Home | History | Annotate | Line # | Download | only in netinet
in_pcb.c revision 1.83.2.2
      1 /*	$NetBSD: in_pcb.c,v 1.83.2.2 2004/09/18 14:54:53 skrll 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. Neither the name of the University nor the names of its contributors
     82  *    may be used to endorse or promote products derived from this software
     83  *    without specific prior written permission.
     84  *
     85  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     86  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     87  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     88  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     89  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     90  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     91  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     92  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     93  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     94  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     95  * SUCH DAMAGE.
     96  *
     97  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
     98  */
     99 
    100 #include <sys/cdefs.h>
    101 __KERNEL_RCSID(0, "$NetBSD: in_pcb.c,v 1.83.2.2 2004/09/18 14:54:53 skrll Exp $");
    102 
    103 #include "opt_inet.h"
    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 INET6
    130 #include <netinet/ip6.h>
    131 #include <netinet6/ip6_var.h>
    132 #include <netinet6/in6_pcb.h>
    133 #endif
    134 
    135 #ifdef IPSEC
    136 #include <netinet6/ipsec.h>
    137 #include <netkey/key.h>
    138 #elif FAST_IPSEC
    139 #include <netipsec/ipsec.h>
    140 #include <netipsec/key.h>
    141 #endif /* IPSEC */
    142 
    143 struct	in_addr zeroin_addr;
    144 
    145 #define	INPCBHASH_PORT(table, lport) \
    146 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
    147 #define	INPCBHASH_BIND(table, laddr, lport) \
    148 	&(table)->inpt_bindhashtbl[ \
    149 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
    150 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
    151 	&(table)->inpt_connecthashtbl[ \
    152 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
    153 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
    154 
    155 int	anonportmin = IPPORT_ANONMIN;
    156 int	anonportmax = IPPORT_ANONMAX;
    157 int	lowportmin  = IPPORT_RESERVEDMIN;
    158 int	lowportmax  = IPPORT_RESERVEDMAX;
    159 
    160 POOL_INIT(inpcb_pool, sizeof(struct inpcb), 0, 0, 0, "inpcbpl", NULL);
    161 
    162 void
    163 in_pcbinit(table, bindhashsize, connecthashsize)
    164 	struct inpcbtable *table;
    165 	int bindhashsize, connecthashsize;
    166 {
    167 
    168 	CIRCLEQ_INIT(&table->inpt_queue);
    169 	table->inpt_porthashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
    170 	    M_WAITOK, &table->inpt_porthash);
    171 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
    172 	    M_WAITOK, &table->inpt_bindhash);
    173 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST,
    174 	    M_PCB, M_WAITOK, &table->inpt_connecthash);
    175 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
    176 	table->inpt_lastport = (u_int16_t)anonportmax;
    177 }
    178 
    179 int
    180 in_pcballoc(so, v)
    181 	struct socket *so;
    182 	void *v;
    183 {
    184 	struct inpcbtable *table = v;
    185 	struct inpcb *inp;
    186 	int s;
    187 #if defined(IPSEC) || defined(FAST_IPSEC)
    188 	int error;
    189 #endif
    190 
    191 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
    192 	if (inp == NULL)
    193 		return (ENOBUFS);
    194 	bzero((caddr_t)inp, sizeof(*inp));
    195 	inp->inp_af = AF_INET;
    196 	inp->inp_table = table;
    197 	inp->inp_socket = so;
    198 	inp->inp_errormtu = -1;
    199 #if defined(IPSEC) || defined(FAST_IPSEC)
    200 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
    201 	if (error != 0) {
    202 		pool_put(&inpcb_pool, inp);
    203 		return error;
    204 	}
    205 #endif
    206 	so->so_pcb = inp;
    207 	s = splnet();
    208 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, &inp->inp_head,
    209 	    inph_queue);
    210 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
    211 	    inph_lhash);
    212 	in_pcbstate(inp, INP_ATTACHED);
    213 	splx(s);
    214 	return (0);
    215 }
    216 
    217 int
    218 in_pcbbind(v, nam, p)
    219 	void *v;
    220 	struct mbuf *nam;
    221 	struct proc *p;
    222 {
    223 	struct in_ifaddr *ia = NULL;
    224 	struct inpcb *inp = v;
    225 	struct socket *so = inp->inp_socket;
    226 	struct inpcbtable *table = inp->inp_table;
    227 	struct sockaddr_in *sin;
    228 	u_int16_t lport = 0;
    229 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
    230 
    231 	if (inp->inp_af != AF_INET)
    232 		return (EINVAL);
    233 
    234 	if (TAILQ_FIRST(&in_ifaddrhead) == 0)
    235 		return (EADDRNOTAVAIL);
    236 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
    237 		return (EINVAL);
    238 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
    239 		wild = 1;
    240 	if (nam == 0)
    241 		goto noname;
    242 	sin = mtod(nam, struct sockaddr_in *);
    243 	if (nam->m_len != sizeof (*sin))
    244 		return (EINVAL);
    245 	if (sin->sin_family != AF_INET)
    246 		return (EAFNOSUPPORT);
    247 	lport = sin->sin_port;
    248 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
    249 		/*
    250 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
    251 		 * allow complete duplication of binding if
    252 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
    253 		 * and a multicast address is bound on both
    254 		 * new and duplicated sockets.
