path: root/sys/netiso/tp_inet.c

/*	$OpenBSD: tp_inet.c,v 1.12 2003/07/09 22:03:16 itojun Exp $	*/
/*	$NetBSD: tp_inet.c,v 1.11 1996/03/16 23:13:49 christos Exp $	*/

/*-
 * Copyright (c) 1991, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)tp_inet.c	8.1 (Berkeley) 6/10/93
 */

/***********************************************************
		Copyright IBM Corporation 1987

                      All Rights Reserved

Permission to use, copy, modify, and distribute this software and its
documentation for any purpose and without fee is hereby granted,
provided that the above copyright notice appear in all copies and that
both that copyright notice and this permission notice appear in
supporting documentation, and that the name of IBM not be
used in advertising or publicity pertaining to distribution of the
software without specific, written prior permission.

IBM DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING
ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL
IBM BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR
ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION,
ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
SOFTWARE.

******************************************************************/

/*
 * ARGO Project, Computer Sciences Dept., University of Wisconsin - Madison
 */
/*
 * Here is where you find the inet-dependent code.  We've tried keep all
 * net-level and (primarily) address-family-dependent stuff out of the tp
 * source, and everthing here is reached indirectly through a switch table
 * (struct nl_protosw *) tpcb->tp_nlproto (see tp_pcb.c). The routines here
 * are: in_getsufx: gets transport suffix out of an inpcb structure.
 * in_putsufx: put transport suffix into an inpcb structure. in_putnetaddr:
 * put a whole net addr into an inpcb. in_getnetaddr: get a whole net addr
 * from an inpcb. in_cmpnetaddr: compare a whole net addr from an isopcb.
 * in_recycle_suffix: clear suffix for reuse in inpcb tpip_mtu: figure out
 * what size tpdu to use tpip_input: take a pkt from ip, strip off its ip
 * header, give to tp tpip_output_dg: package a pkt for ip given 2 addresses
 * & some data tpip_output: package a pkt for ip given an inpcb & some data
 */

#ifdef INET

#include <sys/param.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/mbuf.h>
#include <sys/errno.h>
#include <sys/time.h>
#include <sys/systm.h>

#include <net/if.h>

#include <netiso/tp_param.h>
#include <netiso/argo_debug.h>
#include <netiso/tp_stat.h>
#include <netiso/tp_ip.h>
#include <netiso/tp_pcb.h>
#include <netiso/tp_trace.h>
#include <netiso/tp_stat.h>
#include <netiso/tp_tpdu.h>
#include <netiso/tp_var.h>
#include <netinet/in_var.h>

#ifndef ISO
#include <netiso/iso_chksum.c>
#endif

#include <machine/stdarg.h>

/*
 * NAME:		in_getsufx()
 *
 * CALLED FROM: 	pr_usrreq() on PRU_BIND,
 *			PRU_CONNECT, PRU_ACCEPT, and PRU_PEERADDR
 *
 * FUNCTION, ARGUMENTS, and RETURN VALUE:
 * 	Get a transport suffix from an inpcb structure (inp).
 * 	The argument (which) takes the value TP_LOCAL or TP_FOREIGN.
 *
 * RETURNS:		internet port / transport suffix
 *  			(CAST TO AN INT)
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */
void
in_getsufx(v, lenp, data_out, which)
	void	       *v;
	u_short        *lenp;
	caddr_t         data_out;
	int             which;
{
	struct inpcb   *inp = v;
	*lenp = sizeof(u_short);
	switch (which) {
	case TP_LOCAL:
		*(u_short *) data_out = inp->inp_lport;
		return;

	case TP_FOREIGN:
		*(u_short *) data_out = inp->inp_fport;
	}

}

/*
 * NAME:	in_putsufx()
 *
 * CALLED FROM: tp_newsocket(); i.e., when a connection
 *		is being established by an incoming CR_TPDU.
 *
 * FUNCTION, ARGUMENTS:
 * 	Put a transport suffix (found in name) into an inpcb structure (inp).
 * 	The argument (which) takes the value TP_LOCAL or TP_FOREIGN.
 *
 * RETURNS:	Nada
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */
/* ARGSUSED */
void
in_putsufx(v, sufxloc, sufxlen, which)
	void	       *v;
	caddr_t         sufxloc;
	int		sufxlen;
	int             which;
{
	struct inpcb   *inp = v;
	if (which == TP_FOREIGN) {
		bcopy(sufxloc, (caddr_t) & inp->inp_fport, sizeof(inp->inp_fport));
	}
}

