Home | History | Annotate | Line # | Download | only in netinet6
in6_pcb.c revision 1.35
      1 /*	$NetBSD: in6_pcb.c,v 1.35 2001/02/11 06:49:52 itojun Exp $	*/
      2 /*	$KAME: in6_pcb.c,v 1.84 2001/02/08 18:02:08 itojun Exp $	*/
      3 
      4 /*
      5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      6  * All rights reserved.
      7  *
      8  * Redistribution and use in source and binary forms, with or without
      9  * modification, are permitted provided that the following conditions
     10  * are met:
     11  * 1. Redistributions of source code must retain the above copyright
     12  *    notice, this list of conditions and the following disclaimer.
     13  * 2. Redistributions in binary form must reproduce the above copyright
     14  *    notice, this list of conditions and the following disclaimer in the
     15  *    documentation and/or other materials provided with the distribution.
     16  * 3. Neither the name of the project nor the names of its contributors
     17  *    may be used to endorse or promote products derived from this software
     18  *    without specific prior written permission.
     19  *
     20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     30  * SUCH DAMAGE.
     31  */
     32 
     33 /*
     34  * Copyright (c) 1982, 1986, 1991, 1993
     35  *	The Regents of the University of California.  All rights reserved.
     36  *
     37  * Redistribution and use in source and binary forms, with or without
     38  * modification, are permitted provided that the following conditions
     39  * are met:
     40  * 1. Redistributions of source code must retain the above copyright
     41  *    notice, this list of conditions and the following disclaimer.
     42  * 2. Redistributions in binary form must reproduce the above copyright
     43  *    notice, this list of conditions and the following disclaimer in the
     44  *    documentation and/or other materials provided with the distribution.
     45  * 3. All advertising materials mentioning features or use of this software
     46  *    must display the following acknowledgement:
     47  *	This product includes software developed by the University of
     48  *	California, Berkeley and its contributors.
     49  * 4. Neither the name of the University nor the names of its contributors
     50  *    may be used to endorse or promote products derived from this software
     51  *    without specific prior written permission.
     52  *
     53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     63  * SUCH DAMAGE.
     64  *
     65  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
     66  */
     67 
     68 #include "opt_ipsec.h"
     69 
     70 #include <sys/param.h>
     71 #include <sys/systm.h>
     72 #include <sys/malloc.h>
     73 #include <sys/mbuf.h>
     74 #include <sys/protosw.h>
     75 #include <sys/socket.h>
     76 #include <sys/socketvar.h>
     77 #include <sys/ioctl.h>
     78 #include <sys/errno.h>
     79 #include <sys/time.h>
     80 #include <sys/proc.h>
     81 
     82 #include <net/if.h>
     83 #include <net/route.h>
     84 
     85 #include <netinet/in.h>
     86 #include <netinet/in_var.h>
     87 #include <netinet/in_systm.h>
     88 #include <netinet/ip.h>
     89 #include <netinet/in_pcb.h>
     90 #include <netinet/ip6.h>
     91 #include <netinet6/ip6_var.h>
     92 #include <netinet6/in6_pcb.h>
     93 #include <netinet6/nd6.h>
     94 
     95 #include "loop.h"
     96 extern struct ifnet loif[NLOOP];
     97 #include "faith.h"
     98 
     99 #ifdef IPSEC
    100 #include <netinet6/ipsec.h>
    101 #include <netkey/key.h>
    102 #endif /* IPSEC */
    103 
    104 struct in6_addr zeroin6_addr;
    105 
    106 int ip6_anonportmin = IPV6PORT_ANONMIN;
    107 int ip6_anonportmax = IPV6PORT_ANONMAX;
    108 int ip6_lowportmin  = IPV6PORT_RESERVEDMIN;
    109 int ip6_lowportmax  = IPV6PORT_RESERVEDMAX;
    110 
    111 int
    112 in6_pcballoc(so, head)
    113 	struct socket *so;
    114 	struct in6pcb *head;
