Home | History | Annotate | Line # | Download | only in netinet
in_pcb.c revision 1.59
      1 /*	$NetBSD: in_pcb.c,v 1.59 1999/07/01 08:12:50 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 <sys/param.h>
    105 #include <sys/systm.h>
    106 #include <sys/malloc.h>
    107 #include <sys/mbuf.h>
    108 #include <sys/protosw.h>
    109 #include <sys/socket.h>
    110 #include <sys/socketvar.h>
    111 #include <sys/ioctl.h>
    112 #include <sys/errno.h>
    113 #include <sys/time.h>
    114 #include <sys/pool.h>
    115 #include <sys/proc.h>
    116 
    117 #include <net/if.h>
    118 #include <net/route.h>
    119 
    120 #include <netinet/in.h>
    121 #include <netinet/in_systm.h>
    122 #include <netinet/ip.h>
    123 #include <netinet/in_pcb.h>
    124 #include <netinet/in_var.h>
    125 #include <netinet/ip_var.h>
    126 
    127 #ifdef IPSEC
    128 #include <netinet6/ipsec.h>
    129 #include <netkey/key.h>
    130 #include <netkey/key_debug.h>
    131 #endif /* IPSEC */
    132 
    133 struct	in_addr zeroin_addr;
    134 
    135 #define	INPCBHASH_BIND(table, laddr, lport) \
    136 	&(table)->inpt_bindhashtbl[ \
    137 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
    138 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
    139 	&(table)->inpt_connecthashtbl[ \
    140 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
    141 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
    142 
    143 struct inpcb *
    144 	in_pcblookup_port __P((struct inpcbtable *,
    145 	    struct in_addr, u_int, int));
    146 
    147 int	anonportmin = IPPORT_ANONMIN;
    148 int	anonportmax = IPPORT_ANONMAX;
    149 
    150 struct pool inpcb_pool;
    151 
    152 void
    153 in_pcbinit(table, bindhashsize, connecthashsize)
    154 	struct inpcbtable *table;
    155 	int bindhashsize, connecthashsize;
    156 {
    157 	static int inpcb_pool_initialized;
    158 
    159 	if (inpcb_pool_initialized == 0) {
    160 		pool_init(&inpcb_pool, sizeof(struct inpcb), 0, 0, 0,
    161 		    "inpcbpl", 0, NULL, NULL, M_PCB);
    162 		inpcb_pool_initialized = 1;
    163 	}
    164 
    165 	CIRCLEQ_INIT(&table->inpt_queue);
    166 	table->inpt_bindhashtbl =
    167 	    hashinit(bindhashsize, M_PCB, M_WAITOK, &table->inpt_bindhash);
    168 	table->inpt_connecthashtbl =
    169 	    hashinit(connecthashsize, M_PCB, M_WAITOK, &table->inpt_connecthash);
    170 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
    171 	table->inpt_lastport = (u_int16_t)anonportmax;
    172 }
    173 
    174 int
    175 in_pcballoc(so, v)
    176 	struct socket *so;
    177 	void *v;
    178 {
    179 	struct inpcbtable *table = v;
    180 	register struct inpcb *inp;
    181 	int s;
    182 
    183 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
    184 	if (inp == NULL)
    185 		return (ENOBUFS);
    186 	bzero((caddr_t)inp, sizeof(*inp));
    187 	inp->inp_table = table;
    188 	inp->inp_socket = so;
    189 	inp->inp_errormtu = -1;
    190 	so->so_pcb = inp;
    191 	s = splnet();
    192 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, inp, inp_queue);
    193 	in_pcbstate(inp, INP_ATTACHED);
    194 	splx(s);
    195 	return (0);
    196 }
    197 
    198 int
    199 in_pcbbind(v, nam, p)
    200 	void *v;
    201 	struct mbuf *nam;
    202 	struct proc *p;
    203 {
    204 	register struct inpcb *inp = v;
    205 	register struct socket *so = inp->inp_socket;
    206 	register struct inpcbtable *table = inp->inp_table;
    207 	register struct sockaddr_in *sin;
    208 	u_int16_t lport = 0;
    209 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
    210 #ifndef IPNOPRIVPORTS
    211 	int error;
    212 #endif
    213 
    214 	if (in_ifaddr.tqh_first == 0)
    215 		return (EADDRNOTAVAIL);
    216 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
    217 		return (EINVAL);
    218 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
    219 		wild = 1;
    220 	if (nam == 0)
    221 		goto noname;
    222 	sin = mtod(nam, struct sockaddr_in *);
    223 	if (nam->m_len != sizeof (*sin))
    224 		return (EINVAL);
    225 #ifdef notdef
    226 	/*
    227 	 * We should check the family, but old programs
    228 	 * incorrectly fail to initialize it.
