Home | History | Annotate | Line # | Download | only in netinet
in_pcb.c revision 1.65.4.1
      1 /*	$NetBSD: in_pcb.c,v 1.65.4.1 2000/08/26 16:38:32 tron Exp $	*/
      2 
      3 /*
      4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      5  * All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the project nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  */
     31 
     32 /*-
     33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
     34  * All rights reserved.
     35  *
     36  * This code is derived from software contributed to The NetBSD Foundation
     37  * by Public Access Networks Corporation ("Panix").  It was developed under
     38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
     39  *
     40  * Redistribution and use in source and binary forms, with or without
     41  * modification, are permitted provided that the following conditions
     42  * are met:
     43  * 1. Redistributions of source code must retain the above copyright
     44  *    notice, this list of conditions and the following disclaimer.
     45  * 2. Redistributions in binary form must reproduce the above copyright
     46  *    notice, this list of conditions and the following disclaimer in the
     47  *    documentation and/or other materials provided with the distribution.
     48  * 3. All advertising materials mentioning features or use of this software
     49  *    must display the following acknowledgement:
     50  *	This product includes software developed by the NetBSD
     51  *	Foundation, Inc. and its contributors.
     52  * 4. Neither the name of The NetBSD Foundation nor the names of its
     53  *    contributors may be used to endorse or promote products derived
     54  *    from this software without specific prior written permission.
     55  *
     56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     66  * POSSIBILITY OF SUCH DAMAGE.
     67  */
     68 
     69 /*
     70  * Copyright (c) 1982, 1986, 1991, 1993, 1995
     71  *	The Regents of the University of California.  All rights reserved.
     72  *
     73  * Redistribution and use in source and binary forms, with or without
     74  * modification, are permitted provided that the following conditions
     75  * are met:
     76  * 1. Redistributions of source code must retain the above copyright
     77  *    notice, this list of conditions and the following disclaimer.
     78  * 2. Redistributions in binary form must reproduce the above copyright
     79  *    notice, this list of conditions and the following disclaimer in the
     80  *    documentation and/or other materials provided with the distribution.
     81  * 3. All advertising materials mentioning features or use of this software
     82  *    must display the following acknowledgement:
     83  *	This product includes software developed by the University of
     84  *	California, Berkeley and its contributors.
     85  * 4. Neither the name of the University nor the names of its contributors
     86  *    may be used to endorse or promote products derived from this software
     87  *    without specific prior written permission.
     88  *
     89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     99  * SUCH DAMAGE.
    100  *
    101  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
    102  */
    103 
    104 #include "opt_ipsec.h"
    105 
    106 #include <sys/param.h>
    107 #include <sys/systm.h>
    108 #include <sys/malloc.h>
    109 #include <sys/mbuf.h>
    110 #include <sys/protosw.h>
    111 #include <sys/socket.h>
    112 #include <sys/socketvar.h>
    113 #include <sys/ioctl.h>
    114 #include <sys/errno.h>
    115 #include <sys/time.h>
    116 #include <sys/pool.h>
    117 #include <sys/proc.h>
    118 
    119 #include <net/if.h>
    120 #include <net/route.h>
    121 
    122 #include <netinet/in.h>
    123 #include <netinet/in_systm.h>
    124 #include <netinet/ip.h>
    125 #include <netinet/in_pcb.h>
    126 #include <netinet/in_var.h>
