Home | History | Annotate | Line # | Download | only in netinet
in_pcb.c revision 1.70
      1 /*	$NetBSD: in_pcb.c,v 1.70 2001/07/25 23:28:02 itojun Exp $	*/
      2 
      3 /*
      4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      5  * All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the project nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  */
     31 
     32 /*-
     33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
     34  * All rights reserved.
     35  *
     36  * This code is derived from software contributed to The NetBSD Foundation
     37  * by Public Access Networks Corporation ("Panix").  It was developed under
     38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
     39  *
     40  * Redistribution and use in source and binary forms, with or without
     41  * modification, are permitted provided that the following conditions
     42  * are met:
     43  * 1. Redistributions of source code must retain the above copyright
     44  *    notice, this list of conditions and the following disclaimer.
     45  * 2. Redistributions in binary form must reproduce the above copyright
     46  *    notice, this list of conditions and the following disclaimer in the
     47  *    documentation and/or other materials provided with the distribution.
     48  * 3. All advertising materials mentioning features or use of this software
     49  *    must display the following acknowledgement:
     50  *	This product includes software developed by the NetBSD
     51  *	Foundation, Inc. and its contributors.
     52  * 4. Neither the name of The NetBSD Foundation nor the names of its
     53  *    contributors may be used to endorse or promote products derived
     54  *    from this software without specific prior written permission.
     55  *
     56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     66  * POSSIBILITY OF SUCH DAMAGE.
     67  */
     68 
     69 /*
     70  * Copyright (c) 1982, 1986, 1991, 1993, 1995
     71  *	The Regents of the University of California.  All rights reserved.
     72  *
     73  * Redistribution and use in source and binary forms, with or without
     74  * modification, are permitted provided that the following conditions
     75  * are met:
     76  * 1. Redistributions of source code must retain the above copyright
     77  *    notice, this list of conditions and the following disclaimer.
     78  * 2. Redistributions in binary form must reproduce the above copyright
     79  *    notice, this list of conditions and the following disclaimer in the
     80  *    documentation and/or other materials provided with the distribution.
     81  * 3. All advertising materials mentioning features or use of this software
     82  *    must display the following acknowledgement:
     83  *	This product includes software developed by the University of
     84  *	California, Berkeley and its contributors.
     85  * 4. Neither the name of the University nor the names of its contributors
     86  *    may be used to endorse or promote products derived from this software
     87  *    without specific prior written permission.
     88  *
     89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     99  * SUCH DAMAGE.
    100  *
    101  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
    102  */
    103 
    104 #include "opt_ipsec.h"
    105 
    106 #include <sys/param.h>
    107 #include <sys/systm.h>
    108 #include <sys/malloc.h>
    109 #include <sys/mbuf.h>
    110 #include <sys/protosw.h>
    111 #include <sys/socket.h>
    112 #include <sys/socketvar.h>
    113 #include <sys/ioctl.h>
    114 #include <sys/errno.h>
    115 #include <sys/time.h>
    116 #include <sys/pool.h>
    117 #include <sys/proc.h>
    118 
    119 #include <net/if.h>
    120 #include <net/route.h>
    121 
    122 #include <netinet/in.h>
    123 #include <netinet/in_systm.h>
    124 #include <netinet/ip.h>
    125 #include <netinet/in_pcb.h>
    126 #include <netinet/in_var.h>
    127 #include <netinet/ip_var.h>
    128 
    129 #ifdef IPSEC
    130 #include <netinet6/ipsec.h>
