ip6_output.c revision 1.161 1 /* $NetBSD: ip6_output.c,v 1.161 2015/01/20 21:27:36 roy Exp $ */
2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */
3
4 /*
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33 /*
34 * Copyright (c) 1982, 1986, 1988, 1990, 1993
35 * The Regents of the University of California. All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 * notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 * notice, this list of conditions and the following disclaimer in the
44 * documentation and/or other materials provided with the distribution.
45 * 3. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
62 */
63
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.161 2015/01/20 21:27:36 roy Exp $");
66
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70
71 #include <sys/param.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/errno.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/systm.h>
79 #include <sys/proc.h>
80 #include <sys/kauth.h>
81
82 #include <net/if.h>
83 #include <net/route.h>
84 #include <net/pfil.h>
85
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip6.h>
89 #include <netinet/ip_var.h>
90 #include <netinet/icmp6.h>
91 #include <netinet/in_offload.h>
92 #include <netinet/portalgo.h>
93 #include <netinet6/in6_offload.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet6/ip6_private.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/ip6protosw.h>
99 #include <netinet6/scope6_var.h>
100
101 #ifdef IPSEC
102 #include <netipsec/ipsec.h>
103 #include <netipsec/ipsec6.h>
104 #include <netipsec/key.h>
105 #include <netipsec/xform.h>
106 #endif
107
108
109 #include <net/net_osdep.h>
110
111 extern pfil_head_t *inet6_pfil_hook; /* XXX */
112
113 struct ip6_exthdrs {
114 struct mbuf *ip6e_ip6;
115 struct mbuf *ip6e_hbh;
116 struct mbuf *ip6e_dest1;
117 struct mbuf *ip6e_rthdr;
118 struct mbuf *ip6e_dest2;
119 };
120
121 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
122 kauth_cred_t, int);
123 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
124 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
125 int, int, int);
126 static int ip6_setmoptions(const struct sockopt *, struct in6pcb *);
127 static int ip6_getmoptions(struct sockopt *, struct in6pcb *);
128 static int ip6_copyexthdr(struct mbuf **, void *, int);
129 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
130 struct ip6_frag **);
131 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
132 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
133 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
134 const struct in6_addr *, u_long *, int *);
135 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
136
137 #ifdef RFC2292
138 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
139 #endif
140
141 /*
142 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
143 * header (with pri, len, nxt, hlim, src, dst).
144 * This function may modify ver and hlim only.
145 * The mbuf chain containing the packet will be freed.
146 * The mbuf opt, if present, will not be freed.
147 *
148 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
149 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
150 * which is rt_rmx.rmx_mtu.
151 */
152 int
153 ip6_output(
154 struct mbuf *m0,
155 struct ip6_pktopts *opt,
156 struct route *ro,
157 int flags,
158 struct ip6_moptions *im6o,
159 struct socket *so,
160 struct ifnet **ifpp /* XXX: just for statistics */
161 )
162 {
163 struct ip6_hdr *ip6, *mhip6;
164 struct ifnet *ifp, *origifp;
165 struct mbuf *m = m0;
166 int hlen, tlen, len, off;
167 bool tso;
168 struct route ip6route;
169 struct rtentry *rt = NULL;
170 const struct sockaddr_in6 *dst = NULL;
171 struct sockaddr_in6 src_sa, dst_sa;
172 int error = 0;
173 struct in6_ifaddr *ia = NULL;
174 u_long mtu;
175 int alwaysfrag, dontfrag;
176 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
177 struct ip6_exthdrs exthdrs;
178 struct in6_addr finaldst, src0, dst0;
179 u_int32_t zone;
180 struct route *ro_pmtu = NULL;
181 int hdrsplit = 0;
182 int needipsec = 0;
183 #ifdef IPSEC
184 struct secpolicy *sp = NULL;
185 #endif
186
187 memset(&ip6route, 0, sizeof(ip6route));
188
189 #ifdef DIAGNOSTIC
190 if ((m->m_flags & M_PKTHDR) == 0)
191 panic("ip6_output: no HDR");
192
193 if ((m->m_pkthdr.csum_flags &
194 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
195 panic("ip6_output: IPv4 checksum offload flags: %d",
196 m->m_pkthdr.csum_flags);
197 }
198
199 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
200 (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
201 panic("ip6_output: conflicting checksum offload flags: %d",
202 m->m_pkthdr.csum_flags);
203 }
204 #endif
205
206 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
207
208 #define MAKE_EXTHDR(hp, mp) \
209 do { \
210 if (hp) { \
211 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
212 error = ip6_copyexthdr((mp), (void *)(hp), \
213 ((eh)->ip6e_len + 1) << 3); \
214 if (error) \
215 goto freehdrs; \
216 } \
217 } while (/*CONSTCOND*/ 0)
218
219 memset(&exthdrs, 0, sizeof(exthdrs));
220 if (opt) {
221 /* Hop-by-Hop options header */
222 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
223 /* Destination options header(1st part) */
224 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
225 /* Routing header */
226 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
227 /* Destination options header(2nd part) */
228 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
229 }
230
231 /*
232 * Calculate the total length of the extension header chain.
233 * Keep the length of the unfragmentable part for fragmentation.
234 */
235 optlen = 0;
236 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
237 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
238 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
239 unfragpartlen = optlen + sizeof(struct ip6_hdr);
240 /* NOTE: we don't add AH/ESP length here. do that later. */
241 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
242
243 #ifdef IPSEC
244 if (ipsec_used) {
245 /* Check the security policy (SP) for the packet */
246
247 sp = ipsec6_check_policy(m, so, flags, &needipsec, &error);
248 if (error != 0) {
249 /*
250 * Hack: -EINVAL is used to signal that a packet
251 * should be silently discarded. This is typically
252 * because we asked key management for an SA and
253 * it was delayed (e.g. kicked up to IKE).
254 */
255 if (error == -EINVAL)
256 error = 0;
257 goto freehdrs;
258 }
259 }
260 #endif /* IPSEC */
261
262
263 if (needipsec &&
264 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
265 in6_delayed_cksum(m);
266 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
267 }
268
269
270 /*
271 * If we need IPsec, or there is at least one extension header,
272 * separate IP6 header from the payload.
273 */
274 if ((needipsec || optlen) && !hdrsplit) {
275 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
276 m = NULL;
277 goto freehdrs;
278 }
279 m = exthdrs.ip6e_ip6;
280 hdrsplit++;
281 }
282
283 /* adjust pointer */
284 ip6 = mtod(m, struct ip6_hdr *);
285
286 /* adjust mbuf packet header length */
287 m->m_pkthdr.len += optlen;
288 plen = m->m_pkthdr.len - sizeof(*ip6);
289
290 /* If this is a jumbo payload, insert a jumbo payload option. */
291 if (plen > IPV6_MAXPACKET) {
292 if (!hdrsplit) {
293 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
294 m = NULL;
295 goto freehdrs;
296 }
297 m = exthdrs.ip6e_ip6;
298 hdrsplit++;
299 }
300 /* adjust pointer */
301 ip6 = mtod(m, struct ip6_hdr *);
302 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
303 goto freehdrs;
304 optlen += 8; /* XXX JUMBOOPTLEN */
305 ip6->ip6_plen = 0;
306 } else
307 ip6->ip6_plen = htons(plen);
308
309 /*
310 * Concatenate headers and fill in next header fields.
311 * Here we have, on "m"
312 * IPv6 payload
313 * and we insert headers accordingly. Finally, we should be getting:
314 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
315 *
316 * during the header composing process, "m" points to IPv6 header.
317 * "mprev" points to an extension header prior to esp.
318 */
319 {
320 u_char *nexthdrp = &ip6->ip6_nxt;
321 struct mbuf *mprev = m;
322
323 /*
324 * we treat dest2 specially. this makes IPsec processing
325 * much easier. the goal here is to make mprev point the
326 * mbuf prior to dest2.
327 *
328 * result: IPv6 dest2 payload
329 * m and mprev will point to IPv6 header.
330 */
331 if (exthdrs.ip6e_dest2) {
332 if (!hdrsplit)
333 panic("assumption failed: hdr not split");
334 exthdrs.ip6e_dest2->m_next = m->m_next;
335 m->m_next = exthdrs.ip6e_dest2;
336 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
337 ip6->ip6_nxt = IPPROTO_DSTOPTS;
338 }
339
340 #define MAKE_CHAIN(m, mp, p, i)\
341 do {\
342 if (m) {\
343 if (!hdrsplit) \
344 panic("assumption failed: hdr not split"); \
345 *mtod((m), u_char *) = *(p);\
346 *(p) = (i);\
347 p = mtod((m), u_char *);\
348 (m)->m_next = (mp)->m_next;\
349 (mp)->m_next = (m);\
350 (mp) = (m);\
351 }\
352 } while (/*CONSTCOND*/ 0)
353 /*
354 * result: IPv6 hbh dest1 rthdr dest2 payload
355 * m will point to IPv6 header. mprev will point to the
356 * extension header prior to dest2 (rthdr in the above case).
357 */
358 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
359 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
360 IPPROTO_DSTOPTS);
361 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
362 IPPROTO_ROUTING);
363
364 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
365 sizeof(struct ip6_hdr) + optlen);
366 }
367
368 /*
369 * If there is a routing header, replace destination address field
370 * with the first hop of the routing header.
371 */
372 if (exthdrs.ip6e_rthdr) {
373 struct ip6_rthdr *rh;
374 struct ip6_rthdr0 *rh0;
375 struct in6_addr *addr;
376 struct sockaddr_in6 sa;
377
378 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
379 struct ip6_rthdr *));
380 finaldst = ip6->ip6_dst;
381 switch (rh->ip6r_type) {
382 case IPV6_RTHDR_TYPE_0:
383 rh0 = (struct ip6_rthdr0 *)rh;
384 addr = (struct in6_addr *)(rh0 + 1);
385
386 /*
387 * construct a sockaddr_in6 form of
388 * the first hop.
389 *
390 * XXX: we may not have enough
391 * information about its scope zone;
392 * there is no standard API to pass
393 * the information from the
394 * application.
395 */
396 sockaddr_in6_init(&sa, addr, 0, 0, 0);
397 if ((error = sa6_embedscope(&sa,
398 ip6_use_defzone)) != 0) {
399 goto bad;
400 }
401 ip6->ip6_dst = sa.sin6_addr;
402 (void)memmove(&addr[0], &addr[1],
403 sizeof(struct in6_addr) *
404 (rh0->ip6r0_segleft - 1));
405 addr[rh0->ip6r0_segleft - 1] = finaldst;
406 /* XXX */
407 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
408 break;
409 default: /* is it possible? */
410 error = EINVAL;
411 goto bad;
412 }
413 }
414
415 /* Source address validation */
416 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
417 (flags & IPV6_UNSPECSRC) == 0) {
418 error = EOPNOTSUPP;
419 IP6_STATINC(IP6_STAT_BADSCOPE);
420 goto bad;
421 }
422 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
423 error = EOPNOTSUPP;
424 IP6_STATINC(IP6_STAT_BADSCOPE);
425 goto bad;
426 }
427
428 IP6_STATINC(IP6_STAT_LOCALOUT);
429
430 /*
431 * Route packet.
432 */
433 /* initialize cached route */
434 if (ro == NULL) {
435 ro = &ip6route;
436 }
437 ro_pmtu = ro;
438 if (opt && opt->ip6po_rthdr)
439 ro = &opt->ip6po_route;
440
441 /*
442 * if specified, try to fill in the traffic class field.
443 * do not override if a non-zero value is already set.
444 * we check the diffserv field and the ecn field separately.
445 */
446 if (opt && opt->ip6po_tclass >= 0) {
447 int mask = 0;
448
449 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
450 mask |= 0xfc;
451 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
452 mask |= 0x03;
453 if (mask != 0)
454 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
455 }
456
457 /* fill in or override the hop limit field, if necessary. */
458 if (opt && opt->ip6po_hlim != -1)
459 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
460 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
461 if (im6o != NULL)
462 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
463 else
464 ip6->ip6_hlim = ip6_defmcasthlim;
465 }
466
467 #ifdef IPSEC
468 if (needipsec) {
469 int s = splsoftnet();
470 error = ipsec6_process_packet(m, sp->req);
471
472 /*
473 * Preserve KAME behaviour: ENOENT can be returned
474 * when an SA acquire is in progress. Don't propagate
475 * this to user-level; it confuses applications.
