xref: /freebsd/sys/netinet6/ip6_output.c (revision a9f5b68837094699a3d5204d79c1cbe59d93ae00)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  */
62 
63 #include <sys/cdefs.h>
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 #include "opt_kern_tls.h"
68 #include "opt_ratelimit.h"
69 #include "opt_route.h"
70 #include "opt_rss.h"
71 #include "opt_sctp.h"
72 
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/ktls.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/errno.h>
79 #include <sys/priv.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/syslog.h>
85 #include <sys/ucred.h>
86 
87 #include <machine/in_cksum.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/if_private.h>
92 #include <net/if_vlan_var.h>
93 #include <net/if_llatbl.h>
94 #include <net/ethernet.h>
95 #include <net/netisr.h>
96 #include <net/route.h>
97 #include <net/route/nhop.h>
98 #include <net/pfil.h>
99 #include <net/rss_config.h>
100 #include <net/vnet.h>
101 
102 #include <netinet/in.h>
103 #include <netinet/in_var.h>
104 #include <netinet/ip_var.h>
105 #include <netinet6/in6_fib.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet/ip6.h>
108 #include <netinet/icmp6.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet/in_pcb.h>
111 #include <netinet/tcp_var.h>
112 #include <netinet6/nd6.h>
113 #include <netinet6/in6_rss.h>
114 
115 #include <netipsec/ipsec_support.h>
116 #if defined(SCTP) || defined(SCTP_SUPPORT)
117 #include <netinet/sctp.h>
118 #include <netinet/sctp_crc32.h>
119 #endif
120 
121 #include <netinet6/scope6_var.h>
122 
123 extern int in6_mcast_loop;
124 
125 struct ip6_exthdrs {
126 	struct mbuf *ip6e_ip6;
127 	struct mbuf *ip6e_hbh;
128 	struct mbuf *ip6e_dest1;
129 	struct mbuf *ip6e_rthdr;
130 	struct mbuf *ip6e_dest2;
131 };
132 
133 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
134 
135 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
136 			   struct ucred *, int);
137 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
138 	struct socket *, struct sockopt *);
139 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
140 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
141 	struct ucred *, int, int, int);
142 
143 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
144 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
145 	struct ip6_frag **);
146 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
147 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
148 static void ip6_getpmtu(struct route_in6 *, int,
149 	struct ifnet *, const struct in6_addr *, u_long *, u_int, u_int);
150 static void ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
151 	u_long *, 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  * Make an extension header from option data.  hp is the source,
157  * mp is the destination, and _ol is the optlen.
158  */
159 #define	MAKE_EXTHDR(hp, mp, _ol)					\
160     do {								\
161 	struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
162 	error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
163 	    ((eh)->ip6e_len + 1) << 3);				\
164 	if (error)						\
165 		goto freehdrs;					\
166 	(_ol) += (*(mp))->m_len;				\
167     } while (/*CONSTCOND*/ 0)
168 
169 /*
170  * Form a chain of extension headers.
171  * m is the extension header mbuf
172  * mp is the previous mbuf in the chain
173  * p is the next header
174  * i is the type of option.
175  */
176 #define MAKE_CHAIN(m, mp, p, i)\
177     do {\
178 	if (m) {\
179 		if (!hdrsplit) \
180 			panic("%s:%d: assumption failed: "\
181 			    "hdr not split: hdrsplit %d exthdrs %p",\
182 			    __func__, __LINE__, hdrsplit, &exthdrs);\
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
in6_delayed_cksum(struct mbuf * m,uint32_t plen,u_short offset)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 static void
ip6_output_delayed_csum(struct mbuf * m,struct ifnet * ifp,int csum_flags,int plen,int optlen)209 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
210     int plen, int optlen)
211 {
212 
213 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
214 	    "csum_flags %#x",
215 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
216 
217 	if (csum_flags & CSUM_DELAY_DATA_IPV6) {
218 		in6_delayed_cksum(m, plen - optlen,
219 		    sizeof(struct ip6_hdr) + optlen);
220 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
221 	}
222 #if defined(SCTP) || defined(SCTP_SUPPORT)
223 	if (csum_flags & CSUM_SCTP_IPV6) {
224 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
225 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
226 	}
227 #endif
228 }
229 
230 int
ip6_fragment(struct ifnet * ifp,struct mbuf * m0,int hlen,u_char nextproto,int fraglen,uint32_t id)231 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
232     int fraglen , uint32_t id)
233 {
234 	struct mbuf *m, **mnext, *m_frgpart;
235 	struct ip6_hdr *ip6, *mhip6;
236 	struct ip6_frag *ip6f;
237 	int off;
238 	int error;
239 	int tlen = m0->m_pkthdr.len;
240 
241 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
242 
243 	m = m0;
244 	ip6 = mtod(m, struct ip6_hdr *);
245 	mnext = &m->m_nextpkt;
246 
247 	for (off = hlen; off < tlen; off += fraglen) {
248 		m = m_gethdr(M_NOWAIT, MT_DATA);
249 		if (!m) {
250 			IP6STAT_INC(ip6s_odropped);
251 			return (ENOBUFS);
252 		}
253 
254 		/*
255 		 * Make sure the complete packet header gets copied
256 		 * from the originating mbuf to the newly created
257 		 * mbuf. This also ensures that existing firewall
258 		 * classification(s), VLAN tags and so on get copied
259 		 * to the resulting fragmented packet(s):
260 		 */
261 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
262 			m_free(m);
263 			IP6STAT_INC(ip6s_odropped);
264 			return (ENOBUFS);
265 		}
266 
267 		*mnext = m;
268 		mnext = &m->m_nextpkt;
269 		m->m_data += max_linkhdr;
270 		mhip6 = mtod(m, struct ip6_hdr *);
271 		*mhip6 = *ip6;
272 		m->m_len = sizeof(*mhip6);
273 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
274 		if (error) {
275 			IP6STAT_INC(ip6s_odropped);
276 			return (error);
277 		}
278 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
279 		if (off + fraglen >= tlen)
280 			fraglen = tlen - off;
281 		else
282 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
283 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
284 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
285 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
286 			IP6STAT_INC(ip6s_odropped);
287 			return (ENOBUFS);
288 		}
289 		m_cat(m, m_frgpart);
290 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
291 		ip6f->ip6f_reserved = 0;
292 		ip6f->ip6f_ident = id;
293 		ip6f->ip6f_nxt = nextproto;
294 		IP6STAT_INC(ip6s_ofragments);
295 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
296 	}
297 
298 	return (0);
299 }
300 
301 static int
ip6_output_send(struct inpcb * inp,struct ifnet * ifp,struct ifnet * origifp,struct mbuf * m,struct sockaddr_in6 * dst,struct route_in6 * ro,bool stamp_tag)302 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
303     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
304     bool stamp_tag)
305 {
306 #ifdef KERN_TLS
307 	struct ktls_session *tls = NULL;
308 #endif
309 	struct m_snd_tag *mst;
310 	int error;
311 
312 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
313 	mst = NULL;
314 
315 #ifdef KERN_TLS
316 	/*
317 	 * If this is an unencrypted TLS record, save a reference to
318 	 * the record.  This local reference is used to call
319 	 * ktls_output_eagain after the mbuf has been freed (thus
320 	 * dropping the mbuf's reference) in if_output.
321 	 */
322 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
323 		tls = ktls_hold(m->m_next->m_epg_tls);
324 		mst = tls->snd_tag;
325 
326 		/*
327 		 * If a TLS session doesn't have a valid tag, it must
328 		 * have had an earlier ifp mismatch, so drop this
329 		 * packet.
330 		 */
331 		if (mst == NULL) {
332 			m_freem(m);
333 			error = EAGAIN;
334 			goto done;
335 		}
336 		/*
337 		 * Always stamp tags that include NIC ktls.
338 		 */
339 		stamp_tag = true;
340 	}
341 #endif
342 #ifdef RATELIMIT
343 	if (inp != NULL && mst == NULL) {
344 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
345 		    (inp->inp_snd_tag != NULL &&
346 		    inp->inp_snd_tag->ifp != ifp))
347 			in_pcboutput_txrtlmt(inp, ifp, m);
348 
349 		if (inp->inp_snd_tag != NULL)
350 			mst = inp->inp_snd_tag;
351 	}
352 #endif
353 	if (stamp_tag && mst != NULL) {
354 		KASSERT(m->m_pkthdr.rcvif == NULL,
355 		    ("trying to add a send tag to a forwarded packet"));
356 		if (mst->ifp != ifp) {
357 			m_freem(m);
358 			error = EAGAIN;
359 			goto done;
360 		}
361 
362 		/* stamp send tag on mbuf */
363 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
364 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
365 	}
366 
367 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
368 
369 done:
370 	/* Check for route change invalidating send tags. */
371 #ifdef KERN_TLS
372 	if (tls != NULL) {
373 		if (error == EAGAIN)
374 			error = ktls_output_eagain(inp, tls);
375 		ktls_free(tls);
376 	}
377 #endif
378 #ifdef RATELIMIT
379 	if (error == EAGAIN)
380 		in_pcboutput_eagain(inp);
381 #endif
382 	return (error);
383 }
384 
385 /*
386  * IP6 output.
387  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
388  * nxt, hlim, src, dst).
389  * This function may modify ver and hlim only.
390  * The mbuf chain containing the packet will be freed.
391  * The mbuf opt, if present, will not be freed.
392  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
393  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
394  * then result of route lookup is stored in ro->ro_nh.
395  *
396  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
397  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
398  *
399  * ifpp - XXX: just for statistics
400  */
401 int
ip6_output(struct mbuf * m0,struct ip6_pktopts * opt,struct route_in6 * ro,int flags,struct ip6_moptions * im6o,struct ifnet ** ifpp,struct inpcb * inp)402 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
403     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
404     struct ifnet **ifpp, struct inpcb *inp)
405 {
406 	struct ip6_hdr *ip6;
407 	struct ifnet *ifp, *origifp;
408 	struct mbuf *m = m0;
409 	struct mbuf *mprev;
410 	struct route_in6 *ro_pmtu;
411 	struct nhop_object *nh;
412 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
413 	struct in6_addr odst;
414 	u_char *nexthdrp;
415 	int tlen, len;
416 	int error = 0;
417 	int vlan_pcp = -1;
418 	struct in6_ifaddr *ia = NULL;
419 	u_long mtu;
420 	int dontfrag;
421 	u_int32_t optlen, plen = 0, unfragpartlen;
422 	struct ip6_exthdrs exthdrs;
423 	struct in6_addr src0, dst0;
424 	u_int32_t zone;
425 	bool hdrsplit;
426 	int sw_csum, tso;
427 	int needfiblookup;
428 	uint32_t fibnum;
429 	struct m_tag *fwd_tag = NULL;
430 	uint32_t id;
431 	uint32_t optvalid;
432 
433 	NET_EPOCH_ASSERT();
434 
435 	if (inp != NULL) {
436 		INP_LOCK_ASSERT(inp);
437 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
438 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
439 			/* Unconditionally set flowid. */
440 			m->m_pkthdr.flowid = inp->inp_flowid;
441 			M_HASHTYPE_SET(m, inp->inp_flowtype);
442 		}
443 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
444 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
445 			    INP_2PCP_SHIFT;
446 #ifdef NUMA
447 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
448 #endif
449 	}
450 
451 	/* Source address validation. */
452 	ip6 = mtod(m, struct ip6_hdr *);
453 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
454 	    (flags & IPV6_UNSPECSRC) == 0) {
455 		error = EOPNOTSUPP;
456 		IP6STAT_INC(ip6s_badscope);
457 		goto bad;
458 	}
459 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
460 		error = EOPNOTSUPP;
461 		IP6STAT_INC(ip6s_badscope);
462 		goto bad;
463 	}
464 
465 	/*
466 	 * If we are given packet options to add extension headers prepare them.
467 	 * Calculate the total length of the extension header chain.
468 	 * Keep the length of the unfragmentable part for fragmentation.
469 	 */
470 	bzero(&exthdrs, sizeof(exthdrs));
471 	optlen = optvalid = 0;
472 	unfragpartlen = sizeof(struct ip6_hdr);
473 	if (opt) {
474 		optvalid = opt->ip6po_valid;
475 
476 		/* Hop-by-Hop options header. */
477 		if ((optvalid & IP6PO_VALID_HBH) != 0)
478 			MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
479 
480 		/* Destination options header (1st part). */
481 		if ((optvalid & IP6PO_VALID_RHINFO) != 0) {
482 #ifndef RTHDR_SUPPORT_IMPLEMENTED
483 			/*
484 			 * If there is a routing header, discard the packet
485 			 * right away here. RH0/1 are obsolete and we do not
486 			 * currently support RH2/3/4.
487 			 * People trying to use RH253/254 may want to disable
488 			 * this check.
489 			 * The moment we do support any routing header (again)
490 			 * this block should check the routing type more
491 			 * selectively.
492 			 */
493 			error = EINVAL;
494 			goto bad;
495 #endif
496 
497 			/*
498 			 * Destination options header (1st part).
499 			 * This only makes sense with a routing header.
500 			 * See Section 9.2 of RFC 3542.
501 			 * Disabling this part just for MIP6 convenience is
502 			 * a bad idea.  We need to think carefully about a
503 			 * way to make the advanced API coexist with MIP6
504 			 * options, which might automatically be inserted in
505 			 * the kernel.
506 			 */
507 			if ((optvalid & IP6PO_VALID_DEST1) != 0)
508 				MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
509 				    optlen);
510 		}
511 		/* Routing header. */
512 		if ((optvalid & IP6PO_VALID_RHINFO) != 0)
513 			MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
514 
515 		unfragpartlen += optlen;
516 
517 		/*
518 		 * NOTE: we don't add AH/ESP length here (done in
519 		 * ip6_ipsec_output()).