    127 #include <netinet/ip_var.h>
    128 
    129 #ifdef IPSEC
    130 #include <netinet6/ipsec.h>
    131 #include <netkey/key.h>
    132 #include <netkey/key_debug.h>
    133 #endif /* IPSEC */
    134 
    135 struct	in_addr zeroin_addr;
    136 
    137 #define	INPCBHASH_BIND(table, laddr, lport) \
    138 	&(table)->inpt_bindhashtbl[ \
    139 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
    140 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
    141 	&(table)->inpt_connecthashtbl[ \
    142 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
    143 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
    144 
    145 struct inpcb *
    146 	in_pcblookup_port __P((struct inpcbtable *,
    147 	    struct in_addr, u_int, int));
    148 
    149 int	anonportmin = IPPORT_ANONMIN;
    150 int	anonportmax = IPPORT_ANONMAX;
    151 int	lowportmin  = IPPORT_RESERVEDMIN;
    152 int	lowportmax  = IPPORT_RESERVEDMAX;
    153 
    154 struct pool inpcb_pool;
    155 
    156 void
    157 in_pcbinit(table, bindhashsize, connecthashsize)
    158 	struct inpcbtable *table;
    159 	int bindhashsize, connecthashsize;
    160 {
    161 	static int inpcb_pool_initialized;
    162 
    163 	if (inpcb_pool_initialized == 0) {
    164 		pool_init(&inpcb_pool, sizeof(struct inpcb), 0, 0, 0,
    165 		    "inpcbpl", 0, NULL, NULL, M_PCB);
    166 		inpcb_pool_initialized = 1;
    167 	}
    168 
    169 	CIRCLEQ_INIT(&table->inpt_queue);
    170 	table->inpt_bindhashtbl =
    171 	    hashinit(bindhashsize, M_PCB, M_WAITOK, &table->inpt_bindhash);
    172 	table->inpt_connecthashtbl =
    173 	    hashinit(connecthashsize, M_PCB, M_WAITOK, &table->inpt_connecthash);
    174 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
    175 	table->inpt_lastport = (u_int16_t)anonportmax;
    176 }
    177 
    178 int
    179 in_pcballoc(so, v)
    180 	struct socket *so;
    181 	void *v;
    182 {
    183 	struct inpcbtable *table = v;
    184 	struct inpcb *inp;
    185 	int s;
    186 
    187 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
    188 	if (inp == NULL)
    189 		return (ENOBUFS);
    190 	bzero((caddr_t)inp, sizeof(*inp));
    191 	inp->inp_table = table;
    192 	inp->inp_socket = so;
    193 	inp->inp_errormtu = -1;
    194 	so->so_pcb = inp;
    195 	s = splnet();
    196 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, inp, inp_queue);
    197 	in_pcbstate(inp, INP_ATTACHED);
    198 	splx(s);
    199 	return (0);
    200 }
    201 
    202 int
    203 in_pcbbind(v, nam, p)
    204 	void *v;
    205 	struct mbuf *nam;
    206 	struct proc *p;
    207 {
    208 	struct inpcb *inp = v;
    209 	struct socket *so = inp->inp_socket;
    210 	struct inpcbtable *table = inp->inp_table;
    211 	struct sockaddr_in *sin;
    212 	u_int16_t lport = 0;
    213 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
    214 #ifndef IPNOPRIVPORTS
    215 	int error;
    216 #endif
    217 
    218 	if (in_ifaddr.tqh_first == 0)
    219 		return (EADDRNOTAVAIL);
    220 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
    221 		return (EINVAL);
    222 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
    223 		wild = 1;
    224 	if (nam == 0)
    225 		goto noname;
    226 	sin = mtod(nam, struct sockaddr_in *);
    227 	if (nam->m_len != sizeof (*sin))
    228 		return (EINVAL);
    229 #ifdef notdef
    230 	/*
    231 	 * We should check the family, but old programs
    232 	 * incorrectly fail to initialize it.
    233 	 */
    234 	if (sin->sin_family != AF_INET)
    235 		return (EAFNOSUPPORT);
    236 #endif
    237 	lport = sin->sin_port;
    238 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
    239 		/*
    240 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
    241 		 * allow complete duplication of binding if
    242 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
    243 		 * and a multicast address is bound on both
    244 		 * new and duplicated sockets.
