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_ipsec.h"
69 #include "opt_sctp.h"
70 #include "opt_route.h"
71 #include "opt_rss.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/if_var.h>
90 #include <net/if_llatbl.h>
91 #include <net/netisr.h>
92 #include <net/route.h>
93 #include <net/pfil.h>
94 #include <net/rss_config.h>
95 #include <net/vnet.h>
96
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip_var.h>
100 #include <netinet6/in6_fib.h>
101 #include <netinet6/in6_var.h>
102 #include <netinet/ip6.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/tcp_var.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/in6_rss.h>
109
110 #include <netipsec/ipsec_support.h>
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 MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
134
135 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
136 struct ucred *, int);
137 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
138 struct socket *, struct sockopt *);
139 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
140 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
141 struct ucred *, int, int, int);
142
143 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
144 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
145 struct ip6_frag **);
146 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
147 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
148 static int ip6_getpmtu(struct route_in6 *, int,
149 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
150 u_int);
151 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
152 u_long *, int *, u_int);
153 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
154 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
155
156
157 /*
158 * Make an extension header from option data. hp is the source,
159 * mp is the destination, and _ol is the optlen.
160 */
161 #define MAKE_EXTHDR(hp, mp, _ol) \
162 do { \
163 if (hp) { \
164 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
165 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
166 ((eh)->ip6e_len + 1) << 3); \
167 if (error) \
168 goto freehdrs; \
169 (_ol) += (*(mp))->m_len; \
170 } \
171 } while (/*CONSTCOND*/ 0)
172
173 /*
174 * Form a chain of extension headers.
175 * m is the extension header mbuf
176 * mp is the previous mbuf in the chain
177 * p is the next header
178 * i is the type of option.
179 */
180 #define MAKE_CHAIN(m, mp, p, i)\
181 do {\
182 if (m) {\
183 if (!hdrsplit) \
184 panic("%s:%d: assumption failed: "\
185 "hdr not split: hdrsplit %d exthdrs %p",\
186 __func__, __LINE__, hdrsplit, &exthdrs);\
187 *mtod((m), u_char *) = *(p);\
188 *(p) = (i);\
189 p = mtod((m), u_char *);\
190 (m)->m_next = (mp)->m_next;\
191 (mp)->m_next = (m);\
192 (mp) = (m);\
193 }\
194 } while (/*CONSTCOND*/ 0)
195
196 void
197 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
198 {
199 u_short csum;
200
201 csum = in_cksum_skip(m, offset + plen, offset);
202 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
203 csum = 0xffff;
204 offset += m->m_pkthdr.csum_data; /* checksum offset */
205
206 if (offset + sizeof(csum) > m->m_len)
207 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
208 else
209 *(u_short *)mtodo(m, offset) = csum;
210 }
211
212 static int
213 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
214 int plen, int optlen, bool frag __unused)
215 {
216
217 KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
218 "csum_flags %#x frag %d\n",
219 __func__, __LINE__, plen, optlen, m, ifp, csum_flags, frag));
220
221 if ((csum_flags & CSUM_DELAY_DATA_IPV6) ||
222 #ifdef SCTP
223 (csum_flags & CSUM_SCTP_IPV6) ||
224 #endif
225 false) {
226 if (csum_flags & CSUM_DELAY_DATA_IPV6) {
227 in6_delayed_cksum(m, plen - optlen,
228 sizeof(struct ip6_hdr) + optlen);
229 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
230 }
231 #ifdef SCTP
232 if (csum_flags & CSUM_SCTP_IPV6) {
233 sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
234 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
235 }
236 #endif
237 }
238
239 return (0);
