xref: /freebsd/sys/netinet6/raw_ip6.c (revision 2ed3236082a4473c1da8f72c1ebc071a7b54321f)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *	The Regents of the University of California.
35  * 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. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)raw_ip.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include <sys/cdefs.h>
65 __FBSDID("$FreeBSD$");
66 
67 #include "opt_ipsec.h"
68 #include "opt_inet6.h"
69 #include "opt_route.h"
70 
71 #include <sys/param.h>
72 #include <sys/errno.h>
73 #include <sys/jail.h>
74 #include <sys/kernel.h>
75 #include <sys/lock.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/signalvar.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/sx.h>
85 #include <sys/syslog.h>
86 
87 #include <net/if.h>
88 #include <net/if_var.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 #include <net/vnet.h>
92 
93 #include <netinet/in.h>
94 #include <netinet/in_var.h>
95 #include <netinet/in_systm.h>
96 #include <netinet/in_pcb.h>
97 
98 #include <netinet/icmp6.h>
99 #include <netinet/ip6.h>
100 #include <netinet/ip_var.h>
101 #include <netinet6/ip6protosw.h>
102 #include <netinet6/ip6_mroute.h>
103 #include <netinet6/in6_pcb.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/nd6.h>
106 #include <netinet6/raw_ip6.h>
107 #include <netinet6/in6_fib.h>
108 #include <netinet6/scope6_var.h>
109 #include <netinet6/send.h>
110 
111 #include <netipsec/ipsec_support.h>
112 
113 #include <machine/stdarg.h>
114 
115 #define	satosin6(sa)	((struct sockaddr_in6 *)(sa))
116 #define	ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
117 
118 /*
119  * Raw interface to IP6 protocol.
120  */
121 
122 VNET_DECLARE(struct inpcbhead, ripcb);
123 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
124 #define	V_ripcb				VNET(ripcb)
125 #define	V_ripcbinfo			VNET(ripcbinfo)
126 
127 extern u_long	rip_sendspace;
128 extern u_long	rip_recvspace;
129 
130 VNET_PCPUSTAT_DEFINE(struct rip6stat, rip6stat);
131 VNET_PCPUSTAT_SYSINIT(rip6stat);
132 
133 #ifdef VIMAGE
134 VNET_PCPUSTAT_SYSUNINIT(rip6stat);
135 #endif /* VIMAGE */
136 
137 /*
138  * Hooks for multicast routing. They all default to NULL, so leave them not
139  * initialized and rely on BSS being set to 0.
140  */
141 
142 /*
143  * The socket used to communicate with the multicast routing daemon.
144  */
145 VNET_DEFINE(struct socket *, ip6_mrouter);
146 
147 /*
148  * The various mrouter functions.
149  */
150 int (*ip6_mrouter_set)(struct socket *, struct sockopt *);
151 int (*ip6_mrouter_get)(struct socket *, struct sockopt *);
152 int (*ip6_mrouter_done)(void);
153 int (*ip6_mforward)(struct ip6_hdr *, struct ifnet *, struct mbuf *);
154 int (*mrt6_ioctl)(u_long, caddr_t);
155 
156 /*
157  * Setup generic address and protocol structures for raw_input routine, then
158  * pass them along with mbuf chain.
159  */
160 int
161 rip6_input(struct mbuf **mp, int *offp, int proto)
162 {
163 	struct ifnet *ifp;
164 	struct mbuf *m = *mp;
165 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
166 	struct inpcb *inp;
167 	struct inpcb *last = NULL;
168 	struct mbuf *opts = NULL;
169 	struct sockaddr_in6 fromsa;
170 
171 	NET_EPOCH_ASSERT();
172 
173 	RIP6STAT_INC(rip6s_ipackets);
174 
175 	init_sin6(&fromsa, m, 0); /* general init */
176 
177 	ifp = m->m_pkthdr.rcvif;
178 
179 	CK_LIST_FOREACH(inp, &V_ripcb, inp_list) {
180 		/* XXX inp locking */
181 		if ((inp->inp_vflag & INP_IPV6) == 0)
182 			continue;
183 		if (inp->inp_ip_p &&
184 		    inp->inp_ip_p != proto)
185 			continue;
186 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
187 		    !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
188 			continue;
189 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
190 		    !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
191 			continue;
192 		if (last != NULL) {
193 			struct mbuf *n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
194 
195 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
196 			/*
197 			 * Check AH/ESP integrity.