520 		 */
521 
522 		/* Destination options header (2nd part). */
523 		if ((optvalid & IP6PO_VALID_DEST2) != 0)
524 			MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
525 	}
526 
527 	/*
528 	 * If there is at least one extension header,
529 	 * separate IP6 header from the payload.
530 	 */
531 	hdrsplit = false;
532 	if (optlen) {
533 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
534 			m = NULL;
535 			goto freehdrs;
536 		}
537 		m = exthdrs.ip6e_ip6;
538 		ip6 = mtod(m, struct ip6_hdr *);
539 		hdrsplit = true;
540 	}
541 
542 	/* Adjust mbuf packet header length. */
543 	m->m_pkthdr.len += optlen;
544 	plen = m->m_pkthdr.len - sizeof(*ip6);
545 
546 	/* If this is a jumbo payload, insert a jumbo payload option. */
547 	if (plen > IPV6_MAXPACKET) {
548 		if (!hdrsplit) {
549 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
550 				m = NULL;
551 				goto freehdrs;
552 			}
553 			m = exthdrs.ip6e_ip6;
554 			ip6 = mtod(m, struct ip6_hdr *);
555 			hdrsplit = true;
556 		}
557 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
558 			goto freehdrs;
559 		ip6->ip6_plen = 0;
560 	} else
561 		ip6->ip6_plen = htons(plen);
562 	nexthdrp = &ip6->ip6_nxt;
563 
564 	if (optlen) {
565 		/*
566 		 * Concatenate headers and fill in next header fields.
567 		 * Here we have, on "m"
568 		 *	IPv6 payload
569 		 * and we insert headers accordingly.
570 		 * Finally, we should be getting:
571 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
572 		 *
573 		 * During the header composing process "m" points to IPv6
574 		 * header.  "mprev" points to an extension header prior to esp.
575 		 */
576 		mprev = m;
577 
578 		/*
579 		 * We treat dest2 specially.  This makes IPsec processing
580 		 * much easier.  The goal here is to make mprev point the
581 		 * mbuf prior to dest2.
582 		 *
583 		 * Result: IPv6 dest2 payload.
584 		 * m and mprev will point to IPv6 header.
585 		 */
586 		if (exthdrs.ip6e_dest2) {
587 			if (!hdrsplit)
588 				panic("%s:%d: assumption failed: "
589 				    "hdr not split: hdrsplit %d exthdrs %p",
590 				    __func__, __LINE__, hdrsplit, &exthdrs);
591 			exthdrs.ip6e_dest2->m_next = m->m_next;
592 			m->m_next = exthdrs.ip6e_dest2;
593 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
594 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
595 		}
596 
597 		/*
598 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
599 		 * m will point to IPv6 header.  mprev will point to the
600 		 * extension header prior to dest2 (rthdr in the above case).
601 		 */
602 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
603 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
604 			   IPPROTO_DSTOPTS);
605 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
606 			   IPPROTO_ROUTING);
607 	}
608 
609 	IP6STAT_INC(ip6s_localout);
610 
611 	/* Route packet. */
612 	ro_pmtu = ro;
613 	if ((optvalid & IP6PO_VALID_RHINFO) != 0)
614 		ro = &opt->ip6po_route;
615 	if (ro != NULL)
616 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
617 	else
618 		dst = &sin6;
619 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
620 
621 again:
622 	/*
623 	 * If specified, try to fill in the traffic class field.
624 	 * Do not override if a non-zero value is already set.
625 	 * We check the diffserv field and the ECN field separately.
626 	 */
627 	if ((optvalid & IP6PO_VALID_TC) != 0){
628 		int mask = 0;
629 
630 		if (IPV6_DSCP(ip6) == 0)
631 			mask |= 0xfc;
632 		if (IPV6_ECN(ip6) == 0)
633 			mask |= 0x03;
634 		if (mask != 0)
635 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
636 	}
637 
638 	/* Fill in or override the hop limit field, if necessary. */
639 	if ((optvalid & IP6PO_VALID_HLIM) != 0)
640 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
641 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
642 		if (im6o != NULL)
643 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
644 		else
645 			ip6->ip6_hlim = V_ip6_defmcasthlim;
646 	}
647 
648 	if (ro == NULL || ro->ro_nh == NULL) {
649 		bzero(dst, sizeof(*dst));
650 		dst->sin6_family = AF_INET6;
651 		dst->sin6_len = sizeof(*dst);
652 		dst->sin6_addr = ip6->ip6_dst;
653 	}
654 	/*
655 	 * Validate route against routing table changes.
656 	 * Make sure that the address family is set in route.
657 	 */
658 	nh = NULL;
659 	ifp = NULL;
660 	mtu = 0;
661 	if (ro != NULL) {
662 		if (ro->ro_nh != NULL && inp != NULL) {
663 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
664 			NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
665 			    fibnum);
666 		}
667 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
668 		    (!NH_IS_VALID(ro->ro_nh) ||
669 		    ro->ro_dst.sin6_family != AF_INET6 ||
670 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
671 			RO_INVALIDATE_CACHE(ro);
672 
673 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
674 		    ro->ro_dst.sin6_family == AF_INET6 &&
675 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
676 			/* Nexthop is valid and contains valid ifp */
677 			nh = ro->ro_nh;
678 		} else {
679 			if (ro->ro_lle)
680 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
681 			ro->ro_lle = NULL;
682 			if (fwd_tag == NULL) {
683 				bzero(&dst_sa, sizeof(dst_sa));
684 				dst_sa.sin6_family = AF_INET6;
685 				dst_sa.sin6_len = sizeof(dst_sa);
686 				dst_sa.sin6_addr = ip6->ip6_dst;
687 			}
688 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
689 			    &nh, fibnum, m->m_pkthdr.flowid);
690 			if (error != 0) {
691 				IP6STAT_INC(ip6s_noroute);
692 				if (ifp != NULL)
693 					in6_ifstat_inc(ifp, ifs6_out_discard);
694 				goto bad;
695 			}
696 			/*
697 			 * At this point at least @ifp is not NULL
698 			 * Can be the case when dst is multicast, link-local or
699 			 * interface is explicitly specificed by the caller.
700 			 */
701 		}
702 		if (nh == NULL) {
703 			/*
704 			 * If in6_selectroute() does not return a nexthop
705 			 * dst may not have been updated.
706 			 */
707 			*dst = dst_sa;	/* XXX */
708 			origifp = ifp;
709 			mtu = ifp->if_mtu;
710 		} else {
711 			ifp = nh->nh_ifp;
712 			origifp = nh->nh_aifp;
713 			ia = (struct in6_ifaddr *)(nh->nh_ifa);
714 			counter_u64_add(nh->nh_pksent, 1);
715 		}
716 	} else {
717 		struct nhop_object *nh;
718 		struct in6_addr kdst;
719 		uint32_t scopeid;
720 
721 		if (fwd_tag == NULL) {
722 			bzero(&dst_sa, sizeof(dst_sa));
723 			dst_sa.sin6_family = AF_INET6;
724 			dst_sa.sin6_len = sizeof(dst_sa);
725 			dst_sa.sin6_addr = ip6->ip6_dst;
726 		}
727 
728 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
729 		    im6o != NULL &&
730 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
731 			/* We do not need a route lookup. */
732 			*dst = dst_sa;	/* XXX */
733 			origifp = ifp;
734 			goto nonh6lookup;
735 		}
736 
737 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
738 
739 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
740 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
741 			if (scopeid > 0) {
742 				ifp = in6_getlinkifnet(scopeid);
743 				if (ifp == NULL) {
744 					error = EHOSTUNREACH;
745 					goto bad;
746 				}
747 				*dst = dst_sa;	/* XXX */
748 				origifp = ifp;
749 				goto nonh6lookup;
750 			}
751 		}
752 
753 		nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
754 		    m->m_pkthdr.flowid);
755 		if (nh == NULL) {
756 			IP6STAT_INC(ip6s_noroute);
757 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
758 			error = EHOSTUNREACH;
759 			goto bad;
760 		}
761 
762 		ifp = nh->nh_ifp;
763 		origifp = nh->nh_aifp;
764 		ia = ifatoia6(nh->nh_ifa);
765 		if (nh->nh_flags & NHF_GATEWAY)
766 			dst->sin6_addr = nh->gw6_sa.sin6_addr;
767 		else if (fwd_tag != NULL)
768 			dst->sin6_addr = dst_sa.sin6_addr;
769 nonh6lookup:
770 		;
771 	}
772 	/*
773 	 * At this point ifp MUST be pointing to the valid transmit ifp.
774 	 * origifp MUST be valid and pointing to either the same ifp or,
775 	 * in case of loopback output, to the interface which ip6_src
776 	 * belongs to.
777 	 * Examples:
778 	 *  fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
779 	 *  fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
780 	 *  ::1 -> ::1 -> ifp=lo0, origifp=lo0
781 	 *
782 	 * mtu can be 0 and will be refined later.
783 	 */
784 	KASSERT((ifp != NULL), ("output interface must not be NULL"));
785 	KASSERT((origifp != NULL), ("output address interface must not be NULL"));
786 
787 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
788 	/*
789 	 * IPSec checking which handles several cases.
790 	 * FAST IPSEC: We re-injected the packet.
791 	 * XXX: need scope argument.
792 	 */
793 	if (IPSEC_ENABLED(ipv6)) {
794 		if ((error = IPSEC_OUTPUT(ipv6, ifp, m, inp, mtu == 0 ?
795 		    ifp->if_mtu : mtu)) != 0) {
796 			if (error == EINPROGRESS)
797 				error = 0;
798 			goto done;
799 		}
800 	}
801 #endif /* IPSEC */
802 
803 	if ((flags & IPV6_FORWARDING) == 0) {
804 		/* XXX: the FORWARDING flag can be set for mrouting. */
805 		in6_ifstat_inc(ifp, ifs6_out_request);
806 	}
807 
808 	/* Setup data structures for scope ID checks. */
809 	src0 = ip6->ip6_src;
810 	bzero(&src_sa, sizeof(src_sa));
811 	src_sa.sin6_family = AF_INET6;
812 	src_sa.sin6_len = sizeof(src_sa);
813 	src_sa.sin6_addr = ip6->ip6_src;
814 
815 	dst0 = ip6->ip6_dst;
816 	/* Re-initialize to be sure. */
817 	bzero(&dst_sa, sizeof(dst_sa));
818 	dst_sa.sin6_family = AF_INET6;
819 	dst_sa.sin6_len = sizeof(dst_sa);
820 	dst_sa.sin6_addr = ip6->ip6_dst;
821 
822 	/* Check for valid scope ID. */
823 	if (in6_setscope(&src0, origifp, &zone) == 0 &&
824 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
825 	    in6_setscope(&dst0, origifp, &zone) == 0 &&
826 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
827 		/*
828 		 * The outgoing interface is in the zone of the source
829 		 * and destination addresses.
830 		 *
831 		 */
832 	} else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
833 	    sa6_recoverscope(&src_sa) != 0 ||
834 	    sa6_recoverscope(&dst_sa) != 0 ||
835 	    dst_sa.sin6_scope_id == 0 ||
836 	    (src_sa.sin6_scope_id != 0 &&
837 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
838 	    ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
839 		/*
840 		 * If the destination network interface is not a
841 		 * loopback interface, or the destination network
842 		 * address has no scope ID, or the source address has
843 		 * a scope ID set which is different from the
844 		 * destination address one, or there is no network
845 		 * interface representing this scope ID, the address
846 		 * pair is considered invalid.
847 		 */
848 		IP6STAT_INC(ip6s_badscope);
849 		in6_ifstat_inc(origifp, ifs6_out_discard);
850 		if (error == 0)
851 			error = EHOSTUNREACH; /* XXX */
852 		goto bad;
853 	}
854 	/* All scope ID checks are successful. */
855 
856 	if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
857 		if ((optvalid & IP6PO_VALID_NHINFO) != 0) {
858 			/*
859 			 * The nexthop is explicitly specified by the
860 			 * application.  We assume the next hop is an IPv6
861 			 * address.
862 			 */
863 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
864 		}
865 		else if ((nh->nh_flags & NHF_GATEWAY))
866 			dst = &nh->gw6_sa;
867 	}
868 
869 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
870 		m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
871 	} else {
872 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
873 		in6_ifstat_inc(ifp, ifs6_out_mcast);
874 
875 		/* Confirm that the outgoing interface supports multicast. */
876 		if (!(ifp->if_flags & IFF_MULTICAST)) {
877 			IP6STAT_INC(ip6s_noroute);
878 			in6_ifstat_inc(ifp, ifs6_out_discard);
879 			error = ENETUNREACH;
880 			goto bad;
881 		}
882 		if ((im6o == NULL && in6_mcast_loop) ||
883 		    (im6o && im6o->im6o_multicast_loop)) {
884 			/*
885 			 * Loop back multicast datagram if not expressly
886 			 * forbidden to do so, even if we have not joined
887 			 * the address; protocols will filter it later,
888 			 * thus deferring a hash lookup and lock acquisition
889 			 * at the expense of an m_copym().
890 			 */
891 			ip6_mloopback(ifp, m);
892 		} else {
893 			/*
894 			 * If we are acting as a multicast router, perform
895 			 * multicast forwarding as if the packet had just
896 			 * arrived on the interface to which we are about
897 			 * to send.  The multicast forwarding function
898 			 * recursively calls this function, using the
899 			 * IPV6_FORWARDING flag to prevent infinite recursion.
900 			 *
901 			 * Multicasts that are looped back by ip6_mloopback(),
902 			 * above, will be forwarded by the ip6_input() routine,
903 			 * if necessary.
904 			 */
905 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
906 				/*
907 				 * XXX: ip6_mforward expects that rcvif is NULL
908 				 * when it is called from the originating path.
909 				 * However, it may not always be the case.