    229 	 */
    230 	if (sin->sin_family != AF_INET)
    231 		return (EAFNOSUPPORT);
    232 #endif
    233 	lport = sin->sin_port;
    234 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
    235 		/*
    236 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
    237 		 * allow complete duplication of binding if
    238 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
    239 		 * and a multicast address is bound on both
    240 		 * new and duplicated sockets.
    241 		 */
    242 		if (so->so_options & SO_REUSEADDR)
    243 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
    244 	} else if (!in_nullhost(sin->sin_addr)) {
    245 		sin->sin_port = 0;		/* yech... */
    246 		if (ifa_ifwithaddr(sintosa(sin)) == 0)
    247 			return (EADDRNOTAVAIL);
    248 	}
    249 	if (lport) {
    250 		struct inpcb *t;
    251 #ifndef IPNOPRIVPORTS
    252 		/* GROSS */
    253 		if (ntohs(lport) < IPPORT_RESERVED &&
    254 		    (p == 0 || (error = suser(p->p_ucred, &p->p_acflag))))
    255 			return (EACCES);
    256 #endif
    257 		if (so->so_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
    258 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
    259 		/*
    260 		 * XXX:	investigate ramifications of loosening this
    261 		 *	restriction so that as long as both ports have
    262 		 *	SO_REUSEPORT allow the bind
    263 		 */
    264 			if (t &&
    265 			    (!in_nullhost(sin->sin_addr) ||
    266 			     !in_nullhost(t->inp_laddr) ||
    267 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
    268 			    && (so->so_uid != t->inp_socket->so_uid)) {
    269 				return (EADDRINUSE);
    270 			}
    271 		}
    272 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
    273 		if (t && (reuseport & t->inp_socket->so_options) == 0)
    274 			return (EADDRINUSE);
    275 	}
    276 	inp->inp_laddr = sin->sin_addr;
    277 
    278 noname:
    279 	if (lport == 0) {
    280 		int	   cnt;
    281 		u_int16_t  min, max;
    282 		u_int16_t *lastport;
    283 
    284 		if (inp->inp_flags & INP_LOWPORT) {
    285 #ifndef IPNOPRIVPORTS
    286 			if (p == 0 || (error = suser(p->p_ucred, &p->p_acflag)))
    287 				return (EACCES);
    288 #endif
    289 			min = IPPORT_RESERVEDMIN;
    290 			max = IPPORT_RESERVEDMAX;
    291 			lastport = &table->inpt_lastlow;
    292 		} else {
    293 			min = anonportmin;
    294 			max = anonportmax;
    295 			lastport = &table->inpt_lastport;
    296 		}
    297 		if (min > max) {	/* sanity check */
    298 			u_int16_t swp;
    299 
    300 			swp = min;
    301 			min = max;
    302 			max = swp;
    303 		}
    304 
    305 		lport = *lastport - 1;
    306 		for (cnt = max - min + 1; cnt; cnt--, lport--) {
    307 			if (lport < min || lport > max)
    308 				lport = max;
    309 			if (!in_pcblookup_port(table, inp->inp_laddr,
    310 			    htons(lport), 1))
    311 				goto found;
    312 		}
    313 		if (!in_nullhost(inp->inp_laddr))
    314 			inp->inp_laddr.s_addr = INADDR_ANY;
    315 		return (EAGAIN);
    316 	found:
    317 		inp->inp_flags |= INP_ANONPORT;
    318 		*lastport = lport;
    319 		lport = htons(lport);
    320 	}
    321 	inp->inp_lport = lport;
    322 	in_pcbstate(inp, INP_BOUND);
    323 	return (0);
    324 }
    325 
    326 /*
    327  * Connect from a socket to a specified address.
    328  * Both address and port must be specified in argument sin.
    329  * If don't have a local address for this socket yet,
    330  * then pick one.