476 * XXX this will go away when the SADB is redone.
477 */
478 if (error == ENOENT)
479 error = 0;
480 splx(s);
481 goto done;
482 }
483 #endif /* IPSEC */
484
485 /* adjust pointer */
486 ip6 = mtod(m, struct ip6_hdr *);
487
488 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
489 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
490 &ifp, &rt, 0)) != 0) {
491 if (ifp != NULL)
492 in6_ifstat_inc(ifp, ifs6_out_discard);
493 goto bad;
494 }
495 if (rt == NULL) {
496 /*
497 * If in6_selectroute() does not return a route entry,
498 * dst may not have been updated.
499 */
500 error = rtcache_setdst(ro, sin6tosa(&dst_sa));
501 if (error) {
502 goto bad;
503 }
504 }
505
506 /*
507 * then rt (for unicast) and ifp must be non-NULL valid values.
508 */
509 if ((flags & IPV6_FORWARDING) == 0) {
510 /* XXX: the FORWARDING flag can be set for mrouting. */
511 in6_ifstat_inc(ifp, ifs6_out_request);
512 }
513 if (rt != NULL) {
514 ia = (struct in6_ifaddr *)(rt->rt_ifa);
515 rt->rt_use++;
516 }
517
518 /*
519 * The outgoing interface must be in the zone of source and
520 * destination addresses. We should use ia_ifp to support the
521 * case of sending packets to an address of our own.
522 */
523 if (ia != NULL && ia->ia_ifp)
524 origifp = ia->ia_ifp;
525 else
526 origifp = ifp;
527
528 src0 = ip6->ip6_src;
529 if (in6_setscope(&src0, origifp, &zone))
530 goto badscope;
531 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
532 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
533 goto badscope;
534
535 dst0 = ip6->ip6_dst;
536 if (in6_setscope(&dst0, origifp, &zone))
537 goto badscope;
538 /* re-initialize to be sure */
539 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
540 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
541 goto badscope;
542
543 /* scope check is done. */
544
545 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
546 if (dst == NULL)
547 dst = satocsin6(rtcache_getdst(ro));
548 KASSERT(dst != NULL);
549 } else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
550 /*
551 * The nexthop is explicitly specified by the
552 * application. We assume the next hop is an IPv6
553 * address.
554 */
555 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
556 } else if ((rt->rt_flags & RTF_GATEWAY))
557 dst = (struct sockaddr_in6 *)rt->rt_gateway;
558 else if (dst == NULL)
559 dst = satocsin6(rtcache_getdst(ro));
560
561 /*
562 * XXXXXX: original code follows:
563 */
564 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
565 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
566 else {
567 struct in6_multi *in6m;
568
569 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
570
571 in6_ifstat_inc(ifp, ifs6_out_mcast);
572
573 /*
574 * Confirm that the outgoing interface supports multicast.
575 */
576 if (!(ifp->if_flags & IFF_MULTICAST)) {
577 IP6_STATINC(IP6_STAT_NOROUTE);
578 in6_ifstat_inc(ifp, ifs6_out_discard);
579 error = ENETUNREACH;
580 goto bad;
581 }
582
583 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
584 if (in6m != NULL &&
585 (im6o == NULL || im6o->im6o_multicast_loop)) {
586 /*
587 * If we belong to the destination multicast group
588 * on the outgoing interface, and the caller did not
589 * forbid loopback, loop back a copy.
590 */
591 KASSERT(dst != NULL);
592 ip6_mloopback(ifp, m, dst);
593 } else {
594 /*
595 * If we are acting as a multicast router, perform
596 * multicast forwarding as if the packet had just
597 * arrived on the interface to which we are about
598 * to send. The multicast forwarding function
599 * recursively calls this function, using the
600 * IPV6_FORWARDING flag to prevent infinite recursion.
601 *
602 * Multicasts that are looped back by ip6_mloopback(),
603 * above, will be forwarded by the ip6_input() routine,
604 * if necessary.
605 */
606 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
607 if (ip6_mforward(ip6, ifp, m) != 0) {
608 m_freem(m);
609 goto done;
610 }
611 }
612 }
613 /*
614 * Multicasts with a hoplimit of zero may be looped back,
615 * above, but must not be transmitted on a network.
616 * Also, multicasts addressed to the loopback interface
617 * are not sent -- the above call to ip6_mloopback() will
618 * loop back a copy if this host actually belongs to the
619 * destination group on the loopback interface.
620 */
621 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
622 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
623 m_freem(m);
624 goto done;
625 }
626 }
627
628 /*
629 * Fill the outgoing inteface to tell the upper layer
630 * to increment per-interface statistics.
631 */
632 if (ifpp)
633 *ifpp = ifp;
634
635 /* Determine path MTU. */
636 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
637 &alwaysfrag)) != 0)
638 goto bad;
639
640 /*
641 * The caller of this function may specify to use the minimum MTU
642 * in some cases.
643 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
644 * setting. The logic is a bit complicated; by default, unicast
645 * packets will follow path MTU while multicast packets will be sent at
646 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
647 * including unicast ones will be sent at the minimum MTU. Multicast
648 * packets will always be sent at the minimum MTU unless
649 * IP6PO_MINMTU_DISABLE is explicitly specified.
650 * See RFC 3542 for more details.
651 */
652 if (mtu > IPV6_MMTU) {
653 if ((flags & IPV6_MINMTU))
654 mtu = IPV6_MMTU;
655 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
656 mtu = IPV6_MMTU;
657 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
658 (opt == NULL ||
659 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
660 mtu = IPV6_MMTU;
661 }
662 }
663
664 /*
665 * clear embedded scope identifiers if necessary.
666 * in6_clearscope will touch the addresses only when necessary.
667 */
668 in6_clearscope(&ip6->ip6_src);
669 in6_clearscope(&ip6->ip6_dst);
670
671 /*
672 * If the outgoing packet contains a hop-by-hop options header,
673 * it must be examined and processed even by the source node.
674 * (RFC 2460, section 4.)
675 */
676 if (ip6->ip6_nxt == IPV6_HOPOPTS) {
677 u_int32_t dummy1; /* XXX unused */
678 u_int32_t dummy2; /* XXX unused */
679 int hoff = sizeof(struct ip6_hdr);
680
681 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
682 /* m was already freed at this point */
683 error = EINVAL;/* better error? */
684 goto done;
685 }
686
687 ip6 = mtod(m, struct ip6_hdr *);
688 }
689
690 /*
691 * Run through list of hooks for output packets.
692 */
693 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
694 goto done;
695 if (m == NULL)
696 goto done;
697 ip6 = mtod(m, struct ip6_hdr *);
698
699 /*
700 * Send the packet to the outgoing interface.
701 * If necessary, do IPv6 fragmentation before sending.
702 *
703 * the logic here is rather complex:
704 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
705 * 1-a: send as is if tlen <= path mtu
706 * 1-b: fragment if tlen > path mtu
707 *
708 * 2: if user asks us not to fragment (dontfrag == 1)
709 * 2-a: send as is if tlen <= interface mtu
710 * 2-b: error if tlen > interface mtu
711 *
712 * 3: if we always need to attach fragment header (alwaysfrag == 1)
713 * always fragment
714 *
715 * 4: if dontfrag == 1 && alwaysfrag == 1
716 * error, as we cannot handle this conflicting request
717 */
718 tlen = m->m_pkthdr.len;
719 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
720 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
721 dontfrag = 1;
722 else
723 dontfrag = 0;
724
725 if (dontfrag && alwaysfrag) { /* case 4 */
726 /* conflicting request - can't transmit */
727 error = EMSGSIZE;
728 goto bad;
729 }
730 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */
731 /*
732 * Even if the DONTFRAG option is specified, we cannot send the
733 * packet when the data length is larger than the MTU of the
734 * outgoing interface.
735 * Notify the error by sending IPV6_PATHMTU ancillary data as
736 * well as returning an error code (the latter is not described
737 * in the API spec.)
738 */
739 u_int32_t mtu32;
740 struct ip6ctlparam ip6cp;
741
742 mtu32 = (u_int32_t)mtu;
743 memset(&ip6cp, 0, sizeof(ip6cp));
744 ip6cp.ip6c_cmdarg = (void *)&mtu32;
745 pfctlinput2(PRC_MSGSIZE,
746 rtcache_getdst(ro_pmtu), &ip6cp);
747
748 error = EMSGSIZE;
749 goto bad;
750 }
751
752 /*
753 * transmit packet without fragmentation
754 */
755 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
756 /* case 1-a and 2-a */
757 struct in6_ifaddr *ia6;
758 int sw_csum;
759
760 ip6 = mtod(m, struct ip6_hdr *);
761 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
762 if (ia6) {
763 /* Record statistics for this interface address. */
764 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
765 }
766
767 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
768 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
769 if (IN6_NEED_CHECKSUM(ifp,
770 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
771 in6_delayed_cksum(m);
772 }
773 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
774 }
775
776 KASSERT(dst != NULL);
777 if (__predict_true(!tso ||
778 (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
779 error = nd6_output(ifp, origifp, m, dst, rt);
780 } else {
781 error = ip6_tso_output(ifp, origifp, m, dst, rt);
782 }
783 goto done;
784 }
785
786 if (tso) {
787 error = EINVAL; /* XXX */
788 goto bad;
789 }
790
791 /*
792 * try to fragment the packet. case 1-b and 3
793 */
794 if (mtu < IPV6_MMTU) {
795 /* path MTU cannot be less than IPV6_MMTU */
796 error = EMSGSIZE;
797 in6_ifstat_inc(ifp, ifs6_out_fragfail);
798 goto bad;
799 } else if (ip6->ip6_plen == 0) {
800 /* jumbo payload cannot be fragmented */
801 error = EMSGSIZE;
802 in6_ifstat_inc(ifp, ifs6_out_fragfail);
803 goto bad;
804 } else {
805 struct mbuf **mnext, *m_frgpart;
806 struct ip6_frag *ip6f;
807 u_int32_t id = htonl(ip6_randomid());
808 u_char nextproto;
809 #if 0 /* see below */
810 struct ip6ctlparam ip6cp;
811 u_int32_t mtu32;
812 #endif
813
814 /*
815 * Too large for the destination or interface;
816 * fragment if possible.
817 * Must be able to put at least 8 bytes per fragment.
818 */
819 hlen = unfragpartlen;
820 if (mtu > IPV6_MAXPACKET)
821 mtu = IPV6_MAXPACKET;
822
823 #if 0
824 /*
825 * It is believed this code is a leftover from the
826 * development of the IPV6_RECVPATHMTU sockopt and
827 * associated work to implement RFC3542.
828 * It's not entirely clear what the intent of the API
829 * is at this point, so disable this code for now.
830 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
831 * will send notifications if the application requests.
832 */
833
834 /* Notify a proper path MTU to applications. */
835 mtu32 = (u_int32_t)mtu;
836 memset(&ip6cp, 0, sizeof(ip6cp));
837 ip6cp.ip6c_cmdarg = (void *)&mtu32;
838 pfctlinput2(PRC_MSGSIZE,
839 rtcache_getdst(ro_pmtu), &ip6cp);
840 #endif
841
842 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
843 if (len < 8) {
844 error = EMSGSIZE;
845 in6_ifstat_inc(ifp, ifs6_out_fragfail);
846 goto bad;
847 }
848
849 mnext = &m->m_nextpkt;
850
851 /*
852 * Change the next header field of the last header in the
853 * unfragmentable part.
