Home | History | Annotate | Line # | Download | only in netinet6
in6_pcb.c revision 1.51
      1 /*	$NetBSD: in6_pcb.c,v 1.51 2002/08/26 14:25:00 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 <sys/cdefs.h>
     69 __KERNEL_RCSID(0, "$NetBSD: in6_pcb.c,v 1.51 2002/08/26 14:25:00 itojun Exp $");
     70 
     71 #include "opt_ipsec.h"
     72 
     73 #include <sys/param.h>
     74 #include <sys/systm.h>
     75 #include <sys/malloc.h>
     76 #include <sys/mbuf.h>
     77 #include <sys/protosw.h>
     78 #include <sys/socket.h>
     79 #include <sys/socketvar.h>
     80 #include <sys/ioctl.h>
     81 #include <sys/errno.h>
     82 #include <sys/time.h>
     83 #include <sys/proc.h>
     84 
     85 #include <net/if.h>
     86 #include <net/route.h>
     87 
     88 #include <netinet/in.h>
     89 #include <netinet/in_var.h>
     90 #include <netinet/in_systm.h>
     91 #include <netinet/ip.h>
     92 #include <netinet/in_pcb.h>
     93 #include <netinet/ip6.h>
     94 #include <netinet6/ip6_var.h>
     95 #include <netinet6/in6_pcb.h>
     96 #include <netinet6/nd6.h>
     97 
     98 #include "loop.h"
     99 extern struct ifnet loif[NLOOP];
    100 #include "faith.h"
    101 
    102 #ifdef IPSEC
    103 #include <netinet6/ipsec.h>
    104 #include <netkey/key.h>
    105 #endif /* IPSEC */
    106 
    107 struct in6_addr zeroin6_addr;
    108 
    109 int ip6_anonportmin = IPV6PORT_ANONMIN;
    110 int ip6_anonportmax = IPV6PORT_ANONMAX;
    111 int ip6_lowportmin  = IPV6PORT_RESERVEDMIN;
    112 int ip6_lowportmax  = IPV6PORT_RESERVEDMAX;
    113 
    114 int
    115 in6_pcballoc(so, head)
    116 	struct socket *so;
    117 	struct in6pcb *head;
    118 {
    119 	struct in6pcb *in6p;
    120 	int s;
    121 #ifdef IPSEC
    122 	int error;
    123 #endif
    124 
    125 	MALLOC(in6p, struct in6pcb *, sizeof(*in6p), M_PCB, M_NOWAIT);
    126 	if (in6p == NULL)
    127 		return(ENOBUFS);
    128 	bzero((caddr_t)in6p, sizeof(*in6p));
    129 	in6p->in6p_head = head;
    130 	in6p->in6p_socket = so;
    131 	in6p->in6p_hops = -1;	/* use kernel default */
    132 	in6p->in6p_icmp6filt = NULL;
    133 #ifdef IPSEC
    134 	error = ipsec_init_pcbpolicy(so, &in6p->in6p_sp);
    135 	if (error != 0) {
    136 		FREE(in6p, M_PCB);
    137 		return error;
    138 	}
    139 #endif /* IPSEC */
    140 	s = splnet();
    141 	in6p->in6p_next = head->in6p_next;
    142 	head->in6p_next = in6p;
    143 	in6p->in6p_prev = head;
    144 	in6p->in6p_next->in6p_prev = in6p;
    145 	splx(s);
    146 	if (ip6_v6only)
    147 		in6p->in6p_flags |= IN6P_IPV6_V6ONLY;
    148 	so->so_pcb = (caddr_t)in6p;
    149 	return(0);
    150 }
    151 
    152 int
    153 in6_pcbbind(in6p, nam, p)
    154 	struct in6pcb *in6p;
    155 	struct mbuf *nam;
    156 	struct proc *p;
    157 {
    158 	struct socket *so = in6p->in6p_socket;
    159 	struct in6pcb *head = in6p->in6p_head;
    160 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
    161 	u_int16_t lport = 0;
    162 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
    163 
    164 	if (in6p->in6p_lport || !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
    165 		return(EINVAL);
    166 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
    167 	   ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
    168 	    (so->so_options & SO_ACCEPTCONN) == 0))
    169 		wild = IN6PLOOKUP_WILDCARD;
    170 	if (nam) {
    171 		sin6 = mtod(nam, struct sockaddr_in6 *);
    172 		if (nam->m_len != sizeof(*sin6))
    173 			return(EINVAL);
    174 		/*
    175 		 * We should check the family, but old programs
    176 		 * incorrectly fail to intialize it.