    331  */
    332 int
    333 in_pcbconnect(v, nam)
    334 	register void *v;
    335 	struct mbuf *nam;
    336 {
    337 	register struct inpcb *inp = v;
    338 	struct in_ifaddr *ia;
    339 	struct sockaddr_in *ifaddr = NULL;
    340 	register struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
    341 	int error;
    342 
    343 	if (nam->m_len != sizeof (*sin))
    344 		return (EINVAL);
    345 	if (sin->sin_family != AF_INET)
    346 		return (EAFNOSUPPORT);
    347 	if (sin->sin_port == 0)
    348 		return (EADDRNOTAVAIL);
    349 	if (in_ifaddr.tqh_first != 0) {
    350 		/*
    351 		 * If the destination address is INADDR_ANY,
    352 		 * use any local address (likely loopback).
    353 		 * If the supplied address is INADDR_BROADCAST,
    354 		 * use the broadcast address of an interface
    355 		 * which supports broadcast. (loopback does not)
    356 		 */
    357 
    358 		if (in_nullhost(sin->sin_addr))
    359 			sin->sin_addr = in_ifaddr.tqh_first->ia_addr.sin_addr;
    360 		else if (sin->sin_addr.s_addr == INADDR_BROADCAST)
    361 		    for (ia = in_ifaddr.tqh_first; ia != NULL;
    362 		      ia = ia->ia_list.tqe_next)
    363 			if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
    364 			    sin->sin_addr = ia->ia_broadaddr.sin_addr;
    365 			    break;
    366 			}
    367 	}
    368 	/*
    369 	 * If we haven't bound which network number to use as ours,
    370 	 * we will use the number of the outgoing interface.
    371 	 * This depends on having done a routing lookup, which
    372 	 * we will probably have to do anyway, so we might
    373 	 * as well do it now.  On the other hand if we are
    374 	 * sending to multiple destinations we may have already
    375 	 * done the lookup, so see if we can use the route
    376 	 * from before.  In any case, we only
    377 	 * chose a port number once, even if sending to multiple
    378 	 * destinations.
    379 	 */
    380 	if (in_nullhost(inp->inp_laddr)) {
    381 #if 0
    382 		register struct route *ro;
    383 
    384 		ia = (struct in_ifaddr *)0;
    385 		/*
    386 		 * If route is known or can be allocated now,
    387 		 * our src addr is taken from the i/f, else punt.
    388 		 */
    389 		ro = &inp->inp_route;
    390 		if (ro->ro_rt &&
    391 		    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr,
    392 			sin->sin_addr) ||
    393 		    inp->inp_socket->so_options & SO_DONTROUTE)) {
    394 			RTFREE(ro->ro_rt);
    395 			ro->ro_rt = (struct rtentry *)0;
    396 		}
    397 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/
    398 		    (ro->ro_rt == (struct rtentry *)0 ||
    399 		    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    400 			/* No route yet, so try to acquire one */
    401 			ro->ro_dst.sa_family = AF_INET;
    402 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    403 			satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    404 			rtalloc(ro);
    405 		}
    406 		/*
    407 		 * If we found a route, use the address
    408 		 * corresponding to the outgoing interface
    409 		 * unless it is the loopback (in case a route
    410 		 * to our address on another net goes to loopback).
    411 		 *
    412 		 * XXX Is this still true?  Do we care?
    413 		 */
    414 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    415 			ia = ifatoia(ro->ro_rt->rt_ifa);
    416 		if (ia == 0) {
    417 		    u_int16_t fport = sin->sin_port;
    418 
    419 		    sin->sin_port = 0;
    420 		    ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    421 		    sin->sin_port = fport;
    422 		    if (ia == 0)
    423 			/* Find 1st non-loopback AF_INET address */
    424 			for (ia = in_ifaddr.tqh_first ; ia != NULL ;
    425 				ia = ia->ia_list.tqe_next)
    426 			    if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
    427 				break;
    428 		    if (ia == 0)
    429 			return (EADDRNOTAVAIL);
    430 		}
    431 		/*
    432 		 * If the destination address is multicast and an outgoing
    433 		 * interface has been set as a multicast option, use the
    434 		 * address of that interface as our source address.
