irnet.h

linux-2.6.15.6

/*
 *	IrNET protocol module : Synchronous PPP over an IrDA socket.
 *
 *		Jean II - HPL `00 - <[e-mail]>
 *
 * This file contains definitions and declarations global to the IrNET module,
 * all grouped in one place...
 * This file is a *private* header, so other modules don't want to know
 * what's in there...
 *
 * Note : as most part of the Linux kernel, this module is available
 * under the GNU General Public License (GPL).
 */

#ifndef IRNET_H
#define IRNET_H

/************************** DOCUMENTATION ***************************/
/*
 * What is IrNET
 * -------------
 * IrNET is a protocol allowing to carry TCP/IP traffic between two
 * IrDA peers in an efficient fashion. It is a thin layer, passing PPP
 * packets to IrTTP and vice versa. It uses PPP in synchronous mode,
 * because IrTTP offer a reliable sequenced packet service (as opposed
 * to a byte stream). In fact, you could see IrNET as carrying TCP/IP
 * in a IrDA socket, using PPP to provide the glue.
 *
 * The main difference with traditional PPP over IrCOMM is that we
 * avoid the framing and serial emulation which are a performance
 * bottleneck. It also allows multipoint communications in a sensible
 * fashion.
 *
 * The main difference with IrLAN is that we use PPP for the link
 * management, which is more standard, interoperable and flexible than
 * the IrLAN protocol. For example, PPP adds authentication,
 * encryption, compression, header compression and automated routing
 * setup. And, as IrNET let PPP do the hard work, the implementation
 * is much simpler than IrLAN.
 *
 * The Linux implementation
 * ------------------------
 * IrNET is written on top of the Linux-IrDA stack, and interface with
 * the generic Linux PPP driver. Because IrNET depend on recent
 * changes of the PPP driver interface, IrNET will work only with very
 * recent kernel (2.3.99-pre6 and up).
 * 
 * The present implementation offer the following features :
 *	o simple user interface using pppd
 *	o efficient implementation (interface directly to PPP and IrTTP)
 *	o addressing (you can specify the name of the IrNET recipient)
 *	o multipoint operation (limited by IrLAP specification)
 *	o information in /proc/net/irda/irnet
 *	o IrNET events on /dev/irnet (for user space daemon)
 *	o IrNET daemon (irnetd) to automatically handle incoming requests
 *	o Windows 2000 compatibility (tested, but need more work)
 * Currently missing :
 *	o Lot's of testing (that's your job)
 *	o Connection retries (may be too hard to do)
 *	o Check pppd persist mode
 *	o User space daemon (to automatically handle incoming requests)
 *
 * The setup is not currently the most easy, but this should get much
 * better when everything will get integrated...
 *
 * Acknowledgements
 * ----------------
 * This module is based on :
 *	o The PPP driver (ppp_synctty/ppp_generic) by Paul Mackerras
 *	o The IrLAN protocol (irlan_common/XXX) by Dag Brattli
 *	o The IrSock interface (af_irda) by Dag Brattli
 *	o Some other bits from the kernel and my drivers...
 * Infinite thanks to those brave souls for providing the infrastructure
 * upon which IrNET is built.
 *
 * Thanks to all my collegues in HP for helping me. In particular,
 * thanks to Salil Pradhan and Bill Serra for W2k testing...
 * Thanks to Luiz Magalhaes for irnetd and much testing...
 *
 * Thanks to Alan Cox for answering lot's of my stupid questions, and
 * to Paul Mackerras answering my questions on how to best integrate
 * IrNET and pppd.
 *
 * Jean II
 *
 * Note on some implementations choices...
 * ------------------------------------
 *	1) Direct interface vs tty/socket