    177 		 */
    178 		if (sin6->sin6_family != AF_INET6)
    179 			return(EAFNOSUPPORT);
    180 
    181 		/*
    182 		 * since we do not check port number duplicate with IPv4 space,
    183 		 * we reject it at this moment.  If we leave it, we would
    184 		 * mistakenly allow normal users to hijack tcp/udp ports from
    185 		 * other users.
    186 		 */
    187 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
    188 			return(EADDRNOTAVAIL);
    189 
    190 		/* KAME hack: embed scopeid */
    191 		if (in6_embedscope(&sin6->sin6_addr, sin6, in6p, NULL) != 0)
    192 			return EINVAL;
    193 		/* this must be cleared for ifa_ifwithaddr() */
    194 		sin6->sin6_scope_id = 0;
    195 
    196 		lport = sin6->sin6_port;
    197 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
    198 			/*
    199 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
    200 			 * allow compepte duplication of binding if
    201 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
    202 			 * and a multicast address is bound on both
    203 			 * new and duplicated sockets.
    204 			 */
    205 			if (so->so_options & SO_REUSEADDR)
    206 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
    207 		}
    208 		else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
    209 			struct sockaddr_in sin;
    210 
    211 			bzero(&sin, sizeof(sin));
    212 			sin.sin_len = sizeof(sin);
    213 			sin.sin_family = AF_INET;
    214 			bcopy(&sin6->sin6_addr.s6_addr32[3], &sin.sin_addr,
    215 				sizeof(sin.sin_addr));
    216 			if (ifa_ifwithaddr((struct sockaddr *)&sin) == 0)
    217 				return EADDRNOTAVAIL;
    218 		}
    219 		else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
    220 			struct ifaddr *ia = NULL;
    221 
    222 			sin6->sin6_port = 0;		/* yech... */
    223 			if ((in6p->in6p_flags & IN6P_FAITH) == 0 &&
    224 			    (ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0)
    225 				return(EADDRNOTAVAIL);
    226 
    227 			/*
    228 			 * bind to an anycast address might accidentally
    229 			 * cause sending a packet with an anycast source
    230 			 * address, so we forbid it.
    231 			 *
    232 			 * We should allow to bind to a deprecated address,
    233 			 * since the application dare to use it.
    234 			 * But, can we assume that they are careful enough
    235 			 * to check if the address is deprecated or not?
    236 			 * Maybe, as a safeguard, we should have a setsockopt
    237 			 * flag to control the bind(2) behavior against
    238 			 * deprecated addresses (default: forbid bind(2)).
    239 			 */
    240 			if (ia &&
    241 			    ((struct in6_ifaddr *)ia)->ia6_flags &
    242 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
    243 				return(EADDRNOTAVAIL);
    244 		}
    245 		if (lport) {
    246 #ifndef IPNOPRIVPORTS
    247 			int priv;
    248 
    249 			/*
    250 			 * NOTE: all operating systems use suser() for
    251 			 * privilege check!  do not rewrite it into SS_PRIV.
    252 			 */
    253 			priv = (p && !suser(p->p_ucred, &p->p_acflag)) ? 1 : 0;
    254 			/* GROSS */
    255 			if (ntohs(lport) < IPV6PORT_RESERVED && !priv)
    256 				return(EACCES);
    257 #endif
    258 
    259 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
    260 				/* should check this but we can't ... */
    261 #if 0
    262 				struct inpcb *t;
    263 
    264 				t = in_pcblookup_bind(&tcbtable,
    265 					(struct in_addr *)&sin6->sin6_addr.s6_addr32[3],
    266 					lport);
    267 				if (t && (reuseport & t->inp_socket->so_options) == 0)
    268 					return EADDRINUSE;
    269 #endif
    270 			} else
    271 			{
    272 				struct in6pcb *t;
    273 
    274 				t = in6_pcblookup(head, &zeroin6_addr, 0,
    275 						  &sin6->sin6_addr, lport, wild);
    276 				if (t && (reuseport & t->in6p_socket->so_options) == 0)
    277 					return(EADDRINUSE);
    278 			}
    279 		}
    280 		in6p->in6p_laddr = sin6->sin6_addr;
    281 	}
    282 
    283 	if (lport == 0) {
    284 		int e;
    285 		if ((e = in6_pcbsetport(&in6p->in6p_laddr, in6p, p)) != 0)
    286 			return(e);
    287 	}
    288 	else
    289 		in6p->in6p_lport = lport;
    290 
    291 	in6p->in6p_flowinfo = sin6 ? sin6->sin6_flowinfo : 0;	/*XXX*/
    292 	return(0);
    293 }
    294 
    295 /*
    296  * Connect from a socket to a specified address.
