Home | History | Annotate | Line # | Download | only in netinet
in_pcb.c revision 1.72
      1 /*	$NetBSD: in_pcb.c,v 1.72 2001/11/04 20:55:26 matt 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 (TAILQ_FIRST(&in_ifaddr) == 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 (TAILQ_FIRST(&in_ifaddr) != 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 =
    373 			    TAILQ_FIRST(&in_ifaddr)->ia_addr.sin_addr;
    374 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
    375 			TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
    376 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
    377 					sin->sin_addr =
    378 					    ia->ia_broadaddr.sin_addr;
    379 					break;
    380 				}
    381 			}
    382 		}
    383 	}
    384 	/*
    385 	 * If we haven't bound which network number to use as ours,
    386 	 * we will use the number of the outgoing interface.
    387 	 * This depends on having done a routing lookup, which
    388 	 * we will probably have to do anyway, so we might
    389 	 * as well do it now.  On the other hand if we are
    390 	 * sending to multiple destinations we may have already
    391 	 * done the lookup, so see if we can use the route
    392 	 * from before.  In any case, we only
    393 	 * chose a port number once, even if sending to multiple
    394 	 * destinations.
    395 	 */
    396 	if (in_nullhost(inp->inp_laddr)) {
    397 #if 0
    398 		struct route *ro;
    399 
    400 		ia = (struct in_ifaddr *)0;
    401 		/*
    402 		 * If route is known or can be allocated now,
    403 		 * our src addr is taken from the i/f, else punt.
    404 		 */
    405 		ro = &inp->inp_route;
    406 		if (ro->ro_rt &&
    407 		    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr,
    408 			sin->sin_addr) ||
    409 		    inp->inp_socket->so_options & SO_DONTROUTE)) {
    410 			RTFREE(ro->ro_rt);
    411 			ro->ro_rt = (struct rtentry *)0;
    412 		}
    413 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/
    414 		    (ro->ro_rt == (struct rtentry *)0 ||
    415 		    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    416 			/* No route yet, so try to acquire one */
    417 			ro->ro_dst.sa_family = AF_INET;
    418 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    419 			satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    420 			rtalloc(ro);
    421 		}
    422 		/*
    423 		 * If we found a route, use the address
    424 		 * corresponding to the outgoing interface
    425 		 * unless it is the loopback (in case a route
    426 		 * to our address on another net goes to loopback).
    427 		 *
    428 		 * XXX Is this still true?  Do we care?
    429 		 */
    430 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    431 			ia = ifatoia(ro->ro_rt->rt_ifa);
    432 		if (ia == NULL) {
    433 			u_int16_t fport = sin->sin_port;
    434 
    435 			sin->sin_port = 0;
    436 			ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    437 			sin->sin_port = fport;
    438 			if (ia == 0) {
    439 				/* Find 1st non-loopback AF_INET address */
    440 				TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
    441 					if ((ia->ia_ifp->if_flags &
    442 					     IFF_LOOPBACK) == 0)
    443 						break;
    444 				}
    445 			}
    446 			if (ia == NULL)
    447 				return (EADDRNOTAVAIL);
    448 		}
    449 		/*
    450 		 * If the destination address is multicast and an outgoing
    451 		 * interface has been set as a multicast option, use the
    452 		 * address of that interface as our source address.
    453 		 */
    454 		if (IN_MULTICAST(sin->sin_addr.s_addr) &&
    455 		    inp->inp_moptions != NULL) {
    456 			struct ip_moptions *imo;
    457 			struct ifnet *ifp;
    458 
    459 			imo = inp->inp_moptions;
    460 			if (imo->imo_multicast_ifp != NULL) {
    461 				ifp = imo->imo_multicast_ifp;
    462 				IFP_TO_IA(ifp, ia);		/* XXX */
    463 				if (ia == 0)
    464 					return (EADDRNOTAVAIL);
    465 			}
    466 		}
    467 		ifaddr = satosin(&ia->ia_addr);
    468 #else
    469 		int error;
    470 		ifaddr = in_selectsrc(sin, &inp->inp_route,
    471 			inp->inp_socket->so_options, inp->inp_moptions, &error);
    472 		if (ifaddr == NULL) {
    473 			if (error == 0)
    474 				error = EADDRNOTAVAIL;
    475 			return error;
    476 		}
    477 #endif
    478 	}
    479 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
    480 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
    481 	    inp->inp_lport) != 0)
    482 		return (EADDRINUSE);
    483 	if (in_nullhost(inp->inp_laddr)) {
    484 		if (inp->inp_lport == 0) {
    485 			error = in_pcbbind(inp, (struct mbuf *)0,
    486 			    (struct proc *)0);
    487 			/*
    488 			 * This used to ignore the return value
    489 			 * completely, but we need to check for
    490 			 * ephemeral port shortage.