    115 {
    116 	struct in6pcb *in6p;
    117 
    118 	MALLOC(in6p, struct in6pcb *, sizeof(*in6p), M_PCB, M_NOWAIT);
    119 	if (in6p == NULL)
    120 		return(ENOBUFS);
    121 	bzero((caddr_t)in6p, sizeof(*in6p));
    122 	in6p->in6p_head = head;
    123 	in6p->in6p_socket = so;
    124 	in6p->in6p_hops = -1;	/* use kernel default */
    125 	in6p->in6p_icmp6filt = NULL;
    126 	in6p->in6p_next = head->in6p_next;
    127 	head->in6p_next = in6p;
    128 	in6p->in6p_prev = head;
    129 	in6p->in6p_next->in6p_prev = in6p;
    130 #ifndef INET6_BINDV6ONLY
    131 	if (ip6_bindv6only)
    132 		in6p->in6p_flags |= IN6P_BINDV6ONLY;
    133 #else
    134 	in6p->in6p_flags |= IN6P_BINDV6ONLY;	/*just for safety*/
    135 #endif
    136 	so->so_pcb = (caddr_t)in6p;
    137 	return(0);
    138 }
    139 
    140 int
    141 in6_pcbbind(in6p, nam, p)
    142 	struct in6pcb *in6p;
    143 	struct mbuf *nam;
    144 	struct proc *p;
    145 {
    146 	struct socket *so = in6p->in6p_socket;
    147 	struct in6pcb *head = in6p->in6p_head;
    148 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
    149 	u_int16_t lport = 0;
    150 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
    151 
    152 	if (in6p->in6p_lport || !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
    153 		return(EINVAL);
    154 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
    155 	   ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
    156 	    (so->so_options & SO_ACCEPTCONN) == 0))
    157 		wild = IN6PLOOKUP_WILDCARD;
    158 	if (nam) {
    159 		sin6 = mtod(nam, struct sockaddr_in6 *);
    160 		if (nam->m_len != sizeof(*sin6))
    161 			return(EINVAL);
    162 		/*
    163 		 * We should check the family, but old programs
    164 		 * incorrectly fail to intialize it.
    165 		 */
    166 		if (sin6->sin6_family != AF_INET6)
    167 			return(EAFNOSUPPORT);
    168 
    169 		/*
    170 		 * since we do not check port number duplicate with IPv4 space,
    171 		 * we reject it at this moment.  If we leave it, we make normal
    172 		 * user to hijack port number from other users.
    173 		 */
    174 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
    175 			return(EADDRNOTAVAIL);
    176 
    177 		/* KAME hack: embed scopeid */
    178 		if (in6_embedscope(&sin6->sin6_addr, sin6, in6p, NULL) != 0)
    179 			return EINVAL;
    180 		/* this must be cleared for ifa_ifwithaddr() */
    181 		sin6->sin6_scope_id = 0;
    182 
    183 		lport = sin6->sin6_port;
    184 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
    185 			/*
    186 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
    187 			 * allow compepte duplication of binding if
    188 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
    189 			 * and a multicast address is bound on both
    190 			 * new and duplicated sockets.
    191 			 */
    192 			if (so->so_options & SO_REUSEADDR)
    193 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
    194 		} else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
    195 			struct sockaddr_in sin;
    196 
    197 			bzero(&sin, sizeof(sin));
    198 			sin.sin_len = sizeof(sin);
    199 			sin.sin_family = AF_INET;
    200 			bcopy(&sin6->sin6_addr.s6_addr32[3], &sin.sin_addr,
    201 				sizeof(sin.sin_addr));
    202 			if (ifa_ifwithaddr((struct sockaddr *)&sin) == 0)
    203 				return EADDRNOTAVAIL;
    204 		} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
    205 			struct ifaddr *ia = NULL;
    206 
    207 			sin6->sin6_port = 0;		/* yech... */
    208 #if defined(NFAITH) && NFAITH > 0
    209 			if ((in6p->in6p_flags & IN6P_FAITH) == 0
    210 			 && (ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0)
    211 #else
    212 			if ((ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0)
    213 #endif
    214 				return(EADDRNOTAVAIL);
    215 
    216 			/*
    217 			 * XXX: bind to an anycast address might accidentally
    218 			 * cause sending a packet with anycast source address.