910 				 */
911 				m->m_pkthdr.rcvif = NULL;
912 				if (ip6_mforward(ip6, ifp, m) != 0) {
913 					m_freem(m);
914 					goto done;
915 				}
916 			}
917 		}
918 		/*
919 		 * Multicasts with a hoplimit of zero may be looped back,
920 		 * above, but must not be transmitted on a network.
921 		 * Also, multicasts addressed to the loopback interface
922 		 * are not sent -- the above call to ip6_mloopback() will
923 		 * loop back a copy if this host actually belongs to the
924 		 * destination group on the loopback interface.
925 		 */
926 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
927 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
928 			m_freem(m);
929 			goto done;
930 		}
931 	}
932 
933 	/*
934 	 * Fill the outgoing inteface to tell the upper layer
935 	 * to increment per-interface statistics.
936 	 */
937 	if (ifpp)
938 		*ifpp = ifp;
939 
940 	/* Determine path MTU. */
941 	ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst, &mtu, fibnum,
942 	    *nexthdrp);
943 	KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p fibnum %u",
944 	    __func__, __LINE__, mtu, ro_pmtu, ro, ifp, fibnum));
945 
946 	/*
947 	 * The caller of this function may specify to use the minimum MTU
948 	 * in some cases.
949 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
950 	 * setting.  The logic is a bit complicated; by default, unicast
951 	 * packets will follow path MTU while multicast packets will be sent at
952 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
953 	 * including unicast ones will be sent at the minimum MTU.  Multicast
954 	 * packets will always be sent at the minimum MTU unless
955 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
956 	 * See RFC 3542 for more details.
957 	 */
958 	if (mtu > IPV6_MMTU) {
959 		if ((flags & IPV6_MINMTU))
960 			mtu = IPV6_MMTU;
961 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
962 			mtu = IPV6_MMTU;
963 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
964 			 (opt == NULL ||
965 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
966 			mtu = IPV6_MMTU;
967 		}
968 	}
969 
970 	/*
971 	 * Clear embedded scope identifiers if necessary.
972 	 * in6_clearscope() will touch the addresses only when necessary.
973 	 */
974 	in6_clearscope(&ip6->ip6_src);
975 	in6_clearscope(&ip6->ip6_dst);
976 
977 	/*
978 	 * If the outgoing packet contains a hop-by-hop options header,
979 	 * it must be examined and processed even by the source node.
980 	 * (RFC 2460, section 4.)
981 	 */
982 	if (exthdrs.ip6e_hbh) {
983 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
984 		u_int32_t dummy; /* XXX unused */
985 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
986 
987 #ifdef DIAGNOSTIC
988 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
989 			panic("ip6e_hbh is not contiguous");
990 #endif
991 		/*
992 		 *  XXX: if we have to send an ICMPv6 error to the sender,
993 		 *       we need the M_LOOP flag since icmp6_error() expects
994 		 *       the IPv6 and the hop-by-hop options header are
995 		 *       contiguous unless the flag is set.
996 		 */
997 		m->m_flags |= M_LOOP;
998 		m->m_pkthdr.rcvif = ifp;
999 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1000 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1001 		    &dummy, &plen) < 0) {
1002 			/* m was already freed at this point. */
1003 			error = EINVAL;/* better error? */
1004 			goto done;
1005 		}
1006 		m->m_flags &= ~M_LOOP; /* XXX */
1007 		m->m_pkthdr.rcvif = NULL;
1008 	}
1009 
1010 	/* Jump over all PFIL processing if hooks are not active. */
1011 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1012 		goto passout;
1013 
1014 	odst = ip6->ip6_dst;
1015 	/* Run through list of hooks for output packets. */
1016 	switch (pfil_mbuf_out(V_inet6_pfil_head, &m, ifp, inp)) {
1017 	case PFIL_PASS:
1018 		ip6 = mtod(m, struct ip6_hdr *);
1019 		break;
1020 	case PFIL_DROPPED:
1021 		error = EACCES;
1022 		/* FALLTHROUGH */
1023 	case PFIL_CONSUMED:
1024 		goto done;
1025 	}
1026 
1027 	needfiblookup = 0;
1028 	/* See if destination IP address was changed by packet filter. */
1029 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1030 		m->m_flags |= M_SKIP_FIREWALL;
1031 		/* If destination is now ourself drop to ip6_input(). */
1032 		if (in6_localip(&ip6->ip6_dst)) {
1033 			m->m_flags |= M_FASTFWD_OURS;
1034 			if (m->m_pkthdr.rcvif == NULL)
1035 				m->m_pkthdr.rcvif = V_loif;
1036 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1037 				m->m_pkthdr.csum_flags |=
1038 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1039 				m->m_pkthdr.csum_data = 0xffff;
1040 			}
1041 #if defined(SCTP) || defined(SCTP_SUPPORT)
1042 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1043 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1044 #endif
1045 			error = netisr_queue(NETISR_IPV6, m);
1046 			goto done;
1047 		} else {
1048 			if (ro != NULL)
1049 				RO_INVALIDATE_CACHE(ro);
1050 			needfiblookup = 1; /* Redo the routing table lookup. */
1051 		}
1052 	}
1053 	/* See if fib was changed by packet filter. */
1054 	if (fibnum != M_GETFIB(m)) {
1055 		m->m_flags |= M_SKIP_FIREWALL;
1056 		fibnum = M_GETFIB(m);
1057 		if (ro != NULL)
1058 			RO_INVALIDATE_CACHE(ro);
1059 		needfiblookup = 1;
1060 	}
1061 	if (needfiblookup)
1062 		goto again;
1063 
1064 	/* See if local, if yes, send it to netisr. */
1065 	if (m->m_flags & M_FASTFWD_OURS) {
1066 		if (m->m_pkthdr.rcvif == NULL)
1067 			m->m_pkthdr.rcvif = V_loif;
1068 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1069 			m->m_pkthdr.csum_flags |=
1070 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1071 			m->m_pkthdr.csum_data = 0xffff;
1072 		}
1073 #if defined(SCTP) || defined(SCTP_SUPPORT)
1074 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1075 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1076 #endif
1077 		error = netisr_queue(NETISR_IPV6, m);
1078 		goto done;
1079 	}
1080 	/* Or forward to some other address? */
1081 	if ((m->m_flags & M_IP6_NEXTHOP) &&
1082 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1083 		if (ro != NULL)
1084 			dst = (struct sockaddr_in6 *)&ro->ro_dst;
1085 		else
1086 			dst = &sin6;
1087 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1088 		m->m_flags |= M_SKIP_FIREWALL;
1089 		m->m_flags &= ~M_IP6_NEXTHOP;
1090 		m_tag_delete(m, fwd_tag);
1091 		goto again;
1092 	}
1093 
1094 passout:
1095 	if (vlan_pcp > -1)
1096 		EVL_APPLY_PRI(m, vlan_pcp);
1097 
1098 	/* Ensure the packet data is mapped if the interface requires it. */
1099 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1100 		struct mbuf *m1;
1101 
1102 		error = mb_unmapped_to_ext(m, &m1);
1103 		if (error != 0) {
1104 			if (error == EINVAL) {
1105 				if_printf(ifp, "TLS packet\n");
1106 				/* XXXKIB */
1107 			} else if (error == ENOMEM) {
1108 				error = ENOBUFS;
1109 			}
1110 			IP6STAT_INC(ip6s_odropped);
1111 			return (error);
1112 		} else {
1113 			m = m1;
1114 		}
1115 	}
1116 
1117 	/*
1118 	 * Send the packet to the outgoing interface.
1119 	 * If necessary, do IPv6 fragmentation before sending.
1120 	 *
1121 	 * 1: normal case (dontfrag == 0)
1122 	 * 1-a:	send as is if tlen <= path mtu
1123 	 * 1-b:	fragment if tlen > path mtu
1124 	 *
1125 	 * 2: if user asks us not to fragment (dontfrag == 1)
1126 	 * 2-a:	send as is if tlen <= interface mtu
1127 	 * 2-b:	error if tlen > interface mtu
1128 	 */
1129 	sw_csum = m->m_pkthdr.csum_flags;
1130 	if (!hdrsplit) {
1131 		tso = ((sw_csum & ifp->if_hwassist &
1132 		    (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1133 		sw_csum &= ~ifp->if_hwassist;
1134 	} else
1135 		tso = 0;
1136 	/*
1137 	 * If we added extension headers, we will not do TSO and calculate the
1138 	 * checksums ourselves for now.
1139 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1140 	 * with ext. hdrs.
1141 	 */
1142 	ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1143 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1144 	tlen = m->m_pkthdr.len;
1145 
1146 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1147 		dontfrag = 1;
1148 	else
1149 		dontfrag = 0;
1150 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* Case 2-b. */
1151 		/*
1152 		 * If the DONTFRAG option is specified, we cannot send the
1153 		 * packet when the data length is larger than the MTU of the
1154 		 * outgoing interface.
1155 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1156 		 * application wanted to know the MTU value. Also return an
1157 		 * error code (this is not described in the API spec).
1158 		 */
1159 		if (inp != NULL)
1160 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1161 		error = EMSGSIZE;
1162 		goto bad;
1163 	}
1164 
1165 	/* Transmit packet without fragmentation. */
1166 	if (dontfrag || tlen <= mtu) {	/* Cases 1-a and 2-a. */
1167 		struct in6_ifaddr *ia6;
1168 
1169 		ip6 = mtod(m, struct ip6_hdr *);
1170 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1171 		if (ia6) {
1172 			/* Record statistics for this interface address. */
1173 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1174 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1175 			    m->m_pkthdr.len);
1176 		}
1177 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1178 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
1179 		goto done;
1180 	}
1181 
1182 	/* Try to fragment the packet.  Case 1-b. */
1183 	if (mtu < IPV6_MMTU) {
1184 		/* Path MTU cannot be less than IPV6_MMTU. */
1185 		error = EMSGSIZE;
1186 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1187 		goto bad;
1188 	} else if (ip6->ip6_plen == 0) {
1189 		/* Jumbo payload cannot be fragmented. */
1190 		error = EMSGSIZE;
1191 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1192 		goto bad;
1193 	} else {
1194 		u_char nextproto;
1195 
1196 		/*
1197 		 * Too large for the destination or interface;
1198 		 * fragment if possible.
1199 		 * Must be able to put at least 8 bytes per fragment.
1200 		 */
1201 		if (mtu > IPV6_MAXPACKET)
1202 			mtu = IPV6_MAXPACKET;
1203 
1204 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1205 		if (len < 8) {
1206 			error = EMSGSIZE;
1207 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1208 			goto bad;
1209 		}
1210 
1211 		/*
1212 		 * If the interface will not calculate checksums on
1213 		 * fragmented packets, then do it here.
1214 		 * XXX-BZ handle the hw offloading case.  Need flags.
1215 		 */
1216 		ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1217 		    optlen);
1218 
1219 		/*
1220 		 * Change the next header field of the last header in the
1221 		 * unfragmentable part.
1222 		 */
1223 		if (exthdrs.ip6e_rthdr) {
1224 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1225 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1226 		} else if (exthdrs.ip6e_dest1) {
1227 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1228 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1229 		} else if (exthdrs.ip6e_hbh) {
1230 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1231 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1232 		} else {
1233 			ip6 = mtod(m, struct ip6_hdr *);
1234 			nextproto = ip6->ip6_nxt;
1235 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1236 		}
1237 
1238 		/*
1239 		 * Loop through length of segment after first fragment,
1240 		 * make new header and copy data of each part and link onto
1241 		 * chain.
1242 		 */
1243 		m0 = m;
1244 		id = htonl(ip6_randomid());
1245 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1246 		if (error != 0)
1247 			goto sendorfree;
1248 
1249 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1250 	}
1251 
1252 	/* Remove leading garbage. */
1253 sendorfree:
1254 	m = m0->m_nextpkt;
1255 	m0->m_nextpkt = 0;
1256 	m_freem(m0);
1257 	for (; m; m = m0) {
1258 		m0 = m->m_nextpkt;
1259 		m->m_nextpkt = 0;
1260 		if (error == 0) {
1261 			/* Record statistics for this interface address. */
1262 			if (ia) {
1263 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1264 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1265 				    m->m_pkthdr.len);
1266 			}
1267 			if (vlan_pcp > -1)
1268 				EVL_APPLY_PRI(m, vlan_pcp);
1269 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1270 			    true);
1271 		} else
1272 			m_freem(m);
1273 	}
1274 
1275 	if (error == 0)
1276 		IP6STAT_INC(ip6s_fragmented);
1277 
1278 done:
1279 	return (error);
1280 
1281 freehdrs:
1282 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1283 	m_freem(exthdrs.ip6e_dest1);
1284 	m_freem(exthdrs.ip6e_rthdr);
1285 	m_freem(exthdrs.ip6e_dest2);
1286 	/* FALLTHROUGH */
1287 bad:
1288 	if (m)
1289 		m_freem(m);
1290 	goto done;
1291 }
1292 
1293 static int
ip6_copyexthdr(struct mbuf ** mp,caddr_t hdr,int hlen)1294 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1295 {
1296 	struct mbuf *m;
1297 
1298 	if (hlen > MCLBYTES)
1299 		return (ENOBUFS); /* XXX */
1300 
1301 	if (hlen > MLEN)
1302 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1303 	else
1304 		m = m_get(M_NOWAIT, MT_DATA);
1305 	if (m == NULL)
1306 		return (ENOBUFS);
1307 	m->m_len = hlen;
1308 	if (hdr)
1309 		bcopy(hdr, mtod(m, caddr_t), hlen);
1310 
1311 	*mp = m;
1312 	return (0);
1313 }
1314 
1315 /*
1316  * Insert jumbo payload option.
