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("$FreeBSD$");
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 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=0x%04x\n", __func__, m->m_len, plen, offset,
200 m->m_pkthdr.csum_flags);
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 is not always the case, since
778 * some versions of MGETHDR() does not
779 * initialize the field.
780 */
781 m->m_pkthdr.rcvif = NULL;
782 if (ip6_mforward(ip6, ifp, m) != 0) {
783 m_freem(m);
784 goto done;
785 }
786 }
787 }
788 /*
789 * Multicasts with a hoplimit of zero may be looped back,
790 * above, but must not be transmitted on a network.
791 * Also, multicasts addressed to the loopback interface
792 * are not sent -- the above call to ip6_mloopback() will
793 * loop back a copy if this host actually belongs to the
794 * destination group on the loopback interface.
795 */
796 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
797 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
798 m_freem(m);
799 goto done;
800 }
801 }
802
803 /*
804 * Fill the outgoing inteface to tell the upper layer
805 * to increment per-interface statistics.
806 */
807 if (ifpp)
808 *ifpp = ifp;
809
810 /* Determine path MTU. */
811 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
812 &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
813 goto bad;
814
815 /*
816 * The caller of this function may specify to use the minimum MTU
817 * in some cases.
818 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
819 * setting. The logic is a bit complicated; by default, unicast
820 * packets will follow path MTU while multicast packets will be sent at
821 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
822 * including unicast ones will be sent at the minimum MTU. Multicast
823 * packets will always be sent at the minimum MTU unless
824 * IP6PO_MINMTU_DISABLE is explicitly specified.
825 * See RFC 3542 for more details.
826 */
827 if (mtu > IPV6_MMTU) {
828 if ((flags & IPV6_MINMTU))
829 mtu = IPV6_MMTU;
830 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
831 mtu = IPV6_MMTU;
832 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
833 (opt == NULL ||
834 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
835 mtu = IPV6_MMTU;
836 }
837 }
838
839 /*
840 * clear embedded scope identifiers if necessary.
841 * in6_clearscope will touch the addresses only when necessary.
842 */
843 in6_clearscope(&ip6->ip6_src);
844 in6_clearscope(&ip6->ip6_dst);
845
846 /*
847 * If the outgoing packet contains a hop-by-hop options header,
848 * it must be examined and processed even by the source node.
849 * (RFC 2460, section 4.)
850 */
851 if (exthdrs.ip6e_hbh) {
852 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
853 u_int32_t dummy; /* XXX unused */
854 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
855
856 #ifdef DIAGNOSTIC
857 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
858 panic("ip6e_hbh is not contiguous");
859 #endif
860 /*
861 * XXX: if we have to send an ICMPv6 error to the sender,
862 * we need the M_LOOP flag since icmp6_error() expects
863 * the IPv6 and the hop-by-hop options header are
864 * contiguous unless the flag is set.
865 */
866 m->m_flags |= M_LOOP;
867 m->m_pkthdr.rcvif = ifp;
868 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
869 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
870 &dummy, &plen) < 0) {
871 /* m was already freed at this point */
872 error = EINVAL;/* better error? */
873 goto done;
874 }
875 m->m_flags &= ~M_LOOP; /* XXX */
876 m->m_pkthdr.rcvif = NULL;
877 }
878
879 /* Jump over all PFIL processing if hooks are not active. */
880 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
881 goto passout;
882
883 odst = ip6->ip6_dst;
884 /* Run through list of hooks for output packets. */
885 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
886 if (error != 0 || m == NULL)
887 goto done;
888 ip6 = mtod(m, struct ip6_hdr *);
889
890 /* See if destination IP address was changed by packet filter. */
891 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
892 m->m_flags |= M_SKIP_FIREWALL;
893 /* If destination is now ourself drop to ip6_input(). */
894 if (in6_localip(&ip6->ip6_dst)) {
895 m->m_flags |= M_FASTFWD_OURS;
896 if (m->m_pkthdr.rcvif == NULL)
897 m->m_pkthdr.rcvif = V_loif;
898 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
899 m->m_pkthdr.csum_flags |=
900 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
901 m->m_pkthdr.csum_data = 0xffff;
902 }
903 #ifdef SCTP
904 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
905 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
906 #endif
907 error = netisr_queue(NETISR_IPV6, m);
908 goto done;
909 } else
910 goto again; /* Redo the routing table lookup. */
911 }
912
913 /* See if local, if yes, send it to netisr. */
914 if (m->m_flags & M_FASTFWD_OURS) {
915 if (m->m_pkthdr.rcvif == NULL)
916 m->m_pkthdr.rcvif = V_loif;
917 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
918 m->m_pkthdr.csum_flags |=
919 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
920 m->m_pkthdr.csum_data = 0xffff;
921 }
922 #ifdef SCTP
923 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
924 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
925 #endif
926 error = netisr_queue(NETISR_IPV6, m);
927 goto done;
928 }
929 /* Or forward to some other address? */
930 if ((m->m_flags & M_IP6_NEXTHOP) &&
931 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
932 dst = (struct sockaddr_in6 *)&ro->ro_dst;
933 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
934 m->m_flags |= M_SKIP_FIREWALL;
935 m->m_flags &= ~M_IP6_NEXTHOP;
936 m_tag_delete(m, fwd_tag);
937 goto again;
938 }
939
940 passout:
941 /*
942 * Send the packet to the outgoing interface.
943 * If necessary, do IPv6 fragmentation before sending.
944 *
945 * the logic here is rather complex:
946 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
947 * 1-a: send as is if tlen <= path mtu
948 * 1-b: fragment if tlen > path mtu
949 *
950 * 2: if user asks us not to fragment (dontfrag == 1)
951 * 2-a: send as is if tlen <= interface mtu
952 * 2-b: error if tlen > interface mtu
953 *
954 * 3: if we always need to attach fragment header (alwaysfrag == 1)
955 * always fragment
956 *
957 * 4: if dontfrag == 1 && alwaysfrag == 1
958 * error, as we cannot handle this conflicting request
959 */
960 sw_csum = m->m_pkthdr.csum_flags;
961 if (!hdrsplit) {
962 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
963 sw_csum &= ~ifp->if_hwassist;
964 } else
965 tso = 0;
966 /*
967 * If we added extension headers, we will not do TSO and calculate the
968 * checksums ourselves for now.
969 * XXX-BZ Need a framework to know when the NIC can handle it, even
970 * with ext. hdrs.
971 */
972 if (sw_csum & CSUM_DELAY_DATA_IPV6) {
973 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
974 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
975 }
976 #ifdef SCTP
977 if (sw_csum & CSUM_SCTP_IPV6) {
978 sw_csum &= ~CSUM_SCTP_IPV6;
979 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
980 }
981 #endif
982 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
983 tlen = m->m_pkthdr.len;
984
985 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
986 dontfrag = 1;
987 else
988 dontfrag = 0;
989 if (dontfrag && alwaysfrag) { /* case 4 */
990 /* conflicting request - can't transmit */
991 error = EMSGSIZE;
992 goto bad;
993 }
994 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */
995 /*
996 * Even if the DONTFRAG option is specified, we cannot send the
997 * packet when the data length is larger than the MTU of the
998 * outgoing interface.
999 * Notify the error by sending IPV6_PATHMTU ancillary data if
1000 * application wanted to know the MTU value. Also return an
1001 * error code (this is not described in the API spec).
1002 */
1003 if (inp != NULL)
1004 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1005 error = EMSGSIZE;
1006 goto bad;
1007 }
1008
1009 /*
1010 * transmit packet without fragmentation
1011 */
1012 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1013 struct in6_ifaddr *ia6;
1014
1015 ip6 = mtod(m, struct ip6_hdr *);
1016 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1017 if (ia6) {
1018 /* Record statistics for this interface address. */
1019 ia6->ia_ifa.if_opackets++;
1020 ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1021 ifa_free(&ia6->ia_ifa);
1022 }
1023 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1024 goto done;
1025 }
1026
1027 /*
1028 * try to fragment the packet. case 1-b and 3
1029 */
1030 if (mtu < IPV6_MMTU) {
1031 /* path MTU cannot be less than IPV6_MMTU */
1032 error = EMSGSIZE;
1033 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1034 goto bad;
1035 } else if (ip6->ip6_plen == 0) {
1036 /* jumbo payload cannot be fragmented */
1037 error = EMSGSIZE;
1038 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1039 goto bad;
1040 } else {
1041 struct mbuf **mnext, *m_frgpart;
1042 struct ip6_frag *ip6f;
1043 u_int32_t id = htonl(ip6_randomid());
1044 u_char nextproto;
1045
1046 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1047
1048 /*
1049 * Too large for the destination or interface;
1050 * fragment if possible.
1051 * Must be able to put at least 8 bytes per fragment.