    219 			 */
    220 			if (ia &&
    221 			    ((struct in6_ifaddr *)ia)->ia6_flags &
    222 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|
    223 			     IN6_IFF_DETACHED|IN6_IFF_DEPRECATED)) {
    224 				return(EADDRNOTAVAIL);
    225 			}
    226 		}
    227 		if (lport) {
    228 #ifndef IPNOPRIVPORTS
    229 			int priv;
    230 
    231 			/*
    232 			 * NOTE: all operating systems use suser() for
    233 			 * privilege check!  do not rewrite it into SS_PRIV.
    234 			 */
    235 			priv = (p && !suser(p->p_ucred, &p->p_acflag)) ? 1 : 0;
    236 			/* GROSS */
    237 			if (ntohs(lport) < IPV6PORT_RESERVED && !priv)
    238 				return(EACCES);
    239 #endif
    240 
    241 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
    242 				/* should check this but we can't ... */
    243 #if 0
    244 				struct inpcb *t;
    245 
    246 				t = in_pcblookup_bind(&tcbtable,
    247 					(struct in_addr *)&sin6->sin6_addr.s6_addr32[3],
    248 					lport);
    249 				if (t && (reuseport & t->inp_socket->so_options) == 0)
    250 					return EADDRINUSE;
    251 #endif
    252 			} else {
    253 				struct in6pcb *t;
    254 
    255 				t = in6_pcblookup(head, &zeroin6_addr, 0,
    256 						  &sin6->sin6_addr, lport, wild);
    257 				if (t && (reuseport & t->in6p_socket->so_options) == 0)
    258 					return(EADDRINUSE);
    259 			}
    260 		}
    261 		in6p->in6p_laddr = sin6->sin6_addr;
    262 	}
    263 
    264 	if (lport == 0) {
    265 		int e;
    266 		if ((e = in6_pcbsetport(&in6p->in6p_laddr, in6p)) != 0)
    267 			return(e);
    268 	}
    269 	else
    270 		in6p->in6p_lport = lport;
    271 
    272 	in6p->in6p_flowinfo = sin6 ? sin6->sin6_flowinfo : 0;	/*XXX*/
    273 	return(0);
    274 }
    275 
    276 /*
    277  * Connect from a socket to a specified address.
    278  * Both address and port must be specified in argument sin6.
    279  * If don't have a local address for this socket yet,
    280  * then pick one.
    281  */
    282 int
    283 in6_pcbconnect(in6p, nam)
    284 	struct in6pcb *in6p;
    285 	struct mbuf *nam;
    286 {
    287 	struct in6_addr *in6a = NULL;
    288 	struct sockaddr_in6 *sin6 = mtod(nam, struct sockaddr_in6 *);
    289 	struct ifnet *ifp = NULL;	/* outgoing interface */
    290 	int error = 0;
    291 #ifdef INET
    292 	struct in6_addr mapped;
    293 #endif
    294 	struct sockaddr_in6 tmp;
    295 
    296 	(void)&in6a;				/* XXX fool gcc */
    297 
    298 	if (nam->m_len != sizeof(*sin6))
    299 		return(EINVAL);
    300 	if (sin6->sin6_family != AF_INET6)
    301 		return(EAFNOSUPPORT);
    302 	if (sin6->sin6_port == 0)
    303 		return(EADDRNOTAVAIL);
    304 
    305 	/* sanity check for mapped address case */
    306 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
    307 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
    308 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
    309 		if (!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
    310 			return EINVAL;
    311 	} else {
    312 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
    313 			return EINVAL;
    314 	}
    315 
    316 	/* protect *sin6 from overwrites */
    317 	tmp = *sin6;
    318 	sin6 = &tmp;
    319 
    320 	/* KAME hack: embed scopeid */
    321 	if (in6_embedscope(&sin6->sin6_addr, sin6, in6p, &ifp) != 0)
    322 		return EINVAL;
    323 
    324 	/* Source address selection. */
    325 	if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
    326 	 && in6p->in6p_laddr.s6_addr32[3] == 0) {
    327 #ifdef INET
    328 		struct sockaddr_in sin, *sinp;
    329 
    330 		bzero(&sin, sizeof(sin));
    331 		sin.sin_len = sizeof(sin);
    332 		sin.sin_family = AF_INET;
    333 		bcopy(&sin6->sin6_addr.s6_addr32[3], &sin.sin_addr,
    334 			sizeof(sin.sin_addr));
    335 		sinp = in_selectsrc(&sin, (struct route *)&in6p->in6p_route,
    336 			in6p->in6p_socket->so_options, NULL, &error);
    337 		if (sinp == 0) {
    338 			if (error == 0)
    339 				error = EADDRNOTAVAIL;
    340 			return(error);
    341 		}
    342 		bzero(&mapped, sizeof(mapped));
    343 		mapped.s6_addr16[5] = htons(0xffff);
    344 		bcopy(&sinp->sin_addr, &mapped.s6_addr32[3], sizeof(sinp->sin_addr));
    345 		in6a = &mapped;
    346 #else
    347 		return EADDRNOTAVAIL;
    348 #endif
    349 	} else {
    350 		/*
    351 		 * XXX: in6_selectsrc might replace the bound local address
    352 		 * with the address specified by setsockopt(IPV6_PKTINFO).