1317  */
1318 static int
ip6_insert_jumboopt(struct ip6_exthdrs * exthdrs,u_int32_t plen)1319 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1320 {
1321 	struct mbuf *mopt;
1322 	u_char *optbuf;
1323 	u_int32_t v;
1324 
1325 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1326 
1327 	/*
1328 	 * If there is no hop-by-hop options header, allocate new one.
1329 	 * If there is one but it doesn't have enough space to store the
1330 	 * jumbo payload option, allocate a cluster to store the whole options.
1331 	 * Otherwise, use it to store the options.
1332 	 */
1333 	if (exthdrs->ip6e_hbh == NULL) {
1334 		mopt = m_get(M_NOWAIT, MT_DATA);
1335 		if (mopt == NULL)
1336 			return (ENOBUFS);
1337 		mopt->m_len = JUMBOOPTLEN;
1338 		optbuf = mtod(mopt, u_char *);
1339 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1340 		exthdrs->ip6e_hbh = mopt;
1341 	} else {
1342 		struct ip6_hbh *hbh;
1343 
1344 		mopt = exthdrs->ip6e_hbh;
1345 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1346 			/*
1347 			 * XXX assumption:
1348 			 * - exthdrs->ip6e_hbh is not referenced from places
1349 			 *   other than exthdrs.
1350 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1351 			 */
1352 			int oldoptlen = mopt->m_len;
1353 			struct mbuf *n;
1354 
1355 			/*
1356 			 * XXX: give up if the whole (new) hbh header does
1357 			 * not fit even in an mbuf cluster.
1358 			 */
1359 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1360 				return (ENOBUFS);
1361 
1362 			/*
1363 			 * As a consequence, we must always prepare a cluster
1364 			 * at this point.
1365 			 */
1366 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1367 			if (n == NULL)
1368 				return (ENOBUFS);
1369 			n->m_len = oldoptlen + JUMBOOPTLEN;
1370 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1371 			    oldoptlen);
1372 			optbuf = mtod(n, caddr_t) + oldoptlen;
1373 			m_freem(mopt);
1374 			mopt = exthdrs->ip6e_hbh = n;
1375 		} else {
1376 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1377 			mopt->m_len += JUMBOOPTLEN;
1378 		}
1379 		optbuf[0] = IP6OPT_PADN;
1380 		optbuf[1] = 1;
1381 
1382 		/*
1383 		 * Adjust the header length according to the pad and
1384 		 * the jumbo payload option.
1385 		 */
1386 		hbh = mtod(mopt, struct ip6_hbh *);
1387 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1388 	}
1389 
1390 	/* fill in the option. */
1391 	optbuf[2] = IP6OPT_JUMBO;
1392 	optbuf[3] = 4;
1393 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1394 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1395 
1396 	/* finally, adjust the packet header length */
1397 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1398 
1399 	return (0);
1400 #undef JUMBOOPTLEN
1401 }
1402 
1403 /*
1404  * Insert fragment header and copy unfragmentable header portions.
1405  */
1406 static int
ip6_insertfraghdr(struct mbuf * m0,struct mbuf * m,int hlen,struct ip6_frag ** frghdrp)1407 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1408     struct ip6_frag **frghdrp)
1409 {
1410 	struct mbuf *n, *mlast;
1411 
1412 	if (hlen > sizeof(struct ip6_hdr)) {
1413 		n = m_copym(m0, sizeof(struct ip6_hdr),
1414 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1415 		if (n == NULL)
1416 			return (ENOBUFS);
1417 		m->m_next = n;
1418 	} else
1419 		n = m;
1420 
1421 	/* Search for the last mbuf of unfragmentable part. */
1422 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1423 		;
1424 
1425 	if (M_WRITABLE(mlast) &&
1426 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1427 		/* use the trailing space of the last mbuf for the fragment hdr */
1428 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1429 		    mlast->m_len);
1430 		mlast->m_len += sizeof(struct ip6_frag);
1431 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1432 	} else {
1433 		/* allocate a new mbuf for the fragment header */
1434 		struct mbuf *mfrg;
1435 
1436 		mfrg = m_get(M_NOWAIT, MT_DATA);
1437 		if (mfrg == NULL)
1438 			return (ENOBUFS);
1439 		mfrg->m_len = sizeof(struct ip6_frag);
1440 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1441 		mlast->m_next = mfrg;
1442 	}
1443 
1444 	return (0);
1445 }
1446 
1447 /*
1448  * Calculates IPv6 path mtu for destination @dst.
1449  * Resulting MTU is stored in @mtup.
1450  *
1451  * Returns 0 on success.
1452  */
1453 static int
ip6_getpmtu_ctl(u_int fibnum,const struct in6_addr * dst,u_long * mtup)1454 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1455 {
1456 	struct epoch_tracker et;
1457 	struct nhop_object *nh;
1458 	struct in6_addr kdst;
1459 	uint32_t scopeid;
1460 	int error;
1461 
1462 	in6_splitscope(dst, &kdst, &scopeid);
1463 
1464 	NET_EPOCH_ENTER(et);
1465 	nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1466 	if (nh != NULL) {
1467 		ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, 0);
1468 		error = 0;
1469 	} else
1470 		error = EHOSTUNREACH;
1471 	NET_EPOCH_EXIT(et);
1472 
1473 	return (error);
1474 }
1475 
1476 /*
1477  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1478  * and cached data in @ro_pmtu.
1479  * MTU from (successful) route lookup is saved (along with dst)
1480  * inside @ro_pmtu to avoid subsequent route lookups after packet
1481  * filter processing.
1482  *
1483  * Stores mtu into @mtup.
1484  */
1485 static void
ip6_getpmtu(struct route_in6 * ro_pmtu,int do_lookup,struct ifnet * ifp,const struct in6_addr * dst,u_long * mtup,u_int fibnum,u_int proto)1486 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1487     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1488     u_int fibnum, u_int proto)
1489 {
1490 	struct nhop_object *nh;
1491 	struct in6_addr kdst;
1492 	uint32_t scopeid;
1493 	struct sockaddr_in6 *sa6_dst, sin6;
1494 	u_long mtu;
1495 
1496 	NET_EPOCH_ASSERT();
1497 
1498 	mtu = 0;
1499 	if (ro_pmtu == NULL || do_lookup) {
1500 		/*
1501 		 * Here ro_pmtu has final destination address, while
1502 		 * ro might represent immediate destination.
1503 		 * Use ro_pmtu destination since mtu might differ.
1504 		 */
1505 		if (ro_pmtu != NULL) {
1506 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1507 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1508 				ro_pmtu->ro_mtu = 0;
1509 		} else
1510 			sa6_dst = &sin6;
1511 
1512 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1513 			bzero(sa6_dst, sizeof(*sa6_dst));
1514 			sa6_dst->sin6_family = AF_INET6;
1515 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1516 			sa6_dst->sin6_addr = *dst;
1517 
1518 			in6_splitscope(dst, &kdst, &scopeid);
1519 			nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1520 			if (nh != NULL) {
1521 				mtu = nh->nh_mtu;
1522 				if (ro_pmtu != NULL)
1523 					ro_pmtu->ro_mtu = mtu;
1524 			}
1525 		} else
1526 			mtu = ro_pmtu->ro_mtu;
1527 	}
1528 
1529 	if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1530 		mtu = ro_pmtu->ro_nh->nh_mtu;
1531 
1532 	ip6_calcmtu(ifp, dst, mtu, mtup, proto);
1533 }
1534 
1535 /*
1536  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1537  * hostcache data for @dst.
1538  * Stores mtu into @mtup.
1539  */
1540 static void
ip6_calcmtu(struct ifnet * ifp,const struct in6_addr * dst,u_long rt_mtu,u_long * mtup,u_int proto)1541 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1542     u_long *mtup, u_int proto)
1543 {
1544 	u_long mtu = 0;
1545 
1546 	if (rt_mtu > 0) {
1547 		/* Skip the hostcache if the protocol handles PMTU changes. */
1548 		if (proto != IPPROTO_TCP && proto != IPPROTO_SCTP) {
1549 			struct in_conninfo inc = {
1550 			    .inc_flags = INC_ISIPV6,
1551 			    .inc6_faddr = *dst,
1552 			};
1553 
1554 			mtu = tcp_hc_getmtu(&inc);
1555 		}
1556 
1557 		if (mtu)
1558 			mtu = min(mtu, rt_mtu);
1559 		else
1560 			mtu = rt_mtu;
1561 	}
1562 
1563 	if (mtu == 0)
1564 		mtu = IN6_LINKMTU(ifp);
1565 
1566 	*mtup = mtu;
1567 }
1568 
1569 /*
1570  * IP6 socket option processing.
1571  */
1572 int
ip6_ctloutput(struct socket * so,struct sockopt * sopt)1573 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1574 {
1575 	int optdatalen, uproto;
1576 	void *optdata;
1577 	struct inpcb *inp = sotoinpcb(so);
1578 	int error, optval;
1579 	int level, op, optname;
1580 	int optlen;
1581 	struct thread *td;
1582 #ifdef	RSS
1583 	uint32_t rss_bucket;
1584 	int retval;
1585 #endif
1586 
1587 /*
1588  * Don't use more than a quarter of mbuf clusters.  N.B.:
1589  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1590  * on LP64 architectures, so cast to u_long to avoid undefined
1591  * behavior.  ILP32 architectures cannot have nmbclusters
1592  * large enough to overflow for other reasons.
1593  */
1594 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1595 
1596 	level = sopt->sopt_level;
1597 	op = sopt->sopt_dir;
1598 	optname = sopt->sopt_name;
1599 	optlen = sopt->sopt_valsize;
1600 	td = sopt->sopt_td;
1601 	error = 0;
1602 	optval = 0;
1603 	uproto = (int)so->so_proto->pr_protocol;
1604 
1605 	if (level != IPPROTO_IPV6) {
1606 		error = EINVAL;
1607 
1608 		if (sopt->sopt_level == SOL_SOCKET &&
1609 		    sopt->sopt_dir == SOPT_SET) {
1610 			switch (sopt->sopt_name) {
1611 			case SO_SETFIB:
1612 				error = sooptcopyin(sopt, &optval,
1613 				    sizeof(optval), sizeof(optval));
1614 				if (error != 0)
1615 					break;
1616 
1617 				INP_WLOCK(inp);
1618 				if ((inp->inp_flags & INP_BOUNDFIB) != 0 &&
1619 				    optval != so->so_fibnum) {
1620 					INP_WUNLOCK(inp);
1621 					error = EISCONN;
1622 					break;
1623 				}
1624 				error = sosetfib(inp->inp_socket, optval);
1625 				if (error == 0)
1626 					inp->inp_inc.inc_fibnum = optval;
1627 				INP_WUNLOCK(inp);
1628 				break;
1629 			case SO_MAX_PACING_RATE:
1630 #ifdef RATELIMIT
1631 				INP_WLOCK(inp);
1632 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1633 				INP_WUNLOCK(inp);
1634 				error = 0;
1635 #else
1636 				error = EOPNOTSUPP;
1637 #endif
1638 				break;
1639 			default:
1640 				break;
1641 			}
1642 		}
1643 	} else {		/* level == IPPROTO_IPV6 */
1644 		switch (op) {
1645 		case SOPT_SET:
1646 			switch (optname) {
1647 			case IPV6_2292PKTOPTIONS:
1648 #ifdef IPV6_PKTOPTIONS
1649 			case IPV6_PKTOPTIONS:
1650 #endif
1651 			{
1652 				struct mbuf *m;
1653 
1654 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1655 					printf("ip6_ctloutput: mbuf limit hit\n");
1656 					error = ENOBUFS;
1657 					break;
1658 				}
1659 
1660 				error = soopt_getm(sopt, &m); /* XXX */
1661 				if (error != 0)
1662 					break;
1663 				error = soopt_mcopyin(sopt, m); /* XXX */
1664 				if (error != 0)
1665 					break;
1666 				INP_WLOCK(inp);
1667 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1668 				    so, sopt);
1669 				INP_WUNLOCK(inp);
1670 				m_freem(m); /* XXX */
1671 				break;
1672 			}
1673 
1674 			/*
1675 			 * Use of some Hop-by-Hop options or some
1676 			 * Destination options, might require special
1677 			 * privilege.  That is, normal applications
1678 			 * (without special privilege) might be forbidden
1679 			 * from setting certain options in outgoing packets,
1680 			 * and might never see certain options in received
1681 			 * packets. [RFC 2292 Section 6]
1682 			 * KAME specific note:
1683 			 *  KAME prevents non-privileged users from sending or
1684 			 *  receiving ANY hbh/dst options in order to avoid
1685 			 *  overhead of parsing options in the kernel.