1052 */
1053 hlen = unfragpartlen;
1054 if (mtu > IPV6_MAXPACKET)
1055 mtu = IPV6_MAXPACKET;
1056
1057 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1058 if (len < 8) {
1059 error = EMSGSIZE;
1060 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1061 goto bad;
1062 }
1063
1064 /*
1065 * Verify that we have any chance at all of being able to queue
1066 * the packet or packet fragments
1067 */
1068 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1069 < tlen /* - hlen */)) {
1070 error = ENOBUFS;
1071 IP6STAT_INC(ip6s_odropped);
1072 goto bad;
1073 }
1074
1075
1076 /*
1077 * If the interface will not calculate checksums on
1078 * fragmented packets, then do it here.
1079 * XXX-BZ handle the hw offloading case. Need flags.
1080 */
1081 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1082 in6_delayed_cksum(m, plen, hlen);
1083 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1084 }
1085 #ifdef SCTP
1086 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1087 sctp_delayed_cksum(m, hlen);
1088 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1089 }
1090 #endif
1091 mnext = &m->m_nextpkt;
1092
1093 /*
1094 * Change the next header field of the last header in the
1095 * unfragmentable part.
1096 */
1097 if (exthdrs.ip6e_rthdr) {
1098 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1099 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1100 } else if (exthdrs.ip6e_dest1) {
1101 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1102 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1103 } else if (exthdrs.ip6e_hbh) {
1104 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1105 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1106 } else {
1107 nextproto = ip6->ip6_nxt;
1108 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1109 }
1110
1111 /*
1112 * Loop through length of segment after first fragment,
1113 * make new header and copy data of each part and link onto
1114 * chain.
1115 */
1116 m0 = m;
1117 for (off = hlen; off < tlen; off += len) {
1118 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1119 if (!m) {
1120 error = ENOBUFS;
1121 IP6STAT_INC(ip6s_odropped);
1122 goto sendorfree;
1123 }
1124 m->m_pkthdr.rcvif = NULL;
1125 m->m_flags = m0->m_flags & M_COPYFLAGS; /* incl. FIB */
1126 *mnext = m;
1127 mnext = &m->m_nextpkt;
1128 m->m_data += max_linkhdr;
1129 mhip6 = mtod(m, struct ip6_hdr *);
1130 *mhip6 = *ip6;
1131 m->m_len = sizeof(*mhip6);
1132 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1133 if (error) {
1134 IP6STAT_INC(ip6s_odropped);
1135 goto sendorfree;
1136 }
1137 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1138 if (off + len >= tlen)
1139 len = tlen - off;
1140 else
1141 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1142 mhip6->ip6_plen = htons((u_short)(len + hlen +
1143 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1144 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1145 error = ENOBUFS;
1146 IP6STAT_INC(ip6s_odropped);
1147 goto sendorfree;
1148 }
1149 m_cat(m, m_frgpart);
1150 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1151 m->m_pkthdr.rcvif = NULL;
1152 ip6f->ip6f_reserved = 0;
1153 ip6f->ip6f_ident = id;
1154 ip6f->ip6f_nxt = nextproto;
1155 IP6STAT_INC(ip6s_ofragments);
1156 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1157 }
1158
1159 in6_ifstat_inc(ifp, ifs6_out_fragok);
1160 }
1161
1162 /*
1163 * Remove leading garbages.
1164 */
1165 sendorfree:
1166 m = m0->m_nextpkt;
1167 m0->m_nextpkt = 0;
1168 m_freem(m0);
1169 for (m0 = m; m; m = m0) {
1170 m0 = m->m_nextpkt;
1171 m->m_nextpkt = 0;
1172 if (error == 0) {
1173 /* Record statistics for this interface address. */
1174 if (ia) {
1175 ia->ia_ifa.if_opackets++;
1176 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1177 }
1178 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1179 } else
1180 m_freem(m);
1181 }
1182
1183 if (error == 0)
1184 IP6STAT_INC(ip6s_fragmented);
1185
1186 done:
1187 if (ro == &ip6route)
1188 RO_RTFREE(ro);
1189 if (ro_pmtu == &ip6route)
1190 RO_RTFREE(ro_pmtu);
1191 #ifdef IPSEC
1192 if (sp != NULL)
1193 KEY_FREESP(&sp);
1194 #endif
1195
1196 return (error);
1197
1198 freehdrs:
1199 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1200 m_freem(exthdrs.ip6e_dest1);
1201 m_freem(exthdrs.ip6e_rthdr);
1202 m_freem(exthdrs.ip6e_dest2);
1203 /* FALLTHROUGH */
1204 bad:
1205 if (m)
1206 m_freem(m);
1207 goto done;
1208 }
1209
1210 static int
1211 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1212 {
1213 struct mbuf *m;
1214
1215 if (hlen > MCLBYTES)
1216 return (ENOBUFS); /* XXX */
1217
1218 MGET(m, M_DONTWAIT, MT_DATA);
1219 if (!m)
1220 return (ENOBUFS);
1221
1222 if (hlen > MLEN) {
1223 MCLGET(m, M_DONTWAIT);
1224 if ((m->m_flags & M_EXT) == 0) {
1225 m_free(m);
1226 return (ENOBUFS);
1227 }
1228 }
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 MGET(mopt, M_DONTWAIT, MT_DATA);
1257 if (mopt == 0)
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 MGET(n, M_DONTWAIT, MT_DATA);
1289 if (n) {
1290 MCLGET(n, M_DONTWAIT);
1291 if ((n->m_flags & M_EXT) == 0) {
1292 m_freem(n);
1293 n = NULL;
1294 }
1295 }
1296 if (!n)
1297 return (ENOBUFS);
1298 n->m_len = oldoptlen + JUMBOOPTLEN;
1299 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1300 oldoptlen);
1301 optbuf = mtod(n, caddr_t) + oldoptlen;
1302 m_freem(mopt);
1303 mopt = exthdrs->ip6e_hbh = n;
1304 } else {
1305 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1306 mopt->m_len += JUMBOOPTLEN;
1307 }
1308 optbuf[0] = IP6OPT_PADN;
1309 optbuf[1] = 1;
1310
1311 /*
1312 * Adjust the header length according to the pad and
1313 * the jumbo payload option.
1314 */
1315 hbh = mtod(mopt, struct ip6_hbh *);
1316 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1317 }
1318
1319 /* fill in the option. */
1320 optbuf[2] = IP6OPT_JUMBO;
1321 optbuf[3] = 4;
1322 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1323 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1324
1325 /* finally, adjust the packet header length */
1326 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1327
1328 return (0);
1329 #undef JUMBOOPTLEN
1330 }
1331
1332 /*
1333 * Insert fragment header and copy unfragmentable header portions.
1334 */
1335 static int
1336 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1337 struct ip6_frag **frghdrp)
1338 {
1339 struct mbuf *n, *mlast;
1340
1341 if (hlen > sizeof(struct ip6_hdr)) {
1342 n = m_copym(m0, sizeof(struct ip6_hdr),
1343 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1344 if (n == 0)
1345 return (ENOBUFS);
1346 m->m_next = n;
1347 } else
1348 n = m;
1349
1350 /* Search for the last mbuf of unfragmentable part. */
1351 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1352 ;
1353
1354 if ((mlast->m_flags & M_EXT) == 0 &&
1355 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1356 /* use the trailing space of the last mbuf for the fragment hdr */
1357 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1358 mlast->m_len);
1359 mlast->m_len += sizeof(struct ip6_frag);
1360 m->m_pkthdr.len += sizeof(struct ip6_frag);
1361 } else {
1362 /* allocate a new mbuf for the fragment header */
1363 struct mbuf *mfrg;
1364
1365 MGET(mfrg, M_DONTWAIT, MT_DATA);
1366 if (mfrg == 0)
1367 return (ENOBUFS);
1368 mfrg->m_len = sizeof(struct ip6_frag);
1369 *frghdrp = mtod(mfrg, struct ip6_frag *);
1370 mlast->m_next = mfrg;
1371 }
1372
1373 return (0);
1374 }
1375
1376 static int
1377 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1378 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1379 int *alwaysfragp, u_int fibnum)
1380 {
1381 u_int32_t mtu = 0;
1382 int alwaysfrag = 0;
1383 int error = 0;
1384
1385 if (ro_pmtu != ro) {
1386 /* The first hop and the final destination may differ. */
1387 struct sockaddr_in6 *sa6_dst =
1388 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1389 if (ro_pmtu->ro_rt &&
1390 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1391 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1392 RTFREE(ro_pmtu->ro_rt);
1393 ro_pmtu->ro_rt = (struct rtentry *)NULL;
1394 }
1395 if (ro_pmtu->ro_rt == NULL) {
1396 bzero(sa6_dst, sizeof(*sa6_dst));
1397 sa6_dst->sin6_family = AF_INET6;
1398 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1399 sa6_dst->sin6_addr = *dst;
1400
1401 in6_rtalloc(ro_pmtu, fibnum);
1402 }
1403 }
1404 if (ro_pmtu->ro_rt) {
1405 u_int32_t ifmtu;
1406 struct in_conninfo inc;
1407
1408 bzero(&inc, sizeof(inc));
1409 inc.inc_flags |= INC_ISIPV6;
1410 inc.inc6_faddr = *dst;
1411
1412 if (ifp == NULL)
1413 ifp = ro_pmtu->ro_rt->rt_ifp;
1414 ifmtu = IN6_LINKMTU(ifp);
1415 mtu = tcp_hc_getmtu(&inc);
1416 if (mtu)
1417 mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1418 else
1419 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1420 if (mtu == 0)
1421 mtu = ifmtu;
1422 else if (mtu < IPV6_MMTU) {
1423 /*
1424 * RFC2460 section 5, last paragraph:
1425 * if we record ICMPv6 too big message with
1426 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1427 * or smaller, with framgent header attached.