    353 		 * Is it the intended behavior?
    354 		 */
    355 		in6a = in6_selectsrc(sin6, in6p->in6p_outputopts,
    356 				     in6p->in6p_moptions,
    357 				     &in6p->in6p_route,
    358 				     &in6p->in6p_laddr, &error);
    359 		if (in6a == 0) {
    360 			if (error == 0)
    361 				error = EADDRNOTAVAIL;
    362 			return(error);
    363 		}
    364 	}
    365 	if (in6p->in6p_route.ro_rt)
    366 		ifp = in6p->in6p_route.ro_rt->rt_ifp;
    367 
    368 	in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim(in6p, ifp);
    369 
    370 	if (in6_pcblookup(in6p->in6p_head,
    371 			 &sin6->sin6_addr,
    372 			 sin6->sin6_port,
    373 			 IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ?
    374 			  in6a : &in6p->in6p_laddr,
    375 			 in6p->in6p_lport,
    376 			 0))
    377 		return(EADDRINUSE);
    378 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)
    379 	 || (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
    380 	  && in6p->in6p_laddr.s6_addr32[3] == 0)) {
    381 		if (in6p->in6p_lport == 0) {
    382 			(void)in6_pcbbind(in6p, (struct mbuf *)0,
    383 			    (struct proc *)0);
    384 		}
    385 		in6p->in6p_laddr = *in6a;
    386 	}
    387 	in6p->in6p_faddr = sin6->sin6_addr;
    388 	in6p->in6p_fport = sin6->sin6_port;
    389 	/*
    390 	 * xxx kazu flowlabel is necessary for connect?
    391 	 * but if this line is missing, the garbage value remains.
    392 	 */
    393 	in6p->in6p_flowinfo = sin6->sin6_flowinfo;
    394 	return(0);
    395 }
    396 
    397 void
    398 in6_pcbdisconnect(in6p)
    399 	struct in6pcb *in6p;
    400 {
    401 	bzero((caddr_t)&in6p->in6p_faddr, sizeof(in6p->in6p_faddr));
    402 	in6p->in6p_fport = 0;
    403 	if (in6p->in6p_socket->so_state & SS_NOFDREF)
    404 		in6_pcbdetach(in6p);
    405 }
    406 
    407 void
    408 in6_pcbdetach(in6p)
    409 	struct in6pcb *in6p;
    410 {
    411 	struct socket *so = in6p->in6p_socket;
    412 
    413 #ifdef IPSEC
    414 	ipsec6_delete_pcbpolicy(in6p);
    415 #endif /* IPSEC */
    416 	sotoin6pcb(so) = 0;
    417 	sofree(so);
    418 	if (in6p->in6p_options)
    419 		m_freem(in6p->in6p_options);
    420 	if (in6p->in6p_outputopts) {
    421 		if (in6p->in6p_outputopts->ip6po_rthdr &&
    422 		    in6p->in6p_outputopts->ip6po_route.ro_rt)
    423 			RTFREE(in6p->in6p_outputopts->ip6po_route.ro_rt);
    424 		if (in6p->in6p_outputopts->ip6po_m)
    425 			(void)m_free(in6p->in6p_outputopts->ip6po_m);
    426 		free(in6p->in6p_outputopts, M_IP6OPT);
    427 	}
    428 	if (in6p->in6p_route.ro_rt)
    429 		rtfree(in6p->in6p_route.ro_rt);
    430 	ip6_freemoptions(in6p->in6p_moptions);
    431 	in6p->in6p_next->in6p_prev = in6p->in6p_prev;
    432 	in6p->in6p_prev->in6p_next = in6p->in6p_next;
    433 	in6p->in6p_prev = NULL;
    434 	FREE(in6p, M_PCB);
    435 }
    436 
    437 void
    438 in6_setsockaddr(in6p, nam)
