1 /*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30 */
31
32 /*-
33 * Copyright (c) 1982, 1986, 1988, 1990, 1993
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
61 */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD: releng/10.0/sys/netinet6/ip6_output.c 254804 2013-08-24 19:51:18Z andre $");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_sctp.h"
71 #include "opt_route.h"
72
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85
86 #include <machine/in_cksum.h>
87
88 #include <net/if.h>
89 #include <net/netisr.h>
90 #include <net/route.h>
91 #include <net/pfil.h>
92 #include <net/vnet.h>
93
94 #include <netinet/in.h>
95 #include <netinet/in_var.h>
96 #include <netinet/ip_var.h>
97 #include <netinet6/in6_var.h>
98 #include <netinet/ip6.h>
99 #include <netinet/icmp6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/tcp_var.h>
103 #include <netinet6/nd6.h>
104
105 #ifdef IPSEC
106 #include <netipsec/ipsec.h>
107 #include <netipsec/ipsec6.h>
108 #include <netipsec/key.h>
109 #include <netinet6/ip6_ipsec.h>
110 #endif /* IPSEC */
111 #ifdef SCTP
112 #include <netinet/sctp.h>
113 #include <netinet/sctp_crc32.h>
114 #endif
115
116 #include <netinet6/ip6protosw.h>
117 #include <netinet6/scope6_var.h>
118
119 #ifdef FLOWTABLE
120 #include <net/flowtable.h>
121 #endif
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 int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
134 struct ucred *, int);
135 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
136 struct socket *, struct sockopt *);
137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
138 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
139 struct ucred *, int, int, int);
140
141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
142 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
143 struct ip6_frag **);
144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
146 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
147 struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
148 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
149
150
151 /*
152 * Make an extension header from option data. hp is the source, and
153 * mp is the destination.
154 */
155 #define MAKE_EXTHDR(hp, mp) \
156 do { \
157 if (hp) { \
158 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
159 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
160 ((eh)->ip6e_len + 1) << 3); \
161 if (error) \
162 goto freehdrs; \
163 } \
164 } while (/*CONSTCOND*/ 0)
165
166 /*
167 * Form a chain of extension headers.
168 * m is the extension header mbuf
169 * mp is the previous mbuf in the chain
170 * p is the next header
171 * i is the type of option.
172 */
173 #define MAKE_CHAIN(m, mp, p, i)\
174 do {\
175 if (m) {\
176 if (!hdrsplit) \
177 panic("assumption failed: hdr not split"); \
178 *mtod((m), u_char *) = *(p);\
179 *(p) = (i);\
180 p = mtod((m), u_char *);\
181 (m)->m_next = (mp)->m_next;\
182 (mp)->m_next = (m);\
183 (mp) = (m);\
184 }\
185 } while (/*CONSTCOND*/ 0)
186
187 static void
188 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
189 {
190 u_short csum;
191
192 csum = in_cksum_skip(m, offset + plen, offset);
193 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
194 csum = 0xffff;
195 offset += m->m_pkthdr.csum_data; /* checksum offset */
196
197 if (offset + sizeof(u_short) > m->m_len) {
198 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
199 "csum_flags=%b\n", __func__, m->m_len, plen, offset,
200 (int)m->m_pkthdr.csum_flags, CSUM_BITS);
201 /*
202 * XXX this should not happen, but if it does, the correct
203 * behavior may be to insert the checksum in the appropriate
204 * next mbuf in the chain.
205 */
206 return;
207 }
208 *(u_short *)(m->m_data + offset) = csum;
209 }
210
211 /*
212 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
213 * header (with pri, len, nxt, hlim, src, dst).
214 * This function may modify ver and hlim only.
215 * The mbuf chain containing the packet will be freed.
216 * The mbuf opt, if present, will not be freed.
217 * If route_in6 ro is present and has ro_rt initialized, route lookup would be
218 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
219 * then result of route lookup is stored in ro->ro_rt.
220 *
221 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
222 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
223 * which is rt_rmx.rmx_mtu.
224 *
225 * ifpp - XXX: just for statistics
226 */
227 int
228 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
229 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
230 struct ifnet **ifpp, struct inpcb *inp)
231 {
232 struct ip6_hdr *ip6, *mhip6;
233 struct ifnet *ifp, *origifp;
234 struct mbuf *m = m0;
235 struct mbuf *mprev = NULL;
236 int hlen, tlen, len, off;
237 struct route_in6 ip6route;
238 struct rtentry *rt = NULL;
239 struct sockaddr_in6 *dst, src_sa, dst_sa;
240 struct in6_addr odst;
241 int error = 0;
242 struct in6_ifaddr *ia = NULL;
243 u_long mtu;
244 int alwaysfrag, dontfrag;
245 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
246 struct ip6_exthdrs exthdrs;
247 struct in6_addr finaldst, src0, dst0;
248 u_int32_t zone;
249 struct route_in6 *ro_pmtu = NULL;
250 int hdrsplit = 0;
251 int needipsec = 0;
252 int sw_csum, tso;
253 #ifdef IPSEC
254 struct ipsec_output_state state;
255 struct ip6_rthdr *rh = NULL;
256 int needipsectun = 0;
257 int segleft_org = 0;
258 struct secpolicy *sp = NULL;
259 #endif /* IPSEC */
260 struct m_tag *fwd_tag = NULL;
261
262 ip6 = mtod(m, struct ip6_hdr *);
263 if (ip6 == NULL) {
264 printf ("ip6 is NULL");
265 goto bad;
266 }
267
268 if (inp != NULL)
269 M_SETFIB(m, inp->inp_inc.inc_fibnum);
270
271 finaldst = ip6->ip6_dst;
272 bzero(&exthdrs, sizeof(exthdrs));
273 if (opt) {
274 /* Hop-by-Hop options header */
275 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
276 /* Destination options header(1st part) */
277 if (opt->ip6po_rthdr) {
278 /*
279 * Destination options header(1st part)
280 * This only makes sense with a routing header.
281 * See Section 9.2 of RFC 3542.
282 * Disabling this part just for MIP6 convenience is
283 * a bad idea. We need to think carefully about a
284 * way to make the advanced API coexist with MIP6
285 * options, which might automatically be inserted in
286 * the kernel.
287 */
288 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
289 }
290 /* Routing header */
291 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
292 /* Destination options header(2nd part) */
293 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
294 }
295
296 #ifdef IPSEC
297 /*
298 * IPSec checking which handles several cases.
299 * FAST IPSEC: We re-injected the packet.
300 */
301 switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
302 {
303 case 1: /* Bad packet */
304 goto freehdrs;
305 case -1: /* Do IPSec */
306 needipsec = 1;
307 /*
308 * Do delayed checksums now, as we may send before returning.
309 */
310 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
311 plen = m->m_pkthdr.len - sizeof(*ip6);
312 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
313 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
314 }
315 #ifdef SCTP
316 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
317 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
318 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
319 }
320 #endif
321 case 0: /* No IPSec */
322 default:
323 break;
324 }
325 #endif /* IPSEC */
326
327 /*
328 * Calculate the total length of the extension header chain.
329 * Keep the length of the unfragmentable part for fragmentation.
330 */
331 optlen = 0;
332 if (exthdrs.ip6e_hbh)
333 optlen += exthdrs.ip6e_hbh->m_len;
334 if (exthdrs.ip6e_dest1)
335 optlen += exthdrs.ip6e_dest1->m_len;
336 if (exthdrs.ip6e_rthdr)
337 optlen += exthdrs.ip6e_rthdr->m_len;
338 unfragpartlen = optlen + sizeof(struct ip6_hdr);
339
340 /* NOTE: we don't add AH/ESP length here. do that later. */
341 if (exthdrs.ip6e_dest2)
342 optlen += exthdrs.ip6e_dest2->m_len;
343
344 /*
345 * If we need IPsec, or there is at least one extension header,
346 * separate IP6 header from the payload.
347 */
348 if ((needipsec || optlen) && !hdrsplit) {
349 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
350 m = NULL;
351 goto freehdrs;
352 }
353 m = exthdrs.ip6e_ip6;
354 hdrsplit++;
355 }
356
357 /* adjust pointer */
358 ip6 = mtod(m, struct ip6_hdr *);
359
360 /* adjust mbuf packet header length */
361 m->m_pkthdr.len += optlen;
362 plen = m->m_pkthdr.len - sizeof(*ip6);
363
364 /* If this is a jumbo payload, insert a jumbo payload option. */
365 if (plen > IPV6_MAXPACKET) {
366 if (!hdrsplit) {
367 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
368 m = NULL;
369 goto freehdrs;
370 }
371 m = exthdrs.ip6e_ip6;
372 hdrsplit++;
373 }
374 /* adjust pointer */
375 ip6 = mtod(m, struct ip6_hdr *);
376 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
377 goto freehdrs;
378 ip6->ip6_plen = 0;
379 } else
380 ip6->ip6_plen = htons(plen);
381
382 /*
383 * Concatenate headers and fill in next header fields.
384 * Here we have, on "m"
385 * IPv6 payload
386 * and we insert headers accordingly. Finally, we should be getting:
387 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
388 *
389 * during the header composing process, "m" points to IPv6 header.
390 * "mprev" points to an extension header prior to esp.
391 */
392 u_char *nexthdrp = &ip6->ip6_nxt;
393 mprev = m;
394
395 /*
396 * we treat dest2 specially. this makes IPsec processing
397 * much easier. the goal here is to make mprev point the
398 * mbuf prior to dest2.
399 *
400 * result: IPv6 dest2 payload
401 * m and mprev will point to IPv6 header.
402 */
403 if (exthdrs.ip6e_dest2) {
404 if (!hdrsplit)
405 panic("assumption failed: hdr not split");
406 exthdrs.ip6e_dest2->m_next = m->m_next;
407 m->m_next = exthdrs.ip6e_dest2;
408 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
409 ip6->ip6_nxt = IPPROTO_DSTOPTS;
410 }
411
412 /*
413 * result: IPv6 hbh dest1 rthdr dest2 payload
414 * m will point to IPv6 header. mprev will point to the
415 * extension header prior to dest2 (rthdr in the above case).
416 */
417 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
418 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
419 IPPROTO_DSTOPTS);
420 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
421 IPPROTO_ROUTING);
422
423 #ifdef IPSEC
424 if (!needipsec)
425 goto skip_ipsec2;
426
427 /*
428 * pointers after IPsec headers are not valid any more.
429 * other pointers need a great care too.
430 * (IPsec routines should not mangle mbufs prior to AH/ESP)
431 */
432 exthdrs.ip6e_dest2 = NULL;
433
434 if (exthdrs.ip6e_rthdr) {
435 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
436 segleft_org = rh->ip6r_segleft;
437 rh->ip6r_segleft = 0;
438 }
439
440 bzero(&state, sizeof(state));
441 state.m = m;
442 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
443 &needipsectun);
444 m = state.m;
445 if (error == EJUSTRETURN) {
446 /*
447 * We had a SP with a level of 'use' and no SA. We
448 * will just continue to process the packet without
449 * IPsec processing.
450 */
451 ;
452 } else if (error) {
453 /* mbuf is already reclaimed in ipsec6_output_trans. */
454 m = NULL;
455 switch (error) {
456 case EHOSTUNREACH:
457 case ENETUNREACH:
458 case EMSGSIZE:
459 case ENOBUFS:
460 case ENOMEM:
461 break;
462 default:
463 printf("[%s:%d] (ipsec): error code %d\n",
464 __func__, __LINE__, error);
465 /* FALLTHROUGH */
466 case ENOENT:
467 /* don't show these error codes to the user */
468 error = 0;
469 break;
470 }
471 goto bad;
472 } else if (!needipsectun) {
473 /*
474 * In the FAST IPSec case we have already
475 * re-injected the packet and it has been freed
476 * by the ipsec_done() function. So, just clean
477 * up after ourselves.