854 */
855 if (exthdrs.ip6e_rthdr) {
856 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
857 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
858 } else if (exthdrs.ip6e_dest1) {
859 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
860 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
861 } else if (exthdrs.ip6e_hbh) {
862 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
863 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
864 } else {
865 nextproto = ip6->ip6_nxt;
866 ip6->ip6_nxt = IPPROTO_FRAGMENT;
867 }
868
869 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
870 != 0) {
871 if (IN6_NEED_CHECKSUM(ifp,
872 m->m_pkthdr.csum_flags &
873 (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
874 in6_delayed_cksum(m);
875 }
876 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
877 }
878
879 /*
880 * Loop through length of segment after first fragment,
881 * make new header and copy data of each part and link onto
882 * chain.
883 */
884 m0 = m;
885 for (off = hlen; off < tlen; off += len) {
886 struct mbuf *mlast;
887
888 MGETHDR(m, M_DONTWAIT, MT_HEADER);
889 if (!m) {
890 error = ENOBUFS;
891 IP6_STATINC(IP6_STAT_ODROPPED);
892 goto sendorfree;
893 }
894 m->m_pkthdr.rcvif = NULL;
895 m->m_flags = m0->m_flags & M_COPYFLAGS;
896 *mnext = m;
897 mnext = &m->m_nextpkt;
898 m->m_data += max_linkhdr;
899 mhip6 = mtod(m, struct ip6_hdr *);
900 *mhip6 = *ip6;
901 m->m_len = sizeof(*mhip6);
902 /*
903 * ip6f must be valid if error is 0. But how
904 * can a compiler be expected to infer this?
905 */
906 ip6f = NULL;
907 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
908 if (error) {
909 IP6_STATINC(IP6_STAT_ODROPPED);
910 goto sendorfree;
911 }
912 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
913 if (off + len >= tlen)
914 len = tlen - off;
915 else
916 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
917 mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
918 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
919 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
920 error = ENOBUFS;
921 IP6_STATINC(IP6_STAT_ODROPPED);
922 goto sendorfree;
923 }
924 for (mlast = m; mlast->m_next; mlast = mlast->m_next)
925 ;
926 mlast->m_next = m_frgpart;
927 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
928 m->m_pkthdr.rcvif = NULL;
929 ip6f->ip6f_reserved = 0;
930 ip6f->ip6f_ident = id;
931 ip6f->ip6f_nxt = nextproto;
932 IP6_STATINC(IP6_STAT_OFRAGMENTS);
933 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
934 }
935
936 in6_ifstat_inc(ifp, ifs6_out_fragok);
937 }
938
939 /*
940 * Remove leading garbages.
941 */
942 sendorfree:
943 m = m0->m_nextpkt;
944 m0->m_nextpkt = 0;
945 m_freem(m0);
946 for (m0 = m; m; m = m0) {
947 m0 = m->m_nextpkt;
948 m->m_nextpkt = 0;
949 if (error == 0) {
950 struct in6_ifaddr *ia6;
951 ip6 = mtod(m, struct ip6_hdr *);
952 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
953 if (ia6) {
954 /*
955 * Record statistics for this interface
956 * address.
957 */
958 ia6->ia_ifa.ifa_data.ifad_outbytes +=
959 m->m_pkthdr.len;
960 }
961 KASSERT(dst != NULL);
962 error = nd6_output(ifp, origifp, m, dst, rt);
963 } else
964 m_freem(m);
965 }
966
967 if (error == 0)
968 IP6_STATINC(IP6_STAT_FRAGMENTED);
969
970 done:
971 rtcache_free(&ip6route);
972
973 #ifdef IPSEC
974 if (sp != NULL)
975 KEY_FREESP(&sp);
976 #endif /* IPSEC */
977
978
979 return (error);
980
981 freehdrs:
982 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
983 m_freem(exthdrs.ip6e_dest1);
984 m_freem(exthdrs.ip6e_rthdr);
985 m_freem(exthdrs.ip6e_dest2);
986 /* FALLTHROUGH */
987 bad:
988 m_freem(m);
989 goto done;
990 badscope:
991 IP6_STATINC(IP6_STAT_BADSCOPE);
992 in6_ifstat_inc(origifp, ifs6_out_discard);
993 if (error == 0)
994 error = EHOSTUNREACH; /* XXX */
995 goto bad;
996 }
997
998 static int
999 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1000 {
1001 struct mbuf *m;
1002
1003 if (hlen > MCLBYTES)
1004 return (ENOBUFS); /* XXX */
1005
1006 MGET(m, M_DONTWAIT, MT_DATA);
1007 if (!m)
1008 return (ENOBUFS);
1009
1010 if (hlen > MLEN) {
1011 MCLGET(m, M_DONTWAIT);
1012 if ((m->m_flags & M_EXT) == 0) {
1013 m_free(m);
1014 return (ENOBUFS);
1015 }
1016 }
1017 m->m_len = hlen;
1018 if (hdr)
1019 bcopy(hdr, mtod(m, void *), hlen);
1020
1021 *mp = m;
1022 return (0);
1023 }
1024
1025 /*
1026 * Process a delayed payload checksum calculation.
1027 */
1028 void
1029 in6_delayed_cksum(struct mbuf *m)
1030 {
1031 uint16_t csum, offset;
1032
1033 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1034 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1035 KASSERT((m->m_pkthdr.csum_flags
1036 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1037
1038 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1039 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1040 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1041 csum = 0xffff;
1042 }
1043
1044 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1045 if ((offset + sizeof(csum)) > m->m_len) {
1046 m_copyback(m, offset, sizeof(csum), &csum);
1047 } else {
1048 *(uint16_t *)(mtod(m, char *) + offset) = csum;
1049 }
1050 }
1051
1052 /*
1053 * Insert jumbo payload option.
1054 */
1055 static int
1056 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1057 {
1058 struct mbuf *mopt;
1059 u_int8_t *optbuf;
1060 u_int32_t v;
1061
1062 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1063
1064 /*
1065 * If there is no hop-by-hop options header, allocate new one.
1066 * If there is one but it doesn't have enough space to store the
1067 * jumbo payload option, allocate a cluster to store the whole options.
1068 * Otherwise, use it to store the options.
1069 */
1070 if (exthdrs->ip6e_hbh == 0) {
1071 MGET(mopt, M_DONTWAIT, MT_DATA);
1072 if (mopt == 0)
1073 return (ENOBUFS);
1074 mopt->m_len = JUMBOOPTLEN;
1075 optbuf = mtod(mopt, u_int8_t *);
1076 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1077 exthdrs->ip6e_hbh = mopt;
1078 } else {
1079 struct ip6_hbh *hbh;
1080
1081 mopt = exthdrs->ip6e_hbh;
1082 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1083 /*
1084 * XXX assumption:
1085 * - exthdrs->ip6e_hbh is not referenced from places
1086 * other than exthdrs.
1087 * - exthdrs->ip6e_hbh is not an mbuf chain.
1088 */
1089 int oldoptlen = mopt->m_len;
1090 struct mbuf *n;
1091
1092 /*
1093 * XXX: give up if the whole (new) hbh header does
1094 * not fit even in an mbuf cluster.
1095 */
1096 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1097 return (ENOBUFS);
1098
1099 /*
1100 * As a consequence, we must always prepare a cluster
1101 * at this point.
1102 */
1103 MGET(n, M_DONTWAIT, MT_DATA);
1104 if (n) {
1105 MCLGET(n, M_DONTWAIT);
1106 if ((n->m_flags & M_EXT) == 0) {
1107 m_freem(n);
1108 n = NULL;
1109 }
1110 }
1111 if (!n)
1112 return (ENOBUFS);
1113 n->m_len = oldoptlen + JUMBOOPTLEN;
1114 bcopy(mtod(mopt, void *), mtod(n, void *),
1115 oldoptlen);
1116 optbuf = mtod(n, u_int8_t *) + oldoptlen;
1117 m_freem(mopt);
1118 mopt = exthdrs->ip6e_hbh = n;
1119 } else {
1120 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1121 mopt->m_len += JUMBOOPTLEN;
1122 }
1123 optbuf[0] = IP6OPT_PADN;
1124 optbuf[1] = 0;
1125
1126 /*
1127 * Adjust the header length according to the pad and
1128 * the jumbo payload option.
1129 */
1130 hbh = mtod(mopt, struct ip6_hbh *);
1131 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1132 }
1133
1134 /* fill in the option. */
1135 optbuf[2] = IP6OPT_JUMBO;
1136 optbuf[3] = 4;
1137 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1138 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1139
1140 /* finally, adjust the packet header length */
1141 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1142
1143 return (0);
1144 #undef JUMBOOPTLEN
1145 }
1146
1147 /*
1148 * Insert fragment header and copy unfragmentable header portions.
1149 *
1150 * *frghdrp will not be read, and it is guaranteed that either an
1151 * error is returned or that *frghdrp will point to space allocated
1152 * for the fragment header.
1153 */
1154 static int
1155 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1156 struct ip6_frag **frghdrp)
1157 {
1158 struct mbuf *n, *mlast;
1159
1160 if (hlen > sizeof(struct ip6_hdr)) {
1161 n = m_copym(m0, sizeof(struct ip6_hdr),
1162 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1163 if (n == 0)
1164 return (ENOBUFS);
1165 m->m_next = n;
1166 } else
1167 n = m;
1168
1169 /* Search for the last mbuf of unfragmentable part. */
1170 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1171 ;
1172
1173 if ((mlast->m_flags & M_EXT) == 0 &&
1174 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1175 /* use the trailing space of the last mbuf for the fragment hdr */
1176 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1177 mlast->m_len);
1178 mlast->m_len += sizeof(struct ip6_frag);
1179 m->m_pkthdr.len += sizeof(struct ip6_frag);
1180 } else {
1181 /* allocate a new mbuf for the fragment header */
1182 struct mbuf *mfrg;
1183
1184 MGET(mfrg, M_DONTWAIT, MT_DATA);
1185 if (mfrg == 0)
1186 return (ENOBUFS);
1187 mfrg->m_len = sizeof(struct ip6_frag);
1188 *frghdrp = mtod(mfrg, struct ip6_frag *);
1189 mlast->m_next = mfrg;
1190 }
1191
1192 return (0);
1193 }
1194
1195 static int
1196 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1197 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1198 {
1199 struct rtentry *rt;
1200 u_int32_t mtu = 0;
1201 int alwaysfrag = 0;
1202 int error = 0;
1203
1204 if (ro_pmtu != ro) {
1205 union {
1206 struct sockaddr dst;
1207 struct sockaddr_in6 dst6;
1208 } u;
1209
1210 /* The first hop and the final destination may differ. */
1211 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1212 rt = rtcache_lookup(ro_pmtu, &u.dst);
1213 } else
1214 rt = rtcache_validate(ro_pmtu);
1215 if (rt != NULL) {
1216 u_int32_t ifmtu;
1217
1218 if (ifp == NULL)
1219 ifp = rt->rt_ifp;
1220 ifmtu = IN6_LINKMTU(ifp);
1221 mtu = rt->rt_rmx.rmx_mtu;
1222 if (mtu == 0)
1223 mtu = ifmtu;
1224 else if (mtu < IPV6_MMTU) {
1225 /*
1226 * RFC2460 section 5, last paragraph:
1227 * if we record ICMPv6 too big message with
1228 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1229 * or smaller, with fragment header attached.
1230 * (fragment header is needed regardless from the
1231 * packet size, for translators to identify packets)
1232 */
1233 alwaysfrag = 1;
1234 mtu = IPV6_MMTU;
1235 } else if (mtu > ifmtu) {
1236 /*
1237 * The MTU on the route is larger than the MTU on
1238 * the interface! This shouldn't happen, unless the
1239 * MTU of the interface has been changed after the
1240 * interface was brought up. Change the MTU in the
1241 * route to match the interface MTU (as long as the
1242 * field isn't locked).
1243 */
1244 mtu = ifmtu;
1245 if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1246 rt->rt_rmx.rmx_mtu = mtu;
1247 }
1248 } else if (ifp) {
1249 mtu = IN6_LINKMTU(ifp);
1250 } else
1251 error = EHOSTUNREACH; /* XXX */
1252
1253 *mtup = mtu;
1254 if (alwaysfragp)
1255 *alwaysfragp = alwaysfrag;
1256 return (error);
1257 }
1258
1259 /*
1260 * IP6 socket option processing.