    245 		 */
    246 		if (so->so_options & SO_REUSEADDR)
    247 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
    248 	} else if (!in_nullhost(sin->sin_addr)) {
    249 		sin->sin_port = 0;		/* yech... */
    250 		if (ifa_ifwithaddr(sintosa(sin)) == 0)
    251 			return (EADDRNOTAVAIL);
    252 	}
    253 	if (lport) {
    254 		struct inpcb *t;
    255 #ifndef IPNOPRIVPORTS
    256 		/* GROSS */
    257 		if (ntohs(lport) < IPPORT_RESERVED &&
    258 		    (p == 0 || (error = suser(p->p_ucred, &p->p_acflag))))
    259 			return (EACCES);
    260 #endif
    261 		if (so->so_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
    262 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
    263 		/*
    264 		 * XXX:	investigate ramifications of loosening this
    265 		 *	restriction so that as long as both ports have
    266 		 *	SO_REUSEPORT allow the bind
    267 		 */
    268 			if (t &&
    269 			    (!in_nullhost(sin->sin_addr) ||
    270 			     !in_nullhost(t->inp_laddr) ||
    271 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
    272 			    && (so->so_uid != t->inp_socket->so_uid)) {
    273 				return (EADDRINUSE);
    274 			}
    275 		}
    276 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
    277 		if (t && (reuseport & t->inp_socket->so_options) == 0)
    278 			return (EADDRINUSE);
    279 	}
    280 	inp->inp_laddr = sin->sin_addr;
    281 
    282 noname:
    283 	if (lport == 0) {
    284 		int	   cnt;
    285 		u_int16_t  min, max;
    286 		u_int16_t *lastport;
    287 
    288 		if (inp->inp_flags & INP_LOWPORT) {
    289 #ifndef IPNOPRIVPORTS
    290 			if (p == 0 || (error = suser(p->p_ucred, &p->p_acflag)))
    291 				return (EACCES);
    292 #endif
    293 			min = lowportmin;
    294 			max = lowportmax;
    295 			lastport = &table->inpt_lastlow;
    296 		} else {
    297 			min = anonportmin;
    298 			max = anonportmax;
    299 			lastport = &table->inpt_lastport;
    300 		}
    301 		if (min > max) {	/* sanity check */
    302 			u_int16_t swp;
    303 
    304 			swp = min;
    305 			min = max;
    306 			max = swp;
    307 		}
    308 
    309 		lport = *lastport - 1;
    310 		for (cnt = max - min + 1; cnt; cnt--, lport--) {
    311 			if (lport < min || lport > max)
    312 				lport = max;
    313 			if (!in_pcblookup_port(table, inp->inp_laddr,
    314 			    htons(lport), 1))
    315 				goto found;
    316 		}
    317 		if (!in_nullhost(inp->inp_laddr))
    318 			inp->inp_laddr.s_addr = INADDR_ANY;
    319 		return (EAGAIN);
    320 	found:
    321 		inp->inp_flags |= INP_ANONPORT;
    322 		*lastport = lport;
    323 		lport = htons(lport);
    324 	}
    325 	inp->inp_lport = lport;
    326 	in_pcbstate(inp, INP_BOUND);
    327 	return (0);
    328 }
    329 
    330 /*
    331  * Connect from a socket to a specified address.
    332  * Both address and port must be specified in argument sin.
    333  * If don't have a local address for this socket yet,
    334  * then pick one.
    335  */
    336 int
    337 in_pcbconnect(v, nam)
    338 	void *v;
    339 	struct mbuf *nam;
    340 {
    341 	struct inpcb *inp = v;
    342 	struct in_ifaddr *ia;
    343 	struct sockaddr_in *ifaddr = NULL;
    344 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
    345 	int error;
    346 
    347 	if (nam->m_len != sizeof (*sin))
    348 		return (EINVAL);
    349 	if (sin->sin_family != AF_INET)
    350 		return (EAFNOSUPPORT);
    351 	if (sin->sin_port == 0)
    352 		return (EADDRNOTAVAIL);
    353 	if (in_ifaddr.tqh_first != 0) {
    354 		/*
    355 		 * If the destination address is INADDR_ANY,
    356 		 * use any local address (likely loopback).
    357 		 * If the supplied address is INADDR_BROADCAST,
    358 		 * use the broadcast address of an interface
    359 		 * which supports broadcast. (loopback does not)
    360 		 */
    361 
    362 		if (in_nullhost(sin->sin_addr))
    363 			sin->sin_addr = in_ifaddr.tqh_first->ia_addr.sin_addr;
    364 		else if (sin->sin_addr.s_addr == INADDR_BROADCAST)
    365 		    for (ia = in_ifaddr.tqh_first; ia != NULL;
    366 		      ia = ia->ia_list.tqe_next)
    367 			if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
    368 			    sin->sin_addr = ia->ia_broadaddr.sin_addr;
    369 			    break;
    370 			}
    371 	}
    372 	/*
    373 	 * If we haven't bound which network number to use as ours,
    374 	 * we will use the number of the outgoing interface.