198 			 */
199 			if (IPSEC_ENABLED(ipv6)) {
200 				if (n != NULL &&
201 				    IPSEC_CHECK_POLICY(ipv6, n, last) != 0) {
202 					m_freem(n);
203 					/* Do not inject data into pcb. */
204 					n = NULL;
205 				}
206 			}
207 #endif /* IPSEC */
208 			if (n) {
209 				if (last->inp_flags & INP_CONTROLOPTS ||
210 				    last->inp_socket->so_options & SO_TIMESTAMP)
211 					ip6_savecontrol(last, n, &opts);
212 				/* strip intermediate headers */
213 				m_adj(n, *offp);
214 				if (sbappendaddr(&last->inp_socket->so_rcv,
215 						(struct sockaddr *)&fromsa,
216 						 n, opts) == 0) {
217 					m_freem(n);
218 					if (opts)
219 						m_freem(opts);
220 					RIP6STAT_INC(rip6s_fullsock);
221 				} else
222 					sorwakeup(last->inp_socket);
223 				opts = NULL;
224 			}
225 			INP_RUNLOCK(last);
226 			last = NULL;
227 		}
228 		INP_RLOCK(inp);
229 		if (__predict_false(inp->inp_flags2 & INP_FREED))
230 			goto skip_2;
231 		if (jailed_without_vnet(inp->inp_cred)) {
232 			/*
233 			 * Allow raw socket in jail to receive multicast;
234 			 * assume process had PRIV_NETINET_RAW at attach,
235 			 * and fall through into normal filter path if so.
236 			 */
237 			if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
238 			    prison_check_ip6(inp->inp_cred,
239 			    &ip6->ip6_dst) != 0)
240 				goto skip_2;
241 		}
242 		if (inp->in6p_cksum != -1) {
243 			RIP6STAT_INC(rip6s_isum);
244 			if (m->m_pkthdr.len - (*offp + inp->in6p_cksum) < 2 ||
245 			    in6_cksum(m, proto, *offp,
246 			    m->m_pkthdr.len - *offp)) {
247 				RIP6STAT_INC(rip6s_badsum);
248 				/*
249 				 * Drop the received message, don't send an
250 				 * ICMP6 message. Set proto to IPPROTO_NONE
251 				 * to achieve that.
252 				 */
253 				proto = IPPROTO_NONE;
254 				goto skip_2;
255 			}
256 		}
257 		/*
258 		 * If this raw socket has multicast state, and we
259 		 * have received a multicast, check if this socket
260 		 * should receive it, as multicast filtering is now
261 		 * the responsibility of the transport layer.
262 		 */
263 		if (inp->in6p_moptions &&
264 		    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
265 			/*
266 			 * If the incoming datagram is for MLD, allow it
267 			 * through unconditionally to the raw socket.
268 			 *
269 			 * Use the M_RTALERT_MLD flag to check for MLD
270 			 * traffic without having to inspect the mbuf chain
271 			 * more deeply, as all MLDv1/v2 host messages MUST
272 			 * contain the Router Alert option.
273 			 *
274 			 * In the case of MLDv1, we may not have explicitly
275 			 * joined the group, and may have set IFF_ALLMULTI
276 			 * on the interface. im6o_mc_filter() may discard
277 			 * control traffic we actually need to see.
278 			 *
279 			 * Userland multicast routing daemons should continue
280 			 * filter the control traffic appropriately.
281 			 */
282 			int blocked;
283 
284 			blocked = MCAST_PASS;
285 			if ((m->m_flags & M_RTALERT_MLD) == 0) {
286 				struct sockaddr_in6 mcaddr;
287 
288 				bzero(&mcaddr, sizeof(struct sockaddr_in6));
289 				mcaddr.sin6_len = sizeof(struct sockaddr_in6);
290 				mcaddr.sin6_family = AF_INET6;
291 				mcaddr.sin6_addr = ip6->ip6_dst;
292 
293 				blocked = im6o_mc_filter(inp->in6p_moptions,
294 				    ifp,
295 				    (struct sockaddr *)&mcaddr,
296 				    (struct sockaddr *)&fromsa);
297 			}
298 			if (blocked != MCAST_PASS) {
299 				IP6STAT_INC(ip6s_notmember);
300 				goto skip_2;
301 			}
302 		}
303 		last = inp;
304 		continue;
305 skip_2:
306 		INP_RUNLOCK(inp);
307 	}
308 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
309 	/*
310 	 * Check AH/ESP integrity.
