path: root/drivers/net/appletalk/ltpc.c

/***    ltpc.c -- a driver for the LocalTalk PC card.
 *
 *      Copyright (c) 1995,1996 Bradford W. Johnson <johns393@maroon.tc.umn.edu>
 *
 *      This software may be used and distributed according to the terms
 *      of the GNU General Public License, incorporated herein by reference.
 *
 *      This is ALPHA code at best.  It may not work for you.  It may
 *      damage your equipment.  It may damage your relations with other
 *      users of your network.  Use it at your own risk!
 *
 *      Based in part on:
 *      skeleton.c      by Donald Becker
 *      dummy.c         by Nick Holloway and Alan Cox
 *      loopback.c      by Ross Biro, Fred van Kampen, Donald Becker
 *      the netatalk source code (UMICH)
 *      lots of work on the card...
 *
 *      I do not have access to the (proprietary) SDK that goes with the card.
 *      If you do, I don't want to know about it, and you can probably write
 *      a better driver yourself anyway.  This does mean that the pieces that
 *      talk to the card are guesswork on my part, so use at your own risk!
 *
 *      This is my first try at writing Linux networking code, and is also
 *      guesswork.  Again, use at your own risk!  (Although on this part, I'd
 *      welcome suggestions)
 *
 *      This is a loadable kernel module which seems to work at my site
 *      consisting of a 1.2.13 linux box running netatalk 1.3.3, and with
 *      the kernel support from 1.3.3b2 including patches routing.patch
 *      and ddp.disappears.from.chooser.  In order to run it, you will need
 *      to patch ddp.c and aarp.c in the kernel, but only a little...
 *
 *      I'm fairly confident that while this is arguably badly written, the
 *      problems that people experience will be "higher level", that is, with
 *      complications in the netatalk code.  The driver itself doesn't do
 *      anything terribly complicated -- it pretends to be an ether device
 *      as far as netatalk is concerned, strips the DDP data out of the ether
 *      frame and builds a LLAP packet to send out the card.  In the other
 *      direction, it receives LLAP frames from the card and builds a fake
 *      ether packet that it then tosses up to the networking code.  You can
 *      argue (correctly) that this is an ugly way to do things, but it
 *      requires a minimal amount of fooling with the code in ddp.c and aarp.c.
 *
 *      The card will do a lot more than is used here -- I *think* it has the
 *      layers up through ATP.  Even if you knew how that part works (which I
 *      don't) it would be a big job to carve up the kernel ddp code to insert
 *      things at a higher level, and probably a bad idea...
 *
 *      There are a number of other cards that do LocalTalk on the PC.  If
 *      nobody finds any insurmountable (at the netatalk level) problems
 *      here, this driver should encourage people to put some work into the
 *      other cards (some of which I gather are still commercially available)
 *      and also to put hooks for LocalTalk into the official ddp code.
 *
 *      I welcome comments and suggestions.  This is my first try at Linux
 *      networking stuff, and there are probably lots of things that I did
 *      suboptimally.  
 *
 ***/

/***
 *
 * $Log: ltpc.c,v $
 * Revision 1.1.2.1  2000/03/01 05:35:07  jgarzik
 * at and tr cleanup
 *
 * Revision 1.8  1997/01/28 05:44:54  bradford
 * Clean up for non-module a little.
 * Hacked about a bit to clean things up - Alan Cox 
 * Probably broken it from the origina 1.8
 *

 * 1998/11/09: David Huggins-Daines <dhd@debian.org>
 * Cleaned up the initialization code to use the standard autoirq methods,
   and to probe for things in the standard order of i/o, irq, dma.  This
   removes the "reset the reset" hack, because I couldn't figure out an
   easy way to get the card to trigger an interrupt after it.
 * Added support for passing configuration parameters on the kernel command
   line and through insmod
 * Changed the device name from "ltalk0" to "lt0", both to conform with the
   other localtalk driver, and to clear up the inconsistency between the
   module and the non-module versions of the driver :-)
 * Added a bunch of comments (I was going to make some enums for the state
   codes and the register offsets, but I'm still not sure exactly what their
   semantics are)
 * Don't poll anymore in interrupt-driven mode
 * It seems to work as a module now (as of 2.1.127), but I don't think
   I'm responsible for that...