    131 #include <netkey/key.h>
    132 #endif /* IPSEC */
    133 
    134 struct	in_addr zeroin_addr;
    135 
    136 #define	INPCBHASH_BIND(table, laddr, lport) \
    137 	&(table)->inpt_bindhashtbl[ \
    138 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
    139 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
    140 	&(table)->inpt_connecthashtbl[ \
    141 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
    142 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
    143 
    144 struct inpcb *
    145 	in_pcblookup_port __P((struct inpcbtable *,
    146 	    struct in_addr, u_int, int));
    147 
    148 int	anonportmin = IPPORT_ANONMIN;
    149 int	anonportmax = IPPORT_ANONMAX;
    150 int	lowportmin  = IPPORT_RESERVEDMIN;
    151 int	lowportmax  = IPPORT_RESERVEDMAX;
    152 
    153 struct pool inpcb_pool;
    154 
    155 void
    156 in_pcbinit(table, bindhashsize, connecthashsize)
    157 	struct inpcbtable *table;
    158 	int bindhashsize, connecthashsize;
    159 {
    160 	static int inpcb_pool_initialized;
    161 
    162 	if (inpcb_pool_initialized == 0) {
    163 		pool_init(&inpcb_pool, sizeof(struct inpcb), 0, 0, 0,
    164 		    "inpcbpl", 0, NULL, NULL, M_PCB);
    165 		inpcb_pool_initialized = 1;
    166 	}
    167 
    168 	CIRCLEQ_INIT(&table->inpt_queue);
    169 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
    170 	    M_WAITOK, &table->inpt_bindhash);
    171 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST,
    172 	    M_PCB, M_WAITOK, &table->inpt_connecthash);
    173 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
    174 	table->inpt_lastport = (u_int16_t)anonportmax;
    175 }
    176 
    177 int
    178 in_pcballoc(so, v)
    179 	struct socket *so;
    180 	void *v;
    181 {
    182 	struct inpcbtable *table = v;
    183 	struct inpcb *inp;
    184 	int s;
    185 #ifdef IPSEC
    186 	int error;
    187 #endif
    188 
    189 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
    190 	if (inp == NULL)
    191 		return (ENOBUFS);
    192 	bzero((caddr_t)inp, sizeof(*inp));
    193 	inp->inp_table = table;
    194 	inp->inp_socket = so;
    195 	inp->inp_errormtu = -1;
    196 #ifdef IPSEC
    197 	error = ipsec_init_policy(so, &inp->inp_sp);
    198 	if (error != 0) {
    199 		pool_put(&inpcb_pool, inp);
    200 		return error;
    201 	}
    202 #endif
    203 	so->so_pcb = inp;
    204 	s = splnet();
    205 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, inp, inp_queue);
    206 	in_pcbstate(inp, INP_ATTACHED);
    207 	splx(s);
    208 	return (0);
    209 }
    210 
    211 int
    212 in_pcbbind(v, nam, p)
    213 	void *v;
    214 	struct mbuf *nam;
    215 	struct proc *p;
    216 {
    217 	struct inpcb *inp = v;
    218 	struct socket *so = inp->inp_socket;
    219 	struct inpcbtable *table = inp->inp_table;
    220 	struct sockaddr_in *sin;
    221 	u_int16_t lport = 0;
    222 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
    223 #ifndef IPNOPRIVPORTS
    224 	int error;
    225 #endif
    226 
    227 	if (in_ifaddr.tqh_first == 0)
    228 		return (EADDRNOTAVAIL);
    229 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
    230 		return (EINVAL);
    231 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
    232 		wild = 1;
    233 	if (nam == 0)
    234 		goto noname;
    235 	sin = mtod(nam, struct sockaddr_in *);
    236 	if (nam->m_len != sizeof (*sin))
    237 		return (EINVAL);
    238 #ifdef notdef
    239 	/*
    240 	 * We should check the family, but old programs
    241 	 * incorrectly fail to initialize it.
    242 	 */
    243 	if (sin->sin_family != AF_INET)
    244 		return (EAFNOSUPPORT);
    245 #endif
    246 	lport = sin->sin_port;
    247 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
    248 		/*
    249 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
    250 		 * allow complete duplication of binding if
    251 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
    252 		 * and a multicast address is bound on both
    253 		 * new and duplicated sockets.