311 	 */
312 	if (IPSEC_ENABLED(ipv6) && last != NULL &&
313 	    IPSEC_CHECK_POLICY(ipv6, m, last) != 0) {
314 		m_freem(m);
315 		IP6STAT_DEC(ip6s_delivered);
316 		/* Do not inject data into pcb. */
317 		INP_RUNLOCK(last);
318 	} else
319 #endif /* IPSEC */
320 	if (last != NULL) {
321 		if (last->inp_flags & INP_CONTROLOPTS ||
322 		    last->inp_socket->so_options & SO_TIMESTAMP)
323 			ip6_savecontrol(last, m, &opts);
324 		/* Strip intermediate headers. */
325 		m_adj(m, *offp);
326 		if (sbappendaddr(&last->inp_socket->so_rcv,
327 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
328 			m_freem(m);
329 			if (opts)
330 				m_freem(opts);
331 			RIP6STAT_INC(rip6s_fullsock);
332 		} else
333 			sorwakeup(last->inp_socket);
334 		INP_RUNLOCK(last);
335 	} else {
336 		RIP6STAT_INC(rip6s_nosock);
337 		if (m->m_flags & M_MCAST)
338 			RIP6STAT_INC(rip6s_nosockmcast);
339 		if (proto == IPPROTO_NONE)
340 			m_freem(m);
341 		else
342 			icmp6_error(m, ICMP6_PARAM_PROB,
343 			    ICMP6_PARAMPROB_NEXTHEADER,
344 			    ip6_get_prevhdr(m, *offp));
345 		IP6STAT_DEC(ip6s_delivered);
346 	}
347 	return (IPPROTO_DONE);
348 }
349 
350 void
351 rip6_ctlinput(int cmd, struct sockaddr *sa, void *d)
352 {
353 	struct ip6ctlparam *ip6cp = NULL;
354 	const struct sockaddr_in6 *sa6_src = NULL;
355 	void *cmdarg;
356 	struct inpcb *(*notify)(struct inpcb *, int) = in6_rtchange;
357 
358 	if (sa->sa_family != AF_INET6 ||
359 	    sa->sa_len != sizeof(struct sockaddr_in6))
360 		return;
361 
362 	if ((unsigned)cmd >= PRC_NCMDS)
363 		return;
364 	if (PRC_IS_REDIRECT(cmd))
365 		notify = in6_rtchange, d = NULL;
366 	else if (cmd == PRC_HOSTDEAD)
367 		d = NULL;
368 	else if (inet6ctlerrmap[cmd] == 0)
369 		return;
370 
371 	/*
372 	 * If the parameter is from icmp6, decode it.
373 	 */
374 	if (d != NULL) {
375 		ip6cp = (struct ip6ctlparam *)d;
376 		cmdarg = ip6cp->ip6c_cmdarg;
377 		sa6_src = ip6cp->ip6c_src;
378 	} else {
379 		cmdarg = NULL;
380 		sa6_src = &sa6_any;
381 	}
382 
383 	(void) in6_pcbnotify(&V_ripcbinfo, sa, 0,
384 	    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
385 }
386 
387 /*
388  * Generate IPv6 header and pass packet to ip6_output.  Tack on options user
389  * may have setup with control call.
390  */
391 int
392 rip6_output(struct mbuf *m, struct socket *so, ...)