 * I could have used a tty interface to hook to ppp and use the full
 * socket API to connect to IrDA. The code would have been easier to
 * maintain, and maybe the code would have been smaller...
 * Instead, we hook directly to ppp_generic and to IrTTP, which make
 * things more complicated...
 *
 * The first reason is flexibility : this allow us to create IrNET
 * instances on demand (no /dev/ircommX crap) and to allow linkname
 * specification on pppd command line...
 *
 * Second reason is speed optimisation. If you look closely at the
 * transmit and receive paths, you will notice that they are "super lean"
 * (that's why they look ugly), with no function calls and as little data
 * copy and modification as I could...
 *
 *	2) irnetd in user space
 * irnetd is implemented in user space, which is necessary to call pppd.
 * This also give maximum benefits in term of flexibility and customability,
 * and allow to offer the event channel, useful for other stuff like debug.
 *
 * On the other hand, this require a loose coordination between the
 * present module and irnetd. One critical area is how incoming request
 * are handled.
 * When irnet receive an incoming request, it send an event to irnetd and
 * drop the incoming IrNET socket.
 * irnetd start a pppd instance, which create a new IrNET socket. This new
 * socket is then connected in the originating node to the pppd instance.
 * At this point, in the originating node, the first socket is closed.
 *
 * I admit, this is a bit messy and waste some resources. The alternative
 * is caching incoming socket, and that's also quite messy and waste
 * resources.
 * We also make connection time slower. For example, on a 115 kb/s link it
 * adds 60ms to the connection time (770 ms). However, this is slower than
 * the time it takes to fire up pppd on my P133...
 *
 *
 * History :
 * -------
 *
 * v1 - 15.5.00 - Jean II
 *	o Basic IrNET (hook to ppp_generic & IrTTP - incl. multipoint)
 *	o control channel on /dev/irnet (set name/address)
 *	o event channel on /dev/irnet (for user space daemon)
 *
 * v2 - 5.6.00 - Jean II
 *	o Enable DROP_NOT_READY to avoid PPP timeouts & other weirdness...
 *	o Add DISCONNECT_TO event and rename DISCONNECT_FROM.
 *	o Set official device number alloaction on /dev/irnet
 *
 * v3 - 30.8.00 - Jean II
 *	o Update to latest Linux-IrDA changes :
 *		- queue_t => irda_queue_t
 *	o Update to ppp-2.4.0 :
 *		- move irda_irnet_connect from PPPIOCATTACH to TIOCSETD
 *	o Add EXPIRE event (depend on new IrDA-Linux patch)
 *	o Switch from `hashbin_remove' to `hashbin_remove_this' to fix
 *	  a multilink bug... (depend on new IrDA-Linux patch)
 *	o fix a self->daddr to self->raddr in irda_irnet_connect to fix
 *	  another multilink bug (darn !)
 *	o Remove LINKNAME_IOCTL cruft
 *
 * v3b - 31.8.00 - Jean II
 *	o Dump discovery log at event channel startup
 *
 * v4 - 28.9.00 - Jean II
 *	o Fix interaction between poll/select and dump discovery log
 *	o Add IRNET_BLOCKED_LINK event (depend on new IrDA-Linux patch)
 *	o Add IRNET_NOANSWER_FROM event (mostly to help support)
 *	o Release flow control in disconnect_indication
 *	o Block packets while connecting (speed up connections)
 *
 * v5 - 11.01.01 - Jean II
 *	o Init self->max_header_size, just in case...
 *	o Set up ap->chan.hdrlen, to get zero copy on tx side working.
 *	o avoid tx->ttp->flow->ppp->tx->... loop, by checking flow state
 *		Thanks to Christian Gennerat for finding this bug !
 *	---
 *	o Declare the proper MTU/MRU that we can support
 *		(but PPP doesn't read the MTU value :-()
 *	o Declare hashbin HB_NOLOCK instead of HB_LOCAL to avoid
 *		disabling and enabling irq twice
