Home | History | Annotate | Line # | Download | only in netinet
in_pcb.c revision 1.129.10.1.2.1
      1 /*	$NetBSD: in_pcb.c,v 1.129.10.1.2.1 2010/04/21 00:28:22 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  *
     49  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     50  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     51  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     52  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     53  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     54  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     55  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     56  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     57  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     58  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     59  * POSSIBILITY OF SUCH DAMAGE.
     60  */
     61 
     62 /*
     63  * Copyright (c) 1982, 1986, 1991, 1993, 1995
     64  *	The Regents of the University of California.  All rights reserved.
     65  *
     66  * Redistribution and use in source and binary forms, with or without
     67  * modification, are permitted provided that the following conditions
     68  * are met:
     69  * 1. Redistributions of source code must retain the above copyright
     70  *    notice, this list of conditions and the following disclaimer.
     71  * 2. Redistributions in binary form must reproduce the above copyright
     72  *    notice, this list of conditions and the following disclaimer in the
     73  *    documentation and/or other materials provided with the distribution.
     74  * 3. Neither the name of the University nor the names of its contributors
     75  *    may be used to endorse or promote products derived from this software
     76  *    without specific prior written permission.
     77  *
     78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     88  * SUCH DAMAGE.
     89  *
     90  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
     91  */
     92 
     93 #include <sys/cdefs.h>
     94 __KERNEL_RCSID(0, "$NetBSD: in_pcb.c,v 1.129.10.1.2.1 2010/04/21 00:28:22 matt Exp $");
     95 
     96 #include "opt_inet.h"
     97 #include "opt_ipsec.h"
     98 
     99 #include <sys/param.h>
    100 #include <sys/systm.h>
    101 #include <sys/malloc.h>
    102 #include <sys/mbuf.h>
    103 #include <sys/protosw.h>
    104 #include <sys/socket.h>
    105 #include <sys/socketvar.h>
    106 #include <sys/ioctl.h>
    107 #include <sys/errno.h>
    108 #include <sys/time.h>
    109 #include <sys/once.h>
    110 #include <sys/pool.h>
    111 #include <sys/proc.h>
    112 #include <sys/kauth.h>
    113 #include <sys/uidinfo.h>
    114 
    115 #include <net/if.h>
    116 #include <net/route.h>
    117 
    118 #include <netinet/in.h>
    119 #include <netinet/in_systm.h>
    120 #include <netinet/ip.h>
    121 #include <netinet/in_pcb.h>
    122 #include <netinet/in_var.h>
    123 #include <netinet/ip_var.h>
    124 
    125 #ifdef INET6
    126 #include <netinet/ip6.h>
    127 #include <netinet6/ip6_var.h>
    128 #include <netinet6/in6_pcb.h>
    129 #endif
    130 
    131 #ifdef IPSEC
    132 #include <netinet6/ipsec.h>
    133 #include <netkey/key.h>
    134 #elif FAST_IPSEC
    135 #include <netipsec/ipsec.h>
    136 #include <netipsec/key.h>
    137 #endif /* IPSEC */
    138 
    139 struct	in_addr zeroin_addr;
    140 
    141 #define	INPCBHASH_PORT(table, lport) \
    142 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
    143 #define	INPCBHASH_BIND(table, laddr, lport) \
    144 	&(table)->inpt_bindhashtbl[ \
    145 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
    146 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
    147 	&(table)->inpt_connecthashtbl[ \
    148 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