1428 * (fragment header is needed regardless from the
1429 * packet size, for translators to identify packets)
1430 */
1431 alwaysfrag = 1;
1432 mtu = IPV6_MMTU;
1433 } else if (mtu > ifmtu) {
1434 /*
1435 * The MTU on the route is larger than the MTU on
1436 * the interface! This shouldn't happen, unless the
1437 * MTU of the interface has been changed after the
1438 * interface was brought up. Change the MTU in the
1439 * route to match the interface MTU (as long as the
1440 * field isn't locked).
1441 */
1442 mtu = ifmtu;
1443 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1444 }
1445 } else if (ifp) {
1446 mtu = IN6_LINKMTU(ifp);
1447 } else
1448 error = EHOSTUNREACH; /* XXX */
1449
1450 *mtup = mtu;
1451 if (alwaysfragp)
1452 *alwaysfragp = alwaysfrag;
1453 return (error);
1454 }
1455
1456 /*
1457 * IP6 socket option processing.
1458 */
1459 int
1460 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1461 {
1462 int optdatalen, uproto;
1463 void *optdata;
1464 struct inpcb *in6p = sotoinpcb(so);
1465 int error, optval;
1466 int level, op, optname;
1467 int optlen;
1468 struct thread *td;
1469
1470 level = sopt->sopt_level;
1471 op = sopt->sopt_dir;
1472 optname = sopt->sopt_name;
1473 optlen = sopt->sopt_valsize;
1474 td = sopt->sopt_td;
1475 error = 0;
1476 optval = 0;
1477 uproto = (int)so->so_proto->pr_protocol;
1478
1479 if (level != IPPROTO_IPV6) {
1480 error = EINVAL;
1481
1482 if (sopt->sopt_level == SOL_SOCKET &&
1483 sopt->sopt_dir == SOPT_SET) {
1484 switch (sopt->sopt_name) {
1485 case SO_REUSEADDR:
1486 INP_WLOCK(in6p);
1487 if ((so->so_options & SO_REUSEADDR) != 0)
1488 in6p->inp_flags2 |= INP_REUSEADDR;
1489 else
1490 in6p->inp_flags2 &= ~INP_REUSEADDR;
1491 INP_WUNLOCK(in6p);
1492 error = 0;
1493 break;
1494 case SO_REUSEPORT:
1495 INP_WLOCK(in6p);
1496 if ((so->so_options & SO_REUSEPORT) != 0)
1497 in6p->inp_flags2 |= INP_REUSEPORT;
1498 else
1499 in6p->inp_flags2 &= ~INP_REUSEPORT;
1500 INP_WUNLOCK(in6p);
1501 error = 0;
1502 break;
1503 case SO_SETFIB:
1504 INP_WLOCK(in6p);
1505 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1506 INP_WUNLOCK(in6p);
1507 error = 0;
1508 break;
1509 default:
1510 break;
1511 }
1512 }
1513 } else { /* level == IPPROTO_IPV6 */
1514 switch (op) {
1515
1516 case SOPT_SET:
1517 switch (optname) {
1518 case IPV6_2292PKTOPTIONS:
1519 #ifdef IPV6_PKTOPTIONS
1520 case IPV6_PKTOPTIONS:
1521 #endif
1522 {
1523 struct mbuf *m;
1524
1525 error = soopt_getm(sopt, &m); /* XXX */
1526 if (error != 0)
1527 break;
1528 error = soopt_mcopyin(sopt, m); /* XXX */
1529 if (error != 0)
1530 break;
1531 error = ip6_pcbopts(&in6p->in6p_outputopts,
1532 m, so, sopt);
1533 m_freem(m); /* XXX */
1534 break;
1535 }
1536
1537 /*
1538 * Use of some Hop-by-Hop options or some
1539 * Destination options, might require special
1540 * privilege. That is, normal applications
1541 * (without special privilege) might be forbidden
1542 * from setting certain options in outgoing packets,
1543 * and might never see certain options in received
1544 * packets. [RFC 2292 Section 6]
1545 * KAME specific note:
1546 * KAME prevents non-privileged users from sending or
1547 * receiving ANY hbh/dst options in order to avoid
1548 * overhead of parsing options in the kernel.
1549 */
1550 case IPV6_RECVHOPOPTS:
1551 case IPV6_RECVDSTOPTS:
1552 case IPV6_RECVRTHDRDSTOPTS:
1553 if (td != NULL) {
1554 error = priv_check(td,
1555 PRIV_NETINET_SETHDROPTS);
1556 if (error)
1557 break;
1558 }
1559 /* FALLTHROUGH */
1560 case IPV6_UNICAST_HOPS:
1561 case IPV6_HOPLIMIT:
1562 case IPV6_FAITH:
1563
1564 case IPV6_RECVPKTINFO:
1565 case IPV6_RECVHOPLIMIT:
1566 case IPV6_RECVRTHDR:
1567 case IPV6_RECVPATHMTU:
1568 case IPV6_RECVTCLASS:
1569 case IPV6_V6ONLY:
1570 case IPV6_AUTOFLOWLABEL:
1571 case IPV6_BINDANY:
1572 if (optname == IPV6_BINDANY && td != NULL) {
1573 error = priv_check(td,
1574 PRIV_NETINET_BINDANY);
1575 if (error)
1576 break;
1577 }
1578
1579 if (optlen != sizeof(int)) {
1580 error = EINVAL;
1581 break;
1582 }
1583 error = sooptcopyin(sopt, &optval,
1584 sizeof optval, sizeof optval);
1585 if (error)
1586 break;
1587 switch (optname) {
1588
1589 case IPV6_UNICAST_HOPS:
1590 if (optval < -1 || optval >= 256)
1591 error = EINVAL;
1592 else {
1593 /* -1 = kernel default */
1594 in6p->in6p_hops = optval;
1595 if ((in6p->inp_vflag &
1596 INP_IPV4) != 0)
1597 in6p->inp_ip_ttl = optval;
1598 }
1599 break;
1600 #define OPTSET(bit) \
1601 do { \
1602 INP_WLOCK(in6p); \
1603 if (optval) \
1604 in6p->inp_flags |= (bit); \
1605 else \
1606 in6p->inp_flags &= ~(bit); \
1607 INP_WUNLOCK(in6p); \
1608 } while (/*CONSTCOND*/ 0)
1609 #define OPTSET2292(bit) \
1610 do { \
1611 INP_WLOCK(in6p); \
1612 in6p->inp_flags |= IN6P_RFC2292; \
1613 if (optval) \
1614 in6p->inp_flags |= (bit); \
1615 else \
1616 in6p->inp_flags &= ~(bit); \
1617 INP_WUNLOCK(in6p); \
1618 } while (/*CONSTCOND*/ 0)
1619 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1620
1621 case IPV6_RECVPKTINFO:
1622 /* cannot mix with RFC2292 */
1623 if (OPTBIT(IN6P_RFC2292)) {
1624 error = EINVAL;
1625 break;
1626 }
1627 OPTSET(IN6P_PKTINFO);
1628 break;
1629
1630 case IPV6_HOPLIMIT:
1631 {
1632 struct ip6_pktopts **optp;
1633
1634 /* cannot mix with RFC2292 */
1635 if (OPTBIT(IN6P_RFC2292)) {
1636 error = EINVAL;
1637 break;
1638 }
1639 optp = &in6p->in6p_outputopts;
1640 error = ip6_pcbopt(IPV6_HOPLIMIT,
1641 (u_char *)&optval, sizeof(optval),
1642 optp, (td != NULL) ? td->td_ucred :
1643 NULL, uproto);
1644 break;
1645 }
1646
1647 case IPV6_RECVHOPLIMIT:
1648 /* cannot mix with RFC2292 */
1649 if (OPTBIT(IN6P_RFC2292)) {
1650 error = EINVAL;
1651 break;
1652 }
1653 OPTSET(IN6P_HOPLIMIT);
1654 break;
1655
1656 case IPV6_RECVHOPOPTS:
1657 /* cannot mix with RFC2292 */
1658 if (OPTBIT(IN6P_RFC2292)) {
1659 error = EINVAL;
1660 break;
1661 }
1662 OPTSET(IN6P_HOPOPTS);
1663 break;
1664
1665 case IPV6_RECVDSTOPTS:
1666 /* cannot mix with RFC2292 */
1667 if (OPTBIT(IN6P_RFC2292)) {
1668 error = EINVAL;
1669 break;
1670 }
1671 OPTSET(IN6P_DSTOPTS);
1672 break;
1673
1674 case IPV6_RECVRTHDRDSTOPTS:
1675 /* cannot mix with RFC2292 */
1676 if (OPTBIT(IN6P_RFC2292)) {
1677 error = EINVAL;
1678 break;
1679 }
1680 OPTSET(IN6P_RTHDRDSTOPTS);
1681 break;
1682
1683 case IPV6_RECVRTHDR:
1684 /* cannot mix with RFC2292 */
1685 if (OPTBIT(IN6P_RFC2292)) {
1686 error = EINVAL;
1687 break;
1688 }
1689 OPTSET(IN6P_RTHDR);
1690 break;
1691
1692 case IPV6_FAITH:
1693 OPTSET(INP_FAITH);
1694 break;
1695
1696 case IPV6_RECVPATHMTU:
1697 /*
1698 * We ignore this option for TCP
1699 * sockets.