    297  * Both address and port must be specified in argument sin6.
    298  * If don't have a local address for this socket yet,
    299  * then pick one.
    300  */
    301 int
    302 in6_pcbconnect(in6p, nam)
    303 	struct in6pcb *in6p;
    304 	struct mbuf *nam;
    305 {
    306 	struct in6_addr *in6a = NULL;
    307 	struct sockaddr_in6 *sin6 = mtod(nam, struct sockaddr_in6 *);
    308 	struct ifnet *ifp = NULL;	/* outgoing interface */
    309 	int error = 0;
    310 #ifdef INET
    311 	struct in6_addr mapped;
    312 #endif
    313 	struct sockaddr_in6 tmp;
    314 
    315 	(void)&in6a;				/* XXX fool gcc */
    316 
    317 	if (nam->m_len != sizeof(*sin6))
    318 		return(EINVAL);
    319 	if (sin6->sin6_family != AF_INET6)
    320 		return(EAFNOSUPPORT);
    321 	if (sin6->sin6_port == 0)
    322 		return(EADDRNOTAVAIL);
    323 
    324 	/* sanity check for mapped address case */
    325 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
    326 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
    327 			return EINVAL;
    328 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
    329 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
    330 		if (!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
    331 			return EINVAL;
    332 	} else
    333 	{
    334 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
    335 			return EINVAL;
    336 	}
    337 
    338 	/* protect *sin6 from overwrites */
    339 	tmp = *sin6;
    340 	sin6 = &tmp;
    341 
    342 	/* KAME hack: embed scopeid */
    343 	if (in6_embedscope(&sin6->sin6_addr, sin6, in6p, &ifp) != 0)
    344 		return EINVAL;
    345 
    346 	/* Source address selection. */
    347 	if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
    348 	 && in6p->in6p_laddr.s6_addr32[3] == 0) {
    349 #ifdef INET
    350 		struct sockaddr_in sin, *sinp;
    351 
    352 		bzero(&sin, sizeof(sin));
    353 		sin.sin_len = sizeof(sin);
    354 		sin.sin_family = AF_INET;
    355 		bcopy(&sin6->sin6_addr.s6_addr32[3], &sin.sin_addr,
    356 			sizeof(sin.sin_addr));
    357 		sinp = in_selectsrc(&sin, (struct route *)&in6p->in6p_route,
    358 			in6p->in6p_socket->so_options, NULL, &error);
    359 		if (sinp == 0) {
    360 			if (error == 0)
    361 				error = EADDRNOTAVAIL;
    362 			return(error);
    363 		}
    364 		bzero(&mapped, sizeof(mapped));
    365 		mapped.s6_addr16[5] = htons(0xffff);
    366 		bcopy(&sinp->sin_addr, &mapped.s6_addr32[3], sizeof(sinp->sin_addr));
    367 		in6a = &mapped;
    368 #else
    369 		return EADDRNOTAVAIL;
    370 #endif
    371 	} else
    372 	{
    373 		/*
    374 		 * XXX: in6_selectsrc might replace the bound local address
    375 		 * with the address specified by setsockopt(IPV6_PKTINFO).
    376 		 * Is it the intended behavior?
