xref: /freebsd/sys/netinet6/ip6_output.c (revision 47fb51847fdea3f1cce841b5f2bbbcd6f8a04ee0)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  */
62 
63 #include "opt_inet.h"
64 #include "opt_inet6.h"
65 #include "opt_ipsec.h"
66 #include "opt_kern_tls.h"
67 #include "opt_ratelimit.h"
68 #include "opt_route.h"
69 #include "opt_rss.h"
70 #include "opt_sctp.h"
71 
72 #include <sys/param.h>
73 #include <sys/kernel.h>
74 #include <sys/ktls.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85 
86 #include <machine/in_cksum.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_private.h>
91 #include <net/if_vlan_var.h>
92 #include <net/if_llatbl.h>
93 #include <net/ethernet.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/route/nhop.h>
97 #include <net/pfil.h>
98 #include <net/rss_config.h>
99 #include <net/vnet.h>
100 
101 #include <netinet/in.h>
102 #include <netinet/in_var.h>
103 #include <netinet/ip_var.h>
104 #include <netinet6/in6_fib.h>
105 #include <netinet6/in6_var.h>
106 #include <netinet/ip6.h>
107 #include <netinet/icmp6.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet/in_pcb.h>
110 #include <netinet/tcp_var.h>
111 #include <netinet6/nd6.h>
112 #include <netinet6/in6_rss.h>
113 #include <netinet6/ip6_mroute.h>
114 
115 #include <netipsec/ipsec_support.h>
116 #if defined(SCTP) || defined(SCTP_SUPPORT)
117 #include <netinet/sctp.h>
118 #include <netinet/sctp_crc32.h>
119 #endif
120 
121 #include <netinet6/scope6_var.h>
122 
123 extern int in6_mcast_loop;
124 
125 struct ip6_exthdrs {
126 	struct mbuf *ip6e_ip6;
127 	struct mbuf *ip6e_hbh;
128 	struct mbuf *ip6e_dest1;
129 	struct mbuf *ip6e_rthdr;
130 	struct mbuf *ip6e_dest2;
131 };
132 
133 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
134 
135 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
136 			   struct ucred *, int);
137 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
138 	struct socket *, struct sockopt *);
139 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
140 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
141 	struct ucred *, int, int, int);
142 
143 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
144 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
145 	struct ip6_frag **);
146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
147 static void ip6_getpmtu(struct route_in6 *, int,
148 	struct ifnet *, const struct in6_addr *, u_long *, u_int, u_int);
149 static void ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
150 	u_long *, u_int);
151 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
152 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
153 
154 /*
155  * Make an extension header from option data.  hp is the source,
156  * mp is the destination, and _ol is the optlen.
157  */
158 #define	MAKE_EXTHDR(hp, mp, _ol)					\
159     do {								\
160 	struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
161 	error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
162 	    ((eh)->ip6e_len + 1) << 3);				\
163 	if (error)						\
164 		goto freehdrs;					\
165 	(_ol) += (*(mp))->m_len;				\
166     } while (/*CONSTCOND*/ 0)
167 
168 /*
169  * Form a chain of extension headers.
170  * m is the extension header mbuf
171  * mp is the previous mbuf in the chain
172  * p is the next header
173  * i is the type of option.
174  */
175 #define MAKE_CHAIN(m, mp, p, i)\
176     do {\
177 	if (m) {\
178 		if (!hdrsplit) \
179 			panic("%s:%d: assumption failed: "\
180 			    "hdr not split: hdrsplit %d exthdrs %p",\
181 			    __func__, __LINE__, hdrsplit, &exthdrs);\
182 		*mtod((m), u_char *) = *(p);\
183 		*(p) = (i);\
184 		p = mtod((m), u_char *);\
185 		(m)->m_next = (mp)->m_next;\
186 		(mp)->m_next = (m);\
187 		(mp) = (m);\
188 	}\
189     } while (/*CONSTCOND*/ 0)
190 
191 void
192 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
193 {
194 	u_short csum;
195 
196 	csum = in_cksum_skip(m, offset + plen, offset);
197 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
198 		csum = 0xffff;
199 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
200 
201 	if (offset + sizeof(csum) > m->m_len)
202 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
203 	else
204 		*(u_short *)mtodo(m, offset) = csum;
205 }
206 
207 static void
208 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
209     int plen, int optlen)
210 {
211 
212 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
213 	    "csum_flags %#x",
214 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
215 
216 	if (csum_flags & CSUM_DELAY_DATA_IPV6) {
217 		in6_delayed_cksum(m, plen - optlen,
218 		    sizeof(struct ip6_hdr) + optlen);
219 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
220 	}
221 #if defined(SCTP) || defined(SCTP_SUPPORT)
222 	if (csum_flags & CSUM_SCTP_IPV6) {
223 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
224 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
225 	}
226 #endif
227 }
228 
229 int
230 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
231     int fraglen , uint32_t id)
232 {
233 	struct mbuf *m, **mnext, *m_frgpart;
234 	struct ip6_hdr *ip6, *mhip6;
235 	struct ip6_frag *ip6f;
236 	int off;
237 	int error;
238 	int tlen = m0->m_pkthdr.len;
239 
240 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
241 
242 	m = m0;
243 	ip6 = mtod(m, struct ip6_hdr *);
244 	mnext = &m->m_nextpkt;
245 
246 	for (off = hlen; off < tlen; off += fraglen) {
247 		m = m_gethdr(M_NOWAIT, MT_DATA);
248 		if (!m) {
249 			IP6STAT_INC(ip6s_odropped);
250 			return (ENOBUFS);
251 		}
252 
253 		/*
254 		 * Make sure the complete packet header gets copied
255 		 * from the originating mbuf to the newly created
256 		 * mbuf. This also ensures that existing firewall
257 		 * classification(s), VLAN tags and so on get copied
258 		 * to the resulting fragmented packet(s):
259 		 */
260 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
261 			m_free(m);
262 			IP6STAT_INC(ip6s_odropped);
263 			return (ENOBUFS);
264 		}
265 
266 		*mnext = m;
267 		mnext = &m->m_nextpkt;
268 		m->m_data += max_linkhdr;
269 		mhip6 = mtod(m, struct ip6_hdr *);
270 		*mhip6 = *ip6;
271 		m->m_len = sizeof(*mhip6);
272 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
273 		if (error) {
274 			IP6STAT_INC(ip6s_odropped);
275 			return (error);
276 		}
277 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
278 		if (off + fraglen >= tlen)
279 			fraglen = tlen - off;
280 		else
281 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
282 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
283 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
284 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
285 			IP6STAT_INC(ip6s_odropped);
286 			return (ENOBUFS);
287 		}
288 		m_cat(m, m_frgpart);
289 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
290 		ip6f->ip6f_reserved = 0;
291 		ip6f->ip6f_ident = id;
292 		ip6f->ip6f_nxt = nextproto;
293 		IP6STAT_INC(ip6s_ofragments);
294 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
295 	}
296 
297 	return (0);
298 }
299 
300 static int
301 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
302     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
303     bool stamp_tag)
304 {
305 #ifdef KERN_TLS
306 	struct ktls_session *tls = NULL;
307 #endif
308 	struct m_snd_tag *mst;
309 	int error;
310 
311 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
312 	mst = NULL;
313 
314 #ifdef KERN_TLS
315 	/*
316 	 * If this is an unencrypted TLS record, save a reference to
317 	 * the record.  This local reference is used to call
318 	 * ktls_output_eagain after the mbuf has been freed (thus
319 	 * dropping the mbuf's reference) in if_output.
320 	 */
321 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
322 		tls = ktls_hold(m->m_next->m_epg_tls);
323 		mst = tls->snd_tag;
324 
325 		/*
326 		 * If a TLS session doesn't have a valid tag, it must
327 		 * have had an earlier ifp mismatch, so drop this
328 		 * packet.
329 		 */
330 		if (mst == NULL) {
331 			m_freem(m);
332 			error = EAGAIN;
333 			goto done;
334 		}
335 		/*
336 		 * Always stamp tags that include NIC ktls.
337 		 */
338 		stamp_tag = true;
339 	}
340 #endif
341 #ifdef RATELIMIT
342 	if (inp != NULL && mst == NULL) {
343 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
344 		    (inp->inp_snd_tag != NULL &&
345 		    inp->inp_snd_tag->ifp != ifp))
346 			in_pcboutput_txrtlmt(inp, ifp, m);
347 
348 		if (inp->inp_snd_tag != NULL)
349 			mst = inp->inp_snd_tag;
350 	}
351 #endif
352 	if (stamp_tag && mst != NULL) {
353 		KASSERT(m->m_pkthdr.rcvif == NULL,
354 		    ("trying to add a send tag to a forwarded packet"));
355 		if (mst->ifp != ifp) {
356 			m_freem(m);
357 			error = EAGAIN;
358 			goto done;
359 		}
360 
361 		/* stamp send tag on mbuf */
362 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
363 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
364 	}
365 
366 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
367 
368 done:
369 	/* Check for route change invalidating send tags. */
370 #ifdef KERN_TLS
371 	if (tls != NULL) {
372 		if (error == EAGAIN)
373 			error = ktls_output_eagain(inp, tls);
374 		ktls_free(tls);
375 	}
376 #endif
377 #ifdef RATELIMIT
378 	if (error == EAGAIN)
379 		in_pcboutput_eagain(inp);
380 #endif
381 	return (error);
382 }
383 
384 /*
385  * IP6 output.
386  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
387  * nxt, hlim, src, dst).
388  * This function may modify ver and hlim only.
389  * The mbuf chain containing the packet will be freed.
390  * The mbuf opt, if present, will not be freed.
391  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
392  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
393  * then result of route lookup is stored in ro->ro_nh.
394  *
395  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
396  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
397  *
398  * ifpp - XXX: just for statistics
399  */
400 int
401 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
402     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
403     struct ifnet **ifpp, struct inpcb *inp)
404 {
405 	struct ip6_hdr *ip6;
406 	struct ifnet *ifp, *origifp;
407 	struct mbuf *m = m0;
408 	struct mbuf *mprev;
409 	struct route_in6 *ro_pmtu;
410 	struct nhop_object *nh;
411 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
412 	struct in6_addr odst;
413 	u_char *nexthdrp;
414 	int tlen, len;
415 	int error = 0;
416 	int vlan_pcp = -1;
417 	struct in6_ifaddr *ia = NULL;
418 	u_long mtu;
419 	int dontfrag;
420 	u_int32_t optlen, plen = 0, unfragpartlen;
421 	struct ip6_exthdrs exthdrs;
422 	struct in6_addr src0, dst0;
423 	u_int32_t zone;
424 	bool hdrsplit;
425 	int sw_csum, tso;
426 	int needfiblookup;
427 	uint32_t fibnum;
428 	struct m_tag *fwd_tag = NULL;
429 	uint32_t id;
430 	uint32_t optvalid;
431 
432 	NET_EPOCH_ASSERT();
433 
434 	if (inp != NULL) {
435 		INP_LOCK_ASSERT(inp);
436 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
437 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
438 			/* Unconditionally set flowid. */
439 			m->m_pkthdr.flowid = inp->inp_flowid;
440 			M_HASHTYPE_SET(m, inp->inp_flowtype);
441 		}
442 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
443 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
444 			    INP_2PCP_SHIFT;
445 #ifdef NUMA
446 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
447 #endif
448 	}
449 
450 	/* Source address validation. */
451 	ip6 = mtod(m, struct ip6_hdr *);
452 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
453 	    (flags & IPV6_UNSPECSRC) == 0) {
454 		error = EOPNOTSUPP;
455 		IP6STAT_INC(ip6s_badscope);
456 		goto bad;
457 	}
458 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
459 		error = EOPNOTSUPP;
460 		IP6STAT_INC(ip6s_badscope);
461 		goto bad;
462 	}
463 
464 	/*
465 	 * If we are given packet options to add extension headers prepare them.
466 	 * Calculate the total length of the extension header chain.
467 	 * Keep the length of the unfragmentable part for fragmentation.
468 	 */
469 	bzero(&exthdrs, sizeof(exthdrs));
470 	optlen = optvalid = 0;
471 	unfragpartlen = sizeof(struct ip6_hdr);
472 	if (opt) {
473 		optvalid = opt->ip6po_valid;
474 
475 		/* Hop-by-Hop options header. */
476 		if ((optvalid & IP6PO_VALID_HBH) != 0)
477 			MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
478 
479 		/* Destination options header (1st part). */
480 		if ((optvalid & IP6PO_VALID_RHINFO) != 0) {
481 #ifndef RTHDR_SUPPORT_IMPLEMENTED
482 			/*
483 			 * If there is a routing header, discard the packet
484 			 * right away here. RH0/1 are obsolete and we do not
485 			 * currently support RH2/3/4.
486 			 * People trying to use RH253/254 may want to disable
487 			 * this check.
488 			 * The moment we do support any routing header (again)
489 			 * this block should check the routing type more
490 			 * selectively.
491 			 */
492 			error = EINVAL;
493 			goto bad;
494 #endif
495 
496 			/*
497 			 * Destination options header (1st part).
498 			 * This only makes sense with a routing header.
499 			 * See Section 9.2 of RFC 3542.
