xref: /netbsd-src/sys/netinet6/ip6_output.c (revision f09779b4fc4eea60ed5419ec87a52bdc897fdb73)
1 /*	$NetBSD: ip6_output.c,v 1.235 2024/04/19 00:55:35 riastradh Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 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, 1988, 1990, 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. 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  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.235 2024/04/19 00:55:35 riastradh Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #endif
72 
73 #include <sys/param.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/errno.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/syslog.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/kauth.h>
83 
84 #include <net/if.h>
85 #include <net/route.h>
86 #include <net/pfil.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet/ip_var.h>
92 #include <netinet/icmp6.h>
93 #include <netinet/in_offload.h>
94 #include <netinet/portalgo.h>
95 #include <netinet6/in6_offload.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet6/ip6_private.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/ip6protosw.h>
101 #include <netinet6/scope6_var.h>
102 
103 #ifdef IPSEC
104 #include <netipsec/ipsec.h>
105 #include <netipsec/ipsec6.h>
106 #include <netipsec/key.h>
107 #endif
108 
109 extern pfil_head_t *inet6_pfil_hook;	/* XXX */
110 
111 struct ip6_exthdrs {
112 	struct mbuf *ip6e_ip6;
113 	struct mbuf *ip6e_hbh;
114 	struct mbuf *ip6e_dest1;
115 	struct mbuf *ip6e_rthdr;
116 	struct mbuf *ip6e_dest2;
117 };
118 
119 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
120 	kauth_cred_t, int);
121 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
122 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
123 	int, int, int);
124 static int ip6_setmoptions(const struct sockopt *, struct inpcb *);
125 static int ip6_getmoptions(struct sockopt *, struct inpcb *);
126 static int ip6_copyexthdr(struct mbuf **, void *, int);
127 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
128 	struct ip6_frag **);
129 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
130 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
131 static int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *, int *);
132 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
133 static int ip6_ifaddrvalid(const struct in6_addr *, const struct in6_addr *);
134 static int ip6_handle_rthdr(struct ip6_rthdr *, struct ip6_hdr *);
135 
136 #ifdef RFC2292
137 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
138 #endif
139 
140 static int
ip6_handle_rthdr(struct ip6_rthdr * rh,struct ip6_hdr * ip6)141 ip6_handle_rthdr(struct ip6_rthdr *rh, struct ip6_hdr *ip6)
142 {
143 	int error = 0;
144 
145 	switch (rh->ip6r_type) {
146 	case IPV6_RTHDR_TYPE_0:
147 		/* Dropped, RFC5095. */
148 	default:	/* is it possible? */
149 		error = EINVAL;
150 	}
151 
152 	return error;
153 }
154 
155 /*
156  * Send an IP packet to a host.
157  */
158 int
ip6_if_output(struct ifnet * const ifp,struct ifnet * const origifp,struct mbuf * const m,const struct sockaddr_in6 * const dst,const struct rtentry * rt)159 ip6_if_output(struct ifnet * const ifp, struct ifnet * const origifp,
160     struct mbuf * const m, const struct sockaddr_in6 * const dst,
161     const struct rtentry *rt)
162 {
163 	int error = 0;
164 
165 	if (rt != NULL) {
166 		error = rt_check_reject_route(rt, ifp);
167 		if (error != 0) {
168 			IP6_STATINC(IP6_STAT_RTREJECT);
169 			m_freem(m);
170 			return error;
171 		}
172 	}
173 
174 	/* discard the packet if IPv6 operation is disabled on the interface */
175 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
176 		m_freem(m);
177 		return ENETDOWN; /* better error? */
178 	}
179 
180 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
181 		error = if_output_lock(ifp, origifp, m, sin6tocsa(dst), rt);
182 	else
183 		error = if_output_lock(ifp, ifp, m, sin6tocsa(dst), rt);
184 	return error;
185 }
186 
187 /*
188  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
189  * header (with pri, len, nxt, hlim, src, dst).
190  *
191  * This function may modify ver and hlim only. The mbuf chain containing the
192  * packet will be freed. The mbuf opt, if present, will not be freed.
193  *
194  * Type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
195  * nd_ifinfo.linkmtu is u_int32_t. So we use u_long to hold largest one,
196  * which is rt_rmx.rmx_mtu.
197  */
198 int
ip6_output(struct mbuf * m0,struct ip6_pktopts * opt,struct route * ro,int flags,struct ip6_moptions * im6o,struct inpcb * inp,struct ifnet ** ifpp)199 ip6_output(
200     struct mbuf *m0,
201     struct ip6_pktopts *opt,
202     struct route *ro,
203     int flags,
204     struct ip6_moptions *im6o,
205     struct inpcb *inp,
206     struct ifnet **ifpp		/* XXX: just for statistics */
207 )
208 {
209 	struct ip6_hdr *ip6, *mhip6;
210 	struct ifnet *ifp = NULL, *origifp = NULL;
211 	struct mbuf *m = m0;
212 	int tlen, len, off;
213 	bool tso;
214 	struct route ip6route;
215 	struct rtentry *rt = NULL, *rt_pmtu;
216 	const struct sockaddr_in6 *dst;
217 	struct sockaddr_in6 src_sa, dst_sa;
218 	int error = 0;
219 	struct in6_ifaddr *ia = NULL;
220 	u_long mtu;
221 	int alwaysfrag, dontfrag;
222 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
223 	struct ip6_exthdrs exthdrs;
224 	struct in6_addr finaldst, src0, dst0;
225 	u_int32_t zone;
226 	struct route *ro_pmtu = NULL;
227 	int hdrsplit = 0;
228 	int needipsec = 0;
229 #ifdef IPSEC
230 	struct secpolicy *sp = NULL;
231 #endif
232 	struct psref psref, psref_ia;
233 	int bound = curlwp_bind();
234 	bool release_psref_ia = false;
235 
236 #ifdef DIAGNOSTIC
237 	if ((m->m_flags & M_PKTHDR) == 0)
238 		panic("ip6_output: no HDR");
239 	if ((m->m_pkthdr.csum_flags &
240 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
241 		panic("ip6_output: IPv4 checksum offload flags: %d",
242 		    m->m_pkthdr.csum_flags);
243 	}
244 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
245 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
246 		panic("ip6_output: conflicting checksum offload flags: %d",
247 		    m->m_pkthdr.csum_flags);
248 	}
249 #endif
250 
251 	M_CSUM_DATA_IPv6_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
252 
253 #define MAKE_EXTHDR(hp, mp)						\
254     do {								\
255 	if (hp) {							\
256 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
257 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
258 		    ((eh)->ip6e_len + 1) << 3);				\
259 		if (error)						\
260 			goto freehdrs;					\
261 	}								\
262     } while (/*CONSTCOND*/ 0)
263 
264 	memset(&exthdrs, 0, sizeof(exthdrs));
265 	if (opt) {
266 		/* Hop-by-Hop options header */
267 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
268 		/* Destination options header (1st part) */
269 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
270 		/* Routing header */
271 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
272 		/* Destination options header (2nd part) */
273 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
274 	}
275 
276 	/*
277 	 * Calculate the total length of the extension header chain.
278 	 * Keep the length of the unfragmentable part for fragmentation.
279 	 */
280 	optlen = 0;
281 	if (exthdrs.ip6e_hbh)
282 		optlen += exthdrs.ip6e_hbh->m_len;
283 	if (exthdrs.ip6e_dest1)
284 		optlen += exthdrs.ip6e_dest1->m_len;
285 	if (exthdrs.ip6e_rthdr)
286 		optlen += exthdrs.ip6e_rthdr->m_len;
287 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
288 	/* NOTE: we don't add AH/ESP length here. do that later. */
289 	if (exthdrs.ip6e_dest2)
290 		optlen += exthdrs.ip6e_dest2->m_len;
291 
292 #ifdef IPSEC
293 	if (ipsec_used) {
294 		/* Check the security policy (SP) for the packet */
295 		sp = ipsec6_check_policy(m, inp, flags, &needipsec, &error);
296 		if (error != 0) {
297 			/*
298 			 * Hack: -EINVAL is used to signal that a packet
299 			 * should be silently discarded.  This is typically
300 			 * because we asked key management for an SA and
301 			 * it was delayed (e.g. kicked up to IKE).
302 			 */
303 			if (error == -EINVAL)
304 				error = 0;
305 			IP6_STATINC(IP6_STAT_IPSECDROP_OUT);
306 			goto freehdrs;
307 		}
308 	}
309 #endif
310 
311 	if (needipsec &&
312 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
313 		in6_undefer_cksum_tcpudp(m);
314 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
315 	}
316 
317 	/*
318 	 * If we need IPsec, or there is at least one extension header,
319 	 * separate IP6 header from the payload.
320 	 */
321 	if ((needipsec || optlen) && !hdrsplit) {
322 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
323 			IP6_STATINC(IP6_STAT_ODROPPED);
324 			m = NULL;
325 			goto freehdrs;
326 		}
327 		m = exthdrs.ip6e_ip6;
328 		hdrsplit++;
329 	}
330 
331 	/* adjust pointer */
332 	ip6 = mtod(m, struct ip6_hdr *);
333 
334 	/* adjust mbuf packet header length */
335 	m->m_pkthdr.len += optlen;
336 	plen = m->m_pkthdr.len - sizeof(*ip6);
337 
338 	/* If this is a jumbo payload, insert a jumbo payload option. */
339 	if (plen > IPV6_MAXPACKET) {
340 		if (!hdrsplit) {
341 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
342 				IP6_STATINC(IP6_STAT_ODROPPED);
343 				m = NULL;
344 				goto freehdrs;
345 			}
346 			m = exthdrs.ip6e_ip6;
347 			hdrsplit++;
348 		}
349 		/* adjust pointer */
350 		ip6 = mtod(m, struct ip6_hdr *);
351 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) {
352 			IP6_STATINC(IP6_STAT_ODROPPED);
353 			goto freehdrs;
354 		}
355 		optlen += 8; /* XXX JUMBOOPTLEN */
356 		ip6->ip6_plen = 0;
357 	} else
358 		ip6->ip6_plen = htons(plen);
359 
360 	/*
361 	 * Concatenate headers and fill in next header fields.
362 	 * Here we have, on "m"
363 	 *	IPv6 payload
364 	 * and we insert headers accordingly.  Finally, we should be getting:
365 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
366 	 *
367 	 * during the header composing process, "m" points to IPv6 header.
368 	 * "mprev" points to an extension header prior to esp.
369 	 */
370 	{
371 		u_char *nexthdrp = &ip6->ip6_nxt;
372 		struct mbuf *mprev = m;
373 
374 		/*
375 		 * we treat dest2 specially.  this makes IPsec processing
376 		 * much easier.  the goal here is to make mprev point the
377 		 * mbuf prior to dest2.
378 		 *
379 		 * result: IPv6 dest2 payload
380 		 * m and mprev will point to IPv6 header.
381 		 */
382 		if (exthdrs.ip6e_dest2) {
383 			if (!hdrsplit)
384 				panic("assumption failed: hdr not split");
385 			exthdrs.ip6e_dest2->m_next = m->m_next;
386 			m->m_next = exthdrs.ip6e_dest2;
387 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
388 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
389 		}
390 
391 #define MAKE_CHAIN(m, mp, p, i)\
392     do {\
393 	if (m) {\
394 		if (!hdrsplit) \
395 			panic("assumption failed: hdr not split"); \
396 		*mtod((m), u_char *) = *(p);\
397 		*(p) = (i);\
398 		p = mtod((m), u_char *);\
399 		(m)->m_next = (mp)->m_next;\
400 		(mp)->m_next = (m);\
401 		(mp) = (m);\
402 	}\
403     } while (/*CONSTCOND*/ 0)
404 		/*
405 		 * result: IPv6 hbh dest1 rthdr dest2 payload
406 		 * m will point to IPv6 header.  mprev will point to the
407 		 * extension header prior to dest2 (rthdr in the above case).
408 		 */
409 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
410 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
411 		    IPPROTO_DSTOPTS);
412 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
413 		    IPPROTO_ROUTING);
414 
415 		M_CSUM_DATA_IPv6_SET(m->m_pkthdr.csum_data,
416 		    sizeof(struct ip6_hdr) + optlen);
417 	}
418 
419 	/* Need to save for pmtu */
420 	finaldst = ip6->ip6_dst;
421 
422 	/*
423 	 * If there is a routing header, replace destination address field
424 	 * with the first hop of the routing header.
425 	 */
426 	if (exthdrs.ip6e_rthdr) {
427 		struct ip6_rthdr *rh;
428 
429 		rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
430 
431 		error = ip6_handle_rthdr(rh, ip6);
432 		if (error != 0) {
433 			IP6_STATINC(IP6_STAT_ODROPPED);
434 			goto bad;
435 		}
436 	}
437 
438 	/* Source address validation */
439 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
440 	    (flags & IPV6_UNSPECSRC) == 0) {
441 		error = EOPNOTSUPP;
442 		IP6_STATINC(IP6_STAT_BADSCOPE);
443 		goto bad;
444 	}
445 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
446 		error = EOPNOTSUPP;
447 		IP6_STATINC(IP6_STAT_BADSCOPE);
448 		goto bad;
449 	}
450 
451 	IP6_STATINC(IP6_STAT_LOCALOUT);
452 
453 	/*
454 	 * Route packet.
455 	 */
456 	/* initialize cached route */
457 	if (ro == NULL) {
458 		memset(&ip6route, 0, sizeof(ip6route));
459 		ro = &ip6route;
460 	}
461 	ro_pmtu = ro;
462 	if (opt && opt->ip6po_rthdr)
463 		ro = &opt->ip6po_route;
464 
465 	/*
466 	 * if specified, try to fill in the traffic class field.
467 	 * do not override if a non-zero value is already set.
468 	 * we check the diffserv field and the ecn field separately.
