xref: /freebsd/sys/netinet6/ip6_output.c (revision e430d1ed78d021db4e9760d9800393b627156745)
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  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ratelimit.h"
71 #include "opt_ipsec.h"
72 #include "opt_sctp.h"
73 #include "opt_route.h"
74 #include "opt_rss.h"
75 
76 #include <sys/param.h>
77 #include <sys/kernel.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/errno.h>
81 #include <sys/priv.h>
82 #include <sys/proc.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/syslog.h>
87 #include <sys/ucred.h>
88 
89 #include <machine/in_cksum.h>
90 
91 #include <net/if.h>
92 #include <net/if_var.h>
93 #include <net/if_llatbl.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/pfil.h>
97 #include <net/rss_config.h>
98 #include <net/vnet.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_var.h>
102 #include <netinet/ip_var.h>
103 #include <netinet6/in6_fib.h>
104 #include <netinet6/in6_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/tcp_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/in6_rss.h>
112 
113 #include <netipsec/ipsec_support.h>
114 #ifdef SCTP
115 #include <netinet/sctp.h>
116 #include <netinet/sctp_crc32.h>
117 #endif
118 
119 #include <netinet6/ip6protosw.h>
120 #include <netinet6/scope6_var.h>
121 
122 extern int in6_mcast_loop;
123 
124 struct ip6_exthdrs {
125 	struct mbuf *ip6e_ip6;
126 	struct mbuf *ip6e_hbh;
127 	struct mbuf *ip6e_dest1;
128 	struct mbuf *ip6e_rthdr;
129 	struct mbuf *ip6e_dest2;
130 };
131 
132 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
133 
134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
135 			   struct ucred *, int);
136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
137 	struct socket *, struct sockopt *);
138 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
140 	struct ucred *, int, int, int);
141 
142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
144 	struct ip6_frag **);
145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
147 static int ip6_getpmtu(struct route_in6 *, int,
148 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
149 	u_int);
150 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
151 	u_long *, int *, u_int);
152 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
153 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
154 
155 
156 /*
157  * Make an extension header from option data.  hp is the source, and
158  * mp is the destination.
159  */
160 #define MAKE_EXTHDR(hp, mp)						\
161     do {								\
162 	if (hp) {							\
163 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
164 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
165 		    ((eh)->ip6e_len + 1) << 3);				\
166 		if (error)						\
167 			goto freehdrs;					\
168 	}								\
169     } while (/*CONSTCOND*/ 0)
170 
171 /*
172  * Form a chain of extension headers.
173  * m is the extension header mbuf
174  * mp is the previous mbuf in the chain
175  * p is the next header
176  * i is the type of option.
177  */
178 #define MAKE_CHAIN(m, mp, p, i)\
179     do {\
180 	if (m) {\
181 		if (!hdrsplit) \
182 			panic("assumption failed: hdr not split"); \
183 		*mtod((m), u_char *) = *(p);\
184 		*(p) = (i);\
185 		p = mtod((m), u_char *);\
186 		(m)->m_next = (mp)->m_next;\
187 		(mp)->m_next = (m);\
188 		(mp) = (m);\
189 	}\
190     } while (/*CONSTCOND*/ 0)
191 
192 void
193 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
194 {
195 	u_short csum;
196 
197 	csum = in_cksum_skip(m, offset + plen, offset);
198 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
199 		csum = 0xffff;
200 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
201 
202 	if (offset + sizeof(csum) > m->m_len)
203 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
204 	else
205 		*(u_short *)mtodo(m, offset) = csum;
206 }
207 
208 int
209 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
210     int fraglen , uint32_t id)
211 {
212 	struct mbuf *m, **mnext, *m_frgpart;
213 	struct ip6_hdr *ip6, *mhip6;
214 	struct ip6_frag *ip6f;
215 	int off;
216 	int error;
217 	int tlen = m0->m_pkthdr.len;
218 
219 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
220 
221 	m = m0;
222 	ip6 = mtod(m, struct ip6_hdr *);
223 	mnext = &m->m_nextpkt;
224 
225 	for (off = hlen; off < tlen; off += fraglen) {
226 		m = m_gethdr(M_NOWAIT, MT_DATA);
227 		if (!m) {
228 			IP6STAT_INC(ip6s_odropped);
229 			return (ENOBUFS);
230 		}
231 		m->m_flags = m0->m_flags & M_COPYFLAGS;
232 		*mnext = m;
233 		mnext = &m->m_nextpkt;
234 		m->m_data += max_linkhdr;
235 		mhip6 = mtod(m, struct ip6_hdr *);
236 		*mhip6 = *ip6;
237 		m->m_len = sizeof(*mhip6);
238 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
239 		if (error) {
240 			IP6STAT_INC(ip6s_odropped);
241 			return (error);
242 		}
243 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
244 		if (off + fraglen >= tlen)
245 			fraglen = tlen - off;
246 		else
247 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
248 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
249 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
250 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
251 			IP6STAT_INC(ip6s_odropped);
252 			return (ENOBUFS);
253 		}
254 		m_cat(m, m_frgpart);
255 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
256 		m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
257 		m->m_pkthdr.rcvif = NULL;
258 		ip6f->ip6f_reserved = 0;
259 		ip6f->ip6f_ident = id;
260 		ip6f->ip6f_nxt = nextproto;
261 		IP6STAT_INC(ip6s_ofragments);
262 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
263 	}
264 
265 	return (0);
266 }
267 
268 /*
269  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
270  * header (with pri, len, nxt, hlim, src, dst).
271  * This function may modify ver and hlim only.
272  * The mbuf chain containing the packet will be freed.
273  * The mbuf opt, if present, will not be freed.
274  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
275  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
276  * then result of route lookup is stored in ro->ro_rt.
277  *
278  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
279  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
280  * which is rt_mtu.
281  *
282  * ifpp - XXX: just for statistics
283  */
284 /*
285  * XXX TODO: no flowid is assigned for outbound flows?
286  */
287 int
288 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
289     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
290     struct ifnet **ifpp, struct inpcb *inp)
291 {
292 	struct ip6_hdr *ip6;
293 	struct ifnet *ifp, *origifp;
294 	struct mbuf *m = m0;
295 	struct mbuf *mprev = NULL;
296 	int hlen, tlen, len;
297 	struct route_in6 ip6route;
298 	struct rtentry *rt = NULL;
299 	struct sockaddr_in6 *dst, src_sa, dst_sa;
300 	struct in6_addr odst;
301 	int error = 0;
302 	struct in6_ifaddr *ia = NULL;
303 	u_long mtu;
304 	int alwaysfrag, dontfrag;
305 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
306 	struct ip6_exthdrs exthdrs;
307 	struct in6_addr src0, dst0;
308 	u_int32_t zone;
309 	struct route_in6 *ro_pmtu = NULL;
310 	int hdrsplit = 0;
311 	int sw_csum, tso;
312 	int needfiblookup;
313 	uint32_t fibnum;
314 	struct m_tag *fwd_tag = NULL;
315 	uint32_t id;
316 
317 	if (inp != NULL) {
318 		INP_LOCK_ASSERT(inp);
319 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
320 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
321 			/* unconditionally set flowid */
322 			m->m_pkthdr.flowid = inp->inp_flowid;
323 			M_HASHTYPE_SET(m, inp->inp_flowtype);
324 		}
325 #ifdef NUMA
326 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
327 #endif
328 	}
329 
330 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
331 	/*
332 	 * IPSec checking which handles several cases.
333 	 * FAST IPSEC: We re-injected the packet.
334 	 * XXX: need scope argument.
335 	 */
336 	if (IPSEC_ENABLED(ipv6)) {
337 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
338 			if (error == EINPROGRESS)
339 				error = 0;
340 			goto done;
341 		}
342 	}
343 #endif /* IPSEC */
344 
345 	bzero(&exthdrs, sizeof(exthdrs));
346 	if (opt) {
347 		/* Hop-by-Hop options header */
348 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
349 		/* Destination options header(1st part) */
350 		if (opt->ip6po_rthdr) {
351 			/*
352 			 * Destination options header(1st part)
353 			 * This only makes sense with a routing header.
354 			 * See Section 9.2 of RFC 3542.
355 			 * Disabling this part just for MIP6 convenience is
356 			 * a bad idea.  We need to think carefully about a
357 			 * way to make the advanced API coexist with MIP6
358 			 * options, which might automatically be inserted in
359 			 * the kernel.
360 			 */
361 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
362 		}
363 		/* Routing header */
364 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
365 		/* Destination options header(2nd part) */
366 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
367 	}
368 
369 	/*
370 	 * Calculate the total length of the extension header chain.
371 	 * Keep the length of the unfragmentable part for fragmentation.
372 	 */
373 	optlen = 0;
374 	if (exthdrs.ip6e_hbh)
375 		optlen += exthdrs.ip6e_hbh->m_len;
376 	if (exthdrs.ip6e_dest1)
377 		optlen += exthdrs.ip6e_dest1->m_len;
378 	if (exthdrs.ip6e_rthdr)
379 		optlen += exthdrs.ip6e_rthdr->m_len;
380 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
381 
382 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
383 	if (exthdrs.ip6e_dest2)
384 		optlen += exthdrs.ip6e_dest2->m_len;
385 
386 	/*
387 	 * If there is at least one extension header,
388 	 * separate IP6 header from the payload.
389 	 */
390 	if (optlen && !hdrsplit) {
391 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
392 			m = NULL;
393 			goto freehdrs;
394 		}
395 		m = exthdrs.ip6e_ip6;
396 		hdrsplit++;
397 	}
398 
399 	ip6 = mtod(m, struct ip6_hdr *);
400 
401 	/* adjust mbuf packet header length */
402 	m->m_pkthdr.len += optlen;
403 	plen = m->m_pkthdr.len - sizeof(*ip6);
404 
405 	/* If this is a jumbo payload, insert a jumbo payload option. */
406 	if (plen > IPV6_MAXPACKET) {
407 		if (!hdrsplit) {
408 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
409 				m = NULL;
410 				goto freehdrs;
411 			}
412 			m = exthdrs.ip6e_ip6;
413 			hdrsplit++;
414 		}
415 		/* adjust pointer */
416 		ip6 = mtod(m, struct ip6_hdr *);
417 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
418 			goto freehdrs;
419 		ip6->ip6_plen = 0;
420 	} else
421 		ip6->ip6_plen = htons(plen);
422 
423 	/*
424 	 * Concatenate headers and fill in next header fields.
425 	 * Here we have, on "m"
426 	 *	IPv6 payload
427 	 * and we insert headers accordingly.  Finally, we should be getting:
428 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
429 	 *
430 	 * during the header composing process, "m" points to IPv6 header.
431 	 * "mprev" points to an extension header prior to esp.
432 	 */
433 	u_char *nexthdrp = &ip6->ip6_nxt;
434 	mprev = m;
435 
436 	/*
437 	 * we treat dest2 specially.  this makes IPsec processing
438 	 * much easier.  the goal here is to make mprev point the
439 	 * mbuf prior to dest2.
440 	 *
441 	 * result: IPv6 dest2 payload
442 	 * m and mprev will point to IPv6 header.
443 	 */
444 	if (exthdrs.ip6e_dest2) {
445 		if (!hdrsplit)
446 			panic("assumption failed: hdr not split");
447 		exthdrs.ip6e_dest2->m_next = m->m_next;
448 		m->m_next = exthdrs.ip6e_dest2;
449 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
450 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
451 	}
452 
453 	/*
454 	 * result: IPv6 hbh dest1 rthdr dest2 payload
455 	 * m will point to IPv6 header.  mprev will point to the
456 	 * extension header prior to dest2 (rthdr in the above case).
457 	 */
458 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
459 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
460 		   IPPROTO_DSTOPTS);
461 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
462 		   IPPROTO_ROUTING);
463 
464 	/*
465 	 * If there is a routing header, discard the packet.
466 	 */
467 	if (exthdrs.ip6e_rthdr) {
468 		 error = EINVAL;
469 		 goto bad;
470 	}
471 
472 	/* Source address validation */
473 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
474 	    (flags & IPV6_UNSPECSRC) == 0) {
475 		error = EOPNOTSUPP;
476 		IP6STAT_INC(ip6s_badscope);
477 		goto bad;
478 	}
479 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
480 		error = EOPNOTSUPP;
481 		IP6STAT_INC(ip6s_badscope);
482 		goto bad;
483 	}
484 
485 	IP6STAT_INC(ip6s_localout);
486 
487 	/*
488 	 * Route packet.