    435 		 */
    436 		if (IN_MULTICAST(sin->sin_addr.s_addr) &&
    437 		    inp->inp_moptions != NULL) {
    438 			struct ip_moptions *imo;
    439 			struct ifnet *ifp;
    440 
    441 			imo = inp->inp_moptions;
    442 			if (imo->imo_multicast_ifp != NULL) {
    443 				ifp = imo->imo_multicast_ifp;
    444 				IFP_TO_IA(ifp, ia);		/* XXX */
    445 				if (ia == 0)
    446 					return (EADDRNOTAVAIL);
    447 			}
    448 		}
    449 		ifaddr = satosin(&ia->ia_addr);
    450 #else
    451 		int error;
    452 		ifaddr = in_selectsrc(sin, &inp->inp_route,
    453 			inp->inp_socket->so_options, inp->inp_moptions, &error);
    454 		if (ifaddr == NULL) {
    455 			if (error == 0)
    456 				error = EADDRNOTAVAIL;
    457 			return error;
    458 		}
    459 #endif
    460 	}
    461 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
    462 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
    463 	    inp->inp_lport) != 0)
    464 		return (EADDRINUSE);
    465 	if (in_nullhost(inp->inp_laddr)) {
    466 		if (inp->inp_lport == 0) {
    467 			error = in_pcbbind(inp, (struct mbuf *)0,
    468 			    (struct proc *)0);
    469 			/*
    470 			 * This used to ignore the return value
    471 			 * completely, but we need to check for
    472 			 * ephemeral port shortage.
    473 			 * XXX Should we check for other errors, too?
    474 			 */
    475 			if (error == EAGAIN)
    476 				return (error);
    477 		}
    478 		inp->inp_laddr = ifaddr->sin_addr;
    479 	}
    480 	inp->inp_faddr = sin->sin_addr;
    481 	inp->inp_fport = sin->sin_port;
    482 	in_pcbstate(inp, INP_CONNECTED);
    483 	return (0);
    484 }
    485 
    486 void
    487 in_pcbdisconnect(v)
    488 	void *v;
    489 {
    490 	struct inpcb *inp = v;
    491 
    492 	inp->inp_faddr = zeroin_addr;
    493 	inp->inp_fport = 0;
    494 	in_pcbstate(inp, INP_BOUND);
    495 	if (inp->inp_socket->so_state & SS_NOFDREF)
    496 		in_pcbdetach(inp);
    497 }
    498 
    499 void
    500 in_pcbdetach(v)
    501 	void *v;
    502 {
    503 	struct inpcb *inp = v;
    504 	struct socket *so = inp->inp_socket;
    505 	int s;
    506 
    507 #ifdef IPSEC
    508 	if (so->so_pcb) {
    509 		KEYDEBUG(KEYDEBUG_KEY_STAMP,
    510 			printf("DP call free SO=%p from in_pcbdetach\n", so));
    511 		key_freeso(so);
    512 	}
    513 	ipsec4_delete_pcbpolicy(inp);
    514 #endif /*IPSEC*/
    515 	so->so_pcb = 0;
    516 	sofree(so);
    517 	if (inp->inp_options)
    518 		(void)m_free(inp->inp_options);
    519 	if (inp->inp_route.ro_rt)
    520 		rtfree(inp->inp_route.ro_rt);
    521 	ip_freemoptions(inp->inp_moptions);
    522 	s = splnet();
    523 	in_pcbstate(inp, INP_ATTACHED);
    524 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, inp, inp_queue);
    525 	splx(s);
    526 	pool_put(&inpcb_pool, inp);
    527 }
    528 
    529 void
    530 in_setsockaddr(inp, nam)
    531 	register struct inpcb *inp;
    532 	struct mbuf *nam;
    533 {
    534 	register struct sockaddr_in *sin;
    535 
    536 	nam->m_len = sizeof (*sin);
    537 	sin = mtod(nam, struct sockaddr_in *);
    538 	bzero((caddr_t)sin, sizeof (*sin));
    539 	sin->sin_family = AF_INET;
    540 	sin->sin_len = sizeof(*sin);
    541 	sin->sin_port = inp->inp_lport;
    542 	sin->sin_addr = inp->inp_laddr;
    543 }
    544 
    545 void
    546 in_setpeeraddr(inp, nam)
    547 	struct inpcb *inp;
    548 	struct mbuf *nam;
    549 {
    550 	register struct sockaddr_in *sin;
    551 
    552 	nam->m_len = sizeof (*sin);
    553 	sin = mtod(nam, struct sockaddr_in *);
    554 	bzero((caddr_t)sin, sizeof (*sin));
    555 	sin->sin_family = AF_INET;
    556 	sin->sin_len = sizeof(*sin);
    557 	sin->sin_port = inp->inp_fport;
    558 	sin->sin_addr = inp->inp_faddr;
    559 }
    560 
    561 /*
    562  * Pass some notification to all connections of a protocol
    563  * associated with address dst.  The local address and/or port numbers
    564  * may be specified to limit the search.  The "usual action" will be
    565  * taken, depending on the ctlinput cmd.  The caller must filter any
    566  * cmds that are uninteresting (e.g., no error in the map).