/*
 * NAME:	in_recycle_tsuffix()
 *
 * CALLED FROM:	tp.trans whenever we go into REFWAIT state.
 *
 * FUNCTION and ARGUMENT:
 *	 Called when a ref is frozen, to allow the suffix to be reused.
 * 	(inp) is the net level pcb.
 *
 * RETURNS:			Nada
 *
 * SIDE EFFECTS:
 *
 * NOTES:	This really shouldn't have to be done in a NET level pcb
 *	but... for the internet world that just the way it is done in BSD...
 * 	The alternative is to have the port unusable until the reference
 * 	timer goes off.
 */
void
in_recycle_tsuffix(v)
	void *v;
{
	struct inpcb   *inp = v;
	inp->inp_fport = inp->inp_lport = 0;
}

/*
 * NAME:	in_putnetaddr()
 *
 * CALLED FROM:
 * 	tp_newsocket(); i.e., when a connection is being established by an
 * 	incoming CR_TPDU.
 *
 * FUNCTION and ARGUMENTS:
 * 	Copy a whole net addr from a struct sockaddr (name).
 * 	into an inpcb (inp).
 * 	The argument (which) takes values TP_LOCAL or TP_FOREIGN
 *
 * RETURNS:		Nada
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */
void
in_putnetaddr(v, nm, which)
	register void *v;
	struct sockaddr *nm;
	int             which;
{
	register struct inpcb *inp = v;
	struct sockaddr_in *name = (struct sockaddr_in *) nm;
	switch (which) {
	case TP_LOCAL:
		bcopy((caddr_t) & name->sin_addr,
		      (caddr_t) & inp->inp_laddr, sizeof(struct in_addr));
		/* won't work if the dst address (name) is INADDR_ANY */

		break;
	case TP_FOREIGN:
		if (name != (struct sockaddr_in *) 0) {
			bcopy((caddr_t) & name->sin_addr,
			(caddr_t) & inp->inp_faddr, sizeof(struct in_addr));
		}
	}
}

/*
 * NAME:	in_cmpnetaddr()
 *
 * CALLED FROM:
 * 	tp_input() when a connection is being established by an
 * 	incoming CR_TPDU, and considered for interception.
 *
 * FUNCTION and ARGUMENTS:
 * 	Compare a whole net addr from a struct sockaddr (name),
 * 	with that implicitly stored in an inpcb (inp).
 * 	The argument (which) takes values TP_LOCAL or TP_FOREIGN
 *
 * RETURNS:		Nada
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */
int
in_cmpnetaddr(v, nm, which)
	register void *v;
	register struct sockaddr *nm;
	int             which;
{
	register struct inpcb *inp = v;
	struct sockaddr_in *name = (struct sockaddr_in *) nm;
	if (which == TP_LOCAL) {
		if (name->sin_port && name->sin_port != inp->inp_lport)
			return 0;
		return (name->sin_addr.s_addr == inp->inp_laddr.s_addr);
	}
	if (name->sin_port && name->sin_port != inp->inp_fport)
		return 0;
	return (name->sin_addr.s_addr == inp->inp_faddr.s_addr);
}

/*
 * NAME:	in_getnetaddr()
 *
 * CALLED FROM:
 *  pr_usrreq() PRU_SOCKADDR, PRU_ACCEPT, PRU_PEERADDR
 * FUNCTION and ARGUMENTS:
 * 	Copy a whole net addr from an inpcb (inp) into
 * 	an mbuf (name);
 * 	The argument (which) takes values TP_LOCAL or TP_FOREIGN.
 *
 * RETURNS:		Nada
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */

void
in_getnetaddr(v, name, which)
	void *v;
	register struct mbuf *name;
	int             which;
{
	struct inpcb   *inp = v;
	register struct sockaddr_in *sin = mtod(name, struct sockaddr_in *);
	bzero((caddr_t) sin, sizeof(*sin));
	switch (which) {
	case TP_LOCAL:
		sin->sin_addr = inp->inp_laddr;
		sin->sin_port = inp->inp_lport;
		break;
	case TP_FOREIGN:
		sin->sin_addr = inp->inp_faddr;
		sin->sin_port = inp->inp_fport;
		break;
	default:
		return;
	}
	name->m_len = sin->sin_len = sizeof(*sin);
	sin->sin_family = AF_INET;
}

