
    motherboards and peripheral devices can run at higher
    frequencies. The clock frequency for the ISA bus can usually
    be set in the CMOS setup, by selecting a divisor of the
    mainboard/CPU clock frequency.

    For example, here are some receive speeds as measured by
    the TTCP program on a 40MHz 486, with an old 8 bit WD8003EP
    card, for different ISA bus speeds.

<code>
    ISA Bus Speed (MHz)    Rx TTCP (kB/s)
    -------------------    --------------
    6.7            740
    13.4            970
    20.0            1030
    26.7            1075
</code>

    You would be hard pressed to do better than 1075kB/s with
    <em/any/ 10Mb/s ethernet card, using TCP/IP. However, don't expect
    every system to work at fast ISA bus speeds. Most systems will
    not function properly at speeds above 13MHz. (Also, most
    PCI systems have the ISA bus speed fixed at 8MHz, so that
    the end user does not have the option of increasing it.)

    In addition to faster transfer speeds, one will usually also
    benefit from a reduction in CPU usage due to the shorter
    duration memory and i/o cycles. (Note that hard disks and
    video cards located on the ISA bus will also usually experience
    a performance increase from an increased ISA bus speed.)

<sect1>Setting the TCP Rx Window
<p>

    Once again, cards with small amounts of onboard RAM and
    relatively slow data paths between the card and the computer's
    memory run into trouble. The default TCP Rx
    window setting is 32kB, which means that a fast computer on
    the same subnet as you can dump 32k of data on you without
    stopping to see if you received any of it okay.

    Recent versions of the <tt/route/ command have the ability
    to set the size of this window on the fly. Usually it is only
    for the local net that this window must be reduced, as computers
    that are behind a couple of routers or gateways are `buffered'
    enough to not pose a problem. An example usage would be:

<code>
    route add <whatever> ... window <win_size>
</code>

    where <tt/win_size/ is the size of the window you wish to
    use (in bytes). An 8 bit 3c503 card on an ISA bus operating
    at a speed of 10MHz or less would work well with a window
    size of about 4kB. Too large a window will cause overruns
    and dropped packets, and a drastic reduction in ethernet
    throughput. You can check the operating status by doing
    a <tt>cat /proc/net/dev</tt> which will display any
    dropped or overrun conditions that occurred.


<sect1>Increasing NFS performance
<p>
    Some people have found that using 8 bit
    cards in NFS clients causes poorer than expected performance,
    when using 8kB (native Sun) NFS packet size.

    The possible reason for this could be due to the difference
    in on board buffer size between the 8 bit and the 16 bit cards.
    The maximum ethernet packet size is about 1500 bytes. Now that
    8kB NFS packet will arrive as about 6 back to back maximum size
    ethernet packets. Both the 8 and 16 bit cards have no problem
    Rx'ing back to back packets. The problem arises when the machine
    doesn't remove the packets from the cards buffer in time, and the
    buffer overflows. The fact that 8 bit cards take an extra ISA
    bus cycle per transfer doesn't help either. What you <em/can/ do
    if you have an 8 bit card is either set the NFS transfer
    size to 4kB, or try increasing the ISA bus speed in order to
    get the card's buffer cleared out faster.

<sect>Vendor/Manufacturer/Model Specific Information<label id="card-intro">
<p>

<sect1>3Com<label id="3com">
<p>

    If you are not sure what your card is, but you think it is a
    3Com card, you can probably figure it out from the assembly
    number. 3Com has a document `Identifying 3Com Adapters By
    Assembly Number' (ref 24500002) that would most likely clear
    things up. See
    <ref id="3com-tech" name="Technical Information from 3Com">
    for info on how to get documents from 3Com.

    Also note that 3Com has a FTP site with various goodies:
    <tt/ftp.3Com.com/ that you may want to check out.

    For those of you browsing this document by a WWW browser, you
    can try 3Com's WWW site

<sect2>3c501<label id="3c501">
<p>

    Status -- <em/Semi-Supported/

    Too brain-damaged to use. Available surplus from many
    places. Avoid it like the plague. Again, do not
    purchase this card, even as a joke. It's performance
    is horrible, and it breaks in many ways.

    Cameron L. Spitzer of 3Com said:
    ``I'm speaking only for myself here, of course, but I
    believe 3Com advises against installing a 3C501 in a
    new system, mostly for the same reasons Donald has
    discussed. You probably won't be happy with the
    3C501 in your Linux box. The data sheet is marked
    `(obsolete)' on 3Com's Developers' Order Form, and
    the board is not part of 3Com's program for sending
    free Technical Reference Manuals to people who need
    them. The decade-old things are nearly
    indestructible, but that's about all they've got
    going for them any more.''

    For those not yet convinced, the 3c501 can only do one
    thing at a time -- while you are removing one packet
    from the single-packet buffer it cannot receive
    another packet, nor can it receive a packet while
    loading a transmit packet. This was fine for a
    network between two 8088-based computers where
    processing each packet and replying took 10's of
    msecs, but modern networks send back-to-back
    packets for almost every transaction.

    AutoIRQ works, DMA isn't used, the autoprobe only
    looks at <tt/0x280/ and <tt/0x300/, and the debug level is set
    with the third boot-time argument.

    Once again, the use of a 3c501 is <em/strongly discouraged/!
    Even more so with a IP multicast kernel, as you will
    grind to a halt while listening to <em/all/ multicast
    packets. See the comments at the top of the source code
    for more details.

<sect2>3c503, 3c503/16<label id="3c503">
<p>

    Status -- <em/Supported/

    If you have a 3c503/16 you may be interested to know
    that as of 1.3.37 the driver has the facility to use
    the full 16kB RAM on your card. Previous versions treated
    the 16bit cards as 8bit cards, and only used half of
    the available RAM. This update also detects the newer
    3Com prefix found on newly manufactured cards mentioned
    below.

    Recently made 3c503/16 cards have a new
    base hardware address because 3Com ran out of
    numbers (they made too many cards!) The cards used
    to start with 02 60 8C and the newer ones use
    00 20 AF. Up to 1.3.37, the driver will only check
    for the old address, and skip over the newer cards.
    You can upgrade to a kernel newer than 1.3.37, or
    change the numbers in 3c503.c for older kernels.

    These shared-memory ethercards also have a
    programmed I/O mode that doesn't use the 8390
    facilities (their engineers found too many bugs!)
    It should be about the same speed as the same bus
    width WD80x3, but it is actually a bit slower.
    The 3c503 does not have ``EEPROM setup'',
    so a diagnostic/setup program
    isn't needed before running the card with Linux. The
    shared memory address of the 3c503 is set using jumpers
    that are shared with the boot PROM address. This is
    confusing to people familiar with other ISA cards,
    where you always leave the jumper set to ``disable''
    unless you have a boot PROM.

    The Linux 3c503 driver can also work with the 3c503
    in programmed-I/O mode, but this is slower and less
    reliable than shared memory mode. Also, programmed-I/O
    mode is not as well tested when updating the drivers.
    You still shouldn't use the programmed-I/O mode though,
    unless you need it for MS-DOS compatibility.

    The 3c503's IRQ line is set in software, with no hints
    from an EEPROM. Unlike the MS-DOS drivers, the
    Linux driver has capability to autoIRQ: it uses the
    first available IRQ line in {5,2/9,3,4}, selected each
    time the card is ifconfig'ed. (Older driver versions
    selected the IRQ at boot time.) The ioctl() call
    in `ifconfig' will return EAGAIN if no IRQ line is
    available at that time.

    Some common problems that people have with the 503
    are discussed in <ref id="3com-probs" name="Problems with...">.

    If you intend on using this driver as a loadable module
    you should probably see
    <ref id="modules" name="Using the Ethernet Drivers as Modules">
    and also
    <ref id="mod8390" name="8390 Based Cards as Modules"> for
    module specific information.



<sect2>3c505<label id="3c505">
<p>

    Status -- <em/Semi-Supported/

    This is a driver that was written by Craig Southeren
    <tt/[e-mail]/. These cards also
    use the i82586 chip.
    There are not that many of these cards about.
    It is included in the standard kernel, but it is classed as
    an alpha driver. See <ref id="alfa" name="Alpha Drivers">
    for important information on using alpha-test ethernet drivers
    with Linux.

    There is also the file
    <tt>/usr/src/linux/drivers/net/README.3c505</tt>
    that you should read if you are going to use one of these cards.
    It contains various options that you can enable/disable.
    Technical information is available in
    <ref id="i82586" name="Programming the Intel chips">.

<sect2>3c507<label id="3c507">
<p>
    Status -- <em/Semi-Supported/

    This card uses one of the Intel chips, and the
    development of the driver is closely related to
    the development of the Intel Ether Express driver.
    The driver is included in the standard kernel
    release, but as an alpha driver.

    See <ref id="alfa" name="Alpha Drivers"> for important
    information on using alpha-test ethernet drivers
    with Linux. Technical information is available in
    <ref id="i82586" name="Programming the Intel chips">.

<sect2>3c509 / 3c509B<label id="3c509">
<p>

    Status -- <em/Supported/

    Note: Donald has an updated driver for these cards
    that is not presently in the default kernel. It
    fixes harmless <tt/irq N for unknown device/
    messages at boot, has better handling for
    the 3c509B Plug and Pray cards, and recovers from
    certain adapter failure status values.

    It's fairly inexpensive and has
    excellent performance for a non-bus-master design.
    The drawbacks are that the original 3c509
    requires very low interrupt latency. The 3c509B
    shouldn't suffer from the same problem, due to
    having a larger buffer. (See below.)

    Note that the ISA card detection uses a different method
    than most cards. Basically, you ask the cards to respond
    by sending data to an ID_PORT (port <tt/0x100/). Note that
    if you have some other strange ISA card using an I/O range
    that includes the ID_PORT of the 3c509, it will probably
    not get detected. Note that you can change the ID_PORT to
    <tt/0x110/ or <tt/0x120/ or... in <tt/3c509.c/ if you have
    a conflicting ISA card, and the 3c509 will still be happy.
    Also note that this detection method means that
    a particular card will <em/always/ get detected first
    in a multiple ISA 3c509 configuration.
    The card with the lowest hardware ethernet address
    will <em/always/ end up being <tt/eth0/. This shouldn't matter
    to anyone, except for those people who want to assign
    a 6 byte hardware address to a particular interface.
    If you have multiple 3c509 cards, it is best to append
    <tt/ether=0,0,ethN/ commands without the i/o port specified
    (i.e. use i/o=zero) and allow the probe to sort out which
    card is first, otherwise it may not detect all your cards.

    The original 3c509 has a smaller packet buffer
    (4kB total, 2kB Rx, 2kB Tx), causing the driver to
    occasionally drop a packet if interrupts are masked for
    too long. To minimize this problem, you can try unmasking
    interrupts during IDE disk transfers (see <tt/hdparm(8)/) and/or
    increasing your ISA bus speed so IDE transfers finish sooner.

    (Note that the driver could
    be completely rewritten to use predictive interrupts,
    but performance re-writes of working drivers are low
    priority unless there is some particular incentive or
    need.)

    Cameron Spitzer writes:
    ``Beware that if you put a '509 in EISA addressing mode
    by mistake and save that in the EEPROM, you'll have
    to use an EISA machine or the infamous Test Via to
    get it back to normal, and it will conflict at IO
    location 0 which may hang your ISA machine.
    I believe this problem is corrected in the 3C509B
    version of the board.''

    The newer model 3c509B has 8kB on board, and the buffer
    can be split 4/4, 5/3 or 6/2 for Rx/Tx. This setting
    can also be stored on the EEPROM. This should alleviate the
    above problem with the original 3c509. At this point in
    time, the Linux driver is not aware of this, and treats
    the 3c509B as an older 3c509.

    3c509B users should use the supplied DOS
    utility to disable the <em/plug and play/ support, <em/and/
    to set the output media to what they require.
    Cameron adds:
    ``The 3C509B has 3Com's relocatable I/O port scheme,
    and Microsoft[tm] Plug-and-play ("PnP").  You can't use them
    both at the same time.  Some (broken, IMHO) BIOSes begin a PnP
    sequence by writing to the PnP address (0x279 ?), which causes
    PnP adapters like 3C509B to enter the PnP state, but then they
    (these funny BIOSes) never come back to finish the job.
    The 3C509Bs hang there in the middle of the PnP ID Sequence, where
    they have no idea you didn't mean it and you're going to use the 3Com
    ID sequence after all.  3C5X9CFG /PNPRST clears this hang.
    Disable PnP if your drivers (eg., Linux) don't use it.

    It was a marketing decision to turn PnP on as a factory default setting.
    If it caused you a hassle, or not, please take the time to say so when
    you mail in your warranty card.  The more info they have, the better
    decisions they can make.  Also, check with your motherboard supplier
    to see if you need a BIOS upgrade.''

    With regards to the media detection features, Cameron said:
    ``Autoselect is a feature of the commercial *drivers* for 3C509(B).
    AFAIK nobody ever claimed the Linux driver attempts it.
    When drivers/net/3c509.c recognizes my 3C509B at boot time, it says:
    <tt/eth0: 3c509 at 0x300 tag 1, 10baseT port, .../
    revealing that the card is configured for 10BASE-T.  It finds that out
    by reading the little EEPROM, which IMHO is the Right Way To Do It.''


<sect2>3c523<label id="3c523">
<p>

    Status -- <em/Not Supported/

    This MCA bus card uses the i82586, and now that people are
    actually running Linux on MCA machines, someone may wish
    to try and recycle parts of the 3c507 driver into a
    driver for this card.

<sect2>3c527<label id="3c527">
<p>

    Status -- <em/Not Supported/

    Yes, another MCA card. No, not too much interest in it.
    Better chances with the 3c523 or the 3c529.

<sect2>3c529<label id="3c529">
<p>

    Status -- <em/Not Supported/

    This card actually uses the same chipset as the 3c509.
    Donald actually put hooks into the 3c509 driver to check
    for MCA cards after probing for EISA cards, and before
    probing for ISA cards. But it hasn't evolved much further
    than that. Donald writes:

    ``I don't have access to a MCA machine (nor do I fully understand
    the probing code) so I never wrote the <tt/mca_adaptor_select_mode()/
    or <tt/mca_adaptor_id()/ routines. If you can find a way to get the
    adaptor I/O address that assigned at boot time, you can just
    hard-wire that in place of the commented-out probe. Be sure to
    keep the code that reads the IRQ, if_port, and ethernet address.''

<sect2>3c562
<p>

    Status -- <em/Supported/

    This PCMCIA card is the combination of a 3c589B ethernet card
    with a modem. The modem appears as a standard modem to the
    end user. The only difficulty is getting the two separate
    linux drivers to share one interrupt. There are a couple of new
    registers and some hardware interrupt sharing support.

    Thanks again to Cameron for getting a sample unit and documentation
    sent off to David Hinds. Look for support in David's next
    PCMCIA package release.


<sect2>3c579<label id="3c579">
<p>

    Status -- <em/Supported/

    The EISA version of the 509. The current EISA version
    uses the same 16 bit wide chip rather than a 32 bit
    interface, so the performance increase isn't stunning.
    The EISA probe code was added to 3c509.c for 0.99pl14.
    We would be interested in hearing progress reports
    from any 3c579 users. (Read the above 3c509
    section for info on the driver.)

    Cameron Spitzer writes:
    ``The 3C579 (Etherlink III EISA) should be configured
    as an EISA card. The IO Base Address (window 0
    register 6 bits 4:0) should be 1f, which selects EISA
    addressing mode. Logic outside the ASIC decodes the
    IO address s000, where s is the slot number. I don't
    think it was documented real well. Except for its IO
    Base Address, the '579 should behave EXACTLY like
    the'509 (EL3 ISA), and if it doesn't, I want to hear
    about it (at my work address).''