478 */
479 m = NULL;
480 goto done;
481 }
482 if (exthdrs.ip6e_rthdr) {
483 /* ah6_output doesn't modify mbuf chain */
484 rh->ip6r_segleft = segleft_org;
485 }
486 skip_ipsec2:;
487 #endif /* IPSEC */
488
489 /*
490 * If there is a routing header, discard the packet.
491 */
492 if (exthdrs.ip6e_rthdr) {
493 error = EINVAL;
494 goto bad;
495 }
496
497 /* Source address validation */
498 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
499 (flags & IPV6_UNSPECSRC) == 0) {
500 error = EOPNOTSUPP;
501 IP6STAT_INC(ip6s_badscope);
502 goto bad;
503 }
504 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
505 error = EOPNOTSUPP;
506 IP6STAT_INC(ip6s_badscope);
507 goto bad;
508 }
509
510 IP6STAT_INC(ip6s_localout);
511
512 /*
513 * Route packet.
514 */
515 if (ro == 0) {
516 ro = &ip6route;
517 bzero((caddr_t)ro, sizeof(*ro));
518 }
519 ro_pmtu = ro;
520 if (opt && opt->ip6po_rthdr)
521 ro = &opt->ip6po_route;
522 dst = (struct sockaddr_in6 *)&ro->ro_dst;
523 #ifdef FLOWTABLE
524 if (ro->ro_rt == NULL) {
525 struct flentry *fle;
526
527 /*
528 * The flow table returns route entries valid for up to 30
529 * seconds; we rely on the remainder of ip_output() taking no
530 * longer than that long for the stability of ro_rt. The
531 * flow ID assignment must have happened before this point.
532 */
533 fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6);
534 if (fle != NULL)
535 flow_to_route_in6(fle, ro);
536 }
537 #endif
538 again:
539 /*
540 * if specified, try to fill in the traffic class field.
541 * do not override if a non-zero value is already set.
542 * we check the diffserv field and the ecn field separately.
543 */
544 if (opt && opt->ip6po_tclass >= 0) {
545 int mask = 0;
546
547 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
548 mask |= 0xfc;
549 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
550 mask |= 0x03;
551 if (mask != 0)
552 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
553 }
554
555 /* fill in or override the hop limit field, if necessary. */
556 if (opt && opt->ip6po_hlim != -1)
557 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
558 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
559 if (im6o != NULL)
560 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
561 else
562 ip6->ip6_hlim = V_ip6_defmcasthlim;
563 }
564
565 #ifdef IPSEC
566 /*
567 * We may re-inject packets into the stack here.
568 */
569 if (needipsec && needipsectun) {
570 struct ipsec_output_state state;
571
572 /*
573 * All the extension headers will become inaccessible
574 * (since they can be encrypted).
575 * Don't panic, we need no more updates to extension headers
576 * on inner IPv6 packet (since they are now encapsulated).
577 *
578 * IPv6 [ESP|AH] IPv6 [extension headers] payload
579 */
580 bzero(&exthdrs, sizeof(exthdrs));
581 exthdrs.ip6e_ip6 = m;
582
583 bzero(&state, sizeof(state));
584 state.m = m;
585 state.ro = (struct route *)ro;
586 state.dst = (struct sockaddr *)dst;
587
588 error = ipsec6_output_tunnel(&state, sp, flags);
589
590 m = state.m;
591 ro = (struct route_in6 *)state.ro;
592 dst = (struct sockaddr_in6 *)state.dst;
593 if (error == EJUSTRETURN) {
594 /*
595 * We had a SP with a level of 'use' and no SA. We
596 * will just continue to process the packet without
597 * IPsec processing.
598 */
599 ;
600 } else if (error) {
601 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
602 m0 = m = NULL;
603 m = NULL;
604 switch (error) {
605 case EHOSTUNREACH:
606 case ENETUNREACH:
607 case EMSGSIZE:
608 case ENOBUFS:
609 case ENOMEM:
610 break;
611 default:
612 printf("[%s:%d] (ipsec): error code %d\n",
613 __func__, __LINE__, error);
614 /* FALLTHROUGH */
615 case ENOENT:
616 /* don't show these error codes to the user */
617 error = 0;
618 break;
619 }
620 goto bad;
621 } else {
622 /*
623 * In the FAST IPSec case we have already
624 * re-injected the packet and it has been freed
625 * by the ipsec_done() function. So, just clean
626 * up after ourselves.
627 */
628 m = NULL;
629 goto done;
630 }
631
632 exthdrs.ip6e_ip6 = m;
633 }
634 #endif /* IPSEC */
635
636 /* adjust pointer */
637 ip6 = mtod(m, struct ip6_hdr *);
638
639 if (ro->ro_rt && fwd_tag == NULL) {
640 rt = ro->ro_rt;
641 ifp = ro->ro_rt->rt_ifp;
642 } else {
643 if (fwd_tag == NULL) {
644 bzero(&dst_sa, sizeof(dst_sa));
645 dst_sa.sin6_family = AF_INET6;
646 dst_sa.sin6_len = sizeof(dst_sa);
647 dst_sa.sin6_addr = ip6->ip6_dst;
648 }
649 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
650 &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
651 if (error != 0) {
652 if (ifp != NULL)
653 in6_ifstat_inc(ifp, ifs6_out_discard);
654 goto bad;
655 }
656 }
657 if (rt == NULL) {
658 /*
659 * If in6_selectroute() does not return a route entry,
660 * dst may not have been updated.
661 */
662 *dst = dst_sa; /* XXX */
663 }
664
665 /*
666 * then rt (for unicast) and ifp must be non-NULL valid values.
667 */
668 if ((flags & IPV6_FORWARDING) == 0) {
669 /* XXX: the FORWARDING flag can be set for mrouting. */
670 in6_ifstat_inc(ifp, ifs6_out_request);
671 }
672 if (rt != NULL) {
673 ia = (struct in6_ifaddr *)(rt->rt_ifa);
674 rt->rt_use++;
675 }
676
677
678 /*
679 * The outgoing interface must be in the zone of source and
680 * destination addresses.
681 */
682 origifp = ifp;
683
684 src0 = ip6->ip6_src;
685 if (in6_setscope(&src0, origifp, &zone))
686 goto badscope;
687 bzero(&src_sa, sizeof(src_sa));
688 src_sa.sin6_family = AF_INET6;
689 src_sa.sin6_len = sizeof(src_sa);
690 src_sa.sin6_addr = ip6->ip6_src;
691 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
692 goto badscope;
693
694 dst0 = ip6->ip6_dst;
695 if (in6_setscope(&dst0, origifp, &zone))
696 goto badscope;
697 /* re-initialize to be sure */
698 bzero(&dst_sa, sizeof(dst_sa));
699 dst_sa.sin6_family = AF_INET6;
700 dst_sa.sin6_len = sizeof(dst_sa);
701 dst_sa.sin6_addr = ip6->ip6_dst;
702 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
703 goto badscope;
704 }
705
706 /* We should use ia_ifp to support the case of
707 * sending packets to an address of our own.
708 */
709 if (ia != NULL && ia->ia_ifp)
710 ifp = ia->ia_ifp;
711
712 /* scope check is done. */
713 goto routefound;
714
715 badscope:
716 IP6STAT_INC(ip6s_badscope);
717 in6_ifstat_inc(origifp, ifs6_out_discard);
718 if (error == 0)
719 error = EHOSTUNREACH; /* XXX */
720 goto bad;
721
722 routefound:
723 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
724 if (opt && opt->ip6po_nextroute.ro_rt) {
725 /*
726 * The nexthop is explicitly specified by the
727 * application. We assume the next hop is an IPv6
728 * address.
729 */
730 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
731 }
732 else if ((rt->rt_flags & RTF_GATEWAY))
733 dst = (struct sockaddr_in6 *)rt->rt_gateway;
734 }
735
736 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
737 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
738 } else {
739 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
740 in6_ifstat_inc(ifp, ifs6_out_mcast);
741 /*
742 * Confirm that the outgoing interface supports multicast.
743 */
744 if (!(ifp->if_flags & IFF_MULTICAST)) {
745 IP6STAT_INC(ip6s_noroute);
746 in6_ifstat_inc(ifp, ifs6_out_discard);
747 error = ENETUNREACH;
748 goto bad;
749 }
750 if ((im6o == NULL && in6_mcast_loop) ||
751 (im6o && im6o->im6o_multicast_loop)) {
752 /*
753 * Loop back multicast datagram if not expressly
754 * forbidden to do so, even if we have not joined
755 * the address; protocols will filter it later,
756 * thus deferring a hash lookup and lock acquisition
757 * at the expense of an m_copym().
758 */
759 ip6_mloopback(ifp, m, dst);
760 } else {
761 /*
762 * If we are acting as a multicast router, perform
763 * multicast forwarding as if the packet had just
764 * arrived on the interface to which we are about
765 * to send. The multicast forwarding function
766 * recursively calls this function, using the
767 * IPV6_FORWARDING flag to prevent infinite recursion.
768 *
769 * Multicasts that are looped back by ip6_mloopback(),
770 * above, will be forwarded by the ip6_input() routine,
771 * if necessary.
772 */
773 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
774 /*
775 * XXX: ip6_mforward expects that rcvif is NULL
776 * when it is called from the originating path.
777 * However, it may not always be the case.
778 */
779 m->m_pkthdr.rcvif = NULL;
780 if (ip6_mforward(ip6, ifp, m) != 0) {
781 m_freem(m);
782 goto done;
783 }
784 }
785 }
786 /*
787 * Multicasts with a hoplimit of zero may be looped back,
788 * above, but must not be transmitted on a network.
789 * Also, multicasts addressed to the loopback interface
790 * are not sent -- the above call to ip6_mloopback() will
791 * loop back a copy if this host actually belongs to the
792 * destination group on the loopback interface.
793 */
794 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
795 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
796 m_freem(m);
797 goto done;
798 }
799 }
800
801 /*
802 * Fill the outgoing inteface to tell the upper layer
803 * to increment per-interface statistics.
804 */
805 if (ifpp)
806 *ifpp = ifp;
807
808 /* Determine path MTU. */
809 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
810 &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
811 goto bad;
812
813 /*
814 * The caller of this function may specify to use the minimum MTU
815 * in some cases.
816 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
817 * setting. The logic is a bit complicated; by default, unicast
818 * packets will follow path MTU while multicast packets will be sent at
819 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
820 * including unicast ones will be sent at the minimum MTU. Multicast
821 * packets will always be sent at the minimum MTU unless
822 * IP6PO_MINMTU_DISABLE is explicitly specified.
823 * See RFC 3542 for more details.
824 */
825 if (mtu > IPV6_MMTU) {
826 if ((flags & IPV6_MINMTU))
827 mtu = IPV6_MMTU;
828 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
829 mtu = IPV6_MMTU;
830 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
831 (opt == NULL ||
832 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
833 mtu = IPV6_MMTU;
834 }
835 }
836
837 /*
838 * clear embedded scope identifiers if necessary.
839 * in6_clearscope will touch the addresses only when necessary.
840 */
841 in6_clearscope(&ip6->ip6_src);
842 in6_clearscope(&ip6->ip6_dst);
843
844 /*
845 * If the outgoing packet contains a hop-by-hop options header,
846 * it must be examined and processed even by the source node.
847 * (RFC 2460, section 4.)
848 */
849 if (exthdrs.ip6e_hbh) {
850 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
851 u_int32_t dummy; /* XXX unused */
852 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
853
854 #ifdef DIAGNOSTIC
855 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
856 panic("ip6e_hbh is not contiguous");
857 #endif
858 /*
859 * XXX: if we have to send an ICMPv6 error to the sender,
860 * we need the M_LOOP flag since icmp6_error() expects
861 * the IPv6 and the hop-by-hop options header are
862 * contiguous unless the flag is set.