1686 			 */
1687 			case IPV6_RECVHOPOPTS:
1688 			case IPV6_RECVDSTOPTS:
1689 			case IPV6_RECVRTHDRDSTOPTS:
1690 				if (td != NULL) {
1691 					error = priv_check(td,
1692 					    PRIV_NETINET_SETHDROPTS);
1693 					if (error)
1694 						break;
1695 				}
1696 				/* FALLTHROUGH */
1697 			case IPV6_UNICAST_HOPS:
1698 			case IPV6_HOPLIMIT:
1699 
1700 			case IPV6_RECVPKTINFO:
1701 			case IPV6_RECVHOPLIMIT:
1702 			case IPV6_RECVRTHDR:
1703 			case IPV6_RECVPATHMTU:
1704 			case IPV6_RECVTCLASS:
1705 			case IPV6_RECVFLOWID:
1706 #ifdef	RSS
1707 			case IPV6_RECVRSSBUCKETID:
1708 #endif
1709 			case IPV6_V6ONLY:
1710 			case IPV6_AUTOFLOWLABEL:
1711 			case IPV6_ORIGDSTADDR:
1712 			case IPV6_BINDANY:
1713 			case IPV6_VLAN_PCP:
1714 				if (optname == IPV6_BINDANY && td != NULL) {
1715 					error = priv_check(td,
1716 					    PRIV_NETINET_BINDANY);
1717 					if (error)
1718 						break;
1719 				}
1720 
1721 				if (optlen != sizeof(int)) {
1722 					error = EINVAL;
1723 					break;
1724 				}
1725 				error = sooptcopyin(sopt, &optval,
1726 					sizeof optval, sizeof optval);
1727 				if (error)
1728 					break;
1729 				switch (optname) {
1730 				case IPV6_UNICAST_HOPS:
1731 					if (optval < -1 || optval >= 256)
1732 						error = EINVAL;
1733 					else {
1734 						/* -1 = kernel default */
1735 						inp->in6p_hops = optval;
1736 						if ((inp->inp_vflag &
1737 						     INP_IPV4) != 0)
1738 							inp->inp_ip_ttl = optval;
1739 					}
1740 					break;
1741 #define OPTSET(bit) \
1742 do { \
1743 	INP_WLOCK(inp); \
1744 	if (optval) \
1745 		inp->inp_flags |= (bit); \
1746 	else \
1747 		inp->inp_flags &= ~(bit); \
1748 	INP_WUNLOCK(inp); \
1749 } while (/*CONSTCOND*/ 0)
1750 #define OPTSET2292(bit) \
1751 do { \
1752 	INP_WLOCK(inp); \
1753 	inp->inp_flags |= IN6P_RFC2292; \
1754 	if (optval) \
1755 		inp->inp_flags |= (bit); \
1756 	else \
1757 		inp->inp_flags &= ~(bit); \
1758 	INP_WUNLOCK(inp); \
1759 } while (/*CONSTCOND*/ 0)
1760 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1761 
1762 #define OPTSET2_N(bit, val) do {					\
1763 	if (val)							\
1764 		inp->inp_flags2 |= bit;					\
1765 	else								\
1766 		inp->inp_flags2 &= ~bit;				\
1767 } while (0)
1768 #define OPTSET2(bit, val) do {						\
1769 	INP_WLOCK(inp);							\
1770 	OPTSET2_N(bit, val);						\
1771 	INP_WUNLOCK(inp);						\
1772 } while (0)
1773 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1774 #define OPTSET2292_EXCLUSIVE(bit)					\
1775 do {									\
1776 	INP_WLOCK(inp);							\
1777 	if (OPTBIT(IN6P_RFC2292)) {					\
1778 		error = EINVAL;						\
1779 	} else {							\
1780 		if (optval)						\
1781 			inp->inp_flags |= (bit);			\
1782 		else							\
1783 			inp->inp_flags &= ~(bit);			\
1784 	}								\
1785 	INP_WUNLOCK(inp);						\
1786 } while (/*CONSTCOND*/ 0)
1787 
1788 				case IPV6_RECVPKTINFO:
1789 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1790 					break;
1791 
1792 				case IPV6_HOPLIMIT:
1793 				{
1794 					struct ip6_pktopts **optp;
1795 
1796 					/* cannot mix with RFC2292 */
1797 					if (OPTBIT(IN6P_RFC2292)) {
1798 						error = EINVAL;
1799 						break;
1800 					}
1801 					INP_WLOCK(inp);
1802 					if (inp->inp_flags & INP_DROPPED) {
1803 						INP_WUNLOCK(inp);
1804 						return (ECONNRESET);
1805 					}
1806 					optp = &inp->in6p_outputopts;
1807 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1808 					    (u_char *)&optval, sizeof(optval),
1809 					    optp, (td != NULL) ? td->td_ucred :
1810 					    NULL, uproto);
1811 					INP_WUNLOCK(inp);
1812 					break;
1813 				}
1814 
1815 				case IPV6_RECVHOPLIMIT:
1816 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1817 					break;
1818 
1819 				case IPV6_RECVHOPOPTS:
1820 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1821 					break;
1822 
1823 				case IPV6_RECVDSTOPTS:
1824 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1825 					break;
1826 
1827 				case IPV6_RECVRTHDRDSTOPTS:
1828 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1829 					break;
1830 
1831 				case IPV6_RECVRTHDR:
1832 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1833 					break;
1834 
1835 				case IPV6_RECVPATHMTU:
1836 					/*
1837 					 * We ignore this option for TCP
1838 					 * sockets.
1839 					 * (RFC3542 leaves this case
1840 					 * unspecified.)
1841 					 */
1842 					if (uproto != IPPROTO_TCP)
1843 						OPTSET(IN6P_MTU);
1844 					break;
1845 
1846 				case IPV6_RECVFLOWID:
1847 					OPTSET2(INP_RECVFLOWID, optval);
1848 					break;
1849 
1850 #ifdef	RSS
1851 				case IPV6_RECVRSSBUCKETID:
1852 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1853 					break;
1854 #endif
1855 
1856 				case IPV6_V6ONLY:
1857 					INP_WLOCK(inp);
1858 					if (inp->inp_lport ||
1859 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1860 						/*
1861 						 * The socket is already bound.
1862 						 */
1863 						INP_WUNLOCK(inp);
1864 						error = EINVAL;
1865 						break;
1866 					}
1867 					if (optval) {
1868 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1869 						inp->inp_vflag &= ~INP_IPV4;
1870 					} else {
1871 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1872 						inp->inp_vflag |= INP_IPV4;
1873 					}
1874 					INP_WUNLOCK(inp);
1875 					break;
1876 				case IPV6_RECVTCLASS:
1877 					/* cannot mix with RFC2292 XXX */
1878 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1879 					break;
1880 				case IPV6_AUTOFLOWLABEL:
1881 					OPTSET(IN6P_AUTOFLOWLABEL);
1882 					break;
1883 
1884 				case IPV6_ORIGDSTADDR:
1885 					OPTSET2(INP_ORIGDSTADDR, optval);
1886 					break;
1887 				case IPV6_BINDANY:
1888 					OPTSET(INP_BINDANY);
1889 					break;
1890 				case IPV6_VLAN_PCP:
1891 					if ((optval >= -1) && (optval <=
1892 					    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1893 						if (optval == -1) {
1894 							INP_WLOCK(inp);
1895 							inp->inp_flags2 &=
1896 							    ~(INP_2PCP_SET |
1897 							    INP_2PCP_MASK);
1898 							INP_WUNLOCK(inp);
1899 						} else {
1900 							INP_WLOCK(inp);
1901 							inp->inp_flags2 |=
1902 							    INP_2PCP_SET;
1903 							inp->inp_flags2 &=
1904 							    ~INP_2PCP_MASK;
1905 							inp->inp_flags2 |=
1906 							    optval <<
1907 							    INP_2PCP_SHIFT;
1908 							INP_WUNLOCK(inp);
1909 						}
1910 					} else
1911 						error = EINVAL;
1912 					break;
1913 				}
1914 				break;
1915 
1916 			case IPV6_TCLASS:
1917 			case IPV6_DONTFRAG:
1918 			case IPV6_USE_MIN_MTU:
1919 			case IPV6_PREFER_TEMPADDR:
1920 				if (optlen != sizeof(optval)) {
1921 					error = EINVAL;
1922 					break;
1923 				}
1924 				error = sooptcopyin(sopt, &optval,
1925 					sizeof optval, sizeof optval);
1926 				if (error)
1927 					break;
1928 				{
1929 					struct ip6_pktopts **optp;
1930 					INP_WLOCK(inp);
1931 					if (inp->inp_flags & INP_DROPPED) {
1932 						INP_WUNLOCK(inp);
1933 						return (ECONNRESET);
1934 					}
1935 					optp = &inp->in6p_outputopts;
1936 					error = ip6_pcbopt(optname,
1937 					    (u_char *)&optval, sizeof(optval),
1938 					    optp, (td != NULL) ? td->td_ucred :
1939 					    NULL, uproto);
1940 					INP_WUNLOCK(inp);
1941 					break;
1942 				}
1943 
1944 			case IPV6_2292PKTINFO:
1945 			case IPV6_2292HOPLIMIT:
1946 			case IPV6_2292HOPOPTS:
1947 			case IPV6_2292DSTOPTS:
1948 			case IPV6_2292RTHDR:
1949 				/* RFC 2292 */
1950 				if (optlen != sizeof(int)) {
1951 					error = EINVAL;
1952 					break;
1953 				}
1954 				error = sooptcopyin(sopt, &optval,
1955 					sizeof optval, sizeof optval);
1956 				if (error)
1957 					break;
1958 				switch (optname) {
1959 				case IPV6_2292PKTINFO:
1960 					OPTSET2292(IN6P_PKTINFO);
1961 					break;
1962 				case IPV6_2292HOPLIMIT:
1963 					OPTSET2292(IN6P_HOPLIMIT);
1964 					break;
1965 				case IPV6_2292HOPOPTS:
1966 					/*
1967 					 * Check super-user privilege.
1968 					 * See comments for IPV6_RECVHOPOPTS.
1969 					 */
1970 					if (td != NULL) {
1971 						error = priv_check(td,
1972 						    PRIV_NETINET_SETHDROPTS);
1973 						if (error)
1974 							return (error);
1975 					}
1976 					OPTSET2292(IN6P_HOPOPTS);
1977 					break;
1978 				case IPV6_2292DSTOPTS:
1979 					if (td != NULL) {
1980 						error = priv_check(td,
1981 						    PRIV_NETINET_SETHDROPTS);
1982 						if (error)
1983 							return (error);
1984 					}
1985 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1986 					break;
1987 				case IPV6_2292RTHDR:
1988 					OPTSET2292(IN6P_RTHDR);
1989 					break;
1990 				}
1991 				break;
1992 			case IPV6_PKTINFO:
1993 			case IPV6_HOPOPTS:
1994 			case IPV6_RTHDR:
1995 			case IPV6_DSTOPTS:
1996 			case IPV6_RTHDRDSTOPTS:
1997 			case IPV6_NEXTHOP:
1998 			{
1999 				/* new advanced API (RFC3542) */
2000 				u_char *optbuf;
2001 				u_char optbuf_storage[MCLBYTES];
2002 				int optlen;
2003 				struct ip6_pktopts **optp;
2004 
2005 				/* cannot mix with RFC2292 */
2006 				if (OPTBIT(IN6P_RFC2292)) {
2007 					error = EINVAL;
2008 					break;
2009 				}
2010 
2011 				/*
2012 				 * We only ensure valsize is not too large
2013 				 * here.  Further validation will be done
2014 				 * later.
2015 				 */
2016 				error = sooptcopyin(sopt, optbuf_storage,
2017 				    sizeof(optbuf_storage), 0);
2018 				if (error)
2019 					break;
2020 				optlen = sopt->sopt_valsize;
2021 				optbuf = optbuf_storage;
2022 				INP_WLOCK(inp);
2023 				if (inp->inp_flags & INP_DROPPED) {
2024 					INP_WUNLOCK(inp);
2025 					return (ECONNRESET);
2026 				}
2027 				optp = &inp->in6p_outputopts;
2028 				error = ip6_pcbopt(optname, optbuf, optlen,
2029 				    optp, (td != NULL) ? td->td_ucred : NULL,
2030 				    uproto);
2031 				INP_WUNLOCK(inp);
2032 				break;
2033 			}
2034 #undef OPTSET
2035 
2036 			case IPV6_MULTICAST_IF:
2037 			case IPV6_MULTICAST_HOPS:
2038 			case IPV6_MULTICAST_LOOP:
2039 			case IPV6_JOIN_GROUP:
2040 			case IPV6_LEAVE_GROUP:
2041 			case IPV6_MSFILTER:
2042 			case MCAST_BLOCK_SOURCE:
2043 			case MCAST_UNBLOCK_SOURCE:
2044 			case MCAST_JOIN_GROUP:
2045 			case MCAST_LEAVE_GROUP:
2046 			case MCAST_JOIN_SOURCE_GROUP:
2047 			case MCAST_LEAVE_SOURCE_GROUP:
2048 				error = ip6_setmoptions(inp, sopt);
2049 				break;
2050 
2051 			case IPV6_PORTRANGE:
2052 				error = sooptcopyin(sopt, &optval,
2053 				    sizeof optval, sizeof optval);
2054 				if (error)
2055 					break;
2056 
2057 				INP_WLOCK(inp);
2058 				switch (optval) {
2059 				case IPV6_PORTRANGE_DEFAULT:
2060 					inp->inp_flags &= ~(INP_LOWPORT);
2061 					inp->inp_flags &= ~(INP_HIGHPORT);
2062 					break;
2063 
2064 				case IPV6_PORTRANGE_HIGH:
2065 					inp->inp_flags &= ~(INP_LOWPORT);
2066 					inp->inp_flags |= INP_HIGHPORT;
2067 					break;
2068 
2069 				case IPV6_PORTRANGE_LOW:
2070 					inp->inp_flags &= ~(INP_HIGHPORT);
2071 					inp->inp_flags |= INP_LOWPORT;
2072 					break;
2073 
2074 				default:
2075 					error = EINVAL;
2076 					break;
2077 				}
2078 				INP_WUNLOCK(inp);
2079 				break;
2080 
2081 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2082 			case IPV6_IPSEC_POLICY:
2083 				if (IPSEC_ENABLED(ipv6)) {
2084 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2085 					break;
2086 				}
2087 				/* FALLTHROUGH */
2088 #endif /* IPSEC */
2089 
2090 			default:
2091 				error = ENOPROTOOPT;
2092 				break;
2093 			}
2094 			break;
2095 
2096 		case SOPT_GET:
2097 			switch (optname) {
2098 			case IPV6_2292PKTOPTIONS:
2099 #ifdef IPV6_PKTOPTIONS
2100 			case IPV6_PKTOPTIONS:
2101 #endif
2102 				/*
2103 				 * RFC3542 (effectively) deprecated the
2104 				 * semantics of the 2292-style pktoptions.