469 	 */
470 	if (opt && opt->ip6po_tclass >= 0) {
471 		int mask = 0;
472 
473 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
474 			mask |= 0xfc;
475 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
476 			mask |= 0x03;
477 		if (mask != 0)
478 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
479 	}
480 
481 	/* fill in or override the hop limit field, if necessary. */
482 	if (opt && opt->ip6po_hlim != -1)
483 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
484 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
485 		if (im6o != NULL)
486 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
487 		else
488 			ip6->ip6_hlim = ip6_defmcasthlim;
489 	}
490 
491 #ifdef IPSEC
492 	if (needipsec) {
493 		error = ipsec6_process_packet(m, sp->req, flags);
494 
495 		/*
496 		 * Preserve KAME behaviour: ENOENT can be returned
497 		 * when an SA acquire is in progress.  Don't propagate
498 		 * this to user-level; it confuses applications.
499 		 * XXX this will go away when the SADB is redone.
500 		 */
501 		if (error == ENOENT)
502 			error = 0;
503 
504 		goto done;
505 	}
506 #endif
507 
508 	/* adjust pointer */
509 	ip6 = mtod(m, struct ip6_hdr *);
510 
511 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
512 
513 	/* We do not need a route for multicast */
514 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
515 		struct in6_pktinfo *pi = NULL;
516 
517 		/*
518 		 * If the outgoing interface for the address is specified by
519 		 * the caller, use it.
520 		 */
521 		if (opt && (pi = opt->ip6po_pktinfo) != NULL) {
522 			/* XXX boundary check is assumed to be already done. */
523 			ifp = if_get_byindex(pi->ipi6_ifindex, &psref);
524 		} else if (im6o != NULL) {
525 			ifp = if_get_byindex(im6o->im6o_multicast_if_index,
526 			    &psref);
527 		}
528 	}
529 
530 	if (ifp == NULL) {
531 		error = in6_selectroute(&dst_sa, opt, &ro, &rt, true);
532 		if (error != 0)
533 			goto bad;
534 		ifp = if_get_byindex(rt->rt_ifp->if_index, &psref);
535 	}
536 
537 	if (rt == NULL) {
538 		/*
539 		 * If in6_selectroute() does not return a route entry,
540 		 * dst may not have been updated.
541 		 */
542 		error = rtcache_setdst(ro, sin6tosa(&dst_sa));
543 		if (error) {
544 			IP6_STATINC(IP6_STAT_ODROPPED);
545 			goto bad;
546 		}
547 	}
548 
549 	/*
550 	 * then rt (for unicast) and ifp must be non-NULL valid values.
551 	 */
552 	if ((flags & IPV6_FORWARDING) == 0) {
553 		/* XXX: the FORWARDING flag can be set for mrouting. */
554 		in6_ifstat_inc(ifp, ifs6_out_request);
555 	}
556 	if (rt != NULL) {
557 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
558 		rt->rt_use++;
559 	}
560 
561 	/*
562 	 * The outgoing interface must be in the zone of source and
563 	 * destination addresses.  We should use ia_ifp to support the
564 	 * case of sending packets to an address of our own.
565 	 */
566 	if (ia != NULL) {
567 		origifp = ia->ia_ifp;
568 		if (if_is_deactivated(origifp)) {
569 			IP6_STATINC(IP6_STAT_ODROPPED);
570 			goto bad;
571 		}
572 		if_acquire(origifp, &psref_ia);
573 		release_psref_ia = true;
574 	} else
575 		origifp = ifp;
576 
577 	src0 = ip6->ip6_src;
578 	if (in6_setscope(&src0, origifp, &zone))
579 		goto badscope;
580 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
581 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
582 		goto badscope;
583 
584 	dst0 = ip6->ip6_dst;
585 	if (in6_setscope(&dst0, origifp, &zone))
586 		goto badscope;
587 	/* re-initialize to be sure */
588 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
589 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
590 		goto badscope;
591 
592 	/* scope check is done. */
593 
594 	/* Ensure we only send from a valid address. */
595 	if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
596 	    (flags & IPV6_FORWARDING) == 0 &&
597 	    (error = ip6_ifaddrvalid(&src0, &dst0)) != 0)
598 	{
599 		char ip6buf[INET6_ADDRSTRLEN];
600 		nd6log(LOG_ERR,
601 		    "refusing to send from invalid address %s (pid %d)\n",
602 		    IN6_PRINT(ip6buf, &src0), curproc->p_pid);
603 		IP6_STATINC(IP6_STAT_ODROPPED);
604 		in6_ifstat_inc(origifp, ifs6_out_discard);
605 		if (error == 1)
606 			/*
607 			 * Address exists, but is tentative or detached.
608 			 * We can't send from it because it's invalid,
609 			 * so we drop the packet.
610 			 */
611 			error = 0;
612 		else
613 			error = EADDRNOTAVAIL;
614 		goto bad;
615 	}
616 
617 	if (rt != NULL && (rt->rt_flags & RTF_GATEWAY) &&
618 	    !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
619 		dst = satocsin6(rt->rt_gateway);
620 	else
621 		dst = satocsin6(rtcache_getdst(ro));
622 
623 	/*
624 	 * XXXXXX: original code follows:
625 	 */
626 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
627 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
628 	else {
629 		bool ingroup;
630 
631 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
632 
633 		in6_ifstat_inc(ifp, ifs6_out_mcast);
634 
635 		/*
636 		 * Confirm that the outgoing interface supports multicast.
637 		 */
638 		if (!(ifp->if_flags & IFF_MULTICAST)) {
639 			IP6_STATINC(IP6_STAT_NOROUTE);
640 			in6_ifstat_inc(ifp, ifs6_out_discard);
641 			error = ENETUNREACH;
642 			goto bad;
643 		}
644 
645 		ingroup = in6_multi_group(&ip6->ip6_dst, ifp);
646 		if (ingroup && (im6o == NULL || im6o->im6o_multicast_loop)) {
647 			/*
648 			 * If we belong to the destination multicast group
649 			 * on the outgoing interface, and the caller did not
650 			 * forbid loopback, loop back a copy.
651 			 */
652 			KASSERT(dst != NULL);
653 			ip6_mloopback(ifp, m, dst);
654 		} else {
655 			/*
656 			 * If we are acting as a multicast router, perform
657 			 * multicast forwarding as if the packet had just
658 			 * arrived on the interface to which we are about
659 			 * to send.  The multicast forwarding function
660 			 * recursively calls this function, using the
661 			 * IPV6_FORWARDING flag to prevent infinite recursion.
662 			 *
663 			 * Multicasts that are looped back by ip6_mloopback(),
664 			 * above, will be forwarded by the ip6_input() routine,
665 			 * if necessary.
666 			 */
667 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
668 				if (ip6_mforward(ip6, ifp, m) != 0) {
669 					m_freem(m);
670 					goto done;
671 				}
672 			}
673 		}
674 		/*
675 		 * Multicasts with a hoplimit of zero may be looped back,
676 		 * above, but must not be transmitted on a network.
677 		 * Also, multicasts addressed to the loopback interface
678 		 * are not sent -- the above call to ip6_mloopback() will
679 		 * loop back a copy if this host actually belongs to the
680 		 * destination group on the loopback interface.
681 		 */
682 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
683 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
684 			m_freem(m);
685 			goto done;
686 		}
687 	}
688 
689 	/*
690 	 * Fill the outgoing interface to tell the upper layer
691 	 * to increment per-interface statistics.
692 	 */
693 	if (ifpp)
694 		*ifpp = ifp;
695 
696 	/* Determine path MTU. */
697 	/*
698 	 * ro_pmtu represent final destination while
699 	 * ro might represent immediate destination.
700 	 * Use ro_pmtu destination since MTU might differ.
701 	 */
702 	if (ro_pmtu != ro) {
703 		union {
704 			struct sockaddr		dst;
705 			struct sockaddr_in6	dst6;
706 		} u;
707 
708 		/* ro_pmtu may not have a cache */
709 		sockaddr_in6_init(&u.dst6, &finaldst, 0, 0, 0);
710 		rt_pmtu = rtcache_lookup(ro_pmtu, &u.dst);
711 	} else
712 		rt_pmtu = rt;
713 	error = ip6_getpmtu(rt_pmtu, ifp, &mtu, &alwaysfrag);
714 	if (rt_pmtu != NULL && rt_pmtu != rt)
715 		rtcache_unref(rt_pmtu, ro_pmtu);
716 	KASSERT(error == 0); /* ip6_getpmtu never fail if ifp is passed */
717 
718 	/*
719 	 * The caller of this function may specify to use the minimum MTU
720 	 * in some cases.
721 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
722 	 * setting.  The logic is a bit complicated; by default, unicast
723 	 * packets will follow path MTU while multicast packets will be sent at
724 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
725 	 * including unicast ones will be sent at the minimum MTU.  Multicast
726 	 * packets will always be sent at the minimum MTU unless
727 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
728 	 * See RFC 3542 for more details.
729 	 */
730 	if (mtu > IPV6_MMTU) {
731 		if ((flags & IPV6_MINMTU))
732 			mtu = IPV6_MMTU;
733 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
734 			mtu = IPV6_MMTU;
735 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
736 			 (opt == NULL ||
737 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
738 			mtu = IPV6_MMTU;
739 		}
740 	}
741 
742 	/*
743 	 * clear embedded scope identifiers if necessary.
744 	 * in6_clearscope will touch the addresses only when necessary.
745 	 */
746 	in6_clearscope(&ip6->ip6_src);
747 	in6_clearscope(&ip6->ip6_dst);
748 
749 	/*
750 	 * If the outgoing packet contains a hop-by-hop options header,
751 	 * it must be examined and processed even by the source node.
752 	 * (RFC 2460, section 4.)
753 	 *
754 	 * XXX Is this really necessary?
755 	 */
756 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
757 		u_int32_t dummy1 = 0; /* XXX unused */
758 		u_int32_t dummy2; /* XXX unused */
759 		int hoff = sizeof(struct ip6_hdr);
760 
761 		if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
762 			/* m was already freed at this point */
763 			error = EINVAL;
764 			goto done;
765 		}
766 
767 		ip6 = mtod(m, struct ip6_hdr *);
768 	}
769 
770 	/*
771 	 * Run through list of hooks for output packets.
772 	 */
773 	error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT);
774 	if (error != 0 || m == NULL) {
775 		IP6_STATINC(IP6_STAT_PFILDROP_OUT);
776 		goto done;
777 	}
778 	ip6 = mtod(m, struct ip6_hdr *);
779 
780 	/*
781 	 * Send the packet to the outgoing interface.
782 	 * If necessary, do IPv6 fragmentation before sending.
783 	 *
784 	 * the logic here is rather complex:
785 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
786 	 * 1-a:	send as is if tlen <= path mtu
787 	 * 1-b:	fragment if tlen > path mtu
788 	 *
789 	 * 2: if user asks us not to fragment (dontfrag == 1)
790 	 * 2-a:	send as is if tlen <= interface mtu
791 	 * 2-b:	error if tlen > interface mtu
792 	 *
793 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
794 	 *	always fragment
795 	 *
796 	 * 4: if dontfrag == 1 && alwaysfrag == 1
797 	 *	error, as we cannot handle this conflicting request
798 	 */
799 	tlen = m->m_pkthdr.len;
800 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
801 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
802 		dontfrag = 1;
803 	else
804 		dontfrag = 0;
805 
806 	if (dontfrag && alwaysfrag) {	/* case 4 */
807 		/* conflicting request - can't transmit */
808 		IP6_STATINC(IP6_STAT_CANTFRAG);
809 		error = EMSGSIZE;
810 		goto bad;
811 	}
812 	if (dontfrag && (!tso && tlen > ifp->if_mtu)) {	/* case 2-b */
813 		/*
814 		 * Even if the DONTFRAG option is specified, we cannot send the
815 		 * packet when the data length is larger than the MTU of the
816 		 * outgoing interface.
817 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
818 		 * well as returning an error code (the latter is not described
819 		 * in the API spec.)
820 		 */
821 		u_int32_t mtu32;
822 		struct ip6ctlparam ip6cp;
823 
824 		mtu32 = (u_int32_t)mtu;
825 		memset(&ip6cp, 0, sizeof(ip6cp));
826 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
827 		pfctlinput2(PRC_MSGSIZE,
828 		    rtcache_getdst(ro_pmtu), &ip6cp);
829 
830 		IP6_STATINC(IP6_STAT_CANTFRAG);
831 		error = EMSGSIZE;
832 		goto bad;
833 	}
834 
835 	/*
836 	 * transmit packet without fragmentation
837 	 */
838 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
839 		/* case 1-a and 2-a */
840 		struct in6_ifaddr *ia6;
841 		int sw_csum;
842 		int s;
843 
844 		ip6 = mtod(m, struct ip6_hdr *);
845 		s = pserialize_read_enter();
846 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
847 		if (ia6) {
848 			/* Record statistics for this interface address. */
849 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
850 		}
851 		pserialize_read_exit(s);
852 
853 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
854 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
855 			if (IN6_NEED_CHECKSUM(ifp,
856 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
857 				in6_undefer_cksum_tcpudp(m);
858 			}
859 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
860 		}
861 
862 		KASSERT(dst != NULL);
863 		if (__predict_false(sw_csum & M_CSUM_TSOv6)) {
864 			/*
865 			 * TSO6 is required by a packet, but disabled for
866 			 * the interface.
