xref: /freebsd/sys/netinet6/ip6_output.c (revision 6580f5c38dd5b01aeeaed16b370f1a12423437f0)
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 <sys/cdefs.h>
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 #include "opt_kern_tls.h"
68 #include "opt_ratelimit.h"
69 #include "opt_route.h"
70 #include "opt_rss.h"
71 #include "opt_sctp.h"
72 
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/ktls.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/errno.h>
79 #include <sys/priv.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/syslog.h>
85 #include <sys/ucred.h>
86 
87 #include <machine/in_cksum.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/if_private.h>
92 #include <net/if_vlan_var.h>
93 #include <net/if_llatbl.h>
94 #include <net/ethernet.h>
95 #include <net/netisr.h>
96 #include <net/route.h>
97 #include <net/route/nhop.h>
98 #include <net/pfil.h>
99 #include <net/rss_config.h>
100 #include <net/vnet.h>
101 
102 #include <netinet/in.h>
103 #include <netinet/in_var.h>
104 #include <netinet/ip_var.h>
105 #include <netinet6/in6_fib.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet/ip6.h>
108 #include <netinet/icmp6.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet/in_pcb.h>
111 #include <netinet/tcp_var.h>
112 #include <netinet6/nd6.h>
113 #include <netinet6/in6_rss.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_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
147 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
148 static int ip6_getpmtu(struct route_in6 *, int,
149 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
150 	u_int);
151 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
152 	u_long *, int *, u_int);
153 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
154 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
155 
156 /*
157  * Make an extension header from option data.  hp is the source,
158  * mp is the destination, and _ol is the optlen.
159  */
160 #define	MAKE_EXTHDR(hp, mp, _ol)					\
161     do {								\
162 	struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
163 	error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
164 	    ((eh)->ip6e_len + 1) << 3);				\
165 	if (error)						\
166 		goto freehdrs;					\
167 	(_ol) += (*(mp))->m_len;				\
168     } while (/*CONSTCOND*/ 0)
169 
170 /*
171  * Form a chain of extension headers.
172  * m is the extension header mbuf
173  * mp is the previous mbuf in the chain
174  * p is the next header
175  * i is the type of option.
176  */
177 #define MAKE_CHAIN(m, mp, p, i)\
178     do {\
179 	if (m) {\
180 		if (!hdrsplit) \
181 			panic("%s:%d: assumption failed: "\
182 			    "hdr not split: hdrsplit %d exthdrs %p",\
183 			    __func__, __LINE__, hdrsplit, &exthdrs);\
184 		*mtod((m), u_char *) = *(p);\
185 		*(p) = (i);\
186 		p = mtod((m), u_char *);\
187 		(m)->m_next = (mp)->m_next;\
188 		(mp)->m_next = (m);\
189 		(mp) = (m);\
190 	}\
191     } while (/*CONSTCOND*/ 0)
192 
193 void
194 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
195 {
196 	u_short csum;
197 
198 	csum = in_cksum_skip(m, offset + plen, offset);
199 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
200 		csum = 0xffff;
201 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
202 
203 	if (offset + sizeof(csum) > m->m_len)
204 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
205 	else
206 		*(u_short *)mtodo(m, offset) = csum;
207 }
208 
209 static void
210 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
211     int plen, int optlen)
212 {
213 
214 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
215 	    "csum_flags %#x",
216 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
217 
218 	if (csum_flags & CSUM_DELAY_DATA_IPV6) {
219 		in6_delayed_cksum(m, plen - optlen,
220 		    sizeof(struct ip6_hdr) + optlen);
221 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
222 	}
223 #if defined(SCTP) || defined(SCTP_SUPPORT)
224 	if (csum_flags & CSUM_SCTP_IPV6) {
225 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
226 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
227 	}
228 #endif
229 }
230 
231 int
232 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
233     int fraglen , uint32_t id)
234 {
235 	struct mbuf *m, **mnext, *m_frgpart;
236 	struct ip6_hdr *ip6, *mhip6;
237 	struct ip6_frag *ip6f;
238 	int off;
239 	int error;
240 	int tlen = m0->m_pkthdr.len;
241 
242 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
243 
244 	m = m0;
245 	ip6 = mtod(m, struct ip6_hdr *);
246 	mnext = &m->m_nextpkt;
247 
248 	for (off = hlen; off < tlen; off += fraglen) {
249 		m = m_gethdr(M_NOWAIT, MT_DATA);
250 		if (!m) {
251 			IP6STAT_INC(ip6s_odropped);
252 			return (ENOBUFS);
253 		}
254 
255 		/*
256 		 * Make sure the complete packet header gets copied
257 		 * from the originating mbuf to the newly created
258 		 * mbuf. This also ensures that existing firewall
259 		 * classification(s), VLAN tags and so on get copied
260 		 * to the resulting fragmented packet(s):
261 		 */
262 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
263 			m_free(m);
264 			IP6STAT_INC(ip6s_odropped);
265 			return (ENOBUFS);
266 		}
267 
268 		*mnext = m;
269 		mnext = &m->m_nextpkt;
270 		m->m_data += max_linkhdr;
271 		mhip6 = mtod(m, struct ip6_hdr *);
272 		*mhip6 = *ip6;
273 		m->m_len = sizeof(*mhip6);
274 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
275 		if (error) {
276 			IP6STAT_INC(ip6s_odropped);
277 			return (error);
278 		}
279 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
280 		if (off + fraglen >= tlen)
281 			fraglen = tlen - off;
282 		else
283 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
284 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
285 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
286 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
287 			IP6STAT_INC(ip6s_odropped);
288 			return (ENOBUFS);
289 		}
290 		m_cat(m, m_frgpart);
291 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
292 		ip6f->ip6f_reserved = 0;
293 		ip6f->ip6f_ident = id;
294 		ip6f->ip6f_nxt = nextproto;
295 		IP6STAT_INC(ip6s_ofragments);
296 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
297 	}
298 
299 	return (0);
300 }
301 
302 static int
303 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
304     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
305     bool stamp_tag)
306 {
307 #ifdef KERN_TLS
308 	struct ktls_session *tls = NULL;
309 #endif
310 	struct m_snd_tag *mst;
311 	int error;
312 
313 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
314 	mst = NULL;
315 
316 #ifdef KERN_TLS
317 	/*
318 	 * If this is an unencrypted TLS record, save a reference to
319 	 * the record.  This local reference is used to call
320 	 * ktls_output_eagain after the mbuf has been freed (thus
321 	 * dropping the mbuf's reference) in if_output.
322 	 */
323 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
324 		tls = ktls_hold(m->m_next->m_epg_tls);
325 		mst = tls->snd_tag;
326 
327 		/*
328 		 * If a TLS session doesn't have a valid tag, it must
329 		 * have had an earlier ifp mismatch, so drop this
330 		 * packet.
331 		 */
332 		if (mst == NULL) {
333 			m_freem(m);
334 			error = EAGAIN;
335 			goto done;
336 		}
337 		/*
338 		 * Always stamp tags that include NIC ktls.
339 		 */
340 		stamp_tag = true;
341 	}
342 #endif
343 #ifdef RATELIMIT
344 	if (inp != NULL && mst == NULL) {
345 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
346 		    (inp->inp_snd_tag != NULL &&
347 		    inp->inp_snd_tag->ifp != ifp))
348 			in_pcboutput_txrtlmt(inp, ifp, m);
349 
350 		if (inp->inp_snd_tag != NULL)
351 			mst = inp->inp_snd_tag;
352 	}
353 #endif
354 	if (stamp_tag && mst != NULL) {
355 		KASSERT(m->m_pkthdr.rcvif == NULL,
356 		    ("trying to add a send tag to a forwarded packet"));
357 		if (mst->ifp != ifp) {
358 			m_freem(m);
359 			error = EAGAIN;
360 			goto done;
361 		}
362 
363 		/* stamp send tag on mbuf */
364 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
365 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
366 	}
367 
368 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
369 
370 done:
371 	/* Check for route change invalidating send tags. */
372 #ifdef KERN_TLS
373 	if (tls != NULL) {
374 		if (error == EAGAIN)
375 			error = ktls_output_eagain(inp, tls);
376 		ktls_free(tls);
377 	}
378 #endif
379 #ifdef RATELIMIT
380 	if (error == EAGAIN)
381 		in_pcboutput_eagain(inp);
382 #endif
383 	return (error);
384 }
385 
386 /*
387  * IP6 output.
388  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
389  * nxt, hlim, src, dst).
390  * This function may modify ver and hlim only.
391  * The mbuf chain containing the packet will be freed.
392  * The mbuf opt, if present, will not be freed.
393  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
394  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
395  * then result of route lookup is stored in ro->ro_nh.
396  *
397  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
398  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
399  *
400  * ifpp - XXX: just for statistics
401  */
402 int
403 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
404     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
405     struct ifnet **ifpp, struct inpcb *inp)
406 {
407 	struct ip6_hdr *ip6;
408 	struct ifnet *ifp, *origifp;
409 	struct mbuf *m = m0;
410 	struct mbuf *mprev;
411 	struct route_in6 *ro_pmtu;
412 	struct nhop_object *nh;
413 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
414 	struct in6_addr odst;
415 	u_char *nexthdrp;
416 	int tlen, len;
417 	int error = 0;
418 	int vlan_pcp = -1;
419 	struct in6_ifaddr *ia = NULL;
420 	u_long mtu;
421 	int alwaysfrag, dontfrag;
422 	u_int32_t optlen, plen = 0, unfragpartlen;
423 	struct ip6_exthdrs exthdrs;
424 	struct in6_addr src0, dst0;
425 	u_int32_t zone;
426 	bool hdrsplit;
427 	int sw_csum, tso;
428 	int needfiblookup;
429 	uint32_t fibnum;
430 	struct m_tag *fwd_tag = NULL;
431 	uint32_t id;
432 	uint32_t optvalid;
433 
434 	NET_EPOCH_ASSERT();
435 
436 	if (inp != NULL) {
437 		INP_LOCK_ASSERT(inp);
438 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
439 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
440 			/* Unconditionally set flowid. */
441 			m->m_pkthdr.flowid = inp->inp_flowid;
442 			M_HASHTYPE_SET(m, inp->inp_flowtype);
443 		}
444 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
445 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
446 			    INP_2PCP_SHIFT;
447 #ifdef NUMA
448 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
449 #endif
450 	}
451 
452 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
453 	/*
454 	 * IPSec checking which handles several cases.
455 	 * FAST IPSEC: We re-injected the packet.
456 	 * XXX: need scope argument.
457 	 */
458 	if (IPSEC_ENABLED(ipv6)) {
459 		m = mb_unmapped_to_ext(m);
460 		if (m == NULL) {
461 			IP6STAT_INC(ip6s_odropped);
462 			error = ENOBUFS;
463 			goto bad;
464 		}
465 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
466 			if (error == EINPROGRESS)
467 				error = 0;
468 			goto done;
469 		}
470 	}
471 #endif /* IPSEC */
472 
473 	/* Source address validation. */
474 	ip6 = mtod(m, struct ip6_hdr *);
475 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
476 	    (flags & IPV6_UNSPECSRC) == 0) {
477 		error = EOPNOTSUPP;
478 		IP6STAT_INC(ip6s_badscope);
479 		goto bad;
480 	}
481 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
482 		error = EOPNOTSUPP;
483 		IP6STAT_INC(ip6s_badscope);
484 		goto bad;
485 	}
486 
487 	/*
488 	 * If we are given packet options to add extension headers prepare them.
489 	 * Calculate the total length of the extension header chain.
490 	 * Keep the length of the unfragmentable part for fragmentation.
491 	 */
492 	bzero(&exthdrs, sizeof(exthdrs));
493 	optlen = optvalid = 0;
494 	unfragpartlen = sizeof(struct ip6_hdr);
495 	if (opt) {
496 		optvalid = opt->ip6po_valid;
497 
498 		/* Hop-by-Hop options header. */
499 		if ((optvalid & IP6PO_VALID_HBH) != 0)
500 			MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
501 
502 		/* Destination options header (1st part). */
503 		if ((optvalid & IP6PO_VALID_RHINFO) != 0) {
504 #ifndef RTHDR_SUPPORT_IMPLEMENTED
505 			/*
506 			 * If there is a routing header, discard the packet
507 			 * right away here. RH0/1 are obsolete and we do not
508 			 * currently support RH2/3/4.
509 			 * People trying to use RH253/254 may want to disable
510 			 * this check.
511 			 * The moment we do support any routing header (again)
512 			 * this block should check the routing type more
513 			 * selectively.
514 			 */
515 			error = EINVAL;
516 			goto bad;
517 #endif
518 
519 			/*
520 			 * Destination options header (1st part).
521 			 * This only makes sense with a routing header.
522 			 * See Section 9.2 of RFC 3542.
523 			 * Disabling this part just for MIP6 convenience is
524 			 * a bad idea.  We need to think carefully about a
525 			 * way to make the advanced API coexist with MIP6
526 			 * options, which might automatically be inserted in
527 			 * the kernel.
528 			 */
529 			if ((optvalid & IP6PO_VALID_DEST1) != 0)
530 				MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
531 				    optlen);
532 		}
533 		/* Routing header. */
534 		if ((optvalid & IP6PO_VALID_RHINFO) != 0)
535 			MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
536 
537 		unfragpartlen += optlen;
538 
539 		/*
540 		 * NOTE: we don't add AH/ESP length here (done in
541 		 * ip6_ipsec_output()).