489 	 */
490 	if (ro == NULL) {
491 		ro = &ip6route;
492 		bzero((caddr_t)ro, sizeof(*ro));
493 	}
494 	ro_pmtu = ro;
495 	if (opt && opt->ip6po_rthdr)
496 		ro = &opt->ip6po_route;
497 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
498 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
499 again:
500 	/*
501 	 * if specified, try to fill in the traffic class field.
502 	 * do not override if a non-zero value is already set.
503 	 * we check the diffserv field and the ecn field separately.
504 	 */
505 	if (opt && opt->ip6po_tclass >= 0) {
506 		int mask = 0;
507 
508 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
509 			mask |= 0xfc;
510 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
511 			mask |= 0x03;
512 		if (mask != 0)
513 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
514 	}
515 
516 	/* fill in or override the hop limit field, if necessary. */
517 	if (opt && opt->ip6po_hlim != -1)
518 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
519 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
520 		if (im6o != NULL)
521 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
522 		else
523 			ip6->ip6_hlim = V_ip6_defmcasthlim;
524 	}
525 	/*
526 	 * Validate route against routing table additions;
527 	 * a better/more specific route might have been added.
528 	 * Make sure address family is set in route.
529 	 */
530 	if (inp) {
531 		ro->ro_dst.sin6_family = AF_INET6;
532 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
533 	}
534 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
535 	    ro->ro_dst.sin6_family == AF_INET6 &&
536 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
537 		rt = ro->ro_rt;
538 		ifp = ro->ro_rt->rt_ifp;
539 	} else {
540 		if (ro->ro_lle)
541 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
542 		ro->ro_lle = NULL;
543 		if (fwd_tag == NULL) {
544 			bzero(&dst_sa, sizeof(dst_sa));
545 			dst_sa.sin6_family = AF_INET6;
546 			dst_sa.sin6_len = sizeof(dst_sa);
547 			dst_sa.sin6_addr = ip6->ip6_dst;
548 		}
549 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
550 		    &rt, fibnum);
551 		if (error != 0) {
552 			if (ifp != NULL)
553 				in6_ifstat_inc(ifp, ifs6_out_discard);
554 			goto bad;
555 		}
556 	}
557 	if (rt == NULL) {
558 		/*
559 		 * If in6_selectroute() does not return a route entry,
560 		 * dst may not have been updated.
561 		 */
562 		*dst = dst_sa;	/* XXX */
563 	}
564 
565 	/*
566 	 * then rt (for unicast) and ifp must be non-NULL valid values.
567 	 */
568 	if ((flags & IPV6_FORWARDING) == 0) {
569 		/* XXX: the FORWARDING flag can be set for mrouting. */
570 		in6_ifstat_inc(ifp, ifs6_out_request);
571 	}
572 	if (rt != NULL) {
573 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
574 		counter_u64_add(rt->rt_pksent, 1);
575 	}
576 
577 	/* Setup data structures for scope ID checks. */
578 	src0 = ip6->ip6_src;
579 	bzero(&src_sa, sizeof(src_sa));
580 	src_sa.sin6_family = AF_INET6;
581 	src_sa.sin6_len = sizeof(src_sa);
582 	src_sa.sin6_addr = ip6->ip6_src;
583 
584 	dst0 = ip6->ip6_dst;
585 	/* re-initialize to be sure */
586 	bzero(&dst_sa, sizeof(dst_sa));
587 	dst_sa.sin6_family = AF_INET6;
588 	dst_sa.sin6_len = sizeof(dst_sa);
589 	dst_sa.sin6_addr = ip6->ip6_dst;
590 
591 	/* Check for valid scope ID. */
592 	if (in6_setscope(&src0, ifp, &zone) == 0 &&
593 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
594 	    in6_setscope(&dst0, ifp, &zone) == 0 &&
595 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
596 		/*
597 		 * The outgoing interface is in the zone of the source
598 		 * and destination addresses.
599 		 *
600 		 * Because the loopback interface cannot receive
601 		 * packets with a different scope ID than its own,
602 		 * there is a trick is to pretend the outgoing packet
603 		 * was received by the real network interface, by
604 		 * setting "origifp" different from "ifp". This is
605 		 * only allowed when "ifp" is a loopback network
606 		 * interface. Refer to code in nd6_output_ifp() for
607 		 * more details.
608 		 */
609 		origifp = ifp;
610 
611 		/*
612 		 * We should use ia_ifp to support the case of sending
613 		 * packets to an address of our own.
614 		 */
615 		if (ia != NULL && ia->ia_ifp)
616 			ifp = ia->ia_ifp;
617 
618 	} else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
619 	    sa6_recoverscope(&src_sa) != 0 ||
620 	    sa6_recoverscope(&dst_sa) != 0 ||
621 	    dst_sa.sin6_scope_id == 0 ||
622 	    (src_sa.sin6_scope_id != 0 &&
623 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
624 	    (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
625 		/*
626 		 * If the destination network interface is not a
627 		 * loopback interface, or the destination network
628 		 * address has no scope ID, or the source address has
629 		 * a scope ID set which is different from the
630 		 * destination address one, or there is no network
631 		 * interface representing this scope ID, the address
632 		 * pair is considered invalid.
633 		 */
634 		IP6STAT_INC(ip6s_badscope);
635 		in6_ifstat_inc(ifp, ifs6_out_discard);
636 		if (error == 0)
637 			error = EHOSTUNREACH; /* XXX */
638 		goto bad;
639 	}
640 
641 	/* All scope ID checks are successful. */
642 
643 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
644 		if (opt && opt->ip6po_nextroute.ro_rt) {
645 			/*
646 			 * The nexthop is explicitly specified by the
647 			 * application.  We assume the next hop is an IPv6
648 			 * address.
649 			 */
650 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
651 		}
652 		else if ((rt->rt_flags & RTF_GATEWAY))
653 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
654 	}
655 
656 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
657 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
658 	} else {
659 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
660 		in6_ifstat_inc(ifp, ifs6_out_mcast);
661 		/*
662 		 * Confirm that the outgoing interface supports multicast.
663 		 */
664 		if (!(ifp->if_flags & IFF_MULTICAST)) {
665 			IP6STAT_INC(ip6s_noroute);
666 			in6_ifstat_inc(ifp, ifs6_out_discard);
667 			error = ENETUNREACH;
668 			goto bad;
669 		}
670 		if ((im6o == NULL && in6_mcast_loop) ||
671 		    (im6o && im6o->im6o_multicast_loop)) {
672 			/*
673 			 * Loop back multicast datagram if not expressly
674 			 * forbidden to do so, even if we have not joined
675 			 * the address; protocols will filter it later,
676 			 * thus deferring a hash lookup and lock acquisition
677 			 * at the expense of an m_copym().
678 			 */
679 			ip6_mloopback(ifp, m);
680 		} else {
681 			/*
682 			 * If we are acting as a multicast router, perform
683 			 * multicast forwarding as if the packet had just
684 			 * arrived on the interface to which we are about
685 			 * to send.  The multicast forwarding function
686 			 * recursively calls this function, using the
687 			 * IPV6_FORWARDING flag to prevent infinite recursion.
688 			 *
689 			 * Multicasts that are looped back by ip6_mloopback(),
690 			 * above, will be forwarded by the ip6_input() routine,
691 			 * if necessary.
692 			 */
693 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
694 				/*
695 				 * XXX: ip6_mforward expects that rcvif is NULL
696 				 * when it is called from the originating path.
697 				 * However, it may not always be the case.
698 				 */
699 				m->m_pkthdr.rcvif = NULL;
700 				if (ip6_mforward(ip6, ifp, m) != 0) {
701 					m_freem(m);
702 					goto done;
703 				}
704 			}
705 		}
706 		/*
707 		 * Multicasts with a hoplimit of zero may be looped back,
708 		 * above, but must not be transmitted on a network.
709 		 * Also, multicasts addressed to the loopback interface
710 		 * are not sent -- the above call to ip6_mloopback() will
711 		 * loop back a copy if this host actually belongs to the
712 		 * destination group on the loopback interface.
713 		 */
714 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
715 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
716 			m_freem(m);
717 			goto done;
718 		}
719 	}
720 
721 	/*
722 	 * Fill the outgoing inteface to tell the upper layer
723 	 * to increment per-interface statistics.
724 	 */
725 	if (ifpp)
726 		*ifpp = ifp;
727 
728 	/* Determine path MTU. */
729 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
730 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
731 		goto bad;
732 
733 	/*
734 	 * The caller of this function may specify to use the minimum MTU
735 	 * in some cases.
736 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
737 	 * setting.  The logic is a bit complicated; by default, unicast
738 	 * packets will follow path MTU while multicast packets will be sent at
739 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
740 	 * including unicast ones will be sent at the minimum MTU.  Multicast
741 	 * packets will always be sent at the minimum MTU unless
742 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
743 	 * See RFC 3542 for more details.
744 	 */
745 	if (mtu > IPV6_MMTU) {
746 		if ((flags & IPV6_MINMTU))
747 			mtu = IPV6_MMTU;
748 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
749 			mtu = IPV6_MMTU;
750 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
751 			 (opt == NULL ||
752 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
753 			mtu = IPV6_MMTU;
754 		}
755 	}
756 
757 	/*
758 	 * clear embedded scope identifiers if necessary.
759 	 * in6_clearscope will touch the addresses only when necessary.
760 	 */
761 	in6_clearscope(&ip6->ip6_src);
762 	in6_clearscope(&ip6->ip6_dst);
763 
764 	/*
765 	 * If the outgoing packet contains a hop-by-hop options header,
766 	 * it must be examined and processed even by the source node.
767 	 * (RFC 2460, section 4.)
768 	 */
769 	if (exthdrs.ip6e_hbh) {
770 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
771 		u_int32_t dummy; /* XXX unused */
772 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
773 
774 #ifdef DIAGNOSTIC
775 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
776 			panic("ip6e_hbh is not contiguous");
777 #endif
778 		/*
779 		 *  XXX: if we have to send an ICMPv6 error to the sender,
780 		 *       we need the M_LOOP flag since icmp6_error() expects
781 		 *       the IPv6 and the hop-by-hop options header are
782 		 *       contiguous unless the flag is set.
783 		 */
784 		m->m_flags |= M_LOOP;
785 		m->m_pkthdr.rcvif = ifp;
786 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
787 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
788 		    &dummy, &plen) < 0) {
789 			/* m was already freed at this point */
790 			error = EINVAL;/* better error? */
791 			goto done;
792 		}
793 		m->m_flags &= ~M_LOOP; /* XXX */
794 		m->m_pkthdr.rcvif = NULL;
795 	}
796 
797 	/* Jump over all PFIL processing if hooks are not active. */
798 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
799 		goto passout;
800 
801 	odst = ip6->ip6_dst;
802 	/* Run through list of hooks for output packets. */
803 	switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
804 	case PFIL_PASS:
805 		ip6 = mtod(m, struct ip6_hdr *);
806 		break;
807 	case PFIL_DROPPED:
808 		error = EPERM;
809 		/* FALLTHROUGH */
810 	case PFIL_CONSUMED:
811 		goto done;
812 	}
813 
814 	needfiblookup = 0;
815 	/* See if destination IP address was changed by packet filter. */
816 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
817 		m->m_flags |= M_SKIP_FIREWALL;
818 		/* If destination is now ourself drop to ip6_input(). */
819 		if (in6_localip(&ip6->ip6_dst)) {
820 			m->m_flags |= M_FASTFWD_OURS;
821 			if (m->m_pkthdr.rcvif == NULL)
822 				m->m_pkthdr.rcvif = V_loif;
823 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
824 				m->m_pkthdr.csum_flags |=
825 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
826 				m->m_pkthdr.csum_data = 0xffff;
827 			}
828 #ifdef SCTP
829 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
830 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
831 #endif
832 			error = netisr_queue(NETISR_IPV6, m);
833 			goto done;
834 		} else {
835 			RO_INVALIDATE_CACHE(ro);
836 			needfiblookup = 1; /* Redo the routing table lookup. */
837 		}
838 	}
839 	/* See if fib was changed by packet filter. */
840 	if (fibnum != M_GETFIB(m)) {
841 		m->m_flags |= M_SKIP_FIREWALL;
842 		fibnum = M_GETFIB(m);
843 		RO_INVALIDATE_CACHE(ro);
844 		needfiblookup = 1;
845 	}
846 	if (needfiblookup)
847 		goto again;
848 
849 	/* See if local, if yes, send it to netisr. */
850 	if (m->m_flags & M_FASTFWD_OURS) {
851 		if (m->m_pkthdr.rcvif == NULL)
852 			m->m_pkthdr.rcvif = V_loif;
853 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
854 			m->m_pkthdr.csum_flags |=
855 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
856 			m->m_pkthdr.csum_data = 0xffff;
857 		}
858 #ifdef SCTP
859 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
860 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
861 #endif
862 		error = netisr_queue(NETISR_IPV6, m);
863 		goto done;
864 	}
865 	/* Or forward to some other address? */
866 	if ((m->m_flags & M_IP6_NEXTHOP) &&
867 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
868 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
869 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
870 		m->m_flags |= M_SKIP_FIREWALL;
871 		m->m_flags &= ~M_IP6_NEXTHOP;
872 		m_tag_delete(m, fwd_tag);
873 		goto again;
874 	}
875 
876 passout:
877 	/*
878 	 * Send the packet to the outgoing interface.