    149 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
    150 
    151 int	anonportmin = IPPORT_ANONMIN;
    152 int	anonportmax = IPPORT_ANONMAX;
    153 int	lowportmin  = IPPORT_RESERVEDMIN;
    154 int	lowportmax  = IPPORT_RESERVEDMAX;
    155 
    156 static struct pool inpcb_pool;
    157 
    158 static int
    159 inpcb_poolinit(void)
    160 {
    161 
    162 	pool_init(&inpcb_pool, sizeof(struct inpcb), 0, 0, 0, "inpcbpl", NULL,
    163 	    IPL_NET);
    164 	return 0;
    165 }
    166 
    167 void
    168 in_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
    169 {
    170 	static ONCE_DECL(control);
    171 
    172 	CIRCLEQ_INIT(&table->inpt_queue);
    173 	table->inpt_porthashtbl = hashinit(bindhashsize, HASH_LIST, true,
    174 	    &table->inpt_porthash);
    175 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, true,
    176 	    &table->inpt_bindhash);
    177 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST, true,
    178 	    &table->inpt_connecthash);
    179 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
    180 	table->inpt_lastport = (u_int16_t)anonportmax;
    181 
    182 	RUN_ONCE(&control, inpcb_poolinit);
    183 }
    184 
    185 int
    186 in_pcballoc(struct socket *so, void *v)
    187 {
    188 	struct inpcbtable *table = v;
    189 	struct inpcb *inp;
    190 	int s;
    191 #if defined(IPSEC) || defined(FAST_IPSEC)
    192 	int error;
    193 #endif
    194 
    195 	s = splnet();
    196 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
    197 	splx(s);
    198 	if (inp == NULL)
    199 		return (ENOBUFS);
    200 	bzero((void *)inp, sizeof(*inp));
    201 	inp->inp_af = AF_INET;
    202 	inp->inp_table = table;
    203 	inp->inp_socket = so;
    204 	inp->inp_errormtu = -1;
    205 #if defined(IPSEC) || defined(FAST_IPSEC)
    206 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
    207 	if (error != 0) {
    208 		s = splnet();
    209 		pool_put(&inpcb_pool, inp);
    210 		splx(s);
    211 		return error;
    212 	}
    213 #endif
    214 	so->so_pcb = inp;
    215 	s = splnet();
    216 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, &inp->inp_head,
    217 	    inph_queue);
    218 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
    219 	    inph_lhash);
    220 	in_pcbstate(inp, INP_ATTACHED);
    221 	splx(s);
    222 	return (0);
    223 }
    224 
    225 int
    226 in_pcbbind(void *v, struct mbuf *nam, struct lwp *l)
    227 {
    228 	struct in_ifaddr *ia = NULL;
    229 	struct inpcb *inp = v;
    230 	struct socket *so = inp->inp_socket;
    231 	struct inpcbtable *table = inp->inp_table;
    232 	struct sockaddr_in *sin = NULL; /* XXXGCC */
    233 	u_int16_t lport = 0;
    234 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
    235 #ifndef IPNOPRIVPORTS
    236 	kauth_cred_t cred = l->l_cred;
    237 #endif
    238 
    239 	if (inp->inp_af != AF_INET)
    240 		return (EINVAL);
    241 
    242 	if (TAILQ_FIRST(&in_ifaddrhead) == 0)
    243 		return (EADDRNOTAVAIL);
    244 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
    245 		return (EINVAL);
    246 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
    247 		wild = 1;
    248 	if (nam == 0)
    249 		goto noname;
    250 	sin = mtod(nam, struct sockaddr_in *);
    251 	if (nam->m_len != sizeof (*sin))
    252 		return (EINVAL);
    253 	if (sin->sin_family != AF_INET)
    254 		return (EAFNOSUPPORT);
    255 	lport = sin->sin_port;
    256 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
    257 		/*
    258 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
    259 		 * allow complete duplication of binding if
    260 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
    261 		 * and a multicast address is bound on both
    262 		 * new and duplicated sockets.