1700 * (RFC3542 leaves this case
1701 * unspecified.)
1702 */
1703 if (uproto != IPPROTO_TCP)
1704 OPTSET(IN6P_MTU);
1705 break;
1706
1707 case IPV6_V6ONLY:
1708 /*
1709 * make setsockopt(IPV6_V6ONLY)
1710 * available only prior to bind(2).
1711 * see ipng mailing list, Jun 22 2001.
1712 */
1713 if (in6p->inp_lport ||
1714 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1715 error = EINVAL;
1716 break;
1717 }
1718 OPTSET(IN6P_IPV6_V6ONLY);
1719 if (optval)
1720 in6p->inp_vflag &= ~INP_IPV4;
1721 else
1722 in6p->inp_vflag |= INP_IPV4;
1723 break;
1724 case IPV6_RECVTCLASS:
1725 /* cannot mix with RFC2292 XXX */
1726 if (OPTBIT(IN6P_RFC2292)) {
1727 error = EINVAL;
1728 break;
1729 }
1730 OPTSET(IN6P_TCLASS);
1731 break;
1732 case IPV6_AUTOFLOWLABEL:
1733 OPTSET(IN6P_AUTOFLOWLABEL);
1734 break;
1735
1736 case IPV6_BINDANY:
1737 OPTSET(INP_BINDANY);
1738 break;
1739 }
1740 break;
1741
1742 case IPV6_TCLASS:
1743 case IPV6_DONTFRAG:
1744 case IPV6_USE_MIN_MTU:
1745 case IPV6_PREFER_TEMPADDR:
1746 if (optlen != sizeof(optval)) {
1747 error = EINVAL;
1748 break;
1749 }
1750 error = sooptcopyin(sopt, &optval,
1751 sizeof optval, sizeof optval);
1752 if (error)
1753 break;
1754 {
1755 struct ip6_pktopts **optp;
1756 optp = &in6p->in6p_outputopts;
1757 error = ip6_pcbopt(optname,
1758 (u_char *)&optval, sizeof(optval),
1759 optp, (td != NULL) ? td->td_ucred :
1760 NULL, uproto);
1761 break;
1762 }
1763
1764 case IPV6_2292PKTINFO:
1765 case IPV6_2292HOPLIMIT:
1766 case IPV6_2292HOPOPTS:
1767 case IPV6_2292DSTOPTS:
1768 case IPV6_2292RTHDR:
1769 /* RFC 2292 */
1770 if (optlen != sizeof(int)) {
1771 error = EINVAL;
1772 break;
1773 }
1774 error = sooptcopyin(sopt, &optval,
1775 sizeof optval, sizeof optval);
1776 if (error)
1777 break;
1778 switch (optname) {
1779 case IPV6_2292PKTINFO:
1780 OPTSET2292(IN6P_PKTINFO);
1781 break;
1782 case IPV6_2292HOPLIMIT:
1783 OPTSET2292(IN6P_HOPLIMIT);
1784 break;
1785 case IPV6_2292HOPOPTS:
1786 /*
1787 * Check super-user privilege.
1788 * See comments for IPV6_RECVHOPOPTS.
1789 */
1790 if (td != NULL) {
1791 error = priv_check(td,
1792 PRIV_NETINET_SETHDROPTS);
1793 if (error)
1794 return (error);
1795 }
1796 OPTSET2292(IN6P_HOPOPTS);
1797 break;
1798 case IPV6_2292DSTOPTS:
1799 if (td != NULL) {
1800 error = priv_check(td,
1801 PRIV_NETINET_SETHDROPTS);
1802 if (error)
1803 return (error);
1804 }
1805 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1806 break;
1807 case IPV6_2292RTHDR:
1808 OPTSET2292(IN6P_RTHDR);
1809 break;
1810 }
1811 break;
1812 case IPV6_PKTINFO:
1813 case IPV6_HOPOPTS:
1814 case IPV6_RTHDR:
1815 case IPV6_DSTOPTS:
1816 case IPV6_RTHDRDSTOPTS:
1817 case IPV6_NEXTHOP:
1818 {
1819 /* new advanced API (RFC3542) */
1820 u_char *optbuf;
1821 u_char optbuf_storage[MCLBYTES];
1822 int optlen;
1823 struct ip6_pktopts **optp;
1824
1825 /* cannot mix with RFC2292 */
1826 if (OPTBIT(IN6P_RFC2292)) {
1827 error = EINVAL;
1828 break;
1829 }
1830
1831 /*
1832 * We only ensure valsize is not too large
1833 * here. Further validation will be done
1834 * later.
1835 */
1836 error = sooptcopyin(sopt, optbuf_storage,
1837 sizeof(optbuf_storage), 0);
1838 if (error)
1839 break;
1840 optlen = sopt->sopt_valsize;
1841 optbuf = optbuf_storage;
1842 optp = &in6p->in6p_outputopts;
1843 error = ip6_pcbopt(optname, optbuf, optlen,
1844 optp, (td != NULL) ? td->td_ucred : NULL,
1845 uproto);
1846 break;
1847 }
1848 #undef OPTSET
1849
1850 case IPV6_MULTICAST_IF:
1851 case IPV6_MULTICAST_HOPS:
1852 case IPV6_MULTICAST_LOOP:
1853 case IPV6_JOIN_GROUP:
1854 case IPV6_LEAVE_GROUP:
1855 case IPV6_MSFILTER:
1856 case MCAST_BLOCK_SOURCE:
1857 case MCAST_UNBLOCK_SOURCE:
1858 case MCAST_JOIN_GROUP:
1859 case MCAST_LEAVE_GROUP:
1860 case MCAST_JOIN_SOURCE_GROUP:
1861 case MCAST_LEAVE_SOURCE_GROUP:
1862 error = ip6_setmoptions(in6p, sopt);
1863 break;
1864
1865 case IPV6_PORTRANGE:
1866 error = sooptcopyin(sopt, &optval,
1867 sizeof optval, sizeof optval);
1868 if (error)
1869 break;
1870
1871 INP_WLOCK(in6p);
1872 switch (optval) {
1873 case IPV6_PORTRANGE_DEFAULT:
1874 in6p->inp_flags &= ~(INP_LOWPORT);
1875 in6p->inp_flags &= ~(INP_HIGHPORT);
1876 break;
1877
1878 case IPV6_PORTRANGE_HIGH:
1879 in6p->inp_flags &= ~(INP_LOWPORT);
1880 in6p->inp_flags |= INP_HIGHPORT;
1881 break;
1882
1883 case IPV6_PORTRANGE_LOW:
1884 in6p->inp_flags &= ~(INP_HIGHPORT);
1885 in6p->inp_flags |= INP_LOWPORT;
1886 break;
1887
1888 default:
1889 error = EINVAL;
1890 break;
1891 }
1892 INP_WUNLOCK(in6p);
1893 break;
1894
1895 #ifdef IPSEC
1896 case IPV6_IPSEC_POLICY:
1897 {
1898 caddr_t req;
1899 struct mbuf *m;
1900
1901 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1902 break;
1903 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1904 break;
1905 req = mtod(m, caddr_t);
1906 error = ipsec_set_policy(in6p, optname, req,
1907 m->m_len, (sopt->sopt_td != NULL) ?
1908 sopt->sopt_td->td_ucred : NULL);
1909 m_freem(m);
1910 break;
1911 }
1912 #endif /* IPSEC */
1913
1914 default:
1915 error = ENOPROTOOPT;
1916 break;
1917 }
1918 break;
1919
1920 case SOPT_GET:
1921 switch (optname) {
1922
1923 case IPV6_2292PKTOPTIONS:
1924 #ifdef IPV6_PKTOPTIONS
1925 case IPV6_PKTOPTIONS:
1926 #endif
1927 /*
1928 * RFC3542 (effectively) deprecated the
1929 * semantics of the 2292-style pktoptions.