863 */
864 m->m_flags |= M_LOOP;
865 m->m_pkthdr.rcvif = ifp;
866 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
867 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
868 &dummy, &plen) < 0) {
869 /* m was already freed at this point */
870 error = EINVAL;/* better error? */
871 goto done;
872 }
873 m->m_flags &= ~M_LOOP; /* XXX */
874 m->m_pkthdr.rcvif = NULL;
875 }
876
877 /* Jump over all PFIL processing if hooks are not active. */
878 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
879 goto passout;
880
881 odst = ip6->ip6_dst;
882 /* Run through list of hooks for output packets. */
883 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
884 if (error != 0 || m == NULL)
885 goto done;
886 ip6 = mtod(m, struct ip6_hdr *);
887
888 /* See if destination IP address was changed by packet filter. */
889 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
890 m->m_flags |= M_SKIP_FIREWALL;
891 /* If destination is now ourself drop to ip6_input(). */
892 if (in6_localip(&ip6->ip6_dst)) {
893 m->m_flags |= M_FASTFWD_OURS;
894 if (m->m_pkthdr.rcvif == NULL)
895 m->m_pkthdr.rcvif = V_loif;
896 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
897 m->m_pkthdr.csum_flags |=
898 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
899 m->m_pkthdr.csum_data = 0xffff;
900 }
901 #ifdef SCTP
902 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
903 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
904 #endif
905 error = netisr_queue(NETISR_IPV6, m);
906 goto done;
907 } else
908 goto again; /* Redo the routing table lookup. */
909 }
910
911 /* See if local, if yes, send it to netisr. */
912 if (m->m_flags & M_FASTFWD_OURS) {
913 if (m->m_pkthdr.rcvif == NULL)
914 m->m_pkthdr.rcvif = V_loif;
915 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
916 m->m_pkthdr.csum_flags |=
917 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
918 m->m_pkthdr.csum_data = 0xffff;
919 }
920 #ifdef SCTP
921 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
922 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
923 #endif
924 error = netisr_queue(NETISR_IPV6, m);
925 goto done;
926 }
927 /* Or forward to some other address? */
928 if ((m->m_flags & M_IP6_NEXTHOP) &&
929 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
930 dst = (struct sockaddr_in6 *)&ro->ro_dst;
931 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
932 m->m_flags |= M_SKIP_FIREWALL;
933 m->m_flags &= ~M_IP6_NEXTHOP;
934 m_tag_delete(m, fwd_tag);
935 goto again;
936 }
937
938 passout:
939 /*
940 * Send the packet to the outgoing interface.
941 * If necessary, do IPv6 fragmentation before sending.
942 *
943 * the logic here is rather complex:
944 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
945 * 1-a: send as is if tlen <= path mtu
946 * 1-b: fragment if tlen > path mtu
947 *
948 * 2: if user asks us not to fragment (dontfrag == 1)
949 * 2-a: send as is if tlen <= interface mtu
950 * 2-b: error if tlen > interface mtu
951 *
952 * 3: if we always need to attach fragment header (alwaysfrag == 1)
953 * always fragment
954 *
955 * 4: if dontfrag == 1 && alwaysfrag == 1
956 * error, as we cannot handle this conflicting request
957 */
958 sw_csum = m->m_pkthdr.csum_flags;
959 if (!hdrsplit) {
960 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
961 sw_csum &= ~ifp->if_hwassist;
962 } else
963 tso = 0;
964 /*
965 * If we added extension headers, we will not do TSO and calculate the
966 * checksums ourselves for now.
967 * XXX-BZ Need a framework to know when the NIC can handle it, even
968 * with ext. hdrs.
969 */
970 if (sw_csum & CSUM_DELAY_DATA_IPV6) {
971 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
972 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
973 }
974 #ifdef SCTP
975 if (sw_csum & CSUM_SCTP_IPV6) {
976 sw_csum &= ~CSUM_SCTP_IPV6;
977 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
978 }
979 #endif
980 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
981 tlen = m->m_pkthdr.len;
982
983 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
984 dontfrag = 1;
985 else
986 dontfrag = 0;
987 if (dontfrag && alwaysfrag) { /* case 4 */
988 /* conflicting request - can't transmit */
989 error = EMSGSIZE;
990 goto bad;
991 }
992 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */
993 /*
994 * Even if the DONTFRAG option is specified, we cannot send the
995 * packet when the data length is larger than the MTU of the
996 * outgoing interface.
997 * Notify the error by sending IPV6_PATHMTU ancillary data as
998 * well as returning an error code (the latter is not described
999 * in the API spec.)
1000 */
1001 u_int32_t mtu32;
1002 struct ip6ctlparam ip6cp;
1003
1004 mtu32 = (u_int32_t)mtu;
1005 bzero(&ip6cp, sizeof(ip6cp));
1006 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1007 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1008 (void *)&ip6cp);
1009
1010 error = EMSGSIZE;
1011 goto bad;
1012 }
1013
1014 /*
1015 * transmit packet without fragmentation
1016 */
1017 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1018 struct in6_ifaddr *ia6;
1019
1020 ip6 = mtod(m, struct ip6_hdr *);
1021 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1022 if (ia6) {
1023 /* Record statistics for this interface address. */
1024 ia6->ia_ifa.if_opackets++;
1025 ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1026 ifa_free(&ia6->ia_ifa);
1027 }
1028 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1029 goto done;
1030 }
1031
1032 /*
1033 * try to fragment the packet. case 1-b and 3
1034 */
1035 if (mtu < IPV6_MMTU) {
1036 /* path MTU cannot be less than IPV6_MMTU */
1037 error = EMSGSIZE;
1038 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1039 goto bad;
1040 } else if (ip6->ip6_plen == 0) {
1041 /* jumbo payload cannot be fragmented */
1042 error = EMSGSIZE;
1043 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1044 goto bad;
1045 } else {
1046 struct mbuf **mnext, *m_frgpart;
1047 struct ip6_frag *ip6f;
1048 u_int32_t id = htonl(ip6_randomid());
1049 u_char nextproto;
1050
1051 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1052
1053 /*
1054 * Too large for the destination or interface;
1055 * fragment if possible.
1056 * Must be able to put at least 8 bytes per fragment.
1057 */
1058 hlen = unfragpartlen;
1059 if (mtu > IPV6_MAXPACKET)
1060 mtu = IPV6_MAXPACKET;
1061
1062 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1063 if (len < 8) {
1064 error = EMSGSIZE;
1065 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1066 goto bad;
1067 }
1068
1069 /*
1070 * Verify that we have any chance at all of being able to queue
1071 * the packet or packet fragments
1072 */
1073 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1074 < tlen /* - hlen */)) {
1075 error = ENOBUFS;
1076 IP6STAT_INC(ip6s_odropped);
1077 goto bad;
1078 }
1079
1080
1081 /*
1082 * If the interface will not calculate checksums on
1083 * fragmented packets, then do it here.
1084 * XXX-BZ handle the hw offloading case. Need flags.
1085 */
1086 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1087 in6_delayed_cksum(m, plen, hlen);
1088 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1089 }
1090 #ifdef SCTP
1091 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1092 sctp_delayed_cksum(m, hlen);
1093 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1094 }
1095 #endif
1096 mnext = &m->m_nextpkt;
1097
1098 /*
1099 * Change the next header field of the last header in the
1100 * unfragmentable part.
1101 */
1102 if (exthdrs.ip6e_rthdr) {
1103 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1104 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1105 } else if (exthdrs.ip6e_dest1) {
1106 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1107 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1108 } else if (exthdrs.ip6e_hbh) {
1109 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1110 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1111 } else {
1112 nextproto = ip6->ip6_nxt;
1113 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1114 }
1115
1116 /*
1117 * Loop through length of segment after first fragment,
1118 * make new header and copy data of each part and link onto
1119 * chain.
1120 */
1121 m0 = m;
1122 for (off = hlen; off < tlen; off += len) {
1123 m = m_gethdr(M_NOWAIT, MT_DATA);
1124 if (!m) {
1125 error = ENOBUFS;
1126 IP6STAT_INC(ip6s_odropped);
1127 goto sendorfree;
1128 }
1129 m->m_flags = m0->m_flags & M_COPYFLAGS;
1130 *mnext = m;
1131 mnext = &m->m_nextpkt;
1132 m->m_data += max_linkhdr;
1133 mhip6 = mtod(m, struct ip6_hdr *);
1134 *mhip6 = *ip6;
1135 m->m_len = sizeof(*mhip6);
1136 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1137 if (error) {
1138 IP6STAT_INC(ip6s_odropped);
1139 goto sendorfree;
1140 }
1141 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1142 if (off + len >= tlen)
1143 len = tlen - off;
1144 else
1145 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1146 mhip6->ip6_plen = htons((u_short)(len + hlen +
1147 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1148 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1149 error = ENOBUFS;
1150 IP6STAT_INC(ip6s_odropped);
1151 goto sendorfree;
1152 }
1153 m_cat(m, m_frgpart);
1154 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1155 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
1156 m->m_pkthdr.rcvif = NULL;
1157 ip6f->ip6f_reserved = 0;
1158 ip6f->ip6f_ident = id;
1159 ip6f->ip6f_nxt = nextproto;
1160 IP6STAT_INC(ip6s_ofragments);
1161 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1162 }
1163
1164 in6_ifstat_inc(ifp, ifs6_out_fragok);
1165 }
1166
1167 /*
1168 * Remove leading garbages.
1169 */
1170 sendorfree:
1171 m = m0->m_nextpkt;
1172 m0->m_nextpkt = 0;
1173 m_freem(m0);
1174 for (m0 = m; m; m = m0) {
1175 m0 = m->m_nextpkt;
1176 m->m_nextpkt = 0;
1177 if (error == 0) {
1178 /* Record statistics for this interface address. */
1179 if (ia) {
1180 ia->ia_ifa.if_opackets++;
1181 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1182 }
1183 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1184 } else
1185 m_freem(m);
1186 }
1187
1188 if (error == 0)
1189 IP6STAT_INC(ip6s_fragmented);
1190
1191 done:
1192 if (ro == &ip6route)
1193 RO_RTFREE(ro);
1194 if (ro_pmtu == &ip6route)
1195 RO_RTFREE(ro_pmtu);
1196 #ifdef IPSEC
1197 if (sp != NULL)
1198 KEY_FREESP(&sp);
1199 #endif
1200
1201 return (error);
1202
1203 freehdrs:
1204 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1205 m_freem(exthdrs.ip6e_dest1);
1206 m_freem(exthdrs.ip6e_rthdr);
1207 m_freem(exthdrs.ip6e_dest2);
1208 /* FALLTHROUGH */
1209 bad:
1210 if (m)
1211 m_freem(m);
1212 goto done;
1213 }
1214
1215 static int
1216 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1217 {
1218 struct mbuf *m;
1219
1220 if (hlen > MCLBYTES)
1221 return (ENOBUFS); /* XXX */
1222
1223 if (hlen > MLEN)
1224 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1225 else
1226 m = m_get(M_NOWAIT, MT_DATA);
1227 if (m == NULL)
1228 return (ENOBUFS);
1229 m->m_len = hlen;
1230 if (hdr)
1231 bcopy(hdr, mtod(m, caddr_t), hlen);
1232
1233 *mp = m;
1234 return (0);
1235 }
1236
1237 /*
1238 * Insert jumbo payload option.
1239 */
1240 static int
1241 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1242 {
1243 struct mbuf *mopt;
1244 u_char *optbuf;
1245 u_int32_t v;
1246
1247 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1248
1249 /*
1250 * If there is no hop-by-hop options header, allocate new one.