393 {
394 	struct epoch_tracker et;
395 	struct mbuf *control;
396 	struct m_tag *mtag;
397 	struct sockaddr_in6 *dstsock;
398 	struct ip6_hdr *ip6;
399 	struct inpcb *inp;
400 	u_int	plen = m->m_pkthdr.len;
401 	int error = 0;
402 	struct ip6_pktopts opt, *optp;
403 	struct ifnet *oifp = NULL;
404 	int type = 0, code = 0;		/* for ICMPv6 output statistics only */
405 	int scope_ambiguous = 0;
406 	int use_defzone = 0;
407 	int hlim = 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 	inp = sotoinpcb(so);
417 	INP_WLOCK(inp);
418 
419 	if (control != NULL) {
420 		if ((error = ip6_setpktopts(control, &opt,
421 		    inp->in6p_outputopts, so->so_cred,
422 		    so->so_proto->pr_protocol)) != 0) {
423 			goto bad;
424 		}
425 		optp = &opt;
426 	} else
427 		optp = inp->in6p_outputopts;
428 
429 	/*
430 	 * Check and convert scope zone ID into internal form.
431 	 *
432 	 * XXX: we may still need to determine the zone later.
433 	 */
434 	if (!(so->so_state & SS_ISCONNECTED)) {
435 		if (!optp || !optp->ip6po_pktinfo ||
436 		    !optp->ip6po_pktinfo->ipi6_ifindex)
437 			use_defzone = V_ip6_use_defzone;
438 		if (dstsock->sin6_scope_id == 0 && !use_defzone)
439 			scope_ambiguous = 1;
440 		if ((error = sa6_embedscope(dstsock, use_defzone)) != 0)
441 			goto bad;
442 	}
443 
444 	/*
445 	 * For an ICMPv6 packet, we should know its type and code to update
446 	 * statistics.
447 	 */
448 	if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
449 		struct icmp6_hdr *icmp6;
450 		if (m->m_len < sizeof(struct icmp6_hdr) &&
451 		    (m = m_pullup(m, sizeof(struct icmp6_hdr))) == NULL) {
452 			error = ENOBUFS;
453 			goto bad;
454 		}
455 		icmp6 = mtod(m, struct icmp6_hdr *);
456 		type = icmp6->icmp6_type;
457 		code = icmp6->icmp6_code;
458 	}
459 
460 	M_PREPEND(m, sizeof(*ip6), M_NOWAIT);
461 	if (m == NULL) {
462 		error = ENOBUFS;
463 		goto bad;
464 	}
465 	ip6 = mtod(m, struct ip6_hdr *);
466 
467 #ifdef ROUTE_MPATH
468 	if (CALC_FLOWID_OUTBOUND) {
469 		uint32_t hash_type, hash_val;
470 
471 		hash_val = fib6_calc_software_hash(&inp->in6p_laddr,
472 		    &dstsock->sin6_addr, 0, 0, so->so_proto->pr_protocol,
473 		    &hash_type);
474 		inp->inp_flowid = hash_val;
475 		inp->inp_flowtype = hash_type;
476 	}
477 #endif
478 	/*
479 	 * Source address selection.
480 	 */
481 	error = in6_selectsrc_socket(dstsock, optp, inp, so->so_cred,
482 	    scope_ambiguous, &in6a, &hlim);
483 
484 	if (error)
485 		goto bad;
486 	error = prison_check_ip6(inp->inp_cred, &in6a);
487 	if (error != 0)
488 		goto bad;
489 	ip6->ip6_src = in6a;
490 
491 	ip6->ip6_dst = dstsock->sin6_addr;
492 
493 	/*
494 	 * Fill in the rest of the IPv6 header fields.
495 	 */
496 	ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) |
497 	    (inp->inp_flow & IPV6_FLOWINFO_MASK);
498 	ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) |
499 	    (IPV6_VERSION & IPV6_VERSION_MASK);
500 
501 	/*
502 	 * ip6_plen will be filled in ip6_output, so not fill it here.
503 	 */
504 	ip6->ip6_nxt = inp->inp_ip_p;
505 	ip6->ip6_hlim = hlim;
506 
507 	if (so->so_proto->pr_protocol == IPPROTO_ICMPV6 ||
508 	    inp->in6p_cksum != -1) {
509 		struct mbuf *n;
510 		int off;
511 		u_int16_t *p;
512 
513 		/* Compute checksum. */
514 		if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
515 			off = offsetof(struct icmp6_hdr, icmp6_cksum);
516 		else
517 			off = inp->in6p_cksum;
518 		if (plen < off + 2) {
519 			error = EINVAL;
520 			goto bad;
521 		}
522 		off += sizeof(struct ip6_hdr);
523 
524 		n = m;
525 		while (n && n->m_len <= off) {
526 			off -= n->m_len;
527 			n = n->m_next;
528 		}
529 		if (!n)
530 			goto bad;
531 		p = (u_int16_t *)(mtod(n, caddr_t) + off);
532 		*p = 0;
533 		*p = in6_cksum(m, ip6->ip6_nxt, sizeof(*ip6), plen);
534 	}
535 
536 	/*
537 	 * Send RA/RS messages to user land for protection, before sending
538 	 * them to rtadvd/rtsol.