867 			 */
868 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
869 		} else
870 			error = ip6_if_output(ifp, origifp, m, dst, rt);
871 		goto done;
872 	}
873 
874 	if (tso) {
875 		IP6_STATINC(IP6_STAT_CANTFRAG); /* XXX */
876 		error = EINVAL; /* XXX */
877 		goto bad;
878 	}
879 
880 	/*
881 	 * try to fragment the packet.  case 1-b and 3
882 	 */
883 	if (mtu < IPV6_MMTU) {
884 		/* path MTU cannot be less than IPV6_MMTU */
885 		IP6_STATINC(IP6_STAT_CANTFRAG);
886 		error = EMSGSIZE;
887 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
888 		goto bad;
889 	} else if (ip6->ip6_plen == 0) {
890 		/* jumbo payload cannot be fragmented */
891 		IP6_STATINC(IP6_STAT_CANTFRAG);
892 		error = EMSGSIZE;
893 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
894 		goto bad;
895 	} else {
896 		const uint32_t id = ip6_randomid();
897 		struct mbuf **mnext, *m_frgpart;
898 		const int hlen = unfragpartlen;
899 		struct ip6_frag *ip6f;
900 		u_char nextproto;
901 
902 		if (mtu > IPV6_MAXPACKET)
903 			mtu = IPV6_MAXPACKET;
904 
905 		/*
906 		 * Must be able to put at least 8 bytes per fragment.
907 		 */
908 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
909 		if (len < 8) {
910 			IP6_STATINC(IP6_STAT_CANTFRAG);
911 			error = EMSGSIZE;
912 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
913 			goto bad;
914 		}
915 
916 		mnext = &m->m_nextpkt;
917 
918 		/*
919 		 * Change the next header field of the last header in the
920 		 * unfragmentable part.
921 		 */
922 		if (exthdrs.ip6e_rthdr) {
923 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
924 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
925 		} else if (exthdrs.ip6e_dest1) {
926 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
927 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
928 		} else if (exthdrs.ip6e_hbh) {
929 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
930 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
931 		} else {
932 			nextproto = ip6->ip6_nxt;
933 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
934 		}
935 
936 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
937 		    != 0) {
938 			if (IN6_NEED_CHECKSUM(ifp,
939 			    m->m_pkthdr.csum_flags &
940 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
941 				in6_undefer_cksum_tcpudp(m);
942 			}
943 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
944 		}
945 
946 		/*
947 		 * Loop through length of segment after first fragment,
948 		 * make new header and copy data of each part and link onto
949 		 * chain.
950 		 */
951 		m0 = m;
952 		for (off = hlen; off < tlen; off += len) {
953 			struct mbuf *mlast;
954 
955 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
956 			if (!m) {
957 				error = ENOBUFS;
958 				IP6_STATINC(IP6_STAT_ODROPPED);
959 				goto sendorfree;
960 			}
961 			m_reset_rcvif(m);
962 			m->m_flags = m0->m_flags & M_COPYFLAGS;
963 			*mnext = m;
964 			mnext = &m->m_nextpkt;
965 			m->m_data += max_linkhdr;
966 			mhip6 = mtod(m, struct ip6_hdr *);
967 			*mhip6 = *ip6;
968 			m->m_len = sizeof(*mhip6);
969 
970 			ip6f = NULL;
971 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
972 			if (error) {
973 				IP6_STATINC(IP6_STAT_ODROPPED);
974 				goto sendorfree;
975 			}
976 
977 			/* Fill in the Frag6 Header */
978 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
979 			if (off + len >= tlen)
980 				len = tlen - off;
981 			else
982 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
983 			ip6f->ip6f_reserved = 0;
984 			ip6f->ip6f_ident = id;
985 			ip6f->ip6f_nxt = nextproto;
986 
987 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
988 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
989 			if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL) {
990 				error = ENOBUFS;
991 				IP6_STATINC(IP6_STAT_ODROPPED);
992 				goto sendorfree;
993 			}
994 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
995 				;
996 			mlast->m_next = m_frgpart;
997 
998 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
999 			m_reset_rcvif(m);
1000 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
1001 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1002 		}
1003 
1004 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1005 	}
1006 
1007 sendorfree:
1008 	m = m0->m_nextpkt;
1009 	m0->m_nextpkt = 0;
1010 	m_freem(m0);
1011 	for (m0 = m; m; m = m0) {
1012 		m0 = m->m_nextpkt;
1013 		m->m_nextpkt = 0;
1014 		if (error == 0) {
1015 			struct in6_ifaddr *ia6;
1016 			int s;
1017 			ip6 = mtod(m, struct ip6_hdr *);
1018 			s = pserialize_read_enter();
1019 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1020 			if (ia6) {
1021 				/*
1022 				 * Record statistics for this interface
1023 				 * address.
1024 				 */
1025 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
1026 				    m->m_pkthdr.len;
1027 			}
1028 			pserialize_read_exit(s);
1029 			KASSERT(dst != NULL);
1030 			error = ip6_if_output(ifp, origifp, m, dst, rt);
1031 		} else
1032 			m_freem(m);
1033 	}
1034 
1035 	if (error == 0)
1036 		IP6_STATINC(IP6_STAT_FRAGMENTED);
1037 
1038 done:
1039 	rtcache_unref(rt, ro);
1040 	if (ro == &ip6route)
1041 		rtcache_free(&ip6route);
1042 #ifdef IPSEC
1043 	if (sp != NULL)
1044 		KEY_SP_UNREF(&sp);
1045 #endif
1046 	if_put(ifp, &psref);
1047 	if (release_psref_ia)
1048 		if_put(origifp, &psref_ia);
1049 	curlwp_bindx(bound);
1050 
1051 	return error;
1052 
1053 freehdrs:
1054 	m_freem(exthdrs.ip6e_hbh);
1055 	m_freem(exthdrs.ip6e_dest1);
1056 	m_freem(exthdrs.ip6e_rthdr);
1057 	m_freem(exthdrs.ip6e_dest2);
1058 	/* FALLTHROUGH */
1059 bad:
1060 	m_freem(m);
1061 	goto done;
1062 
1063 badscope:
1064 	IP6_STATINC(IP6_STAT_BADSCOPE);
1065 	in6_ifstat_inc(origifp, ifs6_out_discard);
1066 	if (error == 0)
1067 		error = EHOSTUNREACH; /* XXX */
1068 	goto bad;
1069 }
1070 
1071 static int
ip6_copyexthdr(struct mbuf ** mp,void * hdr,int hlen)1072 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1073 {
1074 	struct mbuf *m;
1075 
1076 	if (hlen > MCLBYTES)
1077 		return ENOBUFS; /* XXX */
1078 
1079 	MGET(m, M_DONTWAIT, MT_DATA);
1080 	if (!m)
1081 		return ENOBUFS;
1082 
1083 	if (hlen > MLEN) {
1084 		MCLGET(m, M_DONTWAIT);
1085 		if ((m->m_flags & M_EXT) == 0) {
1086 			m_free(m);
1087 			return ENOBUFS;
1088 		}
1089 	}
1090 	m->m_len = hlen;
1091 	if (hdr)
1092 		memcpy(mtod(m, void *), hdr, hlen);
1093 
1094 	*mp = m;
1095 	return 0;
1096 }
1097 
1098 /*
1099  * Insert jumbo payload option.
1100  */
1101 static int
ip6_insert_jumboopt(struct ip6_exthdrs * exthdrs,u_int32_t plen)1102 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1103 {
1104 	struct mbuf *mopt;
1105 	u_int8_t *optbuf;
1106 	u_int32_t v;
1107 
1108 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1109 
1110 	/*
1111 	 * If there is no hop-by-hop options header, allocate new one.
1112 	 * If there is one but it doesn't have enough space to store the
1113 	 * jumbo payload option, allocate a cluster to store the whole options.
1114 	 * Otherwise, use it to store the options.
1115 	 */
1116 	if (exthdrs->ip6e_hbh == NULL) {
1117 		MGET(mopt, M_DONTWAIT, MT_DATA);
1118 		if (mopt == 0)
1119 			return (ENOBUFS);
1120 		mopt->m_len = JUMBOOPTLEN;
1121 		optbuf = mtod(mopt, u_int8_t *);
1122 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1123 		exthdrs->ip6e_hbh = mopt;
1124 	} else {
1125 		struct ip6_hbh *hbh;
1126 
1127 		mopt = exthdrs->ip6e_hbh;
1128 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1129 			const int oldoptlen = mopt->m_len;
1130 			struct mbuf *n;
1131 
1132 			/*
1133 			 * Assumptions:
1134 			 * - exthdrs->ip6e_hbh is not referenced from places
1135 			 *   other than exthdrs.
1136 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1137 			 */
1138 			KASSERT(mopt->m_next == NULL);
1139 
1140 			/*
1141 			 * Give up if the whole (new) hbh header does not fit
1142 			 * even in an mbuf cluster.
1143 			 */
1144 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1145 				return ENOBUFS;
1146 
1147 			/*
1148 			 * At this point, we must always prepare a cluster.
1149 			 */
1150 			MGET(n, M_DONTWAIT, MT_DATA);
1151 			if (n) {
1152 				MCLGET(n, M_DONTWAIT);
1153 				if ((n->m_flags & M_EXT) == 0) {
1154 					m_freem(n);
1155 					n = NULL;
1156 				}
1157 			}
1158 			if (!n)
1159 				return ENOBUFS;
1160 
1161 			n->m_len = oldoptlen + JUMBOOPTLEN;
1162 			bcopy(mtod(mopt, void *), mtod(n, void *),
1163 			    oldoptlen);
1164 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1165 			m_freem(mopt);
1166 			mopt = exthdrs->ip6e_hbh = n;
1167 		} else {
1168 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1169 			mopt->m_len += JUMBOOPTLEN;
1170 		}
1171 		optbuf[0] = IP6OPT_PADN;
1172 		optbuf[1] = 0;
1173 
1174 		/*
1175 		 * Adjust the header length according to the pad and
1176 		 * the jumbo payload option.
1177 		 */
1178 		hbh = mtod(mopt, struct ip6_hbh *);
1179 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1180 	}
1181 
1182 	/* fill in the option. */
1183 	optbuf[2] = IP6OPT_JUMBO;
1184 	optbuf[3] = 4;
1185 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1186 	memcpy(&optbuf[4], &v, sizeof(u_int32_t));
1187 
1188 	/* finally, adjust the packet header length */
1189 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1190 
1191 	return 0;
1192 #undef JUMBOOPTLEN
1193 }
1194 
1195 /*
1196  * Insert fragment header and copy unfragmentable header portions.
1197  *
1198  * *frghdrp will not be read, and it is guaranteed that either an
1199  * error is returned or that *frghdrp will point to space allocated
1200  * for the fragment header.
1201  *
1202  * On entry, m contains:
1203  *     IPv6 Header
1204  * On exit, it contains:
1205  *     IPv6 Header -> Unfragmentable Part -> Frag6 Header
1206  */
1207 static int
ip6_insertfraghdr(struct mbuf * m0,struct mbuf * m,int hlen,struct ip6_frag ** frghdrp)1208 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1209 	struct ip6_frag **frghdrp)
1210 {
1211 	struct mbuf *n, *mlast;
1212 
1213 	if (hlen > sizeof(struct ip6_hdr)) {
1214 		n = m_copym(m0, sizeof(struct ip6_hdr),
1215 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1216 		if (n == NULL)
1217 			return ENOBUFS;
1218 		m->m_next = n;
1219 	} else
1220 		n = m;
1221 
1222 	/* Search for the last mbuf of unfragmentable part. */
1223 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1224 		;
1225 
1226 	if ((mlast->m_flags & M_EXT) == 0 &&
1227 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1228 		/* use the trailing space of the last mbuf for the fragment hdr */
1229 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1230 		    mlast->m_len);
1231 		mlast->m_len += sizeof(struct ip6_frag);
1232 	} else {
1233 		/* allocate a new mbuf for the fragment header */
1234 		struct mbuf *mfrg;
1235 
1236 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1237 		if (mfrg == NULL)
1238 			return ENOBUFS;
1239 		mfrg->m_len = sizeof(struct ip6_frag);
1240 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1241 		mlast->m_next = mfrg;
1242 	}
1243 
1244 	return 0;
1245 }
1246 
1247 static int
ip6_getpmtu(struct rtentry * rt,struct ifnet * ifp,u_long * mtup,int * alwaysfragp)1248 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup,
1249     int *alwaysfragp)
1250 {
1251 	u_int32_t mtu = 0;
1252 	int alwaysfrag = 0;
1253 	int error = 0;
1254 
1255 	if (rt != NULL) {
1256 		if (ifp == NULL)
1257 			ifp = rt->rt_ifp;
1258 		mtu = rt->rt_rmx.rmx_mtu;
1259 		if (mtu == 0)
1260 			mtu = ifp->if_mtu;
1261 		else if (mtu < IPV6_MMTU) {
1262 			/*
1263 			 * RFC2460 section 5, last paragraph:
1264 			 * if we record ICMPv6 too big message with
1265 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1266 			 * or smaller, with fragment header attached.
1267 			 * (fragment header is needed regardless from the
1268 			 * packet size, for translators to identify packets)
1269 			 */
1270 			alwaysfrag = 1;
1271 			mtu = IPV6_MMTU;
1272 		} else if (mtu > ifp->if_mtu) {
1273 			/*
1274 			 * The MTU on the route is larger than the MTU on
1275 			 * the interface!  This shouldn't happen, unless the
1276 			 * MTU of the interface has been changed after the
1277 			 * interface was brought up.  Change the MTU in the
1278 			 * route to match the interface MTU (as long as the
1279 			 * field isn't locked).
1280 			 */
1281 			mtu = ifp->if_mtu;
1282 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1283 				rt->rt_rmx.rmx_mtu = mtu;
1284 		}
1285 	} else if (ifp) {
1286 		mtu = ifp->if_mtu;
1287 	} else
1288 		error = EHOSTUNREACH; /* XXX */
1289 
1290 	*mtup = mtu;
1291 	if (alwaysfragp)
1292 		*alwaysfragp = alwaysfrag;
1293 	return (error);
1294 }
1295 
1296 /*
1297  * IP6 socket option processing.