879 	 * If necessary, do IPv6 fragmentation before sending.
880 	 *
881 	 * the logic here is rather complex:
882 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
883 	 * 1-a:	send as is if tlen <= path mtu
884 	 * 1-b:	fragment if tlen > path mtu
885 	 *
886 	 * 2: if user asks us not to fragment (dontfrag == 1)
887 	 * 2-a:	send as is if tlen <= interface mtu
888 	 * 2-b:	error if tlen > interface mtu
889 	 *
890 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
891 	 *	always fragment
892 	 *
893 	 * 4: if dontfrag == 1 && alwaysfrag == 1
894 	 *	error, as we cannot handle this conflicting request
895 	 */
896 	sw_csum = m->m_pkthdr.csum_flags;
897 	if (!hdrsplit) {
898 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
899 		sw_csum &= ~ifp->if_hwassist;
900 	} else
901 		tso = 0;
902 	/*
903 	 * If we added extension headers, we will not do TSO and calculate the
904 	 * checksums ourselves for now.
905 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
906 	 * with ext. hdrs.
907 	 */
908 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
909 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
910 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
911 	}
912 #ifdef SCTP
913 	if (sw_csum & CSUM_SCTP_IPV6) {
914 		sw_csum &= ~CSUM_SCTP_IPV6;
915 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
916 	}
917 #endif
918 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
919 	tlen = m->m_pkthdr.len;
920 
921 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
922 		dontfrag = 1;
923 	else
924 		dontfrag = 0;
925 	if (dontfrag && alwaysfrag) {	/* case 4 */
926 		/* conflicting request - can't transmit */
927 		error = EMSGSIZE;
928 		goto bad;
929 	}
930 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
931 		/*
932 		 * Even if the DONTFRAG option is specified, we cannot send the
933 		 * packet when the data length is larger than the MTU of the
934 		 * outgoing interface.
935 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
936 		 * application wanted to know the MTU value. Also return an
937 		 * error code (this is not described in the API spec).
938 		 */
939 		if (inp != NULL)
940 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
941 		error = EMSGSIZE;
942 		goto bad;
943 	}
944 
945 	/*
946 	 * transmit packet without fragmentation
947 	 */
948 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
949 		struct in6_ifaddr *ia6;
950 
951 		ip6 = mtod(m, struct ip6_hdr *);
952 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
953 		if (ia6) {
954 			/* Record statistics for this interface address. */
955 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
956 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
957 			    m->m_pkthdr.len);
958 			ifa_free(&ia6->ia_ifa);
959 		}
960 #ifdef RATELIMIT
961 		if (inp != NULL) {
962 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
963 				in_pcboutput_txrtlmt(inp, ifp, m);
964 			/* stamp send tag on mbuf */
965 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
966 		} else {
967 			m->m_pkthdr.snd_tag = NULL;
968 		}
969 #endif
970 		error = nd6_output_ifp(ifp, origifp, m, dst,
971 		    (struct route *)ro);
972 #ifdef RATELIMIT
973 		/* check for route change */
974 		if (error == EAGAIN)
975 			in_pcboutput_eagain(inp);
976 #endif
977 		goto done;
978 	}
979 
980 	/*
981 	 * try to fragment the packet.  case 1-b and 3
982 	 */
983 	if (mtu < IPV6_MMTU) {
984 		/* path MTU cannot be less than IPV6_MMTU */
985 		error = EMSGSIZE;
986 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
987 		goto bad;
988 	} else if (ip6->ip6_plen == 0) {
989 		/* jumbo payload cannot be fragmented */
990 		error = EMSGSIZE;
991 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
992 		goto bad;
993 	} else {
994 		u_char nextproto;
995 
996 		/*
997 		 * Too large for the destination or interface;
998 		 * fragment if possible.
999 		 * Must be able to put at least 8 bytes per fragment.
1000 		 */
1001 		hlen = unfragpartlen;
1002 		if (mtu > IPV6_MAXPACKET)
1003 			mtu = IPV6_MAXPACKET;
1004 
1005 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1006 		if (len < 8) {
1007 			error = EMSGSIZE;
1008 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1009 			goto bad;
1010 		}
1011 
1012 		/*
1013 		 * If the interface will not calculate checksums on
1014 		 * fragmented packets, then do it here.
1015 		 * XXX-BZ handle the hw offloading case.  Need flags.
1016 		 */
1017 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1018 			in6_delayed_cksum(m, plen, hlen);
1019 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1020 		}
1021 #ifdef SCTP
1022 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1023 			sctp_delayed_cksum(m, hlen);
1024 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1025 		}
1026 #endif
1027 		/*
1028 		 * Change the next header field of the last header in the
1029 		 * unfragmentable part.
1030 		 */
1031 		if (exthdrs.ip6e_rthdr) {
1032 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1033 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1034 		} else if (exthdrs.ip6e_dest1) {
1035 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1036 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1037 		} else if (exthdrs.ip6e_hbh) {
1038 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1039 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1040 		} else {
1041 			nextproto = ip6->ip6_nxt;
1042 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1043 		}
1044 
1045 		/*
1046 		 * Loop through length of segment after first fragment,
1047 		 * make new header and copy data of each part and link onto
1048 		 * chain.
1049 		 */
1050 		m0 = m;
1051 		id = htonl(ip6_randomid());
1052 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1053 			goto sendorfree;
1054 
1055 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1056 	}
1057 
1058 	/*
1059 	 * Remove leading garbages.
1060 	 */
1061 sendorfree:
1062 	m = m0->m_nextpkt;
1063 	m0->m_nextpkt = 0;
1064 	m_freem(m0);
1065 	for (; m; m = m0) {
1066 		m0 = m->m_nextpkt;
1067 		m->m_nextpkt = 0;
1068 		if (error == 0) {
1069 			/* Record statistics for this interface address. */
1070 			if (ia) {
1071 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1072 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1073 				    m->m_pkthdr.len);
1074 			}
1075 #ifdef RATELIMIT
1076 			if (inp != NULL) {
1077 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1078 					in_pcboutput_txrtlmt(inp, ifp, m);
1079 				/* stamp send tag on mbuf */
1080 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1081 			} else {
1082 				m->m_pkthdr.snd_tag = NULL;
1083 			}
1084 #endif
1085 			error = nd6_output_ifp(ifp, origifp, m, dst,
1086 			    (struct route *)ro);
1087 #ifdef RATELIMIT
1088 			/* check for route change */
1089 			if (error == EAGAIN)
1090 				in_pcboutput_eagain(inp);
1091 #endif
1092 		} else
1093 			m_freem(m);
1094 	}
1095 
1096 	if (error == 0)
1097 		IP6STAT_INC(ip6s_fragmented);
1098 
1099 done:
1100 	if (ro == &ip6route)
1101 		RO_RTFREE(ro);
1102 	return (error);
1103 
1104 freehdrs:
1105 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1106 	m_freem(exthdrs.ip6e_dest1);
1107 	m_freem(exthdrs.ip6e_rthdr);
1108 	m_freem(exthdrs.ip6e_dest2);
1109 	/* FALLTHROUGH */
1110 bad:
1111 	if (m)
1112 		m_freem(m);
1113 	goto done;
1114 }
1115 
1116 static int
1117 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1118 {
1119 	struct mbuf *m;
1120 
1121 	if (hlen > MCLBYTES)
1122 		return (ENOBUFS); /* XXX */
1123 
1124 	if (hlen > MLEN)
1125 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1126 	else
1127 		m = m_get(M_NOWAIT, MT_DATA);
1128 	if (m == NULL)
1129 		return (ENOBUFS);
1130 	m->m_len = hlen;
1131 	if (hdr)
1132 		bcopy(hdr, mtod(m, caddr_t), hlen);
1133 
1134 	*mp = m;
1135 	return (0);
1136 }
1137 
1138 /*
1139  * Insert jumbo payload option.
1140  */
1141 static int
1142 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1143 {
1144 	struct mbuf *mopt;
1145 	u_char *optbuf;
1146 	u_int32_t v;
1147 
1148 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1149 
1150 	/*
1151 	 * If there is no hop-by-hop options header, allocate new one.
1152 	 * If there is one but it doesn't have enough space to store the
1153 	 * jumbo payload option, allocate a cluster to store the whole options.
1154 	 * Otherwise, use it to store the options.
1155 	 */
1156 	if (exthdrs->ip6e_hbh == NULL) {
1157 		mopt = m_get(M_NOWAIT, MT_DATA);
1158 		if (mopt == NULL)
1159 			return (ENOBUFS);
1160 		mopt->m_len = JUMBOOPTLEN;
1161 		optbuf = mtod(mopt, u_char *);
1162 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1163 		exthdrs->ip6e_hbh = mopt;
1164 	} else {
1165 		struct ip6_hbh *hbh;
1166 
1167 		mopt = exthdrs->ip6e_hbh;
1168 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1169 			/*
1170 			 * XXX assumption:
1171 			 * - exthdrs->ip6e_hbh is not referenced from places
1172 			 *   other than exthdrs.
1173 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1174 			 */
1175 			int oldoptlen = mopt->m_len;
1176 			struct mbuf *n;
1177 
1178 			/*
1179 			 * XXX: give up if the whole (new) hbh header does
1180 			 * not fit even in an mbuf cluster.
1181 			 */
1182 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1183 				return (ENOBUFS);
1184 
1185 			/*
1186 			 * As a consequence, we must always prepare a cluster
1187 			 * at this point.
1188 			 */
1189 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1190 			if (n == NULL)
1191 				return (ENOBUFS);
1192 			n->m_len = oldoptlen + JUMBOOPTLEN;
1193 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1194 			    oldoptlen);
1195 			optbuf = mtod(n, caddr_t) + oldoptlen;
1196 			m_freem(mopt);
1197 			mopt = exthdrs->ip6e_hbh = n;
1198 		} else {
1199 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1200 			mopt->m_len += JUMBOOPTLEN;
1201 		}
1202 		optbuf[0] = IP6OPT_PADN;
1203 		optbuf[1] = 1;
1204 
1205 		/*
1206 		 * Adjust the header length according to the pad and
1207 		 * the jumbo payload option.
1208 		 */
1209 		hbh = mtod(mopt, struct ip6_hbh *);
1210 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1211 	}
1212 
1213 	/* fill in the option. */
1214 	optbuf[2] = IP6OPT_JUMBO;
1215 	optbuf[3] = 4;
1216 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1217 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1218 
1219 	/* finally, adjust the packet header length */
1220 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1221 
1222 	return (0);
1223 #undef JUMBOOPTLEN
1224 }
1225 
1226 /*
1227  * Insert fragment header and copy unfragmentable header portions.
1228  */
1229 static int
1230 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1231     struct ip6_frag **frghdrp)
1232 {
1233 	struct mbuf *n, *mlast;
1234 
1235 	if (hlen > sizeof(struct ip6_hdr)) {
1236 		n = m_copym(m0, sizeof(struct ip6_hdr),
1237 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1238 		if (n == NULL)
1239 			return (ENOBUFS);
1240 		m->m_next = n;
1241 	} else
1242 		n = m;
1243 
1244 	/* Search for the last mbuf of unfragmentable part. */
1245 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1246 		;
1247 
1248 	if (M_WRITABLE(mlast) &&
1249 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1250 		/* use the trailing space of the last mbuf for the fragment hdr */
1251 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1252 		    mlast->m_len);
1253 		mlast->m_len += sizeof(struct ip6_frag);
1254 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1255 	} else {
1256 		/* allocate a new mbuf for the fragment header */
1257 		struct mbuf *mfrg;
1258 
1259 		mfrg = m_get(M_NOWAIT, MT_DATA);
1260 		if (mfrg == NULL)
1261 			return (ENOBUFS);
1262 		mfrg->m_len = sizeof(struct ip6_frag);
1263 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1264 		mlast->m_next = mfrg;
1265 	}
1266 
1267 	return (0);
1268 }
1269 
1270 /*
1271  * Calculates IPv6 path mtu for destination @dst.