    254 		 */
    255 		if (so->so_options & SO_REUSEADDR)
    256 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
    257 	} else if (!in_nullhost(sin->sin_addr)) {
    258 		sin->sin_port = 0;		/* yech... */
    259 		if (ifa_ifwithaddr(sintosa(sin)) == 0)
    260 			return (EADDRNOTAVAIL);
    261 	}
    262 	if (lport) {
    263 		struct inpcb *t;
    264 #ifndef IPNOPRIVPORTS
    265 		/* GROSS */
    266 		if (ntohs(lport) < IPPORT_RESERVED &&
    267 		    (p == 0 || (error = suser(p->p_ucred, &p->p_acflag))))
    268 			return (EACCES);
    269 #endif
    270 		if (so->so_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
    271 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
    272 		/*
    273 		 * XXX:	investigate ramifications of loosening this
    274 		 *	restriction so that as long as both ports have
    275 		 *	SO_REUSEPORT allow the bind
    276 		 */
    277 			if (t &&
    278 			    (!in_nullhost(sin->sin_addr) ||
    279 			     !in_nullhost(t->inp_laddr) ||
    280 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
    281 			    && (so->so_uid != t->inp_socket->so_uid)) {
    282 				return (EADDRINUSE);
    283 			}
    284 		}
    285 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
    286 		if (t && (reuseport & t->inp_socket->so_options) == 0)
    287 			return (EADDRINUSE);
    288 	}
    289 	inp->inp_laddr = sin->sin_addr;
    290 
    291 noname:
    292 	if (lport == 0) {
    293 		int	   cnt;
    294 		u_int16_t  min, max;
    295 		u_int16_t *lastport;
    296 
    297 		if (inp->inp_flags & INP_LOWPORT) {
    298 #ifndef IPNOPRIVPORTS
    299 			if (p == 0 || (error = suser(p->p_ucred, &p->p_acflag)))
    300 				return (EACCES);
    301 #endif
    302 			min = lowportmin;
    303 			max = lowportmax;
    304 			lastport = &table->inpt_lastlow;
    305 		} else {
    306 			min = anonportmin;
    307 			max = anonportmax;
    308 			lastport = &table->inpt_lastport;
    309 		}
    310 		if (min > max) {	/* sanity check */
    311 			u_int16_t swp;
    312 
    313 			swp = min;
    314 			min = max;
    315 			max = swp;
    316 		}
    317 
    318 		lport = *lastport - 1;
    319 		for (cnt = max - min + 1; cnt; cnt--, lport--) {
    320 			if (lport < min || lport > max)
    321 				lport = max;
    322 			if (!in_pcblookup_port(table, inp->inp_laddr,
    323 			    htons(lport), 1))
    324 				goto found;
    325 		}
    326 		if (!in_nullhost(inp->inp_laddr))
    327 			inp->inp_laddr.s_addr = INADDR_ANY;
    328 		return (EAGAIN);
    329 	found:
    330 		inp->inp_flags |= INP_ANONPORT;
    331 		*lastport = lport;
    332 		lport = htons(lport);
    333 	}
    334 	inp->inp_lport = lport;
    335 	in_pcbstate(inp, INP_BOUND);
    336 	return (0);
    337 }
    338 
    339 /*
    340  * Connect from a socket to a specified address.
    341  * Both address and port must be specified in argument sin.
    342  * If don't have a local address for this socket yet,
    343  * then pick one.
    344  */
    345 int
    346 in_pcbconnect(v, nam)
    347 	void *v;
    348 	struct mbuf *nam;
    349 {
    350 	struct inpcb *inp = v;
    351 	struct in_ifaddr *ia;
    352 	struct sockaddr_in *ifaddr = NULL;
    353 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
    354 	int error;
    355 
    356 	if (nam->m_len != sizeof (*sin))
    357 		return (EINVAL);
    358 	if (sin->sin_family != AF_INET)
    359 		return (EAFNOSUPPORT);
    360 	if (sin->sin_port == 0)
    361 		return (EADDRNOTAVAIL);
    362 	if (in_ifaddr.tqh_first != 0) {
    363 		/*
    364 		 * If the destination address is INADDR_ANY,
    365 		 * use any local address (likely loopback).
    366 		 * If the supplied address is INADDR_BROADCAST,
    367 		 * use the broadcast address of an interface
    368 		 * which supports broadcast. (loopback does not)
    369 		 */
    370 
    371 		if (in_nullhost(sin->sin_addr))
    372 			sin->sin_addr = in_ifaddr.tqh_first->ia_addr.sin_addr;
    373 		else if (sin->sin_addr.s_addr == INADDR_BROADCAST)
    374 		    for (ia = in_ifaddr.tqh_first; ia != NULL;
    375 		      ia = ia->ia_list.tqe_next)
    376 			if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
    377 			    sin->sin_addr = ia->ia_broadaddr.sin_addr;
    378 			    break;
    379 			}
    380 	}
    381 	/*
    382 	 * If we haven't bound which network number to use as ours,
    383 	 * we will use the number of the outgoing interface.
    384 	 * This depends on having done a routing lookup, which
    385 	 * we will probably have to do anyway, so we might
    386 	 * as well do it now.  On the other hand if we are
    387 	 * sending to multiple destinations we may have already
    388 	 * done the lookup, so see if we can use the route
    389 	 * from before.  In any case, we only
    390 	 * chose a port number once, even if sending to multiple
    391 	 * destinations.