    491 			 * XXX Should we check for other errors, too?
    492 			 */
    493 			if (error == EAGAIN)
    494 				return (error);
    495 		}
    496 		inp->inp_laddr = ifaddr->sin_addr;
    497 	}
    498 	inp->inp_faddr = sin->sin_addr;
    499 	inp->inp_fport = sin->sin_port;
    500 	in_pcbstate(inp, INP_CONNECTED);
    501 #ifdef IPSEC
    502 	if (inp->inp_socket->so_type == SOCK_STREAM)
    503 		ipsec_pcbconn(inp->inp_sp);
    504 #endif
    505 	return (0);
    506 }
    507 
    508 void
    509 in_pcbdisconnect(v)
    510 	void *v;
    511 {
    512 	struct inpcb *inp = v;
    513 
    514 	inp->inp_faddr = zeroin_addr;
    515 	inp->inp_fport = 0;
    516 	in_pcbstate(inp, INP_BOUND);
    517 	if (inp->inp_socket->so_state & SS_NOFDREF)
    518 		in_pcbdetach(inp);
    519 #ifdef IPSEC
    520 	ipsec_pcbdisconn(inp->inp_sp);
    521 #endif
    522 }
    523 
    524 void
    525 in_pcbdetach(v)
    526 	void *v;
    527 {
    528 	struct inpcb *inp = v;
    529 	struct socket *so = inp->inp_socket;
    530 	int s;
    531 
    532 #ifdef IPSEC
    533 	ipsec4_delete_pcbpolicy(inp);
    534 #endif /*IPSEC*/
    535 	so->so_pcb = 0;
    536 	sofree(so);
    537 	if (inp->inp_options)
    538 		(void)m_free(inp->inp_options);
    539 	if (inp->inp_route.ro_rt)
    540 		rtfree(inp->inp_route.ro_rt);
    541 	ip_freemoptions(inp->inp_moptions);
    542 	s = splnet();
    543 	in_pcbstate(inp, INP_ATTACHED);
    544 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, inp, inp_queue);
    545 	splx(s);
    546 	pool_put(&inpcb_pool, inp);
    547 }
    548 
    549 void
    550 in_setsockaddr(inp, nam)
    551 	struct inpcb *inp;
    552 	struct mbuf *nam;
    553 {
    554 	struct sockaddr_in *sin;
    555 
    556 	nam->m_len = sizeof (*sin);
    557 	sin = mtod(nam, struct sockaddr_in *);
    558 	bzero((caddr_t)sin, sizeof (*sin));
    559 	sin->sin_family = AF_INET;
    560 	sin->sin_len = sizeof(*sin);
    561 	sin->sin_port = inp->inp_lport;
    562 	sin->sin_addr = inp->inp_laddr;
    563 }
    564 
    565 void
    566 in_setpeeraddr(inp, nam)
    567 	struct inpcb *inp;
    568 	struct mbuf *nam;
    569 {
    570 	struct sockaddr_in *sin;
    571 
    572 	nam->m_len = sizeof (*sin);
    573 	sin = mtod(nam, struct sockaddr_in *);
    574 	bzero((caddr_t)sin, sizeof (*sin));
    575 	sin->sin_family = AF_INET;
    576 	sin->sin_len = sizeof(*sin);
    577 	sin->sin_port = inp->inp_fport;
    578 	sin->sin_addr = inp->inp_faddr;
    579 }
    580 
    581 /*
    582  * Pass some notification to all connections of a protocol
    583  * associated with address dst.  The local address and/or port numbers
    584  * may be specified to limit the search.  The "usual action" will be
    585  * taken, depending on the ctlinput cmd.  The caller must filter any
    586  * cmds that are uninteresting (e.g., no error in the map).
    587  * Call the protocol specific routine (if any) to report
    588  * any errors for each matching socket.
    589  *
    590  * Must be called at splsoftnet.