1272  * Resulting MTU is stored in @mtup.
1273  *
1274  * Returns 0 on success.
1275  */
1276 static int
1277 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1278 {
1279 	struct nhop6_extended nh6;
1280 	struct in6_addr kdst;
1281 	uint32_t scopeid;
1282 	struct ifnet *ifp;
1283 	u_long mtu;
1284 	int error;
1285 
1286 	in6_splitscope(dst, &kdst, &scopeid);
1287 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1288 		return (EHOSTUNREACH);
1289 
1290 	ifp = nh6.nh_ifp;
1291 	mtu = nh6.nh_mtu;
1292 
1293 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1294 	fib6_free_nh_ext(fibnum, &nh6);
1295 
1296 	return (error);
1297 }
1298 
1299 /*
1300  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1301  * and cached data in @ro_pmtu.
1302  * MTU from (successful) route lookup is saved (along with dst)
1303  * inside @ro_pmtu to avoid subsequent route lookups after packet
1304  * filter processing.
1305  *
1306  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1307  * Returns 0 on success.
1308  */
1309 static int
1310 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1311     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1312     int *alwaysfragp, u_int fibnum, u_int proto)
1313 {
1314 	struct nhop6_basic nh6;
1315 	struct in6_addr kdst;
1316 	uint32_t scopeid;
1317 	struct sockaddr_in6 *sa6_dst;
1318 	u_long mtu;
1319 
1320 	mtu = 0;
1321 	if (do_lookup) {
1322 
1323 		/*
1324 		 * Here ro_pmtu has final destination address, while
1325 		 * ro might represent immediate destination.
1326 		 * Use ro_pmtu destination since mtu might differ.
1327 		 */
1328 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1329 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1330 			ro_pmtu->ro_mtu = 0;
1331 
1332 		if (ro_pmtu->ro_mtu == 0) {
1333 			bzero(sa6_dst, sizeof(*sa6_dst));
1334 			sa6_dst->sin6_family = AF_INET6;
1335 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1336 			sa6_dst->sin6_addr = *dst;
1337 
1338 			in6_splitscope(dst, &kdst, &scopeid);
1339 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1340 			    &nh6) == 0)
1341 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1342 		}
1343 
1344 		mtu = ro_pmtu->ro_mtu;
1345 	}
1346 
1347 	if (ro_pmtu->ro_rt)
1348 		mtu = ro_pmtu->ro_rt->rt_mtu;
1349 
1350 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1351 }
1352 
1353 /*
1354  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1355  * hostcache data for @dst.
1356  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1357  *
1358  * Returns 0 on success.
1359  */
1360 static int
1361 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1362     u_long *mtup, int *alwaysfragp, u_int proto)
1363 {
1364 	u_long mtu = 0;
1365 	int alwaysfrag = 0;
1366 	int error = 0;
1367 
1368 	if (rt_mtu > 0) {
1369 		u_int32_t ifmtu;
1370 		struct in_conninfo inc;
1371 
1372 		bzero(&inc, sizeof(inc));
1373 		inc.inc_flags |= INC_ISIPV6;
1374 		inc.inc6_faddr = *dst;
1375 
1376 		ifmtu = IN6_LINKMTU(ifp);
1377 
1378 		/* TCP is known to react to pmtu changes so skip hc */
1379 		if (proto != IPPROTO_TCP)
1380 			mtu = tcp_hc_getmtu(&inc);
1381 
1382 		if (mtu)
1383 			mtu = min(mtu, rt_mtu);
1384 		else
1385 			mtu = rt_mtu;
1386 		if (mtu == 0)
1387 			mtu = ifmtu;
1388 		else if (mtu < IPV6_MMTU) {
1389 			/*
1390 			 * RFC2460 section 5, last paragraph:
1391 			 * if we record ICMPv6 too big message with
1392 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1393 			 * or smaller, with framgent header attached.
1394 			 * (fragment header is needed regardless from the
1395 			 * packet size, for translators to identify packets)
1396 			 */
1397 			alwaysfrag = 1;
1398 			mtu = IPV6_MMTU;
1399 		}
1400 	} else if (ifp) {
1401 		mtu = IN6_LINKMTU(ifp);
1402 	} else
1403 		error = EHOSTUNREACH; /* XXX */
1404 
1405 	*mtup = mtu;
1406 	if (alwaysfragp)
1407 		*alwaysfragp = alwaysfrag;
1408 	return (error);
1409 }
1410 
1411 /*
1412  * IP6 socket option processing.
1413  */
1414 int
1415 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1416 {
1417 	int optdatalen, uproto;
1418 	void *optdata;
1419 	struct inpcb *in6p = sotoinpcb(so);
1420 	int error, optval;
1421 	int level, op, optname;
1422 	int optlen;
1423 	struct thread *td;
1424 #ifdef	RSS
1425 	uint32_t rss_bucket;
1426 	int retval;
1427 #endif
1428 
1429 /*
1430  * Don't use more than a quarter of mbuf clusters.  N.B.:
1431  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1432  * on LP64 architectures, so cast to u_long to avoid undefined
1433  * behavior.  ILP32 architectures cannot have nmbclusters
1434  * large enough to overflow for other reasons.
1435  */
1436 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1437 
1438 	level = sopt->sopt_level;
1439 	op = sopt->sopt_dir;
1440 	optname = sopt->sopt_name;
1441 	optlen = sopt->sopt_valsize;
1442 	td = sopt->sopt_td;
1443 	error = 0;
1444 	optval = 0;
1445 	uproto = (int)so->so_proto->pr_protocol;
1446 
1447 	if (level != IPPROTO_IPV6) {
1448 		error = EINVAL;
1449 
1450 		if (sopt->sopt_level == SOL_SOCKET &&
1451 		    sopt->sopt_dir == SOPT_SET) {
1452 			switch (sopt->sopt_name) {
1453 			case SO_REUSEADDR:
1454 				INP_WLOCK(in6p);
1455 				if ((so->so_options & SO_REUSEADDR) != 0)
1456 					in6p->inp_flags2 |= INP_REUSEADDR;
1457 				else
1458 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1459 				INP_WUNLOCK(in6p);
1460 				error = 0;
1461 				break;
1462 			case SO_REUSEPORT:
1463 				INP_WLOCK(in6p);
1464 				if ((so->so_options & SO_REUSEPORT) != 0)
1465 					in6p->inp_flags2 |= INP_REUSEPORT;
1466 				else
1467 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1468 				INP_WUNLOCK(in6p);
1469 				error = 0;
1470 				break;
1471 			case SO_REUSEPORT_LB:
1472 				INP_WLOCK(in6p);
1473 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1474 					in6p->inp_flags2 |= INP_REUSEPORT_LB;
1475 				else
1476 					in6p->inp_flags2 &= ~INP_REUSEPORT_LB;
1477 				INP_WUNLOCK(in6p);
1478 				error = 0;
1479 				break;
1480 			case SO_SETFIB:
1481 				INP_WLOCK(in6p);
1482 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1483 				INP_WUNLOCK(in6p);
1484 				error = 0;
1485 				break;
1486 			case SO_MAX_PACING_RATE:
1487 #ifdef RATELIMIT
1488 				INP_WLOCK(in6p);
1489 				in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1490 				INP_WUNLOCK(in6p);
1491 				error = 0;
1492 #else
1493 				error = EOPNOTSUPP;
1494 #endif
1495 				break;
1496 			default:
1497 				break;
1498 			}
1499 		}
1500 	} else {		/* level == IPPROTO_IPV6 */
1501 		switch (op) {
1502 
1503 		case SOPT_SET:
1504 			switch (optname) {
1505 			case IPV6_2292PKTOPTIONS:
1506 #ifdef IPV6_PKTOPTIONS
1507 			case IPV6_PKTOPTIONS:
1508 #endif
1509 			{
1510 				struct mbuf *m;
1511 
1512 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1513 					printf("ip6_ctloutput: mbuf limit hit\n");
1514 					error = ENOBUFS;
1515 					break;
1516 				}
1517 
1518 				error = soopt_getm(sopt, &m); /* XXX */
1519 				if (error != 0)
1520 					break;
1521 				error = soopt_mcopyin(sopt, m); /* XXX */
1522 				if (error != 0)
1523 					break;
1524 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1525 						    m, so, sopt);
1526 				m_freem(m); /* XXX */
1527 				break;
1528 			}
1529 
1530 			/*
1531 			 * Use of some Hop-by-Hop options or some
1532 			 * Destination options, might require special
1533 			 * privilege.  That is, normal applications
1534 			 * (without special privilege) might be forbidden
1535 			 * from setting certain options in outgoing packets,
1536 			 * and might never see certain options in received
1537 			 * packets. [RFC 2292 Section 6]
1538 			 * KAME specific note:
1539 			 *  KAME prevents non-privileged users from sending or
1540 			 *  receiving ANY hbh/dst options in order to avoid
1541 			 *  overhead of parsing options in the kernel.
1542 			 */
1543 			case IPV6_RECVHOPOPTS:
1544 			case IPV6_RECVDSTOPTS:
1545 			case IPV6_RECVRTHDRDSTOPTS:
1546 				if (td != NULL) {
1547 					error = priv_check(td,
1548 					    PRIV_NETINET_SETHDROPTS);
1549 					if (error)
1550 						break;
1551 				}
1552 				/* FALLTHROUGH */
1553 			case IPV6_UNICAST_HOPS:
1554 			case IPV6_HOPLIMIT:
1555 
1556 			case IPV6_RECVPKTINFO:
1557 			case IPV6_RECVHOPLIMIT:
1558 			case IPV6_RECVRTHDR:
1559 			case IPV6_RECVPATHMTU:
1560 			case IPV6_RECVTCLASS:
1561 			case IPV6_RECVFLOWID:
1562 #ifdef	RSS
1563 			case IPV6_RECVRSSBUCKETID:
1564 #endif
1565 			case IPV6_V6ONLY:
1566 			case IPV6_AUTOFLOWLABEL:
1567 			case IPV6_ORIGDSTADDR:
1568 			case IPV6_BINDANY:
1569 			case IPV6_BINDMULTI:
1570 #ifdef	RSS
1571 			case IPV6_RSS_LISTEN_BUCKET:
1572 #endif
1573 				if (optname == IPV6_BINDANY && td != NULL) {
1574 					error = priv_check(td,
1575 					    PRIV_NETINET_BINDANY);
1576 					if (error)
1577 						break;
1578 				}
1579 
1580 				if (optlen != sizeof(int)) {
1581 					error = EINVAL;
1582 					break;
1583 				}
1584 				error = sooptcopyin(sopt, &optval,
1585 					sizeof optval, sizeof optval);
1586 				if (error)
1587 					break;
1588 				switch (optname) {
1589 
1590 				case IPV6_UNICAST_HOPS:
1591 					if (optval < -1 || optval >= 256)
1592 						error = EINVAL;
1593 					else {
1594 						/* -1 = kernel default */
1595 						in6p->in6p_hops = optval;
1596 						if ((in6p->inp_vflag &
1597 						     INP_IPV4) != 0)
1598 							in6p->inp_ip_ttl = optval;
1599 					}
1600 					break;
1601 #define OPTSET(bit) \
1602 do { \
1603 	INP_WLOCK(in6p); \
1604 	if (optval) \
1605 		in6p->inp_flags |= (bit); \
1606 	else \
1607 		in6p->inp_flags &= ~(bit); \
1608 	INP_WUNLOCK(in6p); \
1609 } while (/*CONSTCOND*/ 0)
1610 #define OPTSET2292(bit) \
1611 do { \
1612 	INP_WLOCK(in6p); \
1613 	in6p->inp_flags |= IN6P_RFC2292; \
1614 	if (optval) \
1615 		in6p->inp_flags |= (bit); \
1616 	else \
1617 		in6p->inp_flags &= ~(bit); \
1618 	INP_WUNLOCK(in6p); \
1619 } while (/*CONSTCOND*/ 0)
1620 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1621 
1622 #define OPTSET2_N(bit, val) do {					\
1623 	if (val)							\
1624 		in6p->inp_flags2 |= bit;				\
1625 	else								\
1626 		in6p->inp_flags2 &= ~bit;				\
1627 } while (0)
1628 #define OPTSET2(bit, val) do {						\
1629 	INP_WLOCK(in6p);						\
1630 	OPTSET2_N(bit, val);						\
1631 	INP_WUNLOCK(in6p);						\
1632 } while (0)
1633 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1634 #define OPTSET2292_EXCLUSIVE(bit)					\
1635 do {									\
1636 	INP_WLOCK(in6p);						\
1637 	if (OPTBIT(IN6P_RFC2292)) {					\
1638 		error = EINVAL;						\
1639 	} else {							\
1640 		if (optval)						\
1641 			in6p->inp_flags |= (bit);			\
1642 		else							\
1643 			in6p->inp_flags &= ~(bit);			\
1644 	}								\
1645 	INP_WUNLOCK(in6p);						\
1646 } while (/*CONSTCOND*/ 0)
1647 
1648 				case IPV6_RECVPKTINFO:
1649 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1650 					break;
1651 
1652 				case IPV6_HOPLIMIT:
1653 				{
1654 					struct ip6_pktopts **optp;
1655 
1656 					/* cannot mix with RFC2292 */
1657 					if (OPTBIT(IN6P_RFC2292)) {
1658 						error = EINVAL;
1659 						break;
1660 					}
1661 					INP_WLOCK(in6p);
1662 					if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1663 						INP_WUNLOCK(in6p);
1664 						return (ECONNRESET);
1665 					}
1666 					optp = &in6p->in6p_outputopts;
1667 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1668 					    (u_char *)&optval, sizeof(optval),
1669 					    optp, (td != NULL) ? td->td_ucred :
1670 					    NULL, uproto);
1671 					INP_WUNLOCK(in6p);
1672 					break;
1673 				}
1674 
1675 				case IPV6_RECVHOPLIMIT:
1676 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1677 					break;
1678 
1679 				case IPV6_RECVHOPOPTS:
1680 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1681 					break;
1682 
1683 				case IPV6_RECVDSTOPTS:
1684 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1685 					break;
1686 
1687 				case IPV6_RECVRTHDRDSTOPTS:
1688 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1689 					break;
1690 
1691 				case IPV6_RECVRTHDR:
1692 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1693 					break;
1694 
1695 				case IPV6_RECVPATHMTU:
1696 					/*
1697 					 * We ignore this option for TCP
1698 					 * sockets.