    375 	 * This depends on having done a routing lookup, which
    376 	 * we will probably have to do anyway, so we might
    377 	 * as well do it now.  On the other hand if we are
    378 	 * sending to multiple destinations we may have already
    379 	 * done the lookup, so see if we can use the route
    380 	 * from before.  In any case, we only
    381 	 * chose a port number once, even if sending to multiple
    382 	 * destinations.
    383 	 */
    384 	if (in_nullhost(inp->inp_laddr)) {
    385 #if 0
    386 		struct route *ro;
    387 
    388 		ia = (struct in_ifaddr *)0;
    389 		/*
    390 		 * If route is known or can be allocated now,
    391 		 * our src addr is taken from the i/f, else punt.
    392 		 */
    393 		ro = &inp->inp_route;
    394 		if (ro->ro_rt &&
    395 		    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr,
    396 			sin->sin_addr) ||
    397 		    inp->inp_socket->so_options & SO_DONTROUTE)) {
    398 			RTFREE(ro->ro_rt);
    399 			ro->ro_rt = (struct rtentry *)0;
    400 		}
    401 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/
    402 		    (ro->ro_rt == (struct rtentry *)0 ||
    403 		    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    404 			/* No route yet, so try to acquire one */
    405 			ro->ro_dst.sa_family = AF_INET;
    406 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    407 			satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    408 			rtalloc(ro);
    409 		}
    410 		/*
    411 		 * If we found a route, use the address
    412 		 * corresponding to the outgoing interface
    413 		 * unless it is the loopback (in case a route
    414 		 * to our address on another net goes to loopback).
    415 		 *
    416 		 * XXX Is this still true?  Do we care?
    417 		 */
    418 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    419 			ia = ifatoia(ro->ro_rt->rt_ifa);
    420 		if (ia == NULL) {
    421 			u_int16_t fport = sin->sin_port;
    422 
    423 			sin->sin_port = 0;
    424 			ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    425 			sin->sin_port = fport;
    426 			if (ia == 0) {
    427 				/* Find 1st non-loopback AF_INET address */
    428 				for (ia = in_ifaddr.tqh_first ; ia != NULL;
    429 				     ia = ia->ia_list.tqe_next) {
    430 					if ((ia->ia_ifp->if_flags &
    431 					     IFF_LOOPBACK) == 0)
    432 						break;
    433 				}
    434 			}
    435 			if (ia == NULL)
    436 				return (EADDRNOTAVAIL);
    437 		}
    438 		/*
    439 		 * If the destination address is multicast and an outgoing
    440 		 * interface has been set as a multicast option, use the
    441 		 * address of that interface as our source address.
    442 		 */
    443 		if (IN_MULTICAST(sin->sin_addr.s_addr) &&
    444 		    inp->inp_moptions != NULL) {
    445 			struct ip_moptions *imo;
    446 			struct ifnet *ifp;
    447 
    448 			imo = inp->inp_moptions;
    449 			if (imo->imo_multicast_ifp != NULL) {
    450 				ifp = imo->imo_multicast_ifp;
    451 				IFP_TO_IA(ifp, ia);		/* XXX */
    452 				if (ia == 0)
    453 					return (EADDRNOTAVAIL);
    454 			}
    455 		}
    456 		ifaddr = satosin(&ia->ia_addr);
    457 #else
    458 		int error;
    459 		ifaddr = in_selectsrc(sin, &inp->inp_route,
    460 			inp->inp_socket->so_options, inp->inp_moptions, &error);
    461 		if (ifaddr == NULL) {
    462 			if (error == 0)
    463 				error = EADDRNOTAVAIL;
    464 			return error;
    465 		}
    466 #endif
    467 	}
    468 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
    469 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
    470 	    inp->inp_lport) != 0)
    471 		return (EADDRINUSE);
    472 	if (in_nullhost(inp->inp_laddr)) {
    473 		if (inp->inp_lport == 0) {
    474 			error = in_pcbbind(inp, (struct mbuf *)0,
    475 			    (struct proc *)0);
    476 			/*
    477 			 * This used to ignore the return value
    478 			 * completely, but we need to check for
    479 			 * ephemeral port shortage.
    480 			 * XXX Should we check for other errors, too?