    567  * Call the protocol specific routine (if any) to report
    568  * any errors for each matching socket.
    569  *
    570  * Must be called at splsoftnet.
    571  */
    572 int
    573 in_pcbnotify(table, faddr, fport_arg, laddr, lport_arg, errno, notify)
    574 	struct inpcbtable *table;
    575 	struct in_addr faddr, laddr;
    576 	u_int fport_arg, lport_arg;
    577 	int errno;
    578 	void (*notify) __P((struct inpcb *, int));
    579 {
    580 	struct inpcbhead *head;
    581 	register struct inpcb *inp, *ninp;
    582 	u_int16_t fport = fport_arg, lport = lport_arg;
    583 	int nmatch;
    584 
    585 	if (in_nullhost(faddr) || notify == 0)
    586 		return (0);
    587 
    588 	nmatch = 0;
    589 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    590 	for (inp = head->lh_first; inp != NULL; inp = ninp) {
    591 		ninp = inp->inp_hash.le_next;
    592 		if (in_hosteq(inp->inp_faddr, faddr) &&
    593 		    inp->inp_fport == fport &&
    594 		    inp->inp_lport == lport &&
    595 		    in_hosteq(inp->inp_laddr, laddr)) {
    596 			(*notify)(inp, errno);
    597 			nmatch++;
    598 		}
    599 	}
    600 	return (nmatch);
    601 }
    602 
    603 void
    604 in_pcbnotifyall(table, faddr, errno, notify)
    605 	struct inpcbtable *table;
    606 	struct in_addr faddr;
    607 	int errno;
    608 	void (*notify) __P((struct inpcb *, int));
    609 {
    610 	register struct inpcb *inp, *ninp;
    611 
    612 	if (in_nullhost(faddr) || notify == 0)
    613 		return;
    614 
    615 	for (inp = table->inpt_queue.cqh_first;
    616 	    inp != (struct inpcb *)&table->inpt_queue;
    617 	    inp = ninp) {
    618 		ninp = inp->inp_queue.cqe_next;
    619 		if (in_hosteq(inp->inp_faddr, faddr))
    620 			(*notify)(inp, errno);
    621 	}
    622 }
    623 
    624 /*
    625  * Check for alternatives when higher level complains
    626  * about service problems.  For now, invalidate cached
    627  * routing information.  If the route was created dynamically
    628  * (by a redirect), time to try a default gateway again.
    629  */
    630 void
    631 in_losing(inp)
    632 	struct inpcb *inp;
    633 {
    634 	register struct rtentry *rt;
    635 	struct rt_addrinfo info;
    636 
    637 	if ((rt = inp->inp_route.ro_rt)) {
    638 		inp->inp_route.ro_rt = 0;
    639 		bzero((caddr_t)&info, sizeof(info));
    640 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
    641 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    642 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    643 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
    644 		if (rt->rt_flags & RTF_DYNAMIC)
    645 			(void) rtrequest(RTM_DELETE, rt_key(rt),
    646 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
    647 				(struct rtentry **)0);
    648 		else
    649 		/*
    650 		 * A new route can be allocated
    651 		 * the next time output is attempted.
    652 		 */
    653 			rtfree(rt);
    654 	}
    655 }
    656 
    657 /*
    658  * After a routing change, flush old routing
    659  * and allocate a (hopefully) better one.
    660  */
    661 void
    662 in_rtchange(inp, errno)
    663 	register struct inpcb *inp;
    664 	int errno;
    665 {
    666 
    667 	if (inp->inp_route.ro_rt) {
    668 		rtfree(inp->inp_route.ro_rt);
    669 		inp->inp_route.ro_rt = 0;
    670 		/*
    671 		 * A new route can be allocated the next time
    672 		 * output is attempted.