<sect2>3c589 / 3c589B<label id="3c589">
<p>

    Status -- <em/Semi-Supported/

    Many people have been using this PCMCIA card for quite some time
    now. Note that support for it is not (at present) included
    in the default kernel source tree. You will also need
    a supported PCMCIA controller chipset. There are drivers
    available on Donald's ftp site:

<verb>
    cesdis.gsfc.nasa.gov:/pub/linux/pcmcia/README.3c589
    cesdis.gsfc.nasa.gov:/pub/linux/pcmcia/3c589.c
    cesdis.gsfc.nasa.gov:/pub/linux/pcmcia/dbether.c
</verb>

    Or for those that are <em/net-surfing/ you can try:

    <url url="http://cesdis.gsfc.nasa.gov/linux/pcmcia.html"
        name="Don's PCMCIA Stuff">

    You will still need a PCMCIA socket enabler as well.

    See <ref id="pcmcia" name="PCMCIA Support"> for more
    info on PCMCIA chipsets, socket enablers, etc.

    The "B" in the name means the same here as it does for
    the 3c509 case.

<sect2>3c590 / 3c595<label id="vortex">
<p>

    Status -- <em/Supported/

    These ``Vortex'' cards are for PCI bus machines, with the '590
    being 10Mbps and the '595 being 3Com's 100Mbs offering.
    Also note that you can run the '595 as a '590 (i.e. in a 10Mbps mode).
    The driver will eventually appear in the 1.3.x kernel, but you
    can get the driver now from the following URL:

    <url url="http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html"
        name="Vortex">

    3Com has been offering an evaluation deal where you get 2 '595
    cards for a total price of $250. Check with your local
    3Com distributor to see if this is still available.

    A thanks to Cameron Spitzer and Terry Murphy of 3Com for
    sending cards and documentation to Donald so he could write
    the driver.

<sect2>3c597
<p>

    Status -- <em/Supported/

    This is the recently released EISA version of the 3c59x
    series of cards. The 3c597 (aka Demon) should work with
    the vortex driver discussed above.

<sect1>Accton<label id="accton">
<p>

<sect2>Accton MPX
<p>

    Status -- <em/Supported/

    Don't let the name fool you. This is still supposed to be a
    NE2000 compatible card. The MPX is supposed to stand for
    MultiPacket Accelerator, which, according to Accton, increases
    throughput substantially. But if you are already sending
    back-to-back packets, how can you get any faster...

<sect2>Accton EN1203, EN 1207, EtherDuo-PCI
<p>

    Status -- <em/Supported/

    This is another implementation of the DEC 21040 PCI chip.

    See <ref id="dec-21040" name="DEC 21040">
    for more information on these cards, and the present driver
    situation.


<sect2>Accton EN2212 PCMCIA Card
<p>

    Status -- <em/Semi-Supported/

    David Hinds has been working on a driver for this card, and
    you are best to check the latest release of his PCMCIA
    package to see what the present status is.

<sect1>Allied Telesyn/Telesis<label id="allied-telesis">
<p>

<sect2>AT1500<label id="at-1500">
<p>

    Status --<em/Supported/

    These are a series of low-cost ethercards using the 79C960 version
    of the AMD LANCE. These are bus-master cards, and hence one of
    the faster ISA bus ethercards available.

    DMA selection and chip numbering information can be found in
    <ref id="lance" name="AMD LANCE">.

    More technical information on AMD LANCE based Ethernet cards
    can be found in <ref id="amd-notes" name="Notes on AMD...">.

<sect2>AT1700<label id="at1700">
<p>

    Status -- <em/Supported/

    The Allied Telesis AT1700 series ethercards are based
    on the Fujitsu MB86965. This chip uses a programmed
    I/O interface, and a pair of fixed-size transmit
    buffers. This allows small groups of packets to
    be sent back-to-back, with a short pause while
    switching buffers.

    A unique feature is the ability to drive 150ohm STP
    (Shielded Twisted Pair) cable commonly installed for
    Token Ring, in addition to 10baseT 100ohm UTP
    (unshielded twisted pair).

    The Fujitsu chip used on the AT1700 has a design flaw:
    it can only be fully reset by doing a power cycle of the machine.
    Pressing the reset button doesn't reset the bus interface. This
    wouldn't be so bad, except that it can only be reliably detected
    when it has been freshly reset. The solution/work-around is to
    power-cycle the machine if the kernel has a problem detecting
    the AT1700.

    Some production runs of the AT1700 had another problem:
    they are permanently wired to DMA channel 5.  This is
    undocumented, there are no jumpers to disable the "feature",
    and no driver dares use the DMA capability because of
    compatibility problems. No device driver will be
    written using DMA if installing a second card into
    the machine breaks both, and the only way to disable
    the DMA is with a knife.

    The at1700 driver is included in the standard
    kernel source tree.

<sect2>AT2450<label id="at2450">
<p>

    Status -- <em/Supported/

    This is the PCI version of the AT1500, and it doesn't suffer
    from the problems that the Boca 79c970 PCI card does.
    Allied Telsyn was still `beta testing' the card in
    early/mid 1995, so it may not have spread to various
    retailers yet (but it doesn't hurt to ask.)

    DMA selection and chip numbering information can be found in
    <ref id="lance" name="AMD LANCE">.

    More technical information on AMD LANCE based Ethernet cards
    can be found in <ref id="amd-notes" name="Notes on AMD...">.

<sect1>AMD / Advanced Micro Devices<label id="amd">
<p>

<sect2>AMD LANCE (7990, 79C960, PCnet-ISA)<label id="lance">
<p>

    Status -- <em/Supported/

    There really is no AMD ethernet card. You are probably reading this
    because the only markings you could find on your card said AMD
    and the above number. The 7990 is the original `LANCE' chip,
    but most stuff (including this document) refer to all these
    similar chips as `LANCE' chips. (...incorrectly, I might add.)

    These above numbers refer to chips from AMD
    that are the heart of many ethernet cards.
    For example, the Allied Telesis AT1500 (see
    <ref id="at-1500" name="AT1500">) the NE1500/2100 (see
    <ref id="ne1500" name="NE1500">) and the Boca-VLB/PCI cards (see
    <ref id="boca-pci" name="Boca-VLB/PCI">)

    The 79C960 (a.k.a. PCnet-ISA) contains enhancements and bug fixes
    over the original 7990 LANCE design.

    One common problem people have is the `busmaster arbitration
    failure' message. This is printed out when the LANCE driver
    can't get access to the bus after a reasonable amount of time
    has elapsed (50us). This usually indicates that the motherboard
    implementation of bus-mastering DMA is broken, or some other device
    is hogging the bus, or there is a DMA channel conflict. If your BIOS
    setup has the `GAT option' (for Guaranteed Access Time) then try
    toggling/altering that setting to see if it helps.

    Chances are that the existing LANCE driver will work
    with all AMD LANCE based cards. (except perhaps some of the
    original 7990 designs with shared memory.)
    This driver should also work with NE1500 and NE2100 clones.

    For the ISA bus master mode all structures used
    directly by the LANCE, the initialization block,
    Rx and Tx rings, and data buffers, must be accessible
    from the ISA bus, i.e. in the lower 16M of real memory.
    If more than 16MB of memory is installed, low-memory `bounce-buffers'
    are used when needed.

    The DMA channel can be set with the low bits
    of the otherwise-unused dev->mem_start value (a.k.a. PARAM_1).
    (see <ref id="ether" name="PARAM_1">)
    If unset it is probed for by enabling each free DMA channel
    in turn and checking if initialization succeeds.

    The HP-J2405A board is an exception: with this board it's easy
    to read the EEPROM-set values for the IRQ, and DMA.

    See <ref id="amd-notes" name="Notes on AMD...">
    for more info on these chips.

<sect2>AMD 79C961 (PCnet-ISA+)
<p>

    Status -- <em/Supported/

    This is the PCnet-ISA+  -- an enhanced version of the 79C960.
    It has support for jumper-less configuration and Plug and
    Play.  See the info in the above section.

<sect2>AMD 79C965 (PCnet-32)<label id="pcnet-32">
<p>

    Status -- <em/Supported/

    This is the PCnet-32 -- a 32 bit bus-master version of the
    original LANCE chip for VL-bus and local bus systems.
    Minor cleanups were added to the original lance driver
    around v1.1.50 to support these 32 bit versions of the LANCE
    chip.  The main problem was that the
    current versions of the '965 and '970 chips have a minor bug.
    They clear the Rx buffer length field in the Rx ring when they
    are explicitly documented not to. Again, see the above info.

<sect2>AMD 79C970 (PCnet-PCI)<label id="pcnet-pci">
<p>

    Status -- <em/Supported/

    This is the PCnet-PCI -- similar to the PCnet-32, but designed
    for PCI bus based systems. Again, see the above info.
    Donald has modified
    the LANCE driver to use the PCI BIOS structure
    that was introduced by Drew Eckhardt for the PCI-NCR SCSI
    driver. This means that you need to build a kernel with
    PCI BIOS support enabled.

    Note that the Boca implementation of the 79C970 fails on
    fast Pentium machines. This is a hardware problem, as it
    affects DOS users as well. See the Boca section for more
    details.

<sect2>AMD 79C974 (PCnet-SCSI)
<p>

    Status -- <em/Supported/

    This is the PCnet-SCSI --  which is basically treated like
    a '970 from an Ethernet point of view. A minor '974 specific
    fix was added to the 1.1.8x kernels, so get a 1.1.90 or newer
    kernel. Also see the above info. Don't ask if the
    SCSI half of the chip is supported -- this is the
    <em/Ethernet-Howto/, not the SCSI-Howto.

<sect1>Ansel Communications<label id="ansel">
<p>

<sect2>AC3200 EISA
<p>

    Status -- <em/Semi-Supported/

    This driver is included in the present kernel as an
    alpha test driver. It is based on the common NS8390
    chip used in the ne2000 and wd80x3 cards.
    Please see <ref id="alfa" name="Alpha Drivers"> in
    this document for important information regarding
    alpha drivers.
    If you use it, let Donald know how things work out,
    as not too many people have this card and feedback
    has been low, even though the driver has been in the
    kernel since v1.1.25.

    If you intend on using this driver as a loadable module
    you should probably see
    <ref id="modules" name="Using the Ethernet Drivers as Modules">
    and also
    <ref id="mod8390" name="8390 Based Cards as Modules"> for
    module specific information.

<sect1>Apricot
<p>

<sect2>Apricot Xen-II On Board Ethernet
<p>

    Status -- <em/Supported/

    This on board ethernet uses an i82596 bus-master chip.
    It can only be at i/o address <tt/0x300/. The author of this
    driver is Mark Evans. By looking at the driver source,
    it appears that the IRQ is hardwired to 10.

    Earlier versions of the driver had a tendency to think
    that anything living at <tt/0x300/ was an apricot NIC.
    Since then the hardware address is checked to avoid these
    false detections.

<sect1>Arcnet<label id="arcnet">
<p>

    Status -- <em/Supported/

    With the very low cost and better performance of ethernet,
    chances are that most places will be giving away their Arcnet
    hardware for free, resulting in a lot of home systems with Arcnet.

    An advantage of Arcnet is that all of the cards have identical
    interfaces, so one driver will work for everyone. It also has
    built in error handling so that it supposedly never loses a packet.
    (Great for UDP traffic!)

    Avery Pennarun's arcnet driver has been in the
    default kernel sources since 1.1.80. The arcnet driver
    uses `arc0' as its name instead of the usual `eth0' for
    ethernet devices.
    Bug reports and success stories can be mailed to:

    <tt>[e-mail]</tt>

    There are information files contained in the standard kernel for
    setting jumpers and general hints.

    Supposedly the driver also works with the 100Mbs ARCnet cards
    as well!

<sect1>AT&T
<p>

    Note that AT&T's StarLAN is an orphaned technology, like
    SynOptics LattisNet, and can't be used in a standard 10Base-T
    environment, without a hub that `speaks' both.

<sect2>AT&T T7231 (LanPACER+)
<p>

    Status -- <em/Not Supported/

    These StarLAN cards use an interface similar to the i82586
    chip. At one point, Matthijs Melchior
    (<tt/[e-mail]/) was playing with the 3c507
    driver, and almost had something useable working. Haven't
    heard much since that.

<sect1>AT-Lan-Tec / RealTek
<p>

<sect2>AT-Lan-Tec / RealTek Pocket adaptor<label id="aep-100">
<p>

    Status -- <em/Supported/

    This is a generic, low-cost OEM pocket adaptor being sold by
    AT-Lan-Tec, and (likely) a number of other suppliers. A
    driver for it is included in the standard kernel.
    Note that there is substantial information contained in the
    driver source file `atp.c'.
    BTW, the adaptor (AEP-100L) has both 10baseT and BNC connections!
    You can reach AT-Lan-Tec at 1-301-948-7070. Ask for the model
    that works with Linux, or ask for tech support.

    In the Netherlands a compatible adaptor is sold under the name SHI-TEC
    PE-NET/CT, and sells for about $125. The vendor was Megasellers.
    They state that they do not sell to private persons, but
    this doesn't appear to be strictly adhered to.
    They are: Megasellers, Vianen, The Netherlands. They always
    advertise in Dutch computer magazines.  Note that the
    newer model EPP-NET/CT appears to be significantly different
    than the PE-NET/CT, and will not work with the present driver.
    Hopefully someone will come up with the programming information
    and this will be fixed up.

    In Germany, a similar
    adaptor comes as a no-brand-name product. Prolan 890b, no
    brand on the casing, only a roman II. Resellers can get a price
    of about $130, including a small wall transformer for the power.

    The adaptor is `normal size' for the product class, about 57mm wide,
    22mm high tapering to 15mm high at the DB25 connector, and 105mm long
    (120mm including the BNC socket). It's switchable between the RJ45
    and BNC jacks with a small slide switch positioned between the two:
    a very intuitive design.

    Donald performed some power draw measurements, and determined
    that the average current draw was only about 100mA @ 5V.
    This power draw is low enough
    that you could buy or build a cable to take the 5V directly from the
    keyboard/mouse port available on many laptops. (Bonus points here
    for using a standardized power connector instead of a
    proprietary one.)

    Note that the device name that you pass to <tt/ifconfig/
    is <em/not/ <tt/eth0/ but <tt/atp0/ for this device.


<sect1>Boca Research<label id="boca">
<p>

    Yes, they make more than just multi-port serial cards.  :-)

<sect2>Boca BEN (PCI, VLB)<label id="boca-pci">
<p>

    Status -- <em/Supported/

    These cards are based on AMD's PCnet chips, used in the AT1500 and
    the like. You can pick up a combo (10BaseT and 10Base2) PCI
    card for under $70 at the moment.

    Perspective buyers should be warned that many users have had
    endless problems with these cards. Owners of fast Pentium
    systems have been especially hit. Note that this is not a driver
    problem, as it hits DOS/Win/NT users as well.
    Boca's technical support number is [phone], and you
    can also reach them at <tt/[e-mail]/.

    Donald did a comparitive test with the above Boca PCI card and
    a similar Allied Telsyn PCnet/PCI implementation, which showed
    that the problem lies in Boca's implementation of the PCnet/PCI
    chip. These test results can be accessed on Don's www server.