500 			 * Disabling this part just for MIP6 convenience is
501 			 * a bad idea.  We need to think carefully about a
502 			 * way to make the advanced API coexist with MIP6
503 			 * options, which might automatically be inserted in
504 			 * the kernel.
505 			 */
506 			if ((optvalid & IP6PO_VALID_DEST1) != 0)
507 				MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
508 				    optlen);
509 		}
510 		/* Routing header. */
511 		if ((optvalid & IP6PO_VALID_RHINFO) != 0)
512 			MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
513 
514 		unfragpartlen += optlen;
515 
516 		/*
517 		 * NOTE: we don't add AH/ESP length here (done in
518 		 * ip6_ipsec_output()).
519 		 */
520 
521 		/* Destination options header (2nd part). */
522 		if ((optvalid & IP6PO_VALID_DEST2) != 0)
523 			MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
524 	}
525 
526 	/*
527 	 * If there is at least one extension header,
528 	 * separate IP6 header from the payload.
529 	 */
530 	hdrsplit = false;
531 	if (optlen) {
532 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
533 			m = NULL;
534 			goto freehdrs;
535 		}
536 		m = exthdrs.ip6e_ip6;
537 		ip6 = mtod(m, struct ip6_hdr *);
538 		hdrsplit = true;
539 	}
540 
541 	/* Adjust mbuf packet header length. */
542 	m->m_pkthdr.len += optlen;
543 	plen = m->m_pkthdr.len - sizeof(*ip6);
544 
545 	if (plen > IPV6_MAXPACKET) {
546 		error = EMSGSIZE;
547 		goto freehdrs;
548 	} else
549 		ip6->ip6_plen = htons(plen);
550 	nexthdrp = &ip6->ip6_nxt;
551 
552 	if (optlen) {
553 		/*
554 		 * Concatenate headers and fill in next header fields.
555 		 * Here we have, on "m"
556 		 *	IPv6 payload
557 		 * and we insert headers accordingly.
558 		 * Finally, we should be getting:
559 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
560 		 *
561 		 * During the header composing process "m" points to IPv6
562 		 * header.  "mprev" points to an extension header prior to esp.
563 		 */
564 		mprev = m;
565 
566 		/*
567 		 * We treat dest2 specially.  This makes IPsec processing
568 		 * much easier.  The goal here is to make mprev point the
569 		 * mbuf prior to dest2.
570 		 *
571 		 * Result: IPv6 dest2 payload.
572 		 * m and mprev will point to IPv6 header.
573 		 */
574 		if (exthdrs.ip6e_dest2) {
575 			if (!hdrsplit)
576 				panic("%s:%d: assumption failed: "
577 				    "hdr not split: hdrsplit %d exthdrs %p",
578 				    __func__, __LINE__, hdrsplit, &exthdrs);
579 			exthdrs.ip6e_dest2->m_next = m->m_next;
580 			m->m_next = exthdrs.ip6e_dest2;
581 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
582 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
583 		}
584 
585 		/*
586 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
587 		 * m will point to IPv6 header.  mprev will point to the
588 		 * extension header prior to dest2 (rthdr in the above case).
589 		 */
590 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
591 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
592 			   IPPROTO_DSTOPTS);
593 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
594 			   IPPROTO_ROUTING);
595 	}
596 
597 	IP6STAT_INC(ip6s_localout);
598 
599 	/* Route packet. */
600 	ro_pmtu = ro;
601 	if ((optvalid & IP6PO_VALID_RHINFO) != 0)
602 		ro = &opt->ip6po_route;
603 	if (ro != NULL)
604 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
605 	else
606 		dst = &sin6;
607 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
608 
609 again:
610 	/*
611 	 * If specified, try to fill in the traffic class field.
612 	 * Do not override if a non-zero value is already set.
613 	 * We check the diffserv field and the ECN field separately.
614 	 */
615 	if ((optvalid & IP6PO_VALID_TC) != 0){
616 		int mask = 0;
617 
618 		if (IPV6_DSCP(ip6) == 0)
619 			mask |= 0xfc;
620 		if (IPV6_ECN(ip6) == 0)
621 			mask |= 0x03;
622 		if (mask != 0)
623 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
624 	}
625 
626 	/* Fill in or override the hop limit field, if necessary. */
627 	if ((optvalid & IP6PO_VALID_HLIM) != 0)
628 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
629 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
630 		if (im6o != NULL)
631 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
632 		else
633 			ip6->ip6_hlim = V_ip6_defmcasthlim;
634 	}
635 
636 	if (ro == NULL || ro->ro_nh == NULL) {
637 		bzero(dst, sizeof(*dst));
638 		dst->sin6_family = AF_INET6;
639 		dst->sin6_len = sizeof(*dst);
640 		dst->sin6_addr = ip6->ip6_dst;
641 	}
642 	/*
643 	 * Validate route against routing table changes.
644 	 * Make sure that the address family is set in route.
645 	 */
646 	nh = NULL;
647 	ifp = NULL;
648 	mtu = 0;
649 	if (ro != NULL) {
650 		if (ro->ro_nh != NULL && inp != NULL) {
651 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
652 			NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
653 			    fibnum);
654 		}
655 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
656 		    (!NH_IS_VALID(ro->ro_nh) ||
657 		    ro->ro_dst.sin6_family != AF_INET6 ||
658 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
659 			RO_INVALIDATE_CACHE(ro);
660 
661 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
662 		    ro->ro_dst.sin6_family == AF_INET6 &&
663 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
664 			/* Nexthop is valid and contains valid ifp */
665 			nh = ro->ro_nh;
666 		} else {
667 			if (ro->ro_lle)
668 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
669 			ro->ro_lle = NULL;
670 			if (fwd_tag == NULL) {
671 				bzero(&dst_sa, sizeof(dst_sa));
672 				dst_sa.sin6_family = AF_INET6;
673 				dst_sa.sin6_len = sizeof(dst_sa);
674 				dst_sa.sin6_addr = ip6->ip6_dst;
675 			}
676 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
677 			    &nh, fibnum, m->m_pkthdr.flowid);
678 			if (error != 0) {
679 				IP6STAT_INC(ip6s_noroute);
680 				if (ifp != NULL)
681 					in6_ifstat_inc(ifp, ifs6_out_discard);
682 				goto bad;
683 			}
684 			/*
685 			 * At this point at least @ifp is not NULL
686 			 * Can be the case when dst is multicast, link-local or
687 			 * interface is explicitly specificed by the caller.
688 			 */
689 		}
690 		if (nh == NULL) {
691 			/*
692 			 * If in6_selectroute() does not return a nexthop
693 			 * dst may not have been updated.
694 			 */
695 			*dst = dst_sa;	/* XXX */
696 			origifp = ifp;
697 			mtu = ifp->if_mtu;
698 		} else {
699 			ifp = nh->nh_ifp;
700 			origifp = nh->nh_aifp;
701 			ia = (struct in6_ifaddr *)(nh->nh_ifa);
702 			counter_u64_add(nh->nh_pksent, 1);
703 		}
704 	} else {
705 		struct nhop_object *nh;
706 		struct in6_addr kdst;
707 		uint32_t scopeid;
708 
709 		if (fwd_tag == NULL) {
710 			bzero(&dst_sa, sizeof(dst_sa));
711 			dst_sa.sin6_family = AF_INET6;
712 			dst_sa.sin6_len = sizeof(dst_sa);
713 			dst_sa.sin6_addr = ip6->ip6_dst;
714 		}
715 
716 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
717 		    im6o != NULL &&
718 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
719 			/* We do not need a route lookup. */
720 			*dst = dst_sa;	/* XXX */
721 			origifp = ifp;
722 			goto nonh6lookup;
723 		}
724 
725 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
726 
727 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
728 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
729 			if (scopeid > 0) {
730 				ifp = in6_getlinkifnet(scopeid);
731 				if (ifp == NULL) {
732 					error = EHOSTUNREACH;
733 					goto bad;
734 				}
735 				*dst = dst_sa;	/* XXX */
736 				origifp = ifp;
737 				goto nonh6lookup;
738 			}
739 		}
740 
741 		nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
742 		    m->m_pkthdr.flowid);
743 		if (nh == NULL) {
744 			IP6STAT_INC(ip6s_noroute);
745 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
746 			error = EHOSTUNREACH;
747 			goto bad;
748 		}
749 
750 		ifp = nh->nh_ifp;
751 		origifp = nh->nh_aifp;
752 		ia = ifatoia6(nh->nh_ifa);
753 		if (nh->nh_flags & NHF_GATEWAY)
754 			dst->sin6_addr = nh->gw6_sa.sin6_addr;
755 		else if (fwd_tag != NULL)
756 			dst->sin6_addr = dst_sa.sin6_addr;
757 nonh6lookup:
758 		;
759 	}
760 	/*
761 	 * At this point ifp MUST be pointing to the valid transmit ifp.
762 	 * origifp MUST be valid and pointing to either the same ifp or,
763 	 * in case of loopback output, to the interface which ip6_src
764 	 * belongs to.
765 	 * Examples:
766 	 *  fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
767 	 *  fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
768 	 *  ::1 -> ::1 -> ifp=lo0, origifp=lo0
769 	 *
770 	 * mtu can be 0 and will be refined later.
771 	 */
772 	KASSERT((ifp != NULL), ("output interface must not be NULL"));
773 	KASSERT((origifp != NULL), ("output address interface must not be NULL"));
774 
775 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
776 	/*
777 	 * IPSec checking which handles several cases.
778 	 * FAST IPSEC: We re-injected the packet.
779 	 * XXX: need scope argument.
780 	 */
781 	if (IPSEC_ENABLED(ipv6)) {
782 		if ((error = IPSEC_OUTPUT(ipv6, ifp, m, inp, mtu == 0 ?
783 		    ifp->if_mtu : mtu)) != 0) {
784 			if (error == EINPROGRESS)
785 				error = 0;
786 			goto done;
787 		}
788 	}
789 #endif /* IPSEC */
790 
791 	if ((flags & IPV6_FORWARDING) == 0) {
792 		/* XXX: the FORWARDING flag can be set for mrouting. */
793 		in6_ifstat_inc(ifp, ifs6_out_request);
794 	}
795 
796 	/* Setup data structures for scope ID checks. */
797 	src0 = ip6->ip6_src;
798 	bzero(&src_sa, sizeof(src_sa));
799 	src_sa.sin6_family = AF_INET6;
800 	src_sa.sin6_len = sizeof(src_sa);
801 	src_sa.sin6_addr = ip6->ip6_src;
802 
803 	dst0 = ip6->ip6_dst;
804 	/* Re-initialize to be sure. */
805 	bzero(&dst_sa, sizeof(dst_sa));
806 	dst_sa.sin6_family = AF_INET6;
807 	dst_sa.sin6_len = sizeof(dst_sa);
808 	dst_sa.sin6_addr = ip6->ip6_dst;
809 
810 	/* Check for valid scope ID. */
811 	if (in6_setscope(&src0, origifp, &zone) == 0 &&
812 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
813 	    in6_setscope(&dst0, origifp, &zone) == 0 &&
814 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
815 		/*
816 		 * The outgoing interface is in the zone of the source
817 		 * and destination addresses.
818 		 *
819 		 */
820 	} else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
821 	    sa6_recoverscope(&src_sa) != 0 ||
822 	    sa6_recoverscope(&dst_sa) != 0 ||
823 	    dst_sa.sin6_scope_id == 0 ||
824 	    (src_sa.sin6_scope_id != 0 &&
825 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
826 	    ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
827 		/*
828 		 * If the destination network interface is not a
829 		 * loopback interface, or the destination network
830 		 * address has no scope ID, or the source address has
831 		 * a scope ID set which is different from the
832 		 * destination address one, or there is no network
833 		 * interface representing this scope ID, the address
834 		 * pair is considered invalid.
835 		 */
836 		IP6STAT_INC(ip6s_badscope);
837 		in6_ifstat_inc(origifp, ifs6_out_discard);
838 		if (error == 0)
839 			error = EHOSTUNREACH; /* XXX */
840 		goto bad;
841 	}
842 	/* All scope ID checks are successful. */
843 
844 	if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
845 		if ((optvalid & IP6PO_VALID_NHINFO) != 0) {
846 			/*
847 			 * The nexthop is explicitly specified by the
848 			 * application.  We assume the next hop is an IPv6
849 			 * address.
850 			 */
851 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
852 		}
853 		else if ((nh->nh_flags & NHF_GATEWAY))
854 			dst = &nh->gw6_sa;
855 	}
856 
857 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
858 		m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
859 	} else {
860 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
861 		in6_ifstat_inc(ifp, ifs6_out_mcast);
862 
863 		/* Confirm that the outgoing interface supports multicast. */
864 		if (!(ifp->if_flags & IFF_MULTICAST)) {
865 			IP6STAT_INC(ip6s_noroute);
866 			in6_ifstat_inc(ifp, ifs6_out_discard);
867 			error = ENETUNREACH;
868 			goto bad;
869 		}
870 		if ((im6o == NULL && in6_mcast_loop) ||
871 		    (im6o && im6o->im6o_multicast_loop)) {
872 			/*
873 			 * Loop back multicast datagram if not expressly
874 			 * forbidden to do so, even if we have not joined
875 			 * the address; protocols will filter it later,
876 			 * thus deferring a hash lookup and lock acquisition
877 			 * at the expense of an m_copym().