    673 		 */
    674 	}
    675 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
    676 }
    677 
    678 struct inpcb *
    679 in_pcblookup_port(table, laddr, lport_arg, lookup_wildcard)
    680 	struct inpcbtable *table;
    681 	struct in_addr laddr;
    682 	u_int lport_arg;
    683 	int lookup_wildcard;
    684 {
    685 	register struct inpcb *inp, *match = 0;
    686 	int matchwild = 3, wildcard;
    687 	u_int16_t lport = lport_arg;
    688 
    689 	for (inp = table->inpt_queue.cqh_first;
    690 	    inp != (struct inpcb *)&table->inpt_queue;
    691 	    inp = inp->inp_queue.cqe_next) {
    692 		if (inp->inp_lport != lport)
    693 			continue;
    694 		wildcard = 0;
    695 		if (!in_nullhost(inp->inp_faddr))
    696 			wildcard++;
    697 		if (in_nullhost(inp->inp_laddr)) {
    698 			if (!in_nullhost(laddr))
    699 				wildcard++;
    700 		} else {
    701 			if (in_nullhost(laddr))
    702 				wildcard++;
    703 			else {
    704 				if (!in_hosteq(inp->inp_laddr, laddr))
    705 					continue;
    706 			}
    707 		}
    708 		if (wildcard && !lookup_wildcard)
    709 			continue;
    710 		if (wildcard < matchwild) {
    711 			match = inp;
    712 			matchwild = wildcard;
    713 			if (matchwild == 0)
    714 				break;
    715 		}
    716 	}
    717 	return (match);
    718 }
    719 
    720 #ifdef DIAGNOSTIC
    721 int	in_pcbnotifymiss = 0;
    722 #endif
    723 
    724 struct inpcb *
    725 in_pcblookup_connect(table, faddr, fport_arg, laddr, lport_arg)
    726 	struct inpcbtable *table;
    727 	struct in_addr faddr, laddr;
    728 	u_int fport_arg, lport_arg;
    729 {
    730 	struct inpcbhead *head;
    731 	register struct inpcb *inp;
    732 	u_int16_t fport = fport_arg, lport = lport_arg;
    733 
    734 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    735 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    736 		if (in_hosteq(inp->inp_faddr, faddr) &&
    737 		    inp->inp_fport == fport &&
    738 		    inp->inp_lport == lport &&
    739 		    in_hosteq(inp->inp_laddr, laddr))
    740 			goto out;
    741 	}
    742 #ifdef DIAGNOSTIC
    743 	if (in_pcbnotifymiss) {
    744 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
    745 		    ntohl(faddr.s_addr), ntohs(fport),
    746 		    ntohl(laddr.s_addr), ntohs(lport));
    747 	}
    748 #endif
    749 	return (0);
    750 
    751 out:
    752 	/* Move this PCB to the head of hash chain. */
    753 	if (inp != head->lh_first) {
    754 		LIST_REMOVE(inp, inp_hash);
    755 		LIST_INSERT_HEAD(head, inp, inp_hash);
    756 	}
    757 	return (inp);
    758 }
    759 
    760 struct inpcb *
    761 in_pcblookup_bind(table, laddr, lport_arg)
    762 	struct inpcbtable *table;
    763 	struct in_addr laddr;
    764 	u_int lport_arg;
    765 {
    766 	struct inpcbhead *head;
    767 	register struct inpcb *inp;
    768 	u_int16_t lport = lport_arg;
    769 
    770 	head = INPCBHASH_BIND(table, laddr, lport);
    771 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    772 		if (inp->inp_lport == lport &&
    773 		    in_hosteq(inp->inp_laddr, laddr))
    774 			goto out;
    775 	}
    776 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
    777 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    778 		if (inp->inp_lport == lport &&
    779 		    in_hosteq(inp->inp_laddr, zeroin_addr))
    780 			goto out;
    781 	}
    782 #ifdef DIAGNOSTIC
    783 	if (in_pcbnotifymiss) {
    784 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
    785 		    ntohl(laddr.s_addr), ntohs(lport));
    786 	}
    787 #endif
    788 	return (0);
    789 
    790 out:
    791 	/* Move this PCB to the head of hash chain. */
    792 	if (inp != head->lh_first) {
    793 		LIST_REMOVE(inp, inp_hash);
    794 		LIST_INSERT_HEAD(head, inp, inp_hash);
    795 	}
    796 	return (inp);
    797 }
    798 
    799 void
    800 in_pcbstate(inp, state)
    801 	struct inpcb *inp;
    802 	int state;
    803 {
    804 
    805 	if (inp->inp_state > INP_ATTACHED)
    806 		LIST_REMOVE(inp, inp_hash);
    807 
    808 	switch (state) {
    809 	case INP_BOUND:
    810 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
    811 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
    812 		break;
    813 	case INP_CONNECTED:
    814 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
    815 		    inp->inp_faddr, inp->inp_fport,
    816 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
    817 		break;
    818 	}
    819 
    820 	inp->inp_state = state;
    821 }
    822 
    823 struct rtentry *
    824 in_pcbrtentry(inp)
    825 	struct inpcb *inp;
    826 {
    827 	struct route *ro;
    828 
    829 	ro = &inp->inp_route;
    830 
    831 	if (ro->ro_rt == NULL) {
    832 		/*
    833 		 * No route yet, so try to acquire one.