1930 * Since it was not reliable in nature (i.e.,
1931 * applications had to expect the lack of some
1932 * information after all), it would make sense
1933 * to simplify this part by always returning
1934 * empty data.
1935 */
1936 sopt->sopt_valsize = 0;
1937 break;
1938
1939 case IPV6_RECVHOPOPTS:
1940 case IPV6_RECVDSTOPTS:
1941 case IPV6_RECVRTHDRDSTOPTS:
1942 case IPV6_UNICAST_HOPS:
1943 case IPV6_RECVPKTINFO:
1944 case IPV6_RECVHOPLIMIT:
1945 case IPV6_RECVRTHDR:
1946 case IPV6_RECVPATHMTU:
1947
1948 case IPV6_FAITH:
1949 case IPV6_V6ONLY:
1950 case IPV6_PORTRANGE:
1951 case IPV6_RECVTCLASS:
1952 case IPV6_AUTOFLOWLABEL:
1953 case IPV6_BINDANY:
1954 switch (optname) {
1955
1956 case IPV6_RECVHOPOPTS:
1957 optval = OPTBIT(IN6P_HOPOPTS);
1958 break;
1959
1960 case IPV6_RECVDSTOPTS:
1961 optval = OPTBIT(IN6P_DSTOPTS);
1962 break;
1963
1964 case IPV6_RECVRTHDRDSTOPTS:
1965 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1966 break;
1967
1968 case IPV6_UNICAST_HOPS:
1969 optval = in6p->in6p_hops;
1970 break;
1971
1972 case IPV6_RECVPKTINFO:
1973 optval = OPTBIT(IN6P_PKTINFO);
1974 break;
1975
1976 case IPV6_RECVHOPLIMIT:
1977 optval = OPTBIT(IN6P_HOPLIMIT);
1978 break;
1979
1980 case IPV6_RECVRTHDR:
1981 optval = OPTBIT(IN6P_RTHDR);
1982 break;
1983
1984 case IPV6_RECVPATHMTU:
1985 optval = OPTBIT(IN6P_MTU);
1986 break;
1987
1988 case IPV6_FAITH:
1989 optval = OPTBIT(INP_FAITH);
1990 break;
1991
1992 case IPV6_V6ONLY:
1993 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1994 break;
1995
1996 case IPV6_PORTRANGE:
1997 {
1998 int flags;
1999 flags = in6p->inp_flags;
2000 if (flags & INP_HIGHPORT)
2001 optval = IPV6_PORTRANGE_HIGH;
2002 else if (flags & INP_LOWPORT)
2003 optval = IPV6_PORTRANGE_LOW;
2004 else
2005 optval = 0;
2006 break;
2007 }
2008 case IPV6_RECVTCLASS:
2009 optval = OPTBIT(IN6P_TCLASS);
2010 break;
2011
2012 case IPV6_AUTOFLOWLABEL:
2013 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2014 break;
2015
2016 case IPV6_BINDANY:
2017 optval = OPTBIT(INP_BINDANY);
2018 break;
2019 }
2020 if (error)
2021 break;
2022 error = sooptcopyout(sopt, &optval,
2023 sizeof optval);
2024 break;
2025
2026 case IPV6_PATHMTU:
2027 {
2028 u_long pmtu = 0;
2029 struct ip6_mtuinfo mtuinfo;
2030 struct route_in6 sro;
2031
2032 bzero(&sro, sizeof(sro));
2033
2034 if (!(so->so_state & SS_ISCONNECTED))
2035 return (ENOTCONN);
2036 /*
2037 * XXX: we dot not consider the case of source
2038 * routing, or optional information to specify
2039 * the outgoing interface.
2040 */
2041 error = ip6_getpmtu(&sro, NULL, NULL,
2042 &in6p->in6p_faddr, &pmtu, NULL,
2043 so->so_fibnum);
2044 if (sro.ro_rt)
2045 RTFREE(sro.ro_rt);
2046 if (error)
2047 break;
2048 if (pmtu > IPV6_MAXPACKET)
2049 pmtu = IPV6_MAXPACKET;
2050
2051 bzero(&mtuinfo, sizeof(mtuinfo));
2052 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2053 optdata = (void *)&mtuinfo;
2054 optdatalen = sizeof(mtuinfo);
2055 error = sooptcopyout(sopt, optdata,
2056 optdatalen);
2057 break;
2058 }
2059
2060 case IPV6_2292PKTINFO:
2061 case IPV6_2292HOPLIMIT:
2062 case IPV6_2292HOPOPTS:
2063 case IPV6_2292RTHDR:
2064 case IPV6_2292DSTOPTS:
2065 switch (optname) {
2066 case IPV6_2292PKTINFO:
2067 optval = OPTBIT(IN6P_PKTINFO);
2068 break;
2069 case IPV6_2292HOPLIMIT:
2070 optval = OPTBIT(IN6P_HOPLIMIT);
2071 break;
2072 case IPV6_2292HOPOPTS:
2073 optval = OPTBIT(IN6P_HOPOPTS);
2074 break;
2075 case IPV6_2292RTHDR:
2076 optval = OPTBIT(IN6P_RTHDR);
2077 break;
2078 case IPV6_2292DSTOPTS:
2079 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2080 break;
2081 }
2082 error = sooptcopyout(sopt, &optval,
2083 sizeof optval);
2084 break;
2085 case IPV6_PKTINFO:
2086 case IPV6_HOPOPTS:
2087 case IPV6_RTHDR:
2088 case IPV6_DSTOPTS:
2089 case IPV6_RTHDRDSTOPTS:
2090 case IPV6_NEXTHOP:
2091 case IPV6_TCLASS:
2092 case IPV6_DONTFRAG:
2093 case IPV6_USE_MIN_MTU:
2094 case IPV6_PREFER_TEMPADDR:
2095 error = ip6_getpcbopt(in6p->in6p_outputopts,
2096 optname, sopt);
2097 break;
2098
2099 case IPV6_MULTICAST_IF:
2100 case IPV6_MULTICAST_HOPS:
2101 case IPV6_MULTICAST_LOOP:
2102 case IPV6_MSFILTER:
2103 error = ip6_getmoptions(in6p, sopt);
2104 break;
2105
2106 #ifdef IPSEC
2107 case IPV6_IPSEC_POLICY:
2108 {
2109 caddr_t req = NULL;
2110 size_t len = 0;
2111 struct mbuf *m = NULL;
2112 struct mbuf **mp = &m;
2113 size_t ovalsize = sopt->sopt_valsize;
2114 caddr_t oval = (caddr_t)sopt->sopt_val;
2115
2116 error = soopt_getm(sopt, &m); /* XXX */
2117 if (error != 0)
2118 break;
2119 error = soopt_mcopyin(sopt, m); /* XXX */
2120 if (error != 0)
2121 break;
2122 sopt->sopt_valsize = ovalsize;
2123 sopt->sopt_val = oval;
2124 if (m) {
2125 req = mtod(m, caddr_t);
2126 len = m->m_len;
2127 }
2128 error = ipsec_get_policy(in6p, req, len, mp);
2129 if (error == 0)
2130 error = soopt_mcopyout(sopt, m); /* XXX */
2131 if (error == 0 && m)
2132 m_freem(m);
2133 break;
2134 }
2135 #endif /* IPSEC */
2136
2137 default:
2138 error = ENOPROTOOPT;
2139 break;
2140 }
2141 break;
2142 }
2143 }
2144 return (error);
2145 }
2146
2147 int
2148 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2149 {
2150 int error = 0, optval, optlen;
2151 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2152 struct inpcb *in6p = sotoinpcb(so);
2153 int level, op, optname;
2154
2155 level = sopt->sopt_level;
2156 op = sopt->sopt_dir;
2157 optname = sopt->sopt_name;
2158 optlen = sopt->sopt_valsize;
2159
2160 if (level != IPPROTO_IPV6) {
2161 return (EINVAL);
2162 }
2163
2164 switch (optname) {
2165 case IPV6_CHECKSUM:
2166 /*
2167 * For ICMPv6 sockets, no modification allowed for checksum
2168 * offset, permit "no change" values to help existing apps.
2169 *
2170 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2171 * for an ICMPv6 socket will fail."
2172 * The current behavior does not meet RFC3542.
2173 */
2174 switch (op) {
2175 case SOPT_SET:
2176 if (optlen != sizeof(int)) {
2177 error = EINVAL;
2178 break;
2179 }
2180 error = sooptcopyin(sopt, &optval, sizeof(optval),
2181 sizeof(optval));
2182 if (error)
2183 break;
2184 if ((optval % 2) != 0) {
2185 /* the API assumes even offset values */
2186 error = EINVAL;
2187 } else if (so->so_proto->pr_protocol ==
2188 IPPROTO_ICMPV6) {
2189 if (optval != icmp6off)
2190 error = EINVAL;
2191 } else
2192 in6p->in6p_cksum = optval;
2193 break;
2194
2195 case SOPT_GET:
2196 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2197 optval = icmp6off;
2198 else
2199 optval = in6p->in6p_cksum;
2200
2201 error = sooptcopyout(sopt, &optval, sizeof(optval));
2202 break;
2203
2204 default:
2205 error = EINVAL;
2206 break;
2207 }
2208 break;
2209
2210 default:
2211 error = ENOPROTOOPT;
2212 break;
2213 }
2214
2215 return (error);
2216 }
2217
2218 /*
2219 * Set up IP6 options in pcb for insertion in output packets or
2220 * specifying behavior of outgoing packets.