    255 		 */
    256 		if (so->so_options & SO_REUSEADDR)
    257 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
    258 	} else if (!in_nullhost(sin->sin_addr)) {
    259 		sin->sin_port = 0;		/* yech... */
    260 		INADDR_TO_IA(sin->sin_addr, ia);
    261 		/* check for broadcast addresses */
    262 		if (ia == NULL)
    263 			ia = ifatoia(ifa_ifwithaddr(sintosa(sin)));
    264 		if (ia == NULL)
    265 			return (EADDRNOTAVAIL);
    266 	}
    267 	if (lport) {
    268 		struct inpcb *t;
    269 #ifdef INET6
    270 		struct in6pcb *t6;
    271 		struct in6_addr mapped;
    272 #endif
    273 #ifndef IPNOPRIVPORTS
    274 		/* GROSS */
    275 		if (ntohs(lport) < IPPORT_RESERVED &&
    276 		    (p == 0 || suser(p->p_ucred, &p->p_acflag)))
    277 			return (EACCES);
    278 #endif
    279 #ifdef INET6
    280 		memset(&mapped, 0, sizeof(mapped));
    281 		mapped.s6_addr16[5] = 0xffff;
    282 		memcpy(&mapped.s6_addr32[3], &sin->sin_addr,
    283 		    sizeof(mapped.s6_addr32[3]));
    284 		t6 = in6_pcblookup_port(table, &mapped, lport, wild);
    285 		if (t6 && (reuseport & t6->in6p_socket->so_options) == 0)
    286 			return (EADDRINUSE);
    287 #endif
    288 		if (so->so_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
    289 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
    290 		/*
    291 		 * XXX:	investigate ramifications of loosening this
    292 		 *	restriction so that as long as both ports have
    293 		 *	SO_REUSEPORT allow the bind
    294 		 */
    295 			if (t &&
    296 			    (!in_nullhost(sin->sin_addr) ||
    297 			     !in_nullhost(t->inp_laddr) ||
    298 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
    299 			    && (so->so_uid != t->inp_socket->so_uid)) {
    300 				return (EADDRINUSE);
    301 			}
    302 		}
    303 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
    304 		if (t && (reuseport & t->inp_socket->so_options) == 0)
    305 			return (EADDRINUSE);
    306 	}
    307 	inp->inp_laddr = sin->sin_addr;
    308 
    309 noname:
    310 	if (lport == 0) {
    311 		int	   cnt;
    312 		u_int16_t  min, max;
    313 		u_int16_t *lastport;
    314 
    315 		if (inp->inp_flags & INP_LOWPORT) {
    316 #ifndef IPNOPRIVPORTS
    317 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
    318 				return (EACCES);
    319 #endif
    320 			min = lowportmin;
    321 			max = lowportmax;
    322 			lastport = &table->inpt_lastlow;
    323 		} else {
    324 			min = anonportmin;
    325 			max = anonportmax;
    326 			lastport = &table->inpt_lastport;
    327 		}
    328 		if (min > max) {	/* sanity check */
    329 			u_int16_t swp;
    330 
    331 			swp = min;
    332 			min = max;
    333 			max = swp;
    334 		}
    335 
    336 		lport = *lastport - 1;
    337 		for (cnt = max - min + 1; cnt; cnt--, lport--) {
    338 			if (lport < min || lport > max)
    339 				lport = max;
    340 			if (!in_pcblookup_port(table, inp->inp_laddr,
    341 			    htons(lport), 1))
    342 				goto found;
    343 		}
    344 		if (!in_nullhost(inp->inp_laddr))
    345 			inp->inp_laddr.s_addr = INADDR_ANY;
    346 		return (EAGAIN);
    347 	found:
    348 		inp->inp_flags |= INP_ANONPORT;
    349 		*lastport = lport;
    350 		lport = htons(lport);
    351 	}
    352 	inp->inp_lport = lport;
    353 	LIST_REMOVE(&inp->inp_head, inph_lhash);
    354 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
    355 	    inph_lhash);
    356 	in_pcbstate(inp, INP_BOUND);
    357 	return (0);
    358 }
    359 
    360 /*
    361  * Connect from a socket to a specified address.