240 }
241
242 int
243 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
244 int mtu, uint32_t id)
245 {
246 struct mbuf *m, **mnext, *m_frgpart;
247 struct ip6_hdr *ip6, *mhip6;
248 struct ip6_frag *ip6f;
249 int off;
250 int error;
251 int tlen = m0->m_pkthdr.len;
252
253 KASSERT(( mtu % 8 == 0), ("Fragment length must be a multiple of 8"));
254
255 m = m0;
256 ip6 = mtod(m, struct ip6_hdr *);
257 mnext = &m->m_nextpkt;
258
259 for (off = hlen; off < tlen; off += mtu) {
260 m = m_gethdr(M_NOWAIT, MT_DATA);
261 if (!m) {
262 IP6STAT_INC(ip6s_odropped);
263 return (ENOBUFS);
264 }
265
266 /*
267 * Make sure the complete packet header gets copied
268 * from the originating mbuf to the newly created
269 * mbuf. This also ensures that existing firewall
270 * classification(s), VLAN tags and so on get copied
271 * to the resulting fragmented packet(s):
272 */
273 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
274 m_free(m);
275 IP6STAT_INC(ip6s_odropped);
276 return (ENOBUFS);
277 }
278
279 *mnext = m;
280 mnext = &m->m_nextpkt;
281 m->m_data += max_linkhdr;
282 mhip6 = mtod(m, struct ip6_hdr *);
283 *mhip6 = *ip6;
284 m->m_len = sizeof(*mhip6);
285 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
286 if (error) {
287 IP6STAT_INC(ip6s_odropped);
288 return (error);
289 }
290 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
291 if (off + mtu >= tlen)
292 mtu = tlen - off;
293 else
294 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
295 mhip6->ip6_plen = htons((u_short)(mtu + hlen +
296 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
297 if ((m_frgpart = m_copy(m0, off, mtu)) == NULL) {
298 IP6STAT_INC(ip6s_odropped);
299 return (ENOBUFS);
300 }
301 m_cat(m, m_frgpart);
302 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
303 ip6f->ip6f_reserved = 0;
304 ip6f->ip6f_ident = id;
305 ip6f->ip6f_nxt = nextproto;
306 IP6STAT_INC(ip6s_ofragments);
307 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
308 }
309
310 return (0);
311 }
312
313 /*
314 * IP6 output.
315 * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
316 * nxt, hlim, src, dst).
317 * This function may modify ver and hlim only.
318 * The mbuf chain containing the packet will be freed.
319 * The mbuf opt, if present, will not be freed.
320 * If route_in6 ro is present and has ro_rt initialized, route lookup would be
321 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
322 * then result of route lookup is stored in ro->ro_rt.
323 *
324 * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
325 * is uint32_t. So we use u_long to hold largest one, which is rt_mtu.
326 *
327 * ifpp - XXX: just for statistics
328 */
329 /*
330 * XXX TODO: no flowid is assigned for outbound flows?
331 */
332 int
333 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
334 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
335 struct ifnet **ifpp, struct inpcb *inp)
336 {
337 struct ip6_hdr *ip6;
338 struct ifnet *ifp, *origifp;
339 struct mbuf *m = m0;
340 struct mbuf *mprev;
341 int tlen, len;
342 struct route_in6 ip6route;
343 struct rtentry *rt = NULL;
344 struct sockaddr_in6 *dst, src_sa, dst_sa;
345 struct in6_addr odst;
346 u_char *nexthdrp;
347 int error = 0;
348 struct in6_ifaddr *ia = NULL;
349 u_long mtu;
350 int alwaysfrag, dontfrag;
351 u_int32_t optlen, plen = 0, unfragpartlen;
352 struct ip6_exthdrs exthdrs;
353 struct in6_addr src0, dst0;
354 u_int32_t zone;
355 struct route_in6 *ro_pmtu = NULL;
356 bool hdrsplit;
357 int sw_csum, tso;
358 int needfiblookup;
359 uint32_t fibnum;
360 struct m_tag *fwd_tag = NULL;
361 uint32_t id;
362
363 if (inp != NULL) {
364 INP_LOCK_ASSERT(inp);
365 M_SETFIB(m, inp->inp_inc.inc_fibnum);
366 if ((flags & IP_NODEFAULTFLOWID) == 0) {
367 /* Unconditionally set flowid. */
368 m->m_pkthdr.flowid = inp->inp_flowid;
369 M_HASHTYPE_SET(m, inp->inp_flowtype);
370 }
371 }
372
373 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
374 /*
375 * IPSec checking which handles several cases.
376 * FAST IPSEC: We re-injected the packet.
377 * XXX: need scope argument.