1699 					 * (RFC3542 leaves this case
1700 					 * unspecified.)
1701 					 */
1702 					if (uproto != IPPROTO_TCP)
1703 						OPTSET(IN6P_MTU);
1704 					break;
1705 
1706 				case IPV6_RECVFLOWID:
1707 					OPTSET2(INP_RECVFLOWID, optval);
1708 					break;
1709 
1710 #ifdef	RSS
1711 				case IPV6_RECVRSSBUCKETID:
1712 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1713 					break;
1714 #endif
1715 
1716 				case IPV6_V6ONLY:
1717 					/*
1718 					 * make setsockopt(IPV6_V6ONLY)
1719 					 * available only prior to bind(2).
1720 					 * see ipng mailing list, Jun 22 2001.
1721 					 */
1722 					if (in6p->inp_lport ||
1723 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1724 						error = EINVAL;
1725 						break;
1726 					}
1727 					OPTSET(IN6P_IPV6_V6ONLY);
1728 					if (optval)
1729 						in6p->inp_vflag &= ~INP_IPV4;
1730 					else
1731 						in6p->inp_vflag |= INP_IPV4;
1732 					break;
1733 				case IPV6_RECVTCLASS:
1734 					/* cannot mix with RFC2292 XXX */
1735 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1736 					break;
1737 				case IPV6_AUTOFLOWLABEL:
1738 					OPTSET(IN6P_AUTOFLOWLABEL);
1739 					break;
1740 
1741 				case IPV6_ORIGDSTADDR:
1742 					OPTSET2(INP_ORIGDSTADDR, optval);
1743 					break;
1744 				case IPV6_BINDANY:
1745 					OPTSET(INP_BINDANY);
1746 					break;
1747 
1748 				case IPV6_BINDMULTI:
1749 					OPTSET2(INP_BINDMULTI, optval);
1750 					break;
1751 #ifdef	RSS
1752 				case IPV6_RSS_LISTEN_BUCKET:
1753 					if ((optval >= 0) &&
1754 					    (optval < rss_getnumbuckets())) {
1755 						INP_WLOCK(in6p);
1756 						in6p->inp_rss_listen_bucket = optval;
1757 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1758 						INP_WUNLOCK(in6p);
1759 					} else {
1760 						error = EINVAL;
1761 					}
1762 					break;
1763 #endif
1764 				}
1765 				break;
1766 
1767 			case IPV6_TCLASS:
1768 			case IPV6_DONTFRAG:
1769 			case IPV6_USE_MIN_MTU:
1770 			case IPV6_PREFER_TEMPADDR:
1771 				if (optlen != sizeof(optval)) {
1772 					error = EINVAL;
1773 					break;
1774 				}
1775 				error = sooptcopyin(sopt, &optval,
1776 					sizeof optval, sizeof optval);
1777 				if (error)
1778 					break;
1779 				{
1780 					struct ip6_pktopts **optp;
1781 					INP_WLOCK(in6p);
1782 					if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1783 						INP_WUNLOCK(in6p);
1784 						return (ECONNRESET);
1785 					}
1786 					optp = &in6p->in6p_outputopts;
1787 					error = ip6_pcbopt(optname,
1788 					    (u_char *)&optval, sizeof(optval),
1789 					    optp, (td != NULL) ? td->td_ucred :
1790 					    NULL, uproto);
1791 					INP_WUNLOCK(in6p);
1792 					break;
1793 				}
1794 
1795 			case IPV6_2292PKTINFO:
1796 			case IPV6_2292HOPLIMIT:
1797 			case IPV6_2292HOPOPTS:
1798 			case IPV6_2292DSTOPTS:
1799 			case IPV6_2292RTHDR:
1800 				/* RFC 2292 */
1801 				if (optlen != sizeof(int)) {
1802 					error = EINVAL;
1803 					break;
1804 				}
1805 				error = sooptcopyin(sopt, &optval,
1806 					sizeof optval, sizeof optval);
1807 				if (error)
1808 					break;
1809 				switch (optname) {
1810 				case IPV6_2292PKTINFO:
1811 					OPTSET2292(IN6P_PKTINFO);
1812 					break;
1813 				case IPV6_2292HOPLIMIT:
1814 					OPTSET2292(IN6P_HOPLIMIT);
1815 					break;
1816 				case IPV6_2292HOPOPTS:
1817 					/*
1818 					 * Check super-user privilege.
1819 					 * See comments for IPV6_RECVHOPOPTS.
1820 					 */
1821 					if (td != NULL) {
1822 						error = priv_check(td,
1823 						    PRIV_NETINET_SETHDROPTS);
1824 						if (error)
1825 							return (error);
1826 					}
1827 					OPTSET2292(IN6P_HOPOPTS);
1828 					break;
1829 				case IPV6_2292DSTOPTS:
1830 					if (td != NULL) {
1831 						error = priv_check(td,
1832 						    PRIV_NETINET_SETHDROPTS);
1833 						if (error)
1834 							return (error);
1835 					}
1836 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1837 					break;
1838 				case IPV6_2292RTHDR:
1839 					OPTSET2292(IN6P_RTHDR);
1840 					break;
1841 				}
1842 				break;
1843 			case IPV6_PKTINFO:
1844 			case IPV6_HOPOPTS:
1845 			case IPV6_RTHDR:
1846 			case IPV6_DSTOPTS:
1847 			case IPV6_RTHDRDSTOPTS:
1848 			case IPV6_NEXTHOP:
1849 			{
1850 				/* new advanced API (RFC3542) */
1851 				u_char *optbuf;
1852 				u_char optbuf_storage[MCLBYTES];
1853 				int optlen;
1854 				struct ip6_pktopts **optp;
1855 
1856 				/* cannot mix with RFC2292 */
1857 				if (OPTBIT(IN6P_RFC2292)) {
1858 					error = EINVAL;
1859 					break;
1860 				}
1861 
1862 				/*
1863 				 * We only ensure valsize is not too large
1864 				 * here.  Further validation will be done
1865 				 * later.
1866 				 */
1867 				error = sooptcopyin(sopt, optbuf_storage,
1868 				    sizeof(optbuf_storage), 0);
1869 				if (error)
1870 					break;
1871 				optlen = sopt->sopt_valsize;
1872 				optbuf = optbuf_storage;
1873 				INP_WLOCK(in6p);
1874 				if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1875 					INP_WUNLOCK(in6p);
1876 					return (ECONNRESET);
1877 				}
1878 				optp = &in6p->in6p_outputopts;
1879 				error = ip6_pcbopt(optname, optbuf, optlen,
1880 				    optp, (td != NULL) ? td->td_ucred : NULL,
1881 				    uproto);
1882 				INP_WUNLOCK(in6p);
1883 				break;
1884 			}
1885 #undef OPTSET
1886 
1887 			case IPV6_MULTICAST_IF:
1888 			case IPV6_MULTICAST_HOPS:
1889 			case IPV6_MULTICAST_LOOP:
1890 			case IPV6_JOIN_GROUP:
1891 			case IPV6_LEAVE_GROUP:
1892 			case IPV6_MSFILTER:
1893 			case MCAST_BLOCK_SOURCE:
1894 			case MCAST_UNBLOCK_SOURCE:
1895 			case MCAST_JOIN_GROUP:
1896 			case MCAST_LEAVE_GROUP:
1897 			case MCAST_JOIN_SOURCE_GROUP:
1898 			case MCAST_LEAVE_SOURCE_GROUP:
1899 				error = ip6_setmoptions(in6p, sopt);
1900 				break;
1901 
1902 			case IPV6_PORTRANGE:
1903 				error = sooptcopyin(sopt, &optval,
1904 				    sizeof optval, sizeof optval);
1905 				if (error)
1906 					break;
1907 
1908 				INP_WLOCK(in6p);
1909 				switch (optval) {
1910 				case IPV6_PORTRANGE_DEFAULT:
1911 					in6p->inp_flags &= ~(INP_LOWPORT);
1912 					in6p->inp_flags &= ~(INP_HIGHPORT);
1913 					break;
1914 
1915 				case IPV6_PORTRANGE_HIGH:
1916 					in6p->inp_flags &= ~(INP_LOWPORT);
1917 					in6p->inp_flags |= INP_HIGHPORT;
1918 					break;
1919 
1920 				case IPV6_PORTRANGE_LOW:
1921 					in6p->inp_flags &= ~(INP_HIGHPORT);
1922 					in6p->inp_flags |= INP_LOWPORT;
1923 					break;
1924 
1925 				default:
1926 					error = EINVAL;
1927 					break;
1928 				}
1929 				INP_WUNLOCK(in6p);
1930 				break;
1931 
1932 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1933 			case IPV6_IPSEC_POLICY:
1934 				if (IPSEC_ENABLED(ipv6)) {
1935 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1936 					break;
1937 				}
1938 				/* FALLTHROUGH */
1939 #endif /* IPSEC */
1940 
1941 			default:
1942 				error = ENOPROTOOPT;
1943 				break;
1944 			}
1945 			break;
1946 
1947 		case SOPT_GET:
1948 			switch (optname) {
1949 
1950 			case IPV6_2292PKTOPTIONS:
1951 #ifdef IPV6_PKTOPTIONS
1952 			case IPV6_PKTOPTIONS:
1953 #endif
1954 				/*
1955 				 * RFC3542 (effectively) deprecated the
1956 				 * semantics of the 2292-style pktoptions.
1957 				 * Since it was not reliable in nature (i.e.,
1958 				 * applications had to expect the lack of some
1959 				 * information after all), it would make sense
1960 				 * to simplify this part by always returning
1961 				 * empty data.