1251 * If there is one but it doesn't have enough space to store the
1252 * jumbo payload option, allocate a cluster to store the whole options.
1253 * Otherwise, use it to store the options.
1254 */
1255 if (exthdrs->ip6e_hbh == 0) {
1256 mopt = m_get(M_NOWAIT, MT_DATA);
1257 if (mopt == NULL)
1258 return (ENOBUFS);
1259 mopt->m_len = JUMBOOPTLEN;
1260 optbuf = mtod(mopt, u_char *);
1261 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1262 exthdrs->ip6e_hbh = mopt;
1263 } else {
1264 struct ip6_hbh *hbh;
1265
1266 mopt = exthdrs->ip6e_hbh;
1267 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1268 /*
1269 * XXX assumption:
1270 * - exthdrs->ip6e_hbh is not referenced from places
1271 * other than exthdrs.
1272 * - exthdrs->ip6e_hbh is not an mbuf chain.
1273 */
1274 int oldoptlen = mopt->m_len;
1275 struct mbuf *n;
1276
1277 /*
1278 * XXX: give up if the whole (new) hbh header does
1279 * not fit even in an mbuf cluster.
1280 */
1281 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1282 return (ENOBUFS);
1283
1284 /*
1285 * As a consequence, we must always prepare a cluster
1286 * at this point.
1287 */
1288 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1289 if (n == NULL)
1290 return (ENOBUFS);
1291 n->m_len = oldoptlen + JUMBOOPTLEN;
1292 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1293 oldoptlen);
1294 optbuf = mtod(n, caddr_t) + oldoptlen;
1295 m_freem(mopt);
1296 mopt = exthdrs->ip6e_hbh = n;
1297 } else {
1298 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1299 mopt->m_len += JUMBOOPTLEN;
1300 }
1301 optbuf[0] = IP6OPT_PADN;
1302 optbuf[1] = 1;
1303
1304 /*
1305 * Adjust the header length according to the pad and
1306 * the jumbo payload option.
1307 */
1308 hbh = mtod(mopt, struct ip6_hbh *);
1309 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1310 }
1311
1312 /* fill in the option. */
1313 optbuf[2] = IP6OPT_JUMBO;
1314 optbuf[3] = 4;
1315 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1316 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1317
1318 /* finally, adjust the packet header length */
1319 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1320
1321 return (0);
1322 #undef JUMBOOPTLEN
1323 }
1324
1325 /*
1326 * Insert fragment header and copy unfragmentable header portions.
1327 */
1328 static int
1329 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1330 struct ip6_frag **frghdrp)
1331 {
1332 struct mbuf *n, *mlast;
1333
1334 if (hlen > sizeof(struct ip6_hdr)) {
1335 n = m_copym(m0, sizeof(struct ip6_hdr),
1336 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1337 if (n == 0)
1338 return (ENOBUFS);
1339 m->m_next = n;
1340 } else
1341 n = m;
1342
1343 /* Search for the last mbuf of unfragmentable part. */
1344 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1345 ;
1346
1347 if ((mlast->m_flags & M_EXT) == 0 &&
1348 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1349 /* use the trailing space of the last mbuf for the fragment hdr */
1350 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1351 mlast->m_len);
1352 mlast->m_len += sizeof(struct ip6_frag);
1353 m->m_pkthdr.len += sizeof(struct ip6_frag);
1354 } else {
1355 /* allocate a new mbuf for the fragment header */
1356 struct mbuf *mfrg;
1357
1358 mfrg = m_get(M_NOWAIT, MT_DATA);
1359 if (mfrg == NULL)
1360 return (ENOBUFS);
1361 mfrg->m_len = sizeof(struct ip6_frag);
1362 *frghdrp = mtod(mfrg, struct ip6_frag *);
1363 mlast->m_next = mfrg;
1364 }
1365
1366 return (0);
1367 }
1368
1369 static int
1370 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1371 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1372 int *alwaysfragp, u_int fibnum)
1373 {
1374 u_int32_t mtu = 0;
1375 int alwaysfrag = 0;
1376 int error = 0;
1377
1378 if (ro_pmtu != ro) {
1379 /* The first hop and the final destination may differ. */
1380 struct sockaddr_in6 *sa6_dst =
1381 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1382 if (ro_pmtu->ro_rt &&
1383 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1384 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1385 RTFREE(ro_pmtu->ro_rt);
1386 ro_pmtu->ro_rt = (struct rtentry *)NULL;
1387 }
1388 if (ro_pmtu->ro_rt == NULL) {
1389 bzero(sa6_dst, sizeof(*sa6_dst));
1390 sa6_dst->sin6_family = AF_INET6;
1391 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1392 sa6_dst->sin6_addr = *dst;
1393
1394 in6_rtalloc(ro_pmtu, fibnum);
1395 }
1396 }
1397 if (ro_pmtu->ro_rt) {
1398 u_int32_t ifmtu;
1399 struct in_conninfo inc;
1400
1401 bzero(&inc, sizeof(inc));
1402 inc.inc_flags |= INC_ISIPV6;
1403 inc.inc6_faddr = *dst;
1404
1405 if (ifp == NULL)
1406 ifp = ro_pmtu->ro_rt->rt_ifp;
1407 ifmtu = IN6_LINKMTU(ifp);
1408 mtu = tcp_hc_getmtu(&inc);
1409 if (mtu)
1410 mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1411 else
1412 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1413 if (mtu == 0)
1414 mtu = ifmtu;
1415 else if (mtu < IPV6_MMTU) {
1416 /*
1417 * RFC2460 section 5, last paragraph:
1418 * if we record ICMPv6 too big message with
1419 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1420 * or smaller, with framgent header attached.
1421 * (fragment header is needed regardless from the
1422 * packet size, for translators to identify packets)
1423 */
1424 alwaysfrag = 1;
1425 mtu = IPV6_MMTU;
1426 } else if (mtu > ifmtu) {
1427 /*
1428 * The MTU on the route is larger than the MTU on
1429 * the interface! This shouldn't happen, unless the
1430 * MTU of the interface has been changed after the
1431 * interface was brought up. Change the MTU in the
1432 * route to match the interface MTU (as long as the
1433 * field isn't locked).
1434 */
1435 mtu = ifmtu;
1436 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1437 }
1438 } else if (ifp) {
1439 mtu = IN6_LINKMTU(ifp);
1440 } else
1441 error = EHOSTUNREACH; /* XXX */
1442
1443 *mtup = mtu;
1444 if (alwaysfragp)
1445 *alwaysfragp = alwaysfrag;
1446 return (error);
1447 }
1448
1449 /*
1450 * IP6 socket option processing.
1451 */
1452 int
1453 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1454 {
1455 int optdatalen, uproto;
1456 void *optdata;
1457 struct inpcb *in6p = sotoinpcb(so);
1458 int error, optval;
1459 int level, op, optname;
1460 int optlen;
1461 struct thread *td;
1462
1463 level = sopt->sopt_level;
1464 op = sopt->sopt_dir;
1465 optname = sopt->sopt_name;
1466 optlen = sopt->sopt_valsize;
1467 td = sopt->sopt_td;
1468 error = 0;
1469 optval = 0;
1470 uproto = (int)so->so_proto->pr_protocol;
1471
1472 if (level != IPPROTO_IPV6) {
1473 error = EINVAL;
1474
1475 if (sopt->sopt_level == SOL_SOCKET &&
1476 sopt->sopt_dir == SOPT_SET) {
1477 switch (sopt->sopt_name) {
1478 case SO_REUSEADDR:
1479 INP_WLOCK(in6p);
1480 if ((so->so_options & SO_REUSEADDR) != 0)
1481 in6p->inp_flags2 |= INP_REUSEADDR;
1482 else
1483 in6p->inp_flags2 &= ~INP_REUSEADDR;
1484 INP_WUNLOCK(in6p);
1485 error = 0;
1486 break;
1487 case SO_REUSEPORT:
1488 INP_WLOCK(in6p);
1489 if ((so->so_options & SO_REUSEPORT) != 0)
1490 in6p->inp_flags2 |= INP_REUSEPORT;
1491 else
1492 in6p->inp_flags2 &= ~INP_REUSEPORT;
1493 INP_WUNLOCK(in6p);
1494 error = 0;
1495 break;
1496 case SO_SETFIB:
1497 INP_WLOCK(in6p);
1498 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1499 INP_WUNLOCK(in6p);
1500 error = 0;
1501 break;
1502 default:
1503 break;
1504 }
1505 }
1506 } else { /* level == IPPROTO_IPV6 */
1507 switch (op) {
1508
1509 case SOPT_SET:
1510 switch (optname) {
1511 case IPV6_2292PKTOPTIONS:
1512 #ifdef IPV6_PKTOPTIONS
1513 case IPV6_PKTOPTIONS:
1514 #endif
1515 {
1516 struct mbuf *m;
1517
1518 error = soopt_getm(sopt, &m); /* XXX */
1519 if (error != 0)
1520 break;
1521 error = soopt_mcopyin(sopt, m); /* XXX */
1522 if (error != 0)
1523 break;
1524 error = ip6_pcbopts(&in6p->in6p_outputopts,
1525 m, so, sopt);
1526 m_freem(m); /* XXX */
1527 break;
1528 }
1529
1530 /*
1531 * Use of some Hop-by-Hop options or some
1532 * Destination options, might require special
1533 * privilege. That is, normal applications
1534 * (without special privilege) might be forbidden
1535 * from setting certain options in outgoing packets,
1536 * and might never see certain options in received
1537 * packets. [RFC 2292 Section 6]
1538 * KAME specific note:
1539 * KAME prevents non-privileged users from sending or
1540 * receiving ANY hbh/dst options in order to avoid
1541 * overhead of parsing options in the kernel.