542 		 */
543 
544 		/* Destination options header (2nd part). */
545 		if ((optvalid & IP6PO_VALID_DEST2) != 0)
546 			MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
547 	}
548 
549 	/*
550 	 * If there is at least one extension header,
551 	 * separate IP6 header from the payload.
552 	 */
553 	hdrsplit = false;
554 	if (optlen) {
555 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
556 			m = NULL;
557 			goto freehdrs;
558 		}
559 		m = exthdrs.ip6e_ip6;
560 		ip6 = mtod(m, struct ip6_hdr *);
561 		hdrsplit = true;
562 	}
563 
564 	/* Adjust mbuf packet header length. */
565 	m->m_pkthdr.len += optlen;
566 	plen = m->m_pkthdr.len - sizeof(*ip6);
567 
568 	/* If this is a jumbo payload, insert a jumbo payload option. */
569 	if (plen > IPV6_MAXPACKET) {
570 		if (!hdrsplit) {
571 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
572 				m = NULL;
573 				goto freehdrs;
574 			}
575 			m = exthdrs.ip6e_ip6;
576 			ip6 = mtod(m, struct ip6_hdr *);
577 			hdrsplit = true;
578 		}
579 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
580 			goto freehdrs;
581 		ip6->ip6_plen = 0;
582 	} else
583 		ip6->ip6_plen = htons(plen);
584 	nexthdrp = &ip6->ip6_nxt;
585 
586 	if (optlen) {
587 		/*
588 		 * Concatenate headers and fill in next header fields.
589 		 * Here we have, on "m"
590 		 *	IPv6 payload
591 		 * and we insert headers accordingly.
592 		 * Finally, we should be getting:
593 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
594 		 *
595 		 * During the header composing process "m" points to IPv6
596 		 * header.  "mprev" points to an extension header prior to esp.
597 		 */
598 		mprev = m;
599 
600 		/*
601 		 * We treat dest2 specially.  This makes IPsec processing
602 		 * much easier.  The goal here is to make mprev point the
603 		 * mbuf prior to dest2.
604 		 *
605 		 * Result: IPv6 dest2 payload.
606 		 * m and mprev will point to IPv6 header.
607 		 */
608 		if (exthdrs.ip6e_dest2) {
609 			if (!hdrsplit)
610 				panic("%s:%d: assumption failed: "
611 				    "hdr not split: hdrsplit %d exthdrs %p",
612 				    __func__, __LINE__, hdrsplit, &exthdrs);
613 			exthdrs.ip6e_dest2->m_next = m->m_next;
614 			m->m_next = exthdrs.ip6e_dest2;
615 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
616 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
617 		}
618 
619 		/*
620 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
621 		 * m will point to IPv6 header.  mprev will point to the
622 		 * extension header prior to dest2 (rthdr in the above case).
623 		 */
624 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
625 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
626 			   IPPROTO_DSTOPTS);
627 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
628 			   IPPROTO_ROUTING);
629 	}
630 
631 	IP6STAT_INC(ip6s_localout);
632 
633 	/* Route packet. */
634 	ro_pmtu = ro;
635 	if ((optvalid & IP6PO_VALID_RHINFO) != 0)
636 		ro = &opt->ip6po_route;
637 	if (ro != NULL)
638 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
639 	else
640 		dst = &sin6;
641 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
642 
643 again:
644 	/*
645 	 * If specified, try to fill in the traffic class field.
646 	 * Do not override if a non-zero value is already set.
647 	 * We check the diffserv field and the ECN field separately.
648 	 */
649 	if ((optvalid & IP6PO_VALID_TC) != 0){
650 		int mask = 0;
651 
652 		if (IPV6_DSCP(ip6) == 0)
653 			mask |= 0xfc;
654 		if (IPV6_ECN(ip6) == 0)
655 			mask |= 0x03;
656 		if (mask != 0)
657 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
658 	}
659 
660 	/* Fill in or override the hop limit field, if necessary. */
661 	if ((optvalid & IP6PO_VALID_HLIM) != 0)
662 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
663 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
664 		if (im6o != NULL)
665 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
666 		else
667 			ip6->ip6_hlim = V_ip6_defmcasthlim;
668 	}
669 
670 	if (ro == NULL || ro->ro_nh == NULL) {
671 		bzero(dst, sizeof(*dst));
672 		dst->sin6_family = AF_INET6;
673 		dst->sin6_len = sizeof(*dst);
674 		dst->sin6_addr = ip6->ip6_dst;
675 	}
676 	/*
677 	 * Validate route against routing table changes.
678 	 * Make sure that the address family is set in route.
679 	 */
680 	nh = NULL;
681 	ifp = NULL;
682 	mtu = 0;
683 	if (ro != NULL) {
684 		if (ro->ro_nh != NULL && inp != NULL) {
685 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
686 			NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
687 			    fibnum);
688 		}
689 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
690 		    (!NH_IS_VALID(ro->ro_nh) ||
691 		    ro->ro_dst.sin6_family != AF_INET6 ||
692 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
693 			RO_INVALIDATE_CACHE(ro);
694 
695 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
696 		    ro->ro_dst.sin6_family == AF_INET6 &&
697 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
698 			/* Nexthop is valid and contains valid ifp */
699 			nh = ro->ro_nh;
700 		} else {
701 			if (ro->ro_lle)
702 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
703 			ro->ro_lle = NULL;
704 			if (fwd_tag == NULL) {
705 				bzero(&dst_sa, sizeof(dst_sa));
706 				dst_sa.sin6_family = AF_INET6;
707 				dst_sa.sin6_len = sizeof(dst_sa);
708 				dst_sa.sin6_addr = ip6->ip6_dst;
709 			}
710 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
711 			    &nh, fibnum, m->m_pkthdr.flowid);
712 			if (error != 0) {
713 				IP6STAT_INC(ip6s_noroute);
714 				if (ifp != NULL)
715 					in6_ifstat_inc(ifp, ifs6_out_discard);
716 				goto bad;
717 			}
718 			/*
719 			 * At this point at least @ifp is not NULL
720 			 * Can be the case when dst is multicast, link-local or
721 			 * interface is explicitly specificed by the caller.
722 			 */
723 		}
724 		if (nh == NULL) {
725 			/*
726 			 * If in6_selectroute() does not return a nexthop
727 			 * dst may not have been updated.
728 			 */
729 			*dst = dst_sa;	/* XXX */
730 			origifp = ifp;
731 			mtu = ifp->if_mtu;
732 		} else {
733 			ifp = nh->nh_ifp;
734 			origifp = nh->nh_aifp;
735 			ia = (struct in6_ifaddr *)(nh->nh_ifa);
736 			counter_u64_add(nh->nh_pksent, 1);
737 		}
738 	} else {
739 		struct nhop_object *nh;
740 		struct in6_addr kdst;
741 		uint32_t scopeid;
742 
743 		if (fwd_tag == NULL) {
744 			bzero(&dst_sa, sizeof(dst_sa));
745 			dst_sa.sin6_family = AF_INET6;
746 			dst_sa.sin6_len = sizeof(dst_sa);
747 			dst_sa.sin6_addr = ip6->ip6_dst;
748 		}
749 
750 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
751 		    im6o != NULL &&
752 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
753 			/* We do not need a route lookup. */
754 			*dst = dst_sa;	/* XXX */
755 			origifp = ifp;
756 			goto nonh6lookup;
757 		}
758 
759 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
760 
761 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
762 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
763 			if (scopeid > 0) {
764 				ifp = in6_getlinkifnet(scopeid);
765 				if (ifp == NULL) {
766 					error = EHOSTUNREACH;
767 					goto bad;
768 				}
769 				*dst = dst_sa;	/* XXX */
770 				origifp = ifp;
771 				goto nonh6lookup;
772 			}
773 		}
774 
775 		nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
776 		    m->m_pkthdr.flowid);
777 		if (nh == NULL) {
778 			IP6STAT_INC(ip6s_noroute);
779 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
780 			error = EHOSTUNREACH;
781 			goto bad;
782 		}
783 
784 		ifp = nh->nh_ifp;
785 		origifp = nh->nh_aifp;
786 		ia = ifatoia6(nh->nh_ifa);
787 		if (nh->nh_flags & NHF_GATEWAY)
788 			dst->sin6_addr = nh->gw6_sa.sin6_addr;
789 		else if (fwd_tag != NULL)
790 			dst->sin6_addr = dst_sa.sin6_addr;
791 nonh6lookup:
792 		;
793 	}
794 	/*
795 	 * At this point ifp MUST be pointing to the valid transmit ifp.
796 	 * origifp MUST be valid and pointing to either the same ifp or,
797 	 * in case of loopback output, to the interface which ip6_src
798 	 * belongs to.
799 	 * Examples:
800 	 *  fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
801 	 *  fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
802 	 *  ::1 -> ::1 -> ifp=lo0, origifp=lo0
803 	 *
804 	 * mtu can be 0 and will be refined later.
805 	 */
806 	KASSERT((ifp != NULL), ("output interface must not be NULL"));
807 	KASSERT((origifp != NULL), ("output address interface must not be NULL"));
808 
809 	if ((flags & IPV6_FORWARDING) == 0) {
810 		/* XXX: the FORWARDING flag can be set for mrouting. */
811 		in6_ifstat_inc(ifp, ifs6_out_request);
812 	}
813 
814 	/* Setup data structures for scope ID checks. */
815 	src0 = ip6->ip6_src;
816 	bzero(&src_sa, sizeof(src_sa));
817 	src_sa.sin6_family = AF_INET6;
818 	src_sa.sin6_len = sizeof(src_sa);
819 	src_sa.sin6_addr = ip6->ip6_src;
820 
821 	dst0 = ip6->ip6_dst;
822 	/* Re-initialize to be sure. */
823 	bzero(&dst_sa, sizeof(dst_sa));
824 	dst_sa.sin6_family = AF_INET6;
825 	dst_sa.sin6_len = sizeof(dst_sa);
826 	dst_sa.sin6_addr = ip6->ip6_dst;
827 
828 	/* Check for valid scope ID. */
829 	if (in6_setscope(&src0, origifp, &zone) == 0 &&
830 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
831 	    in6_setscope(&dst0, origifp, &zone) == 0 &&
832 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
833 		/*
834 		 * The outgoing interface is in the zone of the source
835 		 * and destination addresses.
836 		 *
837 		 */
838 	} else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
839 	    sa6_recoverscope(&src_sa) != 0 ||
840 	    sa6_recoverscope(&dst_sa) != 0 ||
841 	    dst_sa.sin6_scope_id == 0 ||
842 	    (src_sa.sin6_scope_id != 0 &&
843 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
844 	    ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
845 		/*
846 		 * If the destination network interface is not a
847 		 * loopback interface, or the destination network
848 		 * address has no scope ID, or the source address has
849 		 * a scope ID set which is different from the
850 		 * destination address one, or there is no network
851 		 * interface representing this scope ID, the address
852 		 * pair is considered invalid.
853 		 */
854 		IP6STAT_INC(ip6s_badscope);
855 		in6_ifstat_inc(origifp, ifs6_out_discard);
856 		if (error == 0)
857 			error = EHOSTUNREACH; /* XXX */
858 		goto bad;
859 	}
860 	/* All scope ID checks are successful. */
861 
862 	if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
863 		if ((optvalid & IP6PO_VALID_NHINFO) != 0) {
864 			/*
865 			 * The nexthop is explicitly specified by the
866 			 * application.  We assume the next hop is an IPv6
867 			 * address.
868 			 */
869 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
870 		}
871 		else if ((nh->nh_flags & NHF_GATEWAY))
872 			dst = &nh->gw6_sa;
873 	}
874 
875 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
876 		m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
877 	} else {
878 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
879 		in6_ifstat_inc(ifp, ifs6_out_mcast);
880 
881 		/* Confirm that the outgoing interface supports multicast. */
882 		if (!(ifp->if_flags & IFF_MULTICAST)) {
883 			IP6STAT_INC(ip6s_noroute);
884 			in6_ifstat_inc(ifp, ifs6_out_discard);
885 			error = ENETUNREACH;
886 			goto bad;
887 		}
888 		if ((im6o == NULL && in6_mcast_loop) ||
889 		    (im6o && im6o->im6o_multicast_loop)) {
890 			/*
891 			 * Loop back multicast datagram if not expressly
892 			 * forbidden to do so, even if we have not joined
893 			 * the address; protocols will filter it later,
894 			 * thus deferring a hash lookup and lock acquisition
895 			 * at the expense of an m_copym().
896 			 */
897 			ip6_mloopback(ifp, m);
898 		} else {
899 			/*
900 			 * If we are acting as a multicast router, perform
901 			 * multicast forwarding as if the packet had just
902 			 * arrived on the interface to which we are about
903 			 * to send.  The multicast forwarding function
904 			 * recursively calls this function, using the
905 			 * IPV6_FORWARDING flag to prevent infinite recursion.
906 			 *
907 			 * Multicasts that are looped back by ip6_mloopback(),
908 			 * above, will be forwarded by the ip6_input() routine,
909 			 * if necessary.
910 			 */
911 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
912 				/*
913 				 * XXX: ip6_mforward expects that rcvif is NULL
914 				 * when it is called from the originating path.
915 				 * However, it may not always be the case.
916 				 */
917 				m->m_pkthdr.rcvif = NULL;
918 				if (ip6_mforward(ip6, ifp, m) != 0) {
919 					m_freem(m);
920 					goto done;
921 				}
922 			}
923 		}
924 		/*
925 		 * Multicasts with a hoplimit of zero may be looped back,
926 		 * above, but must not be transmitted on a network.