    377 		 */
    378 		in6a = in6_selectsrc(sin6, in6p->in6p_outputopts,
    379 				     in6p->in6p_moptions,
    380 				     &in6p->in6p_route,
    381 				     &in6p->in6p_laddr, &error);
    382 		if (in6a == 0) {
    383 			if (error == 0)
    384 				error = EADDRNOTAVAIL;
    385 			return(error);
    386 		}
    387 	}
    388 	if (in6p->in6p_route.ro_rt)
    389 		ifp = in6p->in6p_route.ro_rt->rt_ifp;
    390 
    391 	in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim(in6p, ifp);
    392 
    393 	if (in6_pcblookup(in6p->in6p_head,
    394 			 &sin6->sin6_addr,
    395 			 sin6->sin6_port,
    396 			 IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ?
    397 			  in6a : &in6p->in6p_laddr,
    398 			 in6p->in6p_lport,
    399 			 0))
    400 		return(EADDRINUSE);
    401 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)
    402 	 || (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
    403 	  && in6p->in6p_laddr.s6_addr32[3] == 0))
    404 	{
    405 		/*
    406 		 * XXX in IPv4 mapped address case, we should grab fresh
    407 		 * local port number by in_pcbbind, not in6_pcbbind.
    408 		 * if we are in bad luck, we may assign conflicting port number
    409 		 * between IPv4 and IPv6 unwillingly.
    410 		 */
    411 		if (in6p->in6p_lport == 0) {
    412 			(void)in6_pcbbind(in6p, (struct mbuf *)0,
    413 			    (struct proc *)0);
    414 		}
    415 		in6p->in6p_laddr = *in6a;
    416 	}
    417 	in6p->in6p_faddr = sin6->sin6_addr;
    418 	in6p->in6p_fport = sin6->sin6_port;
    419 	/*
    420 	 * xxx kazu flowlabel is necessary for connect?
    421 	 * but if this line is missing, the garbage value remains.
    422 	 */
    423 	in6p->in6p_flowinfo = sin6->sin6_flowinfo;
    424 #ifdef IPSEC
    425 	if (in6p->in6p_socket->so_type == SOCK_STREAM)
    426 		ipsec_pcbconn(in6p->in6p_sp);
    427 #endif
    428 	return(0);
    429 }
    430 
    431 void
    432 in6_pcbdisconnect(in6p)
    433 	struct in6pcb *in6p;
    434 {
    435 	bzero((caddr_t)&in6p->in6p_faddr, sizeof(in6p->in6p_faddr));
    436 	in6p->in6p_fport = 0;
    437 	if (in6p->in6p_socket->so_state & SS_NOFDREF)
    438 		in6_pcbdetach(in6p);
    439 #ifdef IPSEC
    440 	ipsec_pcbdisconn(in6p->in6p_sp);
    441 #endif
    442 }
    443 
    444 void
    445 in6_pcbdetach(in6p)
    446 	struct in6pcb *in6p;
    447 {
    448 	struct socket *so = in6p->in6p_socket;
    449 	int s;
    450 
    451 #ifdef IPSEC
    452 	ipsec6_delete_pcbpolicy(in6p);
    453 #endif /* IPSEC */
    454 	sotoin6pcb(so) = 0;
    455 	sofree(so);
    456 	if (in6p->in6p_options)
    457 		m_freem(in6p->in6p_options);
    458 	if (in6p->in6p_outputopts) {
    459 		if (in6p->in6p_outputopts->ip6po_rthdr &&
    460 		    in6p->in6p_outputopts->ip6po_route.ro_rt)
    461 			RTFREE(in6p->in6p_outputopts->ip6po_route.ro_rt);
    462 		if (in6p->in6p_outputopts->ip6po_m)
    463 			(void)m_free(in6p->in6p_outputopts->ip6po_m);
    464 		free(in6p->in6p_outputopts, M_IP6OPT);
    465 	}
    466 	if (in6p->in6p_route.ro_rt)
    467 		rtfree(in6p->in6p_route.ro_rt);
    468 	ip6_freemoptions(in6p->in6p_moptions);
    469 	s = splnet();
    470 	in6p->in6p_next->in6p_prev = in6p->in6p_prev;
    471 	in6p->in6p_prev->in6p_next = in6p->in6p_next;
    472 	in6p->in6p_prev = NULL;
    473 	splx(s);
    474 	FREE(in6p, M_PCB);
    475 }
    476 
    477 void
    478 in6_setsockaddr(in6p, nam)
    479 	struct in6pcb *in6p;
    480 	struct mbuf *nam;
    481 {
    482 	struct sockaddr_in6 *sin6;
    483 
    484 	nam->m_len = sizeof(*sin6);