1542 */
1543 case IPV6_RECVHOPOPTS:
1544 case IPV6_RECVDSTOPTS:
1545 case IPV6_RECVRTHDRDSTOPTS:
1546 if (td != NULL) {
1547 error = priv_check(td,
1548 PRIV_NETINET_SETHDROPTS);
1549 if (error)
1550 break;
1551 }
1552 /* FALLTHROUGH */
1553 case IPV6_UNICAST_HOPS:
1554 case IPV6_HOPLIMIT:
1555 case IPV6_FAITH:
1556
1557 case IPV6_RECVPKTINFO:
1558 case IPV6_RECVHOPLIMIT:
1559 case IPV6_RECVRTHDR:
1560 case IPV6_RECVPATHMTU:
1561 case IPV6_RECVTCLASS:
1562 case IPV6_V6ONLY:
1563 case IPV6_AUTOFLOWLABEL:
1564 case IPV6_BINDANY:
1565 if (optname == IPV6_BINDANY && td != NULL) {
1566 error = priv_check(td,
1567 PRIV_NETINET_BINDANY);
1568 if (error)
1569 break;
1570 }
1571
1572 if (optlen != sizeof(int)) {
1573 error = EINVAL;
1574 break;
1575 }
1576 error = sooptcopyin(sopt, &optval,
1577 sizeof optval, sizeof optval);
1578 if (error)
1579 break;
1580 switch (optname) {
1581
1582 case IPV6_UNICAST_HOPS:
1583 if (optval < -1 || optval >= 256)
1584 error = EINVAL;
1585 else {
1586 /* -1 = kernel default */
1587 in6p->in6p_hops = optval;
1588 if ((in6p->inp_vflag &
1589 INP_IPV4) != 0)
1590 in6p->inp_ip_ttl = optval;
1591 }
1592 break;
1593 #define OPTSET(bit) \
1594 do { \
1595 INP_WLOCK(in6p); \
1596 if (optval) \
1597 in6p->inp_flags |= (bit); \
1598 else \
1599 in6p->inp_flags &= ~(bit); \
1600 INP_WUNLOCK(in6p); \
1601 } while (/*CONSTCOND*/ 0)
1602 #define OPTSET2292(bit) \
1603 do { \
1604 INP_WLOCK(in6p); \
1605 in6p->inp_flags |= IN6P_RFC2292; \
1606 if (optval) \
1607 in6p->inp_flags |= (bit); \
1608 else \
1609 in6p->inp_flags &= ~(bit); \
1610 INP_WUNLOCK(in6p); \
1611 } while (/*CONSTCOND*/ 0)
1612 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1613
1614 case IPV6_RECVPKTINFO:
1615 /* cannot mix with RFC2292 */
1616 if (OPTBIT(IN6P_RFC2292)) {
1617 error = EINVAL;
1618 break;
1619 }
1620 OPTSET(IN6P_PKTINFO);
1621 break;
1622
1623 case IPV6_HOPLIMIT:
1624 {
1625 struct ip6_pktopts **optp;
1626
1627 /* cannot mix with RFC2292 */
1628 if (OPTBIT(IN6P_RFC2292)) {
1629 error = EINVAL;
1630 break;
1631 }
1632 optp = &in6p->in6p_outputopts;
1633 error = ip6_pcbopt(IPV6_HOPLIMIT,
1634 (u_char *)&optval, sizeof(optval),
1635 optp, (td != NULL) ? td->td_ucred :
1636 NULL, uproto);
1637 break;
1638 }
1639
1640 case IPV6_RECVHOPLIMIT:
1641 /* cannot mix with RFC2292 */
1642 if (OPTBIT(IN6P_RFC2292)) {
1643 error = EINVAL;
1644 break;
1645 }
1646 OPTSET(IN6P_HOPLIMIT);
1647 break;
1648
1649 case IPV6_RECVHOPOPTS:
1650 /* cannot mix with RFC2292 */
1651 if (OPTBIT(IN6P_RFC2292)) {
1652 error = EINVAL;
1653 break;
1654 }
1655 OPTSET(IN6P_HOPOPTS);
1656 break;
1657
1658 case IPV6_RECVDSTOPTS:
1659 /* cannot mix with RFC2292 */
1660 if (OPTBIT(IN6P_RFC2292)) {
1661 error = EINVAL;
1662 break;
1663 }
1664 OPTSET(IN6P_DSTOPTS);
1665 break;
1666
1667 case IPV6_RECVRTHDRDSTOPTS:
1668 /* cannot mix with RFC2292 */
1669 if (OPTBIT(IN6P_RFC2292)) {
1670 error = EINVAL;
1671 break;
1672 }
1673 OPTSET(IN6P_RTHDRDSTOPTS);
1674 break;
1675
1676 case IPV6_RECVRTHDR:
1677 /* cannot mix with RFC2292 */
1678 if (OPTBIT(IN6P_RFC2292)) {
1679 error = EINVAL;
1680 break;
1681 }
1682 OPTSET(IN6P_RTHDR);
1683 break;
1684
1685 case IPV6_FAITH:
1686 OPTSET(INP_FAITH);
1687 break;
1688
1689 case IPV6_RECVPATHMTU:
1690 /*
1691 * We ignore this option for TCP
1692 * sockets.
1693 * (RFC3542 leaves this case
1694 * unspecified.)
1695 */
1696 if (uproto != IPPROTO_TCP)
1697 OPTSET(IN6P_MTU);
1698 break;
1699
1700 case IPV6_V6ONLY:
1701 /*
1702 * make setsockopt(IPV6_V6ONLY)
1703 * available only prior to bind(2).
1704 * see ipng mailing list, Jun 22 2001.
1705 */
1706 if (in6p->inp_lport ||
1707 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1708 error = EINVAL;
1709 break;
1710 }
1711 OPTSET(IN6P_IPV6_V6ONLY);
1712 if (optval)
1713 in6p->inp_vflag &= ~INP_IPV4;
1714 else
1715 in6p->inp_vflag |= INP_IPV4;
1716 break;
1717 case IPV6_RECVTCLASS:
1718 /* cannot mix with RFC2292 XXX */
1719 if (OPTBIT(IN6P_RFC2292)) {
1720 error = EINVAL;
1721 break;
1722 }
1723 OPTSET(IN6P_TCLASS);
1724 break;
1725 case IPV6_AUTOFLOWLABEL:
1726 OPTSET(IN6P_AUTOFLOWLABEL);
1727 break;
1728
1729 case IPV6_BINDANY:
1730 OPTSET(INP_BINDANY);
1731 break;
1732 }
1733 break;
1734
1735 case IPV6_TCLASS:
1736 case IPV6_DONTFRAG:
1737 case IPV6_USE_MIN_MTU:
1738 case IPV6_PREFER_TEMPADDR:
1739 if (optlen != sizeof(optval)) {
1740 error = EINVAL;
1741 break;
1742 }
1743 error = sooptcopyin(sopt, &optval,
1744 sizeof optval, sizeof optval);
1745 if (error)
1746 break;
1747 {
1748 struct ip6_pktopts **optp;
1749 optp = &in6p->in6p_outputopts;
1750 error = ip6_pcbopt(optname,
1751 (u_char *)&optval, sizeof(optval),
1752 optp, (td != NULL) ? td->td_ucred :
1753 NULL, uproto);
1754 break;
1755 }
1756
1757 case IPV6_2292PKTINFO:
1758 case IPV6_2292HOPLIMIT:
1759 case IPV6_2292HOPOPTS:
1760 case IPV6_2292DSTOPTS:
1761 case IPV6_2292RTHDR:
1762 /* RFC 2292 */
1763 if (optlen != sizeof(int)) {
1764 error = EINVAL;
1765 break;
1766 }
1767 error = sooptcopyin(sopt, &optval,
1768 sizeof optval, sizeof optval);
1769 if (error)
1770 break;
1771 switch (optname) {
1772 case IPV6_2292PKTINFO:
1773 OPTSET2292(IN6P_PKTINFO);
1774 break;
1775 case IPV6_2292HOPLIMIT:
1776 OPTSET2292(IN6P_HOPLIMIT);
1777 break;
1778 case IPV6_2292HOPOPTS:
1779 /*
1780 * Check super-user privilege.
1781 * See comments for IPV6_RECVHOPOPTS.
1782 */
1783 if (td != NULL) {
1784 error = priv_check(td,
1785 PRIV_NETINET_SETHDROPTS);
1786 if (error)
1787 return (error);
1788 }
1789 OPTSET2292(IN6P_HOPOPTS);
1790 break;
1791 case IPV6_2292DSTOPTS:
1792 if (td != NULL) {
1793 error = priv_check(td,
1794 PRIV_NETINET_SETHDROPTS);
1795 if (error)
1796 return (error);
1797 }
1798 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1799 break;
1800 case IPV6_2292RTHDR:
1801 OPTSET2292(IN6P_RTHDR);
1802 break;
1803 }
1804 break;
1805 case IPV6_PKTINFO:
1806 case IPV6_HOPOPTS:
1807 case IPV6_RTHDR:
1808 case IPV6_DSTOPTS:
1809 case IPV6_RTHDRDSTOPTS:
1810 case IPV6_NEXTHOP:
1811 {
1812 /* new advanced API (RFC3542) */
1813 u_char *optbuf;
1814 u_char optbuf_storage[MCLBYTES];
1815 int optlen;
1816 struct ip6_pktopts **optp;
1817
1818 /* cannot mix with RFC2292 */
1819 if (OPTBIT(IN6P_RFC2292)) {
1820 error = EINVAL;
1821 break;
1822 }
1823
1824 /*
1825 * We only ensure valsize is not too large
1826 * here. Further validation will be done
1827 * later.
1828 */
1829 error = sooptcopyin(sopt, optbuf_storage,
1830 sizeof(optbuf_storage), 0);
1831 if (error)
1832 break;
1833 optlen = sopt->sopt_valsize;
1834 optbuf = optbuf_storage;
1835 optp = &in6p->in6p_outputopts;
1836 error = ip6_pcbopt(optname, optbuf, optlen,
1837 optp, (td != NULL) ? td->td_ucred : NULL,
1838 uproto);
1839 break;
1840 }
1841 #undef OPTSET
1842
1843 case IPV6_MULTICAST_IF:
1844 case IPV6_MULTICAST_HOPS:
1845 case IPV6_MULTICAST_LOOP:
1846 case IPV6_JOIN_GROUP:
1847 case IPV6_LEAVE_GROUP:
1848 case IPV6_MSFILTER:
1849 case MCAST_BLOCK_SOURCE:
1850 case MCAST_UNBLOCK_SOURCE:
1851 case MCAST_JOIN_GROUP:
1852 case MCAST_LEAVE_GROUP:
1853 case MCAST_JOIN_SOURCE_GROUP:
1854 case MCAST_LEAVE_SOURCE_GROUP:
1855 error = ip6_setmoptions(in6p, sopt);
1856 break;
1857
1858 case IPV6_PORTRANGE:
1859 error = sooptcopyin(sopt, &optval,
1860 sizeof optval, sizeof optval);
1861 if (error)
1862 break;
1863
1864 INP_WLOCK(in6p);
1865 switch (optval) {
1866 case IPV6_PORTRANGE_DEFAULT:
1867 in6p->inp_flags &= ~(INP_LOWPORT);
1868 in6p->inp_flags &= ~(INP_HIGHPORT);
1869 break;
1870
1871 case IPV6_PORTRANGE_HIGH:
1872 in6p->inp_flags &= ~(INP_LOWPORT);
1873 in6p->inp_flags |= INP_HIGHPORT;
1874 break;
1875
1876 case IPV6_PORTRANGE_LOW:
1877 in6p->inp_flags &= ~(INP_HIGHPORT);
1878 in6p->inp_flags |= INP_LOWPORT;
1879 break;
1880
1881 default:
1882 error = EINVAL;
1883 break;
1884 }
1885 INP_WUNLOCK(in6p);
1886 break;
1887
1888 #ifdef IPSEC
1889 case IPV6_IPSEC_POLICY:
1890 {
1891 caddr_t req;
1892 struct mbuf *m;
1893
1894 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1895 break;
1896 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1897 break;
1898 req = mtod(m, caddr_t);
1899 error = ipsec_set_policy(in6p, optname, req,
1900 m->m_len, (sopt->sopt_td != NULL) ?
1901 sopt->sopt_td->td_ucred : NULL);
1902 m_freem(m);
1903 break;
1904 }
1905 #endif /* IPSEC */
1906
1907 default:
1908 error = ENOPROTOOPT;
1909 break;
1910 }
1911 break;
1912
1913 case SOPT_GET:
1914 switch (optname) {
1915
1916 case IPV6_2292PKTOPTIONS:
1917 #ifdef IPV6_PKTOPTIONS
1918 case IPV6_PKTOPTIONS:
1919 #endif
1920 /*
1921 * RFC3542 (effectively) deprecated the
1922 * semantics of the 2292-style pktoptions.
1923 * Since it was not reliable in nature (i.e.,
1924 * applications had to expect the lack of some
1925 * information after all), it would make sense
1926 * to simplify this part by always returning
1927 * empty data.