927 		 * Also, multicasts addressed to the loopback interface
928 		 * are not sent -- the above call to ip6_mloopback() will
929 		 * loop back a copy if this host actually belongs to the
930 		 * destination group on the loopback interface.
931 		 */
932 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
933 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
934 			m_freem(m);
935 			goto done;
936 		}
937 	}
938 
939 	/*
940 	 * Fill the outgoing inteface to tell the upper layer
941 	 * to increment per-interface statistics.
942 	 */
943 	if (ifpp)
944 		*ifpp = ifp;
945 
946 	/* Determine path MTU. */
947 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
948 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
949 		goto bad;
950 	KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
951 	    "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
952 	    ifp, alwaysfrag, fibnum));
953 
954 	/*
955 	 * The caller of this function may specify to use the minimum MTU
956 	 * in some cases.
957 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
958 	 * setting.  The logic is a bit complicated; by default, unicast
959 	 * packets will follow path MTU while multicast packets will be sent at
960 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
961 	 * including unicast ones will be sent at the minimum MTU.  Multicast
962 	 * packets will always be sent at the minimum MTU unless
963 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
964 	 * See RFC 3542 for more details.
965 	 */
966 	if (mtu > IPV6_MMTU) {
967 		if ((flags & IPV6_MINMTU))
968 			mtu = IPV6_MMTU;
969 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
970 			mtu = IPV6_MMTU;
971 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
972 			 (opt == NULL ||
973 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
974 			mtu = IPV6_MMTU;
975 		}
976 	}
977 
978 	/*
979 	 * Clear embedded scope identifiers if necessary.
980 	 * in6_clearscope() will touch the addresses only when necessary.
981 	 */
982 	in6_clearscope(&ip6->ip6_src);
983 	in6_clearscope(&ip6->ip6_dst);
984 
985 	/*
986 	 * If the outgoing packet contains a hop-by-hop options header,
987 	 * it must be examined and processed even by the source node.
988 	 * (RFC 2460, section 4.)
989 	 */
990 	if (exthdrs.ip6e_hbh) {
991 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
992 		u_int32_t dummy; /* XXX unused */
993 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
994 
995 #ifdef DIAGNOSTIC
996 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
997 			panic("ip6e_hbh is not contiguous");
998 #endif
999 		/*
1000 		 *  XXX: if we have to send an ICMPv6 error to the sender,
1001 		 *       we need the M_LOOP flag since icmp6_error() expects
1002 		 *       the IPv6 and the hop-by-hop options header are
1003 		 *       contiguous unless the flag is set.
1004 		 */
1005 		m->m_flags |= M_LOOP;
1006 		m->m_pkthdr.rcvif = ifp;
1007 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1008 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1009 		    &dummy, &plen) < 0) {
1010 			/* m was already freed at this point. */
1011 			error = EINVAL;/* better error? */
1012 			goto done;
1013 		}
1014 		m->m_flags &= ~M_LOOP; /* XXX */
1015 		m->m_pkthdr.rcvif = NULL;
1016 	}
1017 
1018 	/* Jump over all PFIL processing if hooks are not active. */
1019 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1020 		goto passout;
1021 
1022 	odst = ip6->ip6_dst;
1023 	/* Run through list of hooks for output packets. */
1024 	switch (pfil_mbuf_out(V_inet6_pfil_head, &m, ifp, inp)) {
1025 	case PFIL_PASS:
1026 		ip6 = mtod(m, struct ip6_hdr *);
1027 		break;
1028 	case PFIL_DROPPED:
1029 		error = EACCES;
1030 		/* FALLTHROUGH */
1031 	case PFIL_CONSUMED:
1032 		goto done;
1033 	}
1034 
1035 	needfiblookup = 0;
1036 	/* See if destination IP address was changed by packet filter. */
1037 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1038 		m->m_flags |= M_SKIP_FIREWALL;
1039 		/* If destination is now ourself drop to ip6_input(). */
1040 		if (in6_localip(&ip6->ip6_dst)) {
1041 			m->m_flags |= M_FASTFWD_OURS;
1042 			if (m->m_pkthdr.rcvif == NULL)
1043 				m->m_pkthdr.rcvif = V_loif;
1044 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1045 				m->m_pkthdr.csum_flags |=
1046 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1047 				m->m_pkthdr.csum_data = 0xffff;
1048 			}
1049 #if defined(SCTP) || defined(SCTP_SUPPORT)
1050 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1051 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1052 #endif
1053 			error = netisr_queue(NETISR_IPV6, m);
1054 			goto done;
1055 		} else {
1056 			if (ro != NULL)
1057 				RO_INVALIDATE_CACHE(ro);
1058 			needfiblookup = 1; /* Redo the routing table lookup. */
1059 		}
1060 	}
1061 	/* See if fib was changed by packet filter. */
1062 	if (fibnum != M_GETFIB(m)) {
1063 		m->m_flags |= M_SKIP_FIREWALL;
1064 		fibnum = M_GETFIB(m);
1065 		if (ro != NULL)
1066 			RO_INVALIDATE_CACHE(ro);
1067 		needfiblookup = 1;
1068 	}
1069 	if (needfiblookup)
1070 		goto again;
1071 
1072 	/* See if local, if yes, send it to netisr. */
1073 	if (m->m_flags & M_FASTFWD_OURS) {
1074 		if (m->m_pkthdr.rcvif == NULL)
1075 			m->m_pkthdr.rcvif = V_loif;
1076 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1077 			m->m_pkthdr.csum_flags |=
1078 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1079 			m->m_pkthdr.csum_data = 0xffff;
1080 		}
1081 #if defined(SCTP) || defined(SCTP_SUPPORT)
1082 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1083 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1084 #endif
1085 		error = netisr_queue(NETISR_IPV6, m);
1086 		goto done;
1087 	}
1088 	/* Or forward to some other address? */
1089 	if ((m->m_flags & M_IP6_NEXTHOP) &&
1090 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1091 		if (ro != NULL)
1092 			dst = (struct sockaddr_in6 *)&ro->ro_dst;
1093 		else
1094 			dst = &sin6;
1095 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1096 		m->m_flags |= M_SKIP_FIREWALL;
1097 		m->m_flags &= ~M_IP6_NEXTHOP;
1098 		m_tag_delete(m, fwd_tag);
1099 		goto again;
1100 	}
1101 
1102 passout:
1103 	if (vlan_pcp > -1)
1104 		EVL_APPLY_PRI(m, vlan_pcp);
1105 
1106 	/* Ensure the packet data is mapped if the interface requires it. */
1107 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1108 		m = mb_unmapped_to_ext(m);
1109 		if (m == NULL) {
1110 			IP6STAT_INC(ip6s_odropped);
1111 			return (ENOBUFS);
1112 		}
1113 	}
1114 
1115 	/*
1116 	 * Send the packet to the outgoing interface.
1117 	 * If necessary, do IPv6 fragmentation before sending.
1118 	 *
1119 	 * The logic here is rather complex:
1120 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1121 	 * 1-a:	send as is if tlen <= path mtu
1122 	 * 1-b:	fragment if tlen > path mtu
1123 	 *
1124 	 * 2: if user asks us not to fragment (dontfrag == 1)
1125 	 * 2-a:	send as is if tlen <= interface mtu
1126 	 * 2-b:	error if tlen > interface mtu
1127 	 *
1128 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1129 	 *	always fragment
1130 	 *
1131 	 * 4: if dontfrag == 1 && alwaysfrag == 1
1132 	 *	error, as we cannot handle this conflicting request.
1133 	 */
1134 	sw_csum = m->m_pkthdr.csum_flags;
1135 	if (!hdrsplit) {
1136 		tso = ((sw_csum & ifp->if_hwassist &
1137 		    (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1138 		sw_csum &= ~ifp->if_hwassist;
1139 	} else
1140 		tso = 0;
1141 	/*
1142 	 * If we added extension headers, we will not do TSO and calculate the
1143 	 * checksums ourselves for now.
1144 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1145 	 * with ext. hdrs.
1146 	 */
1147 	ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1148 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1149 	tlen = m->m_pkthdr.len;
1150 
1151 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1152 		dontfrag = 1;
1153 	else
1154 		dontfrag = 0;
1155 	if (dontfrag && alwaysfrag) {	/* Case 4. */
1156 		/* Conflicting request - can't transmit. */
1157 		error = EMSGSIZE;
1158 		goto bad;
1159 	}
1160 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* Case 2-b. */
1161 		/*
1162 		 * Even if the DONTFRAG option is specified, we cannot send the
1163 		 * packet when the data length is larger than the MTU of the
1164 		 * outgoing interface.
1165 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1166 		 * application wanted to know the MTU value. Also return an
1167 		 * error code (this is not described in the API spec).
1168 		 */
1169 		if (inp != NULL)
1170 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1171 		error = EMSGSIZE;
1172 		goto bad;
1173 	}
1174 
1175 	/* Transmit packet without fragmentation. */
1176 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* Cases 1-a and 2-a. */
1177 		struct in6_ifaddr *ia6;
1178 
1179 		ip6 = mtod(m, struct ip6_hdr *);
1180 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1181 		if (ia6) {
1182 			/* Record statistics for this interface address. */
1183 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1184 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1185 			    m->m_pkthdr.len);
1186 		}
1187 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1188 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
1189 		goto done;
1190 	}
1191 
1192 	/* Try to fragment the packet.  Cases 1-b and 3. */
1193 	if (mtu < IPV6_MMTU) {
1194 		/* Path MTU cannot be less than IPV6_MMTU. */
1195 		error = EMSGSIZE;
1196 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1197 		goto bad;
1198 	} else if (ip6->ip6_plen == 0) {
1199 		/* Jumbo payload cannot be fragmented. */
1200 		error = EMSGSIZE;
1201 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1202 		goto bad;
1203 	} else {
1204 		u_char nextproto;
1205 
1206 		/*
1207 		 * Too large for the destination or interface;
1208 		 * fragment if possible.
1209 		 * Must be able to put at least 8 bytes per fragment.
1210 		 */
1211 		if (mtu > IPV6_MAXPACKET)
1212 			mtu = IPV6_MAXPACKET;
1213 
1214 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1215 		if (len < 8) {
1216 			error = EMSGSIZE;
1217 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1218 			goto bad;
1219 		}
1220 
1221 		/*
1222 		 * If the interface will not calculate checksums on
1223 		 * fragmented packets, then do it here.
1224 		 * XXX-BZ handle the hw offloading case.  Need flags.
1225 		 */
1226 		ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1227 		    optlen);
1228 
1229 		/*
1230 		 * Change the next header field of the last header in the
1231 		 * unfragmentable part.
1232 		 */
1233 		if (exthdrs.ip6e_rthdr) {
1234 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1235 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1236 		} else if (exthdrs.ip6e_dest1) {
1237 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1238 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1239 		} else if (exthdrs.ip6e_hbh) {
1240 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1241 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1242 		} else {
1243 			ip6 = mtod(m, struct ip6_hdr *);
1244 			nextproto = ip6->ip6_nxt;
1245 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1246 		}
1247 
1248 		/*
1249 		 * Loop through length of segment after first fragment,
1250 		 * make new header and copy data of each part and link onto
1251 		 * chain.
1252 		 */
1253 		m0 = m;
1254 		id = htonl(ip6_randomid());
1255 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1256 		if (error != 0)
1257 			goto sendorfree;
1258 
1259 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1260 	}
1261 
1262 	/* Remove leading garbage. */
1263 sendorfree:
1264 	m = m0->m_nextpkt;
1265 	m0->m_nextpkt = 0;
1266 	m_freem(m0);
1267 	for (; m; m = m0) {
1268 		m0 = m->m_nextpkt;
1269 		m->m_nextpkt = 0;
1270 		if (error == 0) {
1271 			/* Record statistics for this interface address. */
1272 			if (ia) {
1273 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1274 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1275 				    m->m_pkthdr.len);
1276 			}
1277 			if (vlan_pcp > -1)
1278 				EVL_APPLY_PRI(m, vlan_pcp);
1279 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1280 			    true);
1281 		} else
1282 			m_freem(m);
1283 	}
1284 
1285 	if (error == 0)
1286 		IP6STAT_INC(ip6s_fragmented);
1287 
1288 done:
1289 	return (error);
1290 
1291 freehdrs:
1292 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1293 	m_freem(exthdrs.ip6e_dest1);
1294 	m_freem(exthdrs.ip6e_rthdr);
1295 	m_freem(exthdrs.ip6e_dest2);
1296 	/* FALLTHROUGH */
1297 bad:
1298 	if (m)
1299 		m_freem(m);
1300 	goto done;
1301 }
1302 
1303 static int
1304 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1305 {
1306 	struct mbuf *m;
1307 
1308 	if (hlen > MCLBYTES)
1309 		return (ENOBUFS); /* XXX */
1310 
1311 	if (hlen > MLEN)
1312 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1313 	else
1314 		m = m_get(M_NOWAIT, MT_DATA);
1315 	if (m == NULL)
1316 		return (ENOBUFS);
1317 	m->m_len = hlen;
1318 	if (hdr)
1319 		bcopy(hdr, mtod(m, caddr_t), hlen);
1320 
1321 	*mp = m;
1322 	return (0);
1323 }
1324 
1325 /*
1326  * Insert jumbo payload option.
