xref: /freebsd/sys/netinet6/raw_ip6.c (revision 3823d5e198425b4f5e5a80267d195769d1063773)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 /*-
31  * Copyright (c) 1982, 1986, 1988, 1993
32  *	The Regents of the University of California.
33  * All rights reserved.
34  *
35  * Redistribution and use in source and binary forms, with or without
36  * modification, are permitted provided that the following conditions
37  * are met:
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 4. Neither the name of the University nor the names of its contributors
44  *    may be used to endorse or promote products derived from this software
45  *    without specific prior written permission.
46  *
47  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
48  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
49  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
50  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
51  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
52  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
53  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
54  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
55  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
56  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
57  * SUCH DAMAGE.
58  *
59  *	@(#)raw_ip.c	8.2 (Berkeley) 1/4/94
60  */
61 
62 #include <sys/cdefs.h>
63 __FBSDID("$FreeBSD$");
64 
65 #include "opt_ipsec.h"
66 #include "opt_inet6.h"
67 
68 #include <sys/param.h>
69 #include <sys/errno.h>
70 #include <sys/jail.h>
71 #include <sys/kernel.h>
72 #include <sys/lock.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/priv.h>
76 #include <sys/proc.h>
77 #include <sys/protosw.h>
78 #include <sys/signalvar.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/sx.h>
82 #include <sys/syslog.h>
83 
84 #include <net/if.h>
85 #include <net/if_var.h>
86 #include <net/if_types.h>
87 #include <net/route.h>
88 #include <net/vnet.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_var.h>
92 #include <netinet/in_systm.h>
93 #include <netinet/in_pcb.h>
94 
95 #include <netinet/icmp6.h>
96 #include <netinet/ip6.h>
97 #include <netinet/ip_var.h>
98 #include <netinet6/ip6protosw.h>
99 #include <netinet6/ip6_mroute.h>
100 #include <netinet6/in6_pcb.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/nd6.h>
103 #include <netinet6/raw_ip6.h>
104 #include <netinet6/scope6_var.h>
105 #include <netinet6/send.h>
106 
107 #ifdef IPSEC
108 #include <netipsec/ipsec.h>
109 #include <netipsec/ipsec6.h>
110 #endif /* IPSEC */
111 
112 #include <machine/stdarg.h>
113 
114 #define	satosin6(sa)	((struct sockaddr_in6 *)(sa))
115 #define	ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
116 
117 /*
118  * Raw interface to IP6 protocol.
119  */
120 
121 VNET_DECLARE(struct inpcbhead, ripcb);
122 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
123 #define	V_ripcb				VNET(ripcb)
124 #define	V_ripcbinfo			VNET(ripcbinfo)
125 
126 extern u_long	rip_sendspace;
127 extern u_long	rip_recvspace;
128 
129 VNET_PCPUSTAT_DEFINE(struct rip6stat, rip6stat);
130 VNET_PCPUSTAT_SYSINIT(rip6stat);
131 
132 #ifdef VIMAGE
133 VNET_PCPUSTAT_SYSUNINIT(rip6stat);
134 #endif /* VIMAGE */
135 
136 /*
137  * Hooks for multicast routing. They all default to NULL, so leave them not
138  * initialized and rely on BSS being set to 0.
139  */
140 
141 /*
142  * The socket used to communicate with the multicast routing daemon.
143  */
144 VNET_DEFINE(struct socket *, ip6_mrouter);
145 
146 /*
147  * The various mrouter functions.
148  */
149 int (*ip6_mrouter_set)(struct socket *, struct sockopt *);
150 int (*ip6_mrouter_get)(struct socket *, struct sockopt *);
151 int (*ip6_mrouter_done)(void);
152 int (*ip6_mforward)(struct ip6_hdr *, struct ifnet *, struct mbuf *);
153 int (*mrt6_ioctl)(u_long, caddr_t);
154 
155 /*
156  * Setup generic address and protocol structures for raw_input routine, then
157  * pass them along with mbuf chain.