1298  */
1299 int
ip6_ctloutput(int op,struct socket * so,struct sockopt * sopt)1300 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1301 {
1302 	int optdatalen, uproto;
1303 	void *optdata;
1304 	struct inpcb *inp = sotoinpcb(so);
1305 	struct ip_moptions **mopts;
1306 	int error, optval;
1307 	int level, optname;
1308 
1309 	KASSERT(solocked(so));
1310 	KASSERT(sopt != NULL);
1311 
1312 	level = sopt->sopt_level;
1313 	optname = sopt->sopt_name;
1314 
1315 	error = optval = 0;
1316 	uproto = (int)so->so_proto->pr_protocol;
1317 
1318 	switch (level) {
1319 	case IPPROTO_IP:
1320 		switch (optname) {
1321 		case IP_ADD_MEMBERSHIP:
1322 		case IP_DROP_MEMBERSHIP:
1323 		case IP_MULTICAST_IF:
1324 		case IP_MULTICAST_LOOP:
1325 		case IP_MULTICAST_TTL:
1326 			mopts = &inp->inp_moptions;
1327 			switch (op) {
1328 			case PRCO_GETOPT:
1329 				return ip_getmoptions(*mopts, sopt);
1330 			case PRCO_SETOPT:
1331 				return ip_setmoptions(mopts, sopt);
1332 			default:
1333 				return EINVAL;
1334 			}
1335 		default:
1336 			return ENOPROTOOPT;
1337 		}
1338 	case IPPROTO_IPV6:
1339 		break;
1340 	default:
1341 		return ENOPROTOOPT;
1342 	}
1343 	switch (op) {
1344 	case PRCO_SETOPT:
1345 		switch (optname) {
1346 #ifdef RFC2292
1347 		case IPV6_2292PKTOPTIONS:
1348 			error = ip6_pcbopts(&in6p_outputopts(inp), so, sopt);
1349 			break;
1350 #endif
1351 
1352 		/*
1353 		 * Use of some Hop-by-Hop options or some
1354 		 * Destination options, might require special
1355 		 * privilege.  That is, normal applications
1356 		 * (without special privilege) might be forbidden
1357 		 * from setting certain options in outgoing packets,
1358 		 * and might never see certain options in received
1359 		 * packets. [RFC 2292 Section 6]
1360 		 * KAME specific note:
1361 		 *  KAME prevents non-privileged users from sending or
1362 		 *  receiving ANY hbh/dst options in order to avoid
1363 		 *  overhead of parsing options in the kernel.
1364 		 */
1365 		case IPV6_RECVHOPOPTS:
1366 		case IPV6_RECVDSTOPTS:
1367 		case IPV6_RECVRTHDRDSTOPTS:
1368 			error = kauth_authorize_network(
1369 			    kauth_cred_get(),
1370 			    KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1371 			    NULL, NULL, NULL);
1372 			if (error)
1373 				break;
1374 			/* FALLTHROUGH */
1375 		case IPV6_UNICAST_HOPS:
1376 		case IPV6_HOPLIMIT:
1377 		case IPV6_FAITH:
1378 
1379 		case IPV6_RECVPKTINFO:
1380 		case IPV6_RECVHOPLIMIT:
1381 		case IPV6_RECVRTHDR:
1382 		case IPV6_RECVPATHMTU:
1383 		case IPV6_RECVTCLASS:
1384 		case IPV6_V6ONLY:
1385 		case IPV6_BINDANY:
1386 			error = sockopt_getint(sopt, &optval);
1387 			if (error)
1388 				break;
1389 			switch (optname) {
1390 			case IPV6_UNICAST_HOPS:
1391 				if (optval < -1 || optval >= 256)
1392 					error = EINVAL;
1393 				else {
1394 					/* -1 = kernel default */
1395 					in6p_hops6(inp) = optval;
1396 				}
1397 				break;
1398 #define OPTSET(bit) \
1399 do { \
1400 if (optval) \
1401 	inp->inp_flags |= (bit); \
1402 else \
1403 	inp->inp_flags &= ~(bit); \
1404 } while (/*CONSTCOND*/ 0)
1405 
1406 #ifdef RFC2292
1407 #define OPTSET2292(bit) 			\
1408 do { 						\
1409 inp->inp_flags |= IN6P_RFC2292; 	\
1410 if (optval) 				\
1411 	inp->inp_flags |= (bit); 	\
1412 else 					\
1413 	inp->inp_flags &= ~(bit); 	\
1414 } while (/*CONSTCOND*/ 0)
1415 #endif
1416 
1417 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1418 
1419 			case IPV6_RECVPKTINFO:
1420 #ifdef RFC2292
1421 				/* cannot mix with RFC2292 */
1422 				if (OPTBIT(IN6P_RFC2292)) {
1423 					error = EINVAL;
1424 					break;
1425 				}
1426 #endif
1427 				OPTSET(IN6P_PKTINFO);
1428 				break;
1429 
1430 			case IPV6_HOPLIMIT:
1431 			{
1432 				struct ip6_pktopts **optp;
1433 
1434 #ifdef RFC2292
1435 				/* cannot mix with RFC2292 */
1436 				if (OPTBIT(IN6P_RFC2292)) {
1437 					error = EINVAL;
1438 					break;
1439 				}
1440 #endif
1441 				optp = &in6p_outputopts(inp);
1442 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1443 						   (u_char *)&optval,
1444 						   sizeof(optval),
1445 						   optp,
1446 						   kauth_cred_get(), uproto);
1447 				break;
1448 			}
1449 
1450 			case IPV6_RECVHOPLIMIT:
1451 #ifdef RFC2292
1452 				/* cannot mix with RFC2292 */
1453 				if (OPTBIT(IN6P_RFC2292)) {
1454 					error = EINVAL;
1455 					break;
1456 				}
1457 #endif
1458 				OPTSET(IN6P_HOPLIMIT);
1459 				break;
1460 
1461 			case IPV6_RECVHOPOPTS:
1462 #ifdef RFC2292
1463 				/* cannot mix with RFC2292 */
1464 				if (OPTBIT(IN6P_RFC2292)) {
1465 					error = EINVAL;
1466 					break;
1467 				}
1468 #endif
1469 				OPTSET(IN6P_HOPOPTS);
1470 				break;
1471 
1472 			case IPV6_RECVDSTOPTS:
1473 #ifdef RFC2292
1474 				/* cannot mix with RFC2292 */
1475 				if (OPTBIT(IN6P_RFC2292)) {
1476 					error = EINVAL;
1477 					break;
1478 				}
1479 #endif
1480 				OPTSET(IN6P_DSTOPTS);
1481 				break;
1482 
1483 			case IPV6_RECVRTHDRDSTOPTS:
1484 #ifdef RFC2292
1485 				/* cannot mix with RFC2292 */
1486 				if (OPTBIT(IN6P_RFC2292)) {
1487 					error = EINVAL;
1488 					break;
1489 				}
1490 #endif
1491 				OPTSET(IN6P_RTHDRDSTOPTS);
1492 				break;
1493 
1494 			case IPV6_RECVRTHDR:
1495 #ifdef RFC2292
1496 				/* cannot mix with RFC2292 */
1497 				if (OPTBIT(IN6P_RFC2292)) {
1498 					error = EINVAL;
1499 					break;
1500 				}
1501 #endif
1502 				OPTSET(IN6P_RTHDR);
1503 				break;
1504 
1505 			case IPV6_FAITH:
1506 				OPTSET(IN6P_FAITH);
1507 				break;
1508 
1509 			case IPV6_RECVPATHMTU:
1510 				/*
1511 				 * We ignore this option for TCP
1512 				 * sockets.
1513 				 * (RFC3542 leaves this case
1514 				 * unspecified.)
1515 				 */
1516 				if (uproto != IPPROTO_TCP)
1517 					OPTSET(IN6P_MTU);
1518 				break;
1519 
1520 			case IPV6_V6ONLY:
1521 				/*
1522 				 * make setsockopt(IPV6_V6ONLY)
1523 				 * available only prior to bind(2).
1524 				 * see ipng mailing list, Jun 22 2001.
1525 				 */
1526 				if (inp->inp_lport ||
1527 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp))) {
1528 					error = EINVAL;
1529 					break;
1530 				}
1531 #ifdef INET6_BINDV6ONLY
1532 				if (!optval)
1533 					error = EINVAL;
1534 #else
1535 				OPTSET(IN6P_IPV6_V6ONLY);
1536 #endif
1537 				break;
1538 
1539 			case IPV6_RECVTCLASS:
1540 #ifdef RFC2292
1541 				/* cannot mix with RFC2292 XXX */
1542 				if (OPTBIT(IN6P_RFC2292)) {
1543 					error = EINVAL;
1544 					break;
1545 				}
1546 #endif
1547 				OPTSET(IN6P_TCLASS);
1548 				break;
1549 
1550 			case IPV6_BINDANY:
1551 				error = kauth_authorize_network(
1552 				    kauth_cred_get(), KAUTH_NETWORK_BIND,
1553 				    KAUTH_REQ_NETWORK_BIND_ANYADDR, so, NULL,
1554 				    NULL);
1555 				if (error)
1556 					break;
1557 				OPTSET(IN6P_BINDANY);
1558 				break;
1559 			}
1560 			break;
1561 
1562 		case IPV6_OTCLASS:
1563 		{
1564 			struct ip6_pktopts **optp;
1565 			u_int8_t tclass;
1566 
1567 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1568 			if (error)
1569 				break;
1570 			optp = &in6p_outputopts(inp);
1571 			error = ip6_pcbopt(optname,
1572 					   (u_char *)&tclass,
1573 					   sizeof(tclass),
1574 					   optp,
1575 					   kauth_cred_get(), uproto);
1576 			break;
1577 		}
1578 
1579 		case IPV6_TCLASS:
1580 		case IPV6_DONTFRAG:
1581 		case IPV6_USE_MIN_MTU:
1582 		case IPV6_PREFER_TEMPADDR:
1583 			error = sockopt_getint(sopt, &optval);
1584 			if (error)
1585 				break;
1586 			{
1587 				struct ip6_pktopts **optp;
1588 				optp = &in6p_outputopts(inp);
1589 				error = ip6_pcbopt(optname,
1590 						   (u_char *)&optval,
1591 						   sizeof(optval),
1592 						   optp,
1593 						   kauth_cred_get(), uproto);
1594 				break;
1595 			}
1596 
1597 #ifdef RFC2292
1598 		case IPV6_2292PKTINFO:
1599 		case IPV6_2292HOPLIMIT:
1600 		case IPV6_2292HOPOPTS:
1601 		case IPV6_2292DSTOPTS:
1602 		case IPV6_2292RTHDR:
1603 			/* RFC 2292 */
1604 			error = sockopt_getint(sopt, &optval);
1605 			if (error)
1606 				break;
1607 
1608 			switch (optname) {
1609 			case IPV6_2292PKTINFO:
1610 				OPTSET2292(IN6P_PKTINFO);
1611 				break;
1612 			case IPV6_2292HOPLIMIT:
1613 				OPTSET2292(IN6P_HOPLIMIT);
1614 				break;
1615 			case IPV6_2292HOPOPTS:
1616 				/*
1617 				 * Check super-user privilege.
1618 				 * See comments for IPV6_RECVHOPOPTS.
1619 				 */
1620 				error = kauth_authorize_network(
1621 				    kauth_cred_get(),
1622 				    KAUTH_NETWORK_IPV6,
1623 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1624 				    NULL, NULL);
1625 				if (error)
1626 					return (error);
1627 				OPTSET2292(IN6P_HOPOPTS);
1628 				break;
1629 			case IPV6_2292DSTOPTS:
1630 				error = kauth_authorize_network(
1631 				    kauth_cred_get(),
1632 				    KAUTH_NETWORK_IPV6,
1633 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1634 				    NULL, NULL);
1635 				if (error)
1636 					return (error);
1637 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1638 				break;
1639 			case IPV6_2292RTHDR:
1640 				OPTSET2292(IN6P_RTHDR);
1641 				break;
1642 			}
1643 			break;
1644 #endif
1645 		case IPV6_PKTINFO:
1646 		case IPV6_HOPOPTS:
1647 		case IPV6_RTHDR:
1648 		case IPV6_DSTOPTS:
1649 		case IPV6_RTHDRDSTOPTS:
1650 		case IPV6_NEXTHOP: {
1651 			/* new advanced API (RFC3542) */
1652 			void *optbuf;
1653 			int optbuflen;
1654 			struct ip6_pktopts **optp;
1655 
1656 #ifdef RFC2292
1657 			/* cannot mix with RFC2292 */
1658 			if (OPTBIT(IN6P_RFC2292)) {
1659 				error = EINVAL;
1660 				break;
1661 			}
1662 #endif
1663 
1664 			optbuflen = sopt->sopt_size;
1665 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1666 			if (optbuf == NULL) {
1667 				error = ENOBUFS;
1668 				break;
1669 			}
1670 
1671 			error = sockopt_get(sopt, optbuf, optbuflen);
1672 			if (error) {
1673 				free(optbuf, M_IP6OPT);
1674 				break;
1675 			}
1676 			optp = &in6p_outputopts(inp);
1677 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1678 			    optp, kauth_cred_get(), uproto);
1679 
1680 			free(optbuf, M_IP6OPT);
1681 			break;
1682 			}
1683 #undef OPTSET
1684 
1685 		case IPV6_MULTICAST_IF:
1686 		case IPV6_MULTICAST_HOPS:
1687 		case IPV6_MULTICAST_LOOP:
1688 		case IPV6_JOIN_GROUP:
1689 		case IPV6_LEAVE_GROUP:
1690 			error = ip6_setmoptions(sopt, inp);
1691 			break;
1692 
1693 		case IPV6_PORTRANGE:
1694 			error = sockopt_getint(sopt, &optval);
1695 			if (error)
1696 				break;
1697 
1698 			switch (optval) {
1699 			case IPV6_PORTRANGE_DEFAULT:
1700 				inp->inp_flags &= ~(IN6P_LOWPORT);
1701 				inp->inp_flags &= ~(IN6P_HIGHPORT);
1702 				break;
1703 
1704 			case IPV6_PORTRANGE_HIGH:
1705 				inp->inp_flags &= ~(IN6P_LOWPORT);
1706 				inp->inp_flags |= IN6P_HIGHPORT;
1707 				break;
1708 
1709 			case IPV6_PORTRANGE_LOW:
1710 				inp->inp_flags &= ~(IN6P_HIGHPORT);
1711 				inp->inp_flags |= IN6P_LOWPORT;
1712 				break;
1713 
1714 			default:
1715 				error = EINVAL;
1716 				break;
1717 			}
1718 			break;
1719 
1720 		case IPV6_PORTALGO:
1721 			error = sockopt_getint(sopt, &optval);
1722 			if (error)
1723 				break;
1724 
1725 			error = portalgo_algo_index_select(inp, optval);
1726 			break;
1727 
1728 #if defined(IPSEC)
1729 		case IPV6_IPSEC_POLICY:
1730 			if (ipsec_enabled) {
1731 				error = ipsec_set_policy(inp,
1732 				    sopt->sopt_data, sopt->sopt_size,
1733 				    kauth_cred_get());
1734 			} else
1735 				error = ENOPROTOOPT;
1736 			break;
1737 #endif /* IPSEC */
1738 
1739 		default:
1740 			error = ENOPROTOOPT;
1741 			break;
1742 		}
1743 		break;
1744 
1745 	case PRCO_GETOPT:
1746 		switch (optname) {
1747 #ifdef RFC2292
1748 		case IPV6_2292PKTOPTIONS:
1749 			/*
1750 			 * RFC3542 (effectively) deprecated the
1751 			 * semantics of the 2292-style pktoptions.