    481 			 */
    482 			if (error == EAGAIN)
    483 				return (error);
    484 		}
    485 		inp->inp_laddr = ifaddr->sin_addr;
    486 	}
    487 	inp->inp_faddr = sin->sin_addr;
    488 	inp->inp_fport = sin->sin_port;
    489 	in_pcbstate(inp, INP_CONNECTED);
    490 	return (0);
    491 }
    492 
    493 void
    494 in_pcbdisconnect(v)
    495 	void *v;
    496 {
    497 	struct inpcb *inp = v;
    498 
    499 	inp->inp_faddr = zeroin_addr;
    500 	inp->inp_fport = 0;
    501 	in_pcbstate(inp, INP_BOUND);
    502 	if (inp->inp_socket->so_state & SS_NOFDREF)
    503 		in_pcbdetach(inp);
    504 }
    505 
    506 void
    507 in_pcbdetach(v)
    508 	void *v;
    509 {
    510 	struct inpcb *inp = v;
    511 	struct socket *so = inp->inp_socket;
    512 	int s;
    513 
    514 #ifdef IPSEC
    515 	ipsec4_delete_pcbpolicy(inp);
    516 #endif /*IPSEC*/
    517 	so->so_pcb = 0;
    518 	sofree(so);
    519 	if (inp->inp_options)
    520 		(void)m_free(inp->inp_options);
    521 	if (inp->inp_route.ro_rt)
    522 		rtfree(inp->inp_route.ro_rt);
    523 	ip_freemoptions(inp->inp_moptions);
    524 	s = splnet();
    525 	in_pcbstate(inp, INP_ATTACHED);
    526 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, inp, inp_queue);
    527 	splx(s);
    528 	pool_put(&inpcb_pool, inp);
    529 }
    530 
    531 void
    532 in_setsockaddr(inp, nam)
    533 	struct inpcb *inp;
    534 	struct mbuf *nam;
    535 {
    536 	struct sockaddr_in *sin;
    537 
    538 	nam->m_len = sizeof (*sin);
    539 	sin = mtod(nam, struct sockaddr_in *);
    540 	bzero((caddr_t)sin, sizeof (*sin));
    541 	sin->sin_family = AF_INET;
    542 	sin->sin_len = sizeof(*sin);
    543 	sin->sin_port = inp->inp_lport;
    544 	sin->sin_addr = inp->inp_laddr;
    545 }
    546 
    547 void
    548 in_setpeeraddr(inp, nam)
    549 	struct inpcb *inp;
    550 	struct mbuf *nam;
    551 {
    552 	struct sockaddr_in *sin;
    553 
    554 	nam->m_len = sizeof (*sin);
    555 	sin = mtod(nam, struct sockaddr_in *);
    556 	bzero((caddr_t)sin, sizeof (*sin));
    557 	sin->sin_family = AF_INET;
    558 	sin->sin_len = sizeof(*sin);
    559 	sin->sin_port = inp->inp_fport;
    560 	sin->sin_addr = inp->inp_faddr;
    561 }
    562 
    563 /*
    564  * Pass some notification to all connections of a protocol
    565  * associated with address dst.  The local address and/or port numbers
    566  * may be specified to limit the search.  The "usual action" will be
    567  * taken, depending on the ctlinput cmd.  The caller must filter any
    568  * cmds that are uninteresting (e.g., no error in the map).
    569  * Call the protocol specific routine (if any) to report
    570  * any errors for each matching socket.
    571  *
    572  * Must be called at splsoftnet.