1327  */
1328 static int
1329 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1330 {
1331 	struct mbuf *mopt;
1332 	u_char *optbuf;
1333 	u_int32_t v;
1334 
1335 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1336 
1337 	/*
1338 	 * If there is no hop-by-hop options header, allocate new one.
1339 	 * If there is one but it doesn't have enough space to store the
1340 	 * jumbo payload option, allocate a cluster to store the whole options.
1341 	 * Otherwise, use it to store the options.
1342 	 */
1343 	if (exthdrs->ip6e_hbh == NULL) {
1344 		mopt = m_get(M_NOWAIT, MT_DATA);
1345 		if (mopt == NULL)
1346 			return (ENOBUFS);
1347 		mopt->m_len = JUMBOOPTLEN;
1348 		optbuf = mtod(mopt, u_char *);
1349 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1350 		exthdrs->ip6e_hbh = mopt;
1351 	} else {
1352 		struct ip6_hbh *hbh;
1353 
1354 		mopt = exthdrs->ip6e_hbh;
1355 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1356 			/*
1357 			 * XXX assumption:
1358 			 * - exthdrs->ip6e_hbh is not referenced from places
1359 			 *   other than exthdrs.
1360 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1361 			 */
1362 			int oldoptlen = mopt->m_len;
1363 			struct mbuf *n;
1364 
1365 			/*
1366 			 * XXX: give up if the whole (new) hbh header does
1367 			 * not fit even in an mbuf cluster.
1368 			 */
1369 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1370 				return (ENOBUFS);
1371 
1372 			/*
1373 			 * As a consequence, we must always prepare a cluster
1374 			 * at this point.
1375 			 */
1376 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1377 			if (n == NULL)
1378 				return (ENOBUFS);
1379 			n->m_len = oldoptlen + JUMBOOPTLEN;
1380 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1381 			    oldoptlen);
1382 			optbuf = mtod(n, caddr_t) + oldoptlen;
1383 			m_freem(mopt);
1384 			mopt = exthdrs->ip6e_hbh = n;
1385 		} else {
1386 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1387 			mopt->m_len += JUMBOOPTLEN;
1388 		}
1389 		optbuf[0] = IP6OPT_PADN;
1390 		optbuf[1] = 1;
1391 
1392 		/*
1393 		 * Adjust the header length according to the pad and
1394 		 * the jumbo payload option.
1395 		 */
1396 		hbh = mtod(mopt, struct ip6_hbh *);
1397 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1398 	}
1399 
1400 	/* fill in the option. */
1401 	optbuf[2] = IP6OPT_JUMBO;
1402 	optbuf[3] = 4;
1403 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1404 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1405 
1406 	/* finally, adjust the packet header length */
1407 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1408 
1409 	return (0);
1410 #undef JUMBOOPTLEN
1411 }
1412 
1413 /*
1414  * Insert fragment header and copy unfragmentable header portions.
1415  */
1416 static int
1417 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1418     struct ip6_frag **frghdrp)
1419 {
1420 	struct mbuf *n, *mlast;
1421 
1422 	if (hlen > sizeof(struct ip6_hdr)) {
1423 		n = m_copym(m0, sizeof(struct ip6_hdr),
1424 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1425 		if (n == NULL)
1426 			return (ENOBUFS);
1427 		m->m_next = n;
1428 	} else
1429 		n = m;
1430 
1431 	/* Search for the last mbuf of unfragmentable part. */
1432 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1433 		;
1434 
1435 	if (M_WRITABLE(mlast) &&
1436 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1437 		/* use the trailing space of the last mbuf for the fragment hdr */
1438 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1439 		    mlast->m_len);
1440 		mlast->m_len += sizeof(struct ip6_frag);
1441 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1442 	} else {
1443 		/* allocate a new mbuf for the fragment header */
1444 		struct mbuf *mfrg;
1445 
1446 		mfrg = m_get(M_NOWAIT, MT_DATA);
1447 		if (mfrg == NULL)
1448 			return (ENOBUFS);
1449 		mfrg->m_len = sizeof(struct ip6_frag);
1450 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1451 		mlast->m_next = mfrg;
1452 	}
1453 
1454 	return (0);
1455 }
1456 
1457 /*
1458  * Calculates IPv6 path mtu for destination @dst.
1459  * Resulting MTU is stored in @mtup.
1460  *
1461  * Returns 0 on success.
1462  */
1463 static int
1464 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1465 {
1466 	struct epoch_tracker et;
1467 	struct nhop_object *nh;
1468 	struct in6_addr kdst;
1469 	uint32_t scopeid;
1470 	int error;
1471 
1472 	in6_splitscope(dst, &kdst, &scopeid);
1473 
1474 	NET_EPOCH_ENTER(et);
1475 	nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1476 	if (nh != NULL)
1477 		error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1478 	else
1479 		error = EHOSTUNREACH;
1480 	NET_EPOCH_EXIT(et);
1481 
1482 	return (error);
1483 }
1484 
1485 /*
1486  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1487  * and cached data in @ro_pmtu.
1488  * MTU from (successful) route lookup is saved (along with dst)
1489  * inside @ro_pmtu to avoid subsequent route lookups after packet
1490  * filter processing.
1491  *
1492  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1493  * Returns 0 on success.
1494  */
1495 static int
1496 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1497     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1498     int *alwaysfragp, u_int fibnum, u_int proto)
1499 {
1500 	struct nhop_object *nh;
1501 	struct in6_addr kdst;
1502 	uint32_t scopeid;
1503 	struct sockaddr_in6 *sa6_dst, sin6;
1504 	u_long mtu;
1505 
1506 	NET_EPOCH_ASSERT();
1507 
1508 	mtu = 0;
1509 	if (ro_pmtu == NULL || do_lookup) {
1510 		/*
1511 		 * Here ro_pmtu has final destination address, while
1512 		 * ro might represent immediate destination.
1513 		 * Use ro_pmtu destination since mtu might differ.
1514 		 */
1515 		if (ro_pmtu != NULL) {
1516 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1517 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1518 				ro_pmtu->ro_mtu = 0;
1519 		} else
1520 			sa6_dst = &sin6;
1521 
1522 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1523 			bzero(sa6_dst, sizeof(*sa6_dst));
1524 			sa6_dst->sin6_family = AF_INET6;
1525 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1526 			sa6_dst->sin6_addr = *dst;
1527 
1528 			in6_splitscope(dst, &kdst, &scopeid);
1529 			nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1530 			if (nh != NULL) {
1531 				mtu = nh->nh_mtu;
1532 				if (ro_pmtu != NULL)
1533 					ro_pmtu->ro_mtu = mtu;
1534 			}
1535 		} else
1536 			mtu = ro_pmtu->ro_mtu;
1537 	}
1538 
1539 	if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1540 		mtu = ro_pmtu->ro_nh->nh_mtu;
1541 
1542 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1543 }
1544 
1545 /*
1546  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1547  * hostcache data for @dst.
1548  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1549  *
1550  * Returns 0 on success.
1551  */
1552 static int
1553 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1554     u_long *mtup, int *alwaysfragp, u_int proto)
1555 {
1556 	u_long mtu = 0;
1557 	int alwaysfrag = 0;
1558 	int error = 0;
1559 
1560 	if (rt_mtu > 0) {
1561 		u_int32_t ifmtu;
1562 		struct in_conninfo inc;
1563 
1564 		bzero(&inc, sizeof(inc));
1565 		inc.inc_flags |= INC_ISIPV6;
1566 		inc.inc6_faddr = *dst;
1567 
1568 		ifmtu = IN6_LINKMTU(ifp);
1569 
1570 		/* TCP is known to react to pmtu changes so skip hc */
1571 		if (proto != IPPROTO_TCP)
1572 			mtu = tcp_hc_getmtu(&inc);
1573 
1574 		if (mtu)
1575 			mtu = min(mtu, rt_mtu);
1576 		else
1577 			mtu = rt_mtu;
1578 		if (mtu == 0)
1579 			mtu = ifmtu;
1580 		else if (mtu < IPV6_MMTU) {
1581 			/*
1582 			 * RFC2460 section 5, last paragraph:
1583 			 * if we record ICMPv6 too big message with
1584 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1585 			 * or smaller, with framgent header attached.
1586 			 * (fragment header is needed regardless from the
1587 			 * packet size, for translators to identify packets)
1588 			 */
1589 			alwaysfrag = 1;
1590 			mtu = IPV6_MMTU;
1591 		}
1592 	} else if (ifp) {
1593 		mtu = IN6_LINKMTU(ifp);
1594 	} else
1595 		error = EHOSTUNREACH; /* XXX */
1596 
1597 	*mtup = mtu;
1598 	if (alwaysfragp)
1599 		*alwaysfragp = alwaysfrag;
1600 	return (error);
1601 }
1602 
1603 /*
1604  * IP6 socket option processing.
1605  */
1606 int
1607 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1608 {
1609 	int optdatalen, uproto;
1610 	void *optdata;
1611 	struct inpcb *inp = sotoinpcb(so);
1612 	int error, optval;
1613 	int level, op, optname;
1614 	int optlen;
1615 	struct thread *td;
1616 #ifdef	RSS
1617 	uint32_t rss_bucket;
1618 	int retval;
1619 #endif
1620 
1621 /*
1622  * Don't use more than a quarter of mbuf clusters.  N.B.:
1623  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1624  * on LP64 architectures, so cast to u_long to avoid undefined
1625  * behavior.  ILP32 architectures cannot have nmbclusters
1626  * large enough to overflow for other reasons.
1627  */
1628 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1629 
1630 	level = sopt->sopt_level;
1631 	op = sopt->sopt_dir;
1632 	optname = sopt->sopt_name;
1633 	optlen = sopt->sopt_valsize;
1634 	td = sopt->sopt_td;
1635 	error = 0;
1636 	optval = 0;
1637 	uproto = (int)so->so_proto->pr_protocol;
1638 
1639 	if (level != IPPROTO_IPV6) {
1640 		error = EINVAL;
1641 
1642 		if (sopt->sopt_level == SOL_SOCKET &&
1643 		    sopt->sopt_dir == SOPT_SET) {
1644 			switch (sopt->sopt_name) {
1645 			case SO_SETFIB:
1646 				INP_WLOCK(inp);
1647 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1648 				INP_WUNLOCK(inp);
1649 				error = 0;
1650 				break;
1651 			case SO_MAX_PACING_RATE:
1652 #ifdef RATELIMIT
1653 				INP_WLOCK(inp);
1654 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1655 				INP_WUNLOCK(inp);
1656 				error = 0;
1657 #else
1658 				error = EOPNOTSUPP;
1659 #endif
1660 				break;
1661 			default:
1662 				break;
1663 			}
1664 		}
1665 	} else {		/* level == IPPROTO_IPV6 */
1666 		switch (op) {
1667 		case SOPT_SET:
1668 			switch (optname) {
1669 			case IPV6_2292PKTOPTIONS:
1670 #ifdef IPV6_PKTOPTIONS
1671 			case IPV6_PKTOPTIONS:
1672 #endif
1673 			{
1674 				struct mbuf *m;
1675 
1676 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1677 					printf("ip6_ctloutput: mbuf limit hit\n");
1678 					error = ENOBUFS;
1679 					break;
1680 				}
1681 
1682 				error = soopt_getm(sopt, &m); /* XXX */
1683 				if (error != 0)
1684 					break;
1685 				error = soopt_mcopyin(sopt, m); /* XXX */
1686 				if (error != 0)
1687 					break;
1688 				INP_WLOCK(inp);
1689 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1690 				    so, sopt);
1691 				INP_WUNLOCK(inp);
1692 				m_freem(m); /* XXX */
1693 				break;
1694 			}
1695 
1696 			/*
1697 			 * Use of some Hop-by-Hop options or some
1698 			 * Destination options, might require special
1699 			 * privilege.  That is, normal applications
1700 			 * (without special privilege) might be forbidden
1701 			 * from setting certain options in outgoing packets,
1702 			 * and might never see certain options in received
1703 			 * packets. [RFC 2292 Section 6]
1704 			 * KAME specific note:
1705 			 *  KAME prevents non-privileged users from sending or
1706 			 *  receiving ANY hbh/dst options in order to avoid
1707 			 *  overhead of parsing options in the kernel.