    485 	sin6 = mtod(nam, struct sockaddr_in6 *);
    486 	bzero((caddr_t)sin6, sizeof(*sin6));
    487 	sin6->sin6_family = AF_INET6;
    488 	sin6->sin6_len = sizeof(struct sockaddr_in6);
    489 	sin6->sin6_port = in6p->in6p_lport;
    490 	/* KAME hack: recover scopeid */
    491 	(void)in6_recoverscope(sin6, &in6p->in6p_laddr, NULL);
    492 }
    493 
    494 void
    495 in6_setpeeraddr(in6p, nam)
    496 	struct in6pcb *in6p;
    497 	struct mbuf *nam;
    498 {
    499 	struct sockaddr_in6 *sin6;
    500 
    501 	nam->m_len = sizeof(*sin6);
    502 	sin6 = mtod(nam, struct sockaddr_in6 *);
    503 	bzero((caddr_t)sin6, sizeof(*sin6));
    504 	sin6->sin6_family = AF_INET6;
    505 	sin6->sin6_len = sizeof(struct sockaddr_in6);
    506 	sin6->sin6_port = in6p->in6p_fport;
    507 	/* KAME hack: recover scopeid */
    508 	(void)in6_recoverscope(sin6, &in6p->in6p_faddr, NULL);
    509 }
    510 
    511 /*
    512  * Pass some notification to all connections of a protocol
    513  * associated with address dst.  The local address and/or port numbers
    514  * may be specified to limit the search.  The "usual action" will be
    515  * taken, depending on the ctlinput cmd.  The caller must filter any
    516  * cmds that are uninteresting (e.g., no error in the map).
    517  * Call the protocol specific routine (if any) to report
    518  * any errors for each matching socket.
    519  *
    520  * Must be called at splsoftnet.
    521  *
    522  * Note: src (4th arg) carries the flowlabel value on the original IPv6
    523  * header, in sin6_flowinfo member.
    524  */
    525 int
    526 in6_pcbnotify(head, dst, fport_arg, src, lport_arg, cmd, cmdarg, notify)
    527 	struct in6pcb *head;
    528 	struct sockaddr *dst, *src;
    529 	u_int fport_arg, lport_arg;
    530 	int cmd;
    531 	void *cmdarg;
    532 	void (*notify) __P((struct in6pcb *, int));
    533 {
    534 	struct in6pcb *in6p, *nin6p;
    535 	struct sockaddr_in6 sa6_src, *sa6_dst;
    536 	u_int16_t fport = fport_arg, lport = lport_arg;
    537 	int errno;
    538 	int nmatch = 0;
    539 	u_int32_t flowinfo;
    540 
    541 	if ((unsigned)cmd > PRC_NCMDS || dst->sa_family != AF_INET6)
    542 		return 0;
    543 
    544 	sa6_dst = (struct sockaddr_in6 *)dst;
    545 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
    546 		return 0;
    547 
    548 	/*
    549 	 * note that src can be NULL when we get notify by local fragmentation.
    550 	 */
    551 	sa6_src = (src == NULL) ? sa6_any : *(struct sockaddr_in6 *)src;
    552 	flowinfo = sa6_src.sin6_flowinfo;
    553 
    554 	/*
    555 	 * Redirects go to all references to the destination,
    556 	 * and use in6_rtchange to invalidate the route cache.
    557 	 * Dead host indications: also use in6_rtchange to invalidate
    558 	 * the cache, and deliver the error to all the sockets.
    559 	 * Otherwise, if we have knowledge of the local port and address,
    560 	 * deliver only to that socket.
    561 	 */
    562 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
    563 		fport = 0;
    564 		lport = 0;
    565 		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
    566 
    567 		if (cmd != PRC_HOSTDEAD)
    568 			notify = in6_rtchange;
    569 	}
    570 
    571 	errno = inet6ctlerrmap[cmd];
    572 	for (in6p = head->in6p_next; in6p != head; in6p = nin6p) {
    573 		nin6p = in6p->in6p_next;
    574 
    575 		/*
    576 		 * Under the following condition, notify of redirects
    577 		 * to the pcb, without making address matches against inpcb.
    578 		 * - redirect notification is arrived.