    392 	 */
    393 	if (in_nullhost(inp->inp_laddr)) {
    394 #if 0
    395 		struct route *ro;
    396 
    397 		ia = (struct in_ifaddr *)0;
    398 		/*
    399 		 * If route is known or can be allocated now,
    400 		 * our src addr is taken from the i/f, else punt.
    401 		 */
    402 		ro = &inp->inp_route;
    403 		if (ro->ro_rt &&
    404 		    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr,
    405 			sin->sin_addr) ||
    406 		    inp->inp_socket->so_options & SO_DONTROUTE)) {
    407 			RTFREE(ro->ro_rt);
    408 			ro->ro_rt = (struct rtentry *)0;
    409 		}
    410 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/
    411 		    (ro->ro_rt == (struct rtentry *)0 ||
    412 		    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    413 			/* No route yet, so try to acquire one */
    414 			ro->ro_dst.sa_family = AF_INET;
    415 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    416 			satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    417 			rtalloc(ro);
    418 		}
    419 		/*
    420 		 * If we found a route, use the address
    421 		 * corresponding to the outgoing interface
    422 		 * unless it is the loopback (in case a route
    423 		 * to our address on another net goes to loopback).
    424 		 *
    425 		 * XXX Is this still true?  Do we care?
    426 		 */
    427 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    428 			ia = ifatoia(ro->ro_rt->rt_ifa);
    429 		if (ia == NULL) {
    430 			u_int16_t fport = sin->sin_port;
    431 
    432 			sin->sin_port = 0;
    433 			ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    434 			sin->sin_port = fport;
    435 			if (ia == 0) {
    436 				/* Find 1st non-loopback AF_INET address */
    437 				for (ia = in_ifaddr.tqh_first ; ia != NULL;
    438 				     ia = ia->ia_list.tqe_next) {
    439 					if ((ia->ia_ifp->if_flags &
    440 					     IFF_LOOPBACK) == 0)
    441 						break;
    442 				}
    443 			}
    444 			if (ia == NULL)
    445 				return (EADDRNOTAVAIL);
    446 		}
    447 		/*
    448 		 * If the destination address is multicast and an outgoing
    449 		 * interface has been set as a multicast option, use the
    450 		 * address of that interface as our source address.
    451 		 */
    452 		if (IN_MULTICAST(sin->sin_addr.s_addr) &&
    453 		    inp->inp_moptions != NULL) {
    454 			struct ip_moptions *imo;
    455 			struct ifnet *ifp;
    456 
    457 			imo = inp->inp_moptions;
    458 			if (imo->imo_multicast_ifp != NULL) {
    459 				ifp = imo->imo_multicast_ifp;
    460 				IFP_TO_IA(ifp, ia);		/* XXX */
    461 				if (ia == 0)
    462 					return (EADDRNOTAVAIL);
    463 			}
    464 		}
    465 		ifaddr = satosin(&ia->ia_addr);
    466 #else
    467 		int error;
    468 		ifaddr = in_selectsrc(sin, &inp->inp_route,
    469 			inp->inp_socket->so_options, inp->inp_moptions, &error);
    470 		if (ifaddr == NULL) {
    471 			if (error == 0)
    472 				error = EADDRNOTAVAIL;
    473 			return error;
    474 		}
    475 #endif
    476 	}
    477 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
    478 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
    479 	    inp->inp_lport) != 0)
    480 		return (EADDRINUSE);
    481 	if (in_nullhost(inp->inp_laddr)) {
    482 		if (inp->inp_lport == 0) {
    483 			error = in_pcbbind(inp, (struct mbuf *)0,
    484 			    (struct proc *)0);
    485 			/*
    486 			 * This used to ignore the return value
    487 			 * completely, but we need to check for
    488 			 * ephemeral port shortage.
    489 			 * XXX Should we check for other errors, too?