1261 */
1262 int
1263 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1264 {
1265 int optdatalen, uproto;
1266 void *optdata;
1267 struct in6pcb *in6p = sotoin6pcb(so);
1268 struct ip_moptions **mopts;
1269 int error, optval;
1270 int level, optname;
1271
1272 KASSERT(sopt != NULL);
1273
1274 level = sopt->sopt_level;
1275 optname = sopt->sopt_name;
1276
1277 error = optval = 0;
1278 uproto = (int)so->so_proto->pr_protocol;
1279
1280 switch (level) {
1281 case IPPROTO_IP:
1282 switch (optname) {
1283 case IP_ADD_MEMBERSHIP:
1284 case IP_DROP_MEMBERSHIP:
1285 case IP_MULTICAST_IF:
1286 case IP_MULTICAST_LOOP:
1287 case IP_MULTICAST_TTL:
1288 mopts = &in6p->in6p_v4moptions;
1289 switch (op) {
1290 case PRCO_GETOPT:
1291 return ip_getmoptions(*mopts, sopt);
1292 case PRCO_SETOPT:
1293 return ip_setmoptions(mopts, sopt);
1294 default:
1295 return EINVAL;
1296 }
1297 default:
1298 return ENOPROTOOPT;
1299 }
1300 case IPPROTO_IPV6:
1301 break;
1302 default:
1303 return ENOPROTOOPT;
1304 }
1305 switch (op) {
1306 case PRCO_SETOPT:
1307 switch (optname) {
1308 #ifdef RFC2292
1309 case IPV6_2292PKTOPTIONS:
1310 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1311 break;
1312 #endif
1313
1314 /*
1315 * Use of some Hop-by-Hop options or some
1316 * Destination options, might require special
1317 * privilege. That is, normal applications
1318 * (without special privilege) might be forbidden
1319 * from setting certain options in outgoing packets,
1320 * and might never see certain options in received
1321 * packets. [RFC 2292 Section 6]
1322 * KAME specific note:
1323 * KAME prevents non-privileged users from sending or
1324 * receiving ANY hbh/dst options in order to avoid
1325 * overhead of parsing options in the kernel.
1326 */
1327 case IPV6_RECVHOPOPTS:
1328 case IPV6_RECVDSTOPTS:
1329 case IPV6_RECVRTHDRDSTOPTS:
1330 error = kauth_authorize_network(kauth_cred_get(),
1331 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1332 NULL, NULL, NULL);
1333 if (error)
1334 break;
1335 /* FALLTHROUGH */
1336 case IPV6_UNICAST_HOPS:
1337 case IPV6_HOPLIMIT:
1338 case IPV6_FAITH:
1339
1340 case IPV6_RECVPKTINFO:
1341 case IPV6_RECVHOPLIMIT:
1342 case IPV6_RECVRTHDR:
1343 case IPV6_RECVPATHMTU:
1344 case IPV6_RECVTCLASS:
1345 case IPV6_V6ONLY:
1346 error = sockopt_getint(sopt, &optval);
1347 if (error)
1348 break;
1349 switch (optname) {
1350 case IPV6_UNICAST_HOPS:
1351 if (optval < -1 || optval >= 256)
1352 error = EINVAL;
1353 else {
1354 /* -1 = kernel default */
1355 in6p->in6p_hops = optval;
1356 }
1357 break;
1358 #define OPTSET(bit) \
1359 do { \
1360 if (optval) \
1361 in6p->in6p_flags |= (bit); \
1362 else \
1363 in6p->in6p_flags &= ~(bit); \
1364 } while (/*CONSTCOND*/ 0)
1365
1366 #ifdef RFC2292
1367 #define OPTSET2292(bit) \
1368 do { \
1369 in6p->in6p_flags |= IN6P_RFC2292; \
1370 if (optval) \
1371 in6p->in6p_flags |= (bit); \
1372 else \
1373 in6p->in6p_flags &= ~(bit); \
1374 } while (/*CONSTCOND*/ 0)
1375 #endif
1376
1377 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1378
1379 case IPV6_RECVPKTINFO:
1380 #ifdef RFC2292
1381 /* cannot mix with RFC2292 */
1382 if (OPTBIT(IN6P_RFC2292)) {
1383 error = EINVAL;
1384 break;
1385 }
1386 #endif
1387 OPTSET(IN6P_PKTINFO);
1388 break;
1389
1390 case IPV6_HOPLIMIT:
1391 {
1392 struct ip6_pktopts **optp;
1393
1394 #ifdef RFC2292
1395 /* cannot mix with RFC2292 */
1396 if (OPTBIT(IN6P_RFC2292)) {
1397 error = EINVAL;
1398 break;
1399 }
1400 #endif
1401 optp = &in6p->in6p_outputopts;
1402 error = ip6_pcbopt(IPV6_HOPLIMIT,
1403 (u_char *)&optval,
1404 sizeof(optval),
1405 optp,
1406 kauth_cred_get(), uproto);
1407 break;
1408 }
1409
1410 case IPV6_RECVHOPLIMIT:
1411 #ifdef RFC2292
1412 /* cannot mix with RFC2292 */
1413 if (OPTBIT(IN6P_RFC2292)) {
1414 error = EINVAL;
1415 break;
1416 }
1417 #endif
1418 OPTSET(IN6P_HOPLIMIT);
1419 break;
1420
1421 case IPV6_RECVHOPOPTS:
1422 #ifdef RFC2292
1423 /* cannot mix with RFC2292 */
1424 if (OPTBIT(IN6P_RFC2292)) {
1425 error = EINVAL;
1426 break;
1427 }
1428 #endif
1429 OPTSET(IN6P_HOPOPTS);
1430 break;
1431
1432 case IPV6_RECVDSTOPTS:
1433 #ifdef RFC2292
1434 /* cannot mix with RFC2292 */
1435 if (OPTBIT(IN6P_RFC2292)) {
1436 error = EINVAL;
1437 break;
1438 }
1439 #endif
1440 OPTSET(IN6P_DSTOPTS);
1441 break;
1442
1443 case IPV6_RECVRTHDRDSTOPTS:
1444 #ifdef RFC2292
1445 /* cannot mix with RFC2292 */
1446 if (OPTBIT(IN6P_RFC2292)) {
1447 error = EINVAL;
1448 break;
1449 }
1450 #endif
1451 OPTSET(IN6P_RTHDRDSTOPTS);
1452 break;
1453
1454 case IPV6_RECVRTHDR:
1455 #ifdef RFC2292
1456 /* cannot mix with RFC2292 */
1457 if (OPTBIT(IN6P_RFC2292)) {
1458 error = EINVAL;
1459 break;
1460 }
1461 #endif
1462 OPTSET(IN6P_RTHDR);
1463 break;
1464
1465 case IPV6_FAITH:
1466 OPTSET(IN6P_FAITH);
1467 break;
1468
1469 case IPV6_RECVPATHMTU:
1470 /*
1471 * We ignore this option for TCP
1472 * sockets.
1473 * (RFC3542 leaves this case
1474 * unspecified.)
1475 */
1476 if (uproto != IPPROTO_TCP)
1477 OPTSET(IN6P_MTU);
1478 break;
1479
1480 case IPV6_V6ONLY:
1481 /*
1482 * make setsockopt(IPV6_V6ONLY)
1483 * available only prior to bind(2).
1484 * see ipng mailing list, Jun 22 2001.
1485 */
1486 if (in6p->in6p_lport ||
1487 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1488 error = EINVAL;
1489 break;
1490 }
1491 #ifdef INET6_BINDV6ONLY
1492 if (!optval)
1493 error = EINVAL;
1494 #else
1495 OPTSET(IN6P_IPV6_V6ONLY);
1496 #endif
1497 break;
1498 case IPV6_RECVTCLASS:
1499 #ifdef RFC2292
1500 /* cannot mix with RFC2292 XXX */
1501 if (OPTBIT(IN6P_RFC2292)) {
1502 error = EINVAL;
1503 break;
1504 }
1505 #endif
1506 OPTSET(IN6P_TCLASS);
1507 break;
1508
1509 }
1510 break;
1511
1512 case IPV6_OTCLASS:
1513 {
1514 struct ip6_pktopts **optp;
1515 u_int8_t tclass;
1516
1517 error = sockopt_get(sopt, &tclass, sizeof(tclass));
1518 if (error)
1519 break;
1520 optp = &in6p->in6p_outputopts;
1521 error = ip6_pcbopt(optname,
1522 (u_char *)&tclass,
1523 sizeof(tclass),
1524 optp,
1525 kauth_cred_get(), uproto);
1526 break;
1527 }
1528
1529 case IPV6_TCLASS:
1530 case IPV6_DONTFRAG:
1531 case IPV6_USE_MIN_MTU:
1532 case IPV6_PREFER_TEMPADDR:
1533 error = sockopt_getint(sopt, &optval);
1534 if (error)
1535 break;
1536 {
1537 struct ip6_pktopts **optp;
1538 optp = &in6p->in6p_outputopts;
1539 error = ip6_pcbopt(optname,
1540 (u_char *)&optval,
1541 sizeof(optval),
1542 optp,
1543 kauth_cred_get(), uproto);
1544 break;
1545 }
1546
1547 #ifdef RFC2292
1548 case IPV6_2292PKTINFO:
1549 case IPV6_2292HOPLIMIT:
1550 case IPV6_2292HOPOPTS:
1551 case IPV6_2292DSTOPTS:
1552 case IPV6_2292RTHDR:
1553 /* RFC 2292 */
1554 error = sockopt_getint(sopt, &optval);
1555 if (error)
1556 break;
1557
1558 switch (optname) {
1559 case IPV6_2292PKTINFO:
1560 OPTSET2292(IN6P_PKTINFO);
1561 break;
1562 case IPV6_2292HOPLIMIT:
1563 OPTSET2292(IN6P_HOPLIMIT);
1564 break;
1565 case IPV6_2292HOPOPTS:
1566 /*
1567 * Check super-user privilege.
1568 * See comments for IPV6_RECVHOPOPTS.