2105 				 * Since it was not reliable in nature (i.e.,
2106 				 * applications had to expect the lack of some
2107 				 * information after all), it would make sense
2108 				 * to simplify this part by always returning
2109 				 * empty data.
2110 				 */
2111 				sopt->sopt_valsize = 0;
2112 				break;
2113 
2114 			case IPV6_RECVHOPOPTS:
2115 			case IPV6_RECVDSTOPTS:
2116 			case IPV6_RECVRTHDRDSTOPTS:
2117 			case IPV6_UNICAST_HOPS:
2118 			case IPV6_RECVPKTINFO:
2119 			case IPV6_RECVHOPLIMIT:
2120 			case IPV6_RECVRTHDR:
2121 			case IPV6_RECVPATHMTU:
2122 
2123 			case IPV6_V6ONLY:
2124 			case IPV6_PORTRANGE:
2125 			case IPV6_RECVTCLASS:
2126 			case IPV6_AUTOFLOWLABEL:
2127 			case IPV6_BINDANY:
2128 			case IPV6_FLOWID:
2129 			case IPV6_FLOWTYPE:
2130 			case IPV6_RECVFLOWID:
2131 #ifdef	RSS
2132 			case IPV6_RSSBUCKETID:
2133 			case IPV6_RECVRSSBUCKETID:
2134 #endif
2135 			case IPV6_VLAN_PCP:
2136 				switch (optname) {
2137 				case IPV6_RECVHOPOPTS:
2138 					optval = OPTBIT(IN6P_HOPOPTS);
2139 					break;
2140 
2141 				case IPV6_RECVDSTOPTS:
2142 					optval = OPTBIT(IN6P_DSTOPTS);
2143 					break;
2144 
2145 				case IPV6_RECVRTHDRDSTOPTS:
2146 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2147 					break;
2148 
2149 				case IPV6_UNICAST_HOPS:
2150 					optval = inp->in6p_hops;
2151 					break;
2152 
2153 				case IPV6_RECVPKTINFO:
2154 					optval = OPTBIT(IN6P_PKTINFO);
2155 					break;
2156 
2157 				case IPV6_RECVHOPLIMIT:
2158 					optval = OPTBIT(IN6P_HOPLIMIT);
2159 					break;
2160 
2161 				case IPV6_RECVRTHDR:
2162 					optval = OPTBIT(IN6P_RTHDR);
2163 					break;
2164 
2165 				case IPV6_RECVPATHMTU:
2166 					optval = OPTBIT(IN6P_MTU);
2167 					break;
2168 
2169 				case IPV6_V6ONLY:
2170 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2171 					break;
2172 
2173 				case IPV6_PORTRANGE:
2174 				    {
2175 					int flags;
2176 					flags = inp->inp_flags;
2177 					if (flags & INP_HIGHPORT)
2178 						optval = IPV6_PORTRANGE_HIGH;
2179 					else if (flags & INP_LOWPORT)
2180 						optval = IPV6_PORTRANGE_LOW;
2181 					else
2182 						optval = 0;
2183 					break;
2184 				    }
2185 				case IPV6_RECVTCLASS:
2186 					optval = OPTBIT(IN6P_TCLASS);
2187 					break;
2188 
2189 				case IPV6_AUTOFLOWLABEL:
2190 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2191 					break;
2192 
2193 				case IPV6_ORIGDSTADDR:
2194 					optval = OPTBIT2(INP_ORIGDSTADDR);
2195 					break;
2196 
2197 				case IPV6_BINDANY:
2198 					optval = OPTBIT(INP_BINDANY);
2199 					break;
2200 
2201 				case IPV6_FLOWID:
2202 					optval = inp->inp_flowid;
2203 					break;
2204 
2205 				case IPV6_FLOWTYPE:
2206 					optval = inp->inp_flowtype;
2207 					break;
2208 
2209 				case IPV6_RECVFLOWID:
2210 					optval = OPTBIT2(INP_RECVFLOWID);
2211 					break;
2212 #ifdef	RSS
2213 				case IPV6_RSSBUCKETID:
2214 					retval =
2215 					    rss_hash2bucket(inp->inp_flowid,
2216 					    inp->inp_flowtype,
2217 					    &rss_bucket);
2218 					if (retval == 0)
2219 						optval = rss_bucket;
2220 					else
2221 						error = EINVAL;
2222 					break;
2223 
2224 				case IPV6_RECVRSSBUCKETID:
2225 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2226 					break;
2227 #endif
2228 
2229 
2230 				case IPV6_VLAN_PCP:
2231 					if (OPTBIT2(INP_2PCP_SET)) {
2232 						optval = (inp->inp_flags2 &
2233 							    INP_2PCP_MASK) >>
2234 							    INP_2PCP_SHIFT;
2235 					} else {
2236 						optval = -1;
2237 					}
2238 					break;
2239 				}
2240 
2241 				if (error)
2242 					break;
2243 				error = sooptcopyout(sopt, &optval,
2244 					sizeof optval);
2245 				break;
2246 
2247 			case IPV6_PATHMTU:
2248 			{
2249 				u_long pmtu = 0;
2250 				struct ip6_mtuinfo mtuinfo;
2251 				struct in6_addr addr;
2252 
2253 				if (!(so->so_state & SS_ISCONNECTED))
2254 					return (ENOTCONN);
2255 				/*
2256 				 * XXX: we dot not consider the case of source
2257 				 * routing, or optional information to specify
2258 				 * the outgoing interface.
2259 				 * Copy faddr out of inp to avoid holding lock
2260 				 * on inp during route lookup.
2261 				 */
2262 				INP_RLOCK(inp);
2263 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2264 				INP_RUNLOCK(inp);
2265 				error = ip6_getpmtu_ctl(so->so_fibnum,
2266 				    &addr, &pmtu);
2267 				if (error)
2268 					break;
2269 				if (pmtu > IPV6_MAXPACKET)
2270 					pmtu = IPV6_MAXPACKET;
2271 
2272 				bzero(&mtuinfo, sizeof(mtuinfo));
2273 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2274 				optdata = (void *)&mtuinfo;
2275 				optdatalen = sizeof(mtuinfo);
2276 				error = sooptcopyout(sopt, optdata,
2277 				    optdatalen);
2278 				break;
2279 			}
2280 
2281 			case IPV6_2292PKTINFO:
2282 			case IPV6_2292HOPLIMIT:
2283 			case IPV6_2292HOPOPTS:
2284 			case IPV6_2292RTHDR:
2285 			case IPV6_2292DSTOPTS:
2286 				switch (optname) {
2287 				case IPV6_2292PKTINFO:
2288 					optval = OPTBIT(IN6P_PKTINFO);
2289 					break;
2290 				case IPV6_2292HOPLIMIT:
2291 					optval = OPTBIT(IN6P_HOPLIMIT);
2292 					break;
2293 				case IPV6_2292HOPOPTS:
2294 					optval = OPTBIT(IN6P_HOPOPTS);
2295 					break;
2296 				case IPV6_2292RTHDR:
2297 					optval = OPTBIT(IN6P_RTHDR);
2298 					break;
2299 				case IPV6_2292DSTOPTS:
2300 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2301 					break;
2302 				}
2303 				error = sooptcopyout(sopt, &optval,
2304 				    sizeof optval);
2305 				break;
2306 			case IPV6_PKTINFO:
2307 			case IPV6_HOPOPTS:
2308 			case IPV6_RTHDR:
2309 			case IPV6_DSTOPTS:
2310 			case IPV6_RTHDRDSTOPTS:
2311 			case IPV6_NEXTHOP:
2312 			case IPV6_TCLASS:
2313 			case IPV6_DONTFRAG:
2314 			case IPV6_USE_MIN_MTU:
2315 			case IPV6_PREFER_TEMPADDR:
2316 				error = ip6_getpcbopt(inp, optname, sopt);
2317 				break;
2318 
2319 			case IPV6_MULTICAST_IF:
2320 			case IPV6_MULTICAST_HOPS:
2321 			case IPV6_MULTICAST_LOOP:
2322 			case IPV6_MSFILTER:
2323 				error = ip6_getmoptions(inp, sopt);
2324 				break;
2325 
2326 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2327 			case IPV6_IPSEC_POLICY:
2328 				if (IPSEC_ENABLED(ipv6)) {
2329 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2330 					break;
2331 				}
2332 				/* FALLTHROUGH */
2333 #endif /* IPSEC */
2334 			default:
2335 				error = ENOPROTOOPT;
2336 				break;
2337 			}
2338 			break;
2339 		}
2340 	}
2341 	return (error);
2342 }
2343 
2344 int
ip6_raw_ctloutput(struct socket * so,struct sockopt * sopt)2345 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2346 {
2347 	int error = 0, optval, optlen;
2348 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2349 	struct inpcb *inp = sotoinpcb(so);
2350 	int level, op, optname;
2351 
2352 	level = sopt->sopt_level;
2353 	op = sopt->sopt_dir;
2354 	optname = sopt->sopt_name;
2355 	optlen = sopt->sopt_valsize;
2356 
2357 	if (level != IPPROTO_IPV6) {
2358 		return (EINVAL);
2359 	}
2360 
2361 	switch (optname) {
2362 	case IPV6_CHECKSUM:
2363 		/*
2364 		 * For ICMPv6 sockets, no modification allowed for checksum
2365 		 * offset, permit "no change" values to help existing apps.
2366 		 *
2367 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2368 		 * for an ICMPv6 socket will fail."
2369 		 * The current behavior does not meet RFC3542.
2370 		 */
2371 		switch (op) {
2372 		case SOPT_SET:
2373 			if (optlen != sizeof(int)) {
2374 				error = EINVAL;
2375 				break;
2376 			}
2377 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2378 					    sizeof(optval));
2379 			if (error)
2380 				break;
2381 			if (optval < -1 || (optval % 2) != 0) {
2382 				/*
2383 				 * The API assumes non-negative even offset
2384 				 * values or -1 as a special value.
2385 				 */
2386 				error = EINVAL;
2387 			} else if (inp->inp_ip_p == IPPROTO_ICMPV6) {
2388 				if (optval != icmp6off)
2389 					error = EINVAL;
2390 			} else
2391 				inp->in6p_cksum = optval;
2392 			break;
2393 
2394 		case SOPT_GET:
2395 			if (inp->inp_ip_p == IPPROTO_ICMPV6)
2396 				optval = icmp6off;
2397 			else
2398 				optval = inp->in6p_cksum;
2399 
2400 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2401 			break;
2402 
2403 		default:
2404 			error = EINVAL;
2405 			break;
2406 		}
2407 		break;
2408 
2409 	default:
2410 		error = ENOPROTOOPT;
2411 		break;
2412 	}
2413 
2414 	return (error);
2415 }
2416 
2417 /*
2418  * Set up IP6 options in pcb for insertion in output packets or
2419  * specifying behavior of outgoing packets.
2420  */
2421 static int
ip6_pcbopts(struct ip6_pktopts ** pktopt,struct mbuf * m,struct socket * so,struct sockopt * sopt)2422 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2423     struct socket *so, struct sockopt *sopt)
2424 {
2425 	struct ip6_pktopts *opt = *pktopt;
2426 	int error = 0;
2427 	struct thread *td = sopt->sopt_td;
2428 	struct epoch_tracker et;
2429 
2430 	/* turn off any old options. */
2431 	if (opt) {
2432 #ifdef DIAGNOSTIC
2433 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2434 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2435 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2436 			printf("ip6_pcbopts: all specified options are cleared.\n");
2437 #endif
2438 		ip6_clearpktopts(opt, -1);
2439 	} else {
2440 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2441 		if (opt == NULL)
2442 			return (ENOMEM);
2443 	}
2444 	*pktopt = NULL;
2445 
2446 	if (!m || m->m_len == 0) {
2447 		/*
2448 		 * Only turning off any previous options, regardless of
2449 		 * whether the opt is just created or given.
2450 		 */
2451 		free(opt, M_IP6OPT);
2452 		return (0);
2453 	}
2454 
2455 	/*  set options specified by user. */
2456 	NET_EPOCH_ENTER(et);
2457 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2458 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2459 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2460 		free(opt, M_IP6OPT);
2461 		NET_EPOCH_EXIT(et);
2462 		return (error);
2463 	}
2464 	NET_EPOCH_EXIT(et);
2465 	*pktopt = opt;
2466 	return (0);
2467 }
2468 
2469 /*
2470  * initialize ip6_pktopts.  beware that there are non-zero default values in
2471  * the struct.