    362  * Both address and port must be specified in argument sin.
    363  * If don't have a local address for this socket yet,
    364  * then pick one.
    365  */
    366 int
    367 in_pcbconnect(v, nam)
    368 	void *v;
    369 	struct mbuf *nam;
    370 {
    371 	struct inpcb *inp = v;
    372 	struct in_ifaddr *ia = NULL;
    373 	struct sockaddr_in *ifaddr = NULL;
    374 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
    375 	int error;
    376 
    377 	if (inp->inp_af != AF_INET)
    378 		return (EINVAL);
    379 
    380 	if (nam->m_len != sizeof (*sin))
    381 		return (EINVAL);
    382 	if (sin->sin_family != AF_INET)
    383 		return (EAFNOSUPPORT);
    384 	if (sin->sin_port == 0)
    385 		return (EADDRNOTAVAIL);
    386 	if (TAILQ_FIRST(&in_ifaddrhead) != 0) {
    387 		/*
    388 		 * If the destination address is INADDR_ANY,
    389 		 * use any local address (likely loopback).
    390 		 * If the supplied address is INADDR_BROADCAST,
    391 		 * use the broadcast address of an interface
    392 		 * which supports broadcast. (loopback does not)
    393 		 */
    394 
    395 		if (in_nullhost(sin->sin_addr)) {
    396 			sin->sin_addr =
    397 			    TAILQ_FIRST(&in_ifaddrhead)->ia_addr.sin_addr;
    398 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
    399 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
    400 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
    401 					sin->sin_addr =
    402 					    ia->ia_broadaddr.sin_addr;
    403 					break;
    404 				}
    405 			}
    406 		}
    407 	}
    408 	/*
    409 	 * If we haven't bound which network number to use as ours,
    410 	 * we will use the number of the outgoing interface.
    411 	 * This depends on having done a routing lookup, which
    412 	 * we will probably have to do anyway, so we might
    413 	 * as well do it now.  On the other hand if we are
    414 	 * sending to multiple destinations we may have already
    415 	 * done the lookup, so see if we can use the route
    416 	 * from before.  In any case, we only
    417 	 * chose a port number once, even if sending to multiple
    418 	 * destinations.
    419 	 */
    420 	if (in_nullhost(inp->inp_laddr)) {
    421 		int error;
    422 		ifaddr = in_selectsrc(sin, &inp->inp_route,
    423 			inp->inp_socket->so_options, inp->inp_moptions, &error);
    424 		if (ifaddr == NULL) {
    425 			if (error == 0)
    426 				error = EADDRNOTAVAIL;
    427 			return error;
    428 		}
    429 		INADDR_TO_IA(ifaddr->sin_addr, ia);
    430 		if (ia == NULL)
    431 			return (EADDRNOTAVAIL);
    432 	}
    433 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
    434 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
    435 	    inp->inp_lport) != 0)
    436 		return (EADDRINUSE);
    437 	if (in_nullhost(inp->inp_laddr)) {
    438 		if (inp->inp_lport == 0) {
    439 			error = in_pcbbind(inp, (struct mbuf *)0,
    440 			    (struct proc *)0);
    441 			/*
    442 			 * This used to ignore the return value
    443 			 * completely, but we need to check for
    444 			 * ephemeral port shortage.