    <url url="http://cesdis.gsfc.nasa.gov/linux/" name="Linux at CESDIS">

    Also, Dave Platt has compared the recommended implementation given
    on the AMD data sheet with the Boca implementation, and has
    determined that Boca has left out a substantial number of
    important filtering capacitors. At the risk of being verbose,
    here is a quote from Dave, which will allow you to assess
    if the problem has been addressed yet, given a card for
    visual inspection.

    ``I just reviewed Appendix B in the 79c970 data sheet.  It
    recommends a _minimum_ of 8 high-frequency bypassing caps
    (.1 uF multilayer ceramic) around the chip, to keep ground
    and power bounce from causing unreliable operation.
    Looking at the card I have here, I see a total of 5 such
    caps on the entire card - only a couple of them are close
    enough to the chip to do a decent job of high-frequency
    bypassing.

    If you hold the card with chips facing up, and the PCI pins
    pointed towards you, the lower-left corner of the '970 has
    the positioning dimple.  The upper-left corner of the chip
    is the `analog corner', where the analog power and ground
    pins are.  Appendix B calls these `the most cricical pins
    in the layout of a PCnet-PCI card'.  There are 4 analog
    power pins, and 2 analog ground pins... all of them are
    supposed to be connected, and properly bypassed.

    On the left side of the chip, the uppermost pin is AVSS1
    (analog ground 1).  On the top edge, the fourth pin from
    the left is AVDD3 (analog power 3).  The Appendix
    specifically recommends having a .1 uF bypass cap tied
    directly to these two pins (<em/not/ going through the common
    ground plane).  On the card I have, there is no such cap -
    the nearest bypass cap is about half-an-inch away over by
    the crystal, and is tied to the ground plane.

    The Appendix also specifically recommends `low-frequency
    bulk capacitors' (by which I assume that they mean multi-uF
    tantalum or aluminum electrolytics) as well as
    high-frequency bypass caps.  I see only two bulk capacitors
    (one 10 uF in the AVSS2/AVDD2 filter circuit, and one 4.7
    uF up above the chip which also appears to be part of a
    filter circuit).  The Appendix recommends `at least one
    low-frequency bulk (e.g. 22 uF) bypass capactor...
    connected directly to the power and ground planes.'  There
    is no capacitor on the board which matches this
    description.

    It appears that Boca ignored _several_ of AMD's
    recommendations, regarding the number, size, placement, and
    wiring of the power supply bypass capacitors.''

    (Thanks Dave.) Boca is offering a `warranty repair' for
    affected owners, which involves adding one of the missing
    capacitors, but it appears that this fix doesn't work 100
    percent for most people, although it helps some.

    If you are <em/still/ thinking of buying one of these cards, then
    at least try and get a 7 day unconditional return policy,
    so that if it doesn't work properly in your system, you
    can return it.

    More general information on the AMD chips can be found in
    <ref id="lance" name="AMD LANCE">.

    More technical information on AMD LANCE based Ethernet cards
    can be found in <ref id="amd-notes" name="Notes on AMD...">.

<sect1>Cabletron<label id="ctron">
<p>

    Donald writes:
    `Yes, another one of these companies that won't release its
    programming information. They waited for months before actually
    confirming that all their information was proprietary, deliberately
    wasting my time. Avoid their cards like the plague if you can.
    Also note that some people have phoned Cabletron, and have been
    told things like `a D. Becker is working on a driver
    for linux' -- making it sound like I work for them. This is
    NOT the case.'

    If you feel like asking them why they don't want to release their
    low level programming info so that people can use their cards, write
    to [e-mail].
    Tell them that you are using Linux, and are disappointed that they
    don't support open systems. And no, the usual driver development
    kit they supply is useless. It is just a DOS object file that
    you are supposed to link against. Which you aren't allowed to
    even reverse engineer.


<sect2>E10**, E10**-x, E20**, E20**-x<label id="e10xx">
<p>

    Status -- <em/Semi-Supported/

    These are NEx000 almost-clones that are reported to
    work with the standard NEx000 drivers, thanks to a
    ctron-specific check during the probe. If there are
    any problems, they are unlikely to be fixed, as the
    programming information is unavailable.

<sect2>E2100<label id="e2100">
<p>

    Status -- <em/Semi-Supported/

    Again, there is not much one can do when the
    programming information is proprietary.
    The E2100 is a poor design. Whenever it maps its
    shared memory in during a packet transfer, it
    maps it into the <em/whole 128K region!/ That means you
    <bf/can't/ safely use another interrupt-driven shared
    memory device in that region, including another E2100.
    It will work most of the time, but every once in
    a while it will bite you. (Yes, this problem can
    be avoided by turning off interrupts while
    transferring packets, but that will almost certainly
    lose clock ticks.) Also, if you mis-program the board,
    or halt the machine at just the wrong moment, even
    the reset button won't bring it back. You will <em/have/
    to turn it off and <em/leave/ it off for about 30 seconds.

    Media selection is automatic, but you can override this
    with the low bits of the dev->mem_end parameter.
    See <ref id="ether" name="PARAM_2">. Module users
    can specify an <tt/xcvr=N/ value on the <tt/insmod/
    command line to do the same.

    Also, don't confuse the E2100 for a NE2100 clone.
    The E2100 is a shared memory NatSemi DP8390 design,
    roughly similar to a brain-damaged WD8013, whereas
    the NE2100 (and NE1500) use a bus-mastering AMD
    LANCE design.

    There is an E2100 driver included in the standard kernel.
    However, seeing as programming info isn't available,
    don't expect bug-fixes. Don't use one
    unless you are already stuck with the card.

    If you intend on using this driver as a loadable module
    you should probably see
    <ref id="modules" name="Using the Ethernet Drivers as Modules">
    and also
    <ref id="mod8390" name="8390 Based Cards as Modules"> for
    module specific information.

<sect1>Cogent
<p>

    Here is where and how to reach them:


<verb>
    Cogent Data Technologies, Inc.
    175 West Street, P.O. Box 926
    Friday Harbour, WA 98250, USA.

    Cogent Sales
    15375 S.E. 30th Place, Suite 310
    Bellevue, WA 98007, USA.

    Technical Support:
    Phone [phone] between 8am and 5pm PST
    Fax [phone]
    Compuserve GO COGENT
    Bulletin Board Service [phone]
    Internet: [e-mail]
</verb>

<sect2>Cogent eMASTER+, EM100, EM400, EM960, EM964
<p>

    Status -- <em/Supported/

    These are yet another DEC 21040 implementation that should
    work fine with the standard 21040 driver.

    The EM400 and the EM964 are four port cards using a
    DEC 21050 bridge and 4 21040 chips.

    See <ref id="dec-21040" name="DEC 21040">
    for more information on these cards, and the present driver
    situation.

<sect1>D-Link<label id="d-link">
<p>

    Some people have had difficulty in finding vendors that
    carry D-link stuff. This should help.

<verb>
        [phone]    in the US
        [phone]    in the UK
        [phone]    in Canada
        [phone]    in Taiwan
</verb>

<sect2>DE-100, DE-200, DE-220-T<label id="de-100">
<p>

    Status -- <em/Supported/

    The manual says that it is 100 &percnt compatible with the
    NE2000. This is not true. You should call them and
    tell them you are using their card with Linux, and they
    should correct their documentation. Some pre-0.99pl12
    driver versions may have trouble recognizing the DE2**
    series as 16 bit cards, and these cards are the most
    widely reported as having the spurious transfer address
    mismatch errors. Note that there are cards from
    Digital (DEC) that are also named DE100 and DE200,
    but the similarity stops there.

<sect2>DE-520<label id="de-520">
<p>

    Status -- <em/Supported/

    This is a PCI card using the PCI version of AMD's LANCE chip.
    DMA selection and chip numbering information can be found in
    <ref id="lance" name="AMD LANCE">.

    More technical information on AMD LANCE based Ethernet cards
    can be found in <ref id="amd-notes" name="Notes on AMD...">.

<sect2>DE-530<label id="de-530">
<p>

    Status -- <em/Supported/

    This is a generic DEC 21040 PCI chip implementation,
    and works with the generic 21040 driver.

    See <ref id="dec-21040" name="DEC 21040">
    for more information on these cards, and the present driver
    situation.

<sect2>DE-600<label id="de-600">
<p>

    Status -- <em/Supported/

    Laptop users and other folk who might want a quick
    way to put their computer onto the ethernet may want
    to use this. The driver is included with the default
    kernel source tree.
    Bjorn Ekwall <tt/[e-mail]/ wrote the driver.
    Expect about 180kb/s transfer speed from this via the
    parallel port. You should read the README.DLINK
    file in the kernel source tree.

    Note that the device name that you pass to <tt/ifconfig/
    is <em/now/ <tt/eth0/ and not the previously
    used <tt/dl0/.

    If your parallel port is <em/not/ at the standard <tt/0x378/
    then you will have to recompile. Bjorn writes:
    ``Since the DE-620 driver tries to sqeeze the last microsecond
    from the loops, I made the irq and port address constants instead
    of variables. This makes for a usable speed, but it also means
    that you can't change these assignements from e.g. lilo;
    you _have_ to recompile...'' Also note that some laptops
    implement the on-board parallel port at <tt/0x3bc/ which
    is where the parallel ports on monochrome cards were/are.

<sect2>DE-620<label id="de-620">
<p>

    Status -- <em/Supported/

    Same as the DE-600, only with two output formats.
    Bjorn has written a driver for this model,
    for kernel versions 1.1 and above. See the above information
    on the DE-600.

<sect2>DE-650<label id="de-650">
<p>

    Status -- <em/Semi-Supported/

    Some people have been using this PCMCIA card for
    some time now with their notebooks. It is a basic
    8390 design, much like a NE2000. The LinkSys PCMCIA
    card and the IC-Card Ethernet (available from
    Midwest Micro) are supposedly DE-650 clones as well.
    Note that at present, this driver is
    <em/not/ part of the standard kernel, and so you will
    have to do some patching.

    See <ref id="pcmcia" name="PCMCIA Support"> in this document,
    and if you can, have a look at:

    <url url="http://cesdis.gsfc.nasa.gov/linux/pcmcia.html"
        name="Don's PCMCIA Stuff">

<sect1>DFI<label id="dfi">
<p>

<sect2>DFINET-300 and DFINET-400<label id="dfi-300">
<p>

    Status -- <em/Supported/

    These cards are now detected (as of 0.99pl15) thanks to
    Eberhard Moenkeberg <tt/[e-mail]/ who noted that
    they use `DFI' in the first 3 bytes of the prom, instead
    of using <tt/0x57/ in bytes 14 and 15, which is what all the
    NE1000 and NE2000 cards use. (The 300 is an 8 bit
    pseudo NE1000 clone, and the 400 is a pseudo NE2000 clone.)


<sect1>Digital / DEC<label id="dec">
<p>

<sect2>DEPCA, DE100/1, DE200/1/2, DE210, DE422<label id="dec-200">
<p>

    Status -- <em/Supported/

    As of linux v1.0, there is a driver included as standard
    for these cards. It was written by David C. Davies.
    There is documentation included in the source file
    `depca.c', which includes info on how to use more than
    one of these cards in a machine. Note that the DE422 is
    an EISA card. These cards are all based on the AMD LANCE chip.
    See <ref id="lance" name="AMD LANCE"> for more info.
    A maximum of two of the ISA cards can be used, because they
    can only be set for <tt/0x300/ and <tt/0x200/ base I/O address.
    If you are intending to do this, please read the notes in
    the driver source file <tt/depca.c/ in the standard kernel
    source tree.

    This driver will also work on Alpha CPU based machines, and
    there are various ioctl()s that the user can play with.

<sect2>Digital EtherWorks 3 (DE203, DE204, DE205)<label id="dec-ewrk3">
<p>

    Status -- <em/Supported/

    Included into kernels v1.1.62 and above is this driver,
    also by David C. Davies of DEC. These cards use a proprietary
    chip from DEC, as opposed to the LANCE chip used in the
    earlier cards like the DE200. These cards support both shared
    memory or programmed I/O, although you take about a 50&percnt
    performance hit if you use PIO mode. The shared memory size can
    be set to 2kB, 32kB or 64kB, but only 2 and 32 have been tested
    with this driver. David says that the performance is virtually
    identical between the 2kB and 32kB mode. There is more information
    (including using the driver as a loadable module) at the top
    of the driver file <tt/ewrk3.c/ and also in <tt/README.ewrk3/.
    Both of these files come with the standard kernel distribution.

    The standard driver has a number
    of interesting ioctl() calls that can be used to get or clear
    packet statistics, read/write the EEPROM, change the
    hardware address, and the like. Hackers can see the source
    code for more info on that one.

    David has also written a configuration utility for this
    card (along the lines of the DOS program <tt/NICSETUP.EXE/)
    along with other tools. These can be found on
    <tt/sunsite.unc.edu/ in the directory
    <tt>/pub/Linux/system/Network/management</tt> -- look for the
    file <tt/ewrk3tools-X.XX.tar.gz/.

    The next release of this driver (v0.40) will have Alpha CPU
    support like depca.c does and is available from David now
    if you require it.

<sect2>DE425 (EISA), DE434, DE435, DE500<label id="dec-eisa">
<p>

    Status -- <em/Supported/

    These cards are based on the 21040 chip mentioned below.
    Included into kernels v1.1.86 and above is this driver,
    also by David C. Davies of DEC. It sure is nice to have
    support from someone on the inside  ;-)
    The DE500 uses the newer 21140 chip to provide 10/100Mbs
    ethernet connections.
    Have a read of the 21040 section below for extra info.

    Note that as of 1.1.91, David has added a compile time option that
    will allow non-DEC cards to work with
    this driver. Have a look at <tt/README.de4x5/ for details.

    All the Digital cards will autoprobe for their media (except,
    temporarily, the DE500 due to a patent issue).

    This driver is also ALPHA CPU ready and supports being loaded
    as a module.  Users can access the driver internals through
    ioctl() calls - see the 'ewrk3' tools and the de4x5.c sources
    for information about how to do this.

<sect2>DEC 21040, 21041, 21140, Tulip<label id="dec-21040">
<p>

    Status -- <em/Supported/

    The DEC 21040 is a bus-mastering single chip ethernet solution
    from Digital, similar to AMD's PCnet chip. The 21040 is
    specifically designed for the PCI bus architecture.
    SMC's new EtherPower PCI card uses this chip.

    You have a choice of <em/two/ drivers for cards based on this
    chip. There is the DE425 driver discussed above, and the
    generic 21040 driver that Donald has written.

    The updated 21041 chip is also found in place of the 21040
    on newly produced SMC EtherPower cards.
    The 21140 is for supporting 100Base-? and
    works with the Linux drivers for the 21040 chip.
    To use David's <tt/de4x5/ driver with non-DEC cards, have a
    look at <tt/README.de4x5/ for details.

    Donald has used SMC EtherPower-10/100 cards to develop
    the `tulip' driver. Note that the driver that is
    in the standard kernel tree at the moment is not the most
    up to date version. If you intend on using this driver,
    you should get the newest version from Donald's ftp/WWW
    site.

    <url url="http://cesdis.gsfc.nasa.gov/linux/drivers/tulip.html"
        name="Tulip Driver">

    The above URL also contains a (non-exhaustive) list of
    various cards/vendors that use the 21040 chip.

    Also note that the tulip driver is still considered an <em/alpha/
    driver (see <ref id="alfa" name="Alpha Drivers">) at the
    moment, and should be treated as such. To use it, you
    will have to edit <tt>arch/i386/config.in</tt> and
    uncomment the line for <tt/CONFIG_DEC_ELCP/ support.