878 			 */
879 			ip6_mloopback(ifp, m);
880 		} else {
881 			/*
882 			 * If we are acting as a multicast router, perform
883 			 * multicast forwarding as if the packet had just
884 			 * arrived on the interface to which we are about
885 			 * to send.  The multicast forwarding function
886 			 * recursively calls this function, using the
887 			 * IPV6_FORWARDING flag to prevent infinite recursion.
888 			 *
889 			 * Multicasts that are looped back by ip6_mloopback(),
890 			 * above, will be forwarded by the ip6_input() routine,
891 			 * if necessary.
892 			 */
893 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
894 				/*
895 				 * XXX: ip6_mforward expects that rcvif is NULL
896 				 * when it is called from the originating path.
897 				 * However, it may not always be the case.
898 				 */
899 				m->m_pkthdr.rcvif = NULL;
900 				if (ip6_mforward(ip6, ifp, m) != 0) {
901 					m_freem(m);
902 					goto done;
903 				}
904 			}
905 		}
906 		/*
907 		 * Multicasts with a hoplimit of zero may be looped back,
908 		 * above, but must not be transmitted on a network.
909 		 * Also, multicasts addressed to the loopback interface
910 		 * are not sent -- the above call to ip6_mloopback() will
911 		 * loop back a copy if this host actually belongs to the
912 		 * destination group on the loopback interface.
913 		 */
914 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
915 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
916 			m_freem(m);
917 			goto done;
918 		}
919 	}
920 
921 	/*
922 	 * Fill the outgoing inteface to tell the upper layer
923 	 * to increment per-interface statistics.
924 	 */
925 	if (ifpp)
926 		*ifpp = ifp;
927 
928 	/* Determine path MTU. */
929 	ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst, &mtu, fibnum,
930 	    *nexthdrp);
931 	KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p fibnum %u",
932 	    __func__, __LINE__, mtu, ro_pmtu, ro, ifp, fibnum));
933 
934 	/*
935 	 * The caller of this function may specify to use the minimum MTU
936 	 * in some cases.
937 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
938 	 * setting.  The logic is a bit complicated; by default, unicast
939 	 * packets will follow path MTU while multicast packets will be sent at
940 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
941 	 * including unicast ones will be sent at the minimum MTU.  Multicast
942 	 * packets will always be sent at the minimum MTU unless
943 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
944 	 * See RFC 3542 for more details.
945 	 */
946 	if (mtu > IPV6_MMTU) {
947 		if ((flags & IPV6_MINMTU))
948 			mtu = IPV6_MMTU;
949 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
950 			mtu = IPV6_MMTU;
951 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
952 			 (opt == NULL ||
953 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
954 			mtu = IPV6_MMTU;
955 		}
956 	}
957 
958 	/*
959 	 * Clear embedded scope identifiers if necessary.
960 	 * in6_clearscope() will touch the addresses only when necessary.
961 	 */
962 	in6_clearscope(&ip6->ip6_src);
963 	in6_clearscope(&ip6->ip6_dst);
964 
965 	/*
966 	 * If the outgoing packet contains a hop-by-hop options header,
967 	 * it must be examined and processed even by the source node.
968 	 * (RFC 2460, section 4.)
969 	 */
970 	if (exthdrs.ip6e_hbh) {
971 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
972 		u_int32_t dummy; /* XXX unused */
973 
974 #ifdef DIAGNOSTIC
975 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
976 			panic("ip6e_hbh is not contiguous");
977 #endif
978 		/*
979 		 *  XXX: if we have to send an ICMPv6 error to the sender,
980 		 *       we need the M_LOOP flag since icmp6_error() expects
981 		 *       the IPv6 and the hop-by-hop options header are
982 		 *       contiguous unless the flag is set.
983 		 */
984 		m->m_flags |= M_LOOP;
985 		m->m_pkthdr.rcvif = ifp;
986 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
987 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
988 		    &dummy) < 0) {
989 			/* m was already freed at this point. */
990 			error = EINVAL;/* better error? */
991 			goto done;
992 		}
993 		m->m_flags &= ~M_LOOP; /* XXX */
994 		m->m_pkthdr.rcvif = NULL;
995 	}
996 
997 	/* Jump over all PFIL processing if hooks are not active. */
998 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
999 		goto passout;
1000 
1001 	odst = ip6->ip6_dst;
1002 	/* Run through list of hooks for output packets. */
1003 	switch (pfil_mbuf_out(V_inet6_pfil_head, &m, ifp, inp)) {
1004 	case PFIL_PASS:
1005 		ip6 = mtod(m, struct ip6_hdr *);
1006 		break;
1007 	case PFIL_DROPPED:
1008 		error = EACCES;
1009 		/* FALLTHROUGH */
1010 	case PFIL_CONSUMED:
1011 		goto done;
1012 	}
1013 
1014 	needfiblookup = 0;
1015 	/* See if destination IP address was changed by packet filter. */
1016 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1017 		m->m_flags |= M_SKIP_FIREWALL;
1018 		/* If destination is now ourself drop to ip6_input(). */
1019 		if (in6_localip(&ip6->ip6_dst)) {
1020 			m->m_flags |= M_FASTFWD_OURS;
1021 			if (m->m_pkthdr.rcvif == NULL)
1022 				m->m_pkthdr.rcvif = V_loif;
1023 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1024 				m->m_pkthdr.csum_flags |=
1025 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1026 				m->m_pkthdr.csum_data = 0xffff;
1027 			}
1028 #if defined(SCTP) || defined(SCTP_SUPPORT)
1029 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1030 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1031 #endif
1032 			error = netisr_queue(NETISR_IPV6, m);
1033 			goto done;
1034 		} else {
1035 			if (ro != NULL)
1036 				RO_INVALIDATE_CACHE(ro);
1037 			needfiblookup = 1; /* Redo the routing table lookup. */
1038 		}
1039 	}
1040 	/* See if fib was changed by packet filter. */
1041 	if (fibnum != M_GETFIB(m)) {
1042 		m->m_flags |= M_SKIP_FIREWALL;
1043 		fibnum = M_GETFIB(m);
1044 		if (ro != NULL)
1045 			RO_INVALIDATE_CACHE(ro);
1046 		needfiblookup = 1;
1047 	}
1048 	if (needfiblookup)
1049 		goto again;
1050 
1051 	/* See if local, if yes, send it to netisr. */
1052 	if (m->m_flags & M_FASTFWD_OURS) {
1053 		if (m->m_pkthdr.rcvif == NULL)
1054 			m->m_pkthdr.rcvif = V_loif;
1055 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1056 			m->m_pkthdr.csum_flags |=
1057 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1058 			m->m_pkthdr.csum_data = 0xffff;
1059 		}
1060 #if defined(SCTP) || defined(SCTP_SUPPORT)
1061 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1062 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1063 #endif
1064 		error = netisr_queue(NETISR_IPV6, m);
1065 		goto done;
1066 	}
1067 	/* Or forward to some other address? */
1068 	if ((m->m_flags & M_IP6_NEXTHOP) &&
1069 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1070 		if (ro != NULL)
1071 			dst = (struct sockaddr_in6 *)&ro->ro_dst;
1072 		else
1073 			dst = &sin6;
1074 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1075 		m->m_flags |= M_SKIP_FIREWALL;
1076 		m->m_flags &= ~M_IP6_NEXTHOP;
1077 		m_tag_delete(m, fwd_tag);
1078 		goto again;
1079 	}
1080 
1081 passout:
1082 	if (vlan_pcp > -1)
1083 		EVL_APPLY_PRI(m, vlan_pcp);
1084 
1085 	/* Ensure the packet data is mapped if the interface requires it. */
1086 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1087 		struct mbuf *m1;
1088 
1089 		error = mb_unmapped_to_ext(m, &m1);
1090 		if (error != 0) {
1091 			if (error == EINVAL) {
1092 				if_printf(ifp, "TLS packet\n");
1093 				/* XXXKIB */
1094 			} else if (error == ENOMEM) {
1095 				error = ENOBUFS;
1096 			}
1097 			IP6STAT_INC(ip6s_odropped);
1098 			return (error);
1099 		} else {
1100 			m = m1;
1101 		}
1102 	}
1103 
1104 	/*
1105 	 * Send the packet to the outgoing interface.
1106 	 * If necessary, do IPv6 fragmentation before sending.
1107 	 *
1108 	 * 1: normal case (dontfrag == 0)
1109 	 * 1-a:	send as is if tlen <= path mtu
1110 	 * 1-b:	fragment if tlen > path mtu
1111 	 *
1112 	 * 2: if user asks us not to fragment (dontfrag == 1)
1113 	 * 2-a:	send as is if tlen <= interface mtu
1114 	 * 2-b:	error if tlen > interface mtu
1115 	 */
1116 	sw_csum = m->m_pkthdr.csum_flags;
1117 	if (!hdrsplit) {
1118 		tso = ((sw_csum & ifp->if_hwassist &
1119 		    (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1120 		sw_csum &= ~ifp->if_hwassist;
1121 	} else
1122 		tso = 0;
1123 	/*
1124 	 * If we added extension headers, we will not do TSO and calculate the
1125 	 * checksums ourselves for now.
1126 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1127 	 * with ext. hdrs.
1128 	 */
1129 	ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1130 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1131 	tlen = m->m_pkthdr.len;
1132 
1133 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1134 		dontfrag = 1;
1135 	else
1136 		dontfrag = 0;
1137 	if (dontfrag && tlen > in6_ifmtu(ifp) && !tso) {	/* Case 2-b. */
1138 		/*
1139 		 * If the DONTFRAG option is specified, we cannot send the
1140 		 * packet when the data length is larger than the MTU of the
1141 		 * outgoing interface.
1142 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1143 		 * application wanted to know the MTU value. Also return an
1144 		 * error code (this is not described in the API spec).
1145 		 */
1146 		if (inp != NULL)
1147 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1148 		error = EMSGSIZE;
1149 		goto bad;
1150 	}
1151 
1152 	/* Transmit packet without fragmentation. */
1153 	if (dontfrag || tlen <= mtu) {	/* Cases 1-a and 2-a. */
1154 		struct in6_ifaddr *ia6;
1155 
1156 		ip6 = mtod(m, struct ip6_hdr *);
1157 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1158 		if (ia6) {
1159 			/* Record statistics for this interface address. */
1160 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1161 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1162 			    m->m_pkthdr.len);
1163 		}
1164 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1165 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
1166 		goto done;
1167 	}
1168 
1169 	/* Try to fragment the packet.  Case 1-b. */
1170 	if (mtu < IPV6_MMTU) {
1171 		/* Path MTU cannot be less than IPV6_MMTU. */
1172 		error = EMSGSIZE;
1173 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1174 		goto bad;
1175 	} else if (ip6->ip6_plen == 0) {
1176 		/* We do not support jumbo payload. */
1177 		error = EMSGSIZE;
1178 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1179 		goto bad;
1180 	} else {
1181 		u_char nextproto;
1182 
1183 		/*
1184 		 * Too large for the destination or interface;
1185 		 * fragment if possible.
1186 		 * Must be able to put at least 8 bytes per fragment.
1187 		 */
1188 		if (mtu > IPV6_MAXPACKET)
1189 			mtu = IPV6_MAXPACKET;
1190 
1191 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1192 		if (len < 8) {
1193 			error = EMSGSIZE;
1194 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1195 			goto bad;
1196 		}
1197 
1198 		/*
1199 		 * If the interface will not calculate checksums on
1200 		 * fragmented packets, then do it here.
1201 		 * XXX-BZ handle the hw offloading case.  Need flags.
1202 		 */
1203 		ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1204 		    optlen);
1205 
1206 		/*
1207 		 * Change the next header field of the last header in the
1208 		 * unfragmentable part.
1209 		 */
1210 		if (exthdrs.ip6e_rthdr) {
1211 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1212 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1213 		} else if (exthdrs.ip6e_dest1) {
1214 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1215 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1216 		} else if (exthdrs.ip6e_hbh) {
1217 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1218 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1219 		} else {
1220 			ip6 = mtod(m, struct ip6_hdr *);
1221 			nextproto = ip6->ip6_nxt;
1222 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1223 		}
1224 
1225 		/*
1226 		 * Loop through length of segment after first fragment,
1227 		 * make new header and copy data of each part and link onto
1228 		 * chain.