1708 			 */
1709 			case IPV6_RECVHOPOPTS:
1710 			case IPV6_RECVDSTOPTS:
1711 			case IPV6_RECVRTHDRDSTOPTS:
1712 				if (td != NULL) {
1713 					error = priv_check(td,
1714 					    PRIV_NETINET_SETHDROPTS);
1715 					if (error)
1716 						break;
1717 				}
1718 				/* FALLTHROUGH */
1719 			case IPV6_UNICAST_HOPS:
1720 			case IPV6_HOPLIMIT:
1721 
1722 			case IPV6_RECVPKTINFO:
1723 			case IPV6_RECVHOPLIMIT:
1724 			case IPV6_RECVRTHDR:
1725 			case IPV6_RECVPATHMTU:
1726 			case IPV6_RECVTCLASS:
1727 			case IPV6_RECVFLOWID:
1728 #ifdef	RSS
1729 			case IPV6_RECVRSSBUCKETID:
1730 #endif
1731 			case IPV6_V6ONLY:
1732 			case IPV6_AUTOFLOWLABEL:
1733 			case IPV6_ORIGDSTADDR:
1734 			case IPV6_BINDANY:
1735 			case IPV6_VLAN_PCP:
1736 				if (optname == IPV6_BINDANY && td != NULL) {
1737 					error = priv_check(td,
1738 					    PRIV_NETINET_BINDANY);
1739 					if (error)
1740 						break;
1741 				}
1742 
1743 				if (optlen != sizeof(int)) {
1744 					error = EINVAL;
1745 					break;
1746 				}
1747 				error = sooptcopyin(sopt, &optval,
1748 					sizeof optval, sizeof optval);
1749 				if (error)
1750 					break;
1751 				switch (optname) {
1752 				case IPV6_UNICAST_HOPS:
1753 					if (optval < -1 || optval >= 256)
1754 						error = EINVAL;
1755 					else {
1756 						/* -1 = kernel default */
1757 						inp->in6p_hops = optval;
1758 						if ((inp->inp_vflag &
1759 						     INP_IPV4) != 0)
1760 							inp->inp_ip_ttl = optval;
1761 					}
1762 					break;
1763 #define OPTSET(bit) \
1764 do { \
1765 	INP_WLOCK(inp); \
1766 	if (optval) \
1767 		inp->inp_flags |= (bit); \
1768 	else \
1769 		inp->inp_flags &= ~(bit); \
1770 	INP_WUNLOCK(inp); \
1771 } while (/*CONSTCOND*/ 0)
1772 #define OPTSET2292(bit) \
1773 do { \
1774 	INP_WLOCK(inp); \
1775 	inp->inp_flags |= IN6P_RFC2292; \
1776 	if (optval) \
1777 		inp->inp_flags |= (bit); \
1778 	else \
1779 		inp->inp_flags &= ~(bit); \
1780 	INP_WUNLOCK(inp); \
1781 } while (/*CONSTCOND*/ 0)
1782 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1783 
1784 #define OPTSET2_N(bit, val) do {					\
1785 	if (val)							\
1786 		inp->inp_flags2 |= bit;					\
1787 	else								\
1788 		inp->inp_flags2 &= ~bit;				\
1789 } while (0)
1790 #define OPTSET2(bit, val) do {						\
1791 	INP_WLOCK(inp);							\
1792 	OPTSET2_N(bit, val);						\
1793 	INP_WUNLOCK(inp);						\
1794 } while (0)
1795 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1796 #define OPTSET2292_EXCLUSIVE(bit)					\
1797 do {									\
1798 	INP_WLOCK(inp);							\
1799 	if (OPTBIT(IN6P_RFC2292)) {					\
1800 		error = EINVAL;						\
1801 	} else {							\
1802 		if (optval)						\
1803 			inp->inp_flags |= (bit);			\
1804 		else							\
1805 			inp->inp_flags &= ~(bit);			\
1806 	}								\
1807 	INP_WUNLOCK(inp);						\
1808 } while (/*CONSTCOND*/ 0)
1809 
1810 				case IPV6_RECVPKTINFO:
1811 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1812 					break;
1813 
1814 				case IPV6_HOPLIMIT:
1815 				{
1816 					struct ip6_pktopts **optp;
1817 
1818 					/* cannot mix with RFC2292 */
1819 					if (OPTBIT(IN6P_RFC2292)) {
1820 						error = EINVAL;
1821 						break;
1822 					}
1823 					INP_WLOCK(inp);
1824 					if (inp->inp_flags & INP_DROPPED) {
1825 						INP_WUNLOCK(inp);
1826 						return (ECONNRESET);
1827 					}
1828 					optp = &inp->in6p_outputopts;
1829 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1830 					    (u_char *)&optval, sizeof(optval),
1831 					    optp, (td != NULL) ? td->td_ucred :
1832 					    NULL, uproto);
1833 					INP_WUNLOCK(inp);
1834 					break;
1835 				}
1836 
1837 				case IPV6_RECVHOPLIMIT:
1838 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1839 					break;
1840 
1841 				case IPV6_RECVHOPOPTS:
1842 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1843 					break;
1844 
1845 				case IPV6_RECVDSTOPTS:
1846 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1847 					break;
1848 
1849 				case IPV6_RECVRTHDRDSTOPTS:
1850 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1851 					break;
1852 
1853 				case IPV6_RECVRTHDR:
1854 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1855 					break;
1856 
1857 				case IPV6_RECVPATHMTU:
1858 					/*
1859 					 * We ignore this option for TCP
1860 					 * sockets.
1861 					 * (RFC3542 leaves this case
1862 					 * unspecified.)
1863 					 */
1864 					if (uproto != IPPROTO_TCP)
1865 						OPTSET(IN6P_MTU);
1866 					break;
1867 
1868 				case IPV6_RECVFLOWID:
1869 					OPTSET2(INP_RECVFLOWID, optval);
1870 					break;
1871 
1872 #ifdef	RSS
1873 				case IPV6_RECVRSSBUCKETID:
1874 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1875 					break;
1876 #endif
1877 
1878 				case IPV6_V6ONLY:
1879 					INP_WLOCK(inp);
1880 					if (inp->inp_lport ||
1881 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1882 						/*
1883 						 * The socket is already bound.
1884 						 */
1885 						INP_WUNLOCK(inp);
1886 						error = EINVAL;
1887 						break;
1888 					}
1889 					if (optval) {
1890 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1891 						inp->inp_vflag &= ~INP_IPV4;
1892 					} else {
1893 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1894 						inp->inp_vflag |= INP_IPV4;
1895 					}
1896 					INP_WUNLOCK(inp);
1897 					break;
1898 				case IPV6_RECVTCLASS:
1899 					/* cannot mix with RFC2292 XXX */
1900 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1901 					break;
1902 				case IPV6_AUTOFLOWLABEL:
1903 					OPTSET(IN6P_AUTOFLOWLABEL);
1904 					break;
1905 
1906 				case IPV6_ORIGDSTADDR:
1907 					OPTSET2(INP_ORIGDSTADDR, optval);
1908 					break;
1909 				case IPV6_BINDANY:
1910 					OPTSET(INP_BINDANY);
1911 					break;
1912 				case IPV6_VLAN_PCP:
1913 					if ((optval >= -1) && (optval <=
1914 					    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1915 						if (optval == -1) {
1916 							INP_WLOCK(inp);
1917 							inp->inp_flags2 &=
1918 							    ~(INP_2PCP_SET |
1919 							    INP_2PCP_MASK);
1920 							INP_WUNLOCK(inp);
1921 						} else {
1922 							INP_WLOCK(inp);
1923 							inp->inp_flags2 |=
1924 							    INP_2PCP_SET;
1925 							inp->inp_flags2 &=
1926 							    ~INP_2PCP_MASK;
1927 							inp->inp_flags2 |=
1928 							    optval <<
1929 							    INP_2PCP_SHIFT;
1930 							INP_WUNLOCK(inp);
1931 						}
1932 					} else
1933 						error = EINVAL;
1934 					break;
1935 				}
1936 				break;
1937 
1938 			case IPV6_TCLASS:
1939 			case IPV6_DONTFRAG:
1940 			case IPV6_USE_MIN_MTU:
1941 			case IPV6_PREFER_TEMPADDR:
1942 				if (optlen != sizeof(optval)) {
1943 					error = EINVAL;
1944 					break;
1945 				}
1946 				error = sooptcopyin(sopt, &optval,
1947 					sizeof optval, sizeof optval);
1948 				if (error)
1949 					break;
1950 				{
1951 					struct ip6_pktopts **optp;
1952 					INP_WLOCK(inp);
1953 					if (inp->inp_flags & INP_DROPPED) {
1954 						INP_WUNLOCK(inp);
1955 						return (ECONNRESET);
1956 					}
1957 					optp = &inp->in6p_outputopts;
1958 					error = ip6_pcbopt(optname,
1959 					    (u_char *)&optval, sizeof(optval),
1960 					    optp, (td != NULL) ? td->td_ucred :
1961 					    NULL, uproto);
1962 					INP_WUNLOCK(inp);
1963 					break;
1964 				}
1965 
1966 			case IPV6_2292PKTINFO:
1967 			case IPV6_2292HOPLIMIT:
1968 			case IPV6_2292HOPOPTS:
1969 			case IPV6_2292DSTOPTS:
1970 			case IPV6_2292RTHDR:
1971 				/* RFC 2292 */
1972 				if (optlen != sizeof(int)) {
1973 					error = EINVAL;
1974 					break;
1975 				}
1976 				error = sooptcopyin(sopt, &optval,
1977 					sizeof optval, sizeof optval);
1978 				if (error)
1979 					break;
1980 				switch (optname) {
1981 				case IPV6_2292PKTINFO:
1982 					OPTSET2292(IN6P_PKTINFO);
1983 					break;
1984 				case IPV6_2292HOPLIMIT:
1985 					OPTSET2292(IN6P_HOPLIMIT);
1986 					break;
1987 				case IPV6_2292HOPOPTS:
1988 					/*
1989 					 * Check super-user privilege.
1990 					 * See comments for IPV6_RECVHOPOPTS.
1991 					 */
1992 					if (td != NULL) {
1993 						error = priv_check(td,
1994 						    PRIV_NETINET_SETHDROPTS);
1995 						if (error)
1996 							return (error);
1997 					}
1998 					OPTSET2292(IN6P_HOPOPTS);
1999 					break;
2000 				case IPV6_2292DSTOPTS:
2001 					if (td != NULL) {
2002 						error = priv_check(td,
2003 						    PRIV_NETINET_SETHDROPTS);
2004 						if (error)
2005 							return (error);
2006 					}
2007 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2008 					break;
2009 				case IPV6_2292RTHDR:
2010 					OPTSET2292(IN6P_RTHDR);
2011 					break;
2012 				}
2013 				break;
2014 			case IPV6_PKTINFO:
2015 			case IPV6_HOPOPTS:
2016 			case IPV6_RTHDR:
2017 			case IPV6_DSTOPTS:
2018 			case IPV6_RTHDRDSTOPTS:
2019 			case IPV6_NEXTHOP:
2020 			{
2021 				/* new advanced API (RFC3542) */
2022 				u_char *optbuf;
2023 				u_char optbuf_storage[MCLBYTES];
2024 				int optlen;
2025 				struct ip6_pktopts **optp;
2026 
2027 				/* cannot mix with RFC2292 */
2028 				if (OPTBIT(IN6P_RFC2292)) {
2029 					error = EINVAL;
2030 					break;
2031 				}
2032 
2033 				/*
2034 				 * We only ensure valsize is not too large
2035 				 * here.  Further validation will be done
2036 				 * later.
2037 				 */
2038 				error = sooptcopyin(sopt, optbuf_storage,
2039 				    sizeof(optbuf_storage), 0);
2040 				if (error)
2041 					break;
2042 				optlen = sopt->sopt_valsize;
2043 				optbuf = optbuf_storage;
2044 				INP_WLOCK(inp);
2045 				if (inp->inp_flags & INP_DROPPED) {
2046 					INP_WUNLOCK(inp);
2047 					return (ECONNRESET);
2048 				}
2049 				optp = &inp->in6p_outputopts;
2050 				error = ip6_pcbopt(optname, optbuf, optlen,
2051 				    optp, (td != NULL) ? td->td_ucred : NULL,
2052 				    uproto);
2053 				INP_WUNLOCK(inp);
2054 				break;
2055 			}
2056 #undef OPTSET
2057 
2058 			case IPV6_MULTICAST_IF:
2059 			case IPV6_MULTICAST_HOPS:
2060 			case IPV6_MULTICAST_LOOP:
2061 			case IPV6_JOIN_GROUP:
2062 			case IPV6_LEAVE_GROUP:
2063 			case IPV6_MSFILTER:
2064 			case MCAST_BLOCK_SOURCE:
2065 			case MCAST_UNBLOCK_SOURCE:
2066 			case MCAST_JOIN_GROUP:
2067 			case MCAST_LEAVE_GROUP:
2068 			case MCAST_JOIN_SOURCE_GROUP:
2069 			case MCAST_LEAVE_SOURCE_GROUP:
2070 				error = ip6_setmoptions(inp, sopt);
2071 				break;
2072 
2073 			case IPV6_PORTRANGE:
2074 				error = sooptcopyin(sopt, &optval,
2075 				    sizeof optval, sizeof optval);
2076 				if (error)
2077 					break;
2078 
2079 				INP_WLOCK(inp);
2080 				switch (optval) {
2081 				case IPV6_PORTRANGE_DEFAULT:
2082 					inp->inp_flags &= ~(INP_LOWPORT);
2083 					inp->inp_flags &= ~(INP_HIGHPORT);
2084 					break;
2085 
2086 				case IPV6_PORTRANGE_HIGH:
2087 					inp->inp_flags &= ~(INP_LOWPORT);
2088 					inp->inp_flags |= INP_HIGHPORT;
2089 					break;
2090 
2091 				case IPV6_PORTRANGE_LOW:
2092 					inp->inp_flags &= ~(INP_HIGHPORT);
2093 					inp->inp_flags |= INP_LOWPORT;
2094 					break;
2095 
2096 				default:
2097 					error = EINVAL;
2098 					break;
2099 				}
2100 				INP_WUNLOCK(inp);
2101 				break;
2102 
2103 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2104 			case IPV6_IPSEC_POLICY:
2105 				if (IPSEC_ENABLED(ipv6)) {
2106 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2107 					break;
2108 				}
2109 				/* FALLTHROUGH */
2110 #endif /* IPSEC */
2111 
2112 			default:
2113 				error = ENOPROTOOPT;
2114 				break;
2115 			}
2116 			break;
2117 
2118 		case SOPT_GET:
2119 			switch (optname) {
2120 			case IPV6_2292PKTOPTIONS:
2121 #ifdef IPV6_PKTOPTIONS
2122 			case IPV6_PKTOPTIONS:
2123 #endif
2124 				/*
2125 				 * RFC3542 (effectively) deprecated the
2126 				 * semantics of the 2292-style pktoptions.