/*
 * NAME: 	tpip_mtu()
 *
 * CALLED FROM:
 *  tp_route_to() on incoming CR, CC, and pr_usrreq() for PRU_CONNECT
 *
 * FUNCTION, ARGUMENTS, and RETURN VALUE:
 *
 * Perform subnetwork dependent part of determining MTU information.
 * It appears that setting a double pointer to the rtentry associated with
 * the destination, and returning the header size for the network protocol
 * suffices.
 *
 * SIDE EFFECTS:
 * Sets tp_routep pointer in pcb.
 *
 * NOTES:
 */
int
tpip_mtu(v)
	void *v;
{
	register struct tp_pcb *tpcb = v;
	struct inpcb   *inp = (struct inpcb *) tpcb->tp_npcb;

#ifdef ARGO_DEBUG
	if (argo_debug[D_CONN]) {
		printf("tpip_mtu(tpcb %p)\n", tpcb);
		printf("tpip_mtu routing to addr 0x%x\n", inp->inp_faddr.s_addr);
	}
#endif
	tpcb->tp_routep = &(inp->inp_route.ro_rt);
	return (sizeof(struct ip));

}

/*
 * NAME:	tpip_output()
 *
 * CALLED FROM:  tp_emit()
 *
 * FUNCTION and ARGUMENTS:
 *  Take a packet(m0) from tp and package it so that ip will accept it.
 *  This means prepending space for the ip header and filling in a few
 *  of the fields.
 *  inp is the inpcb structure; datalen is the length of the data in the
 *  mbuf string m0.
 * RETURNS:
 *  whatever (E*) is returned form the net layer output routine.
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */

int
tpip_output(struct mbuf *m0, ...)
{
	int             datalen;
	struct inpcb   *inp;
	int             nochksum;
	va_list		ap;

	va_start(ap, m0);
	datalen = va_arg(ap, int);
	inp = va_arg(ap, struct inpcb *);
	nochksum = va_arg(ap, int);
	va_end(ap);

	return tpip_output_dg(m0, datalen, &inp->inp_laddr, &inp->inp_faddr,
			      &inp->inp_route, nochksum);
}

/*
 * NAME:	tpip_output_dg()
 *
 * CALLED FROM:  tp_error_emit()
 *
 * FUNCTION and ARGUMENTS:
 *  This is a copy of tpip_output that takes the addresses
 *  instead of a pcb.  It's used by the tp_error_emit, when we
 *  don't have an in_pcb with which to call the normal output rtn.
 *
 * RETURNS:	 ENOBUFS or  whatever (E*) is
 *	returned form the net layer output routine.
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */

/* ARGSUSED */
int
tpip_output_dg(struct mbuf *m0, ...)
{
	int             datalen;
	struct in_addr *laddr, *faddr;
	struct route   *ro;
	int             nochksum;
	register struct mbuf *m;
	register struct ip *ip;
	int             error;
	va_list ap;

	va_start(ap, m0);
	datalen = va_arg(ap, int);
	laddr = va_arg(ap, struct in_addr *);
	faddr = va_arg(ap, struct in_addr *);
	ro = va_arg(ap, struct route *);
	nochksum = va_arg(ap, int);
	va_end(ap);

#ifdef ARGO_DEBUG
	if (argo_debug[D_EMIT]) {
		printf("tpip_output_dg  datalen 0x%x m0 %p\n", datalen, m0);
	}
#endif