158  */
159 int
160 rip6_input(struct mbuf **mp, int *offp, int proto)
161 {
162 	struct ifnet *ifp;
163 	struct mbuf *m = *mp;
164 	register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
165 	register struct inpcb *in6p;
166 	struct inpcb *last = 0;
167 	struct mbuf *opts = NULL;
168 	struct sockaddr_in6 fromsa;
169 
170 	RIP6STAT_INC(rip6s_ipackets);
171 
172 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
173 		/* XXX Send icmp6 host/port unreach? */
174 		m_freem(m);
175 		return (IPPROTO_DONE);
176 	}
177 
178 	init_sin6(&fromsa, m); /* general init */
179 
180 	ifp = m->m_pkthdr.rcvif;
181 
182 	INP_INFO_RLOCK(&V_ripcbinfo);
183 	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
184 		/* XXX inp locking */
185 		if ((in6p->inp_vflag & INP_IPV6) == 0)
186 			continue;
187 		if (in6p->inp_ip_p &&
188 		    in6p->inp_ip_p != proto)
189 			continue;
190 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
191 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
192 			continue;
193 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
194 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
195 			continue;
196 		if (jailed_without_vnet(in6p->inp_cred)) {
197 			/*
198 			 * Allow raw socket in jail to receive multicast;
199 			 * assume process had PRIV_NETINET_RAW at attach,
200 			 * and fall through into normal filter path if so.
201 			 */
202 			if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
203 			    prison_check_ip6(in6p->inp_cred,
204 			    &ip6->ip6_dst) != 0)
205 				continue;
206 		}
207 		INP_RLOCK(in6p);
208 		if (in6p->in6p_cksum != -1) {
209 			RIP6STAT_INC(rip6s_isum);
210 			if (in6_cksum(m, proto, *offp,
211 			    m->m_pkthdr.len - *offp)) {
212 				INP_RUNLOCK(in6p);
213 				RIP6STAT_INC(rip6s_badsum);
214 				continue;
215 			}
216 		}
217 		/*
218 		 * If this raw socket has multicast state, and we
219 		 * have received a multicast, check if this socket
220 		 * should receive it, as multicast filtering is now
221 		 * the responsibility of the transport layer.
222 		 */
223 		if (in6p->in6p_moptions &&
224 		    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
225 			/*
226 			 * If the incoming datagram is for MLD, allow it
227 			 * through unconditionally to the raw socket.
228 			 *
229 			 * Use the M_RTALERT_MLD flag to check for MLD
230 			 * traffic without having to inspect the mbuf chain
231 			 * more deeply, as all MLDv1/v2 host messages MUST
232 			 * contain the Router Alert option.
233 			 *
234 			 * In the case of MLDv1, we may not have explicitly
235 			 * joined the group, and may have set IFF_ALLMULTI
236 			 * on the interface. im6o_mc_filter() may discard
237 			 * control traffic we actually need to see.
238 			 *
239 			 * Userland multicast routing daemons should continue
240 			 * filter the control traffic appropriately.
241 			 */
242 			int blocked;
243 
244 			blocked = MCAST_PASS;
245 			if ((m->m_flags & M_RTALERT_MLD) == 0) {
246 				struct sockaddr_in6 mcaddr;
247 
248 				bzero(&mcaddr, sizeof(struct sockaddr_in6));
249 				mcaddr.sin6_len = sizeof(struct sockaddr_in6);
250 				mcaddr.sin6_family = AF_INET6;
251 				mcaddr.sin6_addr = ip6->ip6_dst;
252 
253 				blocked = im6o_mc_filter(in6p->in6p_moptions,
254 				    ifp,
255 				    (struct sockaddr *)&mcaddr,
256 				    (struct sockaddr *)&fromsa);
257 			}
258 			if (blocked != MCAST_PASS) {
259 				IP6STAT_INC(ip6s_notmember);
260 				INP_RUNLOCK(in6p);
261 				continue;
262 			}
263 		}
264 		if (last != NULL) {
265 			struct mbuf *n = m_copy(m, 0, (int)M_COPYALL);
266 
267 #ifdef IPSEC
268 			/*
269 			 * Check AH/ESP integrity.