    439 	struct in6pcb *in6p;
    440 	struct mbuf *nam;
    441 {
    442 	struct sockaddr_in6 *sin6;
    443 
    444 	nam->m_len = sizeof(*sin6);
    445 	sin6 = mtod(nam, struct sockaddr_in6 *);
    446 	bzero((caddr_t)sin6, sizeof(*sin6));
    447 	sin6->sin6_family = AF_INET6;
    448 	sin6->sin6_len = sizeof(struct sockaddr_in6);
    449 	sin6->sin6_port = in6p->in6p_lport;
    450 	/* KAME hack: recover scopeid */
    451 	(void)in6_recoverscope(sin6, &in6p->in6p_laddr, NULL);
    452 }
    453 
    454 void
    455 in6_setpeeraddr(in6p, nam)
    456 	struct in6pcb *in6p;
    457 	struct mbuf *nam;
    458 {
    459 	struct sockaddr_in6 *sin6;
    460 
    461 	nam->m_len = sizeof(*sin6);
    462 	sin6 = mtod(nam, struct sockaddr_in6 *);
    463 	bzero((caddr_t)sin6, sizeof(*sin6));
    464 	sin6->sin6_family = AF_INET6;
    465 	sin6->sin6_len = sizeof(struct sockaddr_in6);
    466 	sin6->sin6_port = in6p->in6p_fport;
    467 	/* KAME hack: recover scopeid */
    468 	(void)in6_recoverscope(sin6, &in6p->in6p_faddr, NULL);
    469 }
    470 
    471 /*
    472  * Pass some notification to all connections of a protocol
    473  * associated with address dst.  The local address and/or port numbers
    474  * may be specified to limit the search.  The "usual action" will be
    475  * taken, depending on the ctlinput cmd.  The caller must filter any
    476  * cmds that are uninteresting (e.g., no error in the map).
    477  * Call the protocol specific routine (if any) to report
    478  * any errors for each matching socket.
    479  *
    480  * Must be called at splsoftnet.
    481  *
    482  * Note: src (4th arg) carries the flowlabel value on the original IPv6
    483  * header, in sin6_flowinfo member.
    484  */
    485 int
    486 in6_pcbnotify(head, dst, fport_arg, src, lport_arg, cmd, cmdarg, notify)
    487 	struct in6pcb *head;
    488 	struct sockaddr *dst, *src;
    489 	u_int fport_arg, lport_arg;
    490 	int cmd;
    491 	void *cmdarg;
    492 	void (*notify) __P((struct in6pcb *, int));
    493 {
    494 	struct in6pcb *in6p, *nin6p;
    495 	struct sockaddr_in6 sa6_src, *sa6_dst;
    496 	u_int16_t fport = fport_arg, lport = lport_arg;
    497 	int errno;
    498 	int nmatch = 0;
    499 	u_int32_t flowinfo;
    500 
    501 	if ((unsigned)cmd > PRC_NCMDS || dst->sa_family != AF_INET6)
    502 		return 0;
    503 
    504 	sa6_dst = (struct sockaddr_in6 *)dst;
    505 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
    506 		return 0;
    507 
    508 	/*
    509 	 * note that src can be NULL when we get notify by local fragmentation.
    510 	 */
    511 	sa6_src = (src == NULL) ? sa6_any : *(struct sockaddr_in6 *)src;
    512 	flowinfo = sa6_src.sin6_flowinfo;
    513 
    514 	/*
    515 	 * Redirects go to all references to the destination,
    516 	 * and use in6_rtchange to invalidate the route cache.