	MGETHDR(m, M_DONTWAIT, TPMT_IPHDR);
	if (m == 0) {
		error = ENOBUFS;
		goto bad;
	}
	M_MOVE_HDR(m, m0);
	m->m_next = m0;
	MH_ALIGN(m, sizeof(struct ip));
	m->m_len = sizeof(struct ip);

	ip = mtod(m, struct ip *);
	bzero((caddr_t) ip, sizeof *ip);

	ip->ip_p = IPPROTO_TP;
	m->m_pkthdr.len = = sizeof(struct ip) + datalen;
	ip->ip_len = htons(sizeof(struct ip) + datalen);
	ip->ip_ttl = MAXTTL;
	/*
	 * don't know why you need to set ttl; overlay doesn't even make this
	 * available
	 */

	ip->ip_src = *laddr;
	ip->ip_dst = *faddr;

	IncStat(ts_tpdu_sent);
#ifdef ARGO_DEBUG
	if (argo_debug[D_EMIT]) {
		dump_mbuf(m, "tpip_output_dg before ip_output\n");
	}
#endif

	error = ip_output(m, (struct mbuf *)0, ro, IP_ALLOWBROADCAST, (void *)NULL, (void *)NULL);

#ifdef ARGO_DEBUG
	if (argo_debug[D_EMIT]) {
		printf("tpip_output_dg after ip_output\n");
	}
#endif

	return error;

bad:
	m_freem(m);
	IncStat(ts_send_drop);
	return error;
}

/*
 * NAME:  tpip_input()
 *
 * CALLED FROM:
 * 	ip's input routine, indirectly through the protosw.
 *
 * FUNCTION and ARGUMENTS:
 * Take a packet (m) from ip, strip off the ip header and give it to tp
 *
 * RETURNS:  No return value.
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */
void
tpip_input(struct mbuf *m, ...)
{
	int             iplen;
	struct sockaddr_in src, dst;
	register struct ip *ip;
	int             s = splsoftnet(), hdrlen;
	va_list ap;

	va_start(ap, m);
	iplen = va_arg(ap, int);
	va_end(ap);

	IncStat(ts_pkt_rcvd);

	/*
	 * IP layer has already pulled up the IP header,
	 * but the first byte after the IP header may not be there,
	 * e.g. if you came in via loopback, so you have to do an
	 * m_pullup to before you can even look to see how much you
	 * really need.  The good news is that m_pullup will round
	 * up to almost the next mbuf's worth.
	 */


	if ((m = m_pullup(m, iplen + 1)) == MNULL)
		goto discard;
	CHANGE_MTYPE(m, TPMT_DATA);

	/*
	 * Now pull up the whole tp header:
	 * Unfortunately, there may be IP options to skip past so we
	 * just fetch it as an unsigned char.
	 */
	hdrlen = iplen + 1 + mtod(m, u_char *)[iplen];

	if (m->m_len < hdrlen) {
		if ((m = m_pullup(m, hdrlen)) == MNULL) {
#ifdef ARGO_DEBUG
			if (argo_debug[D_TPINPUT]) {
				printf("tp_input, pullup 2!\n");
			}
#endif
			goto discard;
		}
	}
	/*
	 * cannot use tp_inputprep() here 'cause you don't have quite the
	 * same situation
	 */

#ifdef ARGO_DEBUG
	if (argo_debug[D_TPINPUT]) {
		dump_mbuf(m, "after tpip_input both pullups");
	}
#endif
	/*
	 * m_pullup may have returned a different mbuf
	 */
	ip = mtod(m, struct ip *);

	/*
	 * drop the ip header from the front of the mbuf
	 * this is necessary for the tp checksum
	 */
	m->m_len -= iplen;
	m->m_data += iplen;

	src.sin_addr = *(struct in_addr *) & (ip->ip_src);
	src.sin_family = AF_INET;
	src.sin_len = sizeof(src);
	dst.sin_addr = *(struct in_addr *) & (ip->ip_dst);
	dst.sin_family = AF_INET;
	dst.sin_len = sizeof(dst);

	tp_input(m, sintosa(&src), sintosa(&dst), 0, tpip_output_dg, 0);
	return;

discard:
#ifdef ARGO_DEBUG
	if (argo_debug[D_TPINPUT]) {
		printf("tpip_input DISCARD\n");
	}
#endif
#ifdef TPPT
	if (tp_traceflags[D_TPINPUT]) {
		tptrace(TPPTmisc, "tpip_input DISCARD m", m, 0, 0, 0);
	}
#endif
	IncStat(ts_recv_drop);
	splx(s);
}