    445 			 * XXX Should we check for other errors, too?
    446 			 */
    447 			if (error == EAGAIN)
    448 				return (error);
    449 		}
    450 		inp->inp_laddr = ia->ia_addr.sin_addr;
    451 		inp->inp_laddr = ifaddr->sin_addr;
    452 	}
    453 	inp->inp_faddr = sin->sin_addr;
    454 	inp->inp_fport = sin->sin_port;
    455 	in_pcbstate(inp, INP_CONNECTED);
    456 #if defined(IPSEC) || defined(FAST_IPSEC)
    457 	if (inp->inp_socket->so_type == SOCK_STREAM)
    458 		ipsec_pcbconn(inp->inp_sp);
    459 #endif
    460 	return (0);
    461 }
    462 
    463 void
    464 in_pcbdisconnect(v)
    465 	void *v;
    466 {
    467 	struct inpcb *inp = v;
    468 
    469 	if (inp->inp_af != AF_INET)
    470 		return;
    471 
    472 	inp->inp_faddr = zeroin_addr;
    473 	inp->inp_fport = 0;
    474 	in_pcbstate(inp, INP_BOUND);
    475 #if defined(IPSEC) || defined(FAST_IPSEC)
    476 	ipsec_pcbdisconn(inp->inp_sp);
    477 #endif
    478 	if (inp->inp_socket->so_state & SS_NOFDREF)
    479 		in_pcbdetach(inp);
    480 }
    481 
    482 void
    483 in_pcbdetach(v)
    484 	void *v;
    485 {
    486 	struct inpcb *inp = v;
    487 	struct socket *so = inp->inp_socket;
    488 	int s;
    489 
    490 	if (inp->inp_af != AF_INET)
    491 		return;
    492 
    493 #if defined(IPSEC) || defined(FAST_IPSEC)
    494 	ipsec4_delete_pcbpolicy(inp);
    495 #endif /*IPSEC*/
    496 	so->so_pcb = 0;
    497 	sofree(so);
    498 	if (inp->inp_options)
    499 		(void)m_free(inp->inp_options);
    500 	if (inp->inp_route.ro_rt)
    501 		rtfree(inp->inp_route.ro_rt);
    502 	ip_freemoptions(inp->inp_moptions);
    503 	s = splnet();
    504 	in_pcbstate(inp, INP_ATTACHED);
    505 	LIST_REMOVE(&inp->inp_head, inph_lhash);
    506 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, &inp->inp_head,
    507 	    inph_queue);
    508 	splx(s);
    509 	pool_put(&inpcb_pool, inp);
    510 }
    511 
    512 void
    513 in_setsockaddr(inp, nam)
    514 	struct inpcb *inp;
    515 	struct mbuf *nam;
    516 {
    517 	struct sockaddr_in *sin;
    518 
    519 	if (inp->inp_af != AF_INET)
    520 		return;
    521 
    522 	nam->m_len = sizeof (*sin);
    523 	sin = mtod(nam, struct sockaddr_in *);
    524 	bzero((caddr_t)sin, sizeof (*sin));
    525 	sin->sin_family = AF_INET;
    526 	sin->sin_len = sizeof(*sin);
    527 	sin->sin_port = inp->inp_lport;
    528 	sin->sin_addr = inp->inp_laddr;
    529 }
    530 
    531 void
    532 in_setpeeraddr(inp, nam)
    533 	struct inpcb *inp;
    534 	struct mbuf *nam;
    535 {
    536 	struct sockaddr_in *sin;
    537 
    538 	if (inp->inp_af != AF_INET)
    539 		return;
    540 
    541 	nam->m_len = sizeof (*sin);
    542 	sin = mtod(nam, struct sockaddr_in *);
    543 	bzero((caddr_t)sin, sizeof (*sin));
    544 	sin->sin_family = AF_INET;
    545 	sin->sin_len = sizeof(*sin);
    546 	sin->sin_port = inp->inp_fport;
    547 	sin->sin_addr = inp->inp_faddr;
    548 }
    549 
    550 /*
    551  * Pass some notification to all connections of a protocol
    552  * associated with address dst.  The local address and/or port numbers
    553  * may be specified to limit the search.  The "usual action" will be
    554  * taken, depending on the ctlinput cmd.  The caller must filter any
    555  * cmds that are uninteresting (e.g., no error in the map).