    263 		 */
    264 		if (so->so_options & SO_REUSEADDR)
    265 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
    266 	} else if (!in_nullhost(sin->sin_addr)) {
    267 		sin->sin_port = 0;		/* yech... */
    268 		INADDR_TO_IA(sin->sin_addr, ia);
    269 		/* check for broadcast addresses */
    270 		if (ia == NULL)
    271 			ia = ifatoia(ifa_ifwithaddr(sintosa(sin)));
    272 		if (ia == NULL)
    273 			return (EADDRNOTAVAIL);
    274 	}
    275 	if (lport) {
    276 		struct inpcb *t;
    277 #ifdef INET6
    278 		struct in6pcb *t6;
    279 		struct in6_addr mapped;
    280 #endif
    281 #ifndef IPNOPRIVPORTS
    282 		/* GROSS */
    283 		if (ntohs(lport) < IPPORT_RESERVED &&
    284 		    kauth_authorize_network(cred,
    285 		    KAUTH_NETWORK_BIND,
    286 		    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so, sin,
    287 		    NULL))
    288 			return (EACCES);
    289 #endif
    290 #ifdef INET6
    291 		memset(&mapped, 0, sizeof(mapped));
    292 		mapped.s6_addr16[5] = 0xffff;
    293 		memcpy(&mapped.s6_addr32[3], &sin->sin_addr,
    294 		    sizeof(mapped.s6_addr32[3]));
    295 		t6 = in6_pcblookup_port(table, &mapped, lport, wild);
    296 		if (t6 && (reuseport & t6->in6p_socket->so_options) == 0)
    297 			return (EADDRINUSE);
    298 #endif
    299 		if (so->so_uidinfo->ui_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
    300 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
    301 		/*
    302 		 * XXX:	investigate ramifications of loosening this
    303 		 *	restriction so that as long as both ports have
    304 		 *	SO_REUSEPORT allow the bind
    305 		 */
    306 			if (t &&
    307 			    (!in_nullhost(sin->sin_addr) ||
    308 			     !in_nullhost(t->inp_laddr) ||
    309 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
    310 			    && (so->so_uidinfo->ui_uid != t->inp_socket->so_uidinfo->ui_uid)) {
    311 				return (EADDRINUSE);
    312 			}
    313 		}
    314 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
    315 		if (t && (reuseport & t->inp_socket->so_options) == 0)
    316 			return (EADDRINUSE);
    317 	}
    318 	inp->inp_laddr = sin->sin_addr;
    319 
    320 noname:
    321 	if (lport == 0) {
    322 		int	   cnt;
    323 		u_int16_t  mymin, mymax;
    324 		u_int16_t *lastport;
    325 
    326 		if (inp->inp_flags & INP_LOWPORT) {
    327 #ifndef IPNOPRIVPORTS
    328 			if (kauth_authorize_network(cred,
    329 			    KAUTH_NETWORK_BIND,
    330 			    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so,
    331 			    sin, NULL))
    332 				return (EACCES);
    333 #endif
    334 			mymin = lowportmin;
    335 			mymax = lowportmax;
    336 			lastport = &table->inpt_lastlow;
    337 		} else {
    338 			mymin = anonportmin;
    339 			mymax = anonportmax;
    340 			lastport = &table->inpt_lastport;
    341 		}
    342 		if (mymin > mymax) {	/* sanity check */
    343 			u_int16_t swp;
    344 
    345 			swp = mymin;
    346 			mymin = mymax;
    347 			mymax = swp;
    348 		}
    349 
    350 		lport = *lastport - 1;
    351 		for (cnt = mymax - mymin + 1; cnt; cnt--, lport--) {
    352 			if (lport < mymin || lport > mymax)
    353 				lport = mymax;
    354 			if (!in_pcblookup_port(table, inp->inp_laddr,
    355 			    htons(lport), 1))
    356 				goto found;
    357 		}
    358 		if (!in_nullhost(inp->inp_laddr))
    359 			inp->inp_laddr.s_addr = INADDR_ANY;
    360 		return (EAGAIN);
    361 	found:
    362 		inp->inp_flags |= INP_ANONPORT;
    363 		*lastport = lport;
    364 		lport = htons(lport);
    365 	}
    366 	inp->inp_lport = lport;
    367 	LIST_REMOVE(&inp->inp_head, inph_lhash);
    368 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
    369 	    inph_lhash);
    370 	in_pcbstate(inp, INP_BOUND);
    371 	return (0);
    372 }
    373 
    374 /*
    375  * Connect from a socket to a specified address.
    376  * Both address and port must be specified in argument sin.
    377  * If don't have a local address for this socket yet,
    378  * then pick one.
    379  */
    380 int
    381 in_pcbconnect(void *v, struct mbuf *nam, struct lwp *l)
    382 {
    383 	struct inpcb *inp = v;
    384 	struct in_ifaddr *ia = NULL;
    385 	struct sockaddr_in *ifaddr = NULL;
    386 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
    387 	int error;
    388 
    389 	if (inp->inp_af != AF_INET)
    390 		return (EINVAL);
    391 
    392 	if (nam->m_len != sizeof (*sin))
    393 		return (EINVAL);
    394 	if (sin->sin_family != AF_INET)
    395 		return (EAFNOSUPPORT);
    396 	if (sin->sin_port == 0)
    397 		return (EADDRNOTAVAIL);
    398 	if (TAILQ_FIRST(&in_ifaddrhead) != 0) {
    399 		/*
    400 		 * If the destination address is INADDR_ANY,
    401 		 * use any local address (likely loopback).