<sect1>Farallon
<p>
    Farallon sells EtherWave adaptors and transceivers. This device
        allows multiple 10baseT devices to be daisy-chained.

<sect2>Farallon Etherwave
<p>

    Status -- <em/Supported/

    This is reported to be a 3c509 clone that includes the
    EtherWave transceiver. People have used these successfully
    with Linux and the present 3c509 driver. They are too expensive
    for general use, but are a great option for special cases.  Hublet
    prices start at $125, and Etherwave
    adds $75-$100 to the price of the board -- worth
    it if you have pulled one wire too few, but not if you are two
    network drops short.

<sect1>Hewlett Packard<label id="hp">
<p>

    The 272** cards use programmed I/O, similar to the NE*000 boards,
    but the data transfer port can be `turned off' when you aren't
    accessing it, avoiding problems with autoprobing drivers.

    Thanks to Glenn Talbott for helping clean up the confusion in this
    section regarding the version numbers of the HP hardware.

<sect2>27245A<label id="hp-27245a">
<p>

    Status -- <em/Supported/

    8 Bit 8390 based 10BaseT, not recommended for all the
    8 bit reasons. It was re-designed a couple years
    ago to be highly integrated which caused some
    changes in initialization timing which only
    affected testing programs, not LAN drivers. (The
    new card is not `ready' as soon after switching
    into and out of loopback mode.)

    If you intend on using this driver as a loadable module
    you should probably see
    <ref id="modules" name="Using the Ethernet Drivers as Modules">
    and also
    <ref id="mod8390" name="8390 Based Cards as Modules"> for
    module specific information.

<sect2>HP PC Lan+ (27247, 27252A)
<p>

    Status -- <em/Supported/

    The HP PC Lan+ is different to the standard HP PC Lan
    card. This driver was added to the list of drivers in the standard
    kernel during the v1.1.x development cycle. It can be
    operated in either a PIO mode like a ne2000, or a shared
    memory mode like a wd8013.

    The 47B is a 16 Bit 8390 based 10BaseT w/AUI, and
    the 52A is a 16 Bit 8390 based ThinLAN w/AUI.
    These cards have 32K onboard RAM for Tx/Rx packet buffering
    instead of the usual 16KB, and they both offer LAN
    connector autosense.

    If you intend on using this driver as a loadable module
    you should probably see
    <ref id="modules" name="Using the Ethernet Drivers as Modules">
    and also
    <ref id="mod8390" name="8390 Based Cards as Modules"> for
    module specific information.

<sect2>HP-J2405A
<p>

    Status -- <em/Supported/

    These are lower priced, and slightly faster than the
    27247/27252A, but are missing some features, such
    as AUI, ThinLAN connectivity, and boot PROM socket.
    This is a fairly generic LANCE design, but a minor
    design decision makes it incompatible with a generic
    `NE2100' driver. Special support for it (including
    reading the DMA channel from the board) is included
    thanks to information provided by HP's Glenn
    Talbott.

    More technical information on LANCE based cards can be found in
    <ref id="amd-notes" name="Notes on AMD...">

<sect2>HP-Vectra On Board Ethernet
<p>

    Status -- <em/Supported/

    The HP-Vectra has an AMD PCnet chip on the motherboard.
    Earlier kernel versions would detect it as the HP-J2405A
    but that would fail, as the Vectra doesn't report the
    IRQ and DMA channel like the J2405A.
    Get a kernel newer than v1.1.53 to avoid this
    problem.

    DMA selection and chip numbering information can be found in
    <ref id="lance" name="AMD LANCE">.

    More technical information on LANCE based cards can be found in
    <ref id="amd-notes" name="Notes on AMD...">

<sect2>HP 10/100 VG Any Lan Cards (27248B, J2573, J2577, J2585)
<p>

    Status -- <em/Supported/

    As of early 1.3.x kernels, this driver was made available
    by Jaroslav Kysela, ([e-mail]). Due to the newness of
    the driver and the relatively small number of VG cards in
    use, feedback on this driver has been low.

    Donald has also written a driver for these cards. Unlike the
    above, it is not presently in the standard kernel source tree.
    Check out the following URL for more information on Donald's
    100VG work.

    <url url="http://cesdis.gsfc.nasa.gov/linux/drivers/100vg.html"
        name="Donald's 100VG Page">


<sect1>IBM / International Business Machines<label id="ibm">
<p>

<sect2>IBM Thinkpad 300<label id="thinkpad-300">
<p>

    Status -- <em/Supported/

    This is compatible with the Intel based Zenith Z-note.
    See <ref id="z-note" name="Z-note"> for more info.

    Supposedly this site has a comprehensive database of
    useful stuff for newer versions of the Thinkpad. I haven't
    checked it out myself yet.

    <url url="http://peipa.essex.ac.uk/html/linux-thinkpad.html"
        Name="Thinkpad-info">

    For those without a WWW browser handy, try
    <tt>peipa.essex.ac.uk:/pub/tp750/</tt>

<sect2>IBM Credit Card Adaptor for Ethernet
<p>

    Status -- <em/Semi-Supported/

    People have been using this PCMCIA card with Linux as well.
    Similar points apply, those being that you need a supported
    PCMCIA chipset on your notebook, and that you will have to
    patch the PCMCIA support into the standard kernel.

    See <ref id="pcmcia" name="PCMCIA Support"> in this document,
    and if you can, have a look at:

    <url url="http://cesdis.gsfc.nasa.gov/linux/pcmcia.html"
        name="Don's PCMCIA Stuff">


<sect2>IBM Token Ring
<p>

    Status -- <em/Semi-Supported/

    To support token ring
    requires more than only writing a device driver, it also requires
    writing the source routing routines for token ring. It is the
    source routing that would be the most time comsuming to write.

    Peter De Schrijver has been spending some time on Token Ring
    lately. and has worked with IBM ISA and
    MCA token ring cards.

    The present token ring code has been included into the first
    of the 1.3.x series kernels.

    Peter says that it was originally tested on an MCA 16/4 Megabit Token
    Ring board, but it should work with other Tropic based boards.

<sect1>ICL Ethernet Cards
<p>

<sect2>ICL EtherTeam 16i/32
<p>
    Status -- <em/Supported/

    Mika Kuoppala ([e-mail]) wrote this driver, and
    it was included into early 1.3.4x kernels. It uses the
    Fujitsu MB86965 chip that is also used on the at1700 cards.

<sect1>Intel Ethernet Cards<label id="intel">
<p>

<sect2>Ether Express
<p>

    Status -- <em/Semi-Supported/

    This card uses the intel i82586. (Surprise, huh?)
    The driver is in the standard release of the
    kernel, as an alpha driver. See
    <ref id="alfa" name="Alpha Drivers"> for important
    information on using alpha-test ethernet drivers
    with Linux.

    The reason is that the driver works well with slow machines,
    but the i82586 occasionally hangs from the packet buffer
    contention that a fast machine can cause.
    One reported hack/fix is to change all of the outw()
    calls to outw_p(). Also, the driver is missing promiscuous
    and multicast modes. (See <ref id="promisc" name="Multicast and...">)

    There is also the standard way of using the chip (read slower)
    that is described in the chip manual, and used in
    other i82586 drivers, but this would require a re-write
    of the entire driver.

    There is some technical information available on
    the i82586 in <ref id="i82586" name="Programming the Intel Chips">
    and also in the source code for the driver `eexpress.c'. Don't
    be afraid to read it. ;-)

<sect2>Ether Express PRO/10
<p>

    Status -- <em/Supported/

    Bao Chau Ha has written a driver for these cards that has been
    included into early 1.3.x kernels. It may also work with some of
    the Compaq built-in ethernet systems that are based on the
    i82595 chip.

<sect2>Ether Express PRO/100
<p>

    Status -- <em/Not Supported/

    These PCI and EISA cards use a different chip that the PRO/10
    cards do. Bao is currently trying to get a datasheet out
    of Intel without having to sign a NDA. Good Luck!

<sect1>LinkSys
<p>

<sect2>LinkSys Pocket Ethernet Adapter Plus (PEAEPP)
<p>

    Status -- <em/Supported/

    This is supposedly a DE-620 clone, and is reported to
    work well with that driver. See
    <ref id="de-620" name="DE-620"> for more information.

<sect2>LinkSys PCMCIA Adaptor
<p>

    Status -- <em/Supported/

    This is supposed to be a re-badged DE-650. See
    <ref id="de-650" name="DE-650"> for more information.

<sect1>Microdyne
<p>

<sect2>Microdyne Exos 205T
<p>

    Status -- <em/Semi-Supported/

    Another i82586 based card. Dirk Niggemann
    <tt/[e-mail]/
    has written a driver that he classes as ``pre-alpha''
    that he would like people to test. Mail him for more details.

<sect1>Mylex
<p>

    Mylex can be reached at the following numbers, in case anyone
    wants to ask them anything.

<verb>
    MYLEX CORPORATION, Fremont
    Sales:    800-77-MYLEX, [phone]
    FAX:    [phone].
</verb>

<sect2>Mylex LNP101
<p>

    Status -- <em/Supported/

    This is a PCI card that is based on DEC's 21040 chip.
    It is selectable between 10BaseT, 10Base2 and 10Base5 output.
    The LNP101 card has been verified to work with the generic
    21040 driver.

    See the section on the 21040 chip
    (<ref id="dec-21040" name="DEC 21040">)
    for more information.

<sect2>Mylex LNP104
<p>

    Status -- <em/Not Supported/

    The LNP104 uses the DEC 21050 chip to deliver <em/four/
    independent 10BaseT ports. It may work with 21040 drivers,
    but nobody has tried yet.

<sect2>Mylex M390B EISA
<p>

    Status -- <em/Not Supported/

    Russ writes that "...it's a fairly vanilla
    8390-based adapter.  Only really tricky thing is that, like the
    Interlan es3210, you MUST do only 32-bit transfers to/from the board.
    Look at the Crynwr packet driver and try making the
    changes yourself...''

<sect1>Novell Ethernet, NExxxx and associated clones.<label id="novell">
<p>

    The prefix `NE' came from Novell Ethernet. Novell followed the
    cheapest NatSemi databook design and sold the manufacturing rights
    (spun off?) Eagle, just to get reasonably-priced ethercards into
    the market. (The now ubiquitous NE2000 card.)

<sect2>NE1000, NE2000<label id="ne2k">
<p>

    Status -- <em/Supported/

    NOTE: If you are using a kernel that is older than
    v1.2.9, it is <em>strongly</em> recommended that you
    upgrade to a newer version. There was an important
    bugfix made to the ne driver in 1.2.7, and another
    important bugfix made to the upper layers (dev.c) in
    1.2.9. Both of these bugs can cause a ne2000 card
    to hang your computer.

    The ne2000 is now a generic name for a bare-bones design around
    the NatSemi 8390 chip. They use programmed I/O rather than
    shared memory, leading to easier installation but
    slightly lower performance and a few problems. Again,
    the savings of using an 8 bit NE1000 over the NE2000
    are only warranted if you expect light use. Some
    recently introduced NE2000 clones use the National
    Semiconductor `AT/LANTic' 83905 chip, which offers
    a shared memory mode similar to the 8013 and EEPROM
    or software configuration. Some problems can arise
    with poor NE2000 clones. See
    <ref id="ne2k-probs" name="Problems with...">, and
    <ref id="ne2k-clones" name="Poor NE2000 Clones">
    In general it is not a good idea to put a NE2000
    clone at I/O address <tt/0x300/ because nearly
    <em/every/ device driver probes there at boot. Some
    poor NE2000 clones don't take kindy to being prodded
    in the wrong areas, and will respond by locking your
    machine.

    Donald has written a NE2000 diagnostic program, but it
    is still presently in alpha test. (ne2k)
    See <ref id="diag" name="Diagnostic Programs"> for more
    information.

    If you intend on using this driver as a loadable module
    you should probably see
    <ref id="modules" name="Using the Ethernet Drivers as Modules">
    and also
    <ref id="mod8390" name="8390 Based Cards as Modules"> for
    module specific information.

<sect2>NE1500, NE2100<label id="ne1500">
<p>

    Status -- <em/Supported/

    These cards use the original 7990 LANCE chip from AMD and
    are supported using the Linux lance driver. Newer NE2100
    clones use the updated PCnet/ISA chip from AMD.

    Some earlier versions of the lance driver had problems
    with getting the IRQ line via autoIRQ from the original
    Novell/Eagle 7990 cards. Hopefully this is now fixed.
    If not, then specify the IRQ via LILO, and let us know
    that it still has problems.

    DMA selection and chip numbering information can be found in
    <ref id="lance" name="AMD LANCE">.

    More technical information on LANCE based cards can be found in
    <ref id="amd-notes" name="Notes on AMD...">

<sect2>NE3200<label id="ne3200">
<p>

    Status -- <em/Not Supported/

    This card uses a lowly 8MHz 80186, and hence you are better
    off using a cheap NE2000 clone. Even if a driver was available,
    the NE2000 card would most likely be faster.

<sect1>Pure Data
<p>

<sect2>PDUC8028, PDI8023
<p>

    Status -- <em/Supported/

    The PureData PDUC8028 and PDI8023 series of cards are reported
    to work, thanks to special probe code contributed by Mike
    Jagdis <tt/[e-mail]/. The support is integrated
    with the WD driver.

<sect1>Racal-Interlan
<p>

<sect2>NI5210
<p>

    Status -- <em/Semi-Supported/

    Michael Hipp has written a driver for this card. It is included
    in the standard kernel as an `alpha' driver. Michael would like
    to hear feedback from users that have this card. See
    <ref id="alfa" name="Alpha Drivers"> for important
    information on using alpha-test ethernet drivers
    with Linux.

    Michael says that ``the internal sysbus seems to be slow.
    So we often lose packets because of overruns while receiving
    from a fast remote host.''

    This card also uses one of the Intel chips. See
    <ref id="i82586" name="Programming the Intel Chips">
    for more technical information.


<sect2>NI6510 (not EB)<label id="ni65xx">
<p>

    Status -- <em/Semi-Supported/

    There is also a driver for the LANCE based NI6510, and it
    is also written by Michael Hipp. Again, it is also an
    `alpha' driver. For some reason, this card is not compatible
    with the generic LANCE driver. See
    <ref id="alfa" name="Alpha Drivers"> for important
    information on using alpha-test ethernet drivers
    with Linux.

<sect2>EtherBlaster (aka NI6510EB)
<p>
    Status -- <em/Supported/

    As of kernel 1.3.23, the generic LANCE driver had a check
    added to it for the <tt/0x52, 0x44/ NI6510EB specific signature.
    Others have reported that this signature is not the same
    for all NI6510EB cards however, which will cause the lance
    driver to not detect your card. If this happens to you, you
    can change the probe (at about line 322 in lance.c) to printk()
    out what the values are for your card and then use them instead
    of the <tt/0x52, 0x44/ defaults.

    The cards should probably be run in `high-performance' mode
    and not in the NI6510 compatible mode when using the lance driver.

<sect1>Sager
<p>

<sect2>Sager NP943
<p>

    Status -- <em/Semi-Supported/

    This is just a 3c501 clone, with a different S.A. PROM
    prefix. I assume it is equally as brain dead as the
    original 3c501 as well. Kernels 1.1.53 and up check
    for the NP943 i.d. and then just treat it as a 3c501
    after that. See <ref id="3c501" name="3Com 3c501">
    for all the reasons as to why you really don't want
    to use one of these cards.