1962 				 */
1963 				sopt->sopt_valsize = 0;
1964 				break;
1965 
1966 			case IPV6_RECVHOPOPTS:
1967 			case IPV6_RECVDSTOPTS:
1968 			case IPV6_RECVRTHDRDSTOPTS:
1969 			case IPV6_UNICAST_HOPS:
1970 			case IPV6_RECVPKTINFO:
1971 			case IPV6_RECVHOPLIMIT:
1972 			case IPV6_RECVRTHDR:
1973 			case IPV6_RECVPATHMTU:
1974 
1975 			case IPV6_V6ONLY:
1976 			case IPV6_PORTRANGE:
1977 			case IPV6_RECVTCLASS:
1978 			case IPV6_AUTOFLOWLABEL:
1979 			case IPV6_BINDANY:
1980 			case IPV6_FLOWID:
1981 			case IPV6_FLOWTYPE:
1982 			case IPV6_RECVFLOWID:
1983 #ifdef	RSS
1984 			case IPV6_RSSBUCKETID:
1985 			case IPV6_RECVRSSBUCKETID:
1986 #endif
1987 			case IPV6_BINDMULTI:
1988 				switch (optname) {
1989 
1990 				case IPV6_RECVHOPOPTS:
1991 					optval = OPTBIT(IN6P_HOPOPTS);
1992 					break;
1993 
1994 				case IPV6_RECVDSTOPTS:
1995 					optval = OPTBIT(IN6P_DSTOPTS);
1996 					break;
1997 
1998 				case IPV6_RECVRTHDRDSTOPTS:
1999 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2000 					break;
2001 
2002 				case IPV6_UNICAST_HOPS:
2003 					optval = in6p->in6p_hops;
2004 					break;
2005 
2006 				case IPV6_RECVPKTINFO:
2007 					optval = OPTBIT(IN6P_PKTINFO);
2008 					break;
2009 
2010 				case IPV6_RECVHOPLIMIT:
2011 					optval = OPTBIT(IN6P_HOPLIMIT);
2012 					break;
2013 
2014 				case IPV6_RECVRTHDR:
2015 					optval = OPTBIT(IN6P_RTHDR);
2016 					break;
2017 
2018 				case IPV6_RECVPATHMTU:
2019 					optval = OPTBIT(IN6P_MTU);
2020 					break;
2021 
2022 				case IPV6_V6ONLY:
2023 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2024 					break;
2025 
2026 				case IPV6_PORTRANGE:
2027 				    {
2028 					int flags;
2029 					flags = in6p->inp_flags;
2030 					if (flags & INP_HIGHPORT)
2031 						optval = IPV6_PORTRANGE_HIGH;
2032 					else if (flags & INP_LOWPORT)
2033 						optval = IPV6_PORTRANGE_LOW;
2034 					else
2035 						optval = 0;
2036 					break;
2037 				    }
2038 				case IPV6_RECVTCLASS:
2039 					optval = OPTBIT(IN6P_TCLASS);
2040 					break;
2041 
2042 				case IPV6_AUTOFLOWLABEL:
2043 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2044 					break;
2045 
2046 				case IPV6_ORIGDSTADDR:
2047 					optval = OPTBIT2(INP_ORIGDSTADDR);
2048 					break;
2049 
2050 				case IPV6_BINDANY:
2051 					optval = OPTBIT(INP_BINDANY);
2052 					break;
2053 
2054 				case IPV6_FLOWID:
2055 					optval = in6p->inp_flowid;
2056 					break;
2057 
2058 				case IPV6_FLOWTYPE:
2059 					optval = in6p->inp_flowtype;
2060 					break;
2061 
2062 				case IPV6_RECVFLOWID:
2063 					optval = OPTBIT2(INP_RECVFLOWID);
2064 					break;
2065 #ifdef	RSS
2066 				case IPV6_RSSBUCKETID:
2067 					retval =
2068 					    rss_hash2bucket(in6p->inp_flowid,
2069 					    in6p->inp_flowtype,
2070 					    &rss_bucket);
2071 					if (retval == 0)
2072 						optval = rss_bucket;
2073 					else
2074 						error = EINVAL;
2075 					break;
2076 
2077 				case IPV6_RECVRSSBUCKETID:
2078 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2079 					break;
2080 #endif
2081 
2082 				case IPV6_BINDMULTI:
2083 					optval = OPTBIT2(INP_BINDMULTI);
2084 					break;
2085 
2086 				}
2087 				if (error)
2088 					break;
2089 				error = sooptcopyout(sopt, &optval,
2090 					sizeof optval);
2091 				break;
2092 
2093 			case IPV6_PATHMTU:
2094 			{
2095 				u_long pmtu = 0;
2096 				struct ip6_mtuinfo mtuinfo;
2097 				struct in6_addr addr;
2098 
2099 				if (!(so->so_state & SS_ISCONNECTED))
2100 					return (ENOTCONN);
2101 				/*
2102 				 * XXX: we dot not consider the case of source
2103 				 * routing, or optional information to specify
2104 				 * the outgoing interface.
2105 				 * Copy faddr out of in6p to avoid holding lock
2106 				 * on inp during route lookup.
2107 				 */
2108 				INP_RLOCK(in6p);
2109 				bcopy(&in6p->in6p_faddr, &addr, sizeof(addr));
2110 				INP_RUNLOCK(in6p);
2111 				error = ip6_getpmtu_ctl(so->so_fibnum,
2112 				    &addr, &pmtu);
2113 				if (error)
2114 					break;
2115 				if (pmtu > IPV6_MAXPACKET)
2116 					pmtu = IPV6_MAXPACKET;
2117 
2118 				bzero(&mtuinfo, sizeof(mtuinfo));
2119 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2120 				optdata = (void *)&mtuinfo;
2121 				optdatalen = sizeof(mtuinfo);
2122 				error = sooptcopyout(sopt, optdata,
2123 				    optdatalen);
2124 				break;
2125 			}
2126 
2127 			case IPV6_2292PKTINFO:
2128 			case IPV6_2292HOPLIMIT:
2129 			case IPV6_2292HOPOPTS:
2130 			case IPV6_2292RTHDR:
2131 			case IPV6_2292DSTOPTS:
2132 				switch (optname) {
2133 				case IPV6_2292PKTINFO:
2134 					optval = OPTBIT(IN6P_PKTINFO);
2135 					break;
2136 				case IPV6_2292HOPLIMIT:
2137 					optval = OPTBIT(IN6P_HOPLIMIT);
2138 					break;
2139 				case IPV6_2292HOPOPTS:
2140 					optval = OPTBIT(IN6P_HOPOPTS);
2141 					break;
2142 				case IPV6_2292RTHDR:
2143 					optval = OPTBIT(IN6P_RTHDR);
2144 					break;
2145 				case IPV6_2292DSTOPTS:
2146 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2147 					break;
2148 				}
2149 				error = sooptcopyout(sopt, &optval,
2150 				    sizeof optval);
2151 				break;
2152 			case IPV6_PKTINFO:
2153 			case IPV6_HOPOPTS:
2154 			case IPV6_RTHDR:
2155 			case IPV6_DSTOPTS:
2156 			case IPV6_RTHDRDSTOPTS:
2157 			case IPV6_NEXTHOP:
2158 			case IPV6_TCLASS:
2159 			case IPV6_DONTFRAG:
2160 			case IPV6_USE_MIN_MTU:
2161 			case IPV6_PREFER_TEMPADDR:
2162 				error = ip6_getpcbopt(in6p, optname, sopt);
2163 				break;
2164 
2165 			case IPV6_MULTICAST_IF:
2166 			case IPV6_MULTICAST_HOPS:
2167 			case IPV6_MULTICAST_LOOP:
2168 			case IPV6_MSFILTER:
2169 				error = ip6_getmoptions(in6p, sopt);
2170 				break;
2171 
2172 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2173 			case IPV6_IPSEC_POLICY:
2174 				if (IPSEC_ENABLED(ipv6)) {
2175 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2176 					break;
2177 				}
2178 				/* FALLTHROUGH */
2179 #endif /* IPSEC */
2180 			default:
2181 				error = ENOPROTOOPT;
2182 				break;
2183 			}
2184 			break;
2185 		}
2186 	}
2187 	return (error);
2188 }
2189 
2190 int
2191 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2192 {
2193 	int error = 0, optval, optlen;
2194 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2195 	struct inpcb *in6p = sotoinpcb(so);
2196 	int level, op, optname;
2197 
2198 	level = sopt->sopt_level;
2199 	op = sopt->sopt_dir;
2200 	optname = sopt->sopt_name;
2201 	optlen = sopt->sopt_valsize;
2202 
2203 	if (level != IPPROTO_IPV6) {
2204 		return (EINVAL);
2205 	}
2206 
2207 	switch (optname) {
2208 	case IPV6_CHECKSUM:
2209 		/*
2210 		 * For ICMPv6 sockets, no modification allowed for checksum
2211 		 * offset, permit "no change" values to help existing apps.
2212 		 *
2213 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2214 		 * for an ICMPv6 socket will fail."
2215 		 * The current behavior does not meet RFC3542.
2216 		 */
2217 		switch (op) {
2218 		case SOPT_SET:
2219 			if (optlen != sizeof(int)) {
2220 				error = EINVAL;
2221 				break;
2222 			}
2223 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2224 					    sizeof(optval));
2225 			if (error)
2226 				break;
2227 			if (optval < -1 || (optval % 2) != 0) {
2228 				/*
2229 				 * The API assumes non-negative even offset
2230 				 * values or -1 as a special value.
2231 				 */
2232 				error = EINVAL;
2233 			} else if (so->so_proto->pr_protocol ==
2234 			    IPPROTO_ICMPV6) {
2235 				if (optval != icmp6off)
2236 					error = EINVAL;
2237 			} else
2238 				in6p->in6p_cksum = optval;
2239 			break;
2240 
2241 		case SOPT_GET:
2242 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2243 				optval = icmp6off;
2244 			else
2245 				optval = in6p->in6p_cksum;
2246 
2247 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2248 			break;
2249 
2250 		default:
2251 			error = EINVAL;
2252 			break;
2253 		}
2254 		break;
2255 
2256 	default:
2257 		error = ENOPROTOOPT;
2258 		break;
2259 	}
2260 
2261 	return (error);
2262 }
2263 
2264 /*
2265  * Set up IP6 options in pcb for insertion in output packets or
2266  * specifying behavior of outgoing packets.
2267  */
2268 static int
2269 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2270     struct socket *so, struct sockopt *sopt)
2271 {
2272 	struct ip6_pktopts *opt = *pktopt;
2273 	int error = 0;
2274 	struct thread *td = sopt->sopt_td;
2275 
2276 	/* turn off any old options. */
2277 	if (opt) {
2278 #ifdef DIAGNOSTIC
2279 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2280 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2281 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2282 			printf("ip6_pcbopts: all specified options are cleared.\n");
2283 #endif
2284 		ip6_clearpktopts(opt, -1);
2285 	} else
2286 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2287 	*pktopt = NULL;
2288 
2289 	if (!m || m->m_len == 0) {
2290 		/*
2291 		 * Only turning off any previous options, regardless of
2292 		 * whether the opt is just created or given.
2293 		 */
2294 		free(opt, M_IP6OPT);
2295 		return (0);
2296 	}
2297 
2298 	/*  set options specified by user. */
2299 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2300 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2301 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2302 		free(opt, M_IP6OPT);
2303 		return (error);
2304 	}
2305 	*pktopt = opt;
2306 	return (0);
2307 }
2308 
2309 /*
2310  * initialize ip6_pktopts.  beware that there are non-zero default values in
2311  * the struct.