270 			 */
271 			if (n && ipsec6_in_reject(n, last)) {
272 				m_freem(n);
273 				IPSEC6STAT_INC(ips_in_polvio);
274 				/* Do not inject data into pcb. */
275 			} else
276 #endif /* IPSEC */
277 			if (n) {
278 				if (last->inp_flags & INP_CONTROLOPTS ||
279 				    last->inp_socket->so_options & SO_TIMESTAMP)
280 					ip6_savecontrol(last, n, &opts);
281 				/* strip intermediate headers */
282 				m_adj(n, *offp);
283 				if (sbappendaddr(&last->inp_socket->so_rcv,
284 						(struct sockaddr *)&fromsa,
285 						 n, opts) == 0) {
286 					m_freem(n);
287 					if (opts)
288 						m_freem(opts);
289 					RIP6STAT_INC(rip6s_fullsock);
290 				} else
291 					sorwakeup(last->inp_socket);
292 				opts = NULL;
293 			}
294 			INP_RUNLOCK(last);
295 		}
296 		last = in6p;
297 	}
298 	INP_INFO_RUNLOCK(&V_ripcbinfo);
299 #ifdef IPSEC
300 	/*
301 	 * Check AH/ESP integrity.
302 	 */
303 	if ((last != NULL) && ipsec6_in_reject(m, last)) {
304 		m_freem(m);
305 		IPSEC6STAT_INC(ips_in_polvio);
306 		IP6STAT_DEC(ip6s_delivered);
307 		/* Do not inject data into pcb. */
308 		INP_RUNLOCK(last);
309 	} else
310 #endif /* IPSEC */
311 	if (last != NULL) {
312 		if (last->inp_flags & INP_CONTROLOPTS ||
313 		    last->inp_socket->so_options & SO_TIMESTAMP)
314 			ip6_savecontrol(last, m, &opts);
315 		/* Strip intermediate headers. */
316 		m_adj(m, *offp);
317 		if (sbappendaddr(&last->inp_socket->so_rcv,
318 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
319 			m_freem(m);
320 			if (opts)
321 				m_freem(opts);
322 			RIP6STAT_INC(rip6s_fullsock);
323 		} else
324 			sorwakeup(last->inp_socket);
325 		INP_RUNLOCK(last);
326 	} else {
327 		RIP6STAT_INC(rip6s_nosock);
328 		if (m->m_flags & M_MCAST)
329 			RIP6STAT_INC(rip6s_nosockmcast);
330 		if (proto == IPPROTO_NONE)
331 			m_freem(m);
332 		else {
333 			char *prvnxtp = ip6_get_prevhdr(m, *offp); /* XXX */
334 			icmp6_error(m, ICMP6_PARAM_PROB,
335 			    ICMP6_PARAMPROB_NEXTHEADER,
336 			    prvnxtp - mtod(m, char *));
337 		}
338 		IP6STAT_DEC(ip6s_delivered);
339 	}
340 	return (IPPROTO_DONE);
341 }
342 
343 void
344 rip6_ctlinput(int cmd, struct sockaddr *sa, void *d)
345 {
346 	struct ip6_hdr *ip6;
347 	struct mbuf *m;
348 	int off = 0;
349 	struct ip6ctlparam *ip6cp = NULL;
350 	const struct sockaddr_in6 *sa6_src = NULL;
351 	void *cmdarg;
352 	struct inpcb *(*notify)(struct inpcb *, int) = in6_rtchange;
353 
354 	if (sa->sa_family != AF_INET6 ||
355 	    sa->sa_len != sizeof(struct sockaddr_in6))
356 		return;
357 
358 	if ((unsigned)cmd >= PRC_NCMDS)
359 		return;
360 	if (PRC_IS_REDIRECT(cmd))
361 		notify = in6_rtchange, d = NULL;
362 	else if (cmd == PRC_HOSTDEAD)
363 		d = NULL;
364 	else if (inet6ctlerrmap[cmd] == 0)
365 		return;
366 
367 	/*
368 	 * If the parameter is from icmp6, decode it.