    556  * Call the protocol specific routine (if any) to report
    557  * any errors for each matching socket.
    558  *
    559  * Must be called at splsoftnet.
    560  */
    561 int
    562 in_pcbnotify(table, faddr, fport_arg, laddr, lport_arg, errno, notify)
    563 	struct inpcbtable *table;
    564 	struct in_addr faddr, laddr;
    565 	u_int fport_arg, lport_arg;
    566 	int errno;
    567 	void (*notify) __P((struct inpcb *, int));
    568 {
    569 	struct inpcbhead *head;
    570 	struct inpcb *inp, *ninp;
    571 	u_int16_t fport = fport_arg, lport = lport_arg;
    572 	int nmatch;
    573 
    574 	if (in_nullhost(faddr) || notify == 0)
    575 		return (0);
    576 
    577 	nmatch = 0;
    578 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    579 	for (inp = (struct inpcb *)LIST_FIRST(head); inp != NULL; inp = ninp) {
    580 		ninp = (struct inpcb *)LIST_NEXT(inp, inp_hash);
    581 		if (inp->inp_af != AF_INET)
    582 			continue;
    583 		if (in_hosteq(inp->inp_faddr, faddr) &&
    584 		    inp->inp_fport == fport &&
    585 		    inp->inp_lport == lport &&
    586 		    in_hosteq(inp->inp_laddr, laddr)) {
    587 			(*notify)(inp, errno);
    588 			nmatch++;
    589 		}
    590 	}
    591 	return (nmatch);
    592 }
    593 
    594 void
    595 in_pcbnotifyall(table, faddr, errno, notify)
    596 	struct inpcbtable *table;
    597 	struct in_addr faddr;
    598 	int errno;
    599 	void (*notify) __P((struct inpcb *, int));
    600 {
    601 	struct inpcb *inp, *ninp;
    602 
    603 	if (in_nullhost(faddr) || notify == 0)
    604 		return;
    605 
    606 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
    607 	    inp != (void *)&table->inpt_queue;
    608 	    inp = ninp) {
    609 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
    610 		if (inp->inp_af != AF_INET)
    611 			continue;
    612 		if (in_hosteq(inp->inp_faddr, faddr))
    613 			(*notify)(inp, errno);
    614 	}
    615 }
    616 
    617 void
    618 in_pcbpurgeif0(table, ifp)
    619 	struct inpcbtable *table;
    620 	struct ifnet *ifp;
    621 {
    622 	struct inpcb *inp, *ninp;
    623 	struct ip_moptions *imo;
    624 	int i, gap;
    625 
    626 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
    627 	    inp != (void *)&table->inpt_queue;
    628 	    inp = ninp) {
    629 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
    630 		if (inp->inp_af != AF_INET)
    631 			continue;
    632 		imo = inp->inp_moptions;
    633 		if (imo != NULL) {
    634 			/*
    635 			 * Unselect the outgoing interface if it is being
    636 			 * detached.
    637 			 */
    638 			if (imo->imo_multicast_ifp == ifp)
    639 				imo->imo_multicast_ifp = NULL;
    640 
    641 			/*
    642 			 * Drop multicast group membership if we joined
    643 			 * through the interface being detached.