1229 		 */
1230 		m0 = m;
1231 		id = htonl(ip6_randomid());
1232 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1233 		if (error != 0)
1234 			goto sendorfree;
1235 
1236 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1237 	}
1238 
1239 	/* Remove leading garbage. */
1240 sendorfree:
1241 	m = m0->m_nextpkt;
1242 	m0->m_nextpkt = 0;
1243 	m_freem(m0);
1244 	for (; m; m = m0) {
1245 		m0 = m->m_nextpkt;
1246 		m->m_nextpkt = 0;
1247 		if (error == 0) {
1248 			/* Record statistics for this interface address. */
1249 			if (ia) {
1250 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1251 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1252 				    m->m_pkthdr.len);
1253 			}
1254 			if (vlan_pcp > -1)
1255 				EVL_APPLY_PRI(m, vlan_pcp);
1256 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1257 			    true);
1258 		} else
1259 			m_freem(m);
1260 	}
1261 
1262 	if (error == 0)
1263 		IP6STAT_INC(ip6s_fragmented);
1264 
1265 done:
1266 	return (error);
1267 
1268 freehdrs:
1269 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1270 	m_freem(exthdrs.ip6e_dest1);
1271 	m_freem(exthdrs.ip6e_rthdr);
1272 	m_freem(exthdrs.ip6e_dest2);
1273 	/* FALLTHROUGH */
1274 bad:
1275 	if (m)
1276 		m_freem(m);
1277 	goto done;
1278 }
1279 
1280 static int
1281 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1282 {
1283 	struct mbuf *m;
1284 
1285 	if (hlen > MCLBYTES)
1286 		return (ENOBUFS); /* XXX */
1287 
1288 	if (hlen > MLEN)
1289 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1290 	else
1291 		m = m_get(M_NOWAIT, MT_DATA);
1292 	if (m == NULL)
1293 		return (ENOBUFS);
1294 	m->m_len = hlen;
1295 	if (hdr)
1296 		bcopy(hdr, mtod(m, caddr_t), hlen);
1297 
1298 	*mp = m;
1299 	return (0);
1300 }
1301 
1302 /*
1303  * Insert fragment header and copy unfragmentable header portions.
1304  */
1305 static int
1306 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1307     struct ip6_frag **frghdrp)
1308 {
1309 	struct mbuf *n, *mlast;
1310 
1311 	if (hlen > sizeof(struct ip6_hdr)) {
1312 		n = m_copym(m0, sizeof(struct ip6_hdr),
1313 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1314 		if (n == NULL)
1315 			return (ENOBUFS);
1316 		m->m_next = n;
1317 	} else
1318 		n = m;
1319 
1320 	/* Search for the last mbuf of unfragmentable part. */
1321 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1322 		;
1323 
1324 	if (M_WRITABLE(mlast) &&
1325 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1326 		/* use the trailing space of the last mbuf for the fragment hdr */
1327 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1328 		    mlast->m_len);
1329 		mlast->m_len += sizeof(struct ip6_frag);
1330 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1331 	} else {
1332 		/* allocate a new mbuf for the fragment header */
1333 		struct mbuf *mfrg;
1334 
1335 		mfrg = m_get(M_NOWAIT, MT_DATA);
1336 		if (mfrg == NULL)
1337 			return (ENOBUFS);
1338 		mfrg->m_len = sizeof(struct ip6_frag);
1339 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1340 		mlast->m_next = mfrg;
1341 	}
1342 
1343 	return (0);
1344 }
1345 
1346 /*
1347  * Calculates IPv6 path mtu for destination @dst.
1348  * Resulting MTU is stored in @mtup.
1349  *
1350  * Returns 0 on success.
1351  */
1352 static int
1353 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1354 {
1355 	struct epoch_tracker et;
1356 	struct nhop_object *nh;
1357 	struct in6_addr kdst;
1358 	uint32_t scopeid;
1359 	int error;
1360 
1361 	in6_splitscope(dst, &kdst, &scopeid);
1362 
1363 	NET_EPOCH_ENTER(et);
1364 	nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1365 	if (nh != NULL) {
1366 		ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, 0);
1367 		error = 0;
1368 	} else
1369 		error = EHOSTUNREACH;
1370 	NET_EPOCH_EXIT(et);
1371 
1372 	return (error);
1373 }
1374 
1375 /*
1376  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1377  * and cached data in @ro_pmtu.
1378  * MTU from (successful) route lookup is saved (along with dst)
1379  * inside @ro_pmtu to avoid subsequent route lookups after packet
1380  * filter processing.
1381  *
1382  * Stores mtu into @mtup.
1383  */
1384 static void
1385 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1386     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1387     u_int fibnum, u_int proto)
1388 {
1389 	struct nhop_object *nh;
1390 	struct in6_addr kdst;
1391 	uint32_t scopeid;
1392 	struct sockaddr_in6 *sa6_dst, sin6;
1393 	u_long mtu;
1394 
1395 	NET_EPOCH_ASSERT();
1396 
1397 	mtu = 0;
1398 	if (ro_pmtu == NULL || do_lookup) {
1399 		/*
1400 		 * Here ro_pmtu has final destination address, while
1401 		 * ro might represent immediate destination.
1402 		 * Use ro_pmtu destination since mtu might differ.
1403 		 */
1404 		if (ro_pmtu != NULL) {
1405 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1406 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1407 				ro_pmtu->ro_mtu = 0;
1408 		} else
1409 			sa6_dst = &sin6;
1410 
1411 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1412 			bzero(sa6_dst, sizeof(*sa6_dst));
1413 			sa6_dst->sin6_family = AF_INET6;
1414 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1415 			sa6_dst->sin6_addr = *dst;
1416 
1417 			in6_splitscope(dst, &kdst, &scopeid);
1418 			nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1419 			if (nh != NULL) {
1420 				mtu = nh->nh_mtu;
1421 				if (ro_pmtu != NULL)
1422 					ro_pmtu->ro_mtu = mtu;
1423 			}
1424 		} else
1425 			mtu = ro_pmtu->ro_mtu;
1426 	}
1427 
1428 	if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1429 		mtu = ro_pmtu->ro_nh->nh_mtu;
1430 
1431 	ip6_calcmtu(ifp, dst, mtu, mtup, proto);
1432 }
1433 
1434 /*
1435  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1436  * hostcache data for @dst.
1437  * Stores mtu into @mtup.
1438  */
1439 static void
1440 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1441     u_long *mtup, u_int proto)
1442 {
1443 	u_long mtu = 0;
1444 
1445 	if (rt_mtu > 0) {
1446 		/* Skip the hostcache if the protocol handles PMTU changes. */
1447 		if (proto != IPPROTO_TCP && proto != IPPROTO_SCTP) {
1448 			struct in_conninfo inc = {
1449 			    .inc_flags = INC_ISIPV6,
1450 			    .inc6_faddr = *dst,
1451 			};
1452 
1453 			mtu = tcp_hc_getmtu(&inc);
1454 		}
1455 
1456 		if (mtu)
1457 			mtu = min(mtu, rt_mtu);
1458 		else
1459 			mtu = rt_mtu;
1460 	}
1461 
1462 	if (mtu == 0)
1463 		mtu = in6_ifmtu(ifp);
1464 
1465 	*mtup = mtu;
1466 }
1467 
1468 /*
1469  * IP6 socket option processing.
1470  */
1471 int
1472 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1473 {
1474 	int optdatalen, uproto;
1475 	void *optdata;
1476 	struct inpcb *inp = sotoinpcb(so);
1477 	int error, optval;
1478 	int level, op, optname;
1479 	int optlen;
1480 	struct thread *td;
1481 #ifdef	RSS
1482 	uint32_t rss_bucket;
1483 	int retval;
1484 #endif
1485 
1486 /*
1487  * Don't use more than a quarter of mbuf clusters.  N.B.:
1488  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1489  * on LP64 architectures, so cast to u_long to avoid undefined
1490  * behavior.  ILP32 architectures cannot have nmbclusters
1491  * large enough to overflow for other reasons.
1492  */
1493 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1494 
1495 	level = sopt->sopt_level;
1496 	op = sopt->sopt_dir;
1497 	optname = sopt->sopt_name;
1498 	optlen = sopt->sopt_valsize;
1499 	td = sopt->sopt_td;
1500 	error = 0;
1501 	optval = 0;
1502 	uproto = (int)so->so_proto->pr_protocol;
1503 
1504 	if (level != IPPROTO_IPV6) {
1505 		error = EINVAL;
1506 
1507 		if (sopt->sopt_level == SOL_SOCKET &&
1508 		    sopt->sopt_dir == SOPT_SET) {
1509 			switch (sopt->sopt_name) {
1510 			case SO_SETFIB:
1511 				error = sooptcopyin(sopt, &optval,
1512 				    sizeof(optval), sizeof(optval));
1513 				if (error != 0)
1514 					break;
1515 
1516 				INP_WLOCK(inp);
1517 				if ((inp->inp_flags & INP_BOUNDFIB) != 0 &&
1518 				    optval != so->so_fibnum) {
1519 					INP_WUNLOCK(inp);
1520 					error = EISCONN;
1521 					break;
1522 				}
1523 				error = sosetfib(inp->inp_socket, optval);
1524 				if (error == 0)
1525 					inp->inp_inc.inc_fibnum = optval;
1526 				INP_WUNLOCK(inp);
1527 				break;
1528 			case SO_MAX_PACING_RATE:
1529 #ifdef RATELIMIT
1530 				INP_WLOCK(inp);
1531 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1532 				INP_WUNLOCK(inp);
1533 				error = 0;
1534 #else
1535 				error = EOPNOTSUPP;
1536 #endif
1537 				break;
1538 			default:
1539 				break;
1540 			}
1541 		}
1542 	} else {		/* level == IPPROTO_IPV6 */
1543 		switch (op) {
1544 		case SOPT_SET:
1545 			switch (optname) {
1546 			case IPV6_2292PKTOPTIONS:
1547 #ifdef IPV6_PKTOPTIONS
1548 			case IPV6_PKTOPTIONS:
1549 #endif
1550 			{
1551 				struct mbuf *m;
1552 
1553 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1554 					printf("ip6_ctloutput: mbuf limit hit\n");
1555 					error = ENOBUFS;
1556 					break;
1557 				}
1558 
1559 				error = soopt_getm(sopt, &m); /* XXX */
1560 				if (error != 0)
1561 					break;
1562 				error = soopt_mcopyin(sopt, m); /* XXX */
1563 				if (error != 0)
1564 					break;
1565 				INP_WLOCK(inp);
1566 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1567 				    so, sopt);
1568 				INP_WUNLOCK(inp);
1569 				m_freem(m); /* XXX */
1570 				break;
1571 			}
1572 
1573 			/*
1574 			 * Use of some Hop-by-Hop options or some
1575 			 * Destination options, might require special
1576 			 * privilege.  That is, normal applications
1577 			 * (without special privilege) might be forbidden
1578 			 * from setting certain options in outgoing packets,
1579 			 * and might never see certain options in received
1580 			 * packets. [RFC 2292 Section 6]
1581 			 * KAME specific note:
1582 			 *  KAME prevents non-privileged users from sending or
1583 			 *  receiving ANY hbh/dst options in order to avoid
1584 			 *  overhead of parsing options in the kernel.