1752 			 * Since it was not reliable in nature (i.e.,
1753 			 * applications had to expect the lack of some
1754 			 * information after all), it would make sense
1755 			 * to simplify this part by always returning
1756 			 * empty data.
1757 			 */
1758 			break;
1759 #endif
1760 
1761 		case IPV6_RECVHOPOPTS:
1762 		case IPV6_RECVDSTOPTS:
1763 		case IPV6_RECVRTHDRDSTOPTS:
1764 		case IPV6_UNICAST_HOPS:
1765 		case IPV6_RECVPKTINFO:
1766 		case IPV6_RECVHOPLIMIT:
1767 		case IPV6_RECVRTHDR:
1768 		case IPV6_RECVPATHMTU:
1769 
1770 		case IPV6_FAITH:
1771 		case IPV6_V6ONLY:
1772 		case IPV6_PORTRANGE:
1773 		case IPV6_RECVTCLASS:
1774 		case IPV6_BINDANY:
1775 			switch (optname) {
1776 
1777 			case IPV6_RECVHOPOPTS:
1778 				optval = OPTBIT(IN6P_HOPOPTS);
1779 				break;
1780 
1781 			case IPV6_RECVDSTOPTS:
1782 				optval = OPTBIT(IN6P_DSTOPTS);
1783 				break;
1784 
1785 			case IPV6_RECVRTHDRDSTOPTS:
1786 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1787 				break;
1788 
1789 			case IPV6_UNICAST_HOPS:
1790 				optval = in6p_hops6(inp);
1791 				break;
1792 
1793 			case IPV6_RECVPKTINFO:
1794 				optval = OPTBIT(IN6P_PKTINFO);
1795 				break;
1796 
1797 			case IPV6_RECVHOPLIMIT:
1798 				optval = OPTBIT(IN6P_HOPLIMIT);
1799 				break;
1800 
1801 			case IPV6_RECVRTHDR:
1802 				optval = OPTBIT(IN6P_RTHDR);
1803 				break;
1804 
1805 			case IPV6_RECVPATHMTU:
1806 				optval = OPTBIT(IN6P_MTU);
1807 				break;
1808 
1809 			case IPV6_FAITH:
1810 				optval = OPTBIT(IN6P_FAITH);
1811 				break;
1812 
1813 			case IPV6_V6ONLY:
1814 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1815 				break;
1816 
1817 			case IPV6_PORTRANGE:
1818 			    {
1819 				int flags;
1820 				flags = inp->inp_flags;
1821 				if (flags & IN6P_HIGHPORT)
1822 					optval = IPV6_PORTRANGE_HIGH;
1823 				else if (flags & IN6P_LOWPORT)
1824 					optval = IPV6_PORTRANGE_LOW;
1825 				else
1826 					optval = 0;
1827 				break;
1828 			    }
1829 			case IPV6_RECVTCLASS:
1830 				optval = OPTBIT(IN6P_TCLASS);
1831 				break;
1832 
1833 			case IPV6_BINDANY:
1834 				optval = OPTBIT(IN6P_BINDANY);
1835 				break;
1836 			}
1837 
1838 			if (error)
1839 				break;
1840 			error = sockopt_setint(sopt, optval);
1841 			break;
1842 
1843 		case IPV6_PATHMTU:
1844 		    {
1845 			u_long pmtu = 0;
1846 			struct ip6_mtuinfo mtuinfo;
1847 			struct route *ro = &inp->inp_route;
1848 			struct rtentry *rt;
1849 			union {
1850 				struct sockaddr		dst;
1851 				struct sockaddr_in6	dst6;
1852 			} u;
1853 
1854 			if (!(so->so_state & SS_ISCONNECTED))
1855 				return (ENOTCONN);
1856 			/*
1857 			 * XXX: we dot not consider the case of source
1858 			 * routing, or optional information to specify
1859 			 * the outgoing interface.
1860 			 */
1861 			sockaddr_in6_init(&u.dst6, &in6p_faddr(inp), 0, 0, 0);
1862 			rt = rtcache_lookup(ro, &u.dst);
1863 			error = ip6_getpmtu(rt, NULL, &pmtu, NULL);
1864 			rtcache_unref(rt, ro);
1865 			if (error)
1866 				break;
1867 			if (pmtu > IPV6_MAXPACKET)
1868 				pmtu = IPV6_MAXPACKET;
1869 
1870 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1871 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1872 			optdata = (void *)&mtuinfo;
1873 			optdatalen = sizeof(mtuinfo);
1874 			if (optdatalen > MCLBYTES)
1875 				return (EMSGSIZE); /* XXX */
1876 			error = sockopt_set(sopt, optdata, optdatalen);
1877 			break;
1878 		    }
1879 
1880 #ifdef RFC2292
1881 		case IPV6_2292PKTINFO:
1882 		case IPV6_2292HOPLIMIT:
1883 		case IPV6_2292HOPOPTS:
1884 		case IPV6_2292RTHDR:
1885 		case IPV6_2292DSTOPTS:
1886 			switch (optname) {
1887 			case IPV6_2292PKTINFO:
1888 				optval = OPTBIT(IN6P_PKTINFO);
1889 				break;
1890 			case IPV6_2292HOPLIMIT:
1891 				optval = OPTBIT(IN6P_HOPLIMIT);
1892 				break;
1893 			case IPV6_2292HOPOPTS:
1894 				optval = OPTBIT(IN6P_HOPOPTS);
1895 				break;
1896 			case IPV6_2292RTHDR:
1897 				optval = OPTBIT(IN6P_RTHDR);
1898 				break;
1899 			case IPV6_2292DSTOPTS:
1900 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1901 				break;
1902 			}
1903 			error = sockopt_setint(sopt, optval);
1904 			break;
1905 #endif
1906 		case IPV6_PKTINFO:
1907 		case IPV6_HOPOPTS:
1908 		case IPV6_RTHDR:
1909 		case IPV6_DSTOPTS:
1910 		case IPV6_RTHDRDSTOPTS:
1911 		case IPV6_NEXTHOP:
1912 		case IPV6_OTCLASS:
1913 		case IPV6_TCLASS:
1914 		case IPV6_DONTFRAG:
1915 		case IPV6_USE_MIN_MTU:
1916 		case IPV6_PREFER_TEMPADDR:
1917 			error = ip6_getpcbopt(in6p_outputopts(inp),
1918 			    optname, sopt);
1919 			break;
1920 
1921 		case IPV6_MULTICAST_IF:
1922 		case IPV6_MULTICAST_HOPS:
1923 		case IPV6_MULTICAST_LOOP:
1924 		case IPV6_JOIN_GROUP:
1925 		case IPV6_LEAVE_GROUP:
1926 			error = ip6_getmoptions(sopt, inp);
1927 			break;
1928 
1929 		case IPV6_PORTALGO:
1930 			optval = inp->inp_portalgo;
1931 			error = sockopt_setint(sopt, optval);
1932 			break;
1933 
1934 #if defined(IPSEC)
1935 		case IPV6_IPSEC_POLICY:
1936 			if (ipsec_used) {
1937 				struct mbuf *m = NULL;
1938 
1939 				/*
1940 				 * XXX: this will return EINVAL as sopt is
1941 				 * empty
1942 				 */
1943 				error = ipsec_get_policy(inp, sopt->sopt_data,
1944 				    sopt->sopt_size, &m);
1945 				if (!error)
1946 					error = sockopt_setmbuf(sopt, m);
1947 			} else
1948 				error = ENOPROTOOPT;
1949 			break;
1950 #endif /* IPSEC */
1951 
1952 		default:
1953 			error = ENOPROTOOPT;
1954 			break;
1955 		}
1956 		break;
1957 	}
1958 	return (error);
1959 }
1960 
1961 int
ip6_raw_ctloutput(int op,struct socket * so,struct sockopt * sopt)1962 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1963 {
1964 	int error = 0, optval;
1965 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1966 	struct inpcb *inp = sotoinpcb(so);
1967 	int level, optname;
1968 
1969 	KASSERT(sopt != NULL);
1970 
1971 	level = sopt->sopt_level;
1972 	optname = sopt->sopt_name;
1973 
1974 	if (level != IPPROTO_IPV6) {
1975 		return ENOPROTOOPT;
1976 	}
1977 
1978 	switch (optname) {
1979 	case IPV6_CHECKSUM:
1980 		/*
1981 		 * For ICMPv6 sockets, no modification allowed for checksum
1982 		 * offset, permit "no change" values to help existing apps.
1983 		 *
1984 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1985 		 * for an ICMPv6 socket will fail."  The current
1986 		 * behavior does not meet RFC3542.
1987 		 */
1988 		switch (op) {
1989 		case PRCO_SETOPT:
1990 			error = sockopt_getint(sopt, &optval);
1991 			if (error)
1992 				break;
1993 			if (optval < -1 ||
1994 			    (optval > 0 && (optval % 2) != 0)) {
1995 				/*
1996 				 * The API assumes non-negative even offset
1997 				 * values or -1 as a special value.
1998 				 */
1999 				error = EINVAL;
2000 			} else if (so->so_proto->pr_protocol ==
2001 			    IPPROTO_ICMPV6) {
2002 				if (optval != icmp6off)
2003 					error = EINVAL;
2004 			} else
2005 				in6p_cksum(inp) = optval;
2006 			break;
2007 
2008 		case PRCO_GETOPT:
2009 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2010 				optval = icmp6off;
2011 			else
2012 				optval = in6p_cksum(inp);
2013 
2014 			error = sockopt_setint(sopt, optval);
2015 			break;
2016 
2017 		default:
2018 			error = EINVAL;
2019 			break;
2020 		}
2021 		break;
2022 
2023 	default:
2024 		error = ENOPROTOOPT;
2025 		break;
2026 	}
2027 
2028 	return (error);
2029 }
2030 
2031 #ifdef RFC2292
2032 /*
2033  * Set up IP6 options in pcb for insertion in output packets or
2034  * specifying behavior of outgoing packets.
2035  */
2036 static int
ip6_pcbopts(struct ip6_pktopts ** pktopt,struct socket * so,struct sockopt * sopt)2037 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
2038     struct sockopt *sopt)
2039 {
2040 	struct ip6_pktopts *opt = *pktopt;
2041 	struct mbuf *m;
2042 	int error = 0;
2043 
2044 	KASSERT(solocked(so));
2045 
2046 	/* turn off any old options. */
2047 	if (opt) {
2048 #ifdef DIAGNOSTIC
2049 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2050 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2051 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2052 		    printf("ip6_pcbopts: all specified options are cleared.\n");
2053 #endif
2054 		ip6_clearpktopts(opt, -1);
2055 	} else {
2056 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2057 		if (opt == NULL)
2058 			return (ENOBUFS);
2059 	}
2060 	*pktopt = NULL;
2061 
2062 	if (sopt == NULL || sopt->sopt_size == 0) {
2063 		/*
2064 		 * Only turning off any previous options, regardless of
2065 		 * whether the opt is just created or given.
2066 		 */
2067 		free(opt, M_IP6OPT);
2068 		return (0);
2069 	}
2070 
2071 	/*  set options specified by user. */
2072 	m = sockopt_getmbuf(sopt);
2073 	if (m == NULL) {
2074 		free(opt, M_IP6OPT);
2075 		return (ENOBUFS);
2076 	}
2077 
2078 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
2079 	    so->so_proto->pr_protocol);
2080 	m_freem(m);
2081 	if (error != 0) {
2082 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2083 		free(opt, M_IP6OPT);
2084 		return (error);
2085 	}
2086 	*pktopt = opt;
2087 	return (0);
2088 }
2089 #endif
2090 
2091 /*
2092  * initialize ip6_pktopts.  beware that there are non-zero default values in
2093  * the struct.