    644 			 */
    645 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
    646 			    i++) {
    647 				if (imo->imo_membership[i]->inm_ifp == ifp) {
    648 					in_delmulti(imo->imo_membership[i]);
    649 					gap++;
    650 				} else if (gap != 0)
    651 					imo->imo_membership[i - gap] =
    652 					    imo->imo_membership[i];
    653 			}
    654 			imo->imo_num_memberships -= gap;
    655 		}
    656 	}
    657 }
    658 
    659 void
    660 in_pcbpurgeif(table, ifp)
    661 	struct inpcbtable *table;
    662 	struct ifnet *ifp;
    663 {
    664 	struct inpcb *inp, *ninp;
    665 
    666 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
    667 	    inp != (void *)&table->inpt_queue;
    668 	    inp = ninp) {
    669 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
    670 		if (inp->inp_af != AF_INET)
    671 			continue;
    672 		if (inp->inp_route.ro_rt != NULL &&
    673 		    inp->inp_route.ro_rt->rt_ifp == ifp)
    674 			in_rtchange(inp, 0);
    675 	}
    676 }
    677 
    678 /*
    679  * Check for alternatives when higher level complains
    680  * about service problems.  For now, invalidate cached
    681  * routing information.  If the route was created dynamically
    682  * (by a redirect), time to try a default gateway again.
    683  */
    684 void
    685 in_losing(inp)
    686 	struct inpcb *inp;
    687 {
    688 	struct rtentry *rt;
    689 	struct rt_addrinfo info;
    690 
    691 	if (inp->inp_af != AF_INET)
    692 		return;
    693 
    694 	if ((rt = inp->inp_route.ro_rt)) {
    695 		inp->inp_route.ro_rt = 0;
    696 		bzero((caddr_t)&info, sizeof(info));
    697 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
    698 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    699 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    700 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
    701 		if (rt->rt_flags & RTF_DYNAMIC)
    702 			(void) rtrequest(RTM_DELETE, rt_key(rt),
    703 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
    704 				(struct rtentry **)0);
    705 		else
    706 		/*
    707 		 * A new route can be allocated
    708 		 * the next time output is attempted.
    709 		 */
    710 			rtfree(rt);
    711 	}
    712 }
    713 
    714 /*
    715  * After a routing change, flush old routing
    716  * and allocate a (hopefully) better one.
    717  */
    718 void
    719 in_rtchange(inp, errno)
    720 	struct inpcb *inp;
    721 	int errno;
    722 {
    723 
    724 	if (inp->inp_af != AF_INET)
    725 		return;
    726 
    727 	if (inp->inp_route.ro_rt) {
    728 		rtfree(inp->inp_route.ro_rt);
    729 		inp->inp_route.ro_rt = 0;
    730 		/*
    731 		 * A new route can be allocated the next time
    732 		 * output is attempted.
    733 		 */
    734 	}
    735 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
    736 }
    737 
    738 struct inpcb *
    739 in_pcblookup_port(table, laddr, lport_arg, lookup_wildcard)
    740 	struct inpcbtable *table;
    741 	struct in_addr laddr;
    742 	u_int lport_arg;
    743 	int lookup_wildcard;
    744 {
    745 	struct inpcbhead *head;
    746 	struct inpcb_hdr *inph;
    747 	struct inpcb *inp, *match = 0;
    748 	int matchwild = 3, wildcard;
    749 	u_int16_t lport = lport_arg;
    750 
    751 	head = INPCBHASH_PORT(table, lport);
    752 	LIST_FOREACH(inph, head, inph_lhash) {
    753 		inp = (struct inpcb *)inph;
    754 		if (inp->inp_af != AF_INET)
    755 			continue;
    756 
    757 		if (inp->inp_lport != lport)
    758 			continue;
    759 		wildcard = 0;
    760 		if (!in_nullhost(inp->inp_faddr))
    761 			wildcard++;
    762 		if (in_nullhost(inp->inp_laddr)) {
    763 			if (!in_nullhost(laddr))
    764 				wildcard++;
    765 		} else {
    766 			if (in_nullhost(laddr))
    767 				wildcard++;
    768 			else {
    769 				if (!in_hosteq(inp->inp_laddr, laddr))
    770 					continue;
    771 			}
    772 		}