2221 */
2222 static int
2223 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2224 struct socket *so, struct sockopt *sopt)
2225 {
2226 struct ip6_pktopts *opt = *pktopt;
2227 int error = 0;
2228 struct thread *td = sopt->sopt_td;
2229
2230 /* turn off any old options. */
2231 if (opt) {
2232 #ifdef DIAGNOSTIC
2233 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2234 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2235 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2236 printf("ip6_pcbopts: all specified options are cleared.\n");
2237 #endif
2238 ip6_clearpktopts(opt, -1);
2239 } else
2240 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2241 *pktopt = NULL;
2242
2243 if (!m || m->m_len == 0) {
2244 /*
2245 * Only turning off any previous options, regardless of
2246 * whether the opt is just created or given.
2247 */
2248 free(opt, M_IP6OPT);
2249 return (0);
2250 }
2251
2252 /* set options specified by user. */
2253 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2254 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2255 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2256 free(opt, M_IP6OPT);
2257 return (error);
2258 }
2259 *pktopt = opt;
2260 return (0);
2261 }
2262
2263 /*
2264 * initialize ip6_pktopts. beware that there are non-zero default values in
2265 * the struct.
2266 */
2267 void
2268 ip6_initpktopts(struct ip6_pktopts *opt)
2269 {
2270
2271 bzero(opt, sizeof(*opt));
2272 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2273 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2274 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2275 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2276 }
2277
2278 static int
2279 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2280 struct ucred *cred, int uproto)
2281 {
2282 struct ip6_pktopts *opt;
2283
2284 if (*pktopt == NULL) {
2285 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2286 M_WAITOK);
2287 ip6_initpktopts(*pktopt);
2288 }
2289 opt = *pktopt;
2290
2291 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2292 }
2293
2294 static int
2295 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2296 {
2297 void *optdata = NULL;
2298 int optdatalen = 0;
2299 struct ip6_ext *ip6e;
2300 int error = 0;
2301 struct in6_pktinfo null_pktinfo;
2302 int deftclass = 0, on;
2303 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2304 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2305
2306 switch (optname) {
2307 case IPV6_PKTINFO:
2308 if (pktopt && pktopt->ip6po_pktinfo)
2309 optdata = (void *)pktopt->ip6po_pktinfo;
2310 else {
2311 /* XXX: we don't have to do this every time... */
2312 bzero(&null_pktinfo, sizeof(null_pktinfo));
2313 optdata = (void *)&null_pktinfo;
2314 }
2315 optdatalen = sizeof(struct in6_pktinfo);
2316 break;
2317 case IPV6_TCLASS:
2318 if (pktopt && pktopt->ip6po_tclass >= 0)
2319 optdata = (void *)&pktopt->ip6po_tclass;
2320 else
2321 optdata = (void *)&deftclass;
2322 optdatalen = sizeof(int);
2323 break;
2324 case IPV6_HOPOPTS:
2325 if (pktopt && pktopt->ip6po_hbh) {
2326 optdata = (void *)pktopt->ip6po_hbh;
2327 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2328 optdatalen = (ip6e->ip6e_len + 1) << 3;
2329 }
2330 break;
2331 case IPV6_RTHDR:
2332 if (pktopt && pktopt->ip6po_rthdr) {
2333 optdata = (void *)pktopt->ip6po_rthdr;
2334 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2335 optdatalen = (ip6e->ip6e_len + 1) << 3;
2336 }
2337 break;
2338 case IPV6_RTHDRDSTOPTS:
2339 if (pktopt && pktopt->ip6po_dest1) {
2340 optdata = (void *)pktopt->ip6po_dest1;
2341 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2342 optdatalen = (ip6e->ip6e_len + 1) << 3;
2343 }
2344 break;
2345 case IPV6_DSTOPTS:
2346 if (pktopt && pktopt->ip6po_dest2) {
2347 optdata = (void *)pktopt->ip6po_dest2;
2348 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2349 optdatalen = (ip6e->ip6e_len + 1) << 3;
2350 }
2351 break;
2352 case IPV6_NEXTHOP:
2353 if (pktopt && pktopt->ip6po_nexthop) {
2354 optdata = (void *)pktopt->ip6po_nexthop;
2355 optdatalen = pktopt->ip6po_nexthop->sa_len;
2356 }
2357 break;
2358 case IPV6_USE_MIN_MTU:
2359 if (pktopt)
2360 optdata = (void *)&pktopt->ip6po_minmtu;
2361 else
2362 optdata = (void *)&defminmtu;
2363 optdatalen = sizeof(int);
2364 break;
2365 case IPV6_DONTFRAG:
2366 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2367 on = 1;
2368 else
2369 on = 0;
2370 optdata = (void *)&on;
2371 optdatalen = sizeof(on);
2372 break;
2373 case IPV6_PREFER_TEMPADDR:
2374 if (pktopt)
2375 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2376 else
2377 optdata = (void *)&defpreftemp;
2378 optdatalen = sizeof(int);
2379 break;
2380 default: /* should not happen */
2381 #ifdef DIAGNOSTIC
2382 panic("ip6_getpcbopt: unexpected option\n");
2383 #endif
2384 return (ENOPROTOOPT);
2385 }
2386
2387 error = sooptcopyout(sopt, optdata, optdatalen);
2388
2389 return (error);
2390 }
2391
2392 void
2393 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2394 {
2395 if (pktopt == NULL)
2396 return;
2397
2398 if (optname == -1 || optname == IPV6_PKTINFO) {
2399 if (pktopt->ip6po_pktinfo)
2400 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2401 pktopt->ip6po_pktinfo = NULL;
2402 }
2403 if (optname == -1 || optname == IPV6_HOPLIMIT)
2404 pktopt->ip6po_hlim = -1;
2405 if (optname == -1 || optname == IPV6_TCLASS)
2406 pktopt->ip6po_tclass = -1;
2407 if (optname == -1 || optname == IPV6_NEXTHOP) {
2408 if (pktopt->ip6po_nextroute.ro_rt) {
2409 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2410 pktopt->ip6po_nextroute.ro_rt = NULL;
2411 }
2412 if (pktopt->ip6po_nexthop)
2413 free(pktopt->ip6po_nexthop, M_IP6OPT);
2414 pktopt->ip6po_nexthop = NULL;
2415 }
2416 if (optname == -1 || optname == IPV6_HOPOPTS) {
2417 if (pktopt->ip6po_hbh)
2418 free(pktopt->ip6po_hbh, M_IP6OPT);
2419 pktopt->ip6po_hbh = NULL;
2420 }
2421 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2422 if (pktopt->ip6po_dest1)
2423 free(pktopt->ip6po_dest1, M_IP6OPT);
2424 pktopt->ip6po_dest1 = NULL;
2425 }
2426 if (optname == -1 || optname == IPV6_RTHDR) {
2427 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2428 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2429 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2430 if (pktopt->ip6po_route.ro_rt) {
2431 RTFREE(pktopt->ip6po_route.ro_rt);
2432 pktopt->ip6po_route.ro_rt = NULL;
2433 }
2434 }
2435 if (optname == -1 || optname == IPV6_DSTOPTS) {
2436 if (pktopt->ip6po_dest2)
2437 free(pktopt->ip6po_dest2, M_IP6OPT);
2438 pktopt->ip6po_dest2 = NULL;
2439 }
2440 }
2441
2442 #define PKTOPT_EXTHDRCPY(type) \
2443 do {\
2444 if (src->type) {\
2445 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2446 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2447 if (dst->type == NULL && canwait == M_NOWAIT)\
2448 goto bad;\
2449 bcopy(src->type, dst->type, hlen);\
2450 }\
2451 } while (/*CONSTCOND*/ 0)
2452
2453 static int
2454 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2455 {
2456 if (dst == NULL || src == NULL) {
2457 printf("ip6_clearpktopts: invalid argument\n");
2458 return (EINVAL);
2459 }
2460
2461 dst->ip6po_hlim = src->ip6po_hlim;
2462 dst->ip6po_tclass = src->ip6po_tclass;
2463 dst->ip6po_flags = src->ip6po_flags;
2464 dst->ip6po_minmtu = src->ip6po_minmtu;
2465 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2466 if (src->ip6po_pktinfo) {
2467 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2468 M_IP6OPT, canwait);
2469 if (dst->ip6po_pktinfo == NULL)
2470 goto bad;
2471 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2472 }
2473 if (src->ip6po_nexthop) {
2474 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2475 M_IP6OPT, canwait);
2476 if (dst->ip6po_nexthop == NULL)
2477 goto bad;
2478 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2479 src->ip6po_nexthop->sa_len);
2480 }
2481 PKTOPT_EXTHDRCPY(ip6po_hbh);
2482 PKTOPT_EXTHDRCPY(ip6po_dest1);
2483 PKTOPT_EXTHDRCPY(ip6po_dest2);
2484 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2485 return (0);
2486
2487 bad:
2488 ip6_clearpktopts(dst, -1);
2489 return (ENOBUFS);
2490 }
2491 #undef PKTOPT_EXTHDRCPY
2492
2493 struct ip6_pktopts *
2494 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2495 {
2496 int error;
2497 struct ip6_pktopts *dst;
2498
2499 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2500 if (dst == NULL)
2501 return (NULL);
2502 ip6_initpktopts(dst);
2503
2504 if ((error = copypktopts(dst, src, canwait)) != 0) {
2505 free(dst, M_IP6OPT);
2506 return (NULL);
2507 }
2508
2509 return (dst);
2510 }
2511
2512 void
2513 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2514 {
2515 if (pktopt == NULL)
2516 return;
2517
2518 ip6_clearpktopts(pktopt, -1);
2519
2520 free(pktopt, M_IP6OPT);
2521 }
2522
2523 /*
2524 * Set IPv6 outgoing packet options based on advanced API.