2094  */
2095 void
ip6_initpktopts(struct ip6_pktopts * opt)2096 ip6_initpktopts(struct ip6_pktopts *opt)
2097 {
2098 
2099 	memset(opt, 0, sizeof(*opt));
2100 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2101 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2102 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2103 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2104 }
2105 
2106 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2107 static int
ip6_pcbopt(int optname,u_char * buf,int len,struct ip6_pktopts ** pktopt,kauth_cred_t cred,int uproto)2108 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2109     kauth_cred_t cred, int uproto)
2110 {
2111 	struct ip6_pktopts *opt;
2112 
2113 	if (*pktopt == NULL) {
2114 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2115 		    M_NOWAIT);
2116 		if (*pktopt == NULL)
2117 			return (ENOBUFS);
2118 
2119 		ip6_initpktopts(*pktopt);
2120 	}
2121 	opt = *pktopt;
2122 
2123 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2124 }
2125 
2126 static int
ip6_getpcbopt(struct ip6_pktopts * pktopt,int optname,struct sockopt * sopt)2127 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2128 {
2129 	void *optdata = NULL;
2130 	int optdatalen = 0;
2131 	struct ip6_ext *ip6e;
2132 	int error = 0;
2133 	struct in6_pktinfo null_pktinfo;
2134 	int deftclass = 0, on;
2135 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2136 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2137 
2138 	switch (optname) {
2139 	case IPV6_PKTINFO:
2140 		if (pktopt && pktopt->ip6po_pktinfo)
2141 			optdata = (void *)pktopt->ip6po_pktinfo;
2142 		else {
2143 			/* XXX: we don't have to do this every time... */
2144 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2145 			optdata = (void *)&null_pktinfo;
2146 		}
2147 		optdatalen = sizeof(struct in6_pktinfo);
2148 		break;
2149 	case IPV6_OTCLASS:
2150 		/* XXX */
2151 		return (EINVAL);
2152 	case IPV6_TCLASS:
2153 		if (pktopt && pktopt->ip6po_tclass >= 0)
2154 			optdata = (void *)&pktopt->ip6po_tclass;
2155 		else
2156 			optdata = (void *)&deftclass;
2157 		optdatalen = sizeof(int);
2158 		break;
2159 	case IPV6_HOPOPTS:
2160 		if (pktopt && pktopt->ip6po_hbh) {
2161 			optdata = (void *)pktopt->ip6po_hbh;
2162 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2163 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2164 		}
2165 		break;
2166 	case IPV6_RTHDR:
2167 		if (pktopt && pktopt->ip6po_rthdr) {
2168 			optdata = (void *)pktopt->ip6po_rthdr;
2169 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2170 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2171 		}
2172 		break;
2173 	case IPV6_RTHDRDSTOPTS:
2174 		if (pktopt && pktopt->ip6po_dest1) {
2175 			optdata = (void *)pktopt->ip6po_dest1;
2176 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2177 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2178 		}
2179 		break;
2180 	case IPV6_DSTOPTS:
2181 		if (pktopt && pktopt->ip6po_dest2) {
2182 			optdata = (void *)pktopt->ip6po_dest2;
2183 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2184 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2185 		}
2186 		break;
2187 	case IPV6_NEXTHOP:
2188 		if (pktopt && pktopt->ip6po_nexthop) {
2189 			optdata = (void *)pktopt->ip6po_nexthop;
2190 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2191 		}
2192 		break;
2193 	case IPV6_USE_MIN_MTU:
2194 		if (pktopt)
2195 			optdata = (void *)&pktopt->ip6po_minmtu;
2196 		else
2197 			optdata = (void *)&defminmtu;
2198 		optdatalen = sizeof(int);
2199 		break;
2200 	case IPV6_DONTFRAG:
2201 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2202 			on = 1;
2203 		else
2204 			on = 0;
2205 		optdata = (void *)&on;
2206 		optdatalen = sizeof(on);
2207 		break;
2208 	case IPV6_PREFER_TEMPADDR:
2209 		if (pktopt)
2210 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2211 		else
2212 			optdata = (void *)&defpreftemp;
2213 		optdatalen = sizeof(int);
2214 		break;
2215 	default:		/* should not happen */
2216 #ifdef DIAGNOSTIC
2217 		panic("ip6_getpcbopt: unexpected option\n");
2218 #endif
2219 		return (ENOPROTOOPT);
2220 	}
2221 
2222 	error = sockopt_set(sopt, optdata, optdatalen);
2223 
2224 	return (error);
2225 }
2226 
2227 void
ip6_clearpktopts(struct ip6_pktopts * pktopt,int optname)2228 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2229 {
2230 	if (optname == -1 || optname == IPV6_PKTINFO) {
2231 		if (pktopt->ip6po_pktinfo)
2232 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2233 		pktopt->ip6po_pktinfo = NULL;
2234 	}
2235 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2236 		pktopt->ip6po_hlim = -1;
2237 	if (optname == -1 || optname == IPV6_TCLASS)
2238 		pktopt->ip6po_tclass = -1;
2239 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2240 		rtcache_free(&pktopt->ip6po_nextroute);
2241 		if (pktopt->ip6po_nexthop)
2242 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2243 		pktopt->ip6po_nexthop = NULL;
2244 	}
2245 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2246 		if (pktopt->ip6po_hbh)
2247 			free(pktopt->ip6po_hbh, M_IP6OPT);
2248 		pktopt->ip6po_hbh = NULL;
2249 	}
2250 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2251 		if (pktopt->ip6po_dest1)
2252 			free(pktopt->ip6po_dest1, M_IP6OPT);
2253 		pktopt->ip6po_dest1 = NULL;
2254 	}
2255 	if (optname == -1 || optname == IPV6_RTHDR) {
2256 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2257 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2258 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2259 		rtcache_free(&pktopt->ip6po_route);
2260 	}
2261 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2262 		if (pktopt->ip6po_dest2)
2263 			free(pktopt->ip6po_dest2, M_IP6OPT);
2264 		pktopt->ip6po_dest2 = NULL;
2265 	}
2266 }
2267 
2268 #define PKTOPT_EXTHDRCPY(type) 					\
2269 do {								\
2270 	if (src->type) {					\
2271 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2272 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2273 		if (dst->type == NULL)				\
2274 			goto bad;				\
2275 		memcpy(dst->type, src->type, hlen);		\
2276 	}							\
2277 } while (/*CONSTCOND*/ 0)
2278 
2279 static int
copypktopts(struct ip6_pktopts * dst,struct ip6_pktopts * src,int canwait)2280 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2281 {
2282 	dst->ip6po_hlim = src->ip6po_hlim;
2283 	dst->ip6po_tclass = src->ip6po_tclass;
2284 	dst->ip6po_flags = src->ip6po_flags;
2285 	dst->ip6po_minmtu = src->ip6po_minmtu;
2286 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2287 	if (src->ip6po_pktinfo) {
2288 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2289 		    M_IP6OPT, canwait);
2290 		if (dst->ip6po_pktinfo == NULL)
2291 			goto bad;
2292 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2293 	}
2294 	if (src->ip6po_nexthop) {
2295 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2296 		    M_IP6OPT, canwait);
2297 		if (dst->ip6po_nexthop == NULL)
2298 			goto bad;
2299 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2300 		    src->ip6po_nexthop->sa_len);
2301 	}
2302 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2303 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2304 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2305 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2306 	return (0);
2307 
2308   bad:
2309 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2310 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2311 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2312 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2313 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2314 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2315 
2316 	return (ENOBUFS);
2317 }
2318 #undef PKTOPT_EXTHDRCPY
2319 
2320 struct ip6_pktopts *
ip6_copypktopts(struct ip6_pktopts * src,int canwait)2321 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2322 {
2323 	int error;
2324 	struct ip6_pktopts *dst;
2325 
2326 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2327 	if (dst == NULL)
2328 		return (NULL);
2329 	ip6_initpktopts(dst);
2330 
2331 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2332 		free(dst, M_IP6OPT);
2333 		return (NULL);
2334 	}
2335 
2336 	return (dst);
2337 }
2338 
2339 void
ip6_freepcbopts(struct ip6_pktopts * pktopt)2340 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2341 {
2342 	if (pktopt == NULL)
2343 		return;
2344 
2345 	ip6_clearpktopts(pktopt, -1);
2346 
2347 	free(pktopt, M_IP6OPT);
2348 }
2349 
2350 int
ip6_get_membership(const struct sockopt * sopt,struct ifnet ** ifp,struct psref * psref,void * v,size_t l)2351 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp,
2352     struct psref *psref, void *v, size_t l)
2353 {
2354 	struct ipv6_mreq mreq;
2355 	int error;
2356 	struct in6_addr *ia = &mreq.ipv6mr_multiaddr;
2357 	struct in_addr *ia4 = (void *)&ia->s6_addr32[3];
2358 
2359 	error = sockopt_get(sopt, &mreq, sizeof(mreq));
2360 	if (error != 0)
2361 		return error;
2362 
2363 	if (IN6_IS_ADDR_UNSPECIFIED(ia)) {
2364 		/*
2365 		 * We use the unspecified address to specify to accept
2366 		 * all multicast addresses. Only super user is allowed
2367 		 * to do this.
2368 		 */
2369 		if (kauth_authorize_network(kauth_cred_get(),
2370 		    KAUTH_NETWORK_IPV6,
2371 		    KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2372 			return EACCES;
2373 	} else if (IN6_IS_ADDR_V4MAPPED(ia)) {
2374 		// Don't bother if we are not going to use ifp.
2375 		if (l == sizeof(*ia)) {
2376 			memcpy(v, ia, l);
2377 			return 0;
2378 		}
2379 	} else if (!IN6_IS_ADDR_MULTICAST(ia)) {
2380 		return EINVAL;
2381 	}
2382 
2383 	/*
2384 	 * If no interface was explicitly specified, choose an
2385 	 * appropriate one according to the given multicast address.
2386 	 */
2387 	if (mreq.ipv6mr_interface == 0) {
2388 		struct rtentry *rt;
2389 		union {
2390 			struct sockaddr		dst;
2391 			struct sockaddr_in	dst4;
2392 			struct sockaddr_in6	dst6;
2393 		} u;
2394 		struct route ro;
2395 
2396 		/*
2397 		 * Look up the routing table for the
2398 		 * address, and choose the outgoing interface.
2399 		 *   XXX: is it a good approach?
2400 		 */
2401 		memset(&ro, 0, sizeof(ro));
2402 		if (IN6_IS_ADDR_V4MAPPED(ia))
2403 			sockaddr_in_init(&u.dst4, ia4, 0);
2404 		else
2405 			sockaddr_in6_init(&u.dst6, ia, 0, 0, 0);
2406 		error = rtcache_setdst(&ro, &u.dst);
2407 		if (error != 0)
2408 			return error;
2409 		rt = rtcache_init(&ro);
2410 		*ifp = rt != NULL ?
2411 		    if_get_byindex(rt->rt_ifp->if_index, psref) : NULL;
2412 		rtcache_unref(rt, &ro);
2413 		rtcache_free(&ro);
2414 	} else {
2415 		/*
2416 		 * If the interface is specified, validate it.
2417 		 */
2418 		*ifp = if_get_byindex(mreq.ipv6mr_interface, psref);
2419 		if (*ifp == NULL)
2420 			return ENXIO;	/* XXX EINVAL? */
2421 	}
2422 	if (sizeof(*ia) == l)
2423 		memcpy(v, ia, l);
2424 	else
2425 		memcpy(v, ia4, l);
2426 	return 0;
2427 }
2428 
2429 /*
2430  * Set the IP6 multicast options in response to user setsockopt().
2431  */
2432 static int
ip6_setmoptions(const struct sockopt * sopt,struct inpcb * inp)2433 ip6_setmoptions(const struct sockopt *sopt, struct inpcb *inp)
2434 {
2435 	int error = 0;
2436 	u_int loop, ifindex;
2437 	struct ipv6_mreq mreq;
2438 	struct in6_addr ia;
2439 	struct ifnet *ifp;
2440 	struct ip6_moptions *im6o = in6p_moptions(inp);
2441 	struct in6_multi_mship *imm;
2442 
2443 	KASSERT(inp_locked(inp));
2444 
2445 	if (im6o == NULL) {
2446 		/*
2447 		 * No multicast option buffer attached to the pcb;
2448 		 * allocate one and initialize to default values.
2449 		 */
2450 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2451 		if (im6o == NULL)
2452 			return (ENOBUFS);
2453 		in6p_moptions(inp) = im6o;
2454 		im6o->im6o_multicast_if_index = 0;
2455 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2456 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2457 		LIST_INIT(&im6o->im6o_memberships);
2458 	}
2459 
2460 	switch (sopt->sopt_name) {
2461 
2462 	case IPV6_MULTICAST_IF: {
2463 		int s;
2464 		/*
2465 		 * Select the interface for outgoing multicast packets.
2466 		 */
2467 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2468 		if (error != 0)
2469 			break;
2470 
2471 		s = pserialize_read_enter();
2472 		if (ifindex != 0) {
2473 			if ((ifp = if_byindex(ifindex)) == NULL) {
2474 				pserialize_read_exit(s);
2475 				error = ENXIO;	/* XXX EINVAL? */
2476 				break;
2477 			}
2478 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2479 				pserialize_read_exit(s);
2480 				error = EADDRNOTAVAIL;
2481 				break;
2482 			}
2483 		} else
2484 			ifp = NULL;
2485 		im6o->im6o_multicast_if_index = if_get_index(ifp);
2486 		pserialize_read_exit(s);
2487 		break;
2488 	    }
2489 
2490 	case IPV6_MULTICAST_HOPS:
2491 	    {
2492 		/*
2493 		 * Set the IP6 hoplimit for outgoing multicast packets.
2494 		 */
2495 		int optval;
2496 
2497 		error = sockopt_getint(sopt, &optval);
2498 		if (error != 0)
2499 			break;
2500 
2501 		if (optval < -1 || optval >= 256)
2502 			error = EINVAL;
2503 		else if (optval == -1)
2504 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2505 		else
2506 			im6o->im6o_multicast_hlim = optval;
2507 		break;
2508 	    }
2509 
2510 	case IPV6_MULTICAST_LOOP:
2511 		/*
2512 		 * Set the loopback flag for outgoing multicast packets.
2513 		 * Must be zero or one.
2514 		 */
2515 		error = sockopt_get(sopt, &loop, sizeof(loop));
2516 		if (error != 0)
2517 			break;
2518 		if (loop > 1) {
2519 			error = EINVAL;
2520 			break;
2521 		}
2522 		im6o->im6o_multicast_loop = loop;
2523 		break;
2524 
2525 	case IPV6_JOIN_GROUP: {
2526 		int bound;
2527 		struct psref psref;
2528 		/*
2529 		 * Add a multicast group membership.