369 	 */
370 	if (d != NULL) {
371 		ip6cp = (struct ip6ctlparam *)d;
372 		m = ip6cp->ip6c_m;
373 		ip6 = ip6cp->ip6c_ip6;
374 		off = ip6cp->ip6c_off;
375 		cmdarg = ip6cp->ip6c_cmdarg;
376 		sa6_src = ip6cp->ip6c_src;
377 	} else {
378 		m = NULL;
379 		ip6 = NULL;
380 		cmdarg = NULL;
381 		sa6_src = &sa6_any;
382 	}
383 
384 	(void) in6_pcbnotify(&V_ripcbinfo, sa, 0,
385 	    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
386 }
387 
388 /*
389  * Generate IPv6 header and pass packet to ip6_output.  Tack on options user
390  * may have setup with control call.
391  */
392 int
393 rip6_output(struct mbuf *m, struct socket *so, ...)
394 {
395 	struct mbuf *control;
396 	struct m_tag *mtag;
397 	struct sockaddr_in6 *dstsock;
398 	struct in6_addr *dst;
399 	struct ip6_hdr *ip6;
400 	struct inpcb *in6p;
401 	u_int	plen = m->m_pkthdr.len;
402 	int error = 0;
403 	struct ip6_pktopts opt, *optp;
404 	struct ifnet *oifp = NULL;
405 	int type = 0, code = 0;		/* for ICMPv6 output statistics only */
406 	int scope_ambiguous = 0;
407 	int use_defzone = 0;
408 	struct in6_addr in6a;
409 	va_list ap;
410 
411 	va_start(ap, so);
412 	dstsock = va_arg(ap, struct sockaddr_in6 *);
413 	control = va_arg(ap, struct mbuf *);
414 	va_end(ap);
415 
416 	in6p = sotoinpcb(so);
417 	INP_WLOCK(in6p);
418 
419 	dst = &dstsock->sin6_addr;
420 	if (control != NULL) {
421 		if ((error = ip6_setpktopts(control, &opt,
422 		    in6p->in6p_outputopts, so->so_cred,
423 		    so->so_proto->pr_protocol)) != 0) {
424 			goto bad;
425 		}
426 		optp = &opt;
427 	} else
428 		optp = in6p->in6p_outputopts;
429 
430 	/*
431 	 * Check and convert scope zone ID into internal form.
432 	 *
433 	 * XXX: we may still need to determine the zone later.
434 	 */
435 	if (!(so->so_state & SS_ISCONNECTED)) {
436 		if (!optp || !optp->ip6po_pktinfo ||
437 		    !optp->ip6po_pktinfo->ipi6_ifindex)
438 			use_defzone = V_ip6_use_defzone;
439 		if (dstsock->sin6_scope_id == 0 && !use_defzone)
440 			scope_ambiguous = 1;
441 		if ((error = sa6_embedscope(dstsock, use_defzone)) != 0)
442 			goto bad;
443 	}
444 
445 	/*
446 	 * For an ICMPv6 packet, we should know its type and code to update
447 	 * statistics.
448 	 */
449 	if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
450 		struct icmp6_hdr *icmp6;
451 		if (m->m_len < sizeof(struct icmp6_hdr) &&
452 		    (m = m_pullup(m, sizeof(struct icmp6_hdr))) == NULL) {
453 			error = ENOBUFS;
454 			goto bad;
455 		}
456 		icmp6 = mtod(m, struct icmp6_hdr *);
457 		type = icmp6->icmp6_type;
458 		code = icmp6->icmp6_code;
459 	}
460 
461 	M_PREPEND(m, sizeof(*ip6), M_NOWAIT);
462 	if (m == NULL) {
463 		error = ENOBUFS;
464 		goto bad;
465 	}
466 	ip6 = mtod(m, struct ip6_hdr *);
467 
468 	/*
469 	 * Source address selection.
470 	 */
471 	error = in6_selectsrc(dstsock, optp, in6p, NULL, so->so_cred,
472 	    &oifp, &in6a);
473 	if (error)
474 		goto bad;
475 	error = prison_check_ip6(in6p->inp_cred, &in6a);
476 	if (error != 0)
477 		goto bad;
478 	ip6->ip6_src = in6a;
479 
480 	if (oifp && scope_ambiguous) {
481 		/*
482 		 * Application should provide a proper zone ID or the use of
483 		 * default zone IDs should be enabled.  Unfortunately, some
484 		 * applications do not behave as it should, so we need a
485 		 * workaround.  Even if an appropriate ID is not determined
486 		 * (when it's required), if we can determine the outgoing
487 		 * interface. determine the zone ID based on the interface.