1585 			 */
1586 			case IPV6_RECVHOPOPTS:
1587 			case IPV6_RECVDSTOPTS:
1588 			case IPV6_RECVRTHDRDSTOPTS:
1589 				if (td != NULL) {
1590 					error = priv_check(td,
1591 					    PRIV_NETINET_SETHDROPTS);
1592 					if (error)
1593 						break;
1594 				}
1595 				/* FALLTHROUGH */
1596 			case IPV6_UNICAST_HOPS:
1597 			case IPV6_HOPLIMIT:
1598 
1599 			case IPV6_RECVPKTINFO:
1600 			case IPV6_RECVHOPLIMIT:
1601 			case IPV6_RECVRTHDR:
1602 			case IPV6_RECVPATHMTU:
1603 			case IPV6_RECVTCLASS:
1604 			case IPV6_RECVFLOWID:
1605 #ifdef	RSS
1606 			case IPV6_RECVRSSBUCKETID:
1607 #endif
1608 			case IPV6_V6ONLY:
1609 			case IPV6_AUTOFLOWLABEL:
1610 			case IPV6_ORIGDSTADDR:
1611 			case IPV6_BINDANY:
1612 			case IPV6_VLAN_PCP:
1613 				if (optname == IPV6_BINDANY && td != NULL) {
1614 					error = priv_check(td,
1615 					    PRIV_NETINET_BINDANY);
1616 					if (error)
1617 						break;
1618 				}
1619 
1620 				if (optlen != sizeof(int)) {
1621 					error = EINVAL;
1622 					break;
1623 				}
1624 				error = sooptcopyin(sopt, &optval,
1625 					sizeof optval, sizeof optval);
1626 				if (error)
1627 					break;
1628 				switch (optname) {
1629 				case IPV6_UNICAST_HOPS:
1630 					if (optval < -1 || optval >= 256)
1631 						error = EINVAL;
1632 					else {
1633 						/* -1 = kernel default */
1634 						inp->in6p_hops = optval;
1635 						if ((inp->inp_vflag &
1636 						     INP_IPV4) != 0)
1637 							inp->inp_ip_ttl = optval;
1638 					}
1639 					break;
1640 #define OPTSET(bit) \
1641 do { \
1642 	INP_WLOCK(inp); \
1643 	if (optval) \
1644 		inp->inp_flags |= (bit); \
1645 	else \
1646 		inp->inp_flags &= ~(bit); \
1647 	INP_WUNLOCK(inp); \
1648 } while (/*CONSTCOND*/ 0)
1649 #define OPTSET2292(bit) \
1650 do { \
1651 	INP_WLOCK(inp); \
1652 	inp->inp_flags |= IN6P_RFC2292; \
1653 	if (optval) \
1654 		inp->inp_flags |= (bit); \
1655 	else \
1656 		inp->inp_flags &= ~(bit); \
1657 	INP_WUNLOCK(inp); \
1658 } while (/*CONSTCOND*/ 0)
1659 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1660 
1661 #define OPTSET2_N(bit, val) do {					\
1662 	if (val)							\
1663 		inp->inp_flags2 |= bit;					\
1664 	else								\
1665 		inp->inp_flags2 &= ~bit;				\
1666 } while (0)
1667 #define OPTSET2(bit, val) do {						\
1668 	INP_WLOCK(inp);							\
1669 	OPTSET2_N(bit, val);						\
1670 	INP_WUNLOCK(inp);						\
1671 } while (0)
1672 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1673 #define OPTSET2292_EXCLUSIVE(bit)					\
1674 do {									\
1675 	INP_WLOCK(inp);							\
1676 	if (OPTBIT(IN6P_RFC2292)) {					\
1677 		error = EINVAL;						\
1678 	} else {							\
1679 		if (optval)						\
1680 			inp->inp_flags |= (bit);			\
1681 		else							\
1682 			inp->inp_flags &= ~(bit);			\
1683 	}								\
1684 	INP_WUNLOCK(inp);						\
1685 } while (/*CONSTCOND*/ 0)
1686 
1687 				case IPV6_RECVPKTINFO:
1688 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1689 					break;
1690 
1691 				case IPV6_HOPLIMIT:
1692 				{
1693 					struct ip6_pktopts **optp;
1694 
1695 					/* cannot mix with RFC2292 */
1696 					if (OPTBIT(IN6P_RFC2292)) {
1697 						error = EINVAL;
1698 						break;
1699 					}
1700 					INP_WLOCK(inp);
1701 					if (inp->inp_flags & INP_DROPPED) {
1702 						INP_WUNLOCK(inp);
1703 						return (ECONNRESET);
1704 					}
1705 					optp = &inp->in6p_outputopts;
1706 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1707 					    (u_char *)&optval, sizeof(optval),
1708 					    optp, (td != NULL) ? td->td_ucred :
1709 					    NULL, uproto);
1710 					INP_WUNLOCK(inp);
1711 					break;
1712 				}
1713 
1714 				case IPV6_RECVHOPLIMIT:
1715 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1716 					break;
1717 
1718 				case IPV6_RECVHOPOPTS:
1719 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1720 					break;
1721 
1722 				case IPV6_RECVDSTOPTS:
1723 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1724 					break;
1725 
1726 				case IPV6_RECVRTHDRDSTOPTS:
1727 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1728 					break;
1729 
1730 				case IPV6_RECVRTHDR:
1731 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1732 					break;
1733 
1734 				case IPV6_RECVPATHMTU:
1735 					/*
1736 					 * We ignore this option for TCP
1737 					 * sockets.
1738 					 * (RFC3542 leaves this case
1739 					 * unspecified.)
1740 					 */
1741 					if (uproto != IPPROTO_TCP)
1742 						OPTSET(IN6P_MTU);
1743 					break;
1744 
1745 				case IPV6_RECVFLOWID:
1746 					OPTSET2(INP_RECVFLOWID, optval);
1747 					break;
1748 
1749 #ifdef	RSS
1750 				case IPV6_RECVRSSBUCKETID:
1751 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1752 					break;
1753 #endif
1754 
1755 				case IPV6_V6ONLY:
1756 					INP_WLOCK(inp);
1757 					if (inp->inp_lport ||
1758 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1759 						/*
1760 						 * The socket is already bound.
1761 						 */
1762 						INP_WUNLOCK(inp);
1763 						error = EINVAL;
1764 						break;
1765 					}
1766 					if (optval) {
1767 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1768 						inp->inp_vflag &= ~INP_IPV4;
1769 					} else {
1770 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1771 						inp->inp_vflag |= INP_IPV4;
1772 					}
1773 					INP_WUNLOCK(inp);
1774 					break;
1775 				case IPV6_RECVTCLASS:
1776 					/* cannot mix with RFC2292 XXX */
1777 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1778 					break;
1779 				case IPV6_AUTOFLOWLABEL:
1780 					OPTSET(IN6P_AUTOFLOWLABEL);
1781 					break;
1782 
1783 				case IPV6_ORIGDSTADDR:
1784 					OPTSET2(INP_ORIGDSTADDR, optval);
1785 					break;
1786 				case IPV6_BINDANY:
1787 					OPTSET(INP_BINDANY);
1788 					break;
1789 				case IPV6_VLAN_PCP:
1790 					if ((optval >= -1) && (optval <=
1791 					    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1792 						if (optval == -1) {
1793 							INP_WLOCK(inp);
1794 							inp->inp_flags2 &=
1795 							    ~(INP_2PCP_SET |
1796 							    INP_2PCP_MASK);
1797 							INP_WUNLOCK(inp);
1798 						} else {
1799 							INP_WLOCK(inp);
1800 							inp->inp_flags2 |=
1801 							    INP_2PCP_SET;
1802 							inp->inp_flags2 &=
1803 							    ~INP_2PCP_MASK;
1804 							inp->inp_flags2 |=
1805 							    optval <<
1806 							    INP_2PCP_SHIFT;
1807 							INP_WUNLOCK(inp);
1808 						}
1809 					} else
1810 						error = EINVAL;
1811 					break;
1812 				}
1813 				break;
1814 
1815 			case IPV6_TCLASS:
1816 			case IPV6_DONTFRAG:
1817 			case IPV6_USE_MIN_MTU:
1818 			case IPV6_PREFER_TEMPADDR:
1819 				if (optlen != sizeof(optval)) {
1820 					error = EINVAL;
1821 					break;
1822 				}
1823 				error = sooptcopyin(sopt, &optval,
1824 					sizeof optval, sizeof optval);
1825 				if (error)
1826 					break;
1827 				{
1828 					struct ip6_pktopts **optp;
1829 					INP_WLOCK(inp);
1830 					if (inp->inp_flags & INP_DROPPED) {
1831 						INP_WUNLOCK(inp);
1832 						return (ECONNRESET);
1833 					}
1834 					optp = &inp->in6p_outputopts;
1835 					error = ip6_pcbopt(optname,
1836 					    (u_char *)&optval, sizeof(optval),
1837 					    optp, (td != NULL) ? td->td_ucred :
1838 					    NULL, uproto);
1839 					INP_WUNLOCK(inp);
1840 					break;
1841 				}
1842 
1843 			case IPV6_2292PKTINFO:
1844 			case IPV6_2292HOPLIMIT:
1845 			case IPV6_2292HOPOPTS:
1846 			case IPV6_2292DSTOPTS:
1847 			case IPV6_2292RTHDR:
1848 				/* RFC 2292 */
1849 				if (optlen != sizeof(int)) {
1850 					error = EINVAL;
1851 					break;
1852 				}
1853 				error = sooptcopyin(sopt, &optval,
1854 					sizeof optval, sizeof optval);
1855 				if (error)
1856 					break;
1857 				switch (optname) {
1858 				case IPV6_2292PKTINFO:
1859 					OPTSET2292(IN6P_PKTINFO);
1860 					break;
1861 				case IPV6_2292HOPLIMIT:
1862 					OPTSET2292(IN6P_HOPLIMIT);
1863 					break;
1864 				case IPV6_2292HOPOPTS:
1865 					/*
1866 					 * Check super-user privilege.
1867 					 * See comments for IPV6_RECVHOPOPTS.
1868 					 */
1869 					if (td != NULL) {
1870 						error = priv_check(td,
1871 						    PRIV_NETINET_SETHDROPTS);
1872 						if (error)
1873 							return (error);
1874 					}
1875 					OPTSET2292(IN6P_HOPOPTS);
1876 					break;
1877 				case IPV6_2292DSTOPTS:
1878 					if (td != NULL) {
1879 						error = priv_check(td,
1880 						    PRIV_NETINET_SETHDROPTS);
1881 						if (error)
1882 							return (error);
1883 					}
1884 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1885 					break;
1886 				case IPV6_2292RTHDR:
1887 					OPTSET2292(IN6P_RTHDR);
1888 					break;
1889 				}
1890 				break;
1891 			case IPV6_PKTINFO:
1892 			case IPV6_HOPOPTS:
1893 			case IPV6_RTHDR:
1894 			case IPV6_DSTOPTS:
1895 			case IPV6_RTHDRDSTOPTS:
1896 			case IPV6_NEXTHOP:
1897 			{
1898 				/* new advanced API (RFC3542) */
1899 				u_char *optbuf;
1900 				u_char optbuf_storage[MCLBYTES];
1901 				int optlen;
1902 				struct ip6_pktopts **optp;
1903 
1904 				/* cannot mix with RFC2292 */
1905 				if (OPTBIT(IN6P_RFC2292)) {
1906 					error = EINVAL;
1907 					break;
1908 				}
1909 
1910 				/*
1911 				 * We only ensure valsize is not too large
1912 				 * here.  Further validation will be done
1913 				 * later.
1914 				 */
1915 				error = sooptcopyin(sopt, optbuf_storage,
1916 				    sizeof(optbuf_storage), 0);
1917 				if (error)
1918 					break;
1919 				optlen = sopt->sopt_valsize;
1920 				optbuf = optbuf_storage;
1921 				INP_WLOCK(inp);
1922 				if (inp->inp_flags & INP_DROPPED) {
1923 					INP_WUNLOCK(inp);
1924 					return (ECONNRESET);
1925 				}
1926 				optp = &inp->in6p_outputopts;
1927 				error = ip6_pcbopt(optname, optbuf, optlen,
1928 				    optp, (td != NULL) ? td->td_ucred : NULL,
1929 				    uproto);
1930 				INP_WUNLOCK(inp);
1931 				break;
1932 			}
1933 #undef OPTSET
1934 
1935 			case IPV6_MULTICAST_IF:
1936 			case IPV6_MULTICAST_HOPS:
1937 			case IPV6_MULTICAST_LOOP:
1938 			case IPV6_JOIN_GROUP:
1939 			case IPV6_LEAVE_GROUP:
1940 			case IPV6_MSFILTER:
1941 			case MCAST_BLOCK_SOURCE:
1942 			case MCAST_UNBLOCK_SOURCE:
1943 			case MCAST_JOIN_GROUP:
1944 			case MCAST_LEAVE_GROUP:
1945 			case MCAST_JOIN_SOURCE_GROUP:
1946 			case MCAST_LEAVE_SOURCE_GROUP:
1947 				error = ip6_setmoptions(inp, sopt);
1948 				break;
1949 
1950 			case IPV6_PORTRANGE:
1951 				error = sooptcopyin(sopt, &optval,
1952 				    sizeof optval, sizeof optval);
1953 				if (error)
1954 					break;
1955 
1956 				INP_WLOCK(inp);
1957 				switch (optval) {
1958 				case IPV6_PORTRANGE_DEFAULT:
1959 					inp->inp_flags &= ~(INP_LOWPORT);
1960 					inp->inp_flags &= ~(INP_HIGHPORT);
1961 					break;
1962 
1963 				case IPV6_PORTRANGE_HIGH:
1964 					inp->inp_flags &= ~(INP_LOWPORT);
1965 					inp->inp_flags |= INP_HIGHPORT;
1966 					break;
1967 
1968 				case IPV6_PORTRANGE_LOW:
1969 					inp->inp_flags &= ~(INP_HIGHPORT);
1970 					inp->inp_flags |= INP_LOWPORT;
1971 					break;
1972 
1973 				default:
1974 					error = EINVAL;
1975 					break;
1976 				}
1977 				INP_WUNLOCK(inp);
1978 				break;
1979 
1980 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1981 			case IPV6_IPSEC_POLICY:
1982 				if (IPSEC_ENABLED(ipv6)) {
1983 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
1984 					break;
1985 				}
1986 				/* FALLTHROUGH */
1987 #endif /* IPSEC */
1988 
1989 			default:
1990 				error = ENOPROTOOPT;
1991 				break;
1992 			}
1993 			break;
1994 
1995 		case SOPT_GET:
1996 			switch (optname) {
1997 			case IPV6_2292PKTOPTIONS:
1998 #ifdef IPV6_PKTOPTIONS
1999 			case IPV6_PKTOPTIONS:
2000 #endif
2001 				/*
2002 				 * RFC3542 (effectively) deprecated the
2003 				 * semantics of the 2292-style pktoptions.
2004 				 * Since it was not reliable in nature (i.e.,
2005 				 * applications had to expect the lack of some
2006 				 * information after all), it would make sense
2007 				 * to simplify this part by always returning
2008 				 * empty data.