2530 		 * Group must be a valid IP6 multicast address.
2531 		 */
2532 		bound = curlwp_bind();
2533 		ifp = NULL;
2534 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
2535 		if (error != 0) {
2536 			KASSERT(ifp == NULL);
2537 			curlwp_bindx(bound);
2538 			return error;
2539 		}
2540 
2541 		if (IN6_IS_ADDR_V4MAPPED(&ia)) {
2542 			error = ip_setmoptions(&inp->inp_moptions, sopt);
2543 			goto put_break;
2544 		}
2545 		/*
2546 		 * See if we found an interface, and confirm that it
2547 		 * supports multicast
2548 		 */
2549 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2550 			error = EADDRNOTAVAIL;
2551 			goto put_break;
2552 		}
2553 
2554 		if (in6_setscope(&ia, ifp, NULL)) {
2555 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2556 			goto put_break;
2557 		}
2558 
2559 		/*
2560 		 * See if the membership already exists.
2561 		 */
2562 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) {
2563 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2564 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2565 			    &ia))
2566 				goto put_break;
2567 		}
2568 		if (imm != NULL) {
2569 			error = EADDRINUSE;
2570 			goto put_break;
2571 		}
2572 		/*
2573 		 * Everything looks good; add a new record to the multicast
2574 		 * address list for the given interface.
2575 		 */
2576 		imm = in6_joingroup(ifp, &ia, &error, 0);
2577 		if (imm == NULL)
2578 			goto put_break;
2579 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2580 	    put_break:
2581 		if_put(ifp, &psref);
2582 		curlwp_bindx(bound);
2583 		break;
2584 	    }
2585 
2586 	case IPV6_LEAVE_GROUP: {
2587 		/*
2588 		 * Drop a multicast group membership.
2589 		 * Group must be a valid IP6 multicast address.
2590 		 */
2591 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2592 		if (error != 0)
2593 			break;
2594 
2595 		if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) {
2596 			error = ip_setmoptions(&inp->inp_moptions, sopt);
2597 			break;
2598 		}
2599 		/*
2600 		 * If an interface address was specified, get a pointer
2601 		 * to its ifnet structure.
2602 		 */
2603 		if (mreq.ipv6mr_interface != 0) {
2604 			if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) {
2605 				error = ENXIO;	/* XXX EINVAL? */
2606 				break;
2607 			}
2608 		} else
2609 			ifp = NULL;
2610 
2611 		/* Fill in the scope zone ID */
2612 		if (ifp) {
2613 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2614 				/* XXX: should not happen */
2615 				error = EADDRNOTAVAIL;
2616 				break;
2617 			}
2618 		} else if (mreq.ipv6mr_interface != 0) {
2619 			/*
2620 			 * XXX: This case would happens when the (positive)
2621 			 * index is in the valid range, but the corresponding
2622 			 * interface has been detached dynamically.  The above
2623 			 * check probably avoids such case to happen here, but
2624 			 * we check it explicitly for safety.
2625 			 */
2626 			error = EADDRNOTAVAIL;
2627 			break;
2628 		} else {	/* ipv6mr_interface == 0 */
2629 			struct sockaddr_in6 sa6_mc;
2630 
2631 			/*
2632 			 * The API spec says as follows:
2633 			 *  If the interface index is specified as 0, the
2634 			 *  system may choose a multicast group membership to
2635 			 *  drop by matching the multicast address only.
2636 			 * On the other hand, we cannot disambiguate the scope
2637 			 * zone unless an interface is provided.  Thus, we
2638 			 * check if there's ambiguity with the default scope
2639 			 * zone as the last resort.
2640 			 */
2641 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2642 			    0, 0, 0);
2643 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2644 			if (error != 0)
2645 				break;
2646 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2647 		}
2648 
2649 		/*
2650 		 * Find the membership in the membership list.
2651 		 */
2652 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) {
2653 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2654 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2655 			    &mreq.ipv6mr_multiaddr))
2656 				break;
2657 		}
2658 		if (imm == NULL) {
2659 			/* Unable to resolve interface */
2660 			error = EADDRNOTAVAIL;
2661 			break;
2662 		}
2663 		/*
2664 		 * Give up the multicast address record to which the
2665 		 * membership points.
2666 		 */
2667 		LIST_REMOVE(imm, i6mm_chain);
2668 		in6_leavegroup(imm);
2669 		/* in6m_ifp should not leave thanks to inp_lock */
2670 		break;
2671 	    }
2672 
2673 	default:
2674 		error = EOPNOTSUPP;
2675 		break;
2676 	}
2677 
2678 	/*
2679 	 * If all options have default values, no need to keep the mbuf.
2680 	 */
2681 	if (im6o->im6o_multicast_if_index == 0 &&
2682 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2683 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2684 	    LIST_EMPTY(&im6o->im6o_memberships)) {
2685 		free(in6p_moptions(inp), M_IPMOPTS);
2686 		in6p_moptions(inp) = NULL;
2687 	}
2688 
2689 	return (error);
2690 }
2691 
2692 /*
2693  * Return the IP6 multicast options in response to user getsockopt().
2694  */
2695 static int
ip6_getmoptions(struct sockopt * sopt,struct inpcb * inp)2696 ip6_getmoptions(struct sockopt *sopt, struct inpcb *inp)
2697 {
2698 	u_int optval;
2699 	int error;
2700 	struct ip6_moptions *im6o = in6p_moptions(inp);
2701 
2702 	switch (sopt->sopt_name) {
2703 	case IPV6_MULTICAST_IF:
2704 		if (im6o == NULL || im6o->im6o_multicast_if_index == 0)
2705 			optval = 0;
2706 		else
2707 			optval = im6o->im6o_multicast_if_index;
2708 
2709 		error = sockopt_set(sopt, &optval, sizeof(optval));
2710 		break;
2711 
2712 	case IPV6_MULTICAST_HOPS:
2713 		if (im6o == NULL)
2714 			optval = ip6_defmcasthlim;
2715 		else
2716 			optval = im6o->im6o_multicast_hlim;
2717 
2718 		error = sockopt_set(sopt, &optval, sizeof(optval));
2719 		break;
2720 
2721 	case IPV6_MULTICAST_LOOP:
2722 		if (im6o == NULL)
2723 			optval = IPV6_DEFAULT_MULTICAST_LOOP;
2724 		else
2725 			optval = im6o->im6o_multicast_loop;
2726 
2727 		error = sockopt_set(sopt, &optval, sizeof(optval));
2728 		break;
2729 
2730 	default:
2731 		error = EOPNOTSUPP;
2732 	}
2733 
2734 	return (error);
2735 }
2736 
2737 /*
2738  * Discard the IP6 multicast options.
2739  */
2740 void
ip6_freemoptions(struct ip6_moptions * im6o)2741 ip6_freemoptions(struct ip6_moptions *im6o)
2742 {
2743 	struct in6_multi_mship *imm, *nimm;
2744 
2745 	if (im6o == NULL)
2746 		return;
2747 
2748 	/* The owner of im6o (inp) should be protected by solock */
2749 	LIST_FOREACH_SAFE(imm, &im6o->im6o_memberships, i6mm_chain, nimm) {
2750 		LIST_REMOVE(imm, i6mm_chain);
2751 		in6_leavegroup(imm);
2752 	}
2753 	free(im6o, M_IPMOPTS);
2754 }
2755 
2756 /*
2757  * Set IPv6 outgoing packet options based on advanced API.
2758  */
2759 int
ip6_setpktopts(struct mbuf * control,struct ip6_pktopts * opt,struct ip6_pktopts * stickyopt,kauth_cred_t cred,int uproto)2760 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2761 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2762 {
2763 	struct cmsghdr *cm = 0;
2764 
2765 	if (control == NULL || opt == NULL)
2766 		return (EINVAL);
2767 
2768 	ip6_initpktopts(opt);
2769 	if (stickyopt) {
2770 		int error;
2771 
2772 		/*
2773 		 * If stickyopt is provided, make a local copy of the options
2774 		 * for this particular packet, then override them by ancillary
2775 		 * objects.
2776 		 * XXX: copypktopts() does not copy the cached route to a next
2777 		 * hop (if any).  This is not very good in terms of efficiency,
2778 		 * but we can allow this since this option should be rarely
2779 		 * used.
2780 		 */
2781 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2782 			return (error);
2783 	}
2784 
2785 	/*
2786 	 * XXX: Currently, we assume all the optional information is stored
2787 	 * in a single mbuf.
2788 	 */
2789 	if (control->m_next)
2790 		return (EINVAL);
2791 
2792 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2793 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2794 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2795 		int error;
2796 
2797 		if (control->m_len < CMSG_LEN(0))
2798 			return (EINVAL);
2799 
2800 		cm = mtod(control, struct cmsghdr *);
2801 		if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > control->m_len)
2802 			return (EINVAL);
2803 		if (cm->cmsg_level != IPPROTO_IPV6)
2804 			continue;
2805 
2806 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2807 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2808 		if (error)
2809 			return (error);
2810 	}
2811 
2812 	return (0);
2813 }
2814 
2815 /*
2816  * Set a particular packet option, as a sticky option or an ancillary data
2817  * item.  "len" can be 0 only when it's a sticky option.
2818  * We have 4 cases of combination of "sticky" and "cmsg":
2819  * "sticky=0, cmsg=0": impossible
2820  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2821  * "sticky=1, cmsg=0": RFC3542 socket option
2822  * "sticky=1, cmsg=1": RFC2292 socket option
2823  */
2824 static int
ip6_setpktopt(int optname,u_char * buf,int len,struct ip6_pktopts * opt,kauth_cred_t cred,int sticky,int cmsg,int uproto)2825 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2826     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2827 {
2828 	int minmtupolicy;
2829 	int error;
2830 
2831 	if (!sticky && !cmsg) {
2832 #ifdef DIAGNOSTIC
2833 		printf("ip6_setpktopt: impossible case\n");
2834 #endif
2835 		return (EINVAL);
2836 	}
2837 
2838 	/*
2839 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2840 	 * not be specified in the context of RFC3542.  Conversely,
2841 	 * RFC3542 types should not be specified in the context of RFC2292.
2842 	 */
2843 	if (!cmsg) {
2844 		switch (optname) {
2845 		case IPV6_2292PKTINFO:
2846 		case IPV6_2292HOPLIMIT:
2847 		case IPV6_2292NEXTHOP:
2848 		case IPV6_2292HOPOPTS:
2849 		case IPV6_2292DSTOPTS:
2850 		case IPV6_2292RTHDR:
2851 		case IPV6_2292PKTOPTIONS:
2852 			return (ENOPROTOOPT);
2853 		}
2854 	}
2855 	if (sticky && cmsg) {
2856 		switch (optname) {
2857 		case IPV6_PKTINFO:
2858 		case IPV6_HOPLIMIT:
2859 		case IPV6_NEXTHOP:
2860 		case IPV6_HOPOPTS:
2861 		case IPV6_DSTOPTS:
2862 		case IPV6_RTHDRDSTOPTS:
2863 		case IPV6_RTHDR:
2864 		case IPV6_USE_MIN_MTU:
2865 		case IPV6_DONTFRAG:
2866 		case IPV6_OTCLASS:
2867 		case IPV6_TCLASS:
2868 		case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */
2869 			return (ENOPROTOOPT);
2870 		}
2871 	}
2872 
2873 	switch (optname) {
2874 #ifdef RFC2292
2875 	case IPV6_2292PKTINFO:
2876 #endif
2877 	case IPV6_PKTINFO:
2878 	{
2879 		struct in6_pktinfo *pktinfo;
2880 
2881 		if (len != sizeof(struct in6_pktinfo))
2882 			return (EINVAL);
2883 
2884 		pktinfo = (struct in6_pktinfo *)buf;
2885 
2886 		/*
2887 		 * An application can clear any sticky IPV6_PKTINFO option by
2888 		 * doing a "regular" setsockopt with ipi6_addr being
2889 		 * in6addr_any and ipi6_ifindex being zero.
2890 		 * [RFC 3542, Section 6]
2891 		 */
2892 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2893 		    pktinfo->ipi6_ifindex == 0 &&
2894 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2895 			ip6_clearpktopts(opt, optname);
2896 			break;
2897 		}
2898 
2899 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2900 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2901 			return (EINVAL);
2902 		}
2903 
2904 		/* Validate the interface index if specified. */
2905 		if (pktinfo->ipi6_ifindex) {
2906 			struct ifnet *ifp;
2907 			int s = pserialize_read_enter();
2908 			ifp = if_byindex(pktinfo->ipi6_ifindex);
2909 			if (ifp == NULL) {
2910 				pserialize_read_exit(s);
2911 				return ENXIO;
2912 			}
2913 			pserialize_read_exit(s);
2914 		}
2915 
2916 		/*
2917 		 * We store the address anyway, and let in6_selectsrc()
2918 		 * validate the specified address.  This is because ipi6_addr
2919 		 * may not have enough information about its scope zone, and
2920 		 * we may need additional information (such as outgoing
2921 		 * interface or the scope zone of a destination address) to
2922 		 * disambiguate the scope.
2923 		 * XXX: the delay of the validation may confuse the
2924 		 * application when it is used as a sticky option.
2925 		 */
2926 		if (opt->ip6po_pktinfo == NULL) {
2927 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2928 			    M_IP6OPT, M_NOWAIT);
2929 			if (opt->ip6po_pktinfo == NULL)
2930 				return (ENOBUFS);
2931 		}
2932 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2933 		break;
2934 	}
2935 
2936 #ifdef RFC2292
2937 	case IPV6_2292HOPLIMIT:
2938 #endif
2939 	case IPV6_HOPLIMIT:
2940 	{
2941 		int *hlimp;
2942 
2943 		/*
2944 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2945 		 * to simplify the ordering among hoplimit options.