1569 */
1570 error =
1571 kauth_authorize_network(kauth_cred_get(),
1572 KAUTH_NETWORK_IPV6,
1573 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1574 NULL, NULL);
1575 if (error)
1576 return (error);
1577 OPTSET2292(IN6P_HOPOPTS);
1578 break;
1579 case IPV6_2292DSTOPTS:
1580 error =
1581 kauth_authorize_network(kauth_cred_get(),
1582 KAUTH_NETWORK_IPV6,
1583 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1584 NULL, NULL);
1585 if (error)
1586 return (error);
1587 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1588 break;
1589 case IPV6_2292RTHDR:
1590 OPTSET2292(IN6P_RTHDR);
1591 break;
1592 }
1593 break;
1594 #endif
1595 case IPV6_PKTINFO:
1596 case IPV6_HOPOPTS:
1597 case IPV6_RTHDR:
1598 case IPV6_DSTOPTS:
1599 case IPV6_RTHDRDSTOPTS:
1600 case IPV6_NEXTHOP: {
1601 /* new advanced API (RFC3542) */
1602 void *optbuf;
1603 int optbuflen;
1604 struct ip6_pktopts **optp;
1605
1606 #ifdef RFC2292
1607 /* cannot mix with RFC2292 */
1608 if (OPTBIT(IN6P_RFC2292)) {
1609 error = EINVAL;
1610 break;
1611 }
1612 #endif
1613
1614 optbuflen = sopt->sopt_size;
1615 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1616 if (optbuf == NULL) {
1617 error = ENOBUFS;
1618 break;
1619 }
1620
1621 error = sockopt_get(sopt, optbuf, optbuflen);
1622 if (error) {
1623 free(optbuf, M_IP6OPT);
1624 break;
1625 }
1626 optp = &in6p->in6p_outputopts;
1627 error = ip6_pcbopt(optname, optbuf, optbuflen,
1628 optp, kauth_cred_get(), uproto);
1629
1630 free(optbuf, M_IP6OPT);
1631 break;
1632 }
1633 #undef OPTSET
1634
1635 case IPV6_MULTICAST_IF:
1636 case IPV6_MULTICAST_HOPS:
1637 case IPV6_MULTICAST_LOOP:
1638 case IPV6_JOIN_GROUP:
1639 case IPV6_LEAVE_GROUP:
1640 error = ip6_setmoptions(sopt, in6p);
1641 break;
1642
1643 case IPV6_PORTRANGE:
1644 error = sockopt_getint(sopt, &optval);
1645 if (error)
1646 break;
1647
1648 switch (optval) {
1649 case IPV6_PORTRANGE_DEFAULT:
1650 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1651 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1652 break;
1653
1654 case IPV6_PORTRANGE_HIGH:
1655 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1656 in6p->in6p_flags |= IN6P_HIGHPORT;
1657 break;
1658
1659 case IPV6_PORTRANGE_LOW:
1660 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1661 in6p->in6p_flags |= IN6P_LOWPORT;
1662 break;
1663
1664 default:
1665 error = EINVAL;
1666 break;
1667 }
1668 break;
1669
1670 case IPV6_PORTALGO:
1671 error = sockopt_getint(sopt, &optval);
1672 if (error)
1673 break;
1674
1675 error = portalgo_algo_index_select(
1676 (struct inpcb_hdr *)in6p, optval);
1677 break;
1678
1679 #if defined(IPSEC)
1680 case IPV6_IPSEC_POLICY:
1681 if (ipsec_enabled) {
1682 error = ipsec6_set_policy(in6p, optname,
1683 sopt->sopt_data, sopt->sopt_size,
1684 kauth_cred_get());
1685 break;
1686 }
1687 /*FALLTHROUGH*/
1688 #endif /* IPSEC */
1689
1690 default:
1691 error = ENOPROTOOPT;
1692 break;
1693 }
1694 break;
1695
1696 case PRCO_GETOPT:
1697 switch (optname) {
1698 #ifdef RFC2292
1699 case IPV6_2292PKTOPTIONS:
1700 /*
1701 * RFC3542 (effectively) deprecated the
1702 * semantics of the 2292-style pktoptions.
1703 * Since it was not reliable in nature (i.e.,
1704 * applications had to expect the lack of some
1705 * information after all), it would make sense
1706 * to simplify this part by always returning
1707 * empty data.
1708 */
1709 break;
1710 #endif
1711
1712 case IPV6_RECVHOPOPTS:
1713 case IPV6_RECVDSTOPTS:
1714 case IPV6_RECVRTHDRDSTOPTS:
1715 case IPV6_UNICAST_HOPS:
1716 case IPV6_RECVPKTINFO:
1717 case IPV6_RECVHOPLIMIT:
1718 case IPV6_RECVRTHDR:
1719 case IPV6_RECVPATHMTU:
1720
1721 case IPV6_FAITH:
1722 case IPV6_V6ONLY:
1723 case IPV6_PORTRANGE:
1724 case IPV6_RECVTCLASS:
1725 switch (optname) {
1726
1727 case IPV6_RECVHOPOPTS:
1728 optval = OPTBIT(IN6P_HOPOPTS);
1729 break;
1730
1731 case IPV6_RECVDSTOPTS:
1732 optval = OPTBIT(IN6P_DSTOPTS);
1733 break;
1734
1735 case IPV6_RECVRTHDRDSTOPTS:
1736 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1737 break;
1738
1739 case IPV6_UNICAST_HOPS:
1740 optval = in6p->in6p_hops;
1741 break;
1742
1743 case IPV6_RECVPKTINFO:
1744 optval = OPTBIT(IN6P_PKTINFO);
1745 break;
1746
1747 case IPV6_RECVHOPLIMIT:
1748 optval = OPTBIT(IN6P_HOPLIMIT);
1749 break;
1750
1751 case IPV6_RECVRTHDR:
1752 optval = OPTBIT(IN6P_RTHDR);
1753 break;
1754
1755 case IPV6_RECVPATHMTU:
1756 optval = OPTBIT(IN6P_MTU);
1757 break;
1758
1759 case IPV6_FAITH:
1760 optval = OPTBIT(IN6P_FAITH);
1761 break;
1762
1763 case IPV6_V6ONLY:
1764 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1765 break;
1766
1767 case IPV6_PORTRANGE:
1768 {
1769 int flags;
1770 flags = in6p->in6p_flags;
1771 if (flags & IN6P_HIGHPORT)
1772 optval = IPV6_PORTRANGE_HIGH;
1773 else if (flags & IN6P_LOWPORT)
1774 optval = IPV6_PORTRANGE_LOW;
1775 else
1776 optval = 0;
1777 break;
1778 }
1779 case IPV6_RECVTCLASS:
1780 optval = OPTBIT(IN6P_TCLASS);
1781 break;
1782
1783 }
1784 if (error)
1785 break;
1786 error = sockopt_setint(sopt, optval);
1787 break;
1788
1789 case IPV6_PATHMTU:
1790 {
1791 u_long pmtu = 0;
1792 struct ip6_mtuinfo mtuinfo;
1793 struct route *ro = &in6p->in6p_route;
1794
1795 if (!(so->so_state & SS_ISCONNECTED))
1796 return (ENOTCONN);
1797 /*
1798 * XXX: we dot not consider the case of source
1799 * routing, or optional information to specify
1800 * the outgoing interface.
1801 */
1802 error = ip6_getpmtu(ro, NULL, NULL,
1803 &in6p->in6p_faddr, &pmtu, NULL);
1804 if (error)
1805 break;
1806 if (pmtu > IPV6_MAXPACKET)
1807 pmtu = IPV6_MAXPACKET;
1808
1809 memset(&mtuinfo, 0, sizeof(mtuinfo));
1810 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1811 optdata = (void *)&mtuinfo;
1812 optdatalen = sizeof(mtuinfo);
1813 if (optdatalen > MCLBYTES)
1814 return (EMSGSIZE); /* XXX */
1815 error = sockopt_set(sopt, optdata, optdatalen);
1816 break;
1817 }
1818
1819 #ifdef RFC2292
1820 case IPV6_2292PKTINFO:
1821 case IPV6_2292HOPLIMIT:
1822 case IPV6_2292HOPOPTS:
1823 case IPV6_2292RTHDR:
1824 case IPV6_2292DSTOPTS:
1825 switch (optname) {
1826 case IPV6_2292PKTINFO:
1827 optval = OPTBIT(IN6P_PKTINFO);
1828 break;
1829 case IPV6_2292HOPLIMIT:
1830 optval = OPTBIT(IN6P_HOPLIMIT);
1831 break;
1832 case IPV6_2292HOPOPTS:
1833 optval = OPTBIT(IN6P_HOPOPTS);
1834 break;
1835 case IPV6_2292RTHDR:
1836 optval = OPTBIT(IN6P_RTHDR);
1837 break;
1838 case IPV6_2292DSTOPTS:
1839 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1840 break;
1841 }
1842 error = sockopt_setint(sopt, optval);
1843 break;
1844 #endif
1845 case IPV6_PKTINFO:
1846 case IPV6_HOPOPTS:
1847 case IPV6_RTHDR:
1848 case IPV6_DSTOPTS:
1849 case IPV6_RTHDRDSTOPTS:
1850 case IPV6_NEXTHOP:
1851 case IPV6_OTCLASS:
1852 case IPV6_TCLASS:
1853 case IPV6_DONTFRAG:
1854 case IPV6_USE_MIN_MTU:
1855 case IPV6_PREFER_TEMPADDR:
1856 error = ip6_getpcbopt(in6p->in6p_outputopts,
1857 optname, sopt);
1858 break;
1859
1860 case IPV6_MULTICAST_IF:
1861 case IPV6_MULTICAST_HOPS:
1862 case IPV6_MULTICAST_LOOP:
1863 case IPV6_JOIN_GROUP:
1864 case IPV6_LEAVE_GROUP:
1865 error = ip6_getmoptions(sopt, in6p);
1866 break;
1867
1868 case IPV6_PORTALGO:
1869 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo;
1870 error = sockopt_setint(sopt, optval);
1871 break;
1872
1873 #if defined(IPSEC)
1874 case IPV6_IPSEC_POLICY:
1875 if (ipsec_used) {
1876 struct mbuf *m = NULL;
1877
1878 /*
1879 * XXX: this will return EINVAL as sopt is
1880 * empty
1881 */
1882 error = ipsec6_get_policy(in6p, sopt->sopt_data,
1883 sopt->sopt_size, &m);
1884 if (!error)
1885 error = sockopt_setmbuf(sopt, m);
1886 break;
1887 }
1888 /*FALLTHROUGH*/
1889 #endif /* IPSEC */
1890
1891 default:
1892 error = ENOPROTOOPT;
1893 break;
1894 }
1895 break;
1896 }
1897 return (error);
1898 }
1899
1900 int
1901 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1902 {
1903 int error = 0, optval;
1904 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1905 struct in6pcb *in6p = sotoin6pcb(so);
1906 int level, optname;
1907
1908 KASSERT(sopt != NULL);
1909
1910 level = sopt->sopt_level;
1911 optname = sopt->sopt_name;
1912
1913 if (level != IPPROTO_IPV6) {
1914 return ENOPROTOOPT;
1915 }
1916
1917 switch (optname) {
1918 case IPV6_CHECKSUM:
1919 /*
1920 * For ICMPv6 sockets, no modification allowed for checksum
1921 * offset, permit "no change" values to help existing apps.
1922 *
1923 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1924 * for an ICMPv6 socket will fail." The current
1925 * behavior does not meet RFC3542.
1926 */
1927 switch (op) {
1928 case PRCO_SETOPT:
1929 error = sockopt_getint(sopt, &optval);
1930 if (error)
1931 break;
1932 if ((optval % 2) != 0) {
1933 /* the API assumes even offset values */
1934 error = EINVAL;
1935 } else if (so->so_proto->pr_protocol ==
1936 IPPROTO_ICMPV6) {
1937 if (optval != icmp6off)
1938 error = EINVAL;
1939 } else
1940 in6p->in6p_cksum = optval;
1941 break;
1942
1943 case PRCO_GETOPT:
1944 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1945 optval = icmp6off;
1946 else
1947 optval = in6p->in6p_cksum;
1948
1949 error = sockopt_setint(sopt, optval);
1950 break;
1951
1952 default:
1953 error = EINVAL;
1954 break;
1955 }
1956 break;
1957
1958 default:
1959 error = ENOPROTOOPT;
1960 break;
1961 }
1962
1963 return (error);
1964 }
1965
1966 #ifdef RFC2292
1967 /*
1968 * Set up IP6 options in pcb for insertion in output packets or
1969 * specifying behavior of outgoing packets.
1970 */
1971 static int
1972 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1973 struct sockopt *sopt)
1974 {
1975 struct ip6_pktopts *opt = *pktopt;
1976 struct mbuf *m;
1977 int error = 0;
1978
1979 /* turn off any old options. */
1980 if (opt) {
1981 #ifdef DIAGNOSTIC
1982 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1983 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1984 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1985 printf("ip6_pcbopts: all specified options are cleared.\n");
1986 #endif
1987 ip6_clearpktopts(opt, -1);
1988 } else {
1989 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
1990 if (opt == NULL)
1991 return (ENOBUFS);
1992 }
1993 *pktopt = NULL;
1994
1995 if (sopt == NULL || sopt->sopt_size == 0) {
1996 /*
1997 * Only turning off any previous options, regardless of
1998 * whether the opt is just created or given.
1999 */
2000 free(opt, M_IP6OPT);
2001 return (0);
2002 }
2003
2004 /* set options specified by user. */
2005 m = sockopt_getmbuf(sopt);
2006 if (m == NULL) {
2007 free(opt, M_IP6OPT);
2008 return (ENOBUFS);
2009 }
2010
2011 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
2012 so->so_proto->pr_protocol);
2013 m_freem(m);
2014 if (error != 0) {
2015 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2016 free(opt, M_IP6OPT);
2017 return (error);
2018 }
2019 *pktopt = opt;
2020 return (0);
2021 }
2022 #endif
2023
2024 /*
2025 * initialize ip6_pktopts. beware that there are non-zero default values in
2026 * the struct.