#include <sys/protosw.h>
#include <netinet/ip_icmp.h>

/*
 * NAME:	tpin_quench()
 *
 * CALLED FROM: tpip_ctlinput()
 *
 * FUNCTION and ARGUMENTS:  find the tpcb pointer and pass it to tp_quench
 *
 * RETURNS:	Nada
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */

void
tpin_quench(inp, dummy)
	struct inpcb   *inp;
	int dummy;
{
	tp_quench((struct inpcb *) inp->inp_socket->so_pcb, PRC_QUENCH);
}

/*
 * NAME:	tpip_ctlinput()
 *
 * CALLED FROM:
 *  The network layer through the protosw table.
 *
 * FUNCTION and ARGUMENTS:
 *	When clnp gets an ICMP msg this gets called.
 *	It either returns an error status to the user or
 *	causes all connections on this address to be aborted
 *	by calling the appropriate xx_notify() routine.
 *	(cmd) is the type of ICMP error.
 * 	(sa) the address of the sender
 *
 * RETURNS:	 Nothing
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */
void *
tpip_ctlinput(cmd, sa, dummy)
	int             cmd;
	struct sockaddr *sa;
	void *dummy;
{
	struct sockaddr_in *sin = (struct sockaddr_in *) sa;
	extern int      inetctlerrmap[];
	void            (*notify)(struct inpcb *, int);
	int             errno;

	if (cmd < 0 || cmd >= PRC_NCMDS)
		return NULL;
	if (sin->sin_family != AF_INET && sin->sin_family != AF_IMPLINK)
		return NULL;
	if (sin->sin_addr.s_addr == INADDR_ANY)
		return NULL;
	errno = inetctlerrmap[cmd];
	switch (cmd) {

	case PRC_QUENCH:
		notify = tp_quench;
		break;

	case PRC_ROUTEDEAD:
	case PRC_HOSTUNREACH:
	case PRC_UNREACH_NET:
	case PRC_IFDOWN:
	case PRC_HOSTDEAD:
		notify = in_rtchange;
		break;

	default:
		/*
		case	PRC_MSGSIZE:
		case	PRC_UNREACH_HOST:
		case	PRC_UNREACH_PROTOCOL:
		case	PRC_UNREACH_PORT:
		case	PRC_UNREACH_NEEDFRAG:
		case	PRC_UNREACH_SRCFAIL:
		case	PRC_REDIRECT_NET:
		case	PRC_REDIRECT_HOST:
		case	PRC_REDIRECT_TOSNET:
		case	PRC_REDIRECT_TOSHOST:
		case	PRC_TIMXCEED_INTRANS:
		case	PRC_TIMXCEED_REASS:
		case	PRC_PARAMPROB:
		*/
		notify = tpin_abort;
		break;
	}
	in_pcbnotifyall(&tp_inpcb, sintosa(sin), errno, notify);
	return NULL;
}

/*
 * NAME:	tpin_abort()
 *
 * CALLED FROM:
 *	xxx_notify() from tp_ctlinput() when
 *  net level gets some ICMP-equiv. type event.
 *
 * FUNCTION and ARGUMENTS:
 *  Cause the connection to be aborted with some sort of error
 *  reason indicating that the network layer caused the abort.
 *  Fakes an ER TPDU so we can go through the driver.
 *
 * RETURNS:	 Nothing
 *
 * SIDE EFFECTS:
 *
 * NOTES:
 */

void
tpin_abort(inp, n)
	struct inpcb   *inp;
	int             n;
{
	struct tp_event e;

	e.ev_number = ER_TPDU;
	e.TPDU_ATTR(ER).e_reason = ENETRESET;
	tp_driver((struct tp_pcb *) inp->inp_ppcb, &e);
}

#ifdef ARGO_DEBUG
void
dump_inaddr(addr)
	register struct sockaddr_in *addr;
{
	printf("INET: port 0x%x; addr 0x%x\n", addr->sin_port, addr->sin_addr.s_addr);
}
#endif	/* ARGO_DEBUG */
#endif	/* INET */