    402 		 * If the supplied address is INADDR_BROADCAST,
    403 		 * use the broadcast address of an interface
    404 		 * which supports broadcast. (loopback does not)
    405 		 */
    406 
    407 		if (in_nullhost(sin->sin_addr)) {
    408 			sin->sin_addr =
    409 			    TAILQ_FIRST(&in_ifaddrhead)->ia_addr.sin_addr;
    410 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
    411 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
    412 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
    413 					sin->sin_addr =
    414 					    ia->ia_broadaddr.sin_addr;
    415 					break;
    416 				}
    417 			}
    418 		}
    419 	}
    420 	/*
    421 	 * If we haven't bound which network number to use as ours,
    422 	 * we will use the number of the outgoing interface.
    423 	 * This depends on having done a routing lookup, which
    424 	 * we will probably have to do anyway, so we might
    425 	 * as well do it now.  On the other hand if we are
    426 	 * sending to multiple destinations we may have already
    427 	 * done the lookup, so see if we can use the route
    428 	 * from before.  In any case, we only
    429 	 * chose a port number once, even if sending to multiple
    430 	 * destinations.
    431 	 */
    432 	if (in_nullhost(inp->inp_laddr)) {
    433 		int xerror;
    434 		ifaddr = in_selectsrc(sin, &inp->inp_route,
    435 		    inp->inp_socket->so_options, inp->inp_moptions, &xerror);
    436 		if (ifaddr == NULL) {
    437 			if (xerror == 0)
    438 				xerror = EADDRNOTAVAIL;
    439 			return xerror;
    440 		}
    441 		INADDR_TO_IA(ifaddr->sin_addr, ia);
    442 		if (ia == NULL)
    443 			return (EADDRNOTAVAIL);
    444 	}
    445 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
    446 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
    447 	    inp->inp_lport) != 0)
    448 		return (EADDRINUSE);
    449 	if (in_nullhost(inp->inp_laddr)) {
    450 		if (inp->inp_lport == 0) {
    451 			error = in_pcbbind(inp, NULL, l);
    452 			/*
    453 			 * This used to ignore the return value
    454 			 * completely, but we need to check for
    455 			 * ephemeral port shortage.
    456 			 * And attempts to request low ports if not root.
    457 			 */
    458 			if (error != 0)
    459 				return (error);
    460 		}
    461 		inp->inp_laddr = ifaddr->sin_addr;
    462 	}
    463 	inp->inp_faddr = sin->sin_addr;
    464 	inp->inp_fport = sin->sin_port;
    465 	in_pcbstate(inp, INP_CONNECTED);
    466 #if defined(IPSEC) || defined(FAST_IPSEC)
    467 	if (inp->inp_socket->so_type == SOCK_STREAM)
    468 		ipsec_pcbconn(inp->inp_sp);
    469 #endif
    470 	return (0);
    471 }
    472 
    473 void
    474 in_pcbdisconnect(void *v)
    475 {
    476 	struct inpcb *inp = v;
    477 
    478 	if (inp->inp_af != AF_INET)
    479 		return;
    480 
    481 	inp->inp_faddr = zeroin_addr;
    482 	inp->inp_fport = 0;
    483 	in_pcbstate(inp, INP_BOUND);
    484 #if defined(IPSEC) || defined(FAST_IPSEC)
    485 	ipsec_pcbdisconn(inp->inp_sp);
    486 #endif
    487 	if (inp->inp_socket->so_state & SS_NOFDREF)
    488 		in_pcbdetach(inp);
    489 }
    490 
    491 void
    492 in_pcbdetach(void *v)
    493 {
    494 	struct inpcb *inp = v;
    495 	struct socket *so = inp->inp_socket;
    496 	int s;
    497 
    498 	if (inp->inp_af != AF_INET)
    499 		return;
    500 
    501 #if defined(IPSEC) || defined(FAST_IPSEC)
    502 	ipsec4_delete_pcbpolicy(inp);
    503 #endif /*IPSEC*/
    504 	so->so_pcb = 0;
    505 	if (inp->inp_options)
    506 		(void)m_free(inp->inp_options);
    507 	rtcache_free(&inp->inp_route);
    508 	ip_freemoptions(inp->inp_moptions);
    509 	s = splnet();
    510 	in_pcbstate(inp, INP_ATTACHED);
    511 	LIST_REMOVE(&inp->inp_head, inph_lhash);