2127 				 * Since it was not reliable in nature (i.e.,
2128 				 * applications had to expect the lack of some
2129 				 * information after all), it would make sense
2130 				 * to simplify this part by always returning
2131 				 * empty data.
2132 				 */
2133 				sopt->sopt_valsize = 0;
2134 				break;
2135 
2136 			case IPV6_RECVHOPOPTS:
2137 			case IPV6_RECVDSTOPTS:
2138 			case IPV6_RECVRTHDRDSTOPTS:
2139 			case IPV6_UNICAST_HOPS:
2140 			case IPV6_RECVPKTINFO:
2141 			case IPV6_RECVHOPLIMIT:
2142 			case IPV6_RECVRTHDR:
2143 			case IPV6_RECVPATHMTU:
2144 
2145 			case IPV6_V6ONLY:
2146 			case IPV6_PORTRANGE:
2147 			case IPV6_RECVTCLASS:
2148 			case IPV6_AUTOFLOWLABEL:
2149 			case IPV6_BINDANY:
2150 			case IPV6_FLOWID:
2151 			case IPV6_FLOWTYPE:
2152 			case IPV6_RECVFLOWID:
2153 #ifdef	RSS
2154 			case IPV6_RSSBUCKETID:
2155 			case IPV6_RECVRSSBUCKETID:
2156 #endif
2157 			case IPV6_VLAN_PCP:
2158 				switch (optname) {
2159 				case IPV6_RECVHOPOPTS:
2160 					optval = OPTBIT(IN6P_HOPOPTS);
2161 					break;
2162 
2163 				case IPV6_RECVDSTOPTS:
2164 					optval = OPTBIT(IN6P_DSTOPTS);
2165 					break;
2166 
2167 				case IPV6_RECVRTHDRDSTOPTS:
2168 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2169 					break;
2170 
2171 				case IPV6_UNICAST_HOPS:
2172 					optval = inp->in6p_hops;
2173 					break;
2174 
2175 				case IPV6_RECVPKTINFO:
2176 					optval = OPTBIT(IN6P_PKTINFO);
2177 					break;
2178 
2179 				case IPV6_RECVHOPLIMIT:
2180 					optval = OPTBIT(IN6P_HOPLIMIT);
2181 					break;
2182 
2183 				case IPV6_RECVRTHDR:
2184 					optval = OPTBIT(IN6P_RTHDR);
2185 					break;
2186 
2187 				case IPV6_RECVPATHMTU:
2188 					optval = OPTBIT(IN6P_MTU);
2189 					break;
2190 
2191 				case IPV6_V6ONLY:
2192 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2193 					break;
2194 
2195 				case IPV6_PORTRANGE:
2196 				    {
2197 					int flags;
2198 					flags = inp->inp_flags;
2199 					if (flags & INP_HIGHPORT)
2200 						optval = IPV6_PORTRANGE_HIGH;
2201 					else if (flags & INP_LOWPORT)
2202 						optval = IPV6_PORTRANGE_LOW;
2203 					else
2204 						optval = 0;
2205 					break;
2206 				    }
2207 				case IPV6_RECVTCLASS:
2208 					optval = OPTBIT(IN6P_TCLASS);
2209 					break;
2210 
2211 				case IPV6_AUTOFLOWLABEL:
2212 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2213 					break;
2214 
2215 				case IPV6_ORIGDSTADDR:
2216 					optval = OPTBIT2(INP_ORIGDSTADDR);
2217 					break;
2218 
2219 				case IPV6_BINDANY:
2220 					optval = OPTBIT(INP_BINDANY);
2221 					break;
2222 
2223 				case IPV6_FLOWID:
2224 					optval = inp->inp_flowid;
2225 					break;
2226 
2227 				case IPV6_FLOWTYPE:
2228 					optval = inp->inp_flowtype;
2229 					break;
2230 
2231 				case IPV6_RECVFLOWID:
2232 					optval = OPTBIT2(INP_RECVFLOWID);
2233 					break;
2234 #ifdef	RSS
2235 				case IPV6_RSSBUCKETID:
2236 					retval =
2237 					    rss_hash2bucket(inp->inp_flowid,
2238 					    inp->inp_flowtype,
2239 					    &rss_bucket);
2240 					if (retval == 0)
2241 						optval = rss_bucket;
2242 					else
2243 						error = EINVAL;
2244 					break;
2245 
2246 				case IPV6_RECVRSSBUCKETID:
2247 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2248 					break;
2249 #endif
2250 
2251 
2252 				case IPV6_VLAN_PCP:
2253 					if (OPTBIT2(INP_2PCP_SET)) {
2254 						optval = (inp->inp_flags2 &
2255 							    INP_2PCP_MASK) >>
2256 							    INP_2PCP_SHIFT;
2257 					} else {
2258 						optval = -1;
2259 					}
2260 					break;
2261 				}
2262 
2263 				if (error)
2264 					break;
2265 				error = sooptcopyout(sopt, &optval,
2266 					sizeof optval);
2267 				break;
2268 
2269 			case IPV6_PATHMTU:
2270 			{
2271 				u_long pmtu = 0;
2272 				struct ip6_mtuinfo mtuinfo;
2273 				struct in6_addr addr;
2274 
2275 				if (!(so->so_state & SS_ISCONNECTED))
2276 					return (ENOTCONN);
2277 				/*
2278 				 * XXX: we dot not consider the case of source
2279 				 * routing, or optional information to specify
2280 				 * the outgoing interface.
2281 				 * Copy faddr out of inp to avoid holding lock
2282 				 * on inp during route lookup.
2283 				 */
2284 				INP_RLOCK(inp);
2285 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2286 				INP_RUNLOCK(inp);
2287 				error = ip6_getpmtu_ctl(so->so_fibnum,
2288 				    &addr, &pmtu);
2289 				if (error)
2290 					break;
2291 				if (pmtu > IPV6_MAXPACKET)
2292 					pmtu = IPV6_MAXPACKET;
2293 
2294 				bzero(&mtuinfo, sizeof(mtuinfo));
2295 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2296 				optdata = (void *)&mtuinfo;
2297 				optdatalen = sizeof(mtuinfo);
2298 				error = sooptcopyout(sopt, optdata,
2299 				    optdatalen);
2300 				break;
2301 			}
2302 
2303 			case IPV6_2292PKTINFO:
2304 			case IPV6_2292HOPLIMIT:
2305 			case IPV6_2292HOPOPTS:
2306 			case IPV6_2292RTHDR:
2307 			case IPV6_2292DSTOPTS:
2308 				switch (optname) {
2309 				case IPV6_2292PKTINFO:
2310 					optval = OPTBIT(IN6P_PKTINFO);
2311 					break;
2312 				case IPV6_2292HOPLIMIT:
2313 					optval = OPTBIT(IN6P_HOPLIMIT);
2314 					break;
2315 				case IPV6_2292HOPOPTS:
2316 					optval = OPTBIT(IN6P_HOPOPTS);
2317 					break;
2318 				case IPV6_2292RTHDR:
2319 					optval = OPTBIT(IN6P_RTHDR);
2320 					break;
2321 				case IPV6_2292DSTOPTS:
2322 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2323 					break;
2324 				}
2325 				error = sooptcopyout(sopt, &optval,
2326 				    sizeof optval);
2327 				break;
2328 			case IPV6_PKTINFO:
2329 			case IPV6_HOPOPTS:
2330 			case IPV6_RTHDR:
2331 			case IPV6_DSTOPTS:
2332 			case IPV6_RTHDRDSTOPTS:
2333 			case IPV6_NEXTHOP:
2334 			case IPV6_TCLASS:
2335 			case IPV6_DONTFRAG:
2336 			case IPV6_USE_MIN_MTU:
2337 			case IPV6_PREFER_TEMPADDR:
2338 				error = ip6_getpcbopt(inp, optname, sopt);
2339 				break;
2340 
2341 			case IPV6_MULTICAST_IF:
2342 			case IPV6_MULTICAST_HOPS:
2343 			case IPV6_MULTICAST_LOOP:
2344 			case IPV6_MSFILTER:
2345 				error = ip6_getmoptions(inp, sopt);
2346 				break;
2347 
2348 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2349 			case IPV6_IPSEC_POLICY:
2350 				if (IPSEC_ENABLED(ipv6)) {
2351 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2352 					break;
2353 				}
2354 				/* FALLTHROUGH */
2355 #endif /* IPSEC */
2356 			default:
2357 				error = ENOPROTOOPT;
2358 				break;
2359 			}
2360 			break;
2361 		}
2362 	}
2363 	return (error);
2364 }
2365 
2366 int
2367 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2368 {
2369 	int error = 0, optval, optlen;
2370 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2371 	struct inpcb *inp = sotoinpcb(so);
2372 	int level, op, optname;
2373 
2374 	level = sopt->sopt_level;
2375 	op = sopt->sopt_dir;
2376 	optname = sopt->sopt_name;
2377 	optlen = sopt->sopt_valsize;
2378 
2379 	if (level != IPPROTO_IPV6) {
2380 		return (EINVAL);
2381 	}
2382 
2383 	switch (optname) {
2384 	case IPV6_CHECKSUM:
2385 		/*
2386 		 * For ICMPv6 sockets, no modification allowed for checksum
2387 		 * offset, permit "no change" values to help existing apps.
2388 		 *
2389 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2390 		 * for an ICMPv6 socket will fail."
2391 		 * The current behavior does not meet RFC3542.
2392 		 */
2393 		switch (op) {
2394 		case SOPT_SET:
2395 			if (optlen != sizeof(int)) {
2396 				error = EINVAL;
2397 				break;
2398 			}
2399 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2400 					    sizeof(optval));
2401 			if (error)
2402 				break;
2403 			if (optval < -1 || (optval % 2) != 0) {
2404 				/*
2405 				 * The API assumes non-negative even offset
2406 				 * values or -1 as a special value.
2407 				 */
2408 				error = EINVAL;
2409 			} else if (inp->inp_ip_p == IPPROTO_ICMPV6) {
2410 				if (optval != icmp6off)
2411 					error = EINVAL;
2412 			} else
2413 				inp->in6p_cksum = optval;
2414 			break;
2415 
2416 		case SOPT_GET:
2417 			if (inp->inp_ip_p == IPPROTO_ICMPV6)
2418 				optval = icmp6off;
2419 			else
2420 				optval = inp->in6p_cksum;
2421 
2422 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2423 			break;
2424 
2425 		default:
2426 			error = EINVAL;
2427 			break;
2428 		}
2429 		break;
2430 
2431 	default:
2432 		error = ENOPROTOOPT;
2433 		break;
2434 	}
2435 
2436 	return (error);
2437 }
2438 
2439 /*
2440  * Set up IP6 options in pcb for insertion in output packets or
2441  * specifying behavior of outgoing packets.
2442  */
2443 static int
2444 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2445     struct socket *so, struct sockopt *sopt)
2446 {
2447 	struct ip6_pktopts *opt = *pktopt;
2448 	int error = 0;
2449 	struct thread *td = sopt->sopt_td;
2450 	struct epoch_tracker et;
2451 
2452 	/* turn off any old options. */
2453 	if (opt) {
2454 #ifdef DIAGNOSTIC
2455 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2456 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2457 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2458 			printf("ip6_pcbopts: all specified options are cleared.\n");
2459 #endif
2460 		ip6_clearpktopts(opt, -1);
2461 	} else {
2462 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2463 		if (opt == NULL)
2464 			return (ENOMEM);
2465 	}
2466 	*pktopt = NULL;
2467 
2468 	if (!m || m->m_len == 0) {
2469 		/*
2470 		 * Only turning off any previous options, regardless of
2471 		 * whether the opt is just created or given.
2472 		 */
2473 		free(opt, M_IP6OPT);
2474 		return (0);
2475 	}
2476 
2477 	/*  set options specified by user. */
2478 	NET_EPOCH_ENTER(et);
2479 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2480 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2481 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2482 		free(opt, M_IP6OPT);
2483 		NET_EPOCH_EXIT(et);
2484 		return (error);
2485 	}
2486 	NET_EPOCH_EXIT(et);
2487 	*pktopt = opt;
2488 	return (0);
2489 }
2490 
2491 /*
2492  * initialize ip6_pktopts.  beware that there are non-zero default values in
2493  * the struct.