1928 */
1929 sopt->sopt_valsize = 0;
1930 break;
1931
1932 case IPV6_RECVHOPOPTS:
1933 case IPV6_RECVDSTOPTS:
1934 case IPV6_RECVRTHDRDSTOPTS:
1935 case IPV6_UNICAST_HOPS:
1936 case IPV6_RECVPKTINFO:
1937 case IPV6_RECVHOPLIMIT:
1938 case IPV6_RECVRTHDR:
1939 case IPV6_RECVPATHMTU:
1940
1941 case IPV6_FAITH:
1942 case IPV6_V6ONLY:
1943 case IPV6_PORTRANGE:
1944 case IPV6_RECVTCLASS:
1945 case IPV6_AUTOFLOWLABEL:
1946 case IPV6_BINDANY:
1947 switch (optname) {
1948
1949 case IPV6_RECVHOPOPTS:
1950 optval = OPTBIT(IN6P_HOPOPTS);
1951 break;
1952
1953 case IPV6_RECVDSTOPTS:
1954 optval = OPTBIT(IN6P_DSTOPTS);
1955 break;
1956
1957 case IPV6_RECVRTHDRDSTOPTS:
1958 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1959 break;
1960
1961 case IPV6_UNICAST_HOPS:
1962 optval = in6p->in6p_hops;
1963 break;
1964
1965 case IPV6_RECVPKTINFO:
1966 optval = OPTBIT(IN6P_PKTINFO);
1967 break;
1968
1969 case IPV6_RECVHOPLIMIT:
1970 optval = OPTBIT(IN6P_HOPLIMIT);
1971 break;
1972
1973 case IPV6_RECVRTHDR:
1974 optval = OPTBIT(IN6P_RTHDR);
1975 break;
1976
1977 case IPV6_RECVPATHMTU:
1978 optval = OPTBIT(IN6P_MTU);
1979 break;
1980
1981 case IPV6_FAITH:
1982 optval = OPTBIT(INP_FAITH);
1983 break;
1984
1985 case IPV6_V6ONLY:
1986 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1987 break;
1988
1989 case IPV6_PORTRANGE:
1990 {
1991 int flags;
1992 flags = in6p->inp_flags;
1993 if (flags & INP_HIGHPORT)
1994 optval = IPV6_PORTRANGE_HIGH;
1995 else if (flags & INP_LOWPORT)
1996 optval = IPV6_PORTRANGE_LOW;
1997 else
1998 optval = 0;
1999 break;
2000 }
2001 case IPV6_RECVTCLASS:
2002 optval = OPTBIT(IN6P_TCLASS);
2003 break;
2004
2005 case IPV6_AUTOFLOWLABEL:
2006 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2007 break;
2008
2009 case IPV6_BINDANY:
2010 optval = OPTBIT(INP_BINDANY);
2011 break;
2012 }
2013 if (error)
2014 break;
2015 error = sooptcopyout(sopt, &optval,
2016 sizeof optval);
2017 break;
2018
2019 case IPV6_PATHMTU:
2020 {
2021 u_long pmtu = 0;
2022 struct ip6_mtuinfo mtuinfo;
2023 struct route_in6 sro;
2024
2025 bzero(&sro, sizeof(sro));
2026
2027 if (!(so->so_state & SS_ISCONNECTED))
2028 return (ENOTCONN);
2029 /*
2030 * XXX: we dot not consider the case of source
2031 * routing, or optional information to specify
2032 * the outgoing interface.
2033 */
2034 error = ip6_getpmtu(&sro, NULL, NULL,
2035 &in6p->in6p_faddr, &pmtu, NULL,
2036 so->so_fibnum);
2037 if (sro.ro_rt)
2038 RTFREE(sro.ro_rt);
2039 if (error)
2040 break;
2041 if (pmtu > IPV6_MAXPACKET)
2042 pmtu = IPV6_MAXPACKET;
2043
2044 bzero(&mtuinfo, sizeof(mtuinfo));
2045 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2046 optdata = (void *)&mtuinfo;
2047 optdatalen = sizeof(mtuinfo);
2048 error = sooptcopyout(sopt, optdata,
2049 optdatalen);
2050 break;
2051 }
2052
2053 case IPV6_2292PKTINFO:
2054 case IPV6_2292HOPLIMIT:
2055 case IPV6_2292HOPOPTS:
2056 case IPV6_2292RTHDR:
2057 case IPV6_2292DSTOPTS:
2058 switch (optname) {
2059 case IPV6_2292PKTINFO:
2060 optval = OPTBIT(IN6P_PKTINFO);
2061 break;
2062 case IPV6_2292HOPLIMIT:
2063 optval = OPTBIT(IN6P_HOPLIMIT);
2064 break;
2065 case IPV6_2292HOPOPTS:
2066 optval = OPTBIT(IN6P_HOPOPTS);
2067 break;
2068 case IPV6_2292RTHDR:
2069 optval = OPTBIT(IN6P_RTHDR);
2070 break;
2071 case IPV6_2292DSTOPTS:
2072 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2073 break;
2074 }
2075 error = sooptcopyout(sopt, &optval,
2076 sizeof optval);
2077 break;
2078 case IPV6_PKTINFO:
2079 case IPV6_HOPOPTS:
2080 case IPV6_RTHDR:
2081 case IPV6_DSTOPTS:
2082 case IPV6_RTHDRDSTOPTS:
2083 case IPV6_NEXTHOP:
2084 case IPV6_TCLASS:
2085 case IPV6_DONTFRAG:
2086 case IPV6_USE_MIN_MTU:
2087 case IPV6_PREFER_TEMPADDR:
2088 error = ip6_getpcbopt(in6p->in6p_outputopts,
2089 optname, sopt);
2090 break;
2091
2092 case IPV6_MULTICAST_IF:
2093 case IPV6_MULTICAST_HOPS:
2094 case IPV6_MULTICAST_LOOP:
2095 case IPV6_MSFILTER:
2096 error = ip6_getmoptions(in6p, sopt);
2097 break;
2098
2099 #ifdef IPSEC
2100 case IPV6_IPSEC_POLICY:
2101 {
2102 caddr_t req = NULL;
2103 size_t len = 0;
2104 struct mbuf *m = NULL;
2105 struct mbuf **mp = &m;
2106 size_t ovalsize = sopt->sopt_valsize;
2107 caddr_t oval = (caddr_t)sopt->sopt_val;
2108
2109 error = soopt_getm(sopt, &m); /* XXX */
2110 if (error != 0)
2111 break;
2112 error = soopt_mcopyin(sopt, m); /* XXX */
2113 if (error != 0)
2114 break;
2115 sopt->sopt_valsize = ovalsize;
2116 sopt->sopt_val = oval;
2117 if (m) {
2118 req = mtod(m, caddr_t);
2119 len = m->m_len;
2120 }
2121 error = ipsec_get_policy(in6p, req, len, mp);
2122 if (error == 0)
2123 error = soopt_mcopyout(sopt, m); /* XXX */
2124 if (error == 0 && m)
2125 m_freem(m);
2126 break;
2127 }
2128 #endif /* IPSEC */
2129
2130 default:
2131 error = ENOPROTOOPT;
2132 break;
2133 }
2134 break;
2135 }
2136 }
2137 return (error);
2138 }
2139
2140 int
2141 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2142 {
2143 int error = 0, optval, optlen;
2144 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2145 struct inpcb *in6p = sotoinpcb(so);
2146 int level, op, optname;
2147
2148 level = sopt->sopt_level;
2149 op = sopt->sopt_dir;
2150 optname = sopt->sopt_name;
2151 optlen = sopt->sopt_valsize;
2152
2153 if (level != IPPROTO_IPV6) {
2154 return (EINVAL);
2155 }
2156
2157 switch (optname) {
2158 case IPV6_CHECKSUM:
2159 /*
2160 * For ICMPv6 sockets, no modification allowed for checksum
2161 * offset, permit "no change" values to help existing apps.
2162 *
2163 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2164 * for an ICMPv6 socket will fail."
2165 * The current behavior does not meet RFC3542.
2166 */
2167 switch (op) {
2168 case SOPT_SET:
2169 if (optlen != sizeof(int)) {
2170 error = EINVAL;
2171 break;
2172 }
2173 error = sooptcopyin(sopt, &optval, sizeof(optval),
2174 sizeof(optval));
2175 if (error)
2176 break;
2177 if ((optval % 2) != 0) {
2178 /* the API assumes even offset values */
2179 error = EINVAL;
2180 } else if (so->so_proto->pr_protocol ==
2181 IPPROTO_ICMPV6) {
2182 if (optval != icmp6off)
2183 error = EINVAL;
2184 } else
2185 in6p->in6p_cksum = optval;
2186 break;
2187
2188 case SOPT_GET:
2189 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2190 optval = icmp6off;
2191 else
2192 optval = in6p->in6p_cksum;
2193
2194 error = sooptcopyout(sopt, &optval, sizeof(optval));
2195 break;
2196
2197 default:
2198 error = EINVAL;
2199 break;
2200 }
2201 break;
2202
2203 default:
2204 error = ENOPROTOOPT;
2205 break;
2206 }
2207
2208 return (error);
2209 }
2210
2211 /*
2212 * Set up IP6 options in pcb for insertion in output packets or
2213 * specifying behavior of outgoing packets.
2214 */
2215 static int
2216 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2217 struct socket *so, struct sockopt *sopt)
2218 {
2219 struct ip6_pktopts *opt = *pktopt;
2220 int error = 0;
2221 struct thread *td = sopt->sopt_td;
2222
2223 /* turn off any old options. */
2224 if (opt) {
2225 #ifdef DIAGNOSTIC
2226 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2227 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2228 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2229 printf("ip6_pcbopts: all specified options are cleared.\n");
2230 #endif
2231 ip6_clearpktopts(opt, -1);
2232 } else
2233 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2234 *pktopt = NULL;
2235
2236 if (!m || m->m_len == 0) {
2237 /*
2238 * Only turning off any previous options, regardless of
2239 * whether the opt is just created or given.
2240 */
2241 free(opt, M_IP6OPT);
2242 return (0);
2243 }
2244
2245 /* set options specified by user. */
2246 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2247 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2248 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2249 free(opt, M_IP6OPT);
2250 return (error);
2251 }
2252 *pktopt = opt;
2253 return (0);
2254 }
2255
2256 /*
2257 * initialize ip6_pktopts. beware that there are non-zero default values in
2258 * the struct.