2946 		 */
2947 		if (optname == IPV6_HOPLIMIT && sticky)
2948 			return (ENOPROTOOPT);
2949 
2950 		if (len != sizeof(int))
2951 			return (EINVAL);
2952 		hlimp = (int *)buf;
2953 		if (*hlimp < -1 || *hlimp > 255)
2954 			return (EINVAL);
2955 
2956 		opt->ip6po_hlim = *hlimp;
2957 		break;
2958 	}
2959 
2960 	case IPV6_OTCLASS:
2961 		if (len != sizeof(u_int8_t))
2962 			return (EINVAL);
2963 
2964 		opt->ip6po_tclass = *(u_int8_t *)buf;
2965 		break;
2966 
2967 	case IPV6_TCLASS:
2968 	{
2969 		int tclass;
2970 
2971 		if (len != sizeof(int))
2972 			return (EINVAL);
2973 		tclass = *(int *)buf;
2974 		if (tclass < -1 || tclass > 255)
2975 			return (EINVAL);
2976 
2977 		opt->ip6po_tclass = tclass;
2978 		break;
2979 	}
2980 
2981 #ifdef RFC2292
2982 	case IPV6_2292NEXTHOP:
2983 #endif
2984 	case IPV6_NEXTHOP:
2985 		error = kauth_authorize_network(cred,
2986 		    KAUTH_NETWORK_IPV6,
2987 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2988 		if (error)
2989 			return (error);
2990 
2991 		if (len == 0) {	/* just remove the option */
2992 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2993 			break;
2994 		}
2995 
2996 		/* check if cmsg_len is large enough for sa_len */
2997 		if (len < sizeof(struct sockaddr) || len < *buf)
2998 			return (EINVAL);
2999 
3000 		switch (((struct sockaddr *)buf)->sa_family) {
3001 		case AF_INET6:
3002 		{
3003 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3004 
3005 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3006 				return (EINVAL);
3007 
3008 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3009 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3010 				return (EINVAL);
3011 			}
3012 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3013 			    != 0) {
3014 				return (error);
3015 			}
3016 			break;
3017 		}
3018 		case AF_LINK:	/* eventually be supported? */
3019 		default:
3020 			return (EAFNOSUPPORT);
3021 		}
3022 
3023 		/* turn off the previous option, then set the new option. */
3024 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3025 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3026 		if (opt->ip6po_nexthop == NULL)
3027 			return (ENOBUFS);
3028 		memcpy(opt->ip6po_nexthop, buf, *buf);
3029 		break;
3030 
3031 #ifdef RFC2292
3032 	case IPV6_2292HOPOPTS:
3033 #endif
3034 	case IPV6_HOPOPTS:
3035 	{
3036 		struct ip6_hbh *hbh;
3037 		int hbhlen;
3038 
3039 		/*
3040 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3041 		 * options, since per-option restriction has too much
3042 		 * overhead.
3043 		 */
3044 		error = kauth_authorize_network(cred,
3045 		    KAUTH_NETWORK_IPV6,
3046 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
3047 		if (error)
3048 			return (error);
3049 
3050 		if (len == 0) {
3051 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3052 			break;	/* just remove the option */
3053 		}
3054 
3055 		/* message length validation */
3056 		if (len < sizeof(struct ip6_hbh))
3057 			return (EINVAL);
3058 		hbh = (struct ip6_hbh *)buf;
3059 		hbhlen = (hbh->ip6h_len + 1) << 3;
3060 		if (len != hbhlen)
3061 			return (EINVAL);
3062 
3063 		/* turn off the previous option, then set the new option. */
3064 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3065 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3066 		if (opt->ip6po_hbh == NULL)
3067 			return (ENOBUFS);
3068 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
3069 
3070 		break;
3071 	}
3072 
3073 #ifdef RFC2292
3074 	case IPV6_2292DSTOPTS:
3075 #endif
3076 	case IPV6_DSTOPTS:
3077 	case IPV6_RTHDRDSTOPTS:
3078 	{
3079 		struct ip6_dest *dest, **newdest = NULL;
3080 		int destlen;
3081 
3082 		/* XXX: see the comment for IPV6_HOPOPTS */
3083 		error = kauth_authorize_network(cred,
3084 		    KAUTH_NETWORK_IPV6,
3085 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
3086 		if (error)
3087 			return (error);
3088 
3089 		if (len == 0) {
3090 			ip6_clearpktopts(opt, optname);
3091 			break;	/* just remove the option */
3092 		}
3093 
3094 		/* message length validation */
3095 		if (len < sizeof(struct ip6_dest))
3096 			return (EINVAL);
3097 		dest = (struct ip6_dest *)buf;
3098 		destlen = (dest->ip6d_len + 1) << 3;
3099 		if (len != destlen)
3100 			return (EINVAL);
3101 		/*
3102 		 * Determine the position that the destination options header
3103 		 * should be inserted; before or after the routing header.
3104 		 */
3105 		switch (optname) {
3106 		case IPV6_2292DSTOPTS:
3107 			/*
3108 			 * The old advanced API is ambiguous on this point.
3109 			 * Our approach is to determine the position based
3110 			 * according to the existence of a routing header.
3111 			 * Note, however, that this depends on the order of the
3112 			 * extension headers in the ancillary data; the 1st
3113 			 * part of the destination options header must appear
3114 			 * before the routing header in the ancillary data,
3115 			 * too.
3116 			 * RFC3542 solved the ambiguity by introducing
3117 			 * separate ancillary data or option types.
3118 			 */
3119 			if (opt->ip6po_rthdr == NULL)
3120 				newdest = &opt->ip6po_dest1;
3121 			else
3122 				newdest = &opt->ip6po_dest2;
3123 			break;
3124 		case IPV6_RTHDRDSTOPTS:
3125 			newdest = &opt->ip6po_dest1;
3126 			break;
3127 		case IPV6_DSTOPTS:
3128 			newdest = &opt->ip6po_dest2;
3129 			break;
3130 		}
3131 
3132 		/* turn off the previous option, then set the new option. */
3133 		ip6_clearpktopts(opt, optname);
3134 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3135 		if (*newdest == NULL)
3136 			return (ENOBUFS);
3137 		memcpy(*newdest, dest, destlen);
3138 
3139 		break;
3140 	}
3141 
3142 #ifdef RFC2292
3143 	case IPV6_2292RTHDR:
3144 #endif
3145 	case IPV6_RTHDR:
3146 	{
3147 		struct ip6_rthdr *rth;
3148 		int rthlen;
3149 
3150 		if (len == 0) {
3151 			ip6_clearpktopts(opt, IPV6_RTHDR);
3152 			break;	/* just remove the option */
3153 		}
3154 
3155 		/* message length validation */
3156 		if (len < sizeof(struct ip6_rthdr))
3157 			return (EINVAL);
3158 		rth = (struct ip6_rthdr *)buf;
3159 		rthlen = (rth->ip6r_len + 1) << 3;
3160 		if (len != rthlen)
3161 			return (EINVAL);
3162 		switch (rth->ip6r_type) {
3163 		case IPV6_RTHDR_TYPE_0:
3164 			/* Dropped, RFC5095. */
3165 		default:
3166 			return (EINVAL);	/* not supported */
3167 		}
3168 		/* turn off the previous option */
3169 		ip6_clearpktopts(opt, IPV6_RTHDR);
3170 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3171 		if (opt->ip6po_rthdr == NULL)
3172 			return (ENOBUFS);
3173 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3174 		break;
3175 	}
3176 
3177 	case IPV6_USE_MIN_MTU:
3178 		if (len != sizeof(int))
3179 			return (EINVAL);
3180 		minmtupolicy = *(int *)buf;
3181 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3182 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3183 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3184 			return (EINVAL);
3185 		}
3186 		opt->ip6po_minmtu = minmtupolicy;
3187 		break;
3188 
3189 	case IPV6_DONTFRAG:
3190 		if (len != sizeof(int))
3191 			return (EINVAL);
3192 
3193 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3194 			/*
3195 			 * we ignore this option for TCP sockets.
3196 			 * (RFC3542 leaves this case unspecified.)
3197 			 */
3198 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3199 		} else
3200 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3201 		break;
3202 
3203 	case IPV6_PREFER_TEMPADDR:
3204 	{
3205 		int preftemp;
3206 
3207 		if (len != sizeof(int))
3208 			return (EINVAL);
3209 		preftemp = *(int *)buf;
3210 		switch (preftemp) {
3211 		case IP6PO_TEMPADDR_SYSTEM:
3212 		case IP6PO_TEMPADDR_NOTPREFER:
3213 		case IP6PO_TEMPADDR_PREFER:
3214 			break;
3215 		default:
3216 			return (EINVAL);
3217 		}
3218 		opt->ip6po_prefer_tempaddr = preftemp;
3219 		break;
3220 	}
3221 
3222 	default:
3223 		return (ENOPROTOOPT);
3224 	} /* end of switch */
3225 
3226 	return (0);
3227 }
3228 
3229 /*
3230  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3231  * packet to the input queue of a specified interface.  Note that this
3232  * calls the output routine of the loopback "driver", but with an interface
3233  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3234  */
3235 void
ip6_mloopback(struct ifnet * ifp,struct mbuf * m,const struct sockaddr_in6 * dst)3236 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3237 	const struct sockaddr_in6 *dst)
3238 {
3239 	struct mbuf *copym;
3240 	struct ip6_hdr *ip6;
3241 
3242 	copym = m_copypacket(m, M_DONTWAIT);
3243 	if (copym == NULL)
3244 		return;
3245 
3246 	/*
3247 	 * Make sure to deep-copy IPv6 header portion in case the data
3248 	 * is in an mbuf cluster, so that we can safely override the IPv6
3249 	 * header portion later.
3250 	 */
3251 	if ((copym->m_flags & M_EXT) != 0 ||
3252 	    copym->m_len < sizeof(struct ip6_hdr)) {
3253 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3254 		if (copym == NULL)
3255 			return;
3256 	}
3257 
3258 #ifdef DIAGNOSTIC
3259 	if (copym->m_len < sizeof(*ip6)) {
3260 		m_freem(copym);
3261 		return;
3262 	}
3263 #endif
3264 
3265 	ip6 = mtod(copym, struct ip6_hdr *);
3266 	/*
3267 	 * clear embedded scope identifiers if necessary.
3268 	 * in6_clearscope will touch the addresses only when necessary.
3269 	 */
3270 	in6_clearscope(&ip6->ip6_src);
3271 	in6_clearscope(&ip6->ip6_dst);
3272 
3273 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3274 }
3275 
3276 /*
3277  * Chop IPv6 header off from the payload.
3278  */
3279 static int
ip6_splithdr(struct mbuf * m,struct ip6_exthdrs * exthdrs)3280 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3281 {
3282 	struct mbuf *mh;
3283 	struct ip6_hdr *ip6;
3284 
3285 	ip6 = mtod(m, struct ip6_hdr *);
3286 	if (m->m_len > sizeof(*ip6)) {
3287 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3288 		if (mh == NULL) {
3289 			m_freem(m);
3290 			return ENOBUFS;
3291 		}
3292 		m_move_pkthdr(mh, m);
3293 		m_align(mh, sizeof(*ip6));
3294 		m->m_len -= sizeof(*ip6);
3295 		m->m_data += sizeof(*ip6);
3296 		mh->m_next = m;
3297 		mh->m_len = sizeof(*ip6);
3298 		memcpy(mtod(mh, void *), (void *)ip6, sizeof(*ip6));
3299 		m = mh;
3300 	}
3301 	exthdrs->ip6e_ip6 = m;
3302 	return 0;
3303 }
3304 
3305 /*
3306  * Compute IPv6 extension header length.
3307  */
3308 int
ip6_optlen(struct inpcb * inp)3309 ip6_optlen(struct inpcb *inp)
3310 {
3311 	int len;
3312 
3313 	if (!in6p_outputopts(inp))
3314 		return 0;
3315 
3316 	len = 0;
3317 #define elen(x) \
3318     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3319 
3320 	len += elen(in6p_outputopts(inp)->ip6po_hbh);
3321 	len += elen(in6p_outputopts(inp)->ip6po_dest1);
3322 	len += elen(in6p_outputopts(inp)->ip6po_rthdr);
3323 	len += elen(in6p_outputopts(inp)->ip6po_dest2);
3324 	return len;
3325 #undef elen
3326 }
3327 
3328 /*
3329  * Ensure sending address is valid.
3330  * Returns 0 on success, -1 if an error should be sent back or 1
3331  * if the packet could be dropped without error (protocol dependent).
3332  */
3333 static int
ip6_ifaddrvalid(const struct in6_addr * src,const struct in6_addr * dst)3334 ip6_ifaddrvalid(const struct in6_addr *src, const struct in6_addr *dst)
3335 {
3336 	struct sockaddr_in6 sin6;
3337 	int s, error;
3338 	struct ifaddr *ifa;
3339 	struct in6_ifaddr *ia6;
3340 
3341 	if (IN6_IS_ADDR_UNSPECIFIED(src))
3342 		return 0;
3343 
3344 	memset(&sin6, 0, sizeof(sin6));
3345 	sin6.sin6_family = AF_INET6;
3346 	sin6.sin6_len = sizeof(sin6);
3347 	sin6.sin6_addr = *src;
3348 
3349 	s = pserialize_read_enter();
3350 	ifa = ifa_ifwithaddr(sin6tosa(&sin6));
3351 	if ((ia6 = ifatoia6(ifa)) == NULL ||
3352 	    ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED))
3353 		error = -1;
3354 	else if (ia6->ia6_flags & IN6_IFF_TENTATIVE)
3355 		error = 1;
3356 	else if (ia6->ia6_flags & IN6_IFF_DETACHED &&
3357 	    (sin6.sin6_addr = *dst, ifa_ifwithaddr(sin6tosa(&sin6)) == NULL))
3358 		/* Allow internal traffic to DETACHED addresses */
3359 		error = 1;
3360 	else
3361 		error = 0;
3362 	pserialize_read_exit(s);
3363 
3364 	return error;
3365 }
3366