    591  */
    592 int
    593 in_pcbnotify(table, faddr, fport_arg, laddr, lport_arg, errno, notify)
    594 	struct inpcbtable *table;
    595 	struct in_addr faddr, laddr;
    596 	u_int fport_arg, lport_arg;
    597 	int errno;
    598 	void (*notify) __P((struct inpcb *, int));
    599 {
    600 	struct inpcbhead *head;
    601 	struct inpcb *inp, *ninp;
    602 	u_int16_t fport = fport_arg, lport = lport_arg;
    603 	int nmatch;
    604 
    605 	if (in_nullhost(faddr) || notify == 0)
    606 		return (0);
    607 
    608 	nmatch = 0;
    609 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    610 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
    611 		ninp = LIST_NEXT(inp, inp_hash);
    612 		if (in_hosteq(inp->inp_faddr, faddr) &&
    613 		    inp->inp_fport == fport &&
    614 		    inp->inp_lport == lport &&
    615 		    in_hosteq(inp->inp_laddr, laddr)) {
    616 			(*notify)(inp, errno);
    617 			nmatch++;
    618 		}
    619 	}
    620 	return (nmatch);
    621 }
    622 
    623 void
    624 in_pcbnotifyall(table, faddr, errno, notify)
    625 	struct inpcbtable *table;
    626 	struct in_addr faddr;
    627 	int errno;
    628 	void (*notify) __P((struct inpcb *, int));
    629 {
    630 	struct inpcb *inp, *ninp;
    631 
    632 	if (in_nullhost(faddr) || notify == 0)
    633 		return;
    634 
    635 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
    636 	    inp != (void *)&table->inpt_queue;
    637 	    inp = ninp) {
    638 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
    639 		if (in_hosteq(inp->inp_faddr, faddr))
    640 			(*notify)(inp, errno);
    641 	}
    642 }
    643 
    644 void
    645 in_pcbpurgeif0(table, ifp)
    646 	struct inpcbtable *table;
    647 	struct ifnet *ifp;
    648 {
    649 	struct inpcb *inp, *ninp;
    650 	struct ip_moptions *imo;
    651 	int i, gap;
    652 
    653 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
    654 	    inp != (void *)&table->inpt_queue;
    655 	    inp = ninp) {
    656 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
    657 		imo = inp->inp_moptions;
    658 		if (imo != NULL) {
    659 			/*
    660 			 * Unselect the outgoing interface if it is being
    661 			 * detached.
    662 			 */
    663 			if (imo->imo_multicast_ifp == ifp)
    664 				imo->imo_multicast_ifp = NULL;
    665 
    666 			/*
    667 			 * Drop multicast group membership if we joined
    668 			 * through the interface being detached.
    669 			 */
    670 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
    671 			    i++) {
    672 				if (imo->imo_membership[i]->inm_ifp == ifp) {
    673 					in_delmulti(imo->imo_membership[i]);
    674 					gap++;
    675 				} else if (gap != 0)
    676 					imo->imo_membership[i - gap] =
    677 					    imo->imo_membership[i];
    678 			}
    679 			imo->imo_num_memberships -= gap;
    680 		}
    681 	}
    682 }
    683 
    684 void
    685 in_pcbpurgeif(table, ifp)
    686 	struct inpcbtable *table;
    687 	struct ifnet *ifp;
    688 {
    689 	struct inpcb *inp, *ninp;
    690 
    691 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
    692 	    inp != (void *)&table->inpt_queue;
    693 	    inp = ninp) {
    694 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
    695 		if (inp->inp_route.ro_rt != NULL &&
    696 		    inp->inp_route.ro_rt->rt_ifp == ifp)
    697 			in_rtchange(inp, 0);
    698 	}
    699 }
    700 
    701 /*
    702  * Check for alternatives when higher level complains
    703  * about service problems.  For now, invalidate cached
    704  * routing information.  If the route was created dynamically
    705  * (by a redirect), time to try a default gateway again.
    706  */
    707 void
    708 in_losing(inp)
    709 	struct inpcb *inp;
    710 {
    711 	struct rtentry *rt;
    712 	struct rt_addrinfo info;
    713 
    714 	if ((rt = inp->inp_route.ro_rt)) {
    715 		inp->inp_route.ro_rt = 0;
    716 		bzero((caddr_t)&info, sizeof(info));
    717 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
    718 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    719 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    720 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
    721 		if (rt->rt_flags & RTF_DYNAMIC)
    722 			(void) rtrequest(RTM_DELETE, rt_key(rt),
    723 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
    724 				(struct rtentry **)0);
    725 		else
    726 		/*
    727 		 * A new route can be allocated
    728 		 * the next time output is attempted.
    729 		 */
    730 			rtfree(rt);
    731 	}
    732 }
    733 
    734 /*
    735  * After a routing change, flush old routing
    736  * and allocate a (hopefully) better one.