2472  */
2473 void
ip6_initpktopts(struct ip6_pktopts * opt)2474 ip6_initpktopts(struct ip6_pktopts *opt)
2475 {
2476 
2477 	bzero(opt, sizeof(*opt));
2478 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2479 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2480 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2481 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2482 }
2483 
2484 static int
ip6_pcbopt(int optname,u_char * buf,int len,struct ip6_pktopts ** pktopt,struct ucred * cred,int uproto)2485 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2486     struct ucred *cred, int uproto)
2487 {
2488 	struct epoch_tracker et;
2489 	struct ip6_pktopts *opt;
2490 	int ret;
2491 
2492 	if (*pktopt == NULL) {
2493 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2494 		    M_NOWAIT);
2495 		if (*pktopt == NULL)
2496 			return (ENOBUFS);
2497 		ip6_initpktopts(*pktopt);
2498 	}
2499 	opt = *pktopt;
2500 
2501 	NET_EPOCH_ENTER(et);
2502 	ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2503 	NET_EPOCH_EXIT(et);
2504 
2505 	return (ret);
2506 }
2507 
2508 #define GET_PKTOPT_VAR(field, lenexpr) do {				\
2509 	if (pktopt && pktopt->field) {					\
2510 		INP_RUNLOCK(inp);					\
2511 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);	\
2512 		malloc_optdata = true;					\
2513 		INP_RLOCK(inp);						\
2514 		if (inp->inp_flags & INP_DROPPED) {			\
2515 			INP_RUNLOCK(inp);				\
2516 			free(optdata, M_TEMP);				\
2517 			return (ECONNRESET);				\
2518 		}							\
2519 		pktopt = inp->in6p_outputopts;				\
2520 		if (pktopt && pktopt->field) {				\
2521 			optdatalen = min(lenexpr, sopt->sopt_valsize);	\
2522 			bcopy(pktopt->field, optdata, optdatalen);	\
2523 		} else {						\
2524 			free(optdata, M_TEMP);				\
2525 			optdata = NULL;					\
2526 			malloc_optdata = false;				\
2527 		}							\
2528 	}								\
2529 } while(0)
2530 
2531 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,			\
2532 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2533 
2534 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,		\
2535 	pktopt->field->sa_len)
2536 
2537 static int
ip6_getpcbopt(struct inpcb * inp,int optname,struct sockopt * sopt)2538 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2539 {
2540 	void *optdata = NULL;
2541 	bool malloc_optdata = false;
2542 	int optdatalen = 0;
2543 	int error = 0;
2544 	struct in6_pktinfo null_pktinfo;
2545 	int deftclass = 0, on;
2546 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2547 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2548 	struct ip6_pktopts *pktopt;
2549 
2550 	INP_RLOCK(inp);
2551 	pktopt = inp->in6p_outputopts;
2552 
2553 	switch (optname) {
2554 	case IPV6_PKTINFO:
2555 		optdata = (void *)&null_pktinfo;
2556 		if (pktopt && pktopt->ip6po_pktinfo) {
2557 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2558 			    sizeof(null_pktinfo));
2559 			in6_clearscope(&null_pktinfo.ipi6_addr);
2560 		} else {
2561 			/* XXX: we don't have to do this every time... */
2562 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2563 		}
2564 		optdatalen = sizeof(struct in6_pktinfo);
2565 		break;
2566 	case IPV6_TCLASS:
2567 		if (pktopt && pktopt->ip6po_tclass >= 0)
2568 			deftclass = pktopt->ip6po_tclass;
2569 		optdata = (void *)&deftclass;
2570 		optdatalen = sizeof(int);
2571 		break;
2572 	case IPV6_HOPOPTS:
2573 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2574 		break;
2575 	case IPV6_RTHDR:
2576 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2577 		break;
2578 	case IPV6_RTHDRDSTOPTS:
2579 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2580 		break;
2581 	case IPV6_DSTOPTS:
2582 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2583 		break;
2584 	case IPV6_NEXTHOP:
2585 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2586 		break;
2587 	case IPV6_USE_MIN_MTU:
2588 		if (pktopt)
2589 			defminmtu = pktopt->ip6po_minmtu;
2590 		optdata = (void *)&defminmtu;
2591 		optdatalen = sizeof(int);
2592 		break;
2593 	case IPV6_DONTFRAG:
2594 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2595 			on = 1;
2596 		else
2597 			on = 0;
2598 		optdata = (void *)&on;
2599 		optdatalen = sizeof(on);
2600 		break;
2601 	case IPV6_PREFER_TEMPADDR:
2602 		if (pktopt)
2603 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2604 		optdata = (void *)&defpreftemp;
2605 		optdatalen = sizeof(int);
2606 		break;
2607 	default:		/* should not happen */
2608 #ifdef DIAGNOSTIC
2609 		panic("ip6_getpcbopt: unexpected option\n");
2610 #endif
2611 		INP_RUNLOCK(inp);
2612 		return (ENOPROTOOPT);
2613 	}
2614 	INP_RUNLOCK(inp);
2615 
2616 	error = sooptcopyout(sopt, optdata, optdatalen);
2617 	if (malloc_optdata)
2618 		free(optdata, M_TEMP);
2619 
2620 	return (error);
2621 }
2622 
2623 void
ip6_clearpktopts(struct ip6_pktopts * pktopt,int optname)2624 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2625 {
2626 	if (pktopt == NULL)
2627 		return;
2628 
2629 	if (optname == -1 || optname == IPV6_PKTINFO) {
2630 		if (pktopt->ip6po_pktinfo)
2631 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2632 		pktopt->ip6po_pktinfo = NULL;
2633 	}
2634 	if (optname == -1 || optname == IPV6_HOPLIMIT) {
2635 		pktopt->ip6po_hlim = -1;
2636 		pktopt->ip6po_valid &= ~IP6PO_VALID_HLIM;
2637 	}
2638 	if (optname == -1 || optname == IPV6_TCLASS) {
2639 		pktopt->ip6po_tclass = -1;
2640 		pktopt->ip6po_valid &= ~IP6PO_VALID_TC;
2641 	}
2642 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2643 		if (pktopt->ip6po_nextroute.ro_nh) {
2644 			NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2645 			pktopt->ip6po_nextroute.ro_nh = NULL;
2646 		}
2647 		if (pktopt->ip6po_nexthop)
2648 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2649 		pktopt->ip6po_nexthop = NULL;
2650 		pktopt->ip6po_valid &= ~IP6PO_VALID_NHINFO;
2651 	}
2652 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2653 		if (pktopt->ip6po_hbh)
2654 			free(pktopt->ip6po_hbh, M_IP6OPT);
2655 		pktopt->ip6po_hbh = NULL;
2656 		pktopt->ip6po_valid &= ~IP6PO_VALID_HBH;
2657 	}
2658 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2659 		if (pktopt->ip6po_dest1)
2660 			free(pktopt->ip6po_dest1, M_IP6OPT);
2661 		pktopt->ip6po_dest1 = NULL;
2662 		pktopt->ip6po_valid &= ~IP6PO_VALID_DEST1;
2663 	}
2664 	if (optname == -1 || optname == IPV6_RTHDR) {
2665 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2666 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2667 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2668 		if (pktopt->ip6po_route.ro_nh) {
2669 			NH_FREE(pktopt->ip6po_route.ro_nh);
2670 			pktopt->ip6po_route.ro_nh = NULL;
2671 		}
2672 		pktopt->ip6po_valid &= ~IP6PO_VALID_RHINFO;
2673 	}
2674 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2675 		if (pktopt->ip6po_dest2)
2676 			free(pktopt->ip6po_dest2, M_IP6OPT);
2677 		pktopt->ip6po_dest2 = NULL;
2678 		pktopt->ip6po_valid &= ~IP6PO_VALID_DEST2;
2679 	}
2680 }
2681 
2682 #define PKTOPT_EXTHDRCPY(type) \
2683 do {\
2684 	if (src->type) {\
2685 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2686 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2687 		if (dst->type == NULL)\
2688 			goto bad;\
2689 		bcopy(src->type, dst->type, hlen);\
2690 	}\
2691 } while (/*CONSTCOND*/ 0)
2692 
2693 static int
copypktopts(struct ip6_pktopts * dst,struct ip6_pktopts * src,int canwait)2694 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2695 {
2696 	if (dst == NULL || src == NULL)  {
2697 		printf("ip6_clearpktopts: invalid argument\n");
2698 		return (EINVAL);
2699 	}
2700 
2701 	dst->ip6po_hlim = src->ip6po_hlim;
2702 	dst->ip6po_tclass = src->ip6po_tclass;
2703 	dst->ip6po_flags = src->ip6po_flags;
2704 	dst->ip6po_minmtu = src->ip6po_minmtu;
2705 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2706 	if (src->ip6po_pktinfo) {
2707 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2708 		    M_IP6OPT, canwait);
2709 		if (dst->ip6po_pktinfo == NULL)
2710 			goto bad;
2711 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2712 	}
2713 	if (src->ip6po_nexthop) {
2714 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2715 		    M_IP6OPT, canwait);
2716 		if (dst->ip6po_nexthop == NULL)
2717 			goto bad;
2718 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2719 		    src->ip6po_nexthop->sa_len);
2720 	}
2721 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2722 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2723 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2724 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2725 	dst->ip6po_valid = src->ip6po_valid;
2726 	return (0);
2727 
2728   bad:
2729 	ip6_clearpktopts(dst, -1);
2730 	return (ENOBUFS);
2731 }
2732 #undef PKTOPT_EXTHDRCPY
2733 
2734 struct ip6_pktopts *
ip6_copypktopts(struct ip6_pktopts * src,int canwait)2735 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2736 {
2737 	int error;
2738 	struct ip6_pktopts *dst;
2739 
2740 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2741 	if (dst == NULL)
2742 		return (NULL);
2743 	ip6_initpktopts(dst);
2744 
2745 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2746 		free(dst, M_IP6OPT);
2747 		return (NULL);
2748 	}
2749 
2750 	return (dst);
2751 }
2752 
2753 void
ip6_freepcbopts(struct ip6_pktopts * pktopt)2754 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2755 {
2756 	if (pktopt == NULL)
2757 		return;
2758 
2759 	ip6_clearpktopts(pktopt, -1);
2760 
2761 	free(pktopt, M_IP6OPT);
2762 }
2763 
2764 /*
2765  * Set IPv6 outgoing packet options based on advanced API.
2766  */
2767 int
ip6_setpktopts(struct mbuf * control,struct ip6_pktopts * opt,struct ip6_pktopts * stickyopt,struct ucred * cred,int uproto)2768 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2769     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2770 {
2771 	struct cmsghdr *cm = NULL;
2772 
2773 	if (control == NULL || opt == NULL)
2774 		return (EINVAL);
2775 
2776 	/*
2777 	 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2778 	 * are in the network epoch here.
2779 	 */
2780 	NET_EPOCH_ASSERT();
2781 
2782 	ip6_initpktopts(opt);
2783 	if (stickyopt) {
2784 		int error;
2785 
2786 		/*
2787 		 * If stickyopt is provided, make a local copy of the options
2788 		 * for this particular packet, then override them by ancillary
2789 		 * objects.
2790 		 * XXX: copypktopts() does not copy the cached route to a next
2791 		 * hop (if any).  This is not very good in terms of efficiency,
2792 		 * but we can allow this since this option should be rarely
2793 		 * used.
2794 		 */
2795 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2796 			return (error);
2797 	}
2798 
2799 	/*
2800 	 * XXX: Currently, we assume all the optional information is stored
2801 	 * in a single mbuf.
2802 	 */
2803 	if (control->m_next)
2804 		return (EINVAL);
2805 
2806 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2807 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2808 		int error;
2809 
2810 		if (control->m_len < CMSG_LEN(0))
2811 			return (EINVAL);
2812 
2813 		cm = mtod(control, struct cmsghdr *);
2814 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2815 			return (EINVAL);
2816 		if (cm->cmsg_level != IPPROTO_IPV6)
2817 			continue;
2818 
2819 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2820 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2821 		if (error)
2822 			return (error);
2823 	}
2824 
2825 	return (0);
2826 }
2827 
2828 /*
2829  * Set a particular packet option, as a sticky option or an ancillary data
2830  * item.  "len" can be 0 only when it's a sticky option.
2831  * We have 4 cases of combination of "sticky" and "cmsg":
2832  * "sticky=0, cmsg=0": impossible
2833  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2834  * "sticky=1, cmsg=0": RFC3542 socket option
2835  * "sticky=1, cmsg=1": RFC2292 socket option
2836  */
2837 static int
ip6_setpktopt(int optname,u_char * buf,int len,struct ip6_pktopts * opt,struct ucred * cred,int sticky,int cmsg,int uproto)2838 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2839     struct ucred *cred, int sticky, int cmsg, int uproto)
2840 {
2841 	int minmtupolicy, preftemp;
2842 	int error;
2843 
2844 	NET_EPOCH_ASSERT();
2845 
2846 	if (!sticky && !cmsg) {
2847 #ifdef DIAGNOSTIC
2848 		printf("ip6_setpktopt: impossible case\n");
2849 #endif
2850 		return (EINVAL);
2851 	}
2852 
2853 	/*
2854 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2855 	 * not be specified in the context of RFC3542.  Conversely,
2856 	 * RFC3542 types should not be specified in the context of RFC2292.
2857 	 */
2858 	if (!cmsg) {
2859 		switch (optname) {
2860 		case IPV6_2292PKTINFO:
2861 		case IPV6_2292HOPLIMIT:
2862 		case IPV6_2292NEXTHOP:
2863 		case IPV6_2292HOPOPTS:
2864 		case IPV6_2292DSTOPTS:
2865 		case IPV6_2292RTHDR:
2866 		case IPV6_2292PKTOPTIONS:
2867 			return (ENOPROTOOPT);
2868 		}
2869 	}
2870 	if (sticky && cmsg) {
2871 		switch (optname) {
2872 		case IPV6_PKTINFO:
2873 		case IPV6_HOPLIMIT:
2874 		case IPV6_NEXTHOP:
2875 		case IPV6_HOPOPTS:
2876 		case IPV6_DSTOPTS:
2877 		case IPV6_RTHDRDSTOPTS:
2878 		case IPV6_RTHDR:
2879 		case IPV6_USE_MIN_MTU:
2880 		case IPV6_DONTFRAG:
2881 		case IPV6_TCLASS:
2882 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2883 			return (ENOPROTOOPT);
2884 		}
2885 	}
2886 
2887 	switch (optname) {
2888 	case IPV6_2292PKTINFO:
2889 	case IPV6_PKTINFO:
2890 	{
2891 		struct ifnet *ifp = NULL;
2892 		struct in6_pktinfo *pktinfo;
2893 
2894 		if (len != sizeof(struct in6_pktinfo))
2895 			return (EINVAL);
2896 
2897 		pktinfo = (struct in6_pktinfo *)buf;
2898 
2899 		/*
2900 		 * An application can clear any sticky IPV6_PKTINFO option by
2901 		 * doing a "regular" setsockopt with ipi6_addr being
2902 		 * in6addr_any and ipi6_ifindex being zero.