2009 				 */
2010 				sopt->sopt_valsize = 0;
2011 				break;
2012 
2013 			case IPV6_RECVHOPOPTS:
2014 			case IPV6_RECVDSTOPTS:
2015 			case IPV6_RECVRTHDRDSTOPTS:
2016 			case IPV6_UNICAST_HOPS:
2017 			case IPV6_RECVPKTINFO:
2018 			case IPV6_RECVHOPLIMIT:
2019 			case IPV6_RECVRTHDR:
2020 			case IPV6_RECVPATHMTU:
2021 
2022 			case IPV6_V6ONLY:
2023 			case IPV6_PORTRANGE:
2024 			case IPV6_RECVTCLASS:
2025 			case IPV6_AUTOFLOWLABEL:
2026 			case IPV6_BINDANY:
2027 			case IPV6_FLOWID:
2028 			case IPV6_FLOWTYPE:
2029 			case IPV6_RECVFLOWID:
2030 #ifdef	RSS
2031 			case IPV6_RSSBUCKETID:
2032 			case IPV6_RECVRSSBUCKETID:
2033 #endif
2034 			case IPV6_VLAN_PCP:
2035 				switch (optname) {
2036 				case IPV6_RECVHOPOPTS:
2037 					optval = OPTBIT(IN6P_HOPOPTS);
2038 					break;
2039 
2040 				case IPV6_RECVDSTOPTS:
2041 					optval = OPTBIT(IN6P_DSTOPTS);
2042 					break;
2043 
2044 				case IPV6_RECVRTHDRDSTOPTS:
2045 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2046 					break;
2047 
2048 				case IPV6_UNICAST_HOPS:
2049 					optval = inp->in6p_hops;
2050 					break;
2051 
2052 				case IPV6_RECVPKTINFO:
2053 					optval = OPTBIT(IN6P_PKTINFO);
2054 					break;
2055 
2056 				case IPV6_RECVHOPLIMIT:
2057 					optval = OPTBIT(IN6P_HOPLIMIT);
2058 					break;
2059 
2060 				case IPV6_RECVRTHDR:
2061 					optval = OPTBIT(IN6P_RTHDR);
2062 					break;
2063 
2064 				case IPV6_RECVPATHMTU:
2065 					optval = OPTBIT(IN6P_MTU);
2066 					break;
2067 
2068 				case IPV6_V6ONLY:
2069 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2070 					break;
2071 
2072 				case IPV6_PORTRANGE:
2073 				    {
2074 					int flags;
2075 					flags = inp->inp_flags;
2076 					if (flags & INP_HIGHPORT)
2077 						optval = IPV6_PORTRANGE_HIGH;
2078 					else if (flags & INP_LOWPORT)
2079 						optval = IPV6_PORTRANGE_LOW;
2080 					else
2081 						optval = 0;
2082 					break;
2083 				    }
2084 				case IPV6_RECVTCLASS:
2085 					optval = OPTBIT(IN6P_TCLASS);
2086 					break;
2087 
2088 				case IPV6_AUTOFLOWLABEL:
2089 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2090 					break;
2091 
2092 				case IPV6_ORIGDSTADDR:
2093 					optval = OPTBIT2(INP_ORIGDSTADDR);
2094 					break;
2095 
2096 				case IPV6_BINDANY:
2097 					optval = OPTBIT(INP_BINDANY);
2098 					break;
2099 
2100 				case IPV6_FLOWID:
2101 					optval = inp->inp_flowid;
2102 					break;
2103 
2104 				case IPV6_FLOWTYPE:
2105 					optval = inp->inp_flowtype;
2106 					break;
2107 
2108 				case IPV6_RECVFLOWID:
2109 					optval = OPTBIT2(INP_RECVFLOWID);
2110 					break;
2111 #ifdef	RSS
2112 				case IPV6_RSSBUCKETID:
2113 					retval =
2114 					    rss_hash2bucket(inp->inp_flowid,
2115 					    inp->inp_flowtype,
2116 					    &rss_bucket);
2117 					if (retval == 0)
2118 						optval = rss_bucket;
2119 					else
2120 						error = EINVAL;
2121 					break;
2122 
2123 				case IPV6_RECVRSSBUCKETID:
2124 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2125 					break;
2126 #endif
2127 
2128 
2129 				case IPV6_VLAN_PCP:
2130 					if (OPTBIT2(INP_2PCP_SET)) {
2131 						optval = (inp->inp_flags2 &
2132 							    INP_2PCP_MASK) >>
2133 							    INP_2PCP_SHIFT;
2134 					} else {
2135 						optval = -1;
2136 					}
2137 					break;
2138 				}
2139 
2140 				if (error)
2141 					break;
2142 				error = sooptcopyout(sopt, &optval,
2143 					sizeof optval);
2144 				break;
2145 
2146 			case IPV6_PATHMTU:
2147 			{
2148 				u_long pmtu = 0;
2149 				struct ip6_mtuinfo mtuinfo;
2150 				struct in6_addr addr;
2151 
2152 				if (!(so->so_state & SS_ISCONNECTED))
2153 					return (ENOTCONN);
2154 				/*
2155 				 * XXX: we dot not consider the case of source
2156 				 * routing, or optional information to specify
2157 				 * the outgoing interface.
2158 				 * Copy faddr out of inp to avoid holding lock
2159 				 * on inp during route lookup.
2160 				 */
2161 				INP_RLOCK(inp);
2162 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2163 				INP_RUNLOCK(inp);
2164 				error = ip6_getpmtu_ctl(so->so_fibnum,
2165 				    &addr, &pmtu);
2166 				if (error)
2167 					break;
2168 				if (pmtu > IPV6_MAXPACKET)
2169 					pmtu = IPV6_MAXPACKET;
2170 
2171 				bzero(&mtuinfo, sizeof(mtuinfo));
2172 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2173 				optdata = (void *)&mtuinfo;
2174 				optdatalen = sizeof(mtuinfo);
2175 				error = sooptcopyout(sopt, optdata,
2176 				    optdatalen);
2177 				break;
2178 			}
2179 
2180 			case IPV6_2292PKTINFO:
2181 			case IPV6_2292HOPLIMIT:
2182 			case IPV6_2292HOPOPTS:
2183 			case IPV6_2292RTHDR:
2184 			case IPV6_2292DSTOPTS:
2185 				switch (optname) {
2186 				case IPV6_2292PKTINFO:
2187 					optval = OPTBIT(IN6P_PKTINFO);
2188 					break;
2189 				case IPV6_2292HOPLIMIT:
2190 					optval = OPTBIT(IN6P_HOPLIMIT);
2191 					break;
2192 				case IPV6_2292HOPOPTS:
2193 					optval = OPTBIT(IN6P_HOPOPTS);
2194 					break;
2195 				case IPV6_2292RTHDR:
2196 					optval = OPTBIT(IN6P_RTHDR);
2197 					break;
2198 				case IPV6_2292DSTOPTS:
2199 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2200 					break;
2201 				}
2202 				error = sooptcopyout(sopt, &optval,
2203 				    sizeof optval);
2204 				break;
2205 			case IPV6_PKTINFO:
2206 			case IPV6_HOPOPTS:
2207 			case IPV6_RTHDR:
2208 			case IPV6_DSTOPTS:
2209 			case IPV6_RTHDRDSTOPTS:
2210 			case IPV6_NEXTHOP:
2211 			case IPV6_TCLASS:
2212 			case IPV6_DONTFRAG:
2213 			case IPV6_USE_MIN_MTU:
2214 			case IPV6_PREFER_TEMPADDR:
2215 				error = ip6_getpcbopt(inp, optname, sopt);
2216 				break;
2217 
2218 			case IPV6_MULTICAST_IF:
2219 			case IPV6_MULTICAST_HOPS:
2220 			case IPV6_MULTICAST_LOOP:
2221 			case IPV6_MSFILTER:
2222 				error = ip6_getmoptions(inp, sopt);
2223 				break;
2224 
2225 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2226 			case IPV6_IPSEC_POLICY:
2227 				if (IPSEC_ENABLED(ipv6)) {
2228 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2229 					break;
2230 				}
2231 				/* FALLTHROUGH */
2232 #endif /* IPSEC */
2233 			default:
2234 				error = ENOPROTOOPT;
2235 				break;
2236 			}
2237 			break;
2238 		}
2239 	}
2240 	return (error);
2241 }
2242 
2243 int
2244 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2245 {
2246 	int error = 0, optval, optlen;
2247 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2248 	struct inpcb *inp = sotoinpcb(so);
2249 	int level, op, optname;
2250 
2251 	level = sopt->sopt_level;
2252 	op = sopt->sopt_dir;
2253 	optname = sopt->sopt_name;
2254 	optlen = sopt->sopt_valsize;
2255 
2256 	if (level != IPPROTO_IPV6) {
2257 		return (EINVAL);
2258 	}
2259 
2260 	switch (optname) {
2261 	case IPV6_CHECKSUM:
2262 		/*
2263 		 * For ICMPv6 sockets, no modification allowed for checksum
2264 		 * offset, permit "no change" values to help existing apps.
2265 		 *
2266 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2267 		 * for an ICMPv6 socket will fail."
2268 		 * The current behavior does not meet RFC3542.
2269 		 */
2270 		switch (op) {
2271 		case SOPT_SET:
2272 			if (optlen != sizeof(int)) {
2273 				error = EINVAL;
2274 				break;
2275 			}
2276 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2277 					    sizeof(optval));
2278 			if (error)
2279 				break;
2280 			if (optval < -1 || (optval % 2) != 0) {
2281 				/*
2282 				 * The API assumes non-negative even offset
2283 				 * values or -1 as a special value.
2284 				 */
2285 				error = EINVAL;
2286 			} else if (inp->inp_ip_p == IPPROTO_ICMPV6) {
2287 				if (optval != icmp6off)
2288 					error = EINVAL;
2289 			} else
2290 				inp->in6p_cksum = optval;
2291 			break;
2292 
2293 		case SOPT_GET:
2294 			if (inp->inp_ip_p == IPPROTO_ICMPV6)
2295 				optval = icmp6off;
2296 			else
2297 				optval = inp->in6p_cksum;
2298 
2299 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2300 			break;
2301 
2302 		default:
2303 			error = EINVAL;
2304 			break;
2305 		}
2306 		break;
2307 
2308 	default:
2309 		error = ENOPROTOOPT;
2310 		break;
2311 	}
2312 
2313 	return (error);
2314 }
2315 
2316 /*
2317  * Set up IP6 options in pcb for insertion in output packets or
2318  * specifying behavior of outgoing packets.
2319  */
2320 static int
2321 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2322     struct socket *so, struct sockopt *sopt)
2323 {
2324 	struct ip6_pktopts *opt = *pktopt;
2325 	int error = 0;
2326 	struct thread *td = sopt->sopt_td;
2327 	struct epoch_tracker et;
2328 
2329 	/* turn off any old options. */
2330 	if (opt) {
2331 #ifdef DIAGNOSTIC
2332 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2333 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2334 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2335 			printf("ip6_pcbopts: all specified options are cleared.\n");
2336 #endif
2337 		ip6_clearpktopts(opt, -1);
2338 	} else {
2339 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2340 		if (opt == NULL)
2341 			return (ENOMEM);
2342 	}
2343 	*pktopt = NULL;
2344 
2345 	if (!m || m->m_len == 0) {
2346 		/*
2347 		 * Only turning off any previous options, regardless of
2348 		 * whether the opt is just created or given.
2349 		 */
2350 		free(opt, M_IP6OPT);
2351 		return (0);
2352 	}
2353 
2354 	/*  set options specified by user. */
2355 	NET_EPOCH_ENTER(et);
2356 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2357 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2358 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2359 		free(opt, M_IP6OPT);
2360 		NET_EPOCH_EXIT(et);
2361 		return (error);
2362 	}
2363 	NET_EPOCH_EXIT(et);
2364 	*pktopt = opt;
2365 	return (0);
2366 }
2367 
2368 /*
2369  * initialize ip6_pktopts.  beware that there are non-zero default values in
2370  * the struct.