    737  */
    738 void
    739 in_rtchange(inp, errno)
    740 	struct inpcb *inp;
    741 	int errno;
    742 {
    743 
    744 	if (inp->inp_route.ro_rt) {
    745 		rtfree(inp->inp_route.ro_rt);
    746 		inp->inp_route.ro_rt = 0;
    747 		/*
    748 		 * A new route can be allocated the next time
    749 		 * output is attempted.
    750 		 */
    751 	}
    752 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
    753 }
    754 
    755 struct inpcb *
    756 in_pcblookup_port(table, laddr, lport_arg, lookup_wildcard)
    757 	struct inpcbtable *table;
    758 	struct in_addr laddr;
    759 	u_int lport_arg;
    760 	int lookup_wildcard;
    761 {
    762 	struct inpcb *inp, *match = 0;
    763 	int matchwild = 3, wildcard;
    764 	u_int16_t lport = lport_arg;
    765 
    766 	CIRCLEQ_FOREACH(inp, &table->inpt_queue, inp_queue) {
    767 		if (inp->inp_lport != lport)
    768 			continue;
    769 		wildcard = 0;
    770 		if (!in_nullhost(inp->inp_faddr))
    771 			wildcard++;
    772 		if (in_nullhost(inp->inp_laddr)) {
    773 			if (!in_nullhost(laddr))
    774 				wildcard++;
    775 		} else {
    776 			if (in_nullhost(laddr))
    777 				wildcard++;
    778 			else {
    779 				if (!in_hosteq(inp->inp_laddr, laddr))
    780 					continue;
    781 			}
    782 		}
    783 		if (wildcard && !lookup_wildcard)
    784 			continue;
    785 		if (wildcard < matchwild) {
    786 			match = inp;
    787 			matchwild = wildcard;
    788 			if (matchwild == 0)
    789 				break;
    790 		}
    791 	}
    792 	return (match);
    793 }
    794 
    795 #ifdef DIAGNOSTIC
    796 int	in_pcbnotifymiss = 0;
    797 #endif
    798 
    799 struct inpcb *
    800 in_pcblookup_connect(table, faddr, fport_arg, laddr, lport_arg)
    801 	struct inpcbtable *table;
    802 	struct in_addr faddr, laddr;
    803 	u_int fport_arg, lport_arg;
    804 {
    805 	struct inpcbhead *head;
    806 	struct inpcb *inp;
    807 	u_int16_t fport = fport_arg, lport = lport_arg;
    808 
    809 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    810 	LIST_FOREACH(inp, head, inp_hash) {
    811 		if (in_hosteq(inp->inp_faddr, faddr) &&
    812 		    inp->inp_fport == fport &&
    813 		    inp->inp_lport == lport &&
    814 		    in_hosteq(inp->inp_laddr, laddr))
    815 			goto out;
    816 	}
    817 #ifdef DIAGNOSTIC
    818 	if (in_pcbnotifymiss) {
    819 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
    820 		    ntohl(faddr.s_addr), ntohs(fport),
    821 		    ntohl(laddr.s_addr), ntohs(lport));
    822 	}
    823 #endif
    824 	return (0);
    825 
    826 out:
    827 	/* Move this PCB to the head of hash chain. */
    828 	if (inp != LIST_FIRST(head)) {
    829 		LIST_REMOVE(inp, inp_hash);
    830 		LIST_INSERT_HEAD(head, inp, inp_hash);
    831 	}
    832 	return (inp);
    833 }
    834 
    835 struct inpcb *
    836 in_pcblookup_bind(table, laddr, lport_arg)
    837 	struct inpcbtable *table;
    838 	struct in_addr laddr;
    839 	u_int lport_arg;
    840 {
    841 	struct inpcbhead *head;
    842 	struct inpcb *inp;
    843 	u_int16_t lport = lport_arg;
    844 
    845 	head = INPCBHASH_BIND(table, laddr, lport);
    846 	LIST_FOREACH(inp, head, inp_hash) {
    847 		if (inp->inp_lport == lport &&
    848 		    in_hosteq(inp->inp_laddr, laddr))
    849 			goto out;
    850 	}
    851 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
    852 	LIST_FOREACH(inp, head, inp_hash) {
    853 		if (inp->inp_lport == lport &&
    854 		    in_hosteq(inp->inp_laddr, zeroin_addr))
    855 			goto out;
    856 	}
    857 #ifdef DIAGNOSTIC
    858 	if (in_pcbnotifymiss) {
    859 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
    860 		    ntohl(laddr.s_addr), ntohs(lport));
    861 	}
    862 #endif
    863 	return (0);
    864 
    865 out:
    866 	/* Move this PCB to the head of hash chain. */
    867 	if (inp != LIST_FIRST(head)) {
    868 		LIST_REMOVE(inp, inp_hash);
    869 		LIST_INSERT_HEAD(head, inp, inp_hash);
    870 	}
    871 	return (inp);