2525 */
2526 int
2527 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2528 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2529 {
2530 struct cmsghdr *cm = 0;
2531
2532 if (control == NULL || opt == NULL)
2533 return (EINVAL);
2534
2535 ip6_initpktopts(opt);
2536 if (stickyopt) {
2537 int error;
2538
2539 /*
2540 * If stickyopt is provided, make a local copy of the options
2541 * for this particular packet, then override them by ancillary
2542 * objects.
2543 * XXX: copypktopts() does not copy the cached route to a next
2544 * hop (if any). This is not very good in terms of efficiency,
2545 * but we can allow this since this option should be rarely
2546 * used.
2547 */
2548 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2549 return (error);
2550 }
2551
2552 /*
2553 * XXX: Currently, we assume all the optional information is stored
2554 * in a single mbuf.
2555 */
2556 if (control->m_next)
2557 return (EINVAL);
2558
2559 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2560 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2561 int error;
2562
2563 if (control->m_len < CMSG_LEN(0))
2564 return (EINVAL);
2565
2566 cm = mtod(control, struct cmsghdr *);
2567 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2568 return (EINVAL);
2569 if (cm->cmsg_level != IPPROTO_IPV6)
2570 continue;
2571
2572 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2573 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2574 if (error)
2575 return (error);
2576 }
2577
2578 return (0);
2579 }
2580
2581 /*
2582 * Set a particular packet option, as a sticky option or an ancillary data
2583 * item. "len" can be 0 only when it's a sticky option.
2584 * We have 4 cases of combination of "sticky" and "cmsg":
2585 * "sticky=0, cmsg=0": impossible
2586 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2587 * "sticky=1, cmsg=0": RFC3542 socket option
2588 * "sticky=1, cmsg=1": RFC2292 socket option
2589 */
2590 static int
2591 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2592 struct ucred *cred, int sticky, int cmsg, int uproto)
2593 {
2594 int minmtupolicy, preftemp;
2595 int error;
2596
2597 if (!sticky && !cmsg) {
2598 #ifdef DIAGNOSTIC
2599 printf("ip6_setpktopt: impossible case\n");
2600 #endif
2601 return (EINVAL);
2602 }
2603
2604 /*
2605 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2606 * not be specified in the context of RFC3542. Conversely,
2607 * RFC3542 types should not be specified in the context of RFC2292.
2608 */
2609 if (!cmsg) {
2610 switch (optname) {
2611 case IPV6_2292PKTINFO:
2612 case IPV6_2292HOPLIMIT:
2613 case IPV6_2292NEXTHOP:
2614 case IPV6_2292HOPOPTS:
2615 case IPV6_2292DSTOPTS:
2616 case IPV6_2292RTHDR:
2617 case IPV6_2292PKTOPTIONS:
2618 return (ENOPROTOOPT);
2619 }
2620 }
2621 if (sticky && cmsg) {
2622 switch (optname) {
2623 case IPV6_PKTINFO:
2624 case IPV6_HOPLIMIT:
2625 case IPV6_NEXTHOP:
2626 case IPV6_HOPOPTS:
2627 case IPV6_DSTOPTS:
2628 case IPV6_RTHDRDSTOPTS:
2629 case IPV6_RTHDR:
2630 case IPV6_USE_MIN_MTU:
2631 case IPV6_DONTFRAG:
2632 case IPV6_TCLASS:
2633 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2634 return (ENOPROTOOPT);
2635 }
2636 }
2637
2638 switch (optname) {
2639 case IPV6_2292PKTINFO:
2640 case IPV6_PKTINFO:
2641 {
2642 struct ifnet *ifp = NULL;
2643 struct in6_pktinfo *pktinfo;
2644
2645 if (len != sizeof(struct in6_pktinfo))
2646 return (EINVAL);
2647
2648 pktinfo = (struct in6_pktinfo *)buf;
2649
2650 /*
2651 * An application can clear any sticky IPV6_PKTINFO option by
2652 * doing a "regular" setsockopt with ipi6_addr being
2653 * in6addr_any and ipi6_ifindex being zero.
2654 * [RFC 3542, Section 6]
2655 */
2656 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2657 pktinfo->ipi6_ifindex == 0 &&
2658 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2659 ip6_clearpktopts(opt, optname);
2660 break;
2661 }
2662
2663 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2664 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2665 return (EINVAL);
2666 }
2667
2668 /* validate the interface index if specified. */
2669 if (pktinfo->ipi6_ifindex > V_if_index ||
2670 pktinfo->ipi6_ifindex < 0) {
2671 return (ENXIO);
2672 }
2673 if (pktinfo->ipi6_ifindex) {
2674 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2675 if (ifp == NULL)
2676 return (ENXIO);
2677 }
2678
2679 /*
2680 * We store the address anyway, and let in6_selectsrc()
2681 * validate the specified address. This is because ipi6_addr
2682 * may not have enough information about its scope zone, and
2683 * we may need additional information (such as outgoing
2684 * interface or the scope zone of a destination address) to
2685 * disambiguate the scope.
2686 * XXX: the delay of the validation may confuse the
2687 * application when it is used as a sticky option.
2688 */
2689 if (opt->ip6po_pktinfo == NULL) {
2690 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2691 M_IP6OPT, M_NOWAIT);
2692 if (opt->ip6po_pktinfo == NULL)
2693 return (ENOBUFS);
2694 }
2695 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2696 break;
2697 }
2698
2699 case IPV6_2292HOPLIMIT:
2700 case IPV6_HOPLIMIT:
2701 {
2702 int *hlimp;
2703
2704 /*
2705 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2706 * to simplify the ordering among hoplimit options.
2707 */
2708 if (optname == IPV6_HOPLIMIT && sticky)
2709 return (ENOPROTOOPT);
2710
2711 if (len != sizeof(int))
2712 return (EINVAL);
2713 hlimp = (int *)buf;
2714 if (*hlimp < -1 || *hlimp > 255)
2715 return (EINVAL);
2716
2717 opt->ip6po_hlim = *hlimp;
2718 break;
2719 }
2720
2721 case IPV6_TCLASS:
2722 {
2723 int tclass;
2724
2725 if (len != sizeof(int))
2726 return (EINVAL);
2727 tclass = *(int *)buf;
2728 if (tclass < -1 || tclass > 255)
2729 return (EINVAL);
2730
2731 opt->ip6po_tclass = tclass;
2732 break;
2733 }
2734
2735 case IPV6_2292NEXTHOP:
2736 case IPV6_NEXTHOP:
2737 if (cred != NULL) {
2738 error = priv_check_cred(cred,
2739 PRIV_NETINET_SETHDROPTS, 0);
2740 if (error)
2741 return (error);
2742 }
2743
2744 if (len == 0) { /* just remove the option */
2745 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2746 break;
2747 }
2748
2749 /* check if cmsg_len is large enough for sa_len */
2750 if (len < sizeof(struct sockaddr) || len < *buf)
2751 return (EINVAL);
2752
2753 switch (((struct sockaddr *)buf)->sa_family) {
2754 case AF_INET6:
2755 {
2756 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2757 int error;
2758
2759 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2760 return (EINVAL);
2761
2762 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2763 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2764 return (EINVAL);
2765 }
2766 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2767 != 0) {
2768 return (error);
2769 }
2770 break;
2771 }
2772 case AF_LINK: /* should eventually be supported */
2773 default:
2774 return (EAFNOSUPPORT);
2775 }
2776
2777 /* turn off the previous option, then set the new option. */
2778 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2779 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2780 if (opt->ip6po_nexthop == NULL)
2781 return (ENOBUFS);
2782 bcopy(buf, opt->ip6po_nexthop, *buf);
2783 break;
2784
2785 case IPV6_2292HOPOPTS:
2786 case IPV6_HOPOPTS:
2787 {
2788 struct ip6_hbh *hbh;
2789 int hbhlen;
2790
2791 /*
2792 * XXX: We don't allow a non-privileged user to set ANY HbH
2793 * options, since per-option restriction has too much
2794 * overhead.