    517 	 * Dead host indications: also use in6_rtchange to invalidate
    518 	 * the cache, and deliver the error to all the sockets.
    519 	 * Otherwise, if we have knowledge of the local port and address,
    520 	 * deliver only to that socket.
    521 	 */
    522 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
    523 		fport = 0;
    524 		lport = 0;
    525 		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
    526 
    527 		if (cmd != PRC_HOSTDEAD)
    528 			notify = in6_rtchange;
    529 	}
    530 
    531 	errno = inet6ctlerrmap[cmd];
    532 	for (in6p = head->in6p_next; in6p != head; in6p = nin6p) {
    533 		nin6p = in6p->in6p_next;
    534 
    535 		/*
    536 		 * Under the following condition, notify of redirects
    537 		 * to the pcb, without making address matches against inpcb.
    538 		 * - redirect notification is arrived.
    539 		 * - the inpcb is unconnected.
    540 		 * - the inpcb is caching !RTF_HOST routing entry.
    541 		 * - the ICMPv6 notification is from the gateway cached in the
    542 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
    543 		 *   the inpcb used very recently.
    544 		 *
    545 		 * This is to improve interaction between netbsd/openbsd
    546 		 * redirect handling code, and inpcb route cache code.
    547 		 * without the clause, !RTF_HOST routing entry (which carries
    548 		 * gateway used by inpcb right before the ICMPv6 redirect)
    549 		 * will be cached forever in unconnected inpcb.
    550 		 *
    551 		 * There still is a question regarding to what is TRT:
    552 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
    553 		 *   generated on packet output.  inpcb will always cache
    554 		 *   RTF_HOST routing entry so there's no need for the clause
    555 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
    556 		 *   and inpcb is caching it already).
    557 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
    558 		 *   due to the cloned routing entries.
    559 		 * - Specwise, "destination cache" is mentioned in RFC2461.
    560 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
    561 		 *   is not really convinced.
    562 		 * - Having hiwat/lowat on # of cloned host route (redirect/
    563 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
    564 		 *   icmp6_mtudisc_update().
    565 		 */
    566 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
    567 		    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
    568 		    in6p->in6p_route.ro_rt &&
    569 		    !(in6p->in6p_route.ro_rt->rt_flags & RTF_HOST)) {
    570 			struct sockaddr_in6 *dst6;
    571 
    572 			dst6 = (struct sockaddr_in6 *)&in6p->in6p_route.ro_dst;
    573 			if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
    574 			    &sa6_dst->sin6_addr))
    575 				goto do_notify;
    576 		}
    577 
    578 		/*
    579 		 * Detect if we should notify the error. If no source and
    580 		 * destination ports are specifed, but non-zero flowinfo and
    581 		 * local address match, notify the error. This is the case
    582 		 * when the error is delivered with an encrypted buffer
    583 		 * by ESP. Otherwise, just compare addresses and ports
    584 		 * as usual.
    585 		 */
    586 		if (lport == 0 && fport == 0 && flowinfo &&
    587 		    in6p->in6p_socket != NULL &&
    588 		    flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
    589 		    IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
    590 			goto do_notify;
    591 		else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
    592 					     &sa6_dst->sin6_addr) ||
    593 		    in6p->in6p_socket == 0 ||
    594 		    (lport && in6p->in6p_lport != lport) ||
    595 		    (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
    596 		     !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
    597 					 &sa6_src.sin6_addr)) ||
    598 		    (fport && in6p->in6p_fport != fport))
    599 			continue;
    600 
    601 	  do_notify:
    602 		if (notify)
    603 			(*notify)(in6p, errno);
    604 		nmatch++;
    605 	}
    606 	return nmatch;
    607 }
    608 
    609 void
    610 in6_pcbpurgeif(head, ifp)
    611 	struct in6pcb *head;
    612 	struct ifnet *ifp;
    613 {
    614 	struct in6pcb *in6p, *nin6p;
    615 
    616 	for (in6p = head->in6p_next; in6p != head; in6p = nin6p) {
    617 		nin6p = in6p->in6p_next;
    618 		if (in6p->in6p_route.ro_rt != NULL &&
    619 		    in6p->in6p_route.ro_rt->rt_ifp == ifp)
    620 			in6_rtchange(in6p, 0);
    621 	}
    622 }
    623 
    624 /*
    625  * Check for alternatives when higher level complains
    626  * about service problems.  For now, invalidate cached
    627  * routing information.  If the route was created dynamically
    628  * (by a redirect), time to try a default gateway again.