<sect1>Schneider & Koch
<p>

<sect2>SK G16
<p>

    Status -- <em/Supported/

    This driver was included into the v1.1 kernels, and it was
    written by PJD Weichmann and SWS Bern. It appears that the
    SK G16 is similar to the NI6510, in that it is based on
    the first edition LANCE chip (the 7990). Once again, it
    appears as though this card won't work with the generic
    LANCE driver.

<sect1>SEEQ
<p>

<sect2>SEEQ 8005
<p>

    Status -- <em/Supported/

    This driver was included into early 1.3.x kernels, and
    was written by Hamish Coleman.  There is little information
    about the card included in the driver, and hence little
    information to be put here. If you have a question, you
    are probably best off e-mailing [e-mail]

<sect1>SMC (Standard Microsystems Corp.)
<p>

    Please see <ref id="smc" name="Western Digital"> for
    information on SMC cards.

<sect1>Thomas Conrad
<p>

<sect2>Thomas Conrad TC-5048
<p>

    This is yet another PCI card that is based on DEC's 21040 chip.

    See the section on the 21040 chip
    (<ref id="dec-21040" name="DEC 21040">)
    for more information.


<sect1>Western Digital / SMC <label id="smc">
<p>

The ethernet part of Western Digital has been bought out by SMC.
One common mistake people make is that the relatively new SMC Elite Ultra
is the same as the older SMC Elite16 models -- this is <bf/not/ the case.

Here is how to contact SMC (not that you should need to.)
<quote>
SMC / Standard Microsystems Corp., 80 Arkay Drive, Hauppage, New York,
11788, USA.
</quote>

Technical Support via phone:
<verb>
    800-992-4762 (USA)
    800-433-5345 (Canada)
    516-435-6250 (Other Countries)
</verb>

Literature requests:
<verb>
    800-SMC-4-YOU (USA)
    800-833-4-SMC (Canada)
    516-435-6255  (Other Countries)
</verb>

Technical Support via E-mail:
<verb>
    [e-mail]
</verb>

FTP Site:
<verb>
    ftp.smc.com
</verb>

<sect2>WD8003, SMC Elite
<p>
    Status -- <em/Supported/

    These are the 8-bit versions of the card. The
    8 bit 8003 is slightly less expensive, but only
    worth the savings for light use. Note that some
    of the non-EEPROM cards (clones with jumpers, or
    old <em/old/ old wd8003 cards) have no way of reporting
    the IRQ line used. In this case, auto-irq is used, and if
    that fails, the driver silently assings IRQ 5.
    Information regarding what the jumpers on old non-EEPROM
    wd8003 cards do can be found in conjunction with the
    SMC setup/driver disks stored on
    <tt/dorm.rutgers.edu/ in the directory
    <tt>/pub/novell/nic_drvs/</tt>. Note that some of the
    newer SMC `SuperDisk' programs will fail to detect
    the old EEPROM-less cards. The file <tt/SMCDSK46.EXE/
    seems to be a good all-round choice. Also the jumper
    settings for old cards are in an ascii text file in the
    aforementioned archive. The latest (greatest?) version
    can be obtained from <tt/ftp.smc.com/.

    As these are basically the
    same as their 16 bit counterparts (WD8013 / SMC Elite16),
    you should see the next section for more information.


<sect2>WD8013, SMC Elite16<label id="8013">
<p>

    Status -- <em/Supported/

    Over the
    years the design has added more registers and an
    EEPROM. Clones usually go by the `8013' name, and
    usually use a non-EEPROM (jumpered) design.
    Late model SMC cards will have two main PLCC chips
    on board; the SMC 83c690 and the SMC 83c694.
    The shared memory design makes the cards 10-20 &percnt faster,
    than PIO cards, especially with larger packets.
    More importantly, from the
    driver's point of view, it avoids a few bugs in the
    programmed-I/O mode of the 8390, allows safe
    multi-threaded access to the packet buffer, and
    it doesn't have a programmed-I/O data register that
    hangs your machine during warm-boot probes.

    Non-EEPROM cards that can't just read the selected
    IRQ will attempt auto-irq, and if that fails, they will
    silently assign IRQ 10. (8 bit versions will assign IRQ 5)

    Cards with a non standard amount of memory on board can
    have the memory size specified at boot (or at `insmod'
    time if using modules). The standard memory size is
    8kB for an 8bit card and 16kB for a 16bit card.
    For example, the older WD8003EBT cards could be jumpered
    for 32kB memory. To make full use of that RAM, you would
    use something like (for i/o=0x280 and IRQ 9):
<code>
    LILO: linux ether=9,0x280,0xd0000,0xd8000,eth0
</code>

    Also see <ref id="8013-probs" name="8013 problems">
    for some of the more common problems that people have.

    If you intend on using this driver as a loadable module
    you should probably see
    <ref id="modules" name="Using the Ethernet Drivers as Modules">
    and also
    <ref id="mod8390" name="8390 Based Cards as Modules"> for
    module specific information.

<sect2>SMC Elite Ultra<label id="ultra">
<p>

    Status -- <em/Supported/

    This ethercard is based on a new chip from SMC, with
    a few new features. While it has a mode that is
    similar to the older SMC ethercards, it's not entirely
    compatible with the old WD80*3 drivers. However, in
    this mode it shares most of its code with the other
    8390 drivers, while operating slightly faster than a
    WD8013 clone.

    Since part of the Ultra <em/looks like/
    an 8013, the Ultra probe is supposed to find an
    Ultra before the wd8013 probe has a chance to
    mistakenly identify it.

    Donald mentioned that it is possible to write a separate
    driver for the Ultra's `Altego' mode which allows
    chaining transmits at the cost of inefficient use of receive
    buffers, but that will probably not happen.

    Bus-Master SCSI host adaptor users take note: In the
    manual that ships with Interactive UNIX, it mentions
    that a bug in the SMC Ultra will cause data corruption
    with SCSI disks being run from an aha-154X host adaptor.
    This will probably bite aha-154X compatible cards, such
    as the BusLogic boards, and the AMI-FastDisk SCSI host
    adaptors as well.

    SMC has acknowledged the problem occurs with
    Interactive, and older Windows NT drivers. It is
    a hardware conflict with early revisions of the card
    that can be worked around in the driver design. The current
    Ultra driver protects against this by only enabling the
    shared memory during data transfers with the card. Make sure
    your kernel version is at least 1.1.84, or that the driver
    version reported at boot is at least <tt/smc-ultra.c:v1.12/
    otherwise you are vulnerable.

    If you intend on using this driver as a loadable module
    you should probably see
    <ref id="modules" name="Using the Ethernet Drivers as Modules">
    and also
    <ref id="mod8390" name="8390 Based Cards as Modules"> for
    module specific information.


<sect2>SMC 8416 (EtherEZ)
<p>

    Status -- <em/Supported/

    This card uses SMC's 83c795 chip and supports the Plug 'n Play
    specification. It also has an <em/SMC Ultra/ compatible mode,
    which allows it to be used with the Linux Ultra driver.
    In this compatibility mode, it uses shared memory
    instead of programmed i/o. Be sure to set your card for
    this compatibility mode. See the above information for
    notes on the Ultra driver.

    Note that the EtherEZ specific checks were added to the
    SMC Ultra driver in 1.1.84, and hence earlier kernel versions
    will not detect or handle these cards correctly.


<sect2>SMC 8432 PCI (EtherPower)<label id="smc-pci">
<p>

    Status -- <em/Supported/

    These cards are
    a basic DEC 21040 implementation, i.e. one big chip
    and a couple of transceivers. Donald has used one
    of these cards for his development of the generic
    21040 driver (aka <tt/tulip.c/). Thanks to Duke Kamstra,
    once again, for supplying a card to do development on.

    Some of the newer revisons of this card use the newer
    DEC 21041 chip, which may cause problems with
    older versions of the tulip driver. Make sure you are
    using the latest driver release.

    See <ref id="dec-21040" name="DEC 21040"> for more
    details on using one of these cards, and the current
    status of the driver.

<sect2>SMC 3008
<p>

    Status -- <em/Not Supported/

    These 8 bit cards are based on the Fujitsu MB86950, which is an
    ancient version of the MB86965 used in the Linux at1700
    driver. Russ says that you could probably hack up a driver
    by looking at the at1700.c code and his DOS packet driver
    for the Tiara card (tiara.asm)

<sect2>SMC 3016
<p>

    Status -- <em/Not Supported/

    These are 16bit i/o mapped 8390 cards, much similar to a generic
    NE2000 card. If you can get the specifications from SMC, then
    porting the NE2000 driver would probably be quite easy.

<sect2>SMC 9000
<p>

    Status -- <em/Not Supported/

    These cards are VLB cards based on the 91c92 chip. They are
    fairly expensive, and hence the demand for a driver is pretty
    low at the moment.

<sect1>Xircom<label id="xircom">
<p>

    Another group that won't release documentation. No cards
    supported. Don't look for any support in the future unless
    they release their programming information. And this is
    highly unlikely, as they <em/forbid/ you from even reverse-
    engineering their drivers. If you are already stuck with one,
    see if you can trade it off on some DOS (l)user.

    And if you just want to verify that this is the case, you can
    reach Xircom at 1-800-874-7875, 1-800-438-4526 or [phone].
    They used to advertise that their products "work with all
    network operating systems", but have since stopped. Wonder
    why...

<sect2>PE1, PE2, PE3-10B*
<p>

    Status -- <em/Not Supported/

    Not to get your hopes up, but if you have one of these parallel
    port adaptors, you may be able to use it in the DOS emulator
    with the Xircom-supplied DOS drivers. You will have to allow
    DOSEMU access to your parallel port, and will probably have
    to play with SIG (DOSEMU's Silly Interrupt Generator).

<sect1>Zenith<label id="zenith">
<p>

<sect2>Z-Note<label id="z-note">
<p>

    Status -- <em/Supported/

    The built-in Z-Note network adaptor is based on the Intel
    i82593 using <em/two/ DMA channels. There is an (alpha?) driver
    available in the present kernel version. As with all notebook
    and pocket adaptors, it is under the `Pocket and portable
    adaptors' section when running <tt/make config/.
    See <ref id="i82586" name="Programming the Intel chips">
    for more technical information.
    Also note that the IBM ThinkPad 300 is compatible with the Z-Note.

<sect1>Znyx<label id="zynx">
<p>

<sect2>Znyx ZX342 (DEC 21040 based)
<p>

    Status -- <em/Supported/

    You have a choice of <em/two/ drivers for cards based on this
    chip. There is the DE425 driver written by David, and the
    generic 21040 driver that Donald has written.

    Note that as of 1.1.91, David has added a compile time option that
    may allow non-DEC cards (such as the Znyx cards) to work with
    this driver. Have a look at <tt/README.de4x5/ for details.

    See <ref id="dec-21040" name="DEC 21040">
    for more information on these cards, and the present driver
    situation.

<sect1>Drivers for Non-Ethernet Devices
<p>

    There are a few other drivers that are in the linux source,
    that present an <em/ethernet-like/ device to network
    programs, while not really being ethernet. These are briefly
    listed here for completeness.

    <tt/dummy.c/ - The purpose of this driver is to provide a device
    to point a route through, but not to actually transmit packets.

    <tt/eql.c/ - Load Equalizer, enslaves multiple devices (usually
    modems) and balances the Tx load across them while presenting
    a single device to the network programs.

    <tt/ibmtr.c/ - IBM Token Ring, which is not really ethernet.
    Broken-Ring requires source routing and other uglies.

    <tt/loopback.c/ - Loopback device, for which all packets
    from you machine and destined for your own machine go.
    It essentially just moves the packet off the Tx queue and
    onto the Rx queue.

    <tt/pi2.c/ - Ottawa Amateur Radio Club PI and PI2 interface.

    <tt/plip.c/ - Parallel Line Internet Protocol, allows two
    computers to send packets to each other over two joined
    parallel ports in a point-to-point fashion.

    <tt/ppp.c/ - Point-to-Point Protocol (RFC1331), for the
    Transmission of Multi-protocol Datagrams over a
    Point-to-Point Link (again usually modems).

    <tt/slip.c/ - Serial Line Internet Protocol, allows two
    computers to send packets to each other over two joined
    serial ports (usually via modems) in a point-to-point fashion.

    <tt/tunnel.c/ - Provides an IP tunnel through which you can
    tunnel network traffic transparently across subnets

    <tt/wavelan.c/ - An Ethernet-like radio transceiver
    controlled by the Intel 82586 coprocessor which is used on
    other ethercards such as the Intel EtherExpress.

<sect>Cables, Coax, Twisted Pair<label id="cable">
<p>
    If you are starting a network from scratch, it's considerably less
    expensive to use thin ethernet, RG58 co-ax cable with BNC connectors,
    than old-fashioned thick ethernet, RG-5 cable with N connectors, or
    10baseT, twisted pair telco-style cables with RJ-45 eight wire `phone'
    connectors. See <ref id="cable-intro" name="Type of cable..."> for
    an introductory look at cables.

    Also note that the FAQ from <em/comp.dcom.lans.ethernet/ has a lot
    of useful information on cables and such. Look in
    <tt/dorm.rutgers.edu/ for the file
    <tt>/pub/novell/info_and_docs/Ethernet.FAQ</tt>

<sect1>Thin Ethernet (thinnet)<label id="bnc">
<p>

    Thin ethernet is the `ether of choice'. The cable is inexpensive. If
    you are making your own cables solid-core RG58A is $0.27/m. and
    stranded RG58AU is $0.45/m. Twist-on BNC
    connectors are < $2 ea.,
    and other misc. pieces are similarly inexpensive. It is essential
    that you properly terminate each end of the cable with 50 ohm
    terminators, so budget $2 ea. for a pair. It's also vital that
    your cable have no `stubs' -- the `T' connectors must be attached
    directly to the ethercards. The only drawback is that if you have
    a big loop of machines connected together, and some bonehead breaks
    the loop by taking one cable off the side of his tee, the whole
    network goes down because it sees an infinite impedance (open
    circuit) instead of the required 50 ohm termination. Note that
    you can remove the tee piece from the card itself without killing
    the whole subnet, as long as you don't remove the cables from the
    tee itself. Of course this will disturb the machine that you
    pull the actual tee off of. 8-) And if you are doing a small
    network of two machines, you <em/still/ need the tees and the 50 ohm
    terminators -- you <em/can't/ just cable them together!


<sect1>Twisted Pair<label id="utp">
<p>

    Twisted pair networks require active hubs,
    which start around $200,
    and the raw cable cost can actually be higher than thinnet. They are
    usually sold using the claim that you can use your existing telephone
    wiring, but it's a rare installation where that turns out to be the
    case. The claim that you can upgrade to higher speeds is also
    suspect, as most proposed schemes use higher-grade
    (read $$) cable and
    more sophisticated termination ($$$)
    than you would likely install on
    speculation. New gizmos are floating around which allow you to
    daisy-chain machines together, and the like. For example,
    Farallon sells EtherWave adaptors and transceivers. This device
    allows multiple 10baseT devices to be daisy-chained. They also
    sell a 3c509 clone that includes the EtherWave transceiver.
    The drawback is that it's more expensive and less reliable than a
    cheap ($100-$150) mini-hub and another ethercard.
    You probably should either go for the hub approach or switch over
    to 10base2 thinnet.

    On the other hand, hubs are rapidly dropping in price, all 100Mb/sec
    ethernet proposals use twisted pair, and most new business
    installations use twisted pair. (This is probably to avoid the
    problem with idiots messing with the BNC's as described above.)

    Also, Russ Nelson adds that `New installations should use Category 5
    wiring. Anything else is a waste of your installer's time, as
    100Base-whatever is going to require Cat 5.'