2259 */
2260 void
2261 ip6_initpktopts(struct ip6_pktopts *opt)
2262 {
2263
2264 bzero(opt, sizeof(*opt));
2265 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2266 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2267 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2268 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2269 }
2270
2271 static int
2272 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2273 struct ucred *cred, int uproto)
2274 {
2275 struct ip6_pktopts *opt;
2276
2277 if (*pktopt == NULL) {
2278 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2279 M_WAITOK);
2280 ip6_initpktopts(*pktopt);
2281 }
2282 opt = *pktopt;
2283
2284 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2285 }
2286
2287 static int
2288 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2289 {
2290 void *optdata = NULL;
2291 int optdatalen = 0;
2292 struct ip6_ext *ip6e;
2293 int error = 0;
2294 struct in6_pktinfo null_pktinfo;
2295 int deftclass = 0, on;
2296 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2297 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2298
2299 switch (optname) {
2300 case IPV6_PKTINFO:
2301 if (pktopt && pktopt->ip6po_pktinfo)
2302 optdata = (void *)pktopt->ip6po_pktinfo;
2303 else {
2304 /* XXX: we don't have to do this every time... */
2305 bzero(&null_pktinfo, sizeof(null_pktinfo));
2306 optdata = (void *)&null_pktinfo;
2307 }
2308 optdatalen = sizeof(struct in6_pktinfo);
2309 break;
2310 case IPV6_TCLASS:
2311 if (pktopt && pktopt->ip6po_tclass >= 0)
2312 optdata = (void *)&pktopt->ip6po_tclass;
2313 else
2314 optdata = (void *)&deftclass;
2315 optdatalen = sizeof(int);
2316 break;
2317 case IPV6_HOPOPTS:
2318 if (pktopt && pktopt->ip6po_hbh) {
2319 optdata = (void *)pktopt->ip6po_hbh;
2320 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2321 optdatalen = (ip6e->ip6e_len + 1) << 3;
2322 }
2323 break;
2324 case IPV6_RTHDR:
2325 if (pktopt && pktopt->ip6po_rthdr) {
2326 optdata = (void *)pktopt->ip6po_rthdr;
2327 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2328 optdatalen = (ip6e->ip6e_len + 1) << 3;
2329 }
2330 break;
2331 case IPV6_RTHDRDSTOPTS:
2332 if (pktopt && pktopt->ip6po_dest1) {
2333 optdata = (void *)pktopt->ip6po_dest1;
2334 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2335 optdatalen = (ip6e->ip6e_len + 1) << 3;
2336 }
2337 break;
2338 case IPV6_DSTOPTS:
2339 if (pktopt && pktopt->ip6po_dest2) {
2340 optdata = (void *)pktopt->ip6po_dest2;
2341 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2342 optdatalen = (ip6e->ip6e_len + 1) << 3;
2343 }
2344 break;
2345 case IPV6_NEXTHOP:
2346 if (pktopt && pktopt->ip6po_nexthop) {
2347 optdata = (void *)pktopt->ip6po_nexthop;
2348 optdatalen = pktopt->ip6po_nexthop->sa_len;
2349 }
2350 break;
2351 case IPV6_USE_MIN_MTU:
2352 if (pktopt)
2353 optdata = (void *)&pktopt->ip6po_minmtu;
2354 else
2355 optdata = (void *)&defminmtu;
2356 optdatalen = sizeof(int);
2357 break;
2358 case IPV6_DONTFRAG:
2359 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2360 on = 1;
2361 else
2362 on = 0;
2363 optdata = (void *)&on;
2364 optdatalen = sizeof(on);
2365 break;
2366 case IPV6_PREFER_TEMPADDR:
2367 if (pktopt)
2368 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2369 else
2370 optdata = (void *)&defpreftemp;
2371 optdatalen = sizeof(int);
2372 break;
2373 default: /* should not happen */
2374 #ifdef DIAGNOSTIC
2375 panic("ip6_getpcbopt: unexpected option\n");
2376 #endif
2377 return (ENOPROTOOPT);
2378 }
2379
2380 error = sooptcopyout(sopt, optdata, optdatalen);
2381
2382 return (error);
2383 }
2384
2385 void
2386 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2387 {
2388 if (pktopt == NULL)
2389 return;
2390
2391 if (optname == -1 || optname == IPV6_PKTINFO) {
2392 if (pktopt->ip6po_pktinfo)
2393 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2394 pktopt->ip6po_pktinfo = NULL;
2395 }
2396 if (optname == -1 || optname == IPV6_HOPLIMIT)
2397 pktopt->ip6po_hlim = -1;
2398 if (optname == -1 || optname == IPV6_TCLASS)
2399 pktopt->ip6po_tclass = -1;
2400 if (optname == -1 || optname == IPV6_NEXTHOP) {
2401 if (pktopt->ip6po_nextroute.ro_rt) {
2402 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2403 pktopt->ip6po_nextroute.ro_rt = NULL;
2404 }
2405 if (pktopt->ip6po_nexthop)
2406 free(pktopt->ip6po_nexthop, M_IP6OPT);
2407 pktopt->ip6po_nexthop = NULL;
2408 }
2409 if (optname == -1 || optname == IPV6_HOPOPTS) {
2410 if (pktopt->ip6po_hbh)
2411 free(pktopt->ip6po_hbh, M_IP6OPT);
2412 pktopt->ip6po_hbh = NULL;
2413 }
2414 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2415 if (pktopt->ip6po_dest1)
2416 free(pktopt->ip6po_dest1, M_IP6OPT);
2417 pktopt->ip6po_dest1 = NULL;
2418 }
2419 if (optname == -1 || optname == IPV6_RTHDR) {
2420 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2421 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2422 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2423 if (pktopt->ip6po_route.ro_rt) {
2424 RTFREE(pktopt->ip6po_route.ro_rt);
2425 pktopt->ip6po_route.ro_rt = NULL;
2426 }
2427 }
2428 if (optname == -1 || optname == IPV6_DSTOPTS) {
2429 if (pktopt->ip6po_dest2)
2430 free(pktopt->ip6po_dest2, M_IP6OPT);
2431 pktopt->ip6po_dest2 = NULL;
2432 }
2433 }
2434
2435 #define PKTOPT_EXTHDRCPY(type) \
2436 do {\
2437 if (src->type) {\
2438 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2439 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2440 if (dst->type == NULL && canwait == M_NOWAIT)\
2441 goto bad;\
2442 bcopy(src->type, dst->type, hlen);\
2443 }\
2444 } while (/*CONSTCOND*/ 0)
2445
2446 static int
2447 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2448 {
2449 if (dst == NULL || src == NULL) {
2450 printf("ip6_clearpktopts: invalid argument\n");
2451 return (EINVAL);
2452 }
2453
2454 dst->ip6po_hlim = src->ip6po_hlim;
2455 dst->ip6po_tclass = src->ip6po_tclass;
2456 dst->ip6po_flags = src->ip6po_flags;
2457 dst->ip6po_minmtu = src->ip6po_minmtu;
2458 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2459 if (src->ip6po_pktinfo) {
2460 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2461 M_IP6OPT, canwait);
2462 if (dst->ip6po_pktinfo == NULL)
2463 goto bad;
2464 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2465 }
2466 if (src->ip6po_nexthop) {
2467 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2468 M_IP6OPT, canwait);
2469 if (dst->ip6po_nexthop == NULL)
2470 goto bad;
2471 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2472 src->ip6po_nexthop->sa_len);
2473 }
2474 PKTOPT_EXTHDRCPY(ip6po_hbh);
2475 PKTOPT_EXTHDRCPY(ip6po_dest1);
2476 PKTOPT_EXTHDRCPY(ip6po_dest2);
2477 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2478 return (0);
2479
2480 bad:
2481 ip6_clearpktopts(dst, -1);
2482 return (ENOBUFS);
2483 }
2484 #undef PKTOPT_EXTHDRCPY
2485
2486 struct ip6_pktopts *
2487 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2488 {
2489 int error;
2490 struct ip6_pktopts *dst;
2491
2492 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2493 if (dst == NULL)
2494 return (NULL);
2495 ip6_initpktopts(dst);
2496
2497 if ((error = copypktopts(dst, src, canwait)) != 0) {
2498 free(dst, M_IP6OPT);
2499 return (NULL);
2500 }
2501
2502 return (dst);
2503 }
2504
2505 void
2506 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2507 {
2508 if (pktopt == NULL)
2509 return;
2510
2511 ip6_clearpktopts(pktopt, -1);
2512
2513 free(pktopt, M_IP6OPT);
2514 }
2515
2516 /*
2517 * Set IPv6 outgoing packet options based on advanced API.
2518 */
2519 int
2520 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2521 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2522 {
2523 struct cmsghdr *cm = 0;
2524
2525 if (control == NULL || opt == NULL)
2526 return (EINVAL);
2527
2528 ip6_initpktopts(opt);
2529 if (stickyopt) {
2530 int error;
2531
2532 /*
2533 * If stickyopt is provided, make a local copy of the options
2534 * for this particular packet, then override them by ancillary
2535 * objects.
2536 * XXX: copypktopts() does not copy the cached route to a next
2537 * hop (if any). This is not very good in terms of efficiency,
2538 * but we can allow this since this option should be rarely
2539 * used.
2540 */
2541 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2542 return (error);
2543 }
2544
2545 /*
2546 * XXX: Currently, we assume all the optional information is stored
2547 * in a single mbuf.
2548 */
2549 if (control->m_next)
2550 return (EINVAL);
2551
2552 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2553 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2554 int error;
2555
2556 if (control->m_len < CMSG_LEN(0))
2557 return (EINVAL);
2558
2559 cm = mtod(control, struct cmsghdr *);
2560 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2561 return (EINVAL);
2562 if (cm->cmsg_level != IPPROTO_IPV6)
2563 continue;
2564
2565 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2566 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2567 if (error)
2568 return (error);
2569 }
2570
2571 return (0);
2572 }
2573
2574 /*
2575 * Set a particular packet option, as a sticky option or an ancillary data
2576 * item. "len" can be 0 only when it's a sticky option.
2577 * We have 4 cases of combination of "sticky" and "cmsg":
2578 * "sticky=0, cmsg=0": impossible
2579 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2580 * "sticky=1, cmsg=0": RFC3542 socket option
2581 * "sticky=1, cmsg=1": RFC2292 socket option
2582 */
2583 static int
2584 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2585 struct ucred *cred, int sticky, int cmsg, int uproto)
2586 {
2587 int minmtupolicy, preftemp;
2588 int error;
2589
2590 if (!sticky && !cmsg) {
2591 #ifdef DIAGNOSTIC
2592 printf("ip6_setpktopt: impossible case\n");
2593 #endif
2594 return (EINVAL);
2595 }
2596
2597 /*
2598 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2599 * not be specified in the context of RFC3542. Conversely,
2600 * RFC3542 types should not be specified in the context of RFC2292.
2601 */
2602 if (!cmsg) {
2603 switch (optname) {
2604 case IPV6_2292PKTINFO:
2605 case IPV6_2292HOPLIMIT:
2606 case IPV6_2292NEXTHOP:
2607 case IPV6_2292HOPOPTS:
2608 case IPV6_2292DSTOPTS:
2609 case IPV6_2292RTHDR:
2610 case IPV6_2292PKTOPTIONS:
2611 return (ENOPROTOOPT);
2612 }
2613 }
2614 if (sticky && cmsg) {
2615 switch (optname) {
2616 case IPV6_PKTINFO:
2617 case IPV6_HOPLIMIT:
2618 case IPV6_NEXTHOP:
2619 case IPV6_HOPOPTS:
2620 case IPV6_DSTOPTS:
2621 case IPV6_RTHDRDSTOPTS:
2622 case IPV6_RTHDR:
2623 case IPV6_USE_MIN_MTU:
2624 case IPV6_DONTFRAG:
2625 case IPV6_TCLASS:
2626 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2627 return (ENOPROTOOPT);
2628 }
2629 }
2630
2631 switch (optname) {
2632 case IPV6_2292PKTINFO:
2633 case IPV6_PKTINFO:
2634 {
2635 struct ifnet *ifp = NULL;
2636 struct in6_pktinfo *pktinfo;
2637
2638 if (len != sizeof(struct in6_pktinfo))
2639 return (EINVAL);
2640
2641 pktinfo = (struct in6_pktinfo *)buf;
2642
2643 /*
2644 * An application can clear any sticky IPV6_PKTINFO option by
2645 * doing a "regular" setsockopt with ipi6_addr being
2646 * in6addr_any and ipi6_ifindex being zero.
2647 * [RFC 3542, Section 6]
2648 */
2649 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2650 pktinfo->ipi6_ifindex == 0 &&
2651 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2652 ip6_clearpktopts(opt, optname);
2653 break;
2654 }
2655
2656 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2657 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2658 return (EINVAL);
2659 }
2660
2661 /* validate the interface index if specified. */
2662 if (pktinfo->ipi6_ifindex > V_if_index ||
2663 pktinfo->ipi6_ifindex < 0) {
2664 return (ENXIO);
2665 }
2666 if (pktinfo->ipi6_ifindex) {
2667 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2668 if (ifp == NULL)
2669 return (ENXIO);
2670 }
2671
2672 /*
2673 * We store the address anyway, and let in6_selectsrc()
2674 * validate the specified address. This is because ipi6_addr
2675 * may not have enough information about its scope zone, and
2676 * we may need additional information (such as outgoing
2677 * interface or the scope zone of a destination address) to
2678 * disambiguate the scope.