    629  */
    630 void
    631 in6_losing(in6p)
    632 	struct in6pcb *in6p;
    633 {
    634 	struct rtentry *rt;
    635 	struct rt_addrinfo info;
    636 
    637 	if ((rt = in6p->in6p_route.ro_rt) != NULL) {
    638 		in6p->in6p_route.ro_rt = 0;
    639 		bzero((caddr_t)&info, sizeof(info));
    640 		info.rti_info[RTAX_DST] =
    641 			(struct sockaddr *)&in6p->in6p_route.ro_dst;
    642 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    643 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    644 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
    645 		if (rt->rt_flags & RTF_DYNAMIC)
    646 			(void)rtrequest(RTM_DELETE, rt_key(rt),
    647 					rt->rt_gateway, rt_mask(rt), rt->rt_flags,
    648 					(struct rtentry **)0);
    649 		else
    650 		/*
    651 		 * A new route can be allocated
    652 		 * the next time output is attempted.
    653 		 */
    654 			rtfree(rt);
    655 	}
    656 }
    657 
    658 /*
    659  * After a routing change, flush old routing
    660  * and allocate a (hopefully) better one.
    661  */
    662 void
    663 in6_rtchange(in6p, errno)
    664 	struct in6pcb *in6p;
    665 	int errno;
    666 {
    667 	if (in6p->in6p_route.ro_rt) {
    668 		rtfree(in6p->in6p_route.ro_rt);
    669 		in6p->in6p_route.ro_rt = 0;
    670 		/*
    671 		 * A new route can be allocated the next time
    672 		 * output is attempted.
    673 		 */
    674 	}
    675 }
    676 
    677 struct in6pcb *
    678 in6_pcblookup(head, faddr6, fport_arg, laddr6, lport_arg, flags)
    679 	struct in6pcb *head;
    680 	struct in6_addr *faddr6, *laddr6;
    681 	u_int fport_arg, lport_arg;
    682 	int flags;
    683 {
    684 	struct in6pcb *in6p, *match = 0;
    685 	int matchwild = 3, wildcard;
    686 	u_int16_t fport = fport_arg, lport = lport_arg;
    687 
    688 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
    689 		if (in6p->in6p_lport != lport)
    690 			continue;
    691 		wildcard = 0;
    692 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
    693 			if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
    694 				wildcard++;
    695 			else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
    696 				continue;
    697 		}
    698 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
    699 			&& in6p->in6p_laddr.s6_addr32[3] == 0) {
    700 			if (!IN6_IS_ADDR_V4MAPPED(laddr6))
    701 				continue;
    702 			if (laddr6->s6_addr32[3] == 0)
    703 				;
    704 			else
    705 				wildcard++;
    706 		}
    707 		else {
    708 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
    709 #ifndef INET6_BINDV6ONLY
    710 				if (in6p->in6p_flags & IN6P_BINDV6ONLY)
    711 					continue;
    712 				else
    713 					wildcard++;
    714 #else
    715 				continue;
    716 #endif
    717 			} else if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
    718 				wildcard++;
    719 		}
    720 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
    721 			if (IN6_IS_ADDR_UNSPECIFIED(faddr6))
    722 				wildcard++;
    723 			else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6)
    724 			      || in6p->in6p_fport != fport)
    725 				continue;
    726 		}
    727 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)
    728 			&& in6p->in6p_faddr.s6_addr32[3] == 0) {
    729 			if (!IN6_IS_ADDR_V4MAPPED(faddr6))
    730 				continue;
    731 			if (faddr6->s6_addr32[3] == 0)
    732 				;
    733 			else
    734 				wildcard++;
    735 		}
    736 		else {
    737 			if (IN6_IS_ADDR_V4MAPPED(faddr6)) {
    738 #ifndef INET6_BINDV6ONLY
    739 				if (in6p->in6p_flags & IN6P_BINDV6ONLY)
    740 					continue;
    741 				else
    742 					wildcard++;
    743 #else