    If you are only connecting two machines, it is possible to avoid
    using a hub, by swapping the Rx and Tx pairs (1-2 and 3-6).

    If you hold the RJ-45 connector facing you (as if you were
    going to plug it into your mouth) with the lock tab on the top,
    then the pins are numbered 1 to 8 from left to right. The pin
    usage is as follows:

<verb>
    Pin Number        Assignment
    ----------        ----------
    1            Output Data (+)
    2            Output Data (-)
    3            Input Data (+)
    4            Reserved for Telephone use
    5            Reserved for Telephone use
    6            Input Data (-)
    7            Reserved for Telephone use
    8            Reserved for Telephone use
</verb>

    Some cards, like the wd8013 can sense reversed polarity, and
    will adjust accordingly. Also note that 3 and 6 <bf/must/ be
    a twisted pair. If you make 3-4 a twisted pair, and 5-6 the
    other twisted pair, your cable may work for lengths less than
    a metre, but will <em/fail miserably/ for longer lengths.

    Note that before 10BaseT was ratified as a standard, there
    existed  other network formats using RJ-45
    connectors, and the same wiring scheme as above. Examples
    are SynOptics's LattisNet, and AT&T's StarLAN.
    In some cases, (as with early 3C503 cards) you could set jumpers
    to get the card to talk to hubs of different types, but in most cases
    cards designed for these older types of networks will not work with
    standard 10BaseT networks/hubs. (Note that if the cards also have
    an AUI port, then there is no reason as to why you can't use that,
    combined with an AUI to 10BaseT transceiver.)

<sect1>Thick Ethernet
<p>
    Thick ethernet is mostly obsolete, and is usually used only to remain
    compatible with an existing implementation. You can stretch the rules
    and connect short spans of thick and thin ethernet together with a
    passive $3 N-to-BNC connector, and that's often the best
    solution to expanding an existing thicknet. A correct (but expensive)
    solution is to use a repeater in this case.
</sect>

<sect>Software Configuration and Card Diagnostics<label id="utils">
<p>

    In most cases, if the configuration is done by software,
    and stored in an EEPROM, you will usually have to boot
    DOS, and use the supplied DOS program to set the cards
    IRQ, I/O, mem_addr and whatnot. Besides, hopefully it is
    something you will only be setting once. For those that
    don't have the DOS utility available, note that a fair
    number of NIC setup/driver disks (e.g. 3Com, SMC/WD and
    Allied Telesis NIC's) are available from <tt/dorm.rutgers.edu/
    in the directory <tt>/pub/novell/nic_drvs/</tt>
    However, there are some cards for which Linux versions of
    the config utils exist, and they are listed here.

    Also, Donald has written a few small card diagnostic
    programs that run under Linux. Most of these are a result
    of debugging tools that he has created while writing the
    various drivers. Don't expect
    fancy menu-driven interfaces. You will have to read the
    source code to use most of these. Even if your particular
    card doesn't have a corresponding diagnostic, you can
    still get lots of information just by typing
    <tt>cat /proc/net/dev</tt> -- assuming that your card
    was at least detected at boot.

    In either case, you will have to run most of these programs
    as root (to allow I/O to the ports) and you probably want
    to shut down the ethercard before doing so by typing
    <tt/ifconfig eth0 down/ (Note: replace <tt/eth0/ with
    <tt/atp0/ or whatever when appropriate.)

<sect1>Configuration Programs for Ethernet Cards<label id="config">
<p>

    For people with wd80x3 cards, there is the program <tt/wdsetup/
    which can be found in <tt/wdsetup-0.6a.tar.gz/ on Linux ftp sites.
    I am not sure if it is being actively maintained or not, as it has
    not been updated for quite a while. If it works fine for you
    then great, if not, use the DOS version that you should have got
    with your card. If you don't have the DOS version, you will be
    glad to know that the SMC setup/driver disks are available
    at the <tt/dorm.rutgers.edu/ site mentioned above.
    Of course, you <em/have/ to have an EEPROM card to use this utility.
    Old, <em/old/ wd8003 cards, and some wd8013 clones use jumpers
    to set up the card instead.

    The Digital EtherWorks 3 card can be configured in a similar
    fashion to the DOS program <tt/NICSETUP.EXE/. David C. Davies
    wrote this and other tools for the EtherWorks 3 in conjunction
    with the driver. Look on <tt/sunsite.unc.edu/ in the directory
    <tt>/pub/linux/system/Network/management</tt> for the file
    that is named <tt/ewrk3tools-X.XX.tar.gz/.

    Some Nat Semi DP83905 implementations (such as the AT/LANTIC
    and the NE2000+) are software configurable. (Note that this
    card can also emulate a wd8013!) You can get the file
    <tt>/pub/linux/setup/atlantic.c</tt> from Donald's ftp
    server, <tt/cesdis.gsfc.nasa.gov/ to configure this card.
    Be careful when configuring NE2000+ cards, as you can give
    them bad setting values which will require you to open the
    case and switch a jumper to force it back to sane settings.

    The 3Com Etherlink III family of cards (i.e. 3c5x9) can
    be configured by using another config utility from Donald.
    You can get the file <tt>/pub/linux/setup/3c5x9setup.c</tt>
    from Donald's ftp server, <tt/cesdis.gsfc.nasa.gov/ to
    configure these cards. (Note that the DOS 3c5x9B config
    utility may have more options pertaining to the new ``B''
    series of the Etherlink III family.)



<sect1>Diagnostic Programs for Ethernet Cards<label id="diag">
<p>

    Any of the diagnostic programs that Donald has written can
    be obtained from this URL.

    <url url="http://cesdis.gsfc.nasa.gov/pub/linux/diag/diagnostic.html"
        name="Ethercard Diagnostics">

    Allied Telesis AT1700 -- look for the file
    <tt>/pub/linux/diag/at1700.c</tt> on <tt/cesdis.gsfc.nasa.gov/.

    Cabletron E21XX -- look for the file
    <tt>/pub/linux/diag/e21.c</tt> on <tt/cesdis.gsfc.nasa.gov/.

    HP PCLAN+ -- look for the file
    <tt>/pub/linux/diag/hp+.c</tt> on <tt/cesdis.gsfc.nasa.gov/.

    Intel EtherExpress -- look for the file
        <tt>/pub/linux/diag/eexpress.c</tt> on <tt/cesdis.gsfc.nasa.gov/.

    NE2000 cards -- look for the file
    <tt>/pub/linux/diag/ne2k.c</tt> on <tt/cesdis.gsfc.nasa.gov/.

    RealTek (ATP) Pocket adaptor -- look for the file
    <tt>/pub/linux/diag/atp-diag.c</tt> on <tt/cesdis.gsfc.nasa.gov/.

    All Other Cards -- try typing <tt>cat /proc/net/dev</tt> and
    see what useful info the kernel has on the card in question.

<sect>Technical Information<label id="tech-intro">
<p>

    For those who want to play with the present drivers, or try to make
    up their own driver for a card that is presently unsupported, this
    information should be useful. If you do not fall into this category,
    then perhaps you will want to skip this section.

<sect1>Probed Addresses<label id="probe">
<p>

    While trying to determine what ethernet card is there, the following
    addresses are autoprobed, assuming the type and specs of the card
    have not been set in the kernel. The file names below are
    in /usr/src/linux/drivers/net/

<code>
        3c501.c         0x280, 0x300
        3c503.c:        0x300, 0x310, 0x330, 0x350, 0x250, 0x280, 0x2a0, 0x2e0
        3c505.c:        0x300, 0x280, 0x310
        3c507.c:        0x300, 0x320, 0x340, 0x280
        3c509.c:        Special ID Port probe
    apricot.c    0x300
        at1700.c:       0x300, 0x280, 0x380, 0x320, 0x340, 0x260, 0x2a0, 0x240
        atp.c:          0x378, 0x278, 0x3bc
        depca.c         0x300, 0x200
        de600.c:        0x378
        de620.c:        0x378
        eexpress.c:     0x300, 0x270, 0x320, 0x340
        hp.c:           0x300, 0x320, 0x340, 0x280, 0x2C0, 0x200, 0x240
    hp-plus.c    0x200, 0x240, 0x280, 0x2C0, 0x300, 0x320, 0x340
        lance.c:        0x300, 0x320, 0x340, 0x360
        ne.c:           0x300, 0x280, 0x320, 0x340, 0x360
    ni52.c        0x300, 0x280, 0x360, 0x320, 0x340
    ni65.c        0x300, 0x320, 0x340, 0x360
        smc-ultra.c:    0x200, 0x220, 0x240, 0x280, 0x300, 0x340, 0x380
        wd.c:           0x300, 0x280, 0x380, 0x240
</code>

    There are some NE2000 clone ethercards out there that are waiting black
    holes for autoprobe drivers. While many NE2000 clones are
    safe until they are enabled, some can't be reset to a safe mode.
    These dangerous ethercards will hang any I/O access to their
    `dataports'. The typical dangerous locations are:
<code>
        Ethercard jumpered base     Dangerous locations (base + 0x10 - 0x1f)
                0x300 *                         0x310-0x317
                0x320                           0x330-0x337
                0x340                           0x350-0x357
                0x360                           0x370-0x377
</code>

    * The 0x300 location is the traditional place to put an ethercard, but
    it's also a popular place to put other devices (often SCSI
    controllers). The 0x320 location is often the next one chosen, but
    that's bad for for the AHA1542 driver probe. The 0x360 location is
    bad, because it conflicts with the parallel port at 0x378.
    If you have two IDE controllers, or two floppy controlers, then
    <tt/0x360/ is also a bad choice, as a NE2000 card will clobber
    them as well.

    Note that kernels > 1.1.7X keep a log of who uses
    which i/o ports, and will not let a driver use i/o ports registered
    by an earlier driver. This may result in probes silently failing.
    You can view who is using what i/o ports by typing
    <tt>cat /proc/ioports</tt> if you have the proc filesystem enabled.

    To avoid these lurking ethercards, here are the things you can do:

<itemize>

<item>    Probe for the device's BIOS in memory space. This is easy
    and always safe, but it only works for cards that always have
    BIOSes, like primary SCSI controllers.

<item>    Avoid probing any of the above locations until you think
    you've located your device. The NE2000 clones have a reset range
    from <base>+0x18 to <base>+0x1f that
    will read as 0xff, so probe
    there first if possible. It's also safe to probe in the 8390
    space at <base>+0x00 - <base>+0x0f, but that area
    will return quasi-random values

<item>    If you must probe in the dangerous range, for instance if your
    target device has only a few port locations, first check that
    there isn't an NE2000 there. You can see how to do this by
    looking at the probe code in /usr/src/linux/net/inet/ne.c

<item>    Use the `reserve' boot time argument to protect volatile
    areas from being probed. See the information on using boot
    time arguments with LILO in
    <ref id="reserve" name="The reserve command">

</itemize>

<sect1>Writing a Driver<label id="skel">
<p>


    The only thing that one needs to use an ethernet card with Linux
    is the appropriate driver. For this, it is essential that the
    manufacturer will release the technical programming information to
    the general public without you (or anyone) having to sign your life
    away. A good guide for the likelihood of getting documentation
    (or, if you aren't writing code, the likelihood that someone
    else will write that driver you really, really need) is the
    availability of the Crynwr (nee Clarkson) packet driver. Russ
    Nelson runs this operation, and has been very helpful in supporting
    the development of drivers for Linux. <em/Net-surfers/ can try this
    URL to look up Russ' software.

    <url url="http://www.crynwr.com/crynwr/home.html"
        name="Russ Nelson's Packet Drivers">

    Given the documentation, you can write a driver for
    your card and use it for Linux (at least in theory).
    Keep in mind that some old hardware that was designed for XT type
    machines will not function very well in a multitasking
    environment such as Linux. Use of these will lead to major
    problems if your network sees a reasonable amount of traffic.

    Most cards come with drivers for MS-DOS interfaces such as
    NDIS and ODI, but these are useless for Linux. Many people
    have suggested directly linking them in or automatic
    translation, but this is nearly impossible. The MS-DOS
    drivers expect to be in 16 bit mode and hook into `software
    interrupts', both incompatible with the Linux kernel. This
    incompatibility is actually a feature, as some Linux drivers
    are considerably better than their MS-DOS counterparts. The
    `8390' series drivers, for instance, use ping-pong transmit
    buffers, which are only now being introduced in the MS-DOS world.

    (Ping-pong Tx buffers means using at least 2 max-size
    packet buffers for Tx packets. One is loaded while the card
    is transmitting the other. The second is then sent as soon
    as the first finished, and so on. In this way, most cards
    are able to continuously send back-to-back packets onto
    the wire.)

    OK. So you have decided that you want to write a driver for the
    Foobar Ethernet card, as you have the programming information,
    and it hasn't been done yet. (...these are the two main
    requirements ;-) You should start with the skeleton
    network driver that is provided
    with the Linux kernel source tree. It can be found in the file
    /usr/src/linux/drivers/net/skeleton.c in all recent kernels.


<sect1>Driver interface to the kernel
<p>

    Here are some notes on the functions that you would have to
    write if creating a new driver. Reading this in conjunction
    with the above skeleton driver may help clear things up.


<sect2>Probe
<p>

    Called at boot to check for existence of card. Best if it
    can check un-obtrsively by reading from memory, etc. Can
    also read from i/o ports. Initial writing to i/o ports in a probe
    is <em/not good/ as it may kill another device.
    Some device initialization is usually done here (allocating
    i/o space, IRQs,filling in the dev->??? fields etc.)
    You need to know what io ports/mem the card can be
    configured to, how to enable shared memory (if used)
    and how to select/enable interrupt generation, etc.

<sect2>Interrupt handler
<p>

    Called by the kernel when the card posts an interrupt.
    This has the job of determining why the card posted
    an interrupt, and acting accordingly. Usual interrupt
    conditions are data to be rec'd, transmit completed,
    error conditions being reported. You need to know
    any relevant interrupt status bits so that you can
    act accordingly.

<sect2>Transmit function
<p>

    Linked to dev->hard_start_xmit() and is called by the
    kernel when there is some data that the kernel wants
    to put out over the device. This puts the data onto
    the card and triggers the transmit. You need to
    know how to bundle the data and how to get it onto the
    card (shared memory copy, PIO transfer, DMA?) and in
    the right place on the card. Then you need to know
    how to tell the card to send the data down the wire, and
    (possibly) post an interrupt when done.
    When the hardware can't accept additional packets it should set
    the dev->tbusy flag. When additional room is available, usually
    during a transmit-complete interrupt, dev->tbusy should be cleared
    and the higher levels informed with <tt/mark_bh(INET_BH)/.

<sect2>Receive function
<p>

    Called by the kernel interrupt handler when the card reports
    that there is data on the card. It pulls the data off
    the card, packages it into a sk_buff and lets the
    kernel know the data is there for it by doing a
    netif_rx(sk_buff). You need to know how to enable
    interrupt generation upon Rx of data, how to check any
    relevant Rx status bits, and how to get that data off the
    card (again sh mem, PIO, DMA, etc.)

<sect2>Open function
<p>

    linked to dev->open and called by the networking layers
    when somebody does <tt/ifconfig eth0 up/ - this
    puts the device on line and enables it for Rx/Tx of
    data. Any special initialization incantations that were
    not done in the probe sequence (enabling IRQ generation, etc.)
    would go in here.

<sect2>Close function (optional)
<p>

    This puts the card in a sane state when someone
    does <tt/ifconfig eth0 down/.
    It should free the IRQs and DMA channels if the hardware permits,
        and turn off anything that will save power (like the transceiver).