378 */
379 if (IPSEC_ENABLED(ipv6)) {
380 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
381 if (error == EINPROGRESS)
382 error = 0;
383 goto done;
384 }
385 }
386 #endif /* IPSEC */
387
388 /* Source address validation. */
389 ip6 = mtod(m, struct ip6_hdr *);
390 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
391 (flags & IPV6_UNSPECSRC) == 0) {
392 error = EOPNOTSUPP;
393 IP6STAT_INC(ip6s_badscope);
394 goto bad;
395 }
396 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
397 error = EOPNOTSUPP;
398 IP6STAT_INC(ip6s_badscope);
399 goto bad;
400 }
401
402 /*
403 * If we are given packet options to add extension headers prepare them.
404 * Calculate the total length of the extension header chain.
405 * Keep the length of the unfragmentable part for fragmentation.
406 */
407 bzero(&exthdrs, sizeof(exthdrs));
408 optlen = 0;
409 unfragpartlen = sizeof(struct ip6_hdr);
410 if (opt) {
411 /* Hop-by-Hop options header. */
412 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
413
414 /* Destination options header (1st part). */
415 if (opt->ip6po_rthdr) {
416 #ifndef RTHDR_SUPPORT_IMPLEMENTED
417 /*
418 * If there is a routing header, discard the packet
419 * right away here. RH0/1 are obsolete and we do not
420 * currently support RH2/3/4.
421 * People trying to use RH253/254 may want to disable
422 * this check.
423 * The moment we do support any routing header (again)
424 * this block should check the routing type more
425 * selectively.
426 */
427 error = EINVAL;
428 goto bad;
429 #endif
430
431 /*
432 * Destination options header (1st part).
433 * This only makes sense with a routing header.
434 * See Section 9.2 of RFC 3542.
435 * Disabling this part just for MIP6 convenience is
436 * a bad idea. We need to think carefully about a
437 * way to make the advanced API coexist with MIP6
438 * options, which might automatically be inserted in
439 * the kernel.
440 */
441 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
442 optlen);
443 }
444 /* Routing header. */
445 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
446
447 unfragpartlen += optlen;
448
449 /*
450 * NOTE: we don't add AH/ESP length here (done in
451 * ip6_ipsec_output()).
452 */
453
454 /* Destination options header (2nd part). */
455 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
456 }
457
458 /*
459 * If there is at least one extension header,
460 * separate IP6 header from the payload.
461 */
462 hdrsplit = false;
463 if (optlen) {
464 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
465 m = NULL;
466 goto freehdrs;
467 }
468 m = exthdrs.ip6e_ip6;
469 ip6 = mtod(m, struct ip6_hdr *);
470 hdrsplit = true;
471 }
472
473 /* Adjust mbuf packet header length. */
474 m->m_pkthdr.len += optlen;
475 plen = m->m_pkthdr.len - sizeof(*ip6);
476
477 /* If this is a jumbo payload, insert a jumbo payload option. */
478 if (plen > IPV6_MAXPACKET) {
479 if (!hdrsplit) {
480 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
481 m = NULL;
482 goto freehdrs;
483 }
484 m = exthdrs.ip6e_ip6;
485 ip6 = mtod(m, struct ip6_hdr *);
486 hdrsplit = true;
487 }
488 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
489 goto freehdrs;
490 ip6->ip6_plen = 0;
491 } else
492 ip6->ip6_plen = htons(plen);
493 nexthdrp = &ip6->ip6_nxt;
494
495 if (optlen) {
496 /*
497 * Concatenate headers and fill in next header fields.
498 * Here we have, on "m"
499 * IPv6 payload
500 * and we insert headers accordingly.
501 * Finally, we should be getting:
502 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
503 *
504 * During the header composing process "m" points to IPv6
505 * header. "mprev" points to an extension header prior to esp.
506 */
507 mprev = m;
508
509 /*
510 * We treat dest2 specially. This makes IPsec processing
511 * much easier. The goal here is to make mprev point the
512 * mbuf prior to dest2.
513 *
514 * Result: IPv6 dest2 payload.
515 * m and mprev will point to IPv6 header.
516 */
517 if (exthdrs.ip6e_dest2) {
518 if (!hdrsplit)
519 panic("%s:%d: assumption failed: "
520 "hdr not split: hdrsplit %d exthdrs %p",
521 __func__, __LINE__, hdrsplit, &exthdrs);
522 exthdrs.ip6e_dest2->m_next = m->m_next;
523 m->m_next = exthdrs.ip6e_dest2;
524 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
525 ip6->ip6_nxt = IPPROTO_DSTOPTS;
526 }
527
528 /*
529 * Result: IPv6 hbh dest1 rthdr dest2 payload.