    490 			 */
    491 			if (error == EAGAIN)
    492 				return (error);
    493 		}
    494 		inp->inp_laddr = ifaddr->sin_addr;
    495 	}
    496 	inp->inp_faddr = sin->sin_addr;
    497 	inp->inp_fport = sin->sin_port;
    498 	in_pcbstate(inp, INP_CONNECTED);
    499 	return (0);
    500 }
    501 
    502 void
    503 in_pcbdisconnect(v)
    504 	void *v;
    505 {
    506 	struct inpcb *inp = v;
    507 
    508 	inp->inp_faddr = zeroin_addr;
    509 	inp->inp_fport = 0;
    510 	in_pcbstate(inp, INP_BOUND);
    511 	if (inp->inp_socket->so_state & SS_NOFDREF)
    512 		in_pcbdetach(inp);
    513 }
    514 
    515 void
    516 in_pcbdetach(v)
    517 	void *v;
    518 {
    519 	struct inpcb *inp = v;
    520 	struct socket *so = inp->inp_socket;
    521 	int s;
    522 
    523 #ifdef IPSEC
    524 	ipsec4_delete_pcbpolicy(inp);
    525 #endif /*IPSEC*/
    526 	so->so_pcb = 0;
    527 	sofree(so);
    528 	if (inp->inp_options)
    529 		(void)m_free(inp->inp_options);
    530 	if (inp->inp_route.ro_rt)
    531 		rtfree(inp->inp_route.ro_rt);
    532 	ip_freemoptions(inp->inp_moptions);
    533 	s = splnet();
    534 	in_pcbstate(inp, INP_ATTACHED);
    535 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, inp, inp_queue);
    536 	splx(s);
    537 	pool_put(&inpcb_pool, inp);
    538 }
    539 
    540 void
    541 in_setsockaddr(inp, nam)
    542 	struct inpcb *inp;
    543 	struct mbuf *nam;
    544 {
    545 	struct sockaddr_in *sin;
    546 
    547 	nam->m_len = sizeof (*sin);
    548 	sin = mtod(nam, struct sockaddr_in *);
    549 	bzero((caddr_t)sin, sizeof (*sin));
    550 	sin->sin_family = AF_INET;
    551 	sin->sin_len = sizeof(*sin);
    552 	sin->sin_port = inp->inp_lport;
    553 	sin->sin_addr = inp->inp_laddr;
    554 }
    555 
    556 void
    557 in_setpeeraddr(inp, nam)
    558 	struct inpcb *inp;
    559 	struct mbuf *nam;
    560 {
    561 	struct sockaddr_in *sin;
    562 
    563 	nam->m_len = sizeof (*sin);
    564 	sin = mtod(nam, struct sockaddr_in *);
    565 	bzero((caddr_t)sin, sizeof (*sin));
    566 	sin->sin_family = AF_INET;
    567 	sin->sin_len = sizeof(*sin);
    568 	sin->sin_port = inp->inp_fport;
    569 	sin->sin_addr = inp->inp_faddr;
    570 }
    571 
    572 /*
    573  * Pass some notification to all connections of a protocol
    574  * associated with address dst.  The local address and/or port numbers
    575  * may be specified to limit the search.  The "usual action" will be
    576  * taken, depending on the ctlinput cmd.  The caller must filter any
    577  * cmds that are uninteresting (e.g., no error in the map).
    578  * Call the protocol specific routine (if any) to report
    579  * any errors for each matching socket.
    580  *
    581  * Must be called at splsoftnet.