    512 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, &inp->inp_head,
    513 	    inph_queue);
    514 	pool_put(&inpcb_pool, inp);
    515 	splx(s);
    516 	sofree(so);			/* drops the socket's lock */
    517 	mutex_enter(softnet_lock);	/* reacquire the softnet_lock */
    518 }
    519 
    520 void
    521 in_setsockaddr(struct inpcb *inp, struct mbuf *nam)
    522 {
    523 	struct sockaddr_in *sin;
    524 
    525 	if (inp->inp_af != AF_INET)
    526 		return;
    527 
    528 	sin = mtod(nam, struct sockaddr_in *);
    529 	sockaddr_in_init(sin, &inp->inp_laddr, inp->inp_lport);
    530 	nam->m_len = sin->sin_len;
    531 }
    532 
    533 void
    534 in_setpeeraddr(struct inpcb *inp, struct mbuf *nam)
    535 {
    536 	struct sockaddr_in *sin;
    537 
    538 	if (inp->inp_af != AF_INET)
    539 		return;
    540 
    541 	sin = mtod(nam, struct sockaddr_in *);
    542 	sockaddr_in_init(sin, &inp->inp_faddr, inp->inp_fport);
    543 	nam->m_len = sin->sin_len;
    544 }
    545 
    546 /*
    547  * Pass some notification to all connections of a protocol
    548  * associated with address dst.  The local address and/or port numbers
    549  * may be specified to limit the search.  The "usual action" will be
    550  * taken, depending on the ctlinput cmd.  The caller must filter any
    551  * cmds that are uninteresting (e.g., no error in the map).
    552  * Call the protocol specific routine (if any) to report
    553  * any errors for each matching socket.
    554  *
    555  * Must be called at splsoftnet.
    556  */
    557 int
    558 in_pcbnotify(struct inpcbtable *table, struct in_addr faddr, u_int fport_arg,
    559     struct in_addr laddr, u_int lport_arg, int errno,
    560     void (*notify)(struct inpcb *, int))
    561 {
    562 	struct inpcbhead *head;
    563 	struct inpcb *inp, *ninp;
    564 	u_int16_t fport = fport_arg, lport = lport_arg;
    565 	int nmatch;
    566 
    567 	if (in_nullhost(faddr) || notify == 0)
    568 		return (0);
    569 
    570 	nmatch = 0;
    571 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    572 	for (inp = (struct inpcb *)LIST_FIRST(head); inp != NULL; inp = ninp) {
    573 		ninp = (struct inpcb *)LIST_NEXT(inp, inp_hash);
    574 		if (inp->inp_af != AF_INET)
    575 			continue;
    576 		if (in_hosteq(inp->inp_faddr, faddr) &&
    577 		    inp->inp_fport == fport &&
    578 		    inp->inp_lport == lport &&
    579 		    in_hosteq(inp->inp_laddr, laddr)) {
    580 			(*notify)(inp, errno);
    581 			nmatch++;
    582 		}
    583 	}
    584 	return (nmatch);
    585 }
    586 
    587 void
    588 in_pcbnotifyall(struct inpcbtable *table, struct in_addr faddr, int errno,
    589     void (*notify)(struct inpcb *, int))
    590 {
    591 	struct inpcb *inp, *ninp;
    592 
    593 	if (in_nullhost(faddr) || notify == 0)
    594 		return;
    595 
    596 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
    597 	    inp != (void *)&table->inpt_queue;
    598 	    inp = ninp) {
    599 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
    600 		if (inp->inp_af != AF_INET)
    601 			continue;
    602 		if (in_hosteq(inp->inp_faddr, faddr))
    603 			(*notify)(inp, errno);
    604 	}
    605 }
    606 
    607 void
    608 in_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
    609 {
    610 	struct inpcb *inp, *ninp;
    611 	struct ip_moptions *imo;
    612 	int i, gap;
    613 
    614 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
    615 	    inp != (void *)&table->inpt_queue;
    616 	    inp = ninp) {
    617 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
    618 		if (inp->inp_af != AF_INET)
    619 			continue;
    620 		imo = inp->inp_moptions;
    621 		if (imo != NULL) {
    622 			/*
    623 			 * Unselect the outgoing interface if it is being
    624 			 * detached.