2903 		 * [RFC 3542, Section 6]
2904 		 */
2905 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2906 		    pktinfo->ipi6_ifindex == 0 &&
2907 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2908 			ip6_clearpktopts(opt, optname);
2909 			break;
2910 		}
2911 
2912 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2913 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2914 			return (EINVAL);
2915 		}
2916 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2917 			return (EINVAL);
2918 		/* validate the interface index if specified. */
2919 		if (pktinfo->ipi6_ifindex) {
2920 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2921 			if (ifp == NULL)
2922 				return (ENXIO);
2923 		}
2924 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2925 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2926 			return (ENETDOWN);
2927 
2928 		if (ifp != NULL &&
2929 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2930 			struct in6_ifaddr *ia;
2931 
2932 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2933 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2934 			if (ia == NULL)
2935 				return (EADDRNOTAVAIL);
2936 			ifa_free(&ia->ia_ifa);
2937 		}
2938 		/*
2939 		 * We store the address anyway, and let in6_selectsrc()
2940 		 * validate the specified address.  This is because ipi6_addr
2941 		 * may not have enough information about its scope zone, and
2942 		 * we may need additional information (such as outgoing
2943 		 * interface or the scope zone of a destination address) to
2944 		 * disambiguate the scope.
2945 		 * XXX: the delay of the validation may confuse the
2946 		 * application when it is used as a sticky option.
2947 		 */
2948 		if (opt->ip6po_pktinfo == NULL) {
2949 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2950 			    M_IP6OPT, M_NOWAIT);
2951 			if (opt->ip6po_pktinfo == NULL)
2952 				return (ENOBUFS);
2953 		}
2954 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2955 		opt->ip6po_valid |= IP6PO_VALID_PKTINFO;
2956 		break;
2957 	}
2958 
2959 	case IPV6_2292HOPLIMIT:
2960 	case IPV6_HOPLIMIT:
2961 	{
2962 		int *hlimp;
2963 
2964 		/*
2965 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2966 		 * to simplify the ordering among hoplimit options.
2967 		 */
2968 		if (optname == IPV6_HOPLIMIT && sticky)
2969 			return (ENOPROTOOPT);
2970 
2971 		if (len != sizeof(int))
2972 			return (EINVAL);
2973 		hlimp = (int *)buf;
2974 		if (*hlimp < -1 || *hlimp > 255)
2975 			return (EINVAL);
2976 
2977 		opt->ip6po_hlim = *hlimp;
2978 		opt->ip6po_valid |= IP6PO_VALID_HLIM;
2979 		break;
2980 	}
2981 
2982 	case IPV6_TCLASS:
2983 	{
2984 		int tclass;
2985 
2986 		if (len != sizeof(int))
2987 			return (EINVAL);
2988 		tclass = *(int *)buf;
2989 		if (tclass < -1 || tclass > 255)
2990 			return (EINVAL);
2991 
2992 		opt->ip6po_tclass = tclass;
2993 		opt->ip6po_valid |= IP6PO_VALID_TC;
2994 		break;
2995 	}
2996 
2997 	case IPV6_2292NEXTHOP:
2998 	case IPV6_NEXTHOP:
2999 		if (cred != NULL) {
3000 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3001 			if (error)
3002 				return (error);
3003 		}
3004 
3005 		if (len == 0) {	/* just remove the option */
3006 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3007 			break;
3008 		}
3009 
3010 		/* check if cmsg_len is large enough for sa_len */
3011 		if (len < sizeof(struct sockaddr) || len < *buf)
3012 			return (EINVAL);
3013 
3014 		switch (((struct sockaddr *)buf)->sa_family) {
3015 		case AF_INET6:
3016 		{
3017 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3018 			int error;
3019 
3020 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3021 				return (EINVAL);
3022 
3023 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3024 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3025 				return (EINVAL);
3026 			}
3027 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3028 			    != 0) {
3029 				return (error);
3030 			}
3031 			break;
3032 		}
3033 		case AF_LINK:	/* should eventually be supported */
3034 		default:
3035 			return (EAFNOSUPPORT);
3036 		}
3037 
3038 		/* turn off the previous option, then set the new option. */
3039 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3040 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3041 		if (opt->ip6po_nexthop == NULL)
3042 			return (ENOBUFS);
3043 		bcopy(buf, opt->ip6po_nexthop, *buf);
3044 		opt->ip6po_valid |= IP6PO_VALID_NHINFO;
3045 		break;
3046 
3047 	case IPV6_2292HOPOPTS:
3048 	case IPV6_HOPOPTS:
3049 	{
3050 		struct ip6_hbh *hbh;
3051 		int hbhlen;
3052 
3053 		/*
3054 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3055 		 * options, since per-option restriction has too much
3056 		 * overhead.
3057 		 */
3058 		if (cred != NULL) {
3059 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3060 			if (error)
3061 				return (error);
3062 		}
3063 
3064 		if (len == 0) {
3065 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3066 			break;	/* just remove the option */
3067 		}
3068 
3069 		/* message length validation */
3070 		if (len < sizeof(struct ip6_hbh))
3071 			return (EINVAL);
3072 		hbh = (struct ip6_hbh *)buf;
3073 		hbhlen = (hbh->ip6h_len + 1) << 3;
3074 		if (len != hbhlen)
3075 			return (EINVAL);
3076 
3077 		/* turn off the previous option, then set the new option. */
3078 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3079 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3080 		if (opt->ip6po_hbh == NULL)
3081 			return (ENOBUFS);
3082 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3083 		opt->ip6po_valid |= IP6PO_VALID_HBH;
3084 
3085 		break;
3086 	}
3087 
3088 	case IPV6_2292DSTOPTS:
3089 	case IPV6_DSTOPTS:
3090 	case IPV6_RTHDRDSTOPTS:
3091 	{
3092 		struct ip6_dest *dest, **newdest = NULL;
3093 		int destlen;
3094 
3095 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3096 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3097 			if (error)
3098 				return (error);
3099 		}
3100 
3101 		if (len == 0) {
3102 			ip6_clearpktopts(opt, optname);
3103 			break;	/* just remove the option */
3104 		}
3105 
3106 		/* message length validation */
3107 		if (len < sizeof(struct ip6_dest))
3108 			return (EINVAL);
3109 		dest = (struct ip6_dest *)buf;
3110 		destlen = (dest->ip6d_len + 1) << 3;
3111 		if (len != destlen)
3112 			return (EINVAL);
3113 
3114 		/*
3115 		 * Determine the position that the destination options header
3116 		 * should be inserted; before or after the routing header.
3117 		 */
3118 		switch (optname) {
3119 		case IPV6_2292DSTOPTS:
3120 			/*
3121 			 * The old advacned API is ambiguous on this point.
3122 			 * Our approach is to determine the position based
3123 			 * according to the existence of a routing header.
3124 			 * Note, however, that this depends on the order of the
3125 			 * extension headers in the ancillary data; the 1st
3126 			 * part of the destination options header must appear
3127 			 * before the routing header in the ancillary data,
3128 			 * too.
3129 			 * RFC3542 solved the ambiguity by introducing
3130 			 * separate ancillary data or option types.
3131 			 */
3132 			if (opt->ip6po_rthdr == NULL)
3133 				newdest = &opt->ip6po_dest1;
3134 			else
3135 				newdest = &opt->ip6po_dest2;
3136 			break;
3137 		case IPV6_RTHDRDSTOPTS:
3138 			newdest = &opt->ip6po_dest1;
3139 			break;
3140 		case IPV6_DSTOPTS:
3141 			newdest = &opt->ip6po_dest2;
3142 			break;
3143 		}
3144 
3145 		/* turn off the previous option, then set the new option. */
3146 		ip6_clearpktopts(opt, optname);
3147 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3148 		if (*newdest == NULL)
3149 			return (ENOBUFS);
3150 		bcopy(dest, *newdest, destlen);
3151 		if (newdest == &opt->ip6po_dest1)
3152 			opt->ip6po_valid |= IP6PO_VALID_DEST1;
3153 		else
3154 			opt->ip6po_valid |= IP6PO_VALID_DEST2;
3155 
3156 		break;
3157 	}
3158 
3159 	case IPV6_2292RTHDR:
3160 	case IPV6_RTHDR:
3161 	{
3162 		struct ip6_rthdr *rth;
3163 		int rthlen;
3164 
3165 		if (len == 0) {
3166 			ip6_clearpktopts(opt, IPV6_RTHDR);
3167 			break;	/* just remove the option */
3168 		}
3169 
3170 		/* message length validation */
3171 		if (len < sizeof(struct ip6_rthdr))
3172 			return (EINVAL);
3173 		rth = (struct ip6_rthdr *)buf;
3174 		rthlen = (rth->ip6r_len + 1) << 3;
3175 		if (len != rthlen)
3176 			return (EINVAL);
3177 
3178 		switch (rth->ip6r_type) {
3179 		case IPV6_RTHDR_TYPE_0:
3180 			if (rth->ip6r_len == 0)	/* must contain one addr */
3181 				return (EINVAL);
3182 			if (rth->ip6r_len % 2) /* length must be even */
3183 				return (EINVAL);
3184 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3185 				return (EINVAL);
3186 			break;
3187 		default:
3188 			return (EINVAL);	/* not supported */
3189 		}
3190 
3191 		/* turn off the previous option */
3192 		ip6_clearpktopts(opt, IPV6_RTHDR);
3193 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3194 		if (opt->ip6po_rthdr == NULL)
3195 			return (ENOBUFS);
3196 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3197 		opt->ip6po_valid |= IP6PO_VALID_RHINFO;
3198 
3199 		break;
3200 	}
3201 
3202 	case IPV6_USE_MIN_MTU:
3203 		if (len != sizeof(int))
3204 			return (EINVAL);
3205 		minmtupolicy = *(int *)buf;
3206 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3207 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3208 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3209 			return (EINVAL);
3210 		}
3211 		opt->ip6po_minmtu = minmtupolicy;
3212 		break;
3213 
3214 	case IPV6_DONTFRAG:
3215 		if (len != sizeof(int))
3216 			return (EINVAL);
3217 
3218 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3219 			/*
3220 			 * we ignore this option for TCP sockets.
3221 			 * (RFC3542 leaves this case unspecified.)
3222 			 */
3223 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3224 		} else
3225 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3226 		break;
3227 
3228 	case IPV6_PREFER_TEMPADDR:
3229 		if (len != sizeof(int))
3230 			return (EINVAL);
3231 		preftemp = *(int *)buf;
3232 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3233 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3234 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3235 			return (EINVAL);
3236 		}
3237 		opt->ip6po_prefer_tempaddr = preftemp;
3238 		break;
3239 
3240 	default:
3241 		return (ENOPROTOOPT);
3242 	} /* end of switch */
3243 
3244 	return (0);
3245 }
3246 
3247 /*
3248  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3249  * packet to the input queue of a specified interface.  Note that this
3250  * calls the output routine of the loopback "driver", but with an interface
3251  * pointer that might NOT be &loif -- easier than replicating that code here.
3252  */
3253 void
ip6_mloopback(struct ifnet * ifp,struct mbuf * m)3254 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3255 {
3256 	struct mbuf *copym;
3257 	struct ip6_hdr *ip6;
3258 
3259 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3260 	if (copym == NULL)
3261 		return;
3262 
3263 	/*
3264 	 * Make sure to deep-copy IPv6 header portion in case the data
3265 	 * is in an mbuf cluster, so that we can safely override the IPv6
3266 	 * header portion later.
3267 	 */
3268 	if (!M_WRITABLE(copym) ||
3269 	    copym->m_len < sizeof(struct ip6_hdr)) {
3270 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3271 		if (copym == NULL)
3272 			return;
3273 	}
3274 	ip6 = mtod(copym, struct ip6_hdr *);
3275 	/*
3276 	 * clear embedded scope identifiers if necessary.
3277 	 * in6_clearscope will touch the addresses only when necessary.
3278 	 */
3279 	in6_clearscope(&ip6->ip6_src);
3280 	in6_clearscope(&ip6->ip6_dst);
3281 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3282 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3283 		    CSUM_PSEUDO_HDR;
3284 		copym->m_pkthdr.csum_data = 0xffff;
3285 	}
3286 	if_simloop(ifp, copym, AF_INET6, 0);
3287 }
3288 
3289 /*
3290  * Chop IPv6 header off from the payload.
3291  */
3292 static int
ip6_splithdr(struct mbuf * m,struct ip6_exthdrs * exthdrs)3293 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3294 {
3295 	struct mbuf *mh;
3296 	struct ip6_hdr *ip6;
3297 
3298 	ip6 = mtod(m, struct ip6_hdr *);
3299 	if (m->m_len > sizeof(*ip6)) {
3300 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3301 		if (mh == NULL) {
3302 			m_freem(m);
3303 			return ENOBUFS;
3304 		}
3305 		m_move_pkthdr(mh, m);
3306 		M_ALIGN(mh, sizeof(*ip6));
3307 		m->m_len -= sizeof(*ip6);
3308 		m->m_data += sizeof(*ip6);
3309 		mh->m_next = m;
3310 		m = mh;
3311 		m->m_len = sizeof(*ip6);
3312 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3313 	}
3314 	exthdrs->ip6e_ip6 = m;
3315 	return 0;
3316 }
3317 
3318 /*
3319  * Compute IPv6 extension header length.
3320  */
3321 int
ip6_optlen(struct inpcb * inp)3322 ip6_optlen(struct inpcb *inp)
3323 {
3324 	int len;
3325 
3326 	if (!inp->in6p_outputopts)
3327 		return 0;
3328 
3329 	len = 0;
3330 #define elen(x) \
3331     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3332 
3333 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3334 	if (inp->in6p_outputopts->ip6po_rthdr)
3335 		/* dest1 is valid with rthdr only */
3336 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3337 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3338 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3339 	return len;
3340 #undef elen
3341 }
3342