    579 		 * - the inpcb is unconnected.
    580 		 * - the inpcb is caching !RTF_HOST routing entry.
    581 		 * - the ICMPv6 notification is from the gateway cached in the
    582 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
    583 		 *   the inpcb used very recently.
    584 		 *
    585 		 * This is to improve interaction between netbsd/openbsd
    586 		 * redirect handling code, and inpcb route cache code.
    587 		 * without the clause, !RTF_HOST routing entry (which carries
    588 		 * gateway used by inpcb right before the ICMPv6 redirect)
    589 		 * will be cached forever in unconnected inpcb.
    590 		 *
    591 		 * There still is a question regarding to what is TRT:
    592 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
    593 		 *   generated on packet output.  inpcb will always cache
    594 		 *   RTF_HOST routing entry so there's no need for the clause
    595 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
    596 		 *   and inpcb is caching it already).
    597 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
    598 		 *   due to the cloned routing entries.
    599 		 * - Specwise, "destination cache" is mentioned in RFC2461.
    600 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
    601 		 *   is not really convinced.
    602 		 * - Having hiwat/lowat on # of cloned host route (redirect/
    603 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
    604 		 *   icmp6_mtudisc_update().
    605 		 */
    606 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
    607 		    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
    608 		    in6p->in6p_route.ro_rt &&
    609 		    !(in6p->in6p_route.ro_rt->rt_flags & RTF_HOST)) {
    610 			struct sockaddr_in6 *dst6;
    611 
    612 			dst6 = (struct sockaddr_in6 *)&in6p->in6p_route.ro_dst;
    613 			if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
    614 			    &sa6_dst->sin6_addr))
    615 				goto do_notify;
    616 		}
    617 
    618 		/*
    619 		 * Detect if we should notify the error. If no source and
    620 		 * destination ports are specified, but non-zero flowinfo and
    621 		 * local address match, notify the error. This is the case
    622 		 * when the error is delivered with an encrypted buffer
    623 		 * by ESP. Otherwise, just compare addresses and ports
    624 		 * as usual.
    625 		 */
    626 		if (lport == 0 && fport == 0 && flowinfo &&
    627 		    in6p->in6p_socket != NULL &&
    628 		    flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
    629 		    IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
    630 			goto do_notify;
    631 		else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
    632 					     &sa6_dst->sin6_addr) ||
    633 		    in6p->in6p_socket == 0 ||
    634 		    (lport && in6p->in6p_lport != lport) ||
    635 		    (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
    636 		     !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
    637 					 &sa6_src.sin6_addr)) ||
    638 		    (fport && in6p->in6p_fport != fport))
    639 			continue;
    640 
    641 	  do_notify:
    642 		if (notify)
    643 			(*notify)(in6p, errno);
    644 		nmatch++;
    645 	}
    646 	return nmatch;
    647 }
    648 
    649 void
    650 in6_pcbpurgeif0(head, ifp)
    651 	struct in6pcb *head;
    652 	struct ifnet *ifp;
    653 {
    654 	struct in6pcb *in6p, *nin6p;
    655 	struct ip6_moptions *im6o;
    656 	struct in6_multi_mship *imm, *nimm;
    657 
    658 	for (in6p = head->in6p_next; in6p != head; in6p = nin6p) {
    659 		nin6p = in6p->in6p_next;
    660 		im6o = in6p->in6p_moptions;
    661 		if (im6o) {
    662 			/*
    663 			 * Unselect the outgoing interface if it is being
    664 			 * detached.
    665 			 */
    666 			if (im6o->im6o_multicast_ifp == ifp)
    667 				im6o->im6o_multicast_ifp = NULL;
    668 
    669 			/*
    670 			 * Drop multicast group membership if we joined
    671 			 * through the interface being detached.
    672 			 * XXX controversial - is it really legal for kernel
    673 			 * to force this?