    573  */
    574 int
    575 in_pcbnotify(table, faddr, fport_arg, laddr, lport_arg, errno, notify)
    576 	struct inpcbtable *table;
    577 	struct in_addr faddr, laddr;
    578 	u_int fport_arg, lport_arg;
    579 	int errno;
    580 	void (*notify) __P((struct inpcb *, int));
    581 {
    582 	struct inpcbhead *head;
    583 	struct inpcb *inp, *ninp;
    584 	u_int16_t fport = fport_arg, lport = lport_arg;
    585 	int nmatch;
    586 
    587 	if (in_nullhost(faddr) || notify == 0)
    588 		return (0);
    589 
    590 	nmatch = 0;
    591 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    592 	for (inp = head->lh_first; inp != NULL; inp = ninp) {
    593 		ninp = inp->inp_hash.le_next;
    594 		if (in_hosteq(inp->inp_faddr, faddr) &&
    595 		    inp->inp_fport == fport &&
    596 		    inp->inp_lport == lport &&
    597 		    in_hosteq(inp->inp_laddr, laddr)) {
    598 			(*notify)(inp, errno);
    599 			nmatch++;
    600 		}
    601 	}
    602 	return (nmatch);
    603 }
    604 
    605 void
    606 in_pcbnotifyall(table, faddr, errno, notify)
    607 	struct inpcbtable *table;
    608 	struct in_addr faddr;
    609 	int errno;
    610 	void (*notify) __P((struct inpcb *, int));
    611 {
    612 	struct inpcb *inp, *ninp;
    613 
    614 	if (in_nullhost(faddr) || notify == 0)
    615 		return;
    616 
    617 	for (inp = table->inpt_queue.cqh_first;
    618 	    inp != (struct inpcb *)&table->inpt_queue;
    619 	    inp = ninp) {
    620 		ninp = inp->inp_queue.cqe_next;
    621 		if (in_hosteq(inp->inp_faddr, faddr))
    622 			(*notify)(inp, errno);
    623 	}
    624 }
    625 
    626 void
    627 in_pcbpurgeif(table, ifp)
    628 	struct inpcbtable *table;
    629 	struct ifnet *ifp;
    630 {
    631 	struct inpcb *inp, *ninp;
    632 	struct ip_moptions *imo;
    633 	int i, gap;
    634 
    635 	for (inp = table->inpt_queue.cqh_first;
    636 	    inp != (struct inpcb *)&table->inpt_queue;
    637 	    inp = ninp) {
    638 		ninp = inp->inp_queue.cqe_next;
    639 		if (inp->inp_route.ro_rt != NULL &&
    640 		    inp->inp_route.ro_rt->rt_ifp == ifp)
    641 			in_rtchange(inp, 0);
    642 		imo = inp->inp_moptions;
    643 		if (imo != NULL) {
    644 			/*
    645 			 * Unselect the outgoing interface if it is being
    646 			 * detached.
    647 			 */
    648 			if (imo->imo_multicast_ifp == ifp)
    649 				imo->imo_multicast_ifp = NULL;
    650 
    651 			/*
    652 			 * Drop multicast group membership if we joined
    653 			 * through the interface being detached.
    654 			 */
    655 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
    656 			    i++) {
    657 				if (imo->imo_membership[i]->inm_ifp == ifp) {
    658 					in_delmulti(imo->imo_membership[i]);
    659 					gap++;
    660 				} else if (gap != 0)
    661 					imo->imo_membership[i - gap] =
    662 					    imo->imo_membership[i];
    663 			}
    664 			imo->imo_num_memberships -= gap;
    665 		}
    666 	}
    667 }
    668 
    669 /*
    670  * Check for alternatives when higher level complains
    671  * about service problems.  For now, invalidate cached
    672  * routing information.  If the route was created dynamically
    673  * (by a redirect), time to try a default gateway again.
    674  */
    675 void
    676 in_losing(inp)
    677 	struct inpcb *inp;
    678 {
    679 	struct rtentry *rt;
    680 	struct rt_addrinfo info;
    681 
    682 	if ((rt = inp->inp_route.ro_rt)) {
    683 		inp->inp_route.ro_rt = 0;
    684 		bzero((caddr_t)&info, sizeof(info));
    685 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
    686 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    687 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    688 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
    689 		if (rt->rt_flags & RTF_DYNAMIC)
    690 			(void) rtrequest(RTM_DELETE, rt_key(rt),
    691 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
    692 				(struct rtentry **)0);
    693 		else
    694 		/*
    695 		 * A new route can be allocated
    696 		 * the next time output is attempted.
    697 		 */
    698 			rtfree(rt);
    699 	}
    700 }
    701 
    702 /*
    703  * After a routing change, flush old routing
    704  * and allocate a (hopefully) better one.
    705  */
    706 void
    707 in_rtchange(inp, errno)
    708 	struct inpcb *inp;
    709 	int errno;
    710 {
    711 
    712 	if (inp->inp_route.ro_rt) {
    713 		rtfree(inp->inp_route.ro_rt);
    714 		inp->inp_route.ro_rt = 0;
    715 		/*
    716 		 * A new route can be allocated the next time
    717 		 * output is attempted.