2027 */
2028 void
2029 ip6_initpktopts(struct ip6_pktopts *opt)
2030 {
2031
2032 memset(opt, 0, sizeof(*opt));
2033 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2034 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2035 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2036 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2037 }
2038
2039 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */
2040 static int
2041 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2042 kauth_cred_t cred, int uproto)
2043 {
2044 struct ip6_pktopts *opt;
2045
2046 if (*pktopt == NULL) {
2047 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2048 M_NOWAIT);
2049 if (*pktopt == NULL)
2050 return (ENOBUFS);
2051
2052 ip6_initpktopts(*pktopt);
2053 }
2054 opt = *pktopt;
2055
2056 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2057 }
2058
2059 static int
2060 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2061 {
2062 void *optdata = NULL;
2063 int optdatalen = 0;
2064 struct ip6_ext *ip6e;
2065 int error = 0;
2066 struct in6_pktinfo null_pktinfo;
2067 int deftclass = 0, on;
2068 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2069 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2070
2071 switch (optname) {
2072 case IPV6_PKTINFO:
2073 if (pktopt && pktopt->ip6po_pktinfo)
2074 optdata = (void *)pktopt->ip6po_pktinfo;
2075 else {
2076 /* XXX: we don't have to do this every time... */
2077 memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2078 optdata = (void *)&null_pktinfo;
2079 }
2080 optdatalen = sizeof(struct in6_pktinfo);
2081 break;
2082 case IPV6_OTCLASS:
2083 /* XXX */
2084 return (EINVAL);
2085 case IPV6_TCLASS:
2086 if (pktopt && pktopt->ip6po_tclass >= 0)
2087 optdata = (void *)&pktopt->ip6po_tclass;
2088 else
2089 optdata = (void *)&deftclass;
2090 optdatalen = sizeof(int);
2091 break;
2092 case IPV6_HOPOPTS:
2093 if (pktopt && pktopt->ip6po_hbh) {
2094 optdata = (void *)pktopt->ip6po_hbh;
2095 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2096 optdatalen = (ip6e->ip6e_len + 1) << 3;
2097 }
2098 break;
2099 case IPV6_RTHDR:
2100 if (pktopt && pktopt->ip6po_rthdr) {
2101 optdata = (void *)pktopt->ip6po_rthdr;
2102 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2103 optdatalen = (ip6e->ip6e_len + 1) << 3;
2104 }
2105 break;
2106 case IPV6_RTHDRDSTOPTS:
2107 if (pktopt && pktopt->ip6po_dest1) {
2108 optdata = (void *)pktopt->ip6po_dest1;
2109 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2110 optdatalen = (ip6e->ip6e_len + 1) << 3;
2111 }
2112 break;
2113 case IPV6_DSTOPTS:
2114 if (pktopt && pktopt->ip6po_dest2) {
2115 optdata = (void *)pktopt->ip6po_dest2;
2116 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2117 optdatalen = (ip6e->ip6e_len + 1) << 3;
2118 }
2119 break;
2120 case IPV6_NEXTHOP:
2121 if (pktopt && pktopt->ip6po_nexthop) {
2122 optdata = (void *)pktopt->ip6po_nexthop;
2123 optdatalen = pktopt->ip6po_nexthop->sa_len;
2124 }
2125 break;
2126 case IPV6_USE_MIN_MTU:
2127 if (pktopt)
2128 optdata = (void *)&pktopt->ip6po_minmtu;
2129 else
2130 optdata = (void *)&defminmtu;
2131 optdatalen = sizeof(int);
2132 break;
2133 case IPV6_DONTFRAG:
2134 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2135 on = 1;
2136 else
2137 on = 0;
2138 optdata = (void *)&on;
2139 optdatalen = sizeof(on);
2140 break;
2141 case IPV6_PREFER_TEMPADDR:
2142 if (pktopt)
2143 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2144 else
2145 optdata = (void *)&defpreftemp;
2146 optdatalen = sizeof(int);
2147 default: /* should not happen */
2148 #ifdef DIAGNOSTIC
2149 panic("ip6_getpcbopt: unexpected option\n");
2150 #endif
2151 return (ENOPROTOOPT);
2152 }
2153
2154 error = sockopt_set(sopt, optdata, optdatalen);
2155
2156 return (error);
2157 }
2158
2159 void
2160 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2161 {
2162 if (optname == -1 || optname == IPV6_PKTINFO) {
2163 if (pktopt->ip6po_pktinfo)
2164 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2165 pktopt->ip6po_pktinfo = NULL;
2166 }
2167 if (optname == -1 || optname == IPV6_HOPLIMIT)
2168 pktopt->ip6po_hlim = -1;
2169 if (optname == -1 || optname == IPV6_TCLASS)
2170 pktopt->ip6po_tclass = -1;
2171 if (optname == -1 || optname == IPV6_NEXTHOP) {
2172 rtcache_free(&pktopt->ip6po_nextroute);
2173 if (pktopt->ip6po_nexthop)
2174 free(pktopt->ip6po_nexthop, M_IP6OPT);
2175 pktopt->ip6po_nexthop = NULL;
2176 }
2177 if (optname == -1 || optname == IPV6_HOPOPTS) {
2178 if (pktopt->ip6po_hbh)
2179 free(pktopt->ip6po_hbh, M_IP6OPT);
2180 pktopt->ip6po_hbh = NULL;
2181 }
2182 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2183 if (pktopt->ip6po_dest1)
2184 free(pktopt->ip6po_dest1, M_IP6OPT);
2185 pktopt->ip6po_dest1 = NULL;
2186 }
2187 if (optname == -1 || optname == IPV6_RTHDR) {
2188 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2189 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2190 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2191 rtcache_free(&pktopt->ip6po_route);
2192 }
2193 if (optname == -1 || optname == IPV6_DSTOPTS) {
2194 if (pktopt->ip6po_dest2)
2195 free(pktopt->ip6po_dest2, M_IP6OPT);
2196 pktopt->ip6po_dest2 = NULL;
2197 }
2198 }
2199
2200 #define PKTOPT_EXTHDRCPY(type) \
2201 do { \
2202 if (src->type) { \
2203 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2204 dst->type = malloc(hlen, M_IP6OPT, canwait); \
2205 if (dst->type == NULL) \
2206 goto bad; \
2207 memcpy(dst->type, src->type, hlen); \
2208 } \
2209 } while (/*CONSTCOND*/ 0)
2210
2211 static int
2212 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2213 {
2214 dst->ip6po_hlim = src->ip6po_hlim;
2215 dst->ip6po_tclass = src->ip6po_tclass;
2216 dst->ip6po_flags = src->ip6po_flags;
2217 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2218 if (src->ip6po_pktinfo) {
2219 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2220 M_IP6OPT, canwait);
2221 if (dst->ip6po_pktinfo == NULL)
2222 goto bad;
2223 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2224 }
2225 if (src->ip6po_nexthop) {
2226 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2227 M_IP6OPT, canwait);
2228 if (dst->ip6po_nexthop == NULL)
2229 goto bad;
2230 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2231 src->ip6po_nexthop->sa_len);
2232 }
2233 PKTOPT_EXTHDRCPY(ip6po_hbh);
2234 PKTOPT_EXTHDRCPY(ip6po_dest1);
2235 PKTOPT_EXTHDRCPY(ip6po_dest2);
2236 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2237 return (0);
2238
2239 bad:
2240 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2241 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2242 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2243 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2244 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2245 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2246
2247 return (ENOBUFS);
2248 }
2249 #undef PKTOPT_EXTHDRCPY
2250
2251 struct ip6_pktopts *
2252 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2253 {
2254 int error;
2255 struct ip6_pktopts *dst;
2256
2257 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2258 if (dst == NULL)
2259 return (NULL);
2260 ip6_initpktopts(dst);
2261
2262 if ((error = copypktopts(dst, src, canwait)) != 0) {
2263 free(dst, M_IP6OPT);
2264 return (NULL);
2265 }
2266
2267 return (dst);
2268 }
2269
2270 void
2271 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2272 {
2273 if (pktopt == NULL)
2274 return;
2275
2276 ip6_clearpktopts(pktopt, -1);
2277
2278 free(pktopt, M_IP6OPT);
2279 }
2280
2281 int
2282 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, void *v,
2283 size_t l)
2284 {
2285 struct ipv6_mreq mreq;
2286 int error;
2287 struct in6_addr *ia = &mreq.ipv6mr_multiaddr;
2288 struct in_addr *ia4 = (void *)&ia->s6_addr32[3];
2289 error = sockopt_get(sopt, &mreq, sizeof(mreq));
2290 if (error != 0)
2291 return error;
2292
2293 if (IN6_IS_ADDR_UNSPECIFIED(ia)) {
2294 /*
2295 * We use the unspecified address to specify to accept
2296 * all multicast addresses. Only super user is allowed
2297 * to do this.
2298 */
2299 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6,
2300 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2301 return EACCES;
2302 } else if (IN6_IS_ADDR_V4MAPPED(ia)) {
2303 // Don't bother if we are not going to use ifp.
2304 if (l == sizeof(*ia)) {
2305 memcpy(v, ia, l);
2306 return 0;
2307 }
2308 } else if (!IN6_IS_ADDR_MULTICAST(ia)) {
2309 return EINVAL;
2310 }
2311
2312 /*
2313 * If no interface was explicitly specified, choose an
2314 * appropriate one according to the given multicast address.
2315 */
2316 if (mreq.ipv6mr_interface == 0) {
2317 struct rtentry *rt;
2318 union {
2319 struct sockaddr dst;
2320 struct sockaddr_in dst4;
2321 struct sockaddr_in6 dst6;
2322 } u;
2323 struct route ro;
2324
2325 /*
2326 * Look up the routing table for the
2327 * address, and choose the outgoing interface.
2328 * XXX: is it a good approach?
2329 */
2330 memset(&ro, 0, sizeof(ro));
2331 if (IN6_IS_ADDR_V4MAPPED(ia))
2332 sockaddr_in_init(&u.dst4, ia4, 0);
2333 else
2334 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0);
2335 rtcache_setdst(&ro, &u.dst);
2336 *ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL;
2337 rtcache_free(&ro);
2338 } else {
2339 /*
2340 * If the interface is specified, validate it.
2341 */
2342 if ((*ifp = if_byindex(mreq.ipv6mr_interface)) == NULL)
2343 return ENXIO; /* XXX EINVAL? */
2344 }
2345 if (sizeof(*ia) == l)
2346 memcpy(v, ia, l);
2347 else
2348 memcpy(v, ia4, l);
2349 return 0;
2350 }
2351
2352 /*
2353 * Set the IP6 multicast options in response to user setsockopt().
2354 */
2355 static int
2356 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p)
2357 {
2358 int error = 0;
2359 u_int loop, ifindex;
2360 struct ipv6_mreq mreq;
2361 struct in6_addr ia;
2362 struct ifnet *ifp;
2363 struct ip6_moptions *im6o = in6p->in6p_moptions;
2364 struct in6_multi_mship *imm;
2365
2366 if (im6o == NULL) {
2367 /*
2368 * No multicast option buffer attached to the pcb;
2369 * allocate one and initialize to default values.
2370 */
2371 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2372 if (im6o == NULL)
2373 return (ENOBUFS);
2374 in6p->in6p_moptions = im6o;
2375 im6o->im6o_multicast_ifp = NULL;
2376 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2377 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2378 LIST_INIT(&im6o->im6o_memberships);
2379 }
2380
2381 switch (sopt->sopt_name) {
2382
2383 case IPV6_MULTICAST_IF:
2384 /*
2385 * Select the interface for outgoing multicast packets.
2386 */
2387 error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2388 if (error != 0)
2389 break;
2390
2391 if (ifindex != 0) {
2392 if ((ifp = if_byindex(ifindex)) == NULL) {
2393 error = ENXIO; /* XXX EINVAL? */
2394 break;
2395 }
2396 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2397 error = EADDRNOTAVAIL;
2398 break;
2399 }
2400 } else
2401 ifp = NULL;
2402 im6o->im6o_multicast_ifp = ifp;
2403 break;
2404
2405 case IPV6_MULTICAST_HOPS:
2406 {
2407 /*
2408 * Set the IP6 hoplimit for outgoing multicast packets.