2679 * XXX: the delay of the validation may confuse the
2680 * application when it is used as a sticky option.
2681 */
2682 if (opt->ip6po_pktinfo == NULL) {
2683 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2684 M_IP6OPT, M_NOWAIT);
2685 if (opt->ip6po_pktinfo == NULL)
2686 return (ENOBUFS);
2687 }
2688 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2689 break;
2690 }
2691
2692 case IPV6_2292HOPLIMIT:
2693 case IPV6_HOPLIMIT:
2694 {
2695 int *hlimp;
2696
2697 /*
2698 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2699 * to simplify the ordering among hoplimit options.
2700 */
2701 if (optname == IPV6_HOPLIMIT && sticky)
2702 return (ENOPROTOOPT);
2703
2704 if (len != sizeof(int))
2705 return (EINVAL);
2706 hlimp = (int *)buf;
2707 if (*hlimp < -1 || *hlimp > 255)
2708 return (EINVAL);
2709
2710 opt->ip6po_hlim = *hlimp;
2711 break;
2712 }
2713
2714 case IPV6_TCLASS:
2715 {
2716 int tclass;
2717
2718 if (len != sizeof(int))
2719 return (EINVAL);
2720 tclass = *(int *)buf;
2721 if (tclass < -1 || tclass > 255)
2722 return (EINVAL);
2723
2724 opt->ip6po_tclass = tclass;
2725 break;
2726 }
2727
2728 case IPV6_2292NEXTHOP:
2729 case IPV6_NEXTHOP:
2730 if (cred != NULL) {
2731 error = priv_check_cred(cred,
2732 PRIV_NETINET_SETHDROPTS, 0);
2733 if (error)
2734 return (error);
2735 }
2736
2737 if (len == 0) { /* just remove the option */
2738 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2739 break;
2740 }
2741
2742 /* check if cmsg_len is large enough for sa_len */
2743 if (len < sizeof(struct sockaddr) || len < *buf)
2744 return (EINVAL);
2745
2746 switch (((struct sockaddr *)buf)->sa_family) {
2747 case AF_INET6:
2748 {
2749 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2750 int error;
2751
2752 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2753 return (EINVAL);
2754
2755 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2756 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2757 return (EINVAL);
2758 }
2759 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2760 != 0) {
2761 return (error);
2762 }
2763 break;
2764 }
2765 case AF_LINK: /* should eventually be supported */
2766 default:
2767 return (EAFNOSUPPORT);
2768 }
2769
2770 /* turn off the previous option, then set the new option. */
2771 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2772 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2773 if (opt->ip6po_nexthop == NULL)
2774 return (ENOBUFS);
2775 bcopy(buf, opt->ip6po_nexthop, *buf);
2776 break;
2777
2778 case IPV6_2292HOPOPTS:
2779 case IPV6_HOPOPTS:
2780 {
2781 struct ip6_hbh *hbh;
2782 int hbhlen;
2783
2784 /*
2785 * XXX: We don't allow a non-privileged user to set ANY HbH
2786 * options, since per-option restriction has too much
2787 * overhead.
2788 */
2789 if (cred != NULL) {
2790 error = priv_check_cred(cred,
2791 PRIV_NETINET_SETHDROPTS, 0);
2792 if (error)
2793 return (error);
2794 }
2795
2796 if (len == 0) {
2797 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2798 break; /* just remove the option */
2799 }
2800
2801 /* message length validation */
2802 if (len < sizeof(struct ip6_hbh))
2803 return (EINVAL);
2804 hbh = (struct ip6_hbh *)buf;
2805 hbhlen = (hbh->ip6h_len + 1) << 3;
2806 if (len != hbhlen)
2807 return (EINVAL);
2808
2809 /* turn off the previous option, then set the new option. */
2810 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2811 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2812 if (opt->ip6po_hbh == NULL)
2813 return (ENOBUFS);
2814 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2815
2816 break;
2817 }
2818
2819 case IPV6_2292DSTOPTS:
2820 case IPV6_DSTOPTS:
2821 case IPV6_RTHDRDSTOPTS:
2822 {
2823 struct ip6_dest *dest, **newdest = NULL;
2824 int destlen;
2825
2826 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2827 error = priv_check_cred(cred,
2828 PRIV_NETINET_SETHDROPTS, 0);
2829 if (error)
2830 return (error);
2831 }
2832
2833 if (len == 0) {
2834 ip6_clearpktopts(opt, optname);
2835 break; /* just remove the option */
2836 }
2837
2838 /* message length validation */
2839 if (len < sizeof(struct ip6_dest))
2840 return (EINVAL);
2841 dest = (struct ip6_dest *)buf;
2842 destlen = (dest->ip6d_len + 1) << 3;
2843 if (len != destlen)
2844 return (EINVAL);
2845
2846 /*
2847 * Determine the position that the destination options header
2848 * should be inserted; before or after the routing header.
2849 */
2850 switch (optname) {
2851 case IPV6_2292DSTOPTS:
2852 /*
2853 * The old advacned API is ambiguous on this point.
2854 * Our approach is to determine the position based
2855 * according to the existence of a routing header.
2856 * Note, however, that this depends on the order of the
2857 * extension headers in the ancillary data; the 1st
2858 * part of the destination options header must appear
2859 * before the routing header in the ancillary data,
2860 * too.
2861 * RFC3542 solved the ambiguity by introducing
2862 * separate ancillary data or option types.
2863 */
2864 if (opt->ip6po_rthdr == NULL)
2865 newdest = &opt->ip6po_dest1;
2866 else
2867 newdest = &opt->ip6po_dest2;
2868 break;
2869 case IPV6_RTHDRDSTOPTS:
2870 newdest = &opt->ip6po_dest1;
2871 break;
2872 case IPV6_DSTOPTS:
2873 newdest = &opt->ip6po_dest2;
2874 break;
2875 }
2876
2877 /* turn off the previous option, then set the new option. */
2878 ip6_clearpktopts(opt, optname);
2879 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2880 if (*newdest == NULL)
2881 return (ENOBUFS);
2882 bcopy(dest, *newdest, destlen);
2883
2884 break;
2885 }
2886
2887 case IPV6_2292RTHDR:
2888 case IPV6_RTHDR:
2889 {
2890 struct ip6_rthdr *rth;
2891 int rthlen;
2892
2893 if (len == 0) {
2894 ip6_clearpktopts(opt, IPV6_RTHDR);
2895 break; /* just remove the option */
2896 }
2897
2898 /* message length validation */
2899 if (len < sizeof(struct ip6_rthdr))
2900 return (EINVAL);
2901 rth = (struct ip6_rthdr *)buf;
2902 rthlen = (rth->ip6r_len + 1) << 3;
2903 if (len != rthlen)
2904 return (EINVAL);
2905
2906 switch (rth->ip6r_type) {
2907 case IPV6_RTHDR_TYPE_0:
2908 if (rth->ip6r_len == 0) /* must contain one addr */
2909 return (EINVAL);
2910 if (rth->ip6r_len % 2) /* length must be even */
2911 return (EINVAL);
2912 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2913 return (EINVAL);
2914 break;
2915 default:
2916 return (EINVAL); /* not supported */
2917 }
2918
2919 /* turn off the previous option */
2920 ip6_clearpktopts(opt, IPV6_RTHDR);
2921 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2922 if (opt->ip6po_rthdr == NULL)
2923 return (ENOBUFS);
2924 bcopy(rth, opt->ip6po_rthdr, rthlen);
2925
2926 break;
2927 }
2928
2929 case IPV6_USE_MIN_MTU:
2930 if (len != sizeof(int))
2931 return (EINVAL);
2932 minmtupolicy = *(int *)buf;
2933 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2934 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2935 minmtupolicy != IP6PO_MINMTU_ALL) {
2936 return (EINVAL);
2937 }
2938 opt->ip6po_minmtu = minmtupolicy;
2939 break;
2940
2941 case IPV6_DONTFRAG:
2942 if (len != sizeof(int))
2943 return (EINVAL);
2944
2945 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2946 /*
2947 * we ignore this option for TCP sockets.
2948 * (RFC3542 leaves this case unspecified.)
2949 */
2950 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2951 } else
2952 opt->ip6po_flags |= IP6PO_DONTFRAG;
2953 break;
2954
2955 case IPV6_PREFER_TEMPADDR:
2956 if (len != sizeof(int))
2957 return (EINVAL);
2958 preftemp = *(int *)buf;
2959 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2960 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2961 preftemp != IP6PO_TEMPADDR_PREFER) {
2962 return (EINVAL);
2963 }
2964 opt->ip6po_prefer_tempaddr = preftemp;
2965 break;
2966
2967 default:
2968 return (ENOPROTOOPT);
2969 } /* end of switch */
2970
2971 return (0);
2972 }
2973
2974 /*
2975 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2976 * packet to the input queue of a specified interface. Note that this
2977 * calls the output routine of the loopback "driver", but with an interface
2978 * pointer that might NOT be &loif -- easier than replicating that code here.
2979 */
2980 void
2981 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2982 {
2983 struct mbuf *copym;
2984 struct ip6_hdr *ip6;
2985
2986 copym = m_copy(m, 0, M_COPYALL);
2987 if (copym == NULL)
2988 return;
2989
2990 /*
2991 * Make sure to deep-copy IPv6 header portion in case the data
2992 * is in an mbuf cluster, so that we can safely override the IPv6
2993 * header portion later.
2994 */
2995 if ((copym->m_flags & M_EXT) != 0 ||
2996 copym->m_len < sizeof(struct ip6_hdr)) {
2997 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2998 if (copym == NULL)
2999 return;
3000 }
3001
3002 #ifdef DIAGNOSTIC
3003 if (copym->m_len < sizeof(*ip6)) {
3004 m_freem(copym);
3005 return;
3006 }
3007 #endif
3008
3009 ip6 = mtod(copym, struct ip6_hdr *);
3010 /*
3011 * clear embedded scope identifiers if necessary.
3012 * in6_clearscope will touch the addresses only when necessary.
3013 */
3014 in6_clearscope(&ip6->ip6_src);
3015 in6_clearscope(&ip6->ip6_dst);
3016
3017 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3018 }
3019
3020 /*
3021 * Chop IPv6 header off from the payload.
3022 */
3023 static int
3024 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3025 {
3026 struct mbuf *mh;
3027 struct ip6_hdr *ip6;
3028
3029 ip6 = mtod(m, struct ip6_hdr *);
3030 if (m->m_len > sizeof(*ip6)) {
3031 mh = m_gethdr(M_NOWAIT, MT_DATA);
3032 if (mh == NULL) {
3033 m_freem(m);
3034 return ENOBUFS;
3035 }
3036 m_move_pkthdr(mh, m);
3037 MH_ALIGN(mh, sizeof(*ip6));
3038 m->m_len -= sizeof(*ip6);
3039 m->m_data += sizeof(*ip6);
3040 mh->m_next = m;
3041 m = mh;
3042 m->m_len = sizeof(*ip6);
3043 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3044 }
3045 exthdrs->ip6e_ip6 = m;
3046 return 0;
3047 }
3048
3049 /*
3050 * Compute IPv6 extension header length.
3051 */
3052 int
3053 ip6_optlen(struct inpcb *in6p)
3054 {
3055 int len;
3056
3057 if (!in6p->in6p_outputopts)
3058 return 0;
3059
3060 len = 0;
3061 #define elen(x) \
3062 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3063
3064 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3065 if (in6p->in6p_outputopts->ip6po_rthdr)
3066 /* dest1 is valid with rthdr only */
3067 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3068 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3069 len += elen(in6p->in6p_outputopts->ip6po_dest2);
3070 return len;
3071 #undef elen
3072 }
Cache object: b765bab1adc6733c3c55bb34c1c9e20f
|