2371  */
2372 void
2373 ip6_initpktopts(struct ip6_pktopts *opt)
2374 {
2375 
2376 	bzero(opt, sizeof(*opt));
2377 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2378 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2379 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2380 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2381 }
2382 
2383 static int
2384 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2385     struct ucred *cred, int uproto)
2386 {
2387 	struct epoch_tracker et;
2388 	struct ip6_pktopts *opt;
2389 	int ret;
2390 
2391 	if (*pktopt == NULL) {
2392 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2393 		    M_NOWAIT);
2394 		if (*pktopt == NULL)
2395 			return (ENOBUFS);
2396 		ip6_initpktopts(*pktopt);
2397 	}
2398 	opt = *pktopt;
2399 
2400 	NET_EPOCH_ENTER(et);
2401 	ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2402 	NET_EPOCH_EXIT(et);
2403 
2404 	return (ret);
2405 }
2406 
2407 #define GET_PKTOPT_VAR(field, lenexpr) do {				\
2408 	if (pktopt && pktopt->field) {					\
2409 		INP_RUNLOCK(inp);					\
2410 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);	\
2411 		malloc_optdata = true;					\
2412 		INP_RLOCK(inp);						\
2413 		if (inp->inp_flags & INP_DROPPED) {			\
2414 			INP_RUNLOCK(inp);				\
2415 			free(optdata, M_TEMP);				\
2416 			return (ECONNRESET);				\
2417 		}							\
2418 		pktopt = inp->in6p_outputopts;				\
2419 		if (pktopt && pktopt->field) {				\
2420 			optdatalen = min(lenexpr, sopt->sopt_valsize);	\
2421 			bcopy(pktopt->field, optdata, optdatalen);	\
2422 		} else {						\
2423 			free(optdata, M_TEMP);				\
2424 			optdata = NULL;					\
2425 			malloc_optdata = false;				\
2426 		}							\
2427 	}								\
2428 } while(0)
2429 
2430 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,			\
2431 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2432 
2433 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,		\
2434 	pktopt->field->sa_len)
2435 
2436 static int
2437 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2438 {
2439 	void *optdata = NULL;
2440 	bool malloc_optdata = false;
2441 	int optdatalen = 0;
2442 	int error = 0;
2443 	struct in6_pktinfo null_pktinfo;
2444 	int deftclass = 0, on;
2445 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2446 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2447 	struct ip6_pktopts *pktopt;
2448 
2449 	INP_RLOCK(inp);
2450 	pktopt = inp->in6p_outputopts;
2451 
2452 	switch (optname) {
2453 	case IPV6_PKTINFO:
2454 		optdata = (void *)&null_pktinfo;
2455 		if (pktopt && pktopt->ip6po_pktinfo) {
2456 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2457 			    sizeof(null_pktinfo));
2458 			in6_clearscope(&null_pktinfo.ipi6_addr);
2459 		} else {
2460 			/* XXX: we don't have to do this every time... */
2461 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2462 		}
2463 		optdatalen = sizeof(struct in6_pktinfo);
2464 		break;
2465 	case IPV6_TCLASS:
2466 		if (pktopt && pktopt->ip6po_tclass >= 0)
2467 			deftclass = pktopt->ip6po_tclass;
2468 		optdata = (void *)&deftclass;
2469 		optdatalen = sizeof(int);
2470 		break;
2471 	case IPV6_HOPOPTS:
2472 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2473 		break;
2474 	case IPV6_RTHDR:
2475 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2476 		break;
2477 	case IPV6_RTHDRDSTOPTS:
2478 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2479 		break;
2480 	case IPV6_DSTOPTS:
2481 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2482 		break;
2483 	case IPV6_NEXTHOP:
2484 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2485 		break;
2486 	case IPV6_USE_MIN_MTU:
2487 		if (pktopt)
2488 			defminmtu = pktopt->ip6po_minmtu;
2489 		optdata = (void *)&defminmtu;
2490 		optdatalen = sizeof(int);
2491 		break;
2492 	case IPV6_DONTFRAG:
2493 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2494 			on = 1;
2495 		else
2496 			on = 0;
2497 		optdata = (void *)&on;
2498 		optdatalen = sizeof(on);
2499 		break;
2500 	case IPV6_PREFER_TEMPADDR:
2501 		if (pktopt)
2502 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2503 		optdata = (void *)&defpreftemp;
2504 		optdatalen = sizeof(int);
2505 		break;
2506 	default:		/* should not happen */
2507 #ifdef DIAGNOSTIC
2508 		panic("ip6_getpcbopt: unexpected option\n");
2509 #endif
2510 		INP_RUNLOCK(inp);
2511 		return (ENOPROTOOPT);
2512 	}
2513 	INP_RUNLOCK(inp);
2514 
2515 	error = sooptcopyout(sopt, optdata, optdatalen);
2516 	if (malloc_optdata)
2517 		free(optdata, M_TEMP);
2518 
2519 	return (error);
2520 }
2521 
2522 void
2523 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2524 {
2525 	if (pktopt == NULL)
2526 		return;
2527 
2528 	if (optname == -1 || optname == IPV6_PKTINFO) {
2529 		if (pktopt->ip6po_pktinfo)
2530 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2531 		pktopt->ip6po_pktinfo = NULL;
2532 	}
2533 	if (optname == -1 || optname == IPV6_HOPLIMIT) {
2534 		pktopt->ip6po_hlim = -1;
2535 		pktopt->ip6po_valid &= ~IP6PO_VALID_HLIM;
2536 	}
2537 	if (optname == -1 || optname == IPV6_TCLASS) {
2538 		pktopt->ip6po_tclass = -1;
2539 		pktopt->ip6po_valid &= ~IP6PO_VALID_TC;
2540 	}
2541 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2542 		if (pktopt->ip6po_nextroute.ro_nh) {
2543 			NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2544 			pktopt->ip6po_nextroute.ro_nh = NULL;
2545 		}
2546 		if (pktopt->ip6po_nexthop)
2547 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2548 		pktopt->ip6po_nexthop = NULL;
2549 		pktopt->ip6po_valid &= ~IP6PO_VALID_NHINFO;
2550 	}
2551 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2552 		if (pktopt->ip6po_hbh)
2553 			free(pktopt->ip6po_hbh, M_IP6OPT);
2554 		pktopt->ip6po_hbh = NULL;
2555 		pktopt->ip6po_valid &= ~IP6PO_VALID_HBH;
2556 	}
2557 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2558 		if (pktopt->ip6po_dest1)
2559 			free(pktopt->ip6po_dest1, M_IP6OPT);
2560 		pktopt->ip6po_dest1 = NULL;
2561 		pktopt->ip6po_valid &= ~IP6PO_VALID_DEST1;
2562 	}
2563 	if (optname == -1 || optname == IPV6_RTHDR) {
2564 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2565 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2566 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2567 		if (pktopt->ip6po_route.ro_nh) {
2568 			NH_FREE(pktopt->ip6po_route.ro_nh);
2569 			pktopt->ip6po_route.ro_nh = NULL;
2570 		}
2571 		pktopt->ip6po_valid &= ~IP6PO_VALID_RHINFO;
2572 	}
2573 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2574 		if (pktopt->ip6po_dest2)
2575 			free(pktopt->ip6po_dest2, M_IP6OPT);
2576 		pktopt->ip6po_dest2 = NULL;
2577 		pktopt->ip6po_valid &= ~IP6PO_VALID_DEST2;
2578 	}
2579 }
2580 
2581 #define PKTOPT_EXTHDRCPY(type) \
2582 do {\
2583 	if (src->type) {\
2584 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2585 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2586 		if (dst->type == NULL)\
2587 			goto bad;\
2588 		bcopy(src->type, dst->type, hlen);\
2589 	}\
2590 } while (/*CONSTCOND*/ 0)
2591 
2592 static int
2593 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2594 {
2595 	if (dst == NULL || src == NULL)  {
2596 		printf("ip6_clearpktopts: invalid argument\n");
2597 		return (EINVAL);
2598 	}
2599 
2600 	dst->ip6po_hlim = src->ip6po_hlim;
2601 	dst->ip6po_tclass = src->ip6po_tclass;
2602 	dst->ip6po_flags = src->ip6po_flags;
2603 	dst->ip6po_minmtu = src->ip6po_minmtu;
2604 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2605 	if (src->ip6po_pktinfo) {
2606 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2607 		    M_IP6OPT, canwait);
2608 		if (dst->ip6po_pktinfo == NULL)
2609 			goto bad;
2610 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2611 	}
2612 	if (src->ip6po_nexthop) {
2613 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2614 		    M_IP6OPT, canwait);
2615 		if (dst->ip6po_nexthop == NULL)
2616 			goto bad;
2617 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2618 		    src->ip6po_nexthop->sa_len);
2619 	}
2620 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2621 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2622 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2623 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2624 	dst->ip6po_valid = src->ip6po_valid;
2625 	return (0);
2626 
2627   bad:
2628 	ip6_clearpktopts(dst, -1);
2629 	return (ENOBUFS);
2630 }
2631 #undef PKTOPT_EXTHDRCPY
2632 
2633 struct ip6_pktopts *
2634 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2635 {
2636 	int error;
2637 	struct ip6_pktopts *dst;
2638 
2639 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2640 	if (dst == NULL)
2641 		return (NULL);
2642 	ip6_initpktopts(dst);
2643 
2644 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2645 		free(dst, M_IP6OPT);
2646 		return (NULL);
2647 	}
2648 
2649 	return (dst);
2650 }
2651 
2652 void
2653 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2654 {
2655 	if (pktopt == NULL)
2656 		return;
2657 
2658 	ip6_clearpktopts(pktopt, -1);
2659 
2660 	free(pktopt, M_IP6OPT);
2661 }
2662 
2663 /*
2664  * Set IPv6 outgoing packet options based on advanced API.
2665  */
2666 int
2667 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2668     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2669 {
2670 	struct cmsghdr *cm = NULL;
2671 
2672 	if (control == NULL || opt == NULL)
2673 		return (EINVAL);
2674 
2675 	/*
2676 	 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2677 	 * are in the network epoch here.
2678 	 */
2679 	NET_EPOCH_ASSERT();
2680 
2681 	ip6_initpktopts(opt);
2682 	if (stickyopt) {
2683 		int error;
2684 
2685 		/*
2686 		 * If stickyopt is provided, make a local copy of the options
2687 		 * for this particular packet, then override them by ancillary
2688 		 * objects.
2689 		 * XXX: copypktopts() does not copy the cached route to a next
2690 		 * hop (if any).  This is not very good in terms of efficiency,
2691 		 * but we can allow this since this option should be rarely
2692 		 * used.
2693 		 */
2694 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2695 			return (error);
2696 	}
2697 
2698 	/*
2699 	 * XXX: Currently, we assume all the optional information is stored
2700 	 * in a single mbuf.
2701 	 */
2702 	if (control->m_next)
2703 		return (EINVAL);
2704 
2705 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2706 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2707 		int error;
2708 
2709 		if (control->m_len < CMSG_LEN(0))
2710 			return (EINVAL);
2711 
2712 		cm = mtod(control, struct cmsghdr *);
2713 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2714 			return (EINVAL);
2715 		if (cm->cmsg_level != IPPROTO_IPV6)
2716 			continue;
2717 
2718 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2719 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2720 		if (error)
2721 			return (error);
2722 	}
2723 
2724 	return (0);
2725 }
2726 
2727 /*
2728  * Set a particular packet option, as a sticky option or an ancillary data
2729  * item.  "len" can be 0 only when it's a sticky option.
2730  * We have 4 cases of combination of "sticky" and "cmsg":
2731  * "sticky=0, cmsg=0": impossible
2732  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2733  * "sticky=1, cmsg=0": RFC3542 socket option
2734  * "sticky=1, cmsg=1": RFC2292 socket option
2735  */
2736 static int
2737 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2738     struct ucred *cred, int sticky, int cmsg, int uproto)
2739 {
2740 	int minmtupolicy, preftemp;
2741 	int error;
2742 
2743 	NET_EPOCH_ASSERT();
2744 
2745 	if (!sticky && !cmsg) {
2746 #ifdef DIAGNOSTIC
2747 		printf("ip6_setpktopt: impossible case\n");
2748 #endif
2749 		return (EINVAL);
2750 	}
2751 
2752 	/*
2753 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2754 	 * not be specified in the context of RFC3542.  Conversely,
2755 	 * RFC3542 types should not be specified in the context of RFC2292.
2756 	 */
2757 	if (!cmsg) {
2758 		switch (optname) {
2759 		case IPV6_2292PKTINFO:
2760 		case IPV6_2292HOPLIMIT:
2761 		case IPV6_2292NEXTHOP:
2762 		case IPV6_2292HOPOPTS:
2763 		case IPV6_2292DSTOPTS:
2764 		case IPV6_2292RTHDR:
2765 		case IPV6_2292PKTOPTIONS:
2766 			return (ENOPROTOOPT);
2767 		}
2768 	}
2769 	if (sticky && cmsg) {
2770 		switch (optname) {
2771 		case IPV6_PKTINFO:
2772 		case IPV6_HOPLIMIT:
2773 		case IPV6_NEXTHOP:
2774 		case IPV6_HOPOPTS:
2775 		case IPV6_DSTOPTS:
2776 		case IPV6_RTHDRDSTOPTS:
2777 		case IPV6_RTHDR:
2778 		case IPV6_USE_MIN_MTU:
2779 		case IPV6_DONTFRAG:
2780 		case IPV6_TCLASS:
2781 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2782 			return (ENOPROTOOPT);
2783 		}
2784 	}
2785 
2786 	switch (optname) {
2787 	case IPV6_2292PKTINFO:
2788 	case IPV6_PKTINFO:
2789 	{
2790 		struct ifnet *ifp = NULL;
2791 		struct in6_pktinfo *pktinfo;
2792 
2793 		if (len != sizeof(struct in6_pktinfo))
2794 			return (EINVAL);
2795 
2796 		pktinfo = (struct in6_pktinfo *)buf;
2797 
2798 		/*
2799 		 * An application can clear any sticky IPV6_PKTINFO option by
2800 		 * doing a "regular" setsockopt with ipi6_addr being
2801 		 * in6addr_any and ipi6_ifindex being zero.