2409 */
2410 int optval;
2411
2412 error = sockopt_getint(sopt, &optval);
2413 if (error != 0)
2414 break;
2415
2416 if (optval < -1 || optval >= 256)
2417 error = EINVAL;
2418 else if (optval == -1)
2419 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2420 else
2421 im6o->im6o_multicast_hlim = optval;
2422 break;
2423 }
2424
2425 case IPV6_MULTICAST_LOOP:
2426 /*
2427 * Set the loopback flag for outgoing multicast packets.
2428 * Must be zero or one.
2429 */
2430 error = sockopt_get(sopt, &loop, sizeof(loop));
2431 if (error != 0)
2432 break;
2433 if (loop > 1) {
2434 error = EINVAL;
2435 break;
2436 }
2437 im6o->im6o_multicast_loop = loop;
2438 break;
2439
2440 case IPV6_JOIN_GROUP:
2441 /*
2442 * Add a multicast group membership.
2443 * Group must be a valid IP6 multicast address.
2444 */
2445 if ((error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia))))
2446 return error;
2447
2448 if (IN6_IS_ADDR_V4MAPPED(&ia)) {
2449 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt);
2450 break;
2451 }
2452 /*
2453 * See if we found an interface, and confirm that it
2454 * supports multicast
2455 */
2456 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2457 error = EADDRNOTAVAIL;
2458 break;
2459 }
2460
2461 if (in6_setscope(&ia, ifp, NULL)) {
2462 error = EADDRNOTAVAIL; /* XXX: should not happen */
2463 break;
2464 }
2465
2466 /*
2467 * See if the membership already exists.
2468 */
2469 for (imm = im6o->im6o_memberships.lh_first;
2470 imm != NULL; imm = imm->i6mm_chain.le_next)
2471 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2472 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2473 &ia))
2474 break;
2475 if (imm != NULL) {
2476 error = EADDRINUSE;
2477 break;
2478 }
2479 /*
2480 * Everything looks good; add a new record to the multicast
2481 * address list for the given interface.
2482 */
2483 imm = in6_joingroup(ifp, &ia, &error, 0);
2484 if (imm == NULL)
2485 break;
2486 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2487 break;
2488
2489 case IPV6_LEAVE_GROUP:
2490 /*
2491 * Drop a multicast group membership.
2492 * Group must be a valid IP6 multicast address.
2493 */
2494 error = sockopt_get(sopt, &mreq, sizeof(mreq));
2495 if (error != 0)
2496 break;
2497
2498 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) {
2499 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt);
2500 break;
2501 }
2502 /*
2503 * If an interface address was specified, get a pointer
2504 * to its ifnet structure.
2505 */
2506 if (mreq.ipv6mr_interface != 0) {
2507 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) {
2508 error = ENXIO; /* XXX EINVAL? */
2509 break;
2510 }
2511 } else
2512 ifp = NULL;
2513
2514 /* Fill in the scope zone ID */
2515 if (ifp) {
2516 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2517 /* XXX: should not happen */
2518 error = EADDRNOTAVAIL;
2519 break;
2520 }
2521 } else if (mreq.ipv6mr_interface != 0) {
2522 /*
2523 * XXX: This case would happens when the (positive)
2524 * index is in the valid range, but the corresponding
2525 * interface has been detached dynamically. The above
2526 * check probably avoids such case to happen here, but
2527 * we check it explicitly for safety.
2528 */
2529 error = EADDRNOTAVAIL;
2530 break;
2531 } else { /* ipv6mr_interface == 0 */
2532 struct sockaddr_in6 sa6_mc;
2533
2534 /*
2535 * The API spec says as follows:
2536 * If the interface index is specified as 0, the
2537 * system may choose a multicast group membership to
2538 * drop by matching the multicast address only.
2539 * On the other hand, we cannot disambiguate the scope
2540 * zone unless an interface is provided. Thus, we
2541 * check if there's ambiguity with the default scope
2542 * zone as the last resort.
2543 */
2544 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2545 0, 0, 0);
2546 error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2547 if (error != 0)
2548 break;
2549 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2550 }
2551
2552 /*
2553 * Find the membership in the membership list.
2554 */
2555 for (imm = im6o->im6o_memberships.lh_first;
2556 imm != NULL; imm = imm->i6mm_chain.le_next) {
2557 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2558 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2559 &mreq.ipv6mr_multiaddr))
2560 break;
2561 }
2562 if (imm == NULL) {
2563 /* Unable to resolve interface */
2564 error = EADDRNOTAVAIL;
2565 break;
2566 }
2567 /*
2568 * Give up the multicast address record to which the
2569 * membership points.
2570 */
2571 LIST_REMOVE(imm, i6mm_chain);
2572 in6_leavegroup(imm);
2573 break;
2574
2575 default:
2576 error = EOPNOTSUPP;
2577 break;
2578 }
2579
2580 /*
2581 * If all options have default values, no need to keep the mbuf.
2582 */
2583 if (im6o->im6o_multicast_ifp == NULL &&
2584 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2585 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2586 im6o->im6o_memberships.lh_first == NULL) {
2587 free(in6p->in6p_moptions, M_IPMOPTS);
2588 in6p->in6p_moptions = NULL;
2589 }
2590
2591 return (error);
2592 }
2593
2594 /*
2595 * Return the IP6 multicast options in response to user getsockopt().
2596 */
2597 static int
2598 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p)
2599 {
2600 u_int optval;
2601 int error;
2602 struct ip6_moptions *im6o = in6p->in6p_moptions;
2603
2604 switch (sopt->sopt_name) {
2605 case IPV6_MULTICAST_IF:
2606 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2607 optval = 0;
2608 else
2609 optval = im6o->im6o_multicast_ifp->if_index;
2610
2611 error = sockopt_set(sopt, &optval, sizeof(optval));
2612 break;
2613
2614 case IPV6_MULTICAST_HOPS:
2615 if (im6o == NULL)
2616 optval = ip6_defmcasthlim;
2617 else
2618 optval = im6o->im6o_multicast_hlim;
2619
2620 error = sockopt_set(sopt, &optval, sizeof(optval));
2621 break;
2622
2623 case IPV6_MULTICAST_LOOP:
2624 if (im6o == NULL)
2625 optval = IPV6_DEFAULT_MULTICAST_LOOP;
2626 else
2627 optval = im6o->im6o_multicast_loop;
2628
2629 error = sockopt_set(sopt, &optval, sizeof(optval));
2630 break;
2631
2632 default:
2633 error = EOPNOTSUPP;
2634 }
2635
2636 return (error);
2637 }
2638
2639 /*
2640 * Discard the IP6 multicast options.
2641 */
2642 void
2643 ip6_freemoptions(struct ip6_moptions *im6o)
2644 {
2645 struct in6_multi_mship *imm;
2646
2647 if (im6o == NULL)
2648 return;
2649
2650 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2651 LIST_REMOVE(imm, i6mm_chain);
2652 in6_leavegroup(imm);
2653 }
2654 free(im6o, M_IPMOPTS);
2655 }
2656
2657 /*
2658 * Set IPv6 outgoing packet options based on advanced API.
2659 */
2660 int
2661 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2662 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2663 {
2664 struct cmsghdr *cm = 0;
2665
2666 if (control == NULL || opt == NULL)
2667 return (EINVAL);
2668
2669 ip6_initpktopts(opt);
2670 if (stickyopt) {
2671 int error;
2672
2673 /*
2674 * If stickyopt is provided, make a local copy of the options
2675 * for this particular packet, then override them by ancillary
2676 * objects.
2677 * XXX: copypktopts() does not copy the cached route to a next
2678 * hop (if any). This is not very good in terms of efficiency,
2679 * but we can allow this since this option should be rarely
2680 * used.
2681 */
2682 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2683 return (error);
2684 }
2685
2686 /*
2687 * XXX: Currently, we assume all the optional information is stored
2688 * in a single mbuf.
2689 */
2690 if (control->m_next)
2691 return (EINVAL);
2692
2693 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2694 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2695 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2696 int error;
2697
2698 if (control->m_len < CMSG_LEN(0))
2699 return (EINVAL);
2700
2701 cm = mtod(control, struct cmsghdr *);
2702 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2703 return (EINVAL);
2704 if (cm->cmsg_level != IPPROTO_IPV6)
2705 continue;
2706
2707 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2708 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2709 if (error)
2710 return (error);
2711 }
2712
2713 return (0);
2714 }
2715
2716 /*
2717 * Set a particular packet option, as a sticky option or an ancillary data
2718 * item. "len" can be 0 only when it's a sticky option.
2719 * We have 4 cases of combination of "sticky" and "cmsg":
2720 * "sticky=0, cmsg=0": impossible
2721 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2722 * "sticky=1, cmsg=0": RFC3542 socket option
2723 * "sticky=1, cmsg=1": RFC2292 socket option
2724 */
2725 static int
2726 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2727 kauth_cred_t cred, int sticky, int cmsg, int uproto)
2728 {
2729 int minmtupolicy;
2730 int error;
2731
2732 if (!sticky && !cmsg) {
2733 #ifdef DIAGNOSTIC
2734 printf("ip6_setpktopt: impossible case\n");
2735 #endif
2736 return (EINVAL);
2737 }
2738
2739 /*
2740 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2741 * not be specified in the context of RFC3542. Conversely,
2742 * RFC3542 types should not be specified in the context of RFC2292.
2743 */
2744 if (!cmsg) {
2745 switch (optname) {
2746 case IPV6_2292PKTINFO:
2747 case IPV6_2292HOPLIMIT:
2748 case IPV6_2292NEXTHOP:
2749 case IPV6_2292HOPOPTS:
2750 case IPV6_2292DSTOPTS:
2751 case IPV6_2292RTHDR:
2752 case IPV6_2292PKTOPTIONS:
2753 return (ENOPROTOOPT);
2754 }
2755 }
2756 if (sticky && cmsg) {
2757 switch (optname) {
2758 case IPV6_PKTINFO:
2759 case IPV6_HOPLIMIT:
2760 case IPV6_NEXTHOP:
2761 case IPV6_HOPOPTS:
2762 case IPV6_DSTOPTS:
2763 case IPV6_RTHDRDSTOPTS:
2764 case IPV6_RTHDR:
2765 case IPV6_USE_MIN_MTU:
2766 case IPV6_DONTFRAG:
2767 case IPV6_OTCLASS:
2768 case IPV6_TCLASS:
2769 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */
2770 return (ENOPROTOOPT);
2771 }
2772 }
2773
2774 switch (optname) {
2775 #ifdef RFC2292
2776 case IPV6_2292PKTINFO:
2777 #endif
2778 case IPV6_PKTINFO:
2779 {
2780 struct ifnet *ifp = NULL;
2781 struct in6_pktinfo *pktinfo;
2782
2783 if (len != sizeof(struct in6_pktinfo))
2784 return (EINVAL);
2785
2786 pktinfo = (struct in6_pktinfo *)buf;
2787
2788 /*
2789 * An application can clear any sticky IPV6_PKTINFO option by
2790 * doing a "regular" setsockopt with ipi6_addr being
2791 * in6addr_any and ipi6_ifindex being zero.
2792 * [RFC 3542, Section 6]
2793 */
2794 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2795 pktinfo->ipi6_ifindex == 0 &&
2796 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2797 ip6_clearpktopts(opt, optname);
2798 break;
2799 }
2800
2801 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2802 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2803 return (EINVAL);
2804 }
2805
2806 /* Validate the interface index if specified. */
2807 if (pktinfo->ipi6_ifindex) {
2808 ifp = if_byindex(pktinfo->ipi6_ifindex);
2809 if (ifp == NULL)
2810 return (ENXIO);
2811 }
2812
2813 /*
2814 * We store the address anyway, and let in6_selectsrc()
2815 * validate the specified address. This is because ipi6_addr
2816 * may not have enough information about its scope zone, and
2817 * we may need additional information (such as outgoing
2818 * interface or the scope zone of a destination address) to
2819 * disambiguate the scope.