2312  */
2313 void
2314 ip6_initpktopts(struct ip6_pktopts *opt)
2315 {
2316 
2317 	bzero(opt, sizeof(*opt));
2318 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2319 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2320 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2321 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2322 }
2323 
2324 static int
2325 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2326     struct ucred *cred, int uproto)
2327 {
2328 	struct ip6_pktopts *opt;
2329 
2330 	if (*pktopt == NULL) {
2331 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2332 		    M_NOWAIT);
2333 		if (*pktopt == NULL)
2334 			return (ENOBUFS);
2335 		ip6_initpktopts(*pktopt);
2336 	}
2337 	opt = *pktopt;
2338 
2339 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2340 }
2341 
2342 #define GET_PKTOPT_VAR(field, lenexpr) do {					\
2343 	if (pktopt && pktopt->field) {						\
2344 		INP_RUNLOCK(in6p);						\
2345 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);		\
2346 		malloc_optdata = true;						\
2347 		INP_RLOCK(in6p);						\
2348 		if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
2349 			INP_RUNLOCK(in6p);					\
2350 			free(optdata, M_TEMP);					\
2351 			return (ECONNRESET);					\
2352 		}								\
2353 		pktopt = in6p->in6p_outputopts;					\
2354 		if (pktopt && pktopt->field) {					\
2355 			optdatalen = min(lenexpr, sopt->sopt_valsize);		\
2356 			bcopy(&pktopt->field, optdata, optdatalen);		\
2357 		} else {							\
2358 			free(optdata, M_TEMP);					\
2359 			optdata = NULL;						\
2360 			malloc_optdata = false;					\
2361 		}								\
2362 	}									\
2363 } while(0)
2364 
2365 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,				\
2366 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2367 
2368 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,			\
2369 	pktopt->field->sa_len)
2370 
2371 static int
2372 ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt)
2373 {
2374 	void *optdata = NULL;
2375 	bool malloc_optdata = false;
2376 	int optdatalen = 0;
2377 	int error = 0;
2378 	struct in6_pktinfo null_pktinfo;
2379 	int deftclass = 0, on;
2380 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2381 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2382 	struct ip6_pktopts *pktopt;
2383 
2384 	INP_RLOCK(in6p);
2385 	pktopt = in6p->in6p_outputopts;
2386 
2387 	switch (optname) {
2388 	case IPV6_PKTINFO:
2389 		optdata = (void *)&null_pktinfo;
2390 		if (pktopt && pktopt->ip6po_pktinfo) {
2391 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2392 			    sizeof(null_pktinfo));
2393 			in6_clearscope(&null_pktinfo.ipi6_addr);
2394 		} else {
2395 			/* XXX: we don't have to do this every time... */
2396 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2397 		}
2398 		optdatalen = sizeof(struct in6_pktinfo);
2399 		break;
2400 	case IPV6_TCLASS:
2401 		if (pktopt && pktopt->ip6po_tclass >= 0)
2402 			deftclass = pktopt->ip6po_tclass;
2403 		optdata = (void *)&deftclass;
2404 		optdatalen = sizeof(int);
2405 		break;
2406 	case IPV6_HOPOPTS:
2407 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2408 		break;
2409 	case IPV6_RTHDR:
2410 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2411 		break;
2412 	case IPV6_RTHDRDSTOPTS:
2413 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2414 		break;
2415 	case IPV6_DSTOPTS:
2416 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2417 		break;
2418 	case IPV6_NEXTHOP:
2419 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2420 		break;
2421 	case IPV6_USE_MIN_MTU:
2422 		if (pktopt)
2423 			defminmtu = pktopt->ip6po_minmtu;
2424 		optdata = (void *)&defminmtu;
2425 		optdatalen = sizeof(int);
2426 		break;
2427 	case IPV6_DONTFRAG:
2428 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2429 			on = 1;
2430 		else
2431 			on = 0;
2432 		optdata = (void *)&on;
2433 		optdatalen = sizeof(on);
2434 		break;
2435 	case IPV6_PREFER_TEMPADDR:
2436 		if (pktopt)
2437 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2438 		optdata = (void *)&defpreftemp;
2439 		optdatalen = sizeof(int);
2440 		break;
2441 	default:		/* should not happen */
2442 #ifdef DIAGNOSTIC
2443 		panic("ip6_getpcbopt: unexpected option\n");
2444 #endif
2445 		INP_RUNLOCK(in6p);
2446 		return (ENOPROTOOPT);
2447 	}
2448 	INP_RUNLOCK(in6p);
2449 
2450 	error = sooptcopyout(sopt, optdata, optdatalen);
2451 	if (malloc_optdata)
2452 		free(optdata, M_TEMP);
2453 
2454 	return (error);
2455 }
2456 
2457 void
2458 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2459 {
2460 	if (pktopt == NULL)
2461 		return;
2462 
2463 	if (optname == -1 || optname == IPV6_PKTINFO) {
2464 		if (pktopt->ip6po_pktinfo)
2465 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2466 		pktopt->ip6po_pktinfo = NULL;
2467 	}
2468 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2469 		pktopt->ip6po_hlim = -1;
2470 	if (optname == -1 || optname == IPV6_TCLASS)
2471 		pktopt->ip6po_tclass = -1;
2472 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2473 		if (pktopt->ip6po_nextroute.ro_rt) {
2474 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2475 			pktopt->ip6po_nextroute.ro_rt = NULL;
2476 		}
2477 		if (pktopt->ip6po_nexthop)
2478 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2479 		pktopt->ip6po_nexthop = NULL;
2480 	}
2481 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2482 		if (pktopt->ip6po_hbh)
2483 			free(pktopt->ip6po_hbh, M_IP6OPT);
2484 		pktopt->ip6po_hbh = NULL;
2485 	}
2486 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2487 		if (pktopt->ip6po_dest1)
2488 			free(pktopt->ip6po_dest1, M_IP6OPT);
2489 		pktopt->ip6po_dest1 = NULL;
2490 	}
2491 	if (optname == -1 || optname == IPV6_RTHDR) {
2492 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2493 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2494 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2495 		if (pktopt->ip6po_route.ro_rt) {
2496 			RTFREE(pktopt->ip6po_route.ro_rt);
2497 			pktopt->ip6po_route.ro_rt = NULL;
2498 		}
2499 	}
2500 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2501 		if (pktopt->ip6po_dest2)
2502 			free(pktopt->ip6po_dest2, M_IP6OPT);
2503 		pktopt->ip6po_dest2 = NULL;
2504 	}
2505 }
2506 
2507 #define PKTOPT_EXTHDRCPY(type) \
2508 do {\
2509 	if (src->type) {\
2510 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2511 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2512 		if (dst->type == NULL)\
2513 			goto bad;\
2514 		bcopy(src->type, dst->type, hlen);\
2515 	}\
2516 } while (/*CONSTCOND*/ 0)
2517 
2518 static int
2519 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2520 {
2521 	if (dst == NULL || src == NULL)  {
2522 		printf("ip6_clearpktopts: invalid argument\n");
2523 		return (EINVAL);
2524 	}
2525 
2526 	dst->ip6po_hlim = src->ip6po_hlim;
2527 	dst->ip6po_tclass = src->ip6po_tclass;
2528 	dst->ip6po_flags = src->ip6po_flags;
2529 	dst->ip6po_minmtu = src->ip6po_minmtu;
2530 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2531 	if (src->ip6po_pktinfo) {
2532 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2533 		    M_IP6OPT, canwait);
2534 		if (dst->ip6po_pktinfo == NULL)
2535 			goto bad;
2536 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2537 	}
2538 	if (src->ip6po_nexthop) {
2539 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2540 		    M_IP6OPT, canwait);
2541 		if (dst->ip6po_nexthop == NULL)
2542 			goto bad;
2543 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2544 		    src->ip6po_nexthop->sa_len);
2545 	}
2546 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2547 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2548 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2549 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2550 	return (0);
2551 
2552   bad:
2553 	ip6_clearpktopts(dst, -1);
2554 	return (ENOBUFS);
2555 }
2556 #undef PKTOPT_EXTHDRCPY
2557 
2558 struct ip6_pktopts *
2559 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2560 {
2561 	int error;
2562 	struct ip6_pktopts *dst;
2563 
2564 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2565 	if (dst == NULL)
2566 		return (NULL);
2567 	ip6_initpktopts(dst);
2568 
2569 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2570 		free(dst, M_IP6OPT);
2571 		return (NULL);
2572 	}
2573 
2574 	return (dst);
2575 }
2576 
2577 void
2578 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2579 {
2580 	if (pktopt == NULL)
2581 		return;
2582 
2583 	ip6_clearpktopts(pktopt, -1);
2584 
2585 	free(pktopt, M_IP6OPT);
2586 }
2587 
2588 /*
2589  * Set IPv6 outgoing packet options based on advanced API.
2590  */
2591 int
2592 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2593     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2594 {
2595 	struct cmsghdr *cm = NULL;
2596 
2597 	if (control == NULL || opt == NULL)
2598 		return (EINVAL);
2599 
2600 	ip6_initpktopts(opt);
2601 	if (stickyopt) {
2602 		int error;
2603 
2604 		/*
2605 		 * If stickyopt is provided, make a local copy of the options
2606 		 * for this particular packet, then override them by ancillary
2607 		 * objects.
2608 		 * XXX: copypktopts() does not copy the cached route to a next
2609 		 * hop (if any).  This is not very good in terms of efficiency,
2610 		 * but we can allow this since this option should be rarely
2611 		 * used.
2612 		 */
2613 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2614 			return (error);
2615 	}
2616 
2617 	/*
2618 	 * XXX: Currently, we assume all the optional information is stored
2619 	 * in a single mbuf.
2620 	 */
2621 	if (control->m_next)
2622 		return (EINVAL);
2623 
2624 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2625 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2626 		int error;
2627 
2628 		if (control->m_len < CMSG_LEN(0))
2629 			return (EINVAL);
2630 
2631 		cm = mtod(control, struct cmsghdr *);
2632 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2633 			return (EINVAL);
2634 		if (cm->cmsg_level != IPPROTO_IPV6)
2635 			continue;
2636 
2637 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2638 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2639 		if (error)
2640 			return (error);
2641 	}
2642 
2643 	return (0);
2644 }
2645 
2646 /*
2647  * Set a particular packet option, as a sticky option or an ancillary data
2648  * item.  "len" can be 0 only when it's a sticky option.
2649  * We have 4 cases of combination of "sticky" and "cmsg":
2650  * "sticky=0, cmsg=0": impossible
2651  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2652  * "sticky=1, cmsg=0": RFC3542 socket option
2653  * "sticky=1, cmsg=1": RFC2292 socket option
2654  */
2655 static int
2656 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2657     struct ucred *cred, int sticky, int cmsg, int uproto)
2658 {
2659 	int minmtupolicy, preftemp;
2660 	int error;
2661 
2662 	if (!sticky && !cmsg) {
2663 #ifdef DIAGNOSTIC
2664 		printf("ip6_setpktopt: impossible case\n");
2665 #endif
2666 		return (EINVAL);
2667 	}
2668 
2669 	/*
2670 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2671 	 * not be specified in the context of RFC3542.  Conversely,
2672 	 * RFC3542 types should not be specified in the context of RFC2292.
2673 	 */
2674 	if (!cmsg) {
2675 		switch (optname) {
2676 		case IPV6_2292PKTINFO:
2677 		case IPV6_2292HOPLIMIT:
2678 		case IPV6_2292NEXTHOP:
2679 		case IPV6_2292HOPOPTS:
2680 		case IPV6_2292DSTOPTS:
2681 		case IPV6_2292RTHDR:
2682 		case IPV6_2292PKTOPTIONS:
2683 			return (ENOPROTOOPT);
2684 		}
2685 	}
2686 	if (sticky && cmsg) {
2687 		switch (optname) {
2688 		case IPV6_PKTINFO:
2689 		case IPV6_HOPLIMIT:
2690 		case IPV6_NEXTHOP:
2691 		case IPV6_HOPOPTS:
2692 		case IPV6_DSTOPTS:
2693 		case IPV6_RTHDRDSTOPTS:
2694 		case IPV6_RTHDR:
2695 		case IPV6_USE_MIN_MTU:
2696 		case IPV6_DONTFRAG:
2697 		case IPV6_TCLASS:
2698 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2699 			return (ENOPROTOOPT);
2700 		}
2701 	}
2702 
2703 	switch (optname) {
2704 	case IPV6_2292PKTINFO:
2705 	case IPV6_PKTINFO:
2706 	{
2707 		struct ifnet *ifp = NULL;
2708 		struct in6_pktinfo *pktinfo;
2709 
2710 		if (len != sizeof(struct in6_pktinfo))
2711 			return (EINVAL);
2712 
2713 		pktinfo = (struct in6_pktinfo *)buf;
2714 
2715 		/*
2716 		 * An application can clear any sticky IPV6_PKTINFO option by
2717 		 * doing a "regular" setsockopt with ipi6_addr being
2718 		 * in6addr_any and ipi6_ifindex being zero.