    834 		 */
    835 		if (!in_nullhost(inp->inp_faddr)) {
    836 			ro->ro_dst.sa_family = AF_INET;
    837 			ro->ro_dst.sa_len = sizeof(ro->ro_dst);
    838 			satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
    839 			rtalloc(ro);
    840 		}
    841 	}
    842 	return (ro->ro_rt);
    843 }
    844 
    845 struct sockaddr_in *
    846 in_selectsrc(sin, ro, soopts, mopts, errorp)
    847 	struct sockaddr_in *sin;
    848 	struct route *ro;
    849 	int soopts;
    850 	struct ip_moptions *mopts;
    851 	int *errorp;
    852 {
    853 	struct in_ifaddr *ia;
    854 
    855 	ia = (struct in_ifaddr *)0;
    856 	/*
    857 	 * If route is known or can be allocated now,
    858 	 * our src addr is taken from the i/f, else punt.
    859 	 */
    860 	if (ro->ro_rt &&
    861 	    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
    862 	    soopts & SO_DONTROUTE)) {
    863 		RTFREE(ro->ro_rt);
    864 		ro->ro_rt = (struct rtentry *)0;
    865 	}
    866 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
    867 	    (ro->ro_rt == (struct rtentry *)0 ||
    868 	    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    869 		/* No route yet, so try to acquire one */
    870 		ro->ro_dst.sa_family = AF_INET;
    871 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    872 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    873 		rtalloc(ro);
    874 	}
    875 	/*
    876 	 * If we found a route, use the address
    877 	 * corresponding to the outgoing interface
    878 	 * unless it is the loopback (in case a route
    879 	 * to our address on another net goes to loopback).
    880 	 *
    881 	 * XXX Is this still true?  Do we care?
    882 	 */
    883 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    884 		ia = ifatoia(ro->ro_rt->rt_ifa);
    885 	if (ia == 0) {
    886 		u_int16_t fport = sin->sin_port;
    887 
    888 		sin->sin_port = 0;
    889 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    890 		sin->sin_port = fport;
    891 		if (ia == 0) {
    892 			/* Find 1st non-loopback AF_INET address */
    893 			for (ia = in_ifaddr.tqh_first;
    894 			     ia != NULL;
    895 			     ia = ia->ia_list.tqe_next)
    896 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
    897 					break;
    898 		}
    899 		if (ia == 0) {
    900 			*errorp = EADDRNOTAVAIL;
    901 			return NULL;
    902 		}
    903 	}
    904 	/*
    905 	 * If the destination address is multicast and an outgoing
    906 	 * interface has been set as a multicast option, use the
    907 	 * address of that interface as our source address.
    908 	 */
    909 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
    910 		struct ip_moptions *imo;
    911 		struct ifnet *ifp;
    912 
    913 		imo = mopts;
    914 		if (imo->imo_multicast_ifp != NULL) {
    915 			ifp = imo->imo_multicast_ifp;
    916 			IFP_TO_IA(ifp, ia);		/* XXX */
    917 			if (ia == 0) {
    918 				*errorp = EADDRNOTAVAIL;
    919 				return NULL;
    920 			}
    921 		}
    922 	}
    923 	return satosin(&ia->ia_addr);
    924 }
    925