2494  */
2495 void
2496 ip6_initpktopts(struct ip6_pktopts *opt)
2497 {
2498 
2499 	bzero(opt, sizeof(*opt));
2500 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2501 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2502 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2503 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2504 }
2505 
2506 static int
2507 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2508     struct ucred *cred, int uproto)
2509 {
2510 	struct epoch_tracker et;
2511 	struct ip6_pktopts *opt;
2512 	int ret;
2513 
2514 	if (*pktopt == NULL) {
2515 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2516 		    M_NOWAIT);
2517 		if (*pktopt == NULL)
2518 			return (ENOBUFS);
2519 		ip6_initpktopts(*pktopt);
2520 	}
2521 	opt = *pktopt;
2522 
2523 	NET_EPOCH_ENTER(et);
2524 	ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2525 	NET_EPOCH_EXIT(et);
2526 
2527 	return (ret);
2528 }
2529 
2530 #define GET_PKTOPT_VAR(field, lenexpr) do {				\
2531 	if (pktopt && pktopt->field) {					\
2532 		INP_RUNLOCK(inp);					\
2533 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);	\
2534 		malloc_optdata = true;					\
2535 		INP_RLOCK(inp);						\
2536 		if (inp->inp_flags & INP_DROPPED) {			\
2537 			INP_RUNLOCK(inp);				\
2538 			free(optdata, M_TEMP);				\
2539 			return (ECONNRESET);				\
2540 		}							\
2541 		pktopt = inp->in6p_outputopts;				\
2542 		if (pktopt && pktopt->field) {				\
2543 			optdatalen = min(lenexpr, sopt->sopt_valsize);	\
2544 			bcopy(pktopt->field, optdata, optdatalen);	\
2545 		} else {						\
2546 			free(optdata, M_TEMP);				\
2547 			optdata = NULL;					\
2548 			malloc_optdata = false;				\
2549 		}							\
2550 	}								\
2551 } while(0)
2552 
2553 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,			\
2554 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2555 
2556 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,		\
2557 	pktopt->field->sa_len)
2558 
2559 static int
2560 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2561 {
2562 	void *optdata = NULL;
2563 	bool malloc_optdata = false;
2564 	int optdatalen = 0;
2565 	int error = 0;
2566 	struct in6_pktinfo null_pktinfo;
2567 	int deftclass = 0, on;
2568 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2569 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2570 	struct ip6_pktopts *pktopt;
2571 
2572 	INP_RLOCK(inp);
2573 	pktopt = inp->in6p_outputopts;
2574 
2575 	switch (optname) {
2576 	case IPV6_PKTINFO:
2577 		optdata = (void *)&null_pktinfo;
2578 		if (pktopt && pktopt->ip6po_pktinfo) {
2579 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2580 			    sizeof(null_pktinfo));
2581 			in6_clearscope(&null_pktinfo.ipi6_addr);
2582 		} else {
2583 			/* XXX: we don't have to do this every time... */
2584 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2585 		}
2586 		optdatalen = sizeof(struct in6_pktinfo);
2587 		break;
2588 	case IPV6_TCLASS:
2589 		if (pktopt && pktopt->ip6po_tclass >= 0)
2590 			deftclass = pktopt->ip6po_tclass;
2591 		optdata = (void *)&deftclass;
2592 		optdatalen = sizeof(int);
2593 		break;
2594 	case IPV6_HOPOPTS:
2595 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2596 		break;
2597 	case IPV6_RTHDR:
2598 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2599 		break;
2600 	case IPV6_RTHDRDSTOPTS:
2601 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2602 		break;
2603 	case IPV6_DSTOPTS:
2604 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2605 		break;
2606 	case IPV6_NEXTHOP:
2607 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2608 		break;
2609 	case IPV6_USE_MIN_MTU:
2610 		if (pktopt)
2611 			defminmtu = pktopt->ip6po_minmtu;
2612 		optdata = (void *)&defminmtu;
2613 		optdatalen = sizeof(int);
2614 		break;
2615 	case IPV6_DONTFRAG:
2616 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2617 			on = 1;
2618 		else
2619 			on = 0;
2620 		optdata = (void *)&on;
2621 		optdatalen = sizeof(on);
2622 		break;
2623 	case IPV6_PREFER_TEMPADDR:
2624 		if (pktopt)
2625 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2626 		optdata = (void *)&defpreftemp;
2627 		optdatalen = sizeof(int);
2628 		break;
2629 	default:		/* should not happen */
2630 #ifdef DIAGNOSTIC
2631 		panic("ip6_getpcbopt: unexpected option\n");
2632 #endif
2633 		INP_RUNLOCK(inp);
2634 		return (ENOPROTOOPT);
2635 	}
2636 	INP_RUNLOCK(inp);
2637 
2638 	error = sooptcopyout(sopt, optdata, optdatalen);
2639 	if (malloc_optdata)
2640 		free(optdata, M_TEMP);
2641 
2642 	return (error);
2643 }
2644 
2645 void
2646 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2647 {
2648 	if (pktopt == NULL)
2649 		return;
2650 
2651 	if (optname == -1 || optname == IPV6_PKTINFO) {
2652 		if (pktopt->ip6po_pktinfo)
2653 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2654 		pktopt->ip6po_pktinfo = NULL;
2655 	}
2656 	if (optname == -1 || optname == IPV6_HOPLIMIT) {
2657 		pktopt->ip6po_hlim = -1;
2658 		pktopt->ip6po_valid &= ~IP6PO_VALID_HLIM;
2659 	}
2660 	if (optname == -1 || optname == IPV6_TCLASS) {
2661 		pktopt->ip6po_tclass = -1;
2662 		pktopt->ip6po_valid &= ~IP6PO_VALID_TC;
2663 	}
2664 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2665 		if (pktopt->ip6po_nextroute.ro_nh) {
2666 			NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2667 			pktopt->ip6po_nextroute.ro_nh = NULL;
2668 		}
2669 		if (pktopt->ip6po_nexthop)
2670 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2671 		pktopt->ip6po_nexthop = NULL;
2672 		pktopt->ip6po_valid &= ~IP6PO_VALID_NHINFO;
2673 	}
2674 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2675 		if (pktopt->ip6po_hbh)
2676 			free(pktopt->ip6po_hbh, M_IP6OPT);
2677 		pktopt->ip6po_hbh = NULL;
2678 		pktopt->ip6po_valid &= ~IP6PO_VALID_HBH;
2679 	}
2680 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2681 		if (pktopt->ip6po_dest1)
2682 			free(pktopt->ip6po_dest1, M_IP6OPT);
2683 		pktopt->ip6po_dest1 = NULL;
2684 		pktopt->ip6po_valid &= ~IP6PO_VALID_DEST1;
2685 	}
2686 	if (optname == -1 || optname == IPV6_RTHDR) {
2687 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2688 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2689 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2690 		if (pktopt->ip6po_route.ro_nh) {
2691 			NH_FREE(pktopt->ip6po_route.ro_nh);
2692 			pktopt->ip6po_route.ro_nh = NULL;
2693 		}
2694 		pktopt->ip6po_valid &= ~IP6PO_VALID_RHINFO;
2695 	}
2696 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2697 		if (pktopt->ip6po_dest2)
2698 			free(pktopt->ip6po_dest2, M_IP6OPT);
2699 		pktopt->ip6po_dest2 = NULL;
2700 		pktopt->ip6po_valid &= ~IP6PO_VALID_DEST2;
2701 	}
2702 }
2703 
2704 #define PKTOPT_EXTHDRCPY(type) \
2705 do {\
2706 	if (src->type) {\
2707 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2708 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2709 		if (dst->type == NULL)\
2710 			goto bad;\
2711 		bcopy(src->type, dst->type, hlen);\
2712 	}\
2713 } while (/*CONSTCOND*/ 0)
2714 
2715 static int
2716 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2717 {
2718 	if (dst == NULL || src == NULL)  {
2719 		printf("ip6_clearpktopts: invalid argument\n");
2720 		return (EINVAL);
2721 	}
2722 
2723 	dst->ip6po_hlim = src->ip6po_hlim;
2724 	dst->ip6po_tclass = src->ip6po_tclass;
2725 	dst->ip6po_flags = src->ip6po_flags;
2726 	dst->ip6po_minmtu = src->ip6po_minmtu;
2727 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2728 	if (src->ip6po_pktinfo) {
2729 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2730 		    M_IP6OPT, canwait);
2731 		if (dst->ip6po_pktinfo == NULL)
2732 			goto bad;
2733 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2734 	}
2735 	if (src->ip6po_nexthop) {
2736 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2737 		    M_IP6OPT, canwait);
2738 		if (dst->ip6po_nexthop == NULL)
2739 			goto bad;
2740 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2741 		    src->ip6po_nexthop->sa_len);
2742 	}
2743 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2744 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2745 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2746 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2747 	dst->ip6po_valid = src->ip6po_valid;
2748 	return (0);
2749 
2750   bad:
2751 	ip6_clearpktopts(dst, -1);
2752 	return (ENOBUFS);
2753 }
2754 #undef PKTOPT_EXTHDRCPY
2755 
2756 struct ip6_pktopts *
2757 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2758 {
2759 	int error;
2760 	struct ip6_pktopts *dst;
2761 
2762 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2763 	if (dst == NULL)
2764 		return (NULL);
2765 	ip6_initpktopts(dst);
2766 
2767 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2768 		free(dst, M_IP6OPT);
2769 		return (NULL);
2770 	}
2771 
2772 	return (dst);
2773 }
2774 
2775 void
2776 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2777 {
2778 	if (pktopt == NULL)
2779 		return;
2780 
2781 	ip6_clearpktopts(pktopt, -1);
2782 
2783 	free(pktopt, M_IP6OPT);
2784 }
2785 
2786 /*
2787  * Set IPv6 outgoing packet options based on advanced API.
2788  */
2789 int
2790 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2791     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2792 {
2793 	struct cmsghdr *cm = NULL;
2794 
2795 	if (control == NULL || opt == NULL)
2796 		return (EINVAL);
2797 
2798 	/*
2799 	 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2800 	 * are in the network epoch here.
2801 	 */
2802 	NET_EPOCH_ASSERT();
2803 
2804 	ip6_initpktopts(opt);
2805 	if (stickyopt) {
2806 		int error;
2807 
2808 		/*
2809 		 * If stickyopt is provided, make a local copy of the options
2810 		 * for this particular packet, then override them by ancillary
2811 		 * objects.
2812 		 * XXX: copypktopts() does not copy the cached route to a next
2813 		 * hop (if any).  This is not very good in terms of efficiency,
2814 		 * but we can allow this since this option should be rarely
2815 		 * used.
2816 		 */
2817 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2818 			return (error);
2819 	}
2820 
2821 	/*
2822 	 * XXX: Currently, we assume all the optional information is stored
2823 	 * in a single mbuf.
2824 	 */
2825 	if (control->m_next)
2826 		return (EINVAL);
2827 
2828 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2829 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2830 		int error;
2831 
2832 		if (control->m_len < CMSG_LEN(0))
2833 			return (EINVAL);
2834 
2835 		cm = mtod(control, struct cmsghdr *);
2836 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2837 			return (EINVAL);
2838 		if (cm->cmsg_level != IPPROTO_IPV6)
2839 			continue;
2840 
2841 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2842 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2843 		if (error)
2844 			return (error);
2845 	}
2846 
2847 	return (0);
2848 }
2849 
2850 /*
2851  * Set a particular packet option, as a sticky option or an ancillary data
2852  * item.  "len" can be 0 only when it's a sticky option.
2853  * We have 4 cases of combination of "sticky" and "cmsg":
2854  * "sticky=0, cmsg=0": impossible
2855  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2856  * "sticky=1, cmsg=0": RFC3542 socket option
2857  * "sticky=1, cmsg=1": RFC2292 socket option
2858  */
2859 static int
2860 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2861     struct ucred *cred, int sticky, int cmsg, int uproto)
2862 {
2863 	int minmtupolicy, preftemp;
2864 	int error;
2865 
2866 	NET_EPOCH_ASSERT();
2867 
2868 	if (!sticky && !cmsg) {
2869 #ifdef DIAGNOSTIC
2870 		printf("ip6_setpktopt: impossible case\n");
2871 #endif
2872 		return (EINVAL);
2873 	}
2874 
2875 	/*
2876 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2877 	 * not be specified in the context of RFC3542.  Conversely,
2878 	 * RFC3542 types should not be specified in the context of RFC2292.
2879 	 */
2880 	if (!cmsg) {
2881 		switch (optname) {
2882 		case IPV6_2292PKTINFO:
2883 		case IPV6_2292HOPLIMIT:
2884 		case IPV6_2292NEXTHOP:
2885 		case IPV6_2292HOPOPTS:
2886 		case IPV6_2292DSTOPTS:
2887 		case IPV6_2292RTHDR:
2888 		case IPV6_2292PKTOPTIONS:
2889 			return (ENOPROTOOPT);
2890 		}
2891 	}
2892 	if (sticky && cmsg) {
2893 		switch (optname) {
2894 		case IPV6_PKTINFO:
2895 		case IPV6_HOPLIMIT:
2896 		case IPV6_NEXTHOP:
2897 		case IPV6_HOPOPTS:
2898 		case IPV6_DSTOPTS:
2899 		case IPV6_RTHDRDSTOPTS:
2900 		case IPV6_RTHDR:
2901 		case IPV6_USE_MIN_MTU:
2902 		case IPV6_DONTFRAG:
2903 		case IPV6_TCLASS:
2904 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2905 			return (ENOPROTOOPT);
2906 		}
2907 	}
2908 
2909 	switch (optname) {
2910 	case IPV6_2292PKTINFO:
2911 	case IPV6_PKTINFO:
2912 	{
2913 		struct ifnet *ifp = NULL;
2914 		struct in6_pktinfo *pktinfo;
2915 
2916 		if (len != sizeof(struct in6_pktinfo))
2917 			return (EINVAL);
2918 
2919 		pktinfo = (struct in6_pktinfo *)buf;
2920 
2921 		/*
2922 		 * An application can clear any sticky IPV6_PKTINFO option by
2923 		 * doing a "regular" setsockopt with ipi6_addr being
2924 		 * in6addr_any and ipi6_ifindex being zero.