2795 */
2796 if (cred != NULL) {
2797 error = priv_check_cred(cred,
2798 PRIV_NETINET_SETHDROPTS, 0);
2799 if (error)
2800 return (error);
2801 }
2802
2803 if (len == 0) {
2804 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2805 break; /* just remove the option */
2806 }
2807
2808 /* message length validation */
2809 if (len < sizeof(struct ip6_hbh))
2810 return (EINVAL);
2811 hbh = (struct ip6_hbh *)buf;
2812 hbhlen = (hbh->ip6h_len + 1) << 3;
2813 if (len != hbhlen)
2814 return (EINVAL);
2815
2816 /* turn off the previous option, then set the new option. */
2817 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2818 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2819 if (opt->ip6po_hbh == NULL)
2820 return (ENOBUFS);
2821 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2822
2823 break;
2824 }
2825
2826 case IPV6_2292DSTOPTS:
2827 case IPV6_DSTOPTS:
2828 case IPV6_RTHDRDSTOPTS:
2829 {
2830 struct ip6_dest *dest, **newdest = NULL;
2831 int destlen;
2832
2833 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2834 error = priv_check_cred(cred,
2835 PRIV_NETINET_SETHDROPTS, 0);
2836 if (error)
2837 return (error);
2838 }
2839
2840 if (len == 0) {
2841 ip6_clearpktopts(opt, optname);
2842 break; /* just remove the option */
2843 }
2844
2845 /* message length validation */
2846 if (len < sizeof(struct ip6_dest))
2847 return (EINVAL);
2848 dest = (struct ip6_dest *)buf;
2849 destlen = (dest->ip6d_len + 1) << 3;
2850 if (len != destlen)
2851 return (EINVAL);
2852
2853 /*
2854 * Determine the position that the destination options header
2855 * should be inserted; before or after the routing header.
2856 */
2857 switch (optname) {
2858 case IPV6_2292DSTOPTS:
2859 /*
2860 * The old advacned API is ambiguous on this point.
2861 * Our approach is to determine the position based
2862 * according to the existence of a routing header.
2863 * Note, however, that this depends on the order of the
2864 * extension headers in the ancillary data; the 1st
2865 * part of the destination options header must appear
2866 * before the routing header in the ancillary data,
2867 * too.
2868 * RFC3542 solved the ambiguity by introducing
2869 * separate ancillary data or option types.
2870 */
2871 if (opt->ip6po_rthdr == NULL)
2872 newdest = &opt->ip6po_dest1;
2873 else
2874 newdest = &opt->ip6po_dest2;
2875 break;
2876 case IPV6_RTHDRDSTOPTS:
2877 newdest = &opt->ip6po_dest1;
2878 break;
2879 case IPV6_DSTOPTS:
2880 newdest = &opt->ip6po_dest2;
2881 break;
2882 }
2883
2884 /* turn off the previous option, then set the new option. */
2885 ip6_clearpktopts(opt, optname);
2886 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2887 if (*newdest == NULL)
2888 return (ENOBUFS);
2889 bcopy(dest, *newdest, destlen);
2890
2891 break;
2892 }
2893
2894 case IPV6_2292RTHDR:
2895 case IPV6_RTHDR:
2896 {
2897 struct ip6_rthdr *rth;
2898 int rthlen;
2899
2900 if (len == 0) {
2901 ip6_clearpktopts(opt, IPV6_RTHDR);
2902 break; /* just remove the option */
2903 }
2904
2905 /* message length validation */
2906 if (len < sizeof(struct ip6_rthdr))
2907 return (EINVAL);
2908 rth = (struct ip6_rthdr *)buf;
2909 rthlen = (rth->ip6r_len + 1) << 3;
2910 if (len != rthlen)
2911 return (EINVAL);
2912
2913 switch (rth->ip6r_type) {
2914 case IPV6_RTHDR_TYPE_0:
2915 if (rth->ip6r_len == 0) /* must contain one addr */
2916 return (EINVAL);
2917 if (rth->ip6r_len % 2) /* length must be even */
2918 return (EINVAL);
2919 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2920 return (EINVAL);
2921 break;
2922 default:
2923 return (EINVAL); /* not supported */
2924 }
2925
2926 /* turn off the previous option */
2927 ip6_clearpktopts(opt, IPV6_RTHDR);
2928 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2929 if (opt->ip6po_rthdr == NULL)
2930 return (ENOBUFS);
2931 bcopy(rth, opt->ip6po_rthdr, rthlen);
2932
2933 break;
2934 }
2935
2936 case IPV6_USE_MIN_MTU:
2937 if (len != sizeof(int))
2938 return (EINVAL);
2939 minmtupolicy = *(int *)buf;
2940 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2941 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2942 minmtupolicy != IP6PO_MINMTU_ALL) {
2943 return (EINVAL);
2944 }
2945 opt->ip6po_minmtu = minmtupolicy;
2946 break;
2947
2948 case IPV6_DONTFRAG:
2949 if (len != sizeof(int))
2950 return (EINVAL);
2951
2952 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2953 /*
2954 * we ignore this option for TCP sockets.
2955 * (RFC3542 leaves this case unspecified.)
2956 */
2957 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2958 } else
2959 opt->ip6po_flags |= IP6PO_DONTFRAG;
2960 break;
2961
2962 case IPV6_PREFER_TEMPADDR:
2963 if (len != sizeof(int))
2964 return (EINVAL);
2965 preftemp = *(int *)buf;
2966 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2967 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2968 preftemp != IP6PO_TEMPADDR_PREFER) {
2969 return (EINVAL);
2970 }
2971 opt->ip6po_prefer_tempaddr = preftemp;
2972 break;
2973
2974 default:
2975 return (ENOPROTOOPT);
2976 } /* end of switch */
2977
2978 return (0);
2979 }
2980
2981 /*
2982 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2983 * packet to the input queue of a specified interface. Note that this
2984 * calls the output routine of the loopback "driver", but with an interface
2985 * pointer that might NOT be &loif -- easier than replicating that code here.
2986 */
2987 void
2988 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2989 {
2990 struct mbuf *copym;
2991 struct ip6_hdr *ip6;
2992
2993 copym = m_copy(m, 0, M_COPYALL);
2994 if (copym == NULL)
2995 return;
2996
2997 /*
2998 * Make sure to deep-copy IPv6 header portion in case the data
2999 * is in an mbuf cluster, so that we can safely override the IPv6
3000 * header portion later.
3001 */
3002 if ((copym->m_flags & M_EXT) != 0 ||
3003 copym->m_len < sizeof(struct ip6_hdr)) {
3004 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3005 if (copym == NULL)
3006 return;
3007 }
3008 ip6 = mtod(copym, struct ip6_hdr *);
3009 /*
3010 * clear embedded scope identifiers if necessary.
3011 * in6_clearscope will touch the addresses only when necessary.
3012 */
3013 in6_clearscope(&ip6->ip6_src);
3014 in6_clearscope(&ip6->ip6_dst);
3015 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3016 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3017 CSUM_PSEUDO_HDR;
3018 copym->m_pkthdr.csum_data = 0xffff;
3019 }
3020 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3021 }
3022
3023 /*
3024 * Chop IPv6 header off from the payload.
3025 */
3026 static int
3027 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3028 {
3029 struct mbuf *mh;
3030 struct ip6_hdr *ip6;
3031
3032 ip6 = mtod(m, struct ip6_hdr *);
3033 if (m->m_len > sizeof(*ip6)) {
3034 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3035 if (mh == 0) {
3036 m_freem(m);
3037 return ENOBUFS;
3038 }
3039 M_MOVE_PKTHDR(mh, m);
3040 MH_ALIGN(mh, sizeof(*ip6));
3041 m->m_len -= sizeof(*ip6);
3042 m->m_data += sizeof(*ip6);
3043 mh->m_next = m;
3044 m = mh;
3045 m->m_len = sizeof(*ip6);
3046 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3047 }
3048 exthdrs->ip6e_ip6 = m;
3049 return 0;
3050 }
3051
3052 /*
3053 * Compute IPv6 extension header length.
3054 */
3055 int
3056 ip6_optlen(struct inpcb *in6p)
3057 {
3058 int len;
3059
3060 if (!in6p->in6p_outputopts)
3061 return 0;
3062
3063 len = 0;
3064 #define elen(x) \
3065 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3066
3067 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3068 if (in6p->in6p_outputopts->ip6po_rthdr)
3069 /* dest1 is valid with rthdr only */
3070 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3071 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3072 len += elen(in6p->in6p_outputopts->ip6po_dest2);
3073 return len;
3074 #undef elen
3075 }
Cache object: 292baea532bfc67452e19878eaa0e394
|