    773 		if (wildcard && !lookup_wildcard)
    774 			continue;
    775 		if (wildcard < matchwild) {
    776 			match = inp;
    777 			matchwild = wildcard;
    778 			if (matchwild == 0)
    779 				break;
    780 		}
    781 	}
    782 	return (match);
    783 }
    784 
    785 #ifdef DIAGNOSTIC
    786 int	in_pcbnotifymiss = 0;
    787 #endif
    788 
    789 struct inpcb *
    790 in_pcblookup_connect(table, faddr, fport_arg, laddr, lport_arg)
    791 	struct inpcbtable *table;
    792 	struct in_addr faddr, laddr;
    793 	u_int fport_arg, lport_arg;
    794 {
    795 	struct inpcbhead *head;
    796 	struct inpcb_hdr *inph;
    797 	struct inpcb *inp;
    798 	u_int16_t fport = fport_arg, lport = lport_arg;
    799 
    800 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    801 	LIST_FOREACH(inph, head, inph_hash) {
    802 		inp = (struct inpcb *)inph;
    803 		if (inp->inp_af != AF_INET)
    804 			continue;
    805 
    806 		if (in_hosteq(inp->inp_faddr, faddr) &&
    807 		    inp->inp_fport == fport &&
    808 		    inp->inp_lport == lport &&
    809 		    in_hosteq(inp->inp_laddr, laddr))
    810 			goto out;
    811 	}
    812 #ifdef DIAGNOSTIC
    813 	if (in_pcbnotifymiss) {
    814 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
    815 		    ntohl(faddr.s_addr), ntohs(fport),
    816 		    ntohl(laddr.s_addr), ntohs(lport));
    817 	}
    818 #endif
    819 	return (0);
    820 
    821 out:
    822 	/* Move this PCB to the head of hash chain. */
    823 	inph = &inp->inp_head;
    824 	if (inph != LIST_FIRST(head)) {
    825 		LIST_REMOVE(inph, inph_hash);
    826 		LIST_INSERT_HEAD(head, inph, inph_hash);
    827 	}
    828 	return (inp);
    829 }
    830 
    831 struct inpcb *
    832 in_pcblookup_bind(table, laddr, lport_arg)
    833 	struct inpcbtable *table;
    834 	struct in_addr laddr;
    835 	u_int lport_arg;
    836 {
    837 	struct inpcbhead *head;
    838 	struct inpcb_hdr *inph;
    839 	struct inpcb *inp;
    840 	u_int16_t lport = lport_arg;
    841 
    842 	head = INPCBHASH_BIND(table, laddr, lport);
    843 	LIST_FOREACH(inph, head, inph_hash) {
    844 		inp = (struct inpcb *)inph;
    845 		if (inp->inp_af != AF_INET)
    846 			continue;
    847 
    848 		if (inp->inp_lport == lport &&
    849 		    in_hosteq(inp->inp_laddr, laddr))
    850 			goto out;
    851 	}
    852 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
    853 	LIST_FOREACH(inph, head, inph_hash) {
    854 		inp = (struct inpcb *)inph;
    855 		if (inp->inp_af != AF_INET)
    856 			continue;
    857 
    858 		if (inp->inp_lport == lport &&
    859 		    in_hosteq(inp->inp_laddr, zeroin_addr))
    860 			goto out;
    861 	}
    862 #ifdef DIAGNOSTIC
    863 	if (in_pcbnotifymiss) {
    864 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
    865 		    ntohl(laddr.s_addr), ntohs(lport));
    866 	}
    867 #endif
    868 	return (0);
    869 
    870 out:
    871 	/* Move this PCB to the head of hash chain. */
    872 	inph = &inp->inp_head;
    873 	if (inph != LIST_FIRST(head)) {
    874 		LIST_REMOVE(inph, inph_hash);
    875 		LIST_INSERT_HEAD(head, inph, inph_hash);
    876 	}
    877 	return (inp);
    878 }
    879 
    880 void
    881 in_pcbstate(inp, state)
    882 	struct inpcb *inp;
    883 	int state;
    884 {
    885 
    886 	if (inp->inp_af != AF_INET)
    887 		return;
    888 
    889 	if (inp->inp_state > INP_ATTACHED)
    890 		LIST_REMOVE(&inp->inp_head, inph_hash);
    891 
    892 	switch (state) {
    893 	case INP_BOUND:
    894 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
    895 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
    896 		    inph_hash);
    897 		break;
    898 	case INP_CONNECTED:
    899 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
    900 		    inp->inp_faddr, inp->inp_fport,
    901 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
    902 		    inph_hash);