    582  */
    583 int
    584 in_pcbnotify(table, faddr, fport_arg, laddr, lport_arg, errno, notify)
    585 	struct inpcbtable *table;
    586 	struct in_addr faddr, laddr;
    587 	u_int fport_arg, lport_arg;
    588 	int errno;
    589 	void (*notify) __P((struct inpcb *, int));
    590 {
    591 	struct inpcbhead *head;
    592 	struct inpcb *inp, *ninp;
    593 	u_int16_t fport = fport_arg, lport = lport_arg;
    594 	int nmatch;
    595 
    596 	if (in_nullhost(faddr) || notify == 0)
    597 		return (0);
    598 
    599 	nmatch = 0;
    600 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    601 	for (inp = head->lh_first; inp != NULL; inp = ninp) {
    602 		ninp = inp->inp_hash.le_next;
    603 		if (in_hosteq(inp->inp_faddr, faddr) &&
    604 		    inp->inp_fport == fport &&
    605 		    inp->inp_lport == lport &&
    606 		    in_hosteq(inp->inp_laddr, laddr)) {
    607 			(*notify)(inp, errno);
    608 			nmatch++;
    609 		}
    610 	}
    611 	return (nmatch);
    612 }
    613 
    614 void
    615 in_pcbnotifyall(table, faddr, errno, notify)
    616 	struct inpcbtable *table;
    617 	struct in_addr faddr;
    618 	int errno;
    619 	void (*notify) __P((struct inpcb *, int));
    620 {
    621 	struct inpcb *inp, *ninp;
    622 
    623 	if (in_nullhost(faddr) || notify == 0)
    624 		return;
    625 
    626 	for (inp = table->inpt_queue.cqh_first;
    627 	    inp != (struct inpcb *)&table->inpt_queue;
    628 	    inp = ninp) {
    629 		ninp = inp->inp_queue.cqe_next;
    630 		if (in_hosteq(inp->inp_faddr, faddr))
    631 			(*notify)(inp, errno);
    632 	}
    633 }
    634 
    635 void
    636 in_pcbpurgeif0(table, ifp)
    637 	struct inpcbtable *table;
    638 	struct ifnet *ifp;
    639 {
    640 	struct inpcb *inp, *ninp;
    641 	struct ip_moptions *imo;
    642 	int i, gap;
    643 
    644 	for (inp = table->inpt_queue.cqh_first;
    645 	    inp != (struct inpcb *)&table->inpt_queue;
    646 	    inp = ninp) {
    647 		ninp = inp->inp_queue.cqe_next;
    648 		imo = inp->inp_moptions;
    649 		if (imo != NULL) {
    650 			/*
    651 			 * Unselect the outgoing interface if it is being
    652 			 * detached.
    653 			 */
    654 			if (imo->imo_multicast_ifp == ifp)
    655 				imo->imo_multicast_ifp = NULL;
    656 
    657 			/*
    658 			 * Drop multicast group membership if we joined
    659 			 * through the interface being detached.
    660 			 */
    661 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
    662 			    i++) {
    663 				if (imo->imo_membership[i]->inm_ifp == ifp) {
    664 					in_delmulti(imo->imo_membership[i]);
    665 					gap++;
    666 				} else if (gap != 0)
    667 					imo->imo_membership[i - gap] =
    668 					    imo->imo_membership[i];
    669 			}
    670 			imo->imo_num_memberships -= gap;
    671 		}
    672 	}
    673 }
    674 
    675 void
    676 in_pcbpurgeif(table, ifp)
    677 	struct inpcbtable *table;
    678 	struct ifnet *ifp;
    679 {
    680 	struct inpcb *inp, *ninp;
    681 
    682 	for (inp = table->inpt_queue.cqh_first;
    683 	    inp != (struct inpcb *)&table->inpt_queue;
    684 	    inp = ninp) {
    685 		ninp = inp->inp_queue.cqe_next;
    686 		if (inp->inp_route.ro_rt != NULL &&
    687 		    inp->inp_route.ro_rt->rt_ifp == ifp)
    688 			in_rtchange(inp, 0);
    689 	}
    690 }
    691 
    692 /*
    693  * Check for alternatives when higher level complains
    694  * about service problems.  For now, invalidate cached
    695  * routing information.  If the route was created dynamically
    696  * (by a redirect), time to try a default gateway again.
    697  */
    698 void
    699 in_losing(inp)
    700 	struct inpcb *inp;
    701 {
    702 	struct rtentry *rt;
    703 	struct rt_addrinfo info;
    704 
    705 	if ((rt = inp->inp_route.ro_rt)) {
    706 		inp->inp_route.ro_rt = 0;
    707 		bzero((caddr_t)&info, sizeof(info));
    708 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
    709 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    710 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    711 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
    712 		if (rt->rt_flags & RTF_DYNAMIC)
    713 			(void) rtrequest(RTM_DELETE, rt_key(rt),
    714 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
    715 				(struct rtentry **)0);
    716 		else
    717 		/*
    718 		 * A new route can be allocated
    719 		 * the next time output is attempted.
    720 		 */
    721 			rtfree(rt);
    722 	}
    723 }
    724 
    725 /*
    726  * After a routing change, flush old routing
    727  * and allocate a (hopefully) better one.
    728  */
    729 void
    730 in_rtchange(inp, errno)
    731 	struct inpcb *inp;
    732 	int errno;
    733 {
    734 
    735 	if (inp->inp_route.ro_rt) {
    736 		rtfree(inp->inp_route.ro_rt);
    737 		inp->inp_route.ro_rt = 0;
    738 		/*
    739 		 * A new route can be allocated the next time
    740 		 * output is attempted.