    718 		 */
    719 	}
    720 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
    721 }
    722 
    723 struct inpcb *
    724 in_pcblookup_port(table, laddr, lport_arg, lookup_wildcard)
    725 	struct inpcbtable *table;
    726 	struct in_addr laddr;
    727 	u_int lport_arg;
    728 	int lookup_wildcard;
    729 {
    730 	struct inpcb *inp, *match = 0;
    731 	int matchwild = 3, wildcard;
    732 	u_int16_t lport = lport_arg;
    733 
    734 	for (inp = table->inpt_queue.cqh_first;
    735 	    inp != (struct inpcb *)&table->inpt_queue;
    736 	    inp = inp->inp_queue.cqe_next) {
    737 		if (inp->inp_lport != lport)
    738 			continue;
    739 		wildcard = 0;
    740 		if (!in_nullhost(inp->inp_faddr))
    741 			wildcard++;
    742 		if (in_nullhost(inp->inp_laddr)) {
    743 			if (!in_nullhost(laddr))
    744 				wildcard++;
    745 		} else {
    746 			if (in_nullhost(laddr))
    747 				wildcard++;
    748 			else {
    749 				if (!in_hosteq(inp->inp_laddr, laddr))
    750 					continue;
    751 			}
    752 		}
    753 		if (wildcard && !lookup_wildcard)
    754 			continue;
    755 		if (wildcard < matchwild) {
    756 			match = inp;
    757 			matchwild = wildcard;
    758 			if (matchwild == 0)
    759 				break;
    760 		}
    761 	}
    762 	return (match);
    763 }
    764 
    765 #ifdef DIAGNOSTIC
    766 int	in_pcbnotifymiss = 0;
    767 #endif
    768 
    769 struct inpcb *
    770 in_pcblookup_connect(table, faddr, fport_arg, laddr, lport_arg)
    771 	struct inpcbtable *table;
    772 	struct in_addr faddr, laddr;
    773 	u_int fport_arg, lport_arg;
    774 {
    775 	struct inpcbhead *head;
    776 	struct inpcb *inp;
    777 	u_int16_t fport = fport_arg, lport = lport_arg;
    778 
    779 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
    780 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    781 		if (in_hosteq(inp->inp_faddr, faddr) &&
    782 		    inp->inp_fport == fport &&
    783 		    inp->inp_lport == lport &&
    784 		    in_hosteq(inp->inp_laddr, laddr))
    785 			goto out;
    786 	}
    787 #ifdef DIAGNOSTIC
    788 	if (in_pcbnotifymiss) {
    789 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
    790 		    ntohl(faddr.s_addr), ntohs(fport),
    791 		    ntohl(laddr.s_addr), ntohs(lport));
    792 	}
    793 #endif
    794 	return (0);
    795 
    796 out:
    797 	/* Move this PCB to the head of hash chain. */
    798 	if (inp != head->lh_first) {
    799 		LIST_REMOVE(inp, inp_hash);
    800 		LIST_INSERT_HEAD(head, inp, inp_hash);
    801 	}
    802 	return (inp);
    803 }
    804 
    805 struct inpcb *
    806 in_pcblookup_bind(table, laddr, lport_arg)
    807 	struct inpcbtable *table;
    808 	struct in_addr laddr;
    809 	u_int lport_arg;
    810 {
    811 	struct inpcbhead *head;
    812 	struct inpcb *inp;
    813 	u_int16_t lport = lport_arg;
    814 
    815 	head = INPCBHASH_BIND(table, laddr, lport);
    816 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    817 		if (inp->inp_lport == lport &&
    818 		    in_hosteq(inp->inp_laddr, laddr))
    819 			goto out;
    820 	}
    821 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
    822 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
    823 		if (inp->inp_lport == lport &&
    824 		    in_hosteq(inp->inp_laddr, zeroin_addr))
    825 			goto out;
    826 	}
    827 #ifdef DIAGNOSTIC
    828 	if (in_pcbnotifymiss) {
    829 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
    830 		    ntohl(laddr.s_addr), ntohs(lport));
    831 	}
    832 #endif
    833 	return (0);
    834 
    835 out:
    836 	/* Move this PCB to the head of hash chain. */
    837 	if (inp != head->lh_first) {
    838 		LIST_REMOVE(inp, inp_hash);
    839 		LIST_INSERT_HEAD(head, inp, inp_hash);
    840 	}
    841 	return (inp);
    842 }
    843 
    844 void
    845 in_pcbstate(inp, state)
    846 	struct inpcb *inp;
    847 	int state;
    848 {
    849 
    850 	if (inp->inp_state > INP_ATTACHED)
    851 		LIST_REMOVE(inp, inp_hash);
    852 
    853 	switch (state) {
    854 	case INP_BOUND:
    855 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
    856 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
    857 		break;
    858 	case INP_CONNECTED:
    859 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
    860 		    inp->inp_faddr, inp->inp_fport,
    861 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
    862 		break;
    863 	}
    864 
    865 	inp->inp_state = state;
    866 }
    867 
    868 struct rtentry *
    869 in_pcbrtentry(inp)
    870 	struct inpcb *inp;
    871 {
    872 	struct route *ro;
    873 
    874 	ro = &inp->inp_route;
    875 
    876 	if (ro->ro_rt == NULL) {
    877 		/*
    878 		 * No route yet, so try to acquire one.