488 		 */
489 		error = in6_setscope(&dstsock->sin6_addr, oifp, NULL);
490 		if (error != 0)
491 			goto bad;
492 	}
493 	ip6->ip6_dst = dstsock->sin6_addr;
494 
495 	/*
496 	 * Fill in the rest of the IPv6 header fields.
497 	 */
498 	ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) |
499 	    (in6p->inp_flow & IPV6_FLOWINFO_MASK);
500 	ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) |
501 	    (IPV6_VERSION & IPV6_VERSION_MASK);
502 
503 	/*
504 	 * ip6_plen will be filled in ip6_output, so not fill it here.
505 	 */
506 	ip6->ip6_nxt = in6p->inp_ip_p;
507 	ip6->ip6_hlim = in6_selecthlim(in6p, oifp);
508 
509 	if (so->so_proto->pr_protocol == IPPROTO_ICMPV6 ||
510 	    in6p->in6p_cksum != -1) {
511 		struct mbuf *n;
512 		int off;
513 		u_int16_t *p;
514 
515 		/* Compute checksum. */
516 		if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
517 			off = offsetof(struct icmp6_hdr, icmp6_cksum);
518 		else
519 			off = in6p->in6p_cksum;
520 		if (plen < off + 1) {
521 			error = EINVAL;
522 			goto bad;
523 		}
524 		off += sizeof(struct ip6_hdr);
525 
526 		n = m;
527 		while (n && n->m_len <= off) {
528 			off -= n->m_len;
529 			n = n->m_next;
530 		}
531 		if (!n)
532 			goto bad;
533 		p = (u_int16_t *)(mtod(n, caddr_t) + off);
534 		*p = 0;
535 		*p = in6_cksum(m, ip6->ip6_nxt, sizeof(*ip6), plen);
536 	}
537 
538 	/*
539 	 * Send RA/RS messages to user land for protection, before sending
540 	 * them to rtadvd/rtsol.
541 	 */
542 	if ((send_sendso_input_hook != NULL) &&
543 	    so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
544 		switch (type) {
545 		case ND_ROUTER_ADVERT:
546 		case ND_ROUTER_SOLICIT:
547 			mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
548 				sizeof(unsigned short), M_NOWAIT);
549 			if (mtag == NULL)
550 				goto bad;
551 			m_tag_prepend(m, mtag);
552 		}
553 	}
554 
555 	error = ip6_output(m, optp, NULL, 0, in6p->in6p_moptions, &oifp, in6p);
556 	if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
557 		if (oifp)
558 			icmp6_ifoutstat_inc(oifp, type, code);
559 		ICMP6STAT_INC(icp6s_outhist[type]);
560 	} else
561 		RIP6STAT_INC(rip6s_opackets);
562 
563 	goto freectl;
564 
565  bad:
566 	if (m)
567 		m_freem(m);
568 
569  freectl:
570 	if (control != NULL) {
571 		ip6_clearpktopts(&opt, -1);
572 		m_freem(control);
573 	}
574 	INP_WUNLOCK(in6p);
575 	return (error);
576 }
577 
578 /*
579  * Raw IPv6 socket option processing.
580  */
581 int
582 rip6_ctloutput(struct socket *so, struct sockopt *sopt)
583 {
584 	struct inpcb *inp;
585 	int error;
586 
587 	if (sopt->sopt_level == IPPROTO_ICMPV6)
588 		/*
589 		 * XXX: is it better to call icmp6_ctloutput() directly
590 		 * from protosw?