2719 		 * [RFC 3542, Section 6]
2720 		 */
2721 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2722 		    pktinfo->ipi6_ifindex == 0 &&
2723 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2724 			ip6_clearpktopts(opt, optname);
2725 			break;
2726 		}
2727 
2728 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2729 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2730 			return (EINVAL);
2731 		}
2732 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2733 			return (EINVAL);
2734 		/* validate the interface index if specified. */
2735 		if (pktinfo->ipi6_ifindex > V_if_index)
2736 			 return (ENXIO);
2737 		if (pktinfo->ipi6_ifindex) {
2738 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2739 			if (ifp == NULL)
2740 				return (ENXIO);
2741 		}
2742 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2743 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2744 			return (ENETDOWN);
2745 
2746 		if (ifp != NULL &&
2747 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2748 			struct in6_ifaddr *ia;
2749 
2750 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2751 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2752 			if (ia == NULL)
2753 				return (EADDRNOTAVAIL);
2754 			ifa_free(&ia->ia_ifa);
2755 		}
2756 		/*
2757 		 * We store the address anyway, and let in6_selectsrc()
2758 		 * validate the specified address.  This is because ipi6_addr
2759 		 * may not have enough information about its scope zone, and
2760 		 * we may need additional information (such as outgoing
2761 		 * interface or the scope zone of a destination address) to
2762 		 * disambiguate the scope.
2763 		 * XXX: the delay of the validation may confuse the
2764 		 * application when it is used as a sticky option.
2765 		 */
2766 		if (opt->ip6po_pktinfo == NULL) {
2767 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2768 			    M_IP6OPT, M_NOWAIT);
2769 			if (opt->ip6po_pktinfo == NULL)
2770 				return (ENOBUFS);
2771 		}
2772 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2773 		break;
2774 	}
2775 
2776 	case IPV6_2292HOPLIMIT:
2777 	case IPV6_HOPLIMIT:
2778 	{
2779 		int *hlimp;
2780 
2781 		/*
2782 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2783 		 * to simplify the ordering among hoplimit options.
2784 		 */
2785 		if (optname == IPV6_HOPLIMIT && sticky)
2786 			return (ENOPROTOOPT);
2787 
2788 		if (len != sizeof(int))
2789 			return (EINVAL);
2790 		hlimp = (int *)buf;
2791 		if (*hlimp < -1 || *hlimp > 255)
2792 			return (EINVAL);
2793 
2794 		opt->ip6po_hlim = *hlimp;
2795 		break;
2796 	}
2797 
2798 	case IPV6_TCLASS:
2799 	{
2800 		int tclass;
2801 
2802 		if (len != sizeof(int))
2803 			return (EINVAL);
2804 		tclass = *(int *)buf;
2805 		if (tclass < -1 || tclass > 255)
2806 			return (EINVAL);
2807 
2808 		opt->ip6po_tclass = tclass;
2809 		break;
2810 	}
2811 
2812 	case IPV6_2292NEXTHOP:
2813 	case IPV6_NEXTHOP:
2814 		if (cred != NULL) {
2815 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2816 			if (error)
2817 				return (error);
2818 		}
2819 
2820 		if (len == 0) {	/* just remove the option */
2821 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2822 			break;
2823 		}
2824 
2825 		/* check if cmsg_len is large enough for sa_len */
2826 		if (len < sizeof(struct sockaddr) || len < *buf)
2827 			return (EINVAL);
2828 
2829 		switch (((struct sockaddr *)buf)->sa_family) {
2830 		case AF_INET6:
2831 		{
2832 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2833 			int error;
2834 
2835 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2836 				return (EINVAL);
2837 
2838 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2839 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2840 				return (EINVAL);
2841 			}
2842 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2843 			    != 0) {
2844 				return (error);
2845 			}
2846 			break;
2847 		}
2848 		case AF_LINK:	/* should eventually be supported */
2849 		default:
2850 			return (EAFNOSUPPORT);
2851 		}
2852 
2853 		/* turn off the previous option, then set the new option. */
2854 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2855 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2856 		if (opt->ip6po_nexthop == NULL)
2857 			return (ENOBUFS);
2858 		bcopy(buf, opt->ip6po_nexthop, *buf);
2859 		break;
2860 
2861 	case IPV6_2292HOPOPTS:
2862 	case IPV6_HOPOPTS:
2863 	{
2864 		struct ip6_hbh *hbh;
2865 		int hbhlen;
2866 
2867 		/*
2868 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2869 		 * options, since per-option restriction has too much
2870 		 * overhead.
2871 		 */
2872 		if (cred != NULL) {
2873 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2874 			if (error)
2875 				return (error);
2876 		}
2877 
2878 		if (len == 0) {
2879 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2880 			break;	/* just remove the option */
2881 		}
2882 
2883 		/* message length validation */
2884 		if (len < sizeof(struct ip6_hbh))
2885 			return (EINVAL);
2886 		hbh = (struct ip6_hbh *)buf;
2887 		hbhlen = (hbh->ip6h_len + 1) << 3;
2888 		if (len != hbhlen)
2889 			return (EINVAL);
2890 
2891 		/* turn off the previous option, then set the new option. */
2892 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2893 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2894 		if (opt->ip6po_hbh == NULL)
2895 			return (ENOBUFS);
2896 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2897 
2898 		break;
2899 	}
2900 
2901 	case IPV6_2292DSTOPTS:
2902 	case IPV6_DSTOPTS:
2903 	case IPV6_RTHDRDSTOPTS:
2904 	{
2905 		struct ip6_dest *dest, **newdest = NULL;
2906 		int destlen;
2907 
2908 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2909 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2910 			if (error)
2911 				return (error);
2912 		}
2913 
2914 		if (len == 0) {
2915 			ip6_clearpktopts(opt, optname);
2916 			break;	/* just remove the option */
2917 		}
2918 
2919 		/* message length validation */
2920 		if (len < sizeof(struct ip6_dest))
2921 			return (EINVAL);
2922 		dest = (struct ip6_dest *)buf;
2923 		destlen = (dest->ip6d_len + 1) << 3;
2924 		if (len != destlen)
2925 			return (EINVAL);
2926 
2927 		/*
2928 		 * Determine the position that the destination options header
2929 		 * should be inserted; before or after the routing header.
2930 		 */
2931 		switch (optname) {
2932 		case IPV6_2292DSTOPTS:
2933 			/*
2934 			 * The old advacned API is ambiguous on this point.
2935 			 * Our approach is to determine the position based
2936 			 * according to the existence of a routing header.
2937 			 * Note, however, that this depends on the order of the
2938 			 * extension headers in the ancillary data; the 1st
2939 			 * part of the destination options header must appear
2940 			 * before the routing header in the ancillary data,
2941 			 * too.
2942 			 * RFC3542 solved the ambiguity by introducing
2943 			 * separate ancillary data or option types.
2944 			 */
2945 			if (opt->ip6po_rthdr == NULL)
2946 				newdest = &opt->ip6po_dest1;
2947 			else
2948 				newdest = &opt->ip6po_dest2;
2949 			break;
2950 		case IPV6_RTHDRDSTOPTS:
2951 			newdest = &opt->ip6po_dest1;
2952 			break;
2953 		case IPV6_DSTOPTS:
2954 			newdest = &opt->ip6po_dest2;
2955 			break;
2956 		}
2957 
2958 		/* turn off the previous option, then set the new option. */
2959 		ip6_clearpktopts(opt, optname);
2960 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2961 		if (*newdest == NULL)
2962 			return (ENOBUFS);
2963 		bcopy(dest, *newdest, destlen);
2964 
2965 		break;
2966 	}
2967 
2968 	case IPV6_2292RTHDR:
2969 	case IPV6_RTHDR:
2970 	{
2971 		struct ip6_rthdr *rth;
2972 		int rthlen;
2973 
2974 		if (len == 0) {
2975 			ip6_clearpktopts(opt, IPV6_RTHDR);
2976 			break;	/* just remove the option */
2977 		}
2978 
2979 		/* message length validation */
2980 		if (len < sizeof(struct ip6_rthdr))
2981 			return (EINVAL);
2982 		rth = (struct ip6_rthdr *)buf;
2983 		rthlen = (rth->ip6r_len + 1) << 3;
2984 		if (len != rthlen)
2985 			return (EINVAL);
2986 
2987 		switch (rth->ip6r_type) {
2988 		case IPV6_RTHDR_TYPE_0:
2989 			if (rth->ip6r_len == 0)	/* must contain one addr */
2990 				return (EINVAL);
2991 			if (rth->ip6r_len % 2) /* length must be even */
2992 				return (EINVAL);
2993 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2994 				return (EINVAL);
2995 			break;
2996 		default:
2997 			return (EINVAL);	/* not supported */
2998 		}
2999 
3000 		/* turn off the previous option */
3001 		ip6_clearpktopts(opt, IPV6_RTHDR);
3002 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3003 		if (opt->ip6po_rthdr == NULL)
3004 			return (ENOBUFS);
3005 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3006 
3007 		break;
3008 	}
3009 
3010 	case IPV6_USE_MIN_MTU:
3011 		if (len != sizeof(int))
3012 			return (EINVAL);
3013 		minmtupolicy = *(int *)buf;
3014 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3015 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3016 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3017 			return (EINVAL);
3018 		}
3019 		opt->ip6po_minmtu = minmtupolicy;
3020 		break;
3021 
3022 	case IPV6_DONTFRAG:
3023 		if (len != sizeof(int))
3024 			return (EINVAL);
3025 
3026 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3027 			/*
3028 			 * we ignore this option for TCP sockets.
3029 			 * (RFC3542 leaves this case unspecified.)
3030 			 */
3031 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3032 		} else
3033 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3034 		break;
3035 
3036 	case IPV6_PREFER_TEMPADDR:
3037 		if (len != sizeof(int))
3038 			return (EINVAL);
3039 		preftemp = *(int *)buf;
3040 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3041 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3042 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3043 			return (EINVAL);
3044 		}
3045 		opt->ip6po_prefer_tempaddr = preftemp;
3046 		break;
3047 
3048 	default:
3049 		return (ENOPROTOOPT);
3050 	} /* end of switch */
3051 
3052 	return (0);
3053 }
3054 
3055 /*
3056  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3057  * packet to the input queue of a specified interface.  Note that this
3058  * calls the output routine of the loopback "driver", but with an interface
3059  * pointer that might NOT be &loif -- easier than replicating that code here.
3060  */
3061 void
3062 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3063 {
3064 	struct mbuf *copym;
3065 	struct ip6_hdr *ip6;
3066 
3067 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3068 	if (copym == NULL)
3069 		return;
3070 
3071 	/*
3072 	 * Make sure to deep-copy IPv6 header portion in case the data
3073 	 * is in an mbuf cluster, so that we can safely override the IPv6
3074 	 * header portion later.
3075 	 */
3076 	if (!M_WRITABLE(copym) ||
3077 	    copym->m_len < sizeof(struct ip6_hdr)) {
3078 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3079 		if (copym == NULL)
3080 			return;
3081 	}
3082 	ip6 = mtod(copym, struct ip6_hdr *);
3083 	/*
3084 	 * clear embedded scope identifiers if necessary.
3085 	 * in6_clearscope will touch the addresses only when necessary.
3086 	 */
3087 	in6_clearscope(&ip6->ip6_src);
3088 	in6_clearscope(&ip6->ip6_dst);
3089 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3090 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3091 		    CSUM_PSEUDO_HDR;
3092 		copym->m_pkthdr.csum_data = 0xffff;
3093 	}
3094 	if_simloop(ifp, copym, AF_INET6, 0);
3095 }
3096 
3097 /*
3098  * Chop IPv6 header off from the payload.
3099  */
3100 static int
3101 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3102 {
3103 	struct mbuf *mh;
3104 	struct ip6_hdr *ip6;
3105 
3106 	ip6 = mtod(m, struct ip6_hdr *);
3107 	if (m->m_len > sizeof(*ip6)) {
3108 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3109 		if (mh == NULL) {
3110 			m_freem(m);
3111 			return ENOBUFS;
3112 		}
3113 		m_move_pkthdr(mh, m);
3114 		M_ALIGN(mh, sizeof(*ip6));
3115 		m->m_len -= sizeof(*ip6);
3116 		m->m_data += sizeof(*ip6);
3117 		mh->m_next = m;
3118 		m = mh;
3119 		m->m_len = sizeof(*ip6);
3120 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3121 	}
3122 	exthdrs->ip6e_ip6 = m;
3123 	return 0;
3124 }
3125 
3126 /*
3127  * Compute IPv6 extension header length.
3128  */
3129 int
3130 ip6_optlen(struct inpcb *in6p)
3131 {
3132 	int len;
3133 
3134 	if (!in6p->in6p_outputopts)
3135 		return 0;
3136 
3137 	len = 0;
3138 #define elen(x) \
3139     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3140 
3141 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3142 	if (in6p->in6p_outputopts->ip6po_rthdr)
3143 		/* dest1 is valid with rthdr only */
3144 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3145 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3146 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3147 	return len;
3148 #undef elen
3149 }
3150