    879 		 */
    880 		if (!in_nullhost(inp->inp_faddr)) {
    881 			ro->ro_dst.sa_family = AF_INET;
    882 			ro->ro_dst.sa_len = sizeof(ro->ro_dst);
    883 			satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
    884 			rtalloc(ro);
    885 		}
    886 	}
    887 	return (ro->ro_rt);
    888 }
    889 
    890 struct sockaddr_in *
    891 in_selectsrc(sin, ro, soopts, mopts, errorp)
    892 	struct sockaddr_in *sin;
    893 	struct route *ro;
    894 	int soopts;
    895 	struct ip_moptions *mopts;
    896 	int *errorp;
    897 {
    898 	struct in_ifaddr *ia;
    899 
    900 	ia = (struct in_ifaddr *)0;
    901 	/*
    902 	 * If route is known or can be allocated now,
    903 	 * our src addr is taken from the i/f, else punt.
    904 	 */
    905 	if (ro->ro_rt &&
    906 	    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
    907 	    soopts & SO_DONTROUTE)) {
    908 		RTFREE(ro->ro_rt);
    909 		ro->ro_rt = (struct rtentry *)0;
    910 	}
    911 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
    912 	    (ro->ro_rt == (struct rtentry *)0 ||
    913 	    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
    914 		/* No route yet, so try to acquire one */
    915 		ro->ro_dst.sa_family = AF_INET;
    916 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
    917 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
    918 		rtalloc(ro);
    919 	}
    920 	/*
    921 	 * If we found a route, use the address
    922 	 * corresponding to the outgoing interface
    923 	 * unless it is the loopback (in case a route
    924 	 * to our address on another net goes to loopback).
    925 	 *
    926 	 * XXX Is this still true?  Do we care?
    927 	 */
    928 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
    929 		ia = ifatoia(ro->ro_rt->rt_ifa);
    930 	if (ia == NULL) {
    931 		u_int16_t fport = sin->sin_port;
    932 
    933 		sin->sin_port = 0;
    934 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
    935 		sin->sin_port = fport;
    936 		if (ia == 0) {
    937 			/* Find 1st non-loopback AF_INET address */
    938 			for (ia = in_ifaddr.tqh_first;
    939 			     ia != NULL;
    940 			     ia = ia->ia_list.tqe_next) {
    941 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
    942 					break;
    943 			}
    944 		}
    945 		if (ia == NULL) {
    946 			*errorp = EADDRNOTAVAIL;
    947 			return NULL;
    948 		}
    949 	}
    950 	/*
    951 	 * If the destination address is multicast and an outgoing
    952 	 * interface has been set as a multicast option, use the
    953 	 * address of that interface as our source address.
    954 	 */
    955 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
    956 		struct ip_moptions *imo;
    957 		struct ifnet *ifp;
    958 
    959 		imo = mopts;
    960 		if (imo->imo_multicast_ifp != NULL) {
    961 			ifp = imo->imo_multicast_ifp;
    962 			IFP_TO_IA(ifp, ia);		/* XXX */
    963 			if (ia == 0) {
    964 				*errorp = EADDRNOTAVAIL;
    965 				return NULL;
    966 			}
    967 		}
    968 	}
    969 	return satosin(&ia->ia_addr);
    970 }
    971