530 * m will point to IPv6 header. mprev will point to the
531 * extension header prior to dest2 (rthdr in the above case).
532 */
533 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
534 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
535 IPPROTO_DSTOPTS);
536 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
537 IPPROTO_ROUTING);
538 }
539
540 IP6STAT_INC(ip6s_localout);
541
542 /* Route packet. */
543 if (ro == NULL) {
544 ro = &ip6route;
545 bzero((caddr_t)ro, sizeof(*ro));
546 }
547 ro_pmtu = ro;
548 if (opt && opt->ip6po_rthdr)
549 ro = &opt->ip6po_route;
550 dst = (struct sockaddr_in6 *)&ro->ro_dst;
551 #ifdef FLOWTABLE
552 if (ro->ro_rt == NULL)
553 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
554 #endif
555 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
556 again:
557 /*
558 * If specified, try to fill in the traffic class field.
559 * Do not override if a non-zero value is already set.
560 * We check the diffserv field and the ECN field separately.
561 */
562 if (opt && opt->ip6po_tclass >= 0) {
563 int mask = 0;
564
565 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
566 mask |= 0xfc;
567 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
568 mask |= 0x03;
569 if (mask != 0)
570 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
571 }
572
573 /* Fill in or override the hop limit field, if necessary. */
574 if (opt && opt->ip6po_hlim != -1)
575 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
576 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
577 if (im6o != NULL)
578 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
579 else
580 ip6->ip6_hlim = V_ip6_defmcasthlim;
581 }
582 /*
583 * Validate route against routing table additions;
584 * a better/more specific route might have been added.
585 * Make sure that the address family is set in route.
586 */
587 if (inp) {
588 ro->ro_dst.sin6_family = AF_INET6;
589 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
590 }
591 if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
592 ro->ro_dst.sin6_family == AF_INET6 &&
593 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
594 rt = ro->ro_rt;
595 ifp = ro->ro_rt->rt_ifp;
596 } else {
597 if (ro->ro_lle)
598 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
599 ro->ro_lle = NULL;
600 if (fwd_tag == NULL) {
601 bzero(&dst_sa, sizeof(dst_sa));
602 dst_sa.sin6_family = AF_INET6;
603 dst_sa.sin6_len = sizeof(dst_sa);
604 dst_sa.sin6_addr = ip6->ip6_dst;
605 }
606 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
607 &rt, fibnum);
608 if (error != 0) {
609 if (ifp != NULL)
610 in6_ifstat_inc(ifp, ifs6_out_discard);
611 goto bad;
612 }
613 }
614 if (rt == NULL) {
615 /*
616 * If in6_selectroute() does not return a route entry
617 * dst may not have been updated.
618 */
619 *dst = dst_sa; /* XXX */
620 }
621
622 /* Then rt (for unicast) and ifp must be non-NULL valid values. */
623 if ((flags & IPV6_FORWARDING) == 0) {
624 /* XXX: the FORWARDING flag can be set for mrouting. */
625 in6_ifstat_inc(ifp, ifs6_out_request);
626 }
627 if (rt != NULL) {
628 ia = (struct in6_ifaddr *)(rt->rt_ifa);
629 counter_u64_add(rt->rt_pksent, 1);
630 }
631
632 /* Setup data structures for scope ID checks. */
633 src0 = ip6->ip6_src;
634 bzero(&src_sa, sizeof(src_sa));
635 src_sa.sin6_family = AF_INET6;
636 src_sa.sin6_len = sizeof(src_sa);
637 src_sa.sin6_addr = ip6->ip6_src;
638
639 dst0 = ip6->ip6_dst;
640 /* Re-initialize to be sure. */
641 bzero(&dst_sa, sizeof(dst_sa));
642 dst_sa.sin6_family = AF_INET6;
643 dst_sa.sin6_len = sizeof(dst_sa);
644 dst_sa.sin6_addr = ip6->ip6_dst;
645
646 /* Check for valid scope ID. */
647 if (in6_setscope(&src0, ifp, &zone) == 0 &&
648 sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
649 in6_setscope(&dst0, ifp, &zone) == 0 &&
650 sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
651 /*
652 * The outgoing interface is in the zone of the source
653 * and destination addresses.