539 	 */
540 	if ((send_sendso_input_hook != NULL) &&
541 	    so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
542 		switch (type) {
543 		case ND_ROUTER_ADVERT:
544 		case ND_ROUTER_SOLICIT:
545 			mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
546 				sizeof(unsigned short), M_NOWAIT);
547 			if (mtag == NULL)
548 				goto bad;
549 			m_tag_prepend(m, mtag);
550 		}
551 	}
552 
553 	NET_EPOCH_ENTER(et);
554 	error = ip6_output(m, optp, NULL, 0, inp->in6p_moptions, &oifp, inp);
555 	NET_EPOCH_EXIT(et);
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(inp);
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 epoch_tracker et;
753 	struct inpcb *inp;
754 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)nam;
755 	struct ifaddr *ifa = NULL;
756 	int error = 0;
757 
758 	inp = sotoinpcb(so);
759 	KASSERT(inp != NULL, ("rip6_bind: inp == NULL"));
760 
761 	if (nam->sa_len != sizeof(*addr))
762 		return (EINVAL);
763 	if ((error = prison_check_ip6(td->td_ucred, &addr->sin6_addr)) != 0)
764 		return (error);
765 	if (CK_STAILQ_EMPTY(&V_ifnet) || addr->sin6_family != AF_INET6)
766 		return (EADDRNOTAVAIL);
767 	if ((error = sa6_embedscope(addr, V_ip6_use_defzone)) != 0)
768 		return (error);
769 
770 	NET_EPOCH_ENTER(et);
771 	if (!IN6_IS_ADDR_UNSPECIFIED(&addr->sin6_addr) &&
772 	    (ifa = ifa_ifwithaddr((struct sockaddr *)addr)) == NULL) {
773 		NET_EPOCH_EXIT(et);
774 		return (EADDRNOTAVAIL);
775 	}
776 	if (ifa != NULL &&
777 	    ((struct in6_ifaddr *)ifa)->ia6_flags &
778 	    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|
779 	     IN6_IFF_DETACHED|IN6_IFF_DEPRECATED)) {
780 		NET_EPOCH_EXIT(et);
781 		return (EADDRNOTAVAIL);
782 	}
783 	NET_EPOCH_EXIT(et);
784 	INP_INFO_WLOCK(&V_ripcbinfo);
785 	INP_WLOCK(inp);
786 	inp->in6p_laddr = addr->sin6_addr;
787 	INP_WUNLOCK(inp);
788 	INP_INFO_WUNLOCK(&V_ripcbinfo);
789 	return (0);
790 }
791 
792 static int
793 rip6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
794 {
795 	struct inpcb *inp;
796 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)nam;
797 	struct in6_addr in6a;
798 	int error = 0, scope_ambiguous = 0;
799 
800 	inp = sotoinpcb(so);
801 	KASSERT(inp != NULL, ("rip6_connect: inp == NULL"));
802 
803 	if (nam->sa_len != sizeof(*addr))
804 		return (EINVAL);
805 	if (CK_STAILQ_EMPTY(&V_ifnet))
806 		return (EADDRNOTAVAIL);
807 	if (addr->sin6_family != AF_INET6)
808 		return (EAFNOSUPPORT);
809 
810 	/*
811 	 * Application should provide a proper zone ID or the use of default
812 	 * zone IDs should be enabled.  Unfortunately, some applications do
813 	 * not behave as it should, so we need a workaround.  Even if an
814 	 * appropriate ID is not determined, we'll see if we can determine
815 	 * the outgoing interface.  If we can, determine the zone ID based on
816 	 * the interface below.