    741 		 */
    742 	}
    743 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
    744 }
    745 
    746 struct inpcb *
    747 in_pcblookup_port(table, laddr, lport_arg, lookup_wildcard)
    748 	struct inpcbtable *table;
    749 	struct in_addr laddr;
    750 	u_int lport_arg;
    751 	int lookup_wildcard;
    752 {
    753 	struct inpcb *inp, *match = 0;
    754 	int matchwild = 3, wildcard;
    755 	u_int16_t lport = lport_arg;
    756 
    757 	for (inp = table->inpt_queue.cqh_first;
    758 	    inp != (struct inpcb *)&table->inpt_queue;
    759 	    inp = inp->inp_queue.cqe_next) {
    760 		if (inp->inp_lport != lport)
    761 			continue;
    762 		wildcard = 0;
    763 		if (!in_nullhost(inp->inp_faddr))
    764 			wildcard++;
    765 		if (in_nullhost(inp->inp_laddr)) {
    766 			if (!in_nullhost(laddr))
    767 				wildcard++;
    768 		} else {
    769 			if (in_nullhost(laddr))
    770 				wildcard++;
    771 			else {
    772 				if (!in_hosteq(inp->inp_laddr, laddr))
    773 					continue;
    774 			}
    775 		}
    776 		if (wildcard && !lookup_wildcard)
    777 			continue;
    778 		if (wildcard < matchwild) {
    779 			match = inp;
    780 			matchwild = wildcard;
    781 			if (matchwild == 0)
    782 				break;
    783 		}
    784 	}
    785 	return (match);
    786 }
    787 
    788 #ifdef DIAGNOSTIC
    789 int	in_pcbnotifymiss = 0;
    790 #endif
    791 
    792 struct inpcb *
    793 in_pcblookup_connect(table, faddr, fport_arg, laddr, lport_arg)
    794 	struct inpcbtable *table;
    795 	struct in_addr faddr, laddr;
    796 	u_int fport_arg, lport_arg;
    797 {
    798 	struct inpcbhead *head;
    799 	struct inpcb *inp;
    800 	u_int16_t fport = fport_arg, lport = lport_arg;
    801 
    802 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    803 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    804 		if (in_hosteq(inp->inp_faddr, faddr) &&
    805 		    inp->inp_fport == fport &&
    806 		    inp->inp_lport == lport &&
    807 		    in_hosteq(inp->inp_laddr, laddr))
    808 			goto out;
    809 	}
    810 #ifdef DIAGNOSTIC
    811 	if (in_pcbnotifymiss) {
    812 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
    813 		    ntohl(faddr.s_addr), ntohs(fport),
    814 		    ntohl(laddr.s_addr), ntohs(lport));
    815 	}
    816 #endif
    817 	return (0);
    818 
    819 out:
    820 	/* Move this PCB to the head of hash chain. */
    821 	if (inp != head->lh_first) {
    822 		LIST_REMOVE(inp, inp_hash);
    823 		LIST_INSERT_HEAD(head, inp, inp_hash);
    824 	}
    825 	return (inp);
    826 }
    827 
    828 struct inpcb *
    829 in_pcblookup_bind(table, laddr, lport_arg)
    830 	struct inpcbtable *table;
    831 	struct in_addr laddr;
    832 	u_int lport_arg;
    833 {
    834 	struct inpcbhead *head;
    835 	struct inpcb *inp;
    836 	u_int16_t lport = lport_arg;
    837 
    838 	head = INPCBHASH_BIND(table, laddr, lport);
    839 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    840 		if (inp->inp_lport == lport &&
    841 		    in_hosteq(inp->inp_laddr, laddr))
    842 			goto out;
    843 	}
    844 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
    845 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    846 		if (inp->inp_lport == lport &&
    847 		    in_hosteq(inp->inp_laddr, zeroin_addr))
    848 			goto out;
    849 	}
    850 #ifdef DIAGNOSTIC
    851 	if (in_pcbnotifymiss) {
    852 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
    853 		    ntohl(laddr.s_addr), ntohs(lport));
    854 	}
    855 #endif
    856 	return (0);
    857 
    858 out:
    859 	/* Move this PCB to the head of hash chain. */
    860 	if (inp != head->lh_first) {
    861 		LIST_REMOVE(inp, inp_hash);
    862 		LIST_INSERT_HEAD(head, inp, inp_hash);
    863 	}
    864 	return (inp);
    865 }
    866 
    867 void
    868 in_pcbstate(inp, state)
    869 	struct inpcb *inp;
    870 	int state;
    871 {
    872 
    873 	if (inp->inp_state > INP_ATTACHED)
    874 		LIST_REMOVE(inp, inp_hash);
    875 
    876 	switch (state) {
    877 	case INP_BOUND:
    878 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
    879 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
    880 		break;
    881 	case INP_CONNECTED:
    882 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
    883 		    inp->inp_faddr, inp->inp_fport,
    884 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
    885 		break;
    886 	}
    887 
    888 	inp->inp_state = state;
    889 }
    890 
    891 struct rtentry *
    892 in_pcbrtentry(inp)
    893 	struct inpcb *inp;
    894 {
    895 	struct route *ro;
    896 
    897 	ro = &inp->inp_route;
    898 
    899 	if (ro->ro_rt == NULL) {
    900 		/*
    901 		 * No route yet, so try to acquire one.