    625 			 */
    626 			if (imo->imo_multicast_ifp == ifp)
    627 				imo->imo_multicast_ifp = NULL;
    628 
    629 			/*
    630 			 * Drop multicast group membership if we joined
    631 			 * through the interface being detached.
    632 			 */
    633 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
    634 			    i++) {
    635 				if (imo->imo_membership[i]->inm_ifp == ifp) {
    636 					in_delmulti(imo->imo_membership[i]);
    637 					gap++;
    638 				} else if (gap != 0)
    639 					imo->imo_membership[i - gap] =
    640 					    imo->imo_membership[i];
    641 			}
    642 			imo->imo_num_memberships -= gap;
    643 		}
    644 	}
    645 }
    646 
    647 void
    648 in_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
    649 {
    650 	struct rtentry *rt;
    651 	struct inpcb *inp, *ninp;
    652 
    653 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
    654 	    inp != (void *)&table->inpt_queue;
    655 	    inp = ninp) {
    656 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
    657 		if (inp->inp_af != AF_INET)
    658 			continue;
    659 		if ((rt = rtcache_validate(&inp->inp_route)) != NULL &&
    660 		    rt->rt_ifp == ifp)
    661 			in_rtchange(inp, 0);
    662 	}
    663 }
    664 
    665 /*
    666  * Check for alternatives when higher level complains
    667  * about service problems.  For now, invalidate cached
    668  * routing information.  If the route was created dynamically
    669  * (by a redirect), time to try a default gateway again.
    670  */
    671 void
    672 in_losing(struct inpcb *inp)
    673 {
    674 	struct rtentry *rt;
    675 	struct rt_addrinfo info;
    676 
    677 	if (inp->inp_af != AF_INET)
    678 		return;
    679 
    680 	if ((rt = rtcache_validate(&inp->inp_route)) == NULL)
    681 		return;
    682 
    683 	memset(&info, 0, sizeof(info));
    684 	info.rti_info[RTAX_DST] = rtcache_getdst(&inp->inp_route);
    685 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    686 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    687 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
    688 	if (rt->rt_flags & RTF_DYNAMIC)
    689 		(void) rtrequest(RTM_DELETE, rt_getkey(rt),
    690 			rt->rt_gateway, rt_mask(rt), rt->rt_flags,
    691 			NULL);
    692 	/*
    693 	 * A new route can be allocated
    694 	 * the next time output is attempted.
    695 	 */
    696 	rtcache_free(&inp->inp_route);
    697 }
    698 
    699 /*
    700  * After a routing change, flush old routing.  A new route can be
    701  * allocated the next time output is attempted.