591 		 */
592 		return (icmp6_ctloutput(so, sopt));
593 	else if (sopt->sopt_level != IPPROTO_IPV6) {
594 		if (sopt->sopt_level == SOL_SOCKET &&
595 		    sopt->sopt_name == SO_SETFIB) {
596 			inp = sotoinpcb(so);
597 			INP_WLOCK(inp);
598 			inp->inp_inc.inc_fibnum = so->so_fibnum;
599 			INP_WUNLOCK(inp);
600 			return (0);
601 		}
602 		return (EINVAL);
603 	}
604 
605 	error = 0;
606 
607 	switch (sopt->sopt_dir) {
608 	case SOPT_GET:
609 		switch (sopt->sopt_name) {
610 		case MRT6_INIT:
611 		case MRT6_DONE:
612 		case MRT6_ADD_MIF:
613 		case MRT6_DEL_MIF:
614 		case MRT6_ADD_MFC:
615 		case MRT6_DEL_MFC:
616 		case MRT6_PIM:
617 			error = ip6_mrouter_get ?  ip6_mrouter_get(so, sopt) :
618 			    EOPNOTSUPP;
619 			break;
620 		case IPV6_CHECKSUM:
621 			error = ip6_raw_ctloutput(so, sopt);
622 			break;
623 		default:
624 			error = ip6_ctloutput(so, sopt);
625 			break;
626 		}
627 		break;
628 
629 	case SOPT_SET:
630 		switch (sopt->sopt_name) {
631 		case MRT6_INIT:
632 		case MRT6_DONE:
633 		case MRT6_ADD_MIF:
634 		case MRT6_DEL_MIF:
635 		case MRT6_ADD_MFC:
636 		case MRT6_DEL_MFC:
637 		case MRT6_PIM:
638 			error = ip6_mrouter_set ?  ip6_mrouter_set(so, sopt) :
639 			    EOPNOTSUPP;
640 			break;
641 		case IPV6_CHECKSUM:
642 			error = ip6_raw_ctloutput(so, sopt);
643 			break;
644 		default:
645 			error = ip6_ctloutput(so, sopt);
646 			break;
647 		}
648 		break;
649 	}
650 
651 	return (error);
652 }
653 
654 static int
655 rip6_attach(struct socket *so, int proto, struct thread *td)
656 {
657 	struct inpcb *inp;
658 	struct icmp6_filter *filter;
659 	int error;
660 
661 	inp = sotoinpcb(so);
662 	KASSERT(inp == NULL, ("rip6_attach: inp != NULL"));
663 
664 	error = priv_check(td, PRIV_NETINET_RAW);
665 	if (error)
666 		return (error);
667 	error = soreserve(so, rip_sendspace, rip_recvspace);
668 	if (error)
669 		return (error);
670 	filter = malloc(sizeof(struct icmp6_filter), M_PCB, M_NOWAIT);
671 	if (filter == NULL)
672 		return (ENOMEM);
673 	INP_INFO_WLOCK(&V_ripcbinfo);
674 	error = in_pcballoc(so, &V_ripcbinfo);
675 	if (error) {
676 		INP_INFO_WUNLOCK(&V_ripcbinfo);
677 		free(filter, M_PCB);
678 		return (error);
679 	}
680 	inp = (struct inpcb *)so->so_pcb;
681 	INP_INFO_WUNLOCK(&V_ripcbinfo);
682 	inp->inp_vflag |= INP_IPV6;
683 	inp->inp_ip_p = (long)proto;
684 	inp->in6p_hops = -1;	/* use kernel default */
685 	inp->in6p_cksum = -1;
686 	inp->in6p_icmp6filt = filter;
687 	ICMP6_FILTER_SETPASSALL(inp->in6p_icmp6filt);
688 	INP_WUNLOCK(inp);
689 	return (0);
690 }
691 
692 static void
693 rip6_detach(struct socket *so)
694 {
695 	struct inpcb *inp;
696 
697 	inp = sotoinpcb(so);
698 	KASSERT(inp != NULL, ("rip6_detach: inp == NULL"));
699 
700 	if (so == V_ip6_mrouter && ip6_mrouter_done)
701 		ip6_mrouter_done();
702 	/* xxx: RSVP */
703 	INP_INFO_WLOCK(&V_ripcbinfo);
704 	INP_WLOCK(inp);
705 	free(inp->in6p_icmp6filt, M_PCB);
706 	in_pcbdetach(inp);
707 	in_pcbfree(inp);
708 	INP_INFO_WUNLOCK(&V_ripcbinfo);
709 }
710 