    902 		 */
    903 		if (!in_nullhost(inp->inp_faddr)) {
    904 			ro->ro_dst.sa_family = AF_INET;
    905 			ro->ro_dst.sa_len = sizeof(ro->ro_dst);
    906 			satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
    907 			rtalloc(ro);
    908 		}
    909 	}
    910 	return (ro->ro_rt);
    911 }
    912 
    913 struct sockaddr_in *
    914 in_selectsrc(sin, ro, soopts, mopts, errorp)
    915 	struct sockaddr_in *sin;
    916 	struct route *ro;
    917 	int soopts;
    918 	struct ip_moptions *mopts;
    919 	int *errorp;
    920 {
    921 	struct in_ifaddr *ia;
    922 
    923 	ia = (struct in_ifaddr *)0;
    924 	/*
    925 	 * If route is known or can be allocated now,
    926 	 * our src addr is taken from the i/f, else punt.
    927 	 */
    928 	if (ro->ro_rt &&
    929 	    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
    930 	    soopts & SO_DONTROUTE)) {
    931 		RTFREE(ro->ro_rt);
    932 		ro->ro_rt = (struct rtentry *)0;
    933 	}
    934 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
    935 	    (ro->ro_rt == (struct rtentry *)0 ||
    936 	    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    937 		/* No route yet, so try to acquire one */
    938 		ro->ro_dst.sa_family = AF_INET;
    939 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    940 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    941 		rtalloc(ro);
    942 	}
    943 	/*
    944 	 * If we found a route, use the address
    945 	 * corresponding to the outgoing interface
    946 	 * unless it is the loopback (in case a route
    947 	 * to our address on another net goes to loopback).
    948 	 *
    949 	 * XXX Is this still true?  Do we care?
    950 	 */
    951 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    952 		ia = ifatoia(ro->ro_rt->rt_ifa);
    953 	if (ia == NULL) {
    954 		u_int16_t fport = sin->sin_port;
    955 
    956 		sin->sin_port = 0;
    957 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    958 		sin->sin_port = fport;
    959 		if (ia == 0) {
    960 			/* Find 1st non-loopback AF_INET address */
    961 			for (ia = in_ifaddr.tqh_first;
    962 			     ia != NULL;
    963 			     ia = ia->ia_list.tqe_next) {
    964 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
    965 					break;
    966 			}
    967 		}
    968 		if (ia == NULL) {
    969 			*errorp = EADDRNOTAVAIL;
    970 			return NULL;
    971 		}
    972 	}
    973 	/*
    974 	 * If the destination address is multicast and an outgoing
    975 	 * interface has been set as a multicast option, use the
    976 	 * address of that interface as our source address.
    977 	 */
    978 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
    979 		struct ip_moptions *imo;
    980 		struct ifnet *ifp;
    981 
    982 		imo = mopts;
    983 		if (imo->imo_multicast_ifp != NULL) {
    984 			ifp = imo->imo_multicast_ifp;
    985 			IFP_TO_IA(ifp, ia);		/* XXX */
    986 			if (ia == 0) {
    987 				*errorp = EADDRNOTAVAIL;
    988 				return NULL;
    989 			}
    990 		}
    991 	}
    992 	return satosin(&ia->ia_addr);
    993 }
    994