2925 		 * [RFC 3542, Section 6]
2926 		 */
2927 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2928 		    pktinfo->ipi6_ifindex == 0 &&
2929 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2930 			ip6_clearpktopts(opt, optname);
2931 			break;
2932 		}
2933 
2934 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2935 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2936 			return (EINVAL);
2937 		}
2938 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2939 			return (EINVAL);
2940 		/* validate the interface index if specified. */
2941 		if (pktinfo->ipi6_ifindex) {
2942 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2943 			if (ifp == NULL)
2944 				return (ENXIO);
2945 		}
2946 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2947 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2948 			return (ENETDOWN);
2949 
2950 		if (ifp != NULL &&
2951 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2952 			struct in6_ifaddr *ia;
2953 
2954 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2955 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2956 			if (ia == NULL)
2957 				return (EADDRNOTAVAIL);
2958 			ifa_free(&ia->ia_ifa);
2959 		}
2960 		/*
2961 		 * We store the address anyway, and let in6_selectsrc()
2962 		 * validate the specified address.  This is because ipi6_addr
2963 		 * may not have enough information about its scope zone, and
2964 		 * we may need additional information (such as outgoing
2965 		 * interface or the scope zone of a destination address) to
2966 		 * disambiguate the scope.
2967 		 * XXX: the delay of the validation may confuse the
2968 		 * application when it is used as a sticky option.
2969 		 */
2970 		if (opt->ip6po_pktinfo == NULL) {
2971 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2972 			    M_IP6OPT, M_NOWAIT);
2973 			if (opt->ip6po_pktinfo == NULL)
2974 				return (ENOBUFS);
2975 		}
2976 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2977 		opt->ip6po_valid |= IP6PO_VALID_PKTINFO;
2978 		break;
2979 	}
2980 
2981 	case IPV6_2292HOPLIMIT:
2982 	case IPV6_HOPLIMIT:
2983 	{
2984 		int *hlimp;
2985 
2986 		/*
2987 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2988 		 * to simplify the ordering among hoplimit options.
2989 		 */
2990 		if (optname == IPV6_HOPLIMIT && sticky)
2991 			return (ENOPROTOOPT);
2992 
2993 		if (len != sizeof(int))
2994 			return (EINVAL);
2995 		hlimp = (int *)buf;
2996 		if (*hlimp < -1 || *hlimp > 255)
2997 			return (EINVAL);
2998 
2999 		opt->ip6po_hlim = *hlimp;
3000 		opt->ip6po_valid |= IP6PO_VALID_HLIM;
3001 		break;
3002 	}
3003 
3004 	case IPV6_TCLASS:
3005 	{
3006 		int tclass;
3007 
3008 		if (len != sizeof(int))
3009 			return (EINVAL);
3010 		tclass = *(int *)buf;
3011 		if (tclass < -1 || tclass > 255)
3012 			return (EINVAL);
3013 
3014 		opt->ip6po_tclass = tclass;
3015 		opt->ip6po_valid |= IP6PO_VALID_TC;
3016 		break;
3017 	}
3018 
3019 	case IPV6_2292NEXTHOP:
3020 	case IPV6_NEXTHOP:
3021 		if (cred != NULL) {
3022 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3023 			if (error)
3024 				return (error);
3025 		}
3026 
3027 		if (len == 0) {	/* just remove the option */
3028 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3029 			break;
3030 		}
3031 
3032 		/* check if cmsg_len is large enough for sa_len */
3033 		if (len < sizeof(struct sockaddr) || len < *buf)
3034 			return (EINVAL);
3035 
3036 		switch (((struct sockaddr *)buf)->sa_family) {
3037 		case AF_INET6:
3038 		{
3039 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3040 			int error;
3041 
3042 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3043 				return (EINVAL);
3044 
3045 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3046 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3047 				return (EINVAL);
3048 			}
3049 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3050 			    != 0) {
3051 				return (error);
3052 			}
3053 			break;
3054 		}
3055 		case AF_LINK:	/* should eventually be supported */
3056 		default:
3057 			return (EAFNOSUPPORT);
3058 		}
3059 
3060 		/* turn off the previous option, then set the new option. */
3061 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3062 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3063 		if (opt->ip6po_nexthop == NULL)
3064 			return (ENOBUFS);
3065 		bcopy(buf, opt->ip6po_nexthop, *buf);
3066 		opt->ip6po_valid |= IP6PO_VALID_NHINFO;
3067 		break;
3068 
3069 	case IPV6_2292HOPOPTS:
3070 	case IPV6_HOPOPTS:
3071 	{
3072 		struct ip6_hbh *hbh;
3073 		int hbhlen;
3074 
3075 		/*
3076 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3077 		 * options, since per-option restriction has too much
3078 		 * overhead.
3079 		 */
3080 		if (cred != NULL) {
3081 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3082 			if (error)
3083 				return (error);
3084 		}
3085 
3086 		if (len == 0) {
3087 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3088 			break;	/* just remove the option */
3089 		}
3090 
3091 		/* message length validation */
3092 		if (len < sizeof(struct ip6_hbh))
3093 			return (EINVAL);
3094 		hbh = (struct ip6_hbh *)buf;
3095 		hbhlen = (hbh->ip6h_len + 1) << 3;
3096 		if (len != hbhlen)
3097 			return (EINVAL);
3098 
3099 		/* turn off the previous option, then set the new option. */
3100 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3101 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3102 		if (opt->ip6po_hbh == NULL)
3103 			return (ENOBUFS);
3104 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3105 		opt->ip6po_valid |= IP6PO_VALID_HBH;
3106 
3107 		break;
3108 	}
3109 
3110 	case IPV6_2292DSTOPTS:
3111 	case IPV6_DSTOPTS:
3112 	case IPV6_RTHDRDSTOPTS:
3113 	{
3114 		struct ip6_dest *dest, **newdest = NULL;
3115 		int destlen;
3116 
3117 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3118 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3119 			if (error)
3120 				return (error);
3121 		}
3122 
3123 		if (len == 0) {
3124 			ip6_clearpktopts(opt, optname);
3125 			break;	/* just remove the option */
3126 		}
3127 
3128 		/* message length validation */
3129 		if (len < sizeof(struct ip6_dest))
3130 			return (EINVAL);
3131 		dest = (struct ip6_dest *)buf;
3132 		destlen = (dest->ip6d_len + 1) << 3;
3133 		if (len != destlen)
3134 			return (EINVAL);
3135 
3136 		/*
3137 		 * Determine the position that the destination options header
3138 		 * should be inserted; before or after the routing header.
3139 		 */
3140 		switch (optname) {
3141 		case IPV6_2292DSTOPTS:
3142 			/*
3143 			 * The old advacned API is ambiguous on this point.
3144 			 * Our approach is to determine the position based
3145 			 * according to the existence of a routing header.
3146 			 * Note, however, that this depends on the order of the
3147 			 * extension headers in the ancillary data; the 1st
3148 			 * part of the destination options header must appear
3149 			 * before the routing header in the ancillary data,
3150 			 * too.
3151 			 * RFC3542 solved the ambiguity by introducing
3152 			 * separate ancillary data or option types.
3153 			 */
3154 			if (opt->ip6po_rthdr == NULL)
3155 				newdest = &opt->ip6po_dest1;
3156 			else
3157 				newdest = &opt->ip6po_dest2;
3158 			break;
3159 		case IPV6_RTHDRDSTOPTS:
3160 			newdest = &opt->ip6po_dest1;
3161 			break;
3162 		case IPV6_DSTOPTS:
3163 			newdest = &opt->ip6po_dest2;
3164 			break;
3165 		}
3166 
3167 		/* turn off the previous option, then set the new option. */
3168 		ip6_clearpktopts(opt, optname);
3169 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3170 		if (*newdest == NULL)
3171 			return (ENOBUFS);
3172 		bcopy(dest, *newdest, destlen);
3173 		if (newdest == &opt->ip6po_dest1)
3174 			opt->ip6po_valid |= IP6PO_VALID_DEST1;
3175 		else
3176 			opt->ip6po_valid |= IP6PO_VALID_DEST2;
3177 
3178 		break;
3179 	}
3180 
3181 	case IPV6_2292RTHDR:
3182 	case IPV6_RTHDR:
3183 	{
3184 		struct ip6_rthdr *rth;
3185 		int rthlen;
3186 
3187 		if (len == 0) {
3188 			ip6_clearpktopts(opt, IPV6_RTHDR);
3189 			break;	/* just remove the option */
3190 		}
3191 
3192 		/* message length validation */
3193 		if (len < sizeof(struct ip6_rthdr))
3194 			return (EINVAL);
3195 		rth = (struct ip6_rthdr *)buf;
3196 		rthlen = (rth->ip6r_len + 1) << 3;
3197 		if (len != rthlen)
3198 			return (EINVAL);
3199 
3200 		switch (rth->ip6r_type) {
3201 		case IPV6_RTHDR_TYPE_0:
3202 			if (rth->ip6r_len == 0)	/* must contain one addr */
3203 				return (EINVAL);
3204 			if (rth->ip6r_len % 2) /* length must be even */
3205 				return (EINVAL);
3206 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3207 				return (EINVAL);
3208 			break;
3209 		default:
3210 			return (EINVAL);	/* not supported */
3211 		}
3212 
3213 		/* turn off the previous option */
3214 		ip6_clearpktopts(opt, IPV6_RTHDR);
3215 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3216 		if (opt->ip6po_rthdr == NULL)
3217 			return (ENOBUFS);
3218 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3219 		opt->ip6po_valid |= IP6PO_VALID_RHINFO;
3220 
3221 		break;
3222 	}
3223 
3224 	case IPV6_USE_MIN_MTU:
3225 		if (len != sizeof(int))
3226 			return (EINVAL);
3227 		minmtupolicy = *(int *)buf;
3228 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3229 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3230 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3231 			return (EINVAL);
3232 		}
3233 		opt->ip6po_minmtu = minmtupolicy;
3234 		break;
3235 
3236 	case IPV6_DONTFRAG:
3237 		if (len != sizeof(int))
3238 			return (EINVAL);
3239 
3240 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3241 			/*
3242 			 * we ignore this option for TCP sockets.
3243 			 * (RFC3542 leaves this case unspecified.)
3244 			 */
3245 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3246 		} else
3247 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3248 		break;
3249 
3250 	case IPV6_PREFER_TEMPADDR:
3251 		if (len != sizeof(int))
3252 			return (EINVAL);
3253 		preftemp = *(int *)buf;
3254 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3255 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3256 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3257 			return (EINVAL);
3258 		}
3259 		opt->ip6po_prefer_tempaddr = preftemp;
3260 		break;
3261 
3262 	default:
3263 		return (ENOPROTOOPT);
3264 	} /* end of switch */
3265 
3266 	return (0);
3267 }
3268 
3269 /*
3270  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3271  * packet to the input queue of a specified interface.  Note that this
3272  * calls the output routine of the loopback "driver", but with an interface
3273  * pointer that might NOT be &loif -- easier than replicating that code here.
3274  */
3275 void
3276 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3277 {
3278 	struct mbuf *copym;
3279 	struct ip6_hdr *ip6;
3280 
3281 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3282 	if (copym == NULL)
3283 		return;
3284 
3285 	/*
3286 	 * Make sure to deep-copy IPv6 header portion in case the data
3287 	 * is in an mbuf cluster, so that we can safely override the IPv6
3288 	 * header portion later.
3289 	 */
3290 	if (!M_WRITABLE(copym) ||
3291 	    copym->m_len < sizeof(struct ip6_hdr)) {
3292 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3293 		if (copym == NULL)
3294 			return;
3295 	}
3296 	ip6 = mtod(copym, struct ip6_hdr *);
3297 	/*
3298 	 * clear embedded scope identifiers if necessary.
3299 	 * in6_clearscope will touch the addresses only when necessary.
3300 	 */
3301 	in6_clearscope(&ip6->ip6_src);
3302 	in6_clearscope(&ip6->ip6_dst);
3303 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3304 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3305 		    CSUM_PSEUDO_HDR;
3306 		copym->m_pkthdr.csum_data = 0xffff;
3307 	}
3308 	if_simloop(ifp, copym, AF_INET6, 0);
3309 }
3310 
3311 /*
3312  * Chop IPv6 header off from the payload.
3313  */
3314 static int
3315 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3316 {
3317 	struct mbuf *mh;
3318 	struct ip6_hdr *ip6;
3319 
3320 	ip6 = mtod(m, struct ip6_hdr *);
3321 	if (m->m_len > sizeof(*ip6)) {
3322 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3323 		if (mh == NULL) {
3324 			m_freem(m);
3325 			return ENOBUFS;
3326 		}
3327 		m_move_pkthdr(mh, m);
3328 		M_ALIGN(mh, sizeof(*ip6));
3329 		m->m_len -= sizeof(*ip6);
3330 		m->m_data += sizeof(*ip6);
3331 		mh->m_next = m;
3332 		m = mh;
3333 		m->m_len = sizeof(*ip6);
3334 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3335 	}
3336 	exthdrs->ip6e_ip6 = m;
3337 	return 0;
3338 }
3339 
3340 /*
3341  * Compute IPv6 extension header length.
3342  */
3343 int
3344 ip6_optlen(struct inpcb *inp)
3345 {
3346 	int len;
3347 
3348 	if (!inp->in6p_outputopts)
3349 		return 0;
3350 
3351 	len = 0;
3352 #define elen(x) \
3353     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3354 
3355 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3356 	if (inp->in6p_outputopts->ip6po_rthdr)
3357 		/* dest1 is valid with rthdr only */
3358 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3359 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3360 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3361 	return len;
3362 #undef elen
3363 }
3364