<sect2>Miscellaneous functions
<p>

    Things like a reset function, so that if things go south,
    the driver can try resetting the card as a last ditch effort.
    Usually done when a Tx times out or similar. Also a function
    to read the statistics registers of the card if so equipped.

<sect1>Interrupts and Linux
<p>

    There are two kinds of interrupt handlers in Linux:
    fast ones and slow ones. You decide what kind you are installing by
    the flags you pass to irqaction(). The fast ones, such as the serial
    interrupt handler, run with _all_ interrupts disabled. The normal
    interrupt handlers, such as the one for ethercard drivers, runs with
    other interrupts enabled.

    There is a two-level interrupt structure. The `fast' part handles the
    device register, removes the packets, and perhaps sets a flag.  After
    it is done, and interrupts are re-enabled, the slow part is run if the
    flag is set.

    The flag between the two parts is set by:

<tscreen>
        mark_bh(INET_BH);
</tscreen>

    Usually this flag is set within dev_rint() during a received-packet
    interrupt, and set directly by the device driver during a
    transmit-complete interrupt.

    You might wonder why all interrupt handlers cannot run in
    `normal mode' with other interrupts enabled. Ross Biro uses this
    scenario to illustrate the problem:

<itemize>

<item>    You get a serial interrupt, and start processing it.
    The serial interrupt is now masked.

<item>    You get a network interrupt, and you start transferring
    a maximum-sized 1500 byte packet from the card.

<item>    Another character comes in, but this time the interrupts
    are masked!

</itemize>

    The `fast' interrupt structure solves this problem by allowing
    bounded-time interrupt handlers to run without the risk of leaving
    their interrupt lines masked by another interrupt request.

    There is an additional distinction between fast and slow interrupt
    handlers -- the arguments passed to the handler. A `slow' handler is
    defined as

<code>

        static void
        handle_interrupt(int reg_ptr)
        {
            int irq = -(((struct pt_regs *)reg_ptr)->orig_eax+2);
            struct device *dev = irq2dev_map[irq];
        ...

</code>

    While a fast handler gets the interrupt number directly

<code>

        static void
        handle_fast_interrupt(int irq)
        {
        ...

</code>

    A final aspect of network performance is latency. The only board
    that really addresses this is the 3c509, which allows a predictive
    interrupt to be posted. It provides an interrupt response timer so
    that the driver can fine-tune how early an interrupt is generated.

<sect1>Programming the Intel chips (i82586 and i82593)<label id="i82586">
<p>

    These chips are used on a number of cards, namely the 3c507 ('86),
    the Intel EtherExpress 16 ('86), Microdyne's exos205t ('86),
    the Z-Note ('93), and the Racal-Interlan ni5210 ('86).

    Russ Nelson writes:
    `Most boards based on the 82586 can reuse quite a bit of their code.
    More, in fact, than the 8390-based adapters. There are only three
    differences between them:

<itemize>

<item>    The code to get the Ethernet address,
<item>    The code to trigger CA on the 82586, and
<item>    The code to reset the 82586.

</itemize>

    The Intel EtherExpress 16 is an exception, as it I/O maps the 82586.
    Yes, I/O maps it. Fairly clunky, but it works.

    Garrett Wollman did an AT&T driver for BSD that uses the BSD
    copyright. The latest version I have (Sep '92) only uses a single
    transmit buffer. You can and should do better than this if you've
    got the memory. The AT&T and 3c507 adapters do; the ni5210 doesn't.

    The people at Intel gave me a very big clue on how you queue up
    multiple transmit packets. You set up a list of
    NOP-> XMIT-> NOP-> XMIT-> NOP-> XMIT-> beginning)
    blocks, then you set the
    `next' pointer of all the NOP blocks to themselves. Now you start
    the command unit on this chain. It continually processes the first
    NOP block. To transmit a packet, you stuff it into the next transmit
    block, then point the NOP to it. To transmit the next packet, you
    stuff the next transmit block and point the previous NOP to <em/it/. In
    this way, you don't have to wait for the previous transmit to finish,
    you can queue up multiple packets without any ambiguity as to whether
    it got accepted, and you can avoid the command unit start-up delay.'

<sect1>Technical information from 3Com<label id="3com-tech">
<p>

    If you are interested in working on drivers for 3Com cards,
    you can get technical documentation from 3Com. Cameron has
    been kind enough to tell us how to go about it below:

    3Com's Ethernet Adapters are documented for driver writers
    in our `Technical References' (TRs). These manuals describe
    the programmer interfaces to the boards but they don't talk
    about the diagnostics, installation programs, etc that end
    users can see.

    The Network Adapter Division marketing department has the
    TRs to give away. To keep this program efficient, we
    centralized it in a thing called `CardFacts.' CardFacts is
    an automated phone system. You call it with a touch-tone
    phone and it faxes you stuff. To get a TR, call CardFacts
    at 408-727-7021. Ask it for Developer's Order Form,
    document number 9070. Have your fax number ready when you
    call. Fill out the order form and fax it to 408-764-5004.
    Manuals are shipped by Federal Express 2nd Day Service.

    After you get a manual, if you still can't figure out how to
    program the board, try our `CardBoard' BBS at
    1-800-876-3266, and if you can't do that, write
    [e-mail] and ask him for alternatives. If
    you have a real stumper that nobody has figured out yet, the
    fellow who needs to know about it is
    [e-mail].

    There are people here who think we are too free with the
    manuals, and they are looking for evidence that the system
    is too expensive, or takes too much time and effort. That's
    why it's important to try to use CardFacts <em/before/ you
    start calling and mailing the people I named here.

    There are even people who think we should be like Diamond
    and Xircom, requiring tight `partnership' with driver
    writers to prevent poorly performing drivers from getting
    written. So far, 3Com customers have been really good about
    this, and there's no problem with the level of requests
    we've been getting. We need your continued cooperation and
    restraint to keep it that way.

<verb>
    Cameron Spitzer, 408-764-6339
    3Com NAD
    Santa Clara
    work: [e-mail]
    home: [e-mail]
</verb>


<sect1>Notes on AMD PCnet / LANCE Based cards<label id="amd-notes">
<p>

    The AMD LANCE (Local Area Network Controller for Ethernet)
    was the original offering, and has since been replaced by
    the `PCnet-ISA' chip, otherwise known as the 79C960.
    A relatively new chip from AMD, the 79C960, is the heart of many
    new cards being released at present. Note that the name `LANCE'
    has stuck, and some people will refer to the new chip by the old
    name. Dave Roberts of the Network Products Division of AMD was kind
    enough to contribute the following information regarding this chip:

    `As for the architecture itself, AMD developed it originally
    and reduced it to a single chip -- the PCnet(tm)-ISA -- over a year
    ago. It's been selling like hotcakes ever since.

    Functionally, it is equivalent to a NE1500. The register set
    is identical to the old LANCE with the 1500/2100 architecture
    additions. Older 1500/2100 drivers will work on the PCnet-ISA.
    The NE1500 and NE2100 architecture is basically the same.
    Initially Novell called it the 2100, but then tried to distinguish
    between coax and 10BASE-T cards. Anything that was 10BASE-T only was
    to be numbered in the 1500 range. That's the only difference.

    Many companies offer PCnet-ISA based products, including HP,
    Racal-Datacom, Allied Telesis, Boca Research, Kingston Technology, etc.
    The cards are basically the same except that some manufacturers
    have added `jumperless' features that allow the card to
    be configured in software. Most have not. AMD offers a standard
    design package for a card that uses the PCnet-ISA and many
    manufacturers use our design without change.
    What this means is that anybody who wants to write drivers for
    most PCnet-ISA based cards can just get the data-sheet from AMD. Call
    our literature distribution center at [phone] and ask for the
    Am79C960, PCnet-ISA data sheet. It's free.

    A quick way to understand whether the card is a `stock' card
    is to just look at it. If it's stock, it should just have one large
    chip on it, a crystal, a small IEEE address PROM, possibly a socket
    for a boot ROM, and a connector (1, 2, or 3, depending on the media
    options offered). Note that if it's a coax card, it will have some
    transceiver stuff built onto it as well, but that should be near the
    connector and away from the PCnet-ISA.'

    There is also some info regarding the LANCE chip in the file
    lance.c which is included in the standard kernel.

    A note to would-be card hackers is that different LANCE
    implementations do `restart' in different ways. Some pick up
    where they left off in the ring, and others start right from
    the beginning of the ring, as if just initialised. This is
    a concern when setting the multicast list.

<sect1>Multicast and Promiscuous Mode<label id="promisc">
<p>

    Another one of the things Donald has worked on is
    implementing multicast and promiscuous mode hooks.
    All of the <em/released/ (i.e. <bf/not/ ALPHA) ISA drivers
    now support promiscuous mode.

    Donald writes:
    `At first I was planning to do it while implementing either
    the /dev/* or DDI interface, but that's not really the
    correct way to do it. We should only enable multicast or
    promiscuous modes when something wants to look at the
    packets, and shut it down when that application is
    finished, neither of which is strongly related to when the
    hardware is opened or released.

    I'll start by discussing promiscuous mode, which is
    conceptually easy to implement. For most hardware you
    only have to set a register bit, and from then on you get
    every packet on the wire. Well, it's almost that easy;
    for some hardware you have to shut the board (potentially
    dropping a few packet), reconfigure it, and then re-enable
    the ethercard. This is grungy and risky, but the
    alternative seems to be to have every application register
    before you open the ethercard at boot-time.

    OK, so that's easy, so I'll move on something that's not
    quite so obvious: Multicast. It can be done two ways:

<enum>

<item>    Use promiscuous mode, and a packet filter like the
    Berkeley packet filter (BPF). The BPF is a pattern matching
    stack language, where you write a program that picks out the
    addresses you are interested in. Its advantage is that it's
    very general and programmable. Its disadvantage is that there
    is no general way for the kernel to avoid turning on promiscuous
    mode and running every packet on the wire through every registered
    packet filter. See <ref id="bpf" name="The Berkeley Packet Filter">
    for more info.

<item>    Using the built-in multicast filter that most etherchips have.

</enum>

    I guess I should list what a few ethercards/chips provide:

<verb>

    Chip/card  Promiscuous  Multicast filter
    ----------------------------------------
    Seeq8001/3c501  Yes     Binary filter (1)
    3Com/3c509      Yes     Binary filter (1)
    8390            Yes     Autodin II six bit hash (2) (3)
    LANCE           Yes     Autodin II six bit hash (2) (3)
    i82586          Yes     Hidden Autodin II six bit hash (2) (4)

</verb>

<enum>

<item>    These cards claim to have a filter, but it's a simple
    yes/no `accept all multicast packets', or `accept no
    multicast packets'.

<item>    AUTODIN II is the standard ethernet CRC (checksum)
    polynomial. In this scheme multicast addresses are hashed
    and looked up in a hash table. If the corresponding bit
    is enabled, this packet is accepted. Ethernet packets are
    laid out so that the hardware to do this is trivial -- you
    just latch six (usually) bits from the CRC circuit (needed
    anyway for error checking) after the first six octets (the
    destination address), and use them as an index into the
    hash table (six bits -- a 64-bit table).

<item>    These chips use the six bit hash, and must have the
    table computed and loaded by the host. This means the
    kernel must include the CRC code.

<item>    The 82586 uses the six bit hash internally, but it
    computes the hash table itself from a list of multicast
    addresses to accept.

</enum>

    Note that none of these chips do perfect filtering, and we
    still need a middle-level module to do the final
    filtering. Also note that in every case we must keep a
    complete list of accepted multicast addresses to recompute
    the hash table when it changes.

    My first pass at device-level support is detailed in the
    outline driver <tt/skeleton.c/

    It looks like the following:

<code>
    #ifdef HAVE_MULTICAST
    static void set_multicast_list(struct device *dev, int num_addrs,
                     void *addrs);
    #endif
    .
    .

    ethercard_open() {
    ...
    #ifdef HAVE_MULTICAST
        dev->set_multicast_list = &ero;set_multicast_list;
    #endif
    ...

    #ifdef HAVE_MULTICAST
    /* Set or clear the multicast filter for this adaptor.
       num_addrs -- -1      Promiscuous mode, receive all packets
       num_addrs -- 0       Normal mode, clear multicast list
       num_addrs > 0        Multicast mode, receive normal and
            MC packets, and do best-effort filtering.
     */
    static void
    set_multicast_list(struct device *dev, int num_addrs, void *addrs)
    {
    ...

</code>

    Any comments, criticism, etc. are welcome.'

<sect1>The Berkeley Packet Filter (BPF)<label id="bpf">
<p>

    The general idea of the developers is
    that the BPF functionality should not be provided
    by the kernel, but should be in a (hopefully little-used)
    compatibility library.

    For those not in the know: BPF (the Berkeley Packet Filter)
    is an mechanism for specifying to the kernel networking layers
    what packets you are interested in. It's implemented as a
    specialized stack language interpreter built into a low level
    of the networking code. An application passes a program
    written in this language to the kernel, and the kernel runs the
    program on each incoming packet. If the kernel has multiple
    BPF applications, each program is run on each packet.

    The problem is that it's difficult to deduce what kind of
    packets the application is really interested in from the packet
    filter program, so the general solution is to always run the
    filter. Imagine a program that registers a BPF program to
    pick up a low data-rate stream sent to a multicast address.
    Most ethernet cards have a hardware multicast address filter
    implemented as a 64 entry hash table that ignores most unwanted
    multicast packets, so the capability exists to make this a very
    inexpensive operation. But with the BFP the kernel must switch
    the interface to promiscuous mode, receive _all_ packets, and
    run them through this filter. This is work, BTW, that's very
    difficult to account back to the process requesting the packets.

<sect>Networking with a Laptop/Notebook Computer<label id="notebook">
<p>

    There are currently only a few ways to put your laptop on a network.
    You can use the SLIP code (and run at serial line speeds);
    you can buy one of the few laptops that come with a NE2000-compatible
    ethercard; you can get a notebook with a supported
    PCMCIA slot built-in; you can get a laptop with a
    docking station and plug in an ISA ethercard; or you can use a
    parallel port Ethernet adapter such as the D-Link DE-600.

<sect1>Using SLIP
<p>

    This is the cheapest solution, but by far the most difficult. Also,
    you will not get very high transmission rates. Since SLIP is not
    really related to ethernet cards, it will not be discussed further
    here. See the NET-2 Howto.

<sect1>Built in NE2000
<p>

    This solution severely limits your laptop choices and is fairly
    expensive. Be sure to read the specifications carefully, as you
    may find that you will have to buy an additional non-standard
    transceiver to actually put the machine on a network. A good
    idea might be to boot the notebook with a kernel that has
    ne2000 support, and make sure it gets detected and works
    before you lay down your cash.


<sect1>PCMCIA Support<label id="pcmcia">
<p>

    As this area of Linux development is fairly young, I'd suggest
    that you join the LAPTOPS mailing channel. See
    <ref id="mailing-lists" name="Mailing lists...">
    which describes how to join a mailing list channel.

    Try and
    determine exactly what hardware you have (ie. card manufacturer,
    PCMCIA chip controller manufacturer) and then ask on the LAPTOPS
    channel. Regardless, don't expect things to be all that simple.
    Expect to have to fiddle around a bit, and patch kernels, etc.
    Maybe someday you will be able to type `make config' 8-)

    At present, the two PCMCIA chipsets that are supported are
    the Databook TCIC/2 and the intel i82365.