2802 		 * [RFC 3542, Section 6]
2803 		 */
2804 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2805 		    pktinfo->ipi6_ifindex == 0 &&
2806 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2807 			ip6_clearpktopts(opt, optname);
2808 			break;
2809 		}
2810 
2811 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2812 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2813 			return (EINVAL);
2814 		}
2815 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2816 			return (EINVAL);
2817 		/* validate the interface index if specified. */
2818 		if (pktinfo->ipi6_ifindex) {
2819 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2820 			if (ifp == NULL)
2821 				return (ENXIO);
2822 		}
2823 		if (ifp != NULL && (ifp->if_inet6 == NULL ||
2824 		    (ifp->if_inet6->nd_flags & ND6_IFF_IFDISABLED) != 0))
2825 			return (ENETDOWN);
2826 
2827 		if (ifp != NULL &&
2828 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2829 			struct in6_ifaddr *ia;
2830 
2831 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2832 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2833 			if (ia == NULL)
2834 				return (EADDRNOTAVAIL);
2835 			ifa_free(&ia->ia_ifa);
2836 		}
2837 		/*
2838 		 * We store the address anyway, and let in6_selectsrc()
2839 		 * validate the specified address.  This is because ipi6_addr
2840 		 * may not have enough information about its scope zone, and
2841 		 * we may need additional information (such as outgoing
2842 		 * interface or the scope zone of a destination address) to
2843 		 * disambiguate the scope.
2844 		 * XXX: the delay of the validation may confuse the
2845 		 * application when it is used as a sticky option.
2846 		 */
2847 		if (opt->ip6po_pktinfo == NULL) {
2848 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2849 			    M_IP6OPT, M_NOWAIT);
2850 			if (opt->ip6po_pktinfo == NULL)
2851 				return (ENOBUFS);
2852 		}
2853 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2854 		opt->ip6po_valid |= IP6PO_VALID_PKTINFO;
2855 		break;
2856 	}
2857 
2858 	case IPV6_2292HOPLIMIT:
2859 	case IPV6_HOPLIMIT:
2860 	{
2861 		int *hlimp;
2862 
2863 		/*
2864 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2865 		 * to simplify the ordering among hoplimit options.
2866 		 */
2867 		if (optname == IPV6_HOPLIMIT && sticky)
2868 			return (ENOPROTOOPT);
2869 
2870 		if (len != sizeof(int))
2871 			return (EINVAL);
2872 		hlimp = (int *)buf;
2873 		if (*hlimp < -1 || *hlimp > 255)
2874 			return (EINVAL);
2875 
2876 		opt->ip6po_hlim = *hlimp;
2877 		opt->ip6po_valid |= IP6PO_VALID_HLIM;
2878 		break;
2879 	}
2880 
2881 	case IPV6_TCLASS:
2882 	{
2883 		int tclass;
2884 
2885 		if (len != sizeof(int))
2886 			return (EINVAL);
2887 		tclass = *(int *)buf;
2888 		if (tclass < -1 || tclass > 255)
2889 			return (EINVAL);
2890 
2891 		opt->ip6po_tclass = tclass;
2892 		opt->ip6po_valid |= IP6PO_VALID_TC;
2893 		break;
2894 	}
2895 
2896 	case IPV6_2292NEXTHOP:
2897 	case IPV6_NEXTHOP:
2898 		if (cred != NULL) {
2899 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2900 			if (error)
2901 				return (error);
2902 		}
2903 
2904 		if (len == 0) {	/* just remove the option */
2905 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2906 			break;
2907 		}
2908 
2909 		/* check if cmsg_len is large enough for sa_len */
2910 		if (len < sizeof(struct sockaddr) || len < *buf)
2911 			return (EINVAL);
2912 
2913 		switch (((struct sockaddr *)buf)->sa_family) {
2914 		case AF_INET6:
2915 		{
2916 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2917 			int error;
2918 
2919 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2920 				return (EINVAL);
2921 
2922 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2923 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2924 				return (EINVAL);
2925 			}
2926 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2927 			    != 0) {
2928 				return (error);
2929 			}
2930 			break;
2931 		}
2932 		case AF_LINK:	/* should eventually be supported */
2933 		default:
2934 			return (EAFNOSUPPORT);
2935 		}
2936 
2937 		/* turn off the previous option, then set the new option. */
2938 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2939 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2940 		if (opt->ip6po_nexthop == NULL)
2941 			return (ENOBUFS);
2942 		bcopy(buf, opt->ip6po_nexthop, *buf);
2943 		opt->ip6po_valid |= IP6PO_VALID_NHINFO;
2944 		break;
2945 
2946 	case IPV6_2292HOPOPTS:
2947 	case IPV6_HOPOPTS:
2948 	{
2949 		struct ip6_hbh *hbh;
2950 		int hbhlen;
2951 
2952 		/*
2953 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2954 		 * options, since per-option restriction has too much
2955 		 * overhead.
2956 		 */
2957 		if (cred != NULL) {
2958 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2959 			if (error)
2960 				return (error);
2961 		}
2962 
2963 		if (len == 0) {
2964 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2965 			break;	/* just remove the option */
2966 		}
2967 
2968 		/* message length validation */
2969 		if (len < sizeof(struct ip6_hbh))
2970 			return (EINVAL);
2971 		hbh = (struct ip6_hbh *)buf;
2972 		hbhlen = (hbh->ip6h_len + 1) << 3;
2973 		if (len != hbhlen)
2974 			return (EINVAL);
2975 
2976 		/* turn off the previous option, then set the new option. */
2977 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2978 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2979 		if (opt->ip6po_hbh == NULL)
2980 			return (ENOBUFS);
2981 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2982 		opt->ip6po_valid |= IP6PO_VALID_HBH;
2983 
2984 		break;
2985 	}
2986 
2987 	case IPV6_2292DSTOPTS:
2988 	case IPV6_DSTOPTS:
2989 	case IPV6_RTHDRDSTOPTS:
2990 	{
2991 		struct ip6_dest *dest, **newdest = NULL;
2992 		int destlen;
2993 
2994 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2995 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2996 			if (error)
2997 				return (error);
2998 		}
2999 
3000 		if (len == 0) {
3001 			ip6_clearpktopts(opt, optname);
3002 			break;	/* just remove the option */
3003 		}
3004 
3005 		/* message length validation */
3006 		if (len < sizeof(struct ip6_dest))
3007 			return (EINVAL);
3008 		dest = (struct ip6_dest *)buf;
3009 		destlen = (dest->ip6d_len + 1) << 3;
3010 		if (len != destlen)
3011 			return (EINVAL);
3012 
3013 		/*
3014 		 * Determine the position that the destination options header
3015 		 * should be inserted; before or after the routing header.
3016 		 */
3017 		switch (optname) {
3018 		case IPV6_2292DSTOPTS:
3019 			/*
3020 			 * The old advacned API is ambiguous on this point.
3021 			 * Our approach is to determine the position based
3022 			 * according to the existence of a routing header.
3023 			 * Note, however, that this depends on the order of the
3024 			 * extension headers in the ancillary data; the 1st
3025 			 * part of the destination options header must appear
3026 			 * before the routing header in the ancillary data,
3027 			 * too.
3028 			 * RFC3542 solved the ambiguity by introducing
3029 			 * separate ancillary data or option types.
3030 			 */
3031 			if (opt->ip6po_rthdr == NULL)
3032 				newdest = &opt->ip6po_dest1;
3033 			else
3034 				newdest = &opt->ip6po_dest2;
3035 			break;
3036 		case IPV6_RTHDRDSTOPTS:
3037 			newdest = &opt->ip6po_dest1;
3038 			break;
3039 		case IPV6_DSTOPTS:
3040 			newdest = &opt->ip6po_dest2;
3041 			break;
3042 		}
3043 
3044 		/* turn off the previous option, then set the new option. */
3045 		ip6_clearpktopts(opt, optname);
3046 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3047 		if (*newdest == NULL)
3048 			return (ENOBUFS);
3049 		bcopy(dest, *newdest, destlen);
3050 		if (newdest == &opt->ip6po_dest1)
3051 			opt->ip6po_valid |= IP6PO_VALID_DEST1;
3052 		else
3053 			opt->ip6po_valid |= IP6PO_VALID_DEST2;
3054 
3055 		break;
3056 	}
3057 
3058 	case IPV6_2292RTHDR:
3059 	case IPV6_RTHDR:
3060 	{
3061 		struct ip6_rthdr *rth;
3062 		int rthlen;
3063 
3064 		if (len == 0) {
3065 			ip6_clearpktopts(opt, IPV6_RTHDR);
3066 			break;	/* just remove the option */
3067 		}
3068 
3069 		/* message length validation */
3070 		if (len < sizeof(struct ip6_rthdr))
3071 			return (EINVAL);
3072 		rth = (struct ip6_rthdr *)buf;
3073 		rthlen = (rth->ip6r_len + 1) << 3;
3074 		if (len != rthlen)
3075 			return (EINVAL);
3076 
3077 		switch (rth->ip6r_type) {
3078 		case IPV6_RTHDR_TYPE_0:
3079 			if (rth->ip6r_len == 0)	/* must contain one addr */
3080 				return (EINVAL);
3081 			if (rth->ip6r_len % 2) /* length must be even */
3082 				return (EINVAL);
3083 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3084 				return (EINVAL);
3085 			break;
3086 		default:
3087 			return (EINVAL);	/* not supported */
3088 		}
3089 
3090 		/* turn off the previous option */
3091 		ip6_clearpktopts(opt, IPV6_RTHDR);
3092 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3093 		if (opt->ip6po_rthdr == NULL)
3094 			return (ENOBUFS);
3095 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3096 		opt->ip6po_valid |= IP6PO_VALID_RHINFO;
3097 
3098 		break;
3099 	}
3100 
3101 	case IPV6_USE_MIN_MTU:
3102 		if (len != sizeof(int))
3103 			return (EINVAL);
3104 		minmtupolicy = *(int *)buf;
3105 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3106 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3107 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3108 			return (EINVAL);
3109 		}
3110 		opt->ip6po_minmtu = minmtupolicy;
3111 		break;
3112 
3113 	case IPV6_DONTFRAG:
3114 		if (len != sizeof(int))
3115 			return (EINVAL);
3116 
3117 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3118 			/*
3119 			 * we ignore this option for TCP sockets.
3120 			 * (RFC3542 leaves this case unspecified.)
3121 			 */
3122 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3123 		} else
3124 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3125 		break;
3126 
3127 	case IPV6_PREFER_TEMPADDR:
3128 		if (len != sizeof(int))
3129 			return (EINVAL);
3130 		preftemp = *(int *)buf;
3131 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3132 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3133 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3134 			return (EINVAL);
3135 		}
3136 		opt->ip6po_prefer_tempaddr = preftemp;
3137 		break;
3138 
3139 	default:
3140 		return (ENOPROTOOPT);
3141 	} /* end of switch */
3142 
3143 	return (0);
3144 }
3145 
3146 /*
3147  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3148  * packet to the input queue of a specified interface.  Note that this
3149  * calls the output routine of the loopback "driver", but with an interface
3150  * pointer that might NOT be &loif -- easier than replicating that code here.
3151  */
3152 void
3153 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3154 {
3155 	struct mbuf *copym;
3156 	struct ip6_hdr *ip6;
3157 
3158 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3159 	if (copym == NULL)
3160 		return;
3161 
3162 	/*
3163 	 * Make sure to deep-copy IPv6 header portion in case the data
3164 	 * is in an mbuf cluster, so that we can safely override the IPv6
3165 	 * header portion later.
3166 	 */
3167 	if (!M_WRITABLE(copym) ||
3168 	    copym->m_len < sizeof(struct ip6_hdr)) {
3169 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3170 		if (copym == NULL)
3171 			return;
3172 	}
3173 	ip6 = mtod(copym, struct ip6_hdr *);
3174 	/*
3175 	 * clear embedded scope identifiers if necessary.
3176 	 * in6_clearscope will touch the addresses only when necessary.
3177 	 */
3178 	in6_clearscope(&ip6->ip6_src);
3179 	in6_clearscope(&ip6->ip6_dst);
3180 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3181 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3182 		    CSUM_PSEUDO_HDR;
3183 		copym->m_pkthdr.csum_data = 0xffff;
3184 	}
3185 	if_simloop(ifp, copym, AF_INET6, 0);
3186 }
3187 
3188 /*
3189  * Chop IPv6 header off from the payload.
3190  */
3191 static int
3192 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3193 {
3194 	struct mbuf *mh;
3195 	struct ip6_hdr *ip6;
3196 
3197 	ip6 = mtod(m, struct ip6_hdr *);
3198 	if (m->m_len > sizeof(*ip6)) {
3199 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3200 		if (mh == NULL) {
3201 			m_freem(m);
3202 			return ENOBUFS;
3203 		}
3204 		m_move_pkthdr(mh, m);
3205 		M_ALIGN(mh, sizeof(*ip6));
3206 		m->m_len -= sizeof(*ip6);
3207 		m->m_data += sizeof(*ip6);
3208 		mh->m_next = m;
3209 		m = mh;
3210 		m->m_len = sizeof(*ip6);
3211 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3212 	}
3213 	exthdrs->ip6e_ip6 = m;
3214 	return 0;
3215 }
3216 
3217 /*
3218  * Compute IPv6 extension header length.
3219  */
3220 int
3221 ip6_optlen(struct inpcb *inp)
3222 {
3223 	int len;
3224 
3225 	if (!inp->in6p_outputopts)
3226 		return 0;
3227 
3228 	len = 0;
3229 #define elen(x) \
3230     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3231 
3232 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3233 	if (inp->in6p_outputopts->ip6po_rthdr)
3234 		/* dest1 is valid with rthdr only */
3235 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3236 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3237 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3238 	return len;
3239 #undef elen
3240 }
3241