711 /* XXXRW: This can't ever be called. */
712 static void
713 rip6_abort(struct socket *so)
714 {
715 	struct inpcb *inp;
716 
717 	inp = sotoinpcb(so);
718 	KASSERT(inp != NULL, ("rip6_abort: inp == NULL"));
719 
720 	soisdisconnected(so);
721 }
722 
723 static void
724 rip6_close(struct socket *so)
725 {
726 	struct inpcb *inp;
727 
728 	inp = sotoinpcb(so);
729 	KASSERT(inp != NULL, ("rip6_close: inp == NULL"));
730 
731 	soisdisconnected(so);
732 }
733 
734 static int
735 rip6_disconnect(struct socket *so)
736 {
737 	struct inpcb *inp;
738 
739 	inp = sotoinpcb(so);
740 	KASSERT(inp != NULL, ("rip6_disconnect: inp == NULL"));
741 
742 	if ((so->so_state & SS_ISCONNECTED) == 0)
743 		return (ENOTCONN);
744 	inp->in6p_faddr = in6addr_any;
745 	rip6_abort(so);
746 	return (0);
747 }
748 
749 static int
750 rip6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
751 {
752 	struct inpcb *inp;
753 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)nam;
754 	struct ifaddr *ifa = NULL;
755 	int error = 0;
756 
757 	inp = sotoinpcb(so);
758 	KASSERT(inp != NULL, ("rip6_bind: inp == NULL"));
759 
760 	if (nam->sa_len != sizeof(*addr))
761 		return (EINVAL);
762 	if ((error = prison_check_ip6(td->td_ucred, &addr->sin6_addr)) != 0)
763 		return (error);
764 	if (TAILQ_EMPTY(&V_ifnet) || addr->sin6_family != AF_INET6)
765 		return (EADDRNOTAVAIL);
766 	if ((error = sa6_embedscope(addr, V_ip6_use_defzone)) != 0)
767 		return (error);
768 
769 	if (!IN6_IS_ADDR_UNSPECIFIED(&addr->sin6_addr) &&
770 	    (ifa = ifa_ifwithaddr((struct sockaddr *)addr)) == NULL)
771 		return (EADDRNOTAVAIL);
772 	if (ifa != NULL &&
773 	    ((struct in6_ifaddr *)ifa)->ia6_flags &
774 	    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|
775 	     IN6_IFF_DETACHED|IN6_IFF_DEPRECATED)) {
776 		ifa_free(ifa);
777 		return (EADDRNOTAVAIL);
778 	}
779 	if (ifa != NULL)
780 		ifa_free(ifa);
781 	INP_INFO_WLOCK(&V_ripcbinfo);
782 	INP_WLOCK(inp);
783 	inp->in6p_laddr = addr->sin6_addr;
784 	INP_WUNLOCK(inp);
785 	INP_INFO_WUNLOCK(&V_ripcbinfo);
786 	return (0);
787 }
788 
789 static int
790 rip6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
791 {
792 	struct inpcb *inp;
793 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)nam;
794 	struct in6_addr in6a;
795 	struct ifnet *ifp = NULL;
796 	int error = 0, scope_ambiguous = 0;
797 
798 	inp = sotoinpcb(so);
799 	KASSERT(inp != NULL, ("rip6_connect: inp == NULL"));
800 
801 	if (nam->sa_len != sizeof(*addr))
802 		return (EINVAL);
803 	if (TAILQ_EMPTY(&V_ifnet))
804 		return (EADDRNOTAVAIL);
805 	if (addr->sin6_family != AF_INET6)
806 		return (EAFNOSUPPORT);
807 
808 	/*
809 	 * Application should provide a proper zone ID or the use of default
810 	 * zone IDs should be enabled.  Unfortunately, some applications do
811 	 * not behave as it should, so we need a workaround.  Even if an
812 	 * appropriate ID is not determined, we'll see if we can determine
813 	 * the outgoing interface.  If we can, determine the zone ID based on
814 	 * the interface below.