    There is a number of programs on tsx-11.mit.edu in
    /pub/linux/packages/laptops/ that you may find useful. These
    range from PCMCIA Ethercard drivers to programs that communicate
    with the PCMCIA controller chip. Note that these drivers are
    usually tied to a specific PCMCIA chip (ie. the intel 82365
    or the TCIC/2)

    For NE2000 compatible cards, some people have had success
    with just configuring the card under DOS, and then booting
    linux from the DOS command prompt via <tt/loadlin/.

    For those that are <em/net-surfing/ you can try:

    <url url="http://cesdis.gsfc.nasa.gov/linux/pcmcia.html"
        name="Don's PCMCIA Stuff">

    Anyway, the PCMCIA driver problem isn't specific to the Linux world.
    It's been a real disaster in the MS-DOS world. In that world
    people expect the hardware to work if they just follow the manual.
    They might not expect it to interoperate with any other hardware
    or software, or operate optimally, but they do expect that the
    software shipped with the product will function. Many PCMCIA
    adaptors don't even pass this test.

    Things are looking up for Linux users that want PCMCIA support, as
    substantial progress is being made. Pioneering this effort is
    David Hinds. His latest PCMCIA support package can be obtained
    from <tt/cb-iris.stanford.edu/ in the directory
    <tt>/pub/pcmcia/</tt>. Look for a file like
    <tt/pcmcia-cs-X.Y.Z.tgz/ where X.Y.Z will be the latest version
    number. This is most likely uploaded to <tt/tsx-11.mit.edu/ as
    well.

    Note that Donald's PCMCIA enabler works as a user-level
    process, and David Hinds' is a kernel-level solution.
    You may be best served by David's package as it is
    much more widely used.

<sect1>ISA Ethercard in the Docking Station.
<p>

    Docking stations for laptops typically cost about $250
    and provide two full-size ISA slots, two serial and one
    parallel port. Most docking stations are powered off of the
    laptop's batteries, and a few allow adding extra batteries in the
    docking station if you use short ISA cards. You can add an inexpensive
    ethercard and enjoy full-speed ethernet performance.

<sect1>Pocket / parallel port adaptors.
<p>

    The `pocket' ethernet adaptors may also fit your need.
    Until recently they actually costed more than a docking station and
    cheap ethercard, and most tie you down with a wall-brick power supply.
    At present, you can choose from the D-Link, or the RealTek adaptor.
    Most other companies, especially Xircom,
    (see <ref id="xircom" name="Xircom">)
    treat the programming
    information as a trade secret, so support will likely be slow in
    coming. (if ever!)

    Note that the transfer speed will not be all that great
    (perhaps 100kB/s tops?) due to the limitations of the
    parallel port interface.

    See <ref id="de-600" name="DE-600 / DE-620"> and
    <ref id="aep-100" name="RealTek"> for supported pocket adaptors.

    You can sometimes avoid the wall-brick with the adaptors by buying
    or making a cable that draws power from the laptop's keyboard
    port. (See <ref id="aep-100" name="keyboard power">)

<sect>Miscellaneous.<label id="misc">
<p>

    Any other associated stuff that didn't fit in anywhere else gets
    dumped here. It may not be relevant, and it may not be of general
    interest but it is here anyway.

<sect1>Passing Ethernet Arguments to the Kernel<label id="lilo">
<p>

    Here are two generic kernel commands  that can be passed to
    the kernel at boot time. This can be done with LILO, loadlin,
    or any other booting utility that accepts optional arguments.

    For example, if the command was `blah' and it expected 3 arguments
    (say 123, 456, and 789) then, with LILO, you would use:

    <tt>LILO: linux blah=123,456,789</tt>

    <em/Note:/ PCI cards have their i/o and IRQ assigned by the BIOS
    at boot. This means that any boot time arguments for a PCI
    card's IRQ or i/o ports are usually ignored.

    For more information on (and a complete list of) boot time
    arguments, please see the
    <url url="http://sunsite.unc.edu/mdw/HOWTO/BootPrompt-HOWTO.html"
        name="BootPrompt-HOWTO">

<sect2>The <tt/ether/ command<label id="ether">
<p>

    In its most generic form, it looks something like this:

<tscreen>
    ether=IRQ,BASE_ADDR,PARAM_1,PARAM_2,NAME
</tscreen>

    All arguments are optional.  The first non-numeric argument
    is taken as the NAME.

    <bf/IRQ:/
    Obvious.  An IRQ value of `0' (usually the default) means to autoIRQ.
    It's a historical accident that the IRQ setting is first rather than
    the base_addr -- this will be fixed whenever something else changes.

    <bf/BASE_ADDR:/
    Also obvious.  A value of `0' (usually the default) means to
    probe a card-type-specific address list for an ethercard.

    <bf/PARAM_1:/
    It was orginally used as an override value for the memory start
    for a shared-memory ethercard, like the WD80*3.
    Some drivers use the low four bits of this value to set the debug
    message level.  0 -- default, 1-7 -- level 1..7, (7 is maximum
    verbosity)  8 -- level 0 (no messages). Also, the LANCE driver
    uses the low four bits of this value to select the DMA channel.
    Otherwise it uses auto-DMA.

    <bf/PARAM_2:/
    The 3c503 driver uses this to select between the internal and external
    transceivers.  0 -- default/internal, 1 -- AUI external.
    The Cabletron E21XX card also uses the low 4 bits of PARAM_2 to
    select the output media. Otherwise it detects automatically.

    <bf/NAME:/
    Selects the network device the values refer to.  The standard kernel
    uses the names `eth0', `eth1', `eth2' and `eth3' for bus-attached
    ethercards, and `atp0' for the parallel port `pocket' ethernet
    adaptor. The arcnet driver uses `arc0' as its name.
    The default setting is for a single ethercard to be probed for as
    `eth0'.  Multiple cards can only be enabled by explicitly setting up
    their base address using these LILO parameters.
    The 1.0 kernel has LANCE-based ethercards as a special case.  LILO
    arguments are ignored, and LANCE cards are always assigned
    `eth<n>' names starting at `eth0'.  Additional non-LANCE ethercards
    must be explicitly assigned to `eth<n+1>', and the usual `eth0'
    probe disabled with something like  `ether=0,-1,eth0'.
    ( Yes, this is bug. )

<sect2> The <tt/reserve/ command<label id="reserve">
<p>

    This next lilo command is used just like `ether=' above, ie. it is
    appended to the name of the boot select specified in lilo.conf

<tscreen>
    reserve=IO-base,extent{,IO-base,extent...}
</tscreen>

    In some machines it may be necessary to prevent device drivers from
    checking for devices (auto-probing) in a specific region. This may be
    because of poorly designed hardware that causes the boot to <em/freeze/
    (such as some ethercards), hardware that is mistakenly identified,
    hardware whose state is changed by an earlier probe, or merely
    hardware you don't want the kernel to initialize.

    The <tt/reserve/ boot-time argument addresses this problem by specifying
    an I/O port region that shouldn't be probed. That region is reserved
    in the kernel's port registration table as if a device has already
    been found in that region. Note that this mechanism shouldn't be
    necessary on most machines. Only when there is a problem or special
    case would it be necessary to use this.

    The I/O ports in the specified region are protected against
    device probes. This was put in to be used when some driver was
    hanging on a NE2000, or misidentifying some other device
    as its own.  A correct device driver shouldn't probe a reserved
    region, unless another boot argument explicitly specifies that
    it do so.  This implies that <tt/reserve/ will most often be used
    with some other boot argument. Hence if you specify a <tt/reserve/
    region to protect a specific device, you must generally specify
    an explicit probe for that device. Most drivers ignore the port
    registration table if they are given an explicit address.

    For example, the boot line

<tscreen>
    LILO: linux  reserve=0x300,32  ether=0,0x300,eth0
</tscreen>

    keeps all device drivers except the ethercard drivers from
    probing 0x300-0x31f.

    As usual with boot-time specifiers there is an 11 parameter limit,
    thus you can only specify 5 reserved regions per <tt/reserve/ keyword.
    Multiple <tt/reserve/ specifiers will work if you have an unusually
    complicated request.

<sect1>Using the Ethernet Drivers as Modules<label id="modules">
<p>

    At present, all the modules are put in the subdirectory
    <tt/modules/ in your Linux kernel source tree (usually in
    the form of symbolic links). To actually generate the modules,
    you have to type <tt/make modules/ after you have finished building
    the kernel proper. Earlier kernels built them automatically,
    which wasn't fair to those compiling on 4MB 386sx-16 machines.

    Most modules accept parameters like <tt/io=0x340/ and
    <tt/irq=12/ on the <tt/insmod/ command line. It
    is <em/STRONGLY ADVISED/ that you supply these parameters
    to avoid probing for the card. Unlike PCI and EISA devices,
    there is no real safe way to do auto-probing for ISA devices,
    and so it should be avoided when using drivers as modules.

    A list of all the parameters that each module accepts can be
    found in the file:

    <tt>/usr/src/linux/Documentation/networking/net-modules.txt</tt>

    It is recommended that you read that to find out what options
    you can use for your particular card.

    Once you have figured out the arguments/options you are going
    to use, you can insert the module by typing as root:
<code>
    insmod mod_name.o [io=val1[,val2,...]] [irq=val7[,val8,...]]
</code>
    The comma separated value lists are used for modules that
    have the capability to handle multiple devices from a single
    module, such as all the 8390 drivers, and the PLIP driver.

    See the <tt/insmod(8)/ manual page for information on passing
    arguments to the module as it is being loaded.
    The command <tt/lsmod/ will show you what modules are
    loaded, and <tt/rmmod/ will remove them.

    Once a module is inserted, then you can use it just like normal,
    and give <tt/ifconfig/ commands. If you set up your networking at
    boot, then make sure your <tt>/etc/rc*</tt> files run the
    <tt/insmod/ command(s) before getting to the <tt/ifconfig/ command.

    Also note that a <em/busy/ module can't be removed. That means
    that you will have to <tt/ifconfig eth0 down/  (shut down the
    ethernet card) before you can remove the module(s).

<sect2>8390 Based Cards as Modules<label id="mod8390">
<p>

    The present list of 8390 based drivers is: 3c503, ac3200, e2100,
    hp, hp-plus, ne, smc-ultra and wd. These cards were not
    supported as modules for kernel versions prior to 1.3.42.
    (This does not include some
    of the separately distributed PCMCIA drivers (e.g. de-650)
    that are also 8390 based, that have had module support for quite
    some time now.)

    If you have an 8390 based card, you may have to insert
    <em/two/ modules, 8390.o and then the module for your card.
    If 8390 support has been built into your kernel, then you
    will not need to insert the 8390 module. (8390 support is
    built in whenever an 8390 based card is selected to be built
    into the kernel.) Doing a <tt>cat /proc/ksyms | grep 8390</tt>
    will tell you if 8390 support is in your kernel.

    For an 8390 based card, you will have to remove the card module
    before removing the 8390 module, as the 8390 module is used by
    the card module, and thus marked as <em/busy/.

    <label id="multmod8390">

    The 8390 series of network drivers now support multiple
    card systems without reloading the same module multiple
    times (memory efficient!) This is done by specifying
    multiple comma separated values, such as:

<code>
    insmod 3c503.o io=0x280,0x300,0x330,0x350 xcvr=0,1,0,1
</code>

    The above would have the one module controlling four
    3c503 cards, with card 2 and 4 using external
    transcievers.

    It is *STRONGLY RECOMMENDED* that you supply "io="
    instead of autoprobing.  If an "io=" argument is not
    supplied, then the ISA 8390 drivers will complain about
    autoprobing being not recommended, and begrudgingly
    autoprobe for a *SINGLE CARD ONLY* -- if you want to use
    multiple cards you *have* to supply an
    "io=0xNNN,0xQQQ,..." argument.

    The ne module is an exception to the above. A NE2000 is
    essentially an 8390 chip, some bus glue and some RAM.
    Because of this, the ne probe is more invasive than the
    rest, and so at boot we make sure the ne probe is done
    last of all the 8390 cards (so that it won't trip over
    other 8390 based cards) With modules we can't ensure
    that all other non-ne 8390 cards have already been
    found. Because of this, the ne module <em/REQUIRES/ an
    <tt/io=0xNNN/ argument passed in via insmod. It
    will <em/refuse/ to autoprobe.

    It is also worth noting that auto-IRQ probably isn't as
    reliable during the flurry of interrupt activity on a
    running machine. Cards such as the ne2000 that can't get
    the IRQ setting from an EEPROM or configuration register
    are probably best supplied with an <tt/irq=M/ argument as
    well. The file

        <tt>/usr/src/linux/Documentation/networking/net-modules.txt</tt>

    also lists how the interrupt settings are determined for the
    various cards if an <tt/irq=N/ value is not given.

<sect1>Mailing Lists and the Linux Newsgroups<label id="mailing-lists">
<p>

    If you have questions about your ethernet card, please READ this
    document first. You may also want to join the NET channel of the
    Linux mailing lists by sending mail to
    <tt/[e-mail]/ to get help with what lists
    are available, and how to join them.

    Furthermore keep in mind that the NET channel is for development
    discussions only. General questions on how to configure your system
    should be directed to comp.os.linux.setup unless you are actively
    involved in the development of part of the networking for Linux.
    We ask that you <em/please/ respect this general guideline for content.

    Also, the news groups
    <em/comp.sys.ibm.pc.hardware.networking/ and
    <em/comp.dcom.lans.ethernet/ should be used for questions that
    are not Linux specific.

<sect1>Related Documentation<label id="other_docs">
<p>

    Much of this info came from saved postings from the comp.os.linux
    groups, which shows that it is a valuable resource of information.
    Other useful information came from a bunch of small files by Donald
    himself. Of course, if you are setting up an Ethernet card,
    then you will want to read the NET-2 Howto so that you can actually
    configure the software you will use.  Also, if you fancy yourself as
    a bit of a hacker, you can always scrounge some additional info
    from the driver source files as well. There is usually a paragraph
    or two in there describing any important points before any actual
    code starts..

    For those looking for information that is not specific in any way
    to Linux (i.e. what is 10BaseT, what is AUI, what does a hub do, etc.)
    I strongly recommend the <bf/Ethernet-FAQ/  that is
    posted regularly to the newsgroup <em/comp.dcom.lans.ethernet/.
    You can grab it from the following URL:

    <url url="ftp://dorm.rutgers.edu/pub/novell/info_and_docs/Ethernet.FAQ"
        name="Ethernet FAQ">

    You can also have a look at the `Ethernet-HomePage' so to speak,
    which is at the following URL:

    <url url="http://wwwhost.ots.utexas.edu/ethernet/ethernet-home.html"
        name="Ethernet-HomePage">

<sect1>Contributors<label id="contrib">
<p>

    Other people who have contributed (directly or indirectly)
    to the Ethernet-Howto are, in alphabetical order:

<verb>
    Ross Biro               <[e-mail]>
    Alan Cox                <[e-mail]>
    David C. Davies         <[e-mail]>
    Bjorn Ekwall            <[e-mail]>
    David Hinds        <[e-mail]>
    Michael Hipp        <[e-mail]>
    Mike Jagdis             <[e-mail]>
    Duke Kamstra            <[e-mail]>
    Russell Nelson          <[e-mail]>
    Cameron Spitzer         <[e-mail]>
    Dave Roberts            <[e-mail]>
    Glenn Talbott           <[e-mail]>
</verb>


    Many thanks to the above people, and all the other unmentioned
    testers out there.

<sect1>Closing
<p>

    If you have found any glaring typos, or outdated info in this
    document, please let one of us know. It's getting big, and it
    is easy to overlook stuff.

    Thanks,

    Paul Gortmaker, <tt/[e-mail]/

    Donald J. Becker, <tt/[e-mail]/


</article>