    872 }
    873 
    874 void
    875 in_pcbstate(inp, state)
    876 	struct inpcb *inp;
    877 	int state;
    878 {
    879 
    880 	if (inp->inp_state > INP_ATTACHED)
    881 		LIST_REMOVE(inp, inp_hash);
    882 
    883 	switch (state) {
    884 	case INP_BOUND:
    885 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
    886 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
    887 		break;
    888 	case INP_CONNECTED:
    889 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
    890 		    inp->inp_faddr, inp->inp_fport,
    891 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
    892 		break;
    893 	}
    894 
    895 	inp->inp_state = state;
    896 }
    897 
    898 struct rtentry *
    899 in_pcbrtentry(inp)
    900 	struct inpcb *inp;
    901 {
    902 	struct route *ro;
    903 
    904 	ro = &inp->inp_route;
    905 
    906 	if (ro->ro_rt == NULL) {
    907 		/*
    908 		 * No route yet, so try to acquire one.
    909 		 */
    910 		if (!in_nullhost(inp->inp_faddr)) {
    911 			ro->ro_dst.sa_family = AF_INET;
    912 			ro->ro_dst.sa_len = sizeof(ro->ro_dst);
    913 			satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
    914 			rtalloc(ro);
    915 		}
    916 	}
    917 	return (ro->ro_rt);
    918 }
    919 
    920 struct sockaddr_in *
    921 in_selectsrc(sin, ro, soopts, mopts, errorp)
    922 	struct sockaddr_in *sin;
    923 	struct route *ro;
    924 	int soopts;
    925 	struct ip_moptions *mopts;
    926 	int *errorp;
    927 {
    928 	struct in_ifaddr *ia;
    929 
    930 	ia = (struct in_ifaddr *)0;
    931 	/*
    932 	 * If route is known or can be allocated now,
    933 	 * our src addr is taken from the i/f, else punt.
    934 	 */
    935 	if (ro->ro_rt &&
    936 	    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
    937 	    soopts & SO_DONTROUTE)) {
    938 		RTFREE(ro->ro_rt);
    939 		ro->ro_rt = (struct rtentry *)0;
    940 	}
    941 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
    942 	    (ro->ro_rt == (struct rtentry *)0 ||
    943 	    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    944 		/* No route yet, so try to acquire one */
    945 		ro->ro_dst.sa_family = AF_INET;
    946 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    947 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    948 		rtalloc(ro);
    949 	}
    950 	/*
    951 	 * If we found a route, use the address
    952 	 * corresponding to the outgoing interface
    953 	 * unless it is the loopback (in case a route
    954 	 * to our address on another net goes to loopback).
    955 	 *
    956 	 * XXX Is this still true?  Do we care?
    957 	 */
    958 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    959 		ia = ifatoia(ro->ro_rt->rt_ifa);
    960 	if (ia == NULL) {
    961 		u_int16_t fport = sin->sin_port;
    962 
    963 		sin->sin_port = 0;
    964 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    965 		sin->sin_port = fport;
    966 		if (ia == 0) {
    967 			/* Find 1st non-loopback AF_INET address */
    968 			TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
    969 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
    970 					break;
    971 			}
    972 		}
    973 		if (ia == NULL) {
    974 			*errorp = EADDRNOTAVAIL;
    975 			return NULL;
    976 		}
    977 	}
    978 	/*
    979 	 * If the destination address is multicast and an outgoing
    980 	 * interface has been set as a multicast option, use the
    981 	 * address of that interface as our source address.
    982 	 */
    983 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
    984 		struct ip_moptions *imo;
    985 		struct ifnet *ifp;
    986 
    987 		imo = mopts;
    988 		if (imo->imo_multicast_ifp != NULL) {
    989 			ifp = imo->imo_multicast_ifp;
    990 			IFP_TO_IA(ifp, ia);		/* XXX */
    991 			if (ia == 0) {
    992 				*errorp = EADDRNOTAVAIL;
    993 				return NULL;
    994 			}
    995 		}
    996 	}
    997 	return satosin(&ia->ia_addr);
    998 }
    999