 *
 * Revision 1.7  1996/12/12 03:42:33  bradford
 * DMA alloc cribbed from 3c505.c.
 *
 * Revision 1.6  1996/12/12 03:18:58  bradford
 * Added virt_to_bus; works in 2.1.13.
 *
 * Revision 1.5  1996/12/12 03:13:22  root
 * xmitQel initialization -- think through better though.
 *
 * Revision 1.4  1996/06/18 14:55:55  root
 * Change names to ltpc. Tabs. Took a shot at dma alloc,
 * although more needs to be done eventually.
 *
 * Revision 1.3  1996/05/22 14:59:39  root
 * Change dev->open, dev->close to track dummy.c in 1.99.(around 7)
 *
 * Revision 1.2  1996/05/22 14:58:24  root
 * Change tabs mostly.
 *
 * Revision 1.1  1996/04/23 04:45:09  root
 * Initial revision
 *
 * Revision 0.16  1996/03/05 15:59:56  root
 * Change ARPHRD_LOCALTLK definition to the "real" one.
 *
 * Revision 0.15  1996/03/05 06:28:30  root
 * Changes for kernel 1.3.70.  Still need a few patches to kernel, but
 * it's getting closer.
 *
 * Revision 0.14  1996/02/25 17:38:32  root
 * More cleanups.  Removed query to card on get_stats.
 *
 * Revision 0.13  1996/02/21  16:27:40  root
 * Refix debug_print_skb.  Fix mac.raw gotcha that appeared in 1.3.65.
 * Clean up receive code a little.
 *
 * Revision 0.12  1996/02/19  16:34:53  root
 * Fix debug_print_skb.  Kludge outgoing snet to 0 when using startup
 * range.  Change debug to mask: 1 for verbose, 2 for higher level stuff
 * including packet printing, 4 for lower level (card i/o) stuff.
 *
 * Revision 0.11  1996/02/12  15:53:38  root
 * Added router sends (requires new aarp.c patch)
 *
 * Revision 0.10  1996/02/11  00:19:35  root
 * Change source LTALK_LOGGING debug switch to insmod ... debug=2.
 *
 * Revision 0.9  1996/02/10  23:59:35  root
 * Fixed those fixes for 1.2 -- DANGER!  The at.h that comes with netatalk
 * has a *different* definition of struct sockaddr_at than the Linux kernel
 * does.  This is an "insidious and invidious" bug...
 * (Actually the preceding comment is false -- it's the atalk.h in the
 * ancient atalk-0.06 that's the problem)
 *
 * Revision 0.8  1996/02/10 19:09:00  root
 * Merge 1.3 changes.  Tested OK under 1.3.60.
 *
 * Revision 0.7  1996/02/10 17:56:56  root
 * Added debug=1 parameter on insmod for debugging prints.  Tried
 * to fix timer unload on rmmod, but I don't think that's the problem.
 *
 * Revision 0.6  1995/12/31  19:01:09  root
 * Clean up rmmod, irq comments per feedback from Corin Anderson (Thanks Corey!)
 * Clean up initial probing -- sometimes the card wakes up latched in reset.
 *
 * Revision 0.5  1995/12/22  06:03:44  root
 * Added comments in front and cleaned up a bit.
 * This version sent out to people.
 *
 * Revision 0.4  1995/12/18  03:46:44  root
 * Return shortDDP to longDDP fake to 0/0.  Added command structs.
 *
 ***/

/* ltpc jumpers are:
*
*	Interrupts -- set at most one.  If none are set, the driver uses
*	polled mode.  Because the card was developed in the XT era, the
*	origina