    702  */
    703 void
    704 in_rtchange(struct inpcb *inp, int errno)
    705 {
    706 
    707 	if (inp->inp_af != AF_INET)
    708 		return;
    709 
    710 	rtcache_free(&inp->inp_route);
    711 
    712 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
    713 }
    714 
    715 struct inpcb *
    716 in_pcblookup_port(struct inpcbtable *table, struct in_addr laddr,
    717     u_int lport_arg, int lookup_wildcard)
    718 {
    719 	struct inpcbhead *head;
    720 	struct inpcb_hdr *inph;
    721 	struct inpcb *inp, *match = 0;
    722 	int matchwild = 3, wildcard;
    723 	u_int16_t lport = lport_arg;
    724 
    725 	head = INPCBHASH_PORT(table, lport);
    726 	LIST_FOREACH(inph, head, inph_lhash) {
    727 		inp = (struct inpcb *)inph;
    728 		if (inp->inp_af != AF_INET)
    729 			continue;
    730 
    731 		if (inp->inp_lport != lport)
    732 			continue;
    733 		wildcard = 0;
    734 		if (!in_nullhost(inp->inp_faddr))
    735 			wildcard++;
    736 		if (in_nullhost(inp->inp_laddr)) {
    737 			if (!in_nullhost(laddr))
    738 				wildcard++;
    739 		} else {
    740 			if (in_nullhost(laddr))
    741 				wildcard++;
    742 			else {
    743 				if (!in_hosteq(inp->inp_laddr, laddr))
    744 					continue;
    745 			}
    746 		}
    747 		if (wildcard && !lookup_wildcard)
    748 			continue;
    749 		if (wildcard < matchwild) {
    750 			match = inp;
    751 			matchwild = wildcard;
    752 			if (matchwild == 0)
    753 				break;
    754 		}
    755 	}
    756 	return (match);
    757 }
    758 
    759 #ifdef DIAGNOSTIC
    760 int	in_pcbnotifymiss = 0;
    761 #endif
    762 
    763 struct inpcb *
    764 in_pcblookup_connect(struct inpcbtable *table,
    765     struct in_addr faddr, u_int fport_arg,
    766     struct in_addr laddr, u_int lport_arg)
    767 {
    768 	struct inpcbhead *head;
    769 	struct inpcb_hdr *inph;
    770 	struct inpcb *inp;
    771 	u_int16_t fport = fport_arg, lport = lport_arg;
    772 
    773 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    774 	LIST_FOREACH(inph, head, inph_hash) {
    775 		inp = (struct inpcb *)inph;
    776 		if (inp->inp_af != AF_INET)
    777 			continue;
    778 
    779 		if (in_hosteq(inp->inp_faddr, faddr) &&
    780 		    inp->inp_fport == fport &&
    781 		    inp->inp_lport == lport &&
    782 		    in_hosteq(inp->inp_laddr, laddr))
    783 			goto out;
    784 	}
    785 #ifdef DIAGNOSTIC
    786 	if (in_pcbnotifymiss) {
    787 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
    788 		    ntohl(faddr.s_addr), ntohs(fport),
    789 		    ntohl(laddr.s_addr), ntohs(lport));
    790 	}
    791 #endif
    792 	return (0);
    793 
    794 out:
    795 	/* Move this PCB to the head of hash chain. */
    796 	inph = &inp->inp_head;
    797 	if (inph != LIST_FIRST(head)) {
    798 		LIST_REMOVE(inph, inph_hash);
    799 		LIST_INSERT_HEAD(head, inph, inph_hash);
    800 	}
    801 	return (inp);
    802 }
    803 
    804 struct inpcb *
    805 in_pcblookup_bind(struct inpcbtable *table,
    806     struct in_addr laddr, u_int lport_arg)
    807 {
    808 	struct inpcbhead *head;
    809 	struct inpcb_hdr *inph;
    810 	struct inpcb *inp;
    811 	u_int16_t lport = lport_arg;
    812 
    813 	head = INPCBHASH_BIND(table, laddr, lport);
    814 	LIST_FOREACH(inph, head, inph_hash) {
    815 		inp = (struct inpcb *)inph;
    816 		if (inp->inp_af != AF_INET)
    817 			continue;
    818 
    819 		if (inp->inp_lport == lport &&
    820 		    in_hosteq(inp->inp_laddr, laddr))
    821 			goto out;
    822 	}
    823 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
    824 	LIST_FOREACH(inph, head, inph_hash) {
    825 		inp = (struct inpcb *)inph;
    826 		if (inp->inp_af != AF_INET)
    827 			continue;
    828 
    829 		if (inp->inp_lport == lport &&
    830 		    in_hosteq(inp->inp_laddr, zeroin_addr))
    831 			goto out;
    832 	}
    833 #ifdef DIAGNOSTIC
    834 	if (in_pcbnotifymiss) {
    835 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
    836 		    ntohl(laddr.s_addr), ntohs(lport));
    837 	}
    838 #endif
    839 	return (0);
    840 
    841 out:
    842 	/* Move this PCB to the head of hash chain. */
    843 	inph = &inp->inp_head;
    844 	if (inph != LIST_FIRST(head)) {
    845 		LIST_REMOVE(inph, inph_hash);
    846 		LIST_INSERT_HEAD(head, inph, inph_hash);
    847 	}
    848 	return (inp);
    849 }
    850 
    851 void
    852 in_pcbstate(struct inpcb *inp, int state)
    853 {
    854 
    855 	if (inp->inp_af != AF_INET)
    856 		return;
    857 
    858 	if (inp->inp_state > INP_ATTACHED)
    859 		LIST_REMOVE(&inp->inp_head, inph_hash);
    860 
    861 	switch (state) {
    862 	case INP_BOUND:
    863 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
    864 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
    865 		    inph_hash);
    866 		break;
    867 	case INP_CONNECTED:
    868 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
    869 		    inp->inp_faddr, inp->inp_fport,
    870 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
    871 		    inph_hash);
    872 		break;
    873 	}
    874 
    875 	inp->inp_state = state;
    876 }
    877 
    878 struct rtentry *
    879 in_pcbrtentry(struct inpcb *inp)
    880 {
    881 	struct route *ro;
    882 	union {
    883 		struct sockaddr		dst;
    884 		struct sockaddr_in	dst4;
    885 	} u;
    886 
    887 	if (inp->inp_af != AF_INET)
    888 		return (NULL);
    889 
    890 	ro = &inp->inp_route;
    891 
    892 	sockaddr_in_init(&u.dst4, &inp->inp_faddr, 0);
    893 	return rtcache_lookup(ro, &u.dst);
    894 }
    895 
    896 struct sockaddr_in *
    897 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
    898     int soopts, struct ip_moptions *mopts, int *errorp)
    899 {
    900 	struct rtentry *rt = NULL;
    901 	struct in_ifaddr *ia = NULL;
    902 
    903 	/*
    904          * If route is known or can be allocated now, take the
    905          * source address from the interface.  Otherwise, punt.
    906 	 */
    907 	if ((soopts & SO_DONTROUTE) != 0)
    908 		rtcache_free(ro);
    909 	else {
    910 		union {
    911 			struct sockaddr		dst;
    912 			struct sockaddr_in	dst4;
    913 		} u;
    914 
    915 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
    916 		rt = rtcache_lookup(ro, &u.dst);
    917 	}
    918 	/*
    919 	 * If we found a route, use the address
    920 	 * corresponding to the outgoing interface
    921 	 * unless it is the loopback (in case a route
    922 	 * to our address on another net goes to loopback).
    923 	 *
    924 	 * XXX Is this still true?  Do we care?
    925 	 */
    926 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
    927 		ia = ifatoia(rt->rt_ifa);
    928 	if (ia == NULL) {
    929 		u_int16_t fport = sin->sin_port;
    930 
    931 		sin->sin_port = 0;
    932 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    933 		sin->sin_port = fport;
    934 		if (ia == NULL) {
    935 			/* Find 1st non-loopback AF_INET address */
    936 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
    937 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
    938 					break;
    939 			}
    940 		}
    941 		if (ia == NULL) {
    942 			*errorp = EADDRNOTAVAIL;
    943 			return NULL;
    944 		}
    945 	}
    946 	/*
    947 	 * If the destination address is multicast and an outgoing
    948 	 * interface has been set as a multicast option, use the
    949 	 * address of that interface as our source address.
    950 	 */
    951 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
    952 		struct ip_moptions *imo;
    953 		struct ifnet *ifp;
    954 
    955 		imo = mopts;
    956 		if (imo->imo_multicast_ifp != NULL) {
    957 			ifp = imo->imo_multicast_ifp;
    958 			IFP_TO_IA(ifp, ia);		/* XXX */
    959 			if (ia == 0) {
    960 				*errorp = EADDRNOTAVAIL;
    961 				return NULL;
    962 			}
    963 		}
    964 	}
    965 	if (ia->ia_ifa.ifa_getifa != NULL) {
    966 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
    967 		                                      sintosa(sin)));
    968 	}
    969 #ifdef GETIFA_DEBUG
    970 	else
    971 		printf("%s: missing ifa_getifa\n", __func__);
    972 #endif
    973 	return satosin(&ia->ia_addr);
    974 }
    975