2820 * XXX: the delay of the validation may confuse the
2821 * application when it is used as a sticky option.
2822 */
2823 if (opt->ip6po_pktinfo == NULL) {
2824 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2825 M_IP6OPT, M_NOWAIT);
2826 if (opt->ip6po_pktinfo == NULL)
2827 return (ENOBUFS);
2828 }
2829 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2830 break;
2831 }
2832
2833 #ifdef RFC2292
2834 case IPV6_2292HOPLIMIT:
2835 #endif
2836 case IPV6_HOPLIMIT:
2837 {
2838 int *hlimp;
2839
2840 /*
2841 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2842 * to simplify the ordering among hoplimit options.
2843 */
2844 if (optname == IPV6_HOPLIMIT && sticky)
2845 return (ENOPROTOOPT);
2846
2847 if (len != sizeof(int))
2848 return (EINVAL);
2849 hlimp = (int *)buf;
2850 if (*hlimp < -1 || *hlimp > 255)
2851 return (EINVAL);
2852
2853 opt->ip6po_hlim = *hlimp;
2854 break;
2855 }
2856
2857 case IPV6_OTCLASS:
2858 if (len != sizeof(u_int8_t))
2859 return (EINVAL);
2860
2861 opt->ip6po_tclass = *(u_int8_t *)buf;
2862 break;
2863
2864 case IPV6_TCLASS:
2865 {
2866 int tclass;
2867
2868 if (len != sizeof(int))
2869 return (EINVAL);
2870 tclass = *(int *)buf;
2871 if (tclass < -1 || tclass > 255)
2872 return (EINVAL);
2873
2874 opt->ip6po_tclass = tclass;
2875 break;
2876 }
2877
2878 #ifdef RFC2292
2879 case IPV6_2292NEXTHOP:
2880 #endif
2881 case IPV6_NEXTHOP:
2882 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2883 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2884 if (error)
2885 return (error);
2886
2887 if (len == 0) { /* just remove the option */
2888 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2889 break;
2890 }
2891
2892 /* check if cmsg_len is large enough for sa_len */
2893 if (len < sizeof(struct sockaddr) || len < *buf)
2894 return (EINVAL);
2895
2896 switch (((struct sockaddr *)buf)->sa_family) {
2897 case AF_INET6:
2898 {
2899 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2900
2901 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2902 return (EINVAL);
2903
2904 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2905 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2906 return (EINVAL);
2907 }
2908 if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2909 != 0) {
2910 return (error);
2911 }
2912 break;
2913 }
2914 case AF_LINK: /* eventually be supported? */
2915 default:
2916 return (EAFNOSUPPORT);
2917 }
2918
2919 /* turn off the previous option, then set the new option. */
2920 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2921 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2922 if (opt->ip6po_nexthop == NULL)
2923 return (ENOBUFS);
2924 memcpy(opt->ip6po_nexthop, buf, *buf);
2925 break;
2926
2927 #ifdef RFC2292
2928 case IPV6_2292HOPOPTS:
2929 #endif
2930 case IPV6_HOPOPTS:
2931 {
2932 struct ip6_hbh *hbh;
2933 int hbhlen;
2934
2935 /*
2936 * XXX: We don't allow a non-privileged user to set ANY HbH
2937 * options, since per-option restriction has too much
2938 * overhead.
2939 */
2940 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2941 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2942 if (error)
2943 return (error);
2944
2945 if (len == 0) {
2946 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2947 break; /* just remove the option */
2948 }
2949
2950 /* message length validation */
2951 if (len < sizeof(struct ip6_hbh))
2952 return (EINVAL);
2953 hbh = (struct ip6_hbh *)buf;
2954 hbhlen = (hbh->ip6h_len + 1) << 3;
2955 if (len != hbhlen)
2956 return (EINVAL);
2957
2958 /* turn off the previous option, then set the new option. */
2959 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2960 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2961 if (opt->ip6po_hbh == NULL)
2962 return (ENOBUFS);
2963 memcpy(opt->ip6po_hbh, hbh, hbhlen);
2964
2965 break;
2966 }
2967
2968 #ifdef RFC2292
2969 case IPV6_2292DSTOPTS:
2970 #endif
2971 case IPV6_DSTOPTS:
2972 case IPV6_RTHDRDSTOPTS:
2973 {
2974 struct ip6_dest *dest, **newdest = NULL;
2975 int destlen;
2976
2977 /* XXX: see the comment for IPV6_HOPOPTS */
2978 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2979 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2980 if (error)
2981 return (error);
2982
2983 if (len == 0) {
2984 ip6_clearpktopts(opt, optname);
2985 break; /* just remove the option */
2986 }
2987
2988 /* message length validation */
2989 if (len < sizeof(struct ip6_dest))
2990 return (EINVAL);
2991 dest = (struct ip6_dest *)buf;
2992 destlen = (dest->ip6d_len + 1) << 3;
2993 if (len != destlen)
2994 return (EINVAL);
2995 /*
2996 * Determine the position that the destination options header
2997 * should be inserted; before or after the routing header.
2998 */
2999 switch (optname) {
3000 case IPV6_2292DSTOPTS:
3001 /*
3002 * The old advanced API is ambiguous on this point.
3003 * Our approach is to determine the position based
3004 * according to the existence of a routing header.
3005 * Note, however, that this depends on the order of the
3006 * extension headers in the ancillary data; the 1st
3007 * part of the destination options header must appear
3008 * before the routing header in the ancillary data,
3009 * too.
3010 * RFC3542 solved the ambiguity by introducing
3011 * separate ancillary data or option types.
3012 */
3013 if (opt->ip6po_rthdr == NULL)
3014 newdest = &opt->ip6po_dest1;
3015 else
3016 newdest = &opt->ip6po_dest2;
3017 break;
3018 case IPV6_RTHDRDSTOPTS:
3019 newdest = &opt->ip6po_dest1;
3020 break;
3021 case IPV6_DSTOPTS:
3022 newdest = &opt->ip6po_dest2;
3023 break;
3024 }
3025
3026 /* turn off the previous option, then set the new option. */
3027 ip6_clearpktopts(opt, optname);
3028 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3029 if (*newdest == NULL)
3030 return (ENOBUFS);
3031 memcpy(*newdest, dest, destlen);
3032
3033 break;
3034 }
3035
3036 #ifdef RFC2292
3037 case IPV6_2292RTHDR:
3038 #endif
3039 case IPV6_RTHDR:
3040 {
3041 struct ip6_rthdr *rth;
3042 int rthlen;
3043
3044 if (len == 0) {
3045 ip6_clearpktopts(opt, IPV6_RTHDR);
3046 break; /* just remove the option */
3047 }
3048
3049 /* message length validation */
3050 if (len < sizeof(struct ip6_rthdr))
3051 return (EINVAL);
3052 rth = (struct ip6_rthdr *)buf;
3053 rthlen = (rth->ip6r_len + 1) << 3;
3054 if (len != rthlen)
3055 return (EINVAL);
3056 switch (rth->ip6r_type) {
3057 case IPV6_RTHDR_TYPE_0:
3058 if (rth->ip6r_len == 0) /* must contain one addr */
3059 return (EINVAL);
3060 if (rth->ip6r_len % 2) /* length must be even */
3061 return (EINVAL);
3062 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3063 return (EINVAL);
3064 break;
3065 default:
3066 return (EINVAL); /* not supported */
3067 }
3068 /* turn off the previous option */
3069 ip6_clearpktopts(opt, IPV6_RTHDR);
3070 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3071 if (opt->ip6po_rthdr == NULL)
3072 return (ENOBUFS);
3073 memcpy(opt->ip6po_rthdr, rth, rthlen);
3074 break;
3075 }
3076
3077 case IPV6_USE_MIN_MTU:
3078 if (len != sizeof(int))
3079 return (EINVAL);
3080 minmtupolicy = *(int *)buf;
3081 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3082 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3083 minmtupolicy != IP6PO_MINMTU_ALL) {
3084 return (EINVAL);
3085 }
3086 opt->ip6po_minmtu = minmtupolicy;
3087 break;
3088
3089 case IPV6_DONTFRAG:
3090 if (len != sizeof(int))
3091 return (EINVAL);
3092
3093 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3094 /*
3095 * we ignore this option for TCP sockets.
3096 * (RFC3542 leaves this case unspecified.)
3097 */
3098 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3099 } else
3100 opt->ip6po_flags |= IP6PO_DONTFRAG;
3101 break;
3102
3103 case IPV6_PREFER_TEMPADDR:
3104 {
3105 int preftemp;
3106
3107 if (len != sizeof(int))
3108 return (EINVAL);
3109 preftemp = *(int *)buf;
3110 switch (preftemp) {
3111 case IP6PO_TEMPADDR_SYSTEM:
3112 case IP6PO_TEMPADDR_NOTPREFER:
3113 case IP6PO_TEMPADDR_PREFER:
3114 break;
3115 default:
3116 return (EINVAL);
3117 }
3118 opt->ip6po_prefer_tempaddr = preftemp;
3119 break;
3120 }
3121
3122 default:
3123 return (ENOPROTOOPT);
3124 } /* end of switch */
3125
3126 return (0);
3127 }
3128
3129 /*
3130 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3131 * packet to the input queue of a specified interface. Note that this
3132 * calls the output routine of the loopback "driver", but with an interface
3133 * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3134 */
3135 void
3136 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3137 const struct sockaddr_in6 *dst)
3138 {
3139 struct mbuf *copym;
3140 struct ip6_hdr *ip6;
3141
3142 copym = m_copy(m, 0, M_COPYALL);
3143 if (copym == NULL)
3144 return;
3145
3146 /*
3147 * Make sure to deep-copy IPv6 header portion in case the data
3148 * is in an mbuf cluster, so that we can safely override the IPv6
3149 * header portion later.
3150 */
3151 if ((copym->m_flags & M_EXT) != 0 ||
3152 copym->m_len < sizeof(struct ip6_hdr)) {
3153 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3154 if (copym == NULL)
3155 return;
3156 }
3157
3158 #ifdef DIAGNOSTIC
3159 if (copym->m_len < sizeof(*ip6)) {
3160 m_freem(copym);
3161 return;
3162 }
3163 #endif
3164
3165 ip6 = mtod(copym, struct ip6_hdr *);
3166 /*
3167 * clear embedded scope identifiers if necessary.
3168 * in6_clearscope will touch the addresses only when necessary.
3169 */
3170 in6_clearscope(&ip6->ip6_src);
3171 in6_clearscope(&ip6->ip6_dst);
3172
3173 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3174 }
3175
3176 /*
3177 * Chop IPv6 header off from the payload.
3178 */
3179 static int
3180 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3181 {
3182 struct mbuf *mh;
3183 struct ip6_hdr *ip6;
3184
3185 ip6 = mtod(m, struct ip6_hdr *);
3186 if (m->m_len > sizeof(*ip6)) {
3187 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3188 if (mh == 0) {
3189 m_freem(m);
3190 return ENOBUFS;
3191 }
3192 M_MOVE_PKTHDR(mh, m);
3193 MH_ALIGN(mh, sizeof(*ip6));
3194 m->m_len -= sizeof(*ip6);
3195 m->m_data += sizeof(*ip6);
3196 mh->m_next = m;
3197 m = mh;
3198 m->m_len = sizeof(*ip6);
3199 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3200 }
3201 exthdrs->ip6e_ip6 = m;
3202 return 0;
3203 }
3204
3205 /*
3206 * Compute IPv6 extension header length.
3207 */
3208 int
3209 ip6_optlen(struct in6pcb *in6p)
3210 {
3211 int len;
3212
3213 if (!in6p->in6p_outputopts)
3214 return 0;
3215
3216 len = 0;
3217 #define elen(x) \
3218 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3219
3220 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3221 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3222 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3223 len += elen(in6p->in6p_outputopts->ip6po_dest2);
3224 return len;
3225 #undef elen
3226 }
3227