    903 		break;
    904 	}
    905 
    906 	inp->inp_state = state;
    907 }
    908 
    909 struct rtentry *
    910 in_pcbrtentry(inp)
    911 	struct inpcb *inp;
    912 {
    913 	struct route *ro;
    914 
    915 	if (inp->inp_af != AF_INET)
    916 		return (NULL);
    917 
    918 	ro = &inp->inp_route;
    919 
    920 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    921 	    !in_hosteq(satosin(&ro->ro_dst)->sin_addr, inp->inp_faddr))) {
    922 		RTFREE(ro->ro_rt);
    923 		ro->ro_rt = (struct rtentry *)NULL;
    924 	}
    925 	if (ro->ro_rt == (struct rtentry *)NULL &&
    926 	    !in_nullhost(inp->inp_faddr)) {
    927 		bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
    928 		ro->ro_dst.sa_family = AF_INET;
    929 		ro->ro_dst.sa_len = sizeof(ro->ro_dst);
    930 		satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
    931 		rtalloc(ro);
    932 	}
    933 	return (ro->ro_rt);
    934 }
    935 
    936 struct sockaddr_in *
    937 in_selectsrc(sin, ro, soopts, mopts, errorp)
    938 	struct sockaddr_in *sin;
    939 	struct route *ro;
    940 	int soopts;
    941 	struct ip_moptions *mopts;
    942 	int *errorp;
    943 {
    944 	struct in_ifaddr *ia;
    945 
    946 	ia = (struct in_ifaddr *)0;
    947 	/*
    948 	 * If route is known or can be allocated now,
    949 	 * our src addr is taken from the i/f, else punt.
    950 	 * Note that we should check the address family of the cached
    951 	 * destination, in case of sharing the cache with IPv6.
    952 	 */
    953 	if (ro->ro_rt &&
    954 	    (ro->ro_dst.sa_family != AF_INET ||
    955 	    !in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
    956 	    soopts & SO_DONTROUTE)) {
    957 		RTFREE(ro->ro_rt);
    958 		ro->ro_rt = (struct rtentry *)0;
    959 	}
    960 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
    961 	    (ro->ro_rt == (struct rtentry *)0 ||
    962 	     ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    963 		/* No route yet, so try to acquire one */
    964 		bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
    965 		ro->ro_dst.sa_family = AF_INET;
    966 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    967 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    968 		rtalloc(ro);
    969 	}
    970 	/*
    971 	 * If we found a route, use the address
    972 	 * corresponding to the outgoing interface
    973 	 * unless it is the loopback (in case a route
    974 	 * to our address on another net goes to loopback).
    975 	 *
    976 	 * XXX Is this still true?  Do we care?
    977 	 */
    978 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    979 		ia = ifatoia(ro->ro_rt->rt_ifa);
    980 	if (ia == NULL) {
    981 		u_int16_t fport = sin->sin_port;
    982 
    983 		sin->sin_port = 0;
    984 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    985 		sin->sin_port = fport;
    986 		if (ia == 0) {
    987 			/* Find 1st non-loopback AF_INET address */
    988 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
    989 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
    990 					break;
    991 			}
    992 		}
    993 		if (ia == NULL) {
    994 			*errorp = EADDRNOTAVAIL;
    995 			return NULL;
    996 		}
    997 	}
    998 	/*
    999 	 * If the destination address is multicast and an outgoing
   1000 	 * interface has been set as a multicast option, use the
   1001 	 * address of that interface as our source address.
   1002 	 */
   1003 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
   1004 		struct ip_moptions *imo;
   1005 		struct ifnet *ifp;
   1006 
   1007 		imo = mopts;
   1008 		if (imo->imo_multicast_ifp != NULL) {
   1009 			ifp = imo->imo_multicast_ifp;
   1010 			IFP_TO_IA(ifp, ia);		/* XXX */
   1011 			if (ia == 0) {
   1012 				*errorp = EADDRNOTAVAIL;
   1013 				return NULL;
   1014 			}
   1015 		}
   1016 	}
   1017 	return satosin(&ia->ia_addr);
   1018 }
   1019