815 	 */
816 	if (addr->sin6_scope_id == 0 && !V_ip6_use_defzone)
817 		scope_ambiguous = 1;
818 	if ((error = sa6_embedscope(addr, V_ip6_use_defzone)) != 0)
819 		return (error);
820 
821 	INP_INFO_WLOCK(&V_ripcbinfo);
822 	INP_WLOCK(inp);
823 	/* Source address selection. XXX: need pcblookup? */
824 	error = in6_selectsrc(addr, inp->in6p_outputopts,
825 	    inp, NULL, so->so_cred, &ifp, &in6a);
826 	if (error) {
827 		INP_WUNLOCK(inp);
828 		INP_INFO_WUNLOCK(&V_ripcbinfo);
829 		return (error);
830 	}
831 
832 	/* XXX: see above */
833 	if (ifp && scope_ambiguous &&
834 	    (error = in6_setscope(&addr->sin6_addr, ifp, NULL)) != 0) {
835 		INP_WUNLOCK(inp);
836 		INP_INFO_WUNLOCK(&V_ripcbinfo);
837 		return (error);
838 	}
839 	inp->in6p_faddr = addr->sin6_addr;
840 	inp->in6p_laddr = in6a;
841 	soisconnected(so);
842 	INP_WUNLOCK(inp);
843 	INP_INFO_WUNLOCK(&V_ripcbinfo);
844 	return (0);
845 }
846 
847 static int
848 rip6_shutdown(struct socket *so)
849 {
850 	struct inpcb *inp;
851 
852 	inp = sotoinpcb(so);
853 	KASSERT(inp != NULL, ("rip6_shutdown: inp == NULL"));
854 
855 	INP_WLOCK(inp);
856 	socantsendmore(so);
857 	INP_WUNLOCK(inp);
858 	return (0);
859 }
860 
861 static int
862 rip6_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
863     struct mbuf *control, struct thread *td)
864 {
865 	struct inpcb *inp;
866 	struct sockaddr_in6 tmp;
867 	struct sockaddr_in6 *dst;
868 	int ret;
869 
870 	inp = sotoinpcb(so);
871 	KASSERT(inp != NULL, ("rip6_send: inp == NULL"));
872 
873 	/* Always copy sockaddr to avoid overwrites. */
874 	/* Unlocked read. */
875 	if (so->so_state & SS_ISCONNECTED) {
876 		if (nam) {
877 			m_freem(m);
878 			return (EISCONN);
879 		}
880 		/* XXX */
881 		bzero(&tmp, sizeof(tmp));
882 		tmp.sin6_family = AF_INET6;
883 		tmp.sin6_len = sizeof(struct sockaddr_in6);
884 		INP_RLOCK(inp);
885 		bcopy(&inp->in6p_faddr, &tmp.sin6_addr,
886 		    sizeof(struct in6_addr));
887 		INP_RUNLOCK(inp);
888 		dst = &tmp;
889 	} else {
890 		if (nam == NULL) {
891 			m_freem(m);
892 			return (ENOTCONN);
893 		}
894 		if (nam->sa_len != sizeof(struct sockaddr_in6)) {
895 			m_freem(m);
896 			return (EINVAL);
897 		}
898 		tmp = *(struct sockaddr_in6 *)nam;
899 		dst = &tmp;
900 
901 		if (dst->sin6_family == AF_UNSPEC) {
902 			/*
903 			 * XXX: we allow this case for backward
904 			 * compatibility to buggy applications that
905 			 * rely on old (and wrong) kernel behavior.
906 			 */
907 			log(LOG_INFO, "rip6 SEND: address family is "
908 			    "unspec. Assume AF_INET6\n");
909 			dst->sin6_family = AF_INET6;
910 		} else if (dst->sin6_family != AF_INET6) {
911 			m_freem(m);
912 			return(EAFNOSUPPORT);
913 		}
914 	}
915 	ret = rip6_output(m, so, dst, control);
916 	return (ret);
917 }
918 
919 struct pr_usrreqs rip6_usrreqs = {
920 	.pru_abort =		rip6_abort,
921 	.pru_attach =		rip6_attach,
922 	.pru_bind =		rip6_bind,
923 	.pru_connect =		rip6_connect,
924 	.pru_control =		in6_control,
925 	.pru_detach =		rip6_detach,
926 	.pru_disconnect =	rip6_disconnect,
927 	.pru_peeraddr =		in6_getpeeraddr,
928 	.pru_send =		rip6_send,
929 	.pru_shutdown =		rip6_shutdown,
930 	.pru_sockaddr =		in6_getsockaddr,
931 	.pru_close =		rip6_close,
932 };
933