    674 			 */
    675 			for (imm = im6o->im6o_memberships.lh_first;
    676 			     imm != NULL; imm = nimm) {
    677 				nimm = imm->i6mm_chain.le_next;
    678 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
    679 					LIST_REMOVE(imm, i6mm_chain);
    680 					in6_leavegroup(imm);
    681 				}
    682 			}
    683 		}
    684 	}
    685 }
    686 
    687 void
    688 in6_pcbpurgeif(head, ifp)
    689 	struct in6pcb *head;
    690 	struct ifnet *ifp;
    691 {
    692 	struct in6pcb *in6p, *nin6p;
    693 
    694 	for (in6p = head->in6p_next; in6p != head; in6p = nin6p) {
    695 		nin6p = in6p->in6p_next;
    696 		if (in6p->in6p_route.ro_rt != NULL &&
    697 		    in6p->in6p_route.ro_rt->rt_ifp == ifp)
    698 			in6_rtchange(in6p, 0);
    699 	}
    700 }
    701 
    702 /*
    703  * Check for alternatives when higher level complains
    704  * about service problems.  For now, invalidate cached
    705  * routing information.  If the route was created dynamically
    706  * (by a redirect), time to try a default gateway again.
    707  */
    708 void
    709 in6_losing(in6p)
    710 	struct in6pcb *in6p;
    711 {
    712 	struct rtentry *rt;
    713 	struct rt_addrinfo info;
    714 
    715 	if ((rt = in6p->in6p_route.ro_rt) != NULL) {
    716 		in6p->in6p_route.ro_rt = 0;
    717 		bzero((caddr_t)&info, sizeof(info));
    718 		info.rti_info[RTAX_DST] =
    719 			(struct sockaddr *)&in6p->in6p_route.ro_dst;
    720 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    721 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    722 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
    723 		if (rt->rt_flags & RTF_DYNAMIC) {
    724 			(void)rtrequest(RTM_DELETE, rt_key(rt),
    725 					rt->rt_gateway, rt_mask(rt), rt->rt_flags,
    726 					(struct rtentry **)0);
    727 		} else {
    728 			/*
    729 			 * A new route can be allocated
    730 			 * the next time output is attempted.
    731 			 */
    732 			rtfree(rt);
    733 		}
    734 	}
    735 }
    736 
    737 /*
    738  * After a routing change, flush old routing
    739  * and allocate a (hopefully) better one.
    740  */
    741 void
    742 in6_rtchange(in6p, errno)
    743 	struct in6pcb *in6p;
    744 	int errno;
    745 {
    746 	if (in6p->in6p_route.ro_rt) {
    747 		rtfree(in6p->in6p_route.ro_rt);
    748 		in6p->in6p_route.ro_rt = 0;
    749 		/*
    750 		 * A new route can be allocated the next time
    751 		 * output is attempted.
    752 		 */
    753 	}
    754 }
    755 
    756 struct in6pcb *
    757 in6_pcblookup(head, faddr6, fport_arg, laddr6, lport_arg, flags)
    758 	struct in6pcb *head;
    759 	struct in6_addr *faddr6, *laddr6;
    760 	u_int fport_arg, lport_arg;
    761 	int flags;
    762 {
    763 	struct in6pcb *in6p, *match = 0;
    764 	int matchwild = 3, wildcard;
    765 	u_int16_t fport = fport_arg, lport = lport_arg;
    766 
    767 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
    768 		if (in6p->in6p_lport != lport)
    769 			continue;
    770 		wildcard = 0;
    771 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
    772 			if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
    773 				wildcard++;
    774 			else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
    775 				continue;
    776 		}
    777 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
    778 			&& in6p->in6p_laddr.s6_addr32[3] == 0) {
    779 			if (!IN6_IS_ADDR_V4MAPPED(laddr6))
    780 				continue;
    781 			if (laddr6->s6_addr32[3] == 0)
    782 				;
    783 			else
    784 				wildcard++;
    785 		}
    786 		else {
    787 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
    788 				if (in6p->in6p_flags & IN6P_IPV6_V6ONLY)
    789 					continue;
    790 				else
    791 					wildcard++;
    792 			} else if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
    793 				wildcard++;
    794 		}
    795 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
    796 			if (IN6_IS_ADDR_UNSPECIFIED(faddr6))
    797 				wildcard++;
    798 			else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6)
    799 			      || in6p->in6p_fport != fport)
    800 				continue;
    801 		}
    802 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)
    803 			&& in6p->in6p_faddr.s6_addr32[3] == 0) {