    744 				continue;
    745 #endif
    746 			} else if (!IN6_IS_ADDR_UNSPECIFIED(faddr6))
    747 				wildcard++;
    748 		}
    749 
    750 		if (wildcard && (flags & IN6PLOOKUP_WILDCARD) == 0)
    751 			continue;
    752 		if (wildcard < matchwild) {
    753 			match = in6p;
    754 			matchwild = wildcard;
    755 			if (matchwild == 0)
    756 				break;
    757 		}
    758 	}
    759 	return(match);
    760 }
    761 
    762 struct rtentry *
    763 in6_pcbrtentry(in6p)
    764 	struct in6pcb *in6p;
    765 {
    766 	struct route_in6 *ro;
    767 	struct sockaddr_in6 *dst6;
    768 
    769 	ro = &in6p->in6p_route;
    770 	dst6 = (struct sockaddr_in6 *)&ro->ro_dst;
    771 
    772 	if (ro->ro_rt == NULL) {
    773 		/*
    774 		 * No route yet, so try to acquire one.
    775 		 */
    776 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
    777 			bzero(dst6, sizeof(*dst6));
    778 			dst6->sin6_family = AF_INET6;
    779 			dst6->sin6_len = sizeof(struct sockaddr_in6);
    780 			dst6->sin6_addr = in6p->in6p_faddr;
    781 			rtalloc((struct route *)ro);
    782 		}
    783 	}
    784 	return (ro->ro_rt);
    785 }
    786 
    787 struct in6pcb *
    788 in6_pcblookup_connect(head, faddr6, fport_arg, laddr6, lport_arg, faith)
    789 	struct in6pcb *head;
    790 	struct in6_addr *faddr6, *laddr6;
    791 	u_int fport_arg, lport_arg;
    792 	int faith;
    793 {
    794 	struct in6pcb *in6p;
    795 	u_int16_t fport = fport_arg, lport = lport_arg;
    796 
    797 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
    798 		/* find exact match on both source and dest */
    799 		if (in6p->in6p_fport != fport)
    800 			continue;
    801 		if (in6p->in6p_lport != lport)
    802 			continue;
    803 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
    804 			continue;
    805 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6))
    806 			continue;
    807 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
    808 			continue;
    809 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
    810 			continue;
    811 		return in6p;
    812 	}
    813 	return NULL;
    814 }
    815 
    816 struct in6pcb *
    817 in6_pcblookup_bind(head, laddr6, lport_arg, faith)
    818 	struct in6pcb *head;
    819 	struct in6_addr *laddr6;
    820 	u_int lport_arg;
    821 	int faith;
    822 {
    823 	struct in6pcb *in6p, *match;
    824 	u_int16_t lport = lport_arg;
    825 
    826 	match = NULL;
    827 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
    828 		/*
    829 	 	 * find destination match.  exact match is preferred
    830 		 * against wildcard match.
    831 		 */
    832 #if defined(NFAITH) && NFAITH > 0
    833 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
    834 			continue;
    835 #endif
    836 		if (in6p->in6p_fport != 0)
    837 			continue;
    838 		if (in6p->in6p_lport != lport)
    839 			continue;
    840 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
    841 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
    842 #ifndef INET6_BINDV6ONLY
    843 				if (in6p->in6p_flags & IN6P_BINDV6ONLY)
    844 					continue;
    845 				else
    846 					match = in6p;
    847 #else
    848 				continue;
    849 #endif
    850 			} else
    851 				match = in6p;
    852 		}
    853 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
    854 			 in6p->in6p_laddr.s6_addr32[3] == 0) {
    855 			if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
    856 			    laddr6->s6_addr32[3] != 0)
    857 				match = in6p;
    858 		}
    859 		else if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
    860 			return in6p;
    861 	}
    862 	return match;
    863 }
    864