817 	 */
818 	if (addr->sin6_scope_id == 0 && !V_ip6_use_defzone)
819 		scope_ambiguous = 1;
820 	if ((error = sa6_embedscope(addr, V_ip6_use_defzone)) != 0)
821 		return (error);
822 
823 	INP_INFO_WLOCK(&V_ripcbinfo);
824 	INP_WLOCK(inp);
825 	/* Source address selection. XXX: need pcblookup? */
826 	error = in6_selectsrc_socket(addr, inp->in6p_outputopts,
827 	    inp, so->so_cred, scope_ambiguous, &in6a, NULL);
828 	if (error) {
829 		INP_WUNLOCK(inp);
830 		INP_INFO_WUNLOCK(&V_ripcbinfo);
831 		return (error);
832 	}
833 
834 	inp->in6p_faddr = addr->sin6_addr;
835 	inp->in6p_laddr = in6a;
836 	soisconnected(so);
837 	INP_WUNLOCK(inp);
838 	INP_INFO_WUNLOCK(&V_ripcbinfo);
839 	return (0);
840 }
841 
842 static int
843 rip6_shutdown(struct socket *so)
844 {
845 	struct inpcb *inp;
846 
847 	inp = sotoinpcb(so);
848 	KASSERT(inp != NULL, ("rip6_shutdown: inp == NULL"));
849 
850 	INP_WLOCK(inp);
851 	socantsendmore(so);
852 	INP_WUNLOCK(inp);
853 	return (0);
854 }
855 
856 static int
857 rip6_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
858     struct mbuf *control, struct thread *td)
859 {
860 	struct inpcb *inp;
861 	struct sockaddr_in6 tmp;
862 	struct sockaddr_in6 *dst;
863 	int ret;
864 
865 	inp = sotoinpcb(so);
866 	KASSERT(inp != NULL, ("rip6_send: inp == NULL"));
867 
868 	/* Always copy sockaddr to avoid overwrites. */
869 	/* Unlocked read. */
870 	if (so->so_state & SS_ISCONNECTED) {
871 		if (nam) {
872 			m_freem(m);
873 			return (EISCONN);
874 		}
875 		/* XXX */
876 		bzero(&tmp, sizeof(tmp));
877 		tmp.sin6_family = AF_INET6;
878 		tmp.sin6_len = sizeof(struct sockaddr_in6);
879 		INP_RLOCK(inp);
880 		bcopy(&inp->in6p_faddr, &tmp.sin6_addr,
881 		    sizeof(struct in6_addr));
882 		INP_RUNLOCK(inp);
883 		dst = &tmp;
884 	} else {
885 		if (nam == NULL) {
886 			m_freem(m);
887 			return (ENOTCONN);
888 		}
889 		if (nam->sa_len != sizeof(struct sockaddr_in6)) {
890 			m_freem(m);
891 			return (EINVAL);
892 		}
893 		tmp = *(struct sockaddr_in6 *)nam;
894 		dst = &tmp;
895 
896 		if (dst->sin6_family == AF_UNSPEC) {
897 			/*
898 			 * XXX: we allow this case for backward
899 			 * compatibility to buggy applications that
900 			 * rely on old (and wrong) kernel behavior.
901 			 */
902 			log(LOG_INFO, "rip6 SEND: address family is "
903 			    "unspec. Assume AF_INET6\n");
904 			dst->sin6_family = AF_INET6;
905 		} else if (dst->sin6_family != AF_INET6) {
906 			m_freem(m);
907 			return(EAFNOSUPPORT);
908 		}
909 	}
910 	ret = rip6_output(m, so, dst, control);
911 	return (ret);
912 }
913 
914 struct pr_usrreqs rip6_usrreqs = {
915 	.pru_abort =		rip6_abort,
916 	.pru_attach =		rip6_attach,
917 	.pru_bind =		rip6_bind,
918 	.pru_connect =		rip6_connect,
919 	.pru_control =		in6_control,
920 	.pru_detach =		rip6_detach,
921 	.pru_disconnect =	rip6_disconnect,
922 	.pru_peeraddr =		in6_getpeeraddr,
923 	.pru_send =		rip6_send,
924 	.pru_shutdown =		rip6_shutdown,
925 	.pru_sockaddr =		in6_getsockaddr,
926 	.pru_close =		rip6_close,
927 };
928