    804 			if (!IN6_IS_ADDR_V4MAPPED(faddr6))
    805 				continue;
    806 			if (faddr6->s6_addr32[3] == 0)
    807 				;
    808 			else
    809 				wildcard++;
    810 		}
    811 		else {
    812 			if (IN6_IS_ADDR_V4MAPPED(faddr6)) {
    813 				if (in6p->in6p_flags & IN6P_IPV6_V6ONLY)
    814 					continue;
    815 				else
    816 					wildcard++;
    817 			} else if (!IN6_IS_ADDR_UNSPECIFIED(faddr6))
    818 				wildcard++;
    819 		}
    820 
    821 		if (wildcard && (flags & IN6PLOOKUP_WILDCARD) == 0)
    822 			continue;
    823 		if (wildcard < matchwild) {
    824 			match = in6p;
    825 			matchwild = wildcard;
    826 			if (matchwild == 0)
    827 				break;
    828 		}
    829 	}
    830 	return(match);
    831 }
    832 
    833 struct rtentry *
    834 in6_pcbrtentry(in6p)
    835 	struct in6pcb *in6p;
    836 {
    837 	struct route_in6 *ro;
    838 	struct sockaddr_in6 *dst6;
    839 
    840 	ro = &in6p->in6p_route;
    841 	dst6 = (struct sockaddr_in6 *)&ro->ro_dst;
    842 
    843 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    844 	    !IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr, &in6p->in6p_faddr))) {
    845 		RTFREE(ro->ro_rt);
    846 		ro->ro_rt = (struct rtentry *)NULL;
    847 	}
    848 	if (ro->ro_rt == (struct rtentry *)NULL &&
    849 	    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
    850 		bzero(dst6, sizeof(*dst6));
    851 		dst6->sin6_family = AF_INET6;
    852 		dst6->sin6_len = sizeof(struct sockaddr_in6);
    853 		dst6->sin6_addr = in6p->in6p_faddr;
    854 		rtalloc((struct route *)ro);
    855 	}
    856 	return (ro->ro_rt);
    857 }
    858 
    859 struct in6pcb *
    860 in6_pcblookup_connect(head, faddr6, fport_arg, laddr6, lport_arg, faith)
    861 	struct in6pcb *head;
    862 	struct in6_addr *faddr6, *laddr6;
    863 	u_int fport_arg, lport_arg;
    864 	int faith;
    865 {
    866 	struct in6pcb *in6p;
    867 	u_int16_t fport = fport_arg, lport = lport_arg;
    868 
    869 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
    870 		/* find exact match on both source and dest */
    871 		if (in6p->in6p_fport != fport)
    872 			continue;
    873 		if (in6p->in6p_lport != lport)
    874 			continue;
    875 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
    876 			continue;
    877 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6))
    878 			continue;
    879 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
    880 			continue;
    881 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
    882 			continue;
    883 		if ((IN6_IS_ADDR_V4MAPPED(laddr6) ||
    884 		     IN6_IS_ADDR_V4MAPPED(faddr6)) &&
    885 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
    886 			continue;
    887 		return in6p;
    888 	}
    889 	return NULL;
    890 }
    891 
    892 struct in6pcb *
    893 in6_pcblookup_bind(head, laddr6, lport_arg, faith)
    894 	struct in6pcb *head;
    895 	struct in6_addr *laddr6;
    896 	u_int lport_arg;
    897 	int faith;
    898 {
    899 	struct in6pcb *in6p, *match;
    900 	u_int16_t lport = lport_arg;
    901 
    902 	match = NULL;
    903 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
    904 		/*
    905 	 	 * find destination match.  exact match is preferred
    906 		 * against wildcard match.
    907 		 */
    908 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
    909 			continue;
    910 		if (in6p->in6p_fport != 0)
    911 			continue;
    912 		if (in6p->in6p_lport != lport)
    913 			continue;
    914 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
    915 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
    916 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY))
    917 					continue;
    918 				else
    919 					match = in6p;
    920 			} else
    921 				match = in6p;
    922 		}
    923 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
    924 			 !(in6p->in6p_flags & IN6P_IPV6_V6ONLY) &&
    925 			 in6p->in6p_laddr.s6_addr32[3] == 0) {
    926 			if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
    927 			    laddr6->s6_addr32[3] != 0)
    928 				match = in6p;
    929 		}
    930 		else if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
    931 			return in6p;
    932 	}
    933 	return match;
    934 }
    935