 *
 * v6 - 31.05.01 - Jean II
 *	o Print source address in Found, Discovery, Expiry & Request events
 *	o Print requested source address in /proc/net/irnet
 *	o Change control channel input. Allow multiple commands in one line.
 *	o Add saddr command to change ap->rsaddr (and use that in IrDA)
 *	---
 *	o Make the IrDA connection procedure totally asynchronous.
 *	  Heavy rewrite of the IAS query code and the whole connection
 *	  procedure. Now, irnet_connect() no longer need to be called from
 *	  a process context...
 *	o Enable IrDA connect retries in ppp_irnet_send(). The good thing
 *	  is that IrDA connect retries are directly driven by PPP LCP
 *	  retries (we retry for each LCP packet), so that everything
 *	  is transparently controlled from pppd lcp-max-configure.
 *	o Add ttp_connect flag to prevent rentry on the connect procedure
 *	o Test and fixups to eliminate side effects of retries
 *
 * v7 - 22.08.01 - Jean II
 *	o Cleanup : Change "saddr = 0x0" to "saddr = DEV_ADDR_ANY"
 *	o Fix bug in BLOCK_WHEN_CONNECT introduced in v6 : due to the
 *	  asynchronous IAS query, self->tsap is NULL when PPP send the
 *	  first packet.  This was preventing "connect-delay 0" to work.
 *	  Change the test in ppp_irnet_send() to self->ttp_connect.
 *
 * v8 - 1.11.01 - Jean II
 *	o Tighten the use of self->ttp_connect and self->ttp_open to
 *	  prevent various race conditions.
 *	o Avoid leaking discovery log and skb
 *	o Replace "self" with "server" in irnet_connect_indication() to
 *	  better detect cut'n'paste error ;-)
 *
 * v9 - 29.11.01 - Jean II
 *	o Fix event generation in disconnect indication that I broke in v8
 *	  It was always generation "No-Answer" because I was testing ttp_open
 *	  just after clearing it. *blush*.
 *	o Use newly created irttp_listen() to fix potential crash when LAP
 *	  destroyed before irnet module removed.
 *
 * v10 - 4.3.2 - Jean II
 *	o When receiving a disconnect indication, don't reenable the
 *	  PPP Tx queue, this will trigger a reconnect. Instead, close
 *	  the channel, which will kill pppd...
 *
 * v11 - 20.3.02 - Jean II
 *	o Oops ! v10 fix disabled IrNET retries and passive behaviour.
 *	  Better fix in irnet_disconnect_indication() :
 *	  - if connected, kill pppd via hangup.
 *	  - if not connected, reenable ppp Tx, which trigger IrNET retry.
 *
 * v12 - 10.4.02 - Jean II
 *	o Fix race condition in irnet_connect_indication().
 *	  If the socket was already trying to connect, drop old connection
 *	  and use new one only if acting as primary. See comments.
 *
 * v13 - 30.5.02 - Jean II
 *	o Update module init code
 *
 * v14 - 20.2.03 - Jean II
 *	o Add discovery hint bits in the control channel.
 *	o Remove obsolete MOD_INC/DEC_USE_COUNT in favor of .owner
 *
 * v15 - 7.4.03 - Jean II
 *	o Replace spin_lock_irqsave() with spin_lock_bh() so that we can
 *	  use ppp_unit_number(). It's probably also better overall...
 *	o Disable call to ppp_unregister_channel(), because we can't do it.
 */

/***************************** INCLUDES *****************************/

#include <linux/module.h>

#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/tty.h>
#include <linux/proc_fs.h>
#include <linux/devfs_fs_kernel.h>
#include <linux/netdevice.h>
#include <linux/miscdevice.h>
#include <linux/poll.h>
#include <linux/config.h>
#include <linux/ctype.h>	/* isspace() */
#include <asm/uaccess.h>
#include <linux/init.h>

#include <linux/ppp_defs.h>
#include <linux/if_ppp.h>
#include <linux/ppp_channel.h>

#include <net/irda/irda.h>
#include <net/irda/iriap.h>
#include <net/irda/irias_object.h>
#include <net/irda/irlmp.h>
#include <net/irda/irttp.h>