654 *
655 * Because the loopback interface cannot receive
656 * packets with a different scope ID than its own,
657 * there is a trick to pretend the outgoing packet
658 * was received by the real network interface, by
659 * setting "origifp" different from "ifp". This is
660 * only allowed when "ifp" is a loopback network
661 * interface. Refer to code in nd6_output_ifp() for
662 * more details.
663 */
664 origifp = ifp;
665
666 /*
667 * We should use ia_ifp to support the case of sending
668 * packets to an address of our own.
669 */
670 if (ia != NULL && ia->ia_ifp)
671 ifp = ia->ia_ifp;
672
673 } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
674 sa6_recoverscope(&src_sa) != 0 ||
675 sa6_recoverscope(&dst_sa) != 0 ||
676 dst_sa.sin6_scope_id == 0 ||
677 (src_sa.sin6_scope_id != 0 &&
678 src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
679 (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
680 /*
681 * If the destination network interface is not a
682 * loopback interface, or the destination network
683 * address has no scope ID, or the source address has
684 * a scope ID set which is different from the
685 * destination address one, or there is no network
686 * interface representing this scope ID, the address
687 * pair is considered invalid.
688 */
689 IP6STAT_INC(ip6s_badscope);
690 in6_ifstat_inc(ifp, ifs6_out_discard);
691 if (error == 0)
692 error = EHOSTUNREACH; /* XXX */
693 goto bad;
694 }
695 /* All scope ID checks are successful. */
696
697 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
698 if (opt && opt->ip6po_nextroute.ro_rt) {
699 /*
700 * The nexthop is explicitly specified by the
701 * application. We assume the next hop is an IPv6
702 * address.
703 */
704 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
705 }
706 else if ((rt->rt_flags & RTF_GATEWAY))
707 dst = (struct sockaddr_in6 *)rt->rt_gateway;
708 }
709
710 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
711 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
712 } else {
713 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
714 in6_ifstat_inc(ifp, ifs6_out_mcast);
715
716 /* Confirm that the outgoing interface supports multicast. */
717 if (!(ifp->if_flags & IFF_MULTICAST)) {
718 IP6STAT_INC(ip6s_noroute);
719 in6_ifstat_inc(ifp, ifs6_out_discard);
720 error = ENETUNREACH;
721 goto bad;
722 }
723 if ((im6o == NULL && in6_mcast_loop) ||
724 (im6o && im6o->im6o_multicast_loop)) {
725 /*
726 * Loop back multicast datagram if not expressly
727 * forbidden to do so, even if we have not joined
728 * the address; protocols will filter it later,
729 * thus deferring a hash lookup and lock acquisition
730 * at the expense of an m_copym().
731 */
732 ip6_mloopback(ifp, m);
733 } else {
734 /*
735 * If we are acting as a multicast router, perform
736 * multicast forwarding as if the packet had just
737 * arrived on the interface to which we are about
738 * to send. The multicast forwarding function
739 * recursively calls this function, using the
740 * IPV6_FORWARDING flag to prevent infinite recursion.
741 *
742 * Multicasts that are looped back by ip6_mloopback(),
743 * above, will be forwarded by the ip6_input() routine,
744 * if necessary.
745 */
746 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
747 /*
748 * XXX: ip6_mforward expects that rcvif is NULL
749 * when it is called from the originating path.
750 * However, it may not always be the case.
751 */
752 m->m_pkthdr.rcvif = NULL;
753 if (ip6_mforward(ip6, ifp, m) != 0) {
754 m_freem(m);
755 goto done;
756 }
757 }
758 }
759 /*
760 * Multicasts with a hoplimit of zero may be looped back,
761 * above, but must not be transmitted on a network.
762 * Also, multicasts addressed to the loopback interface
763 * are not sent -- the above call to ip6_mloopback() will
764 * loop back a copy if this host actually belongs to the
765 * destination group on the loopback interface.
766 */
767 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
768 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
769 m_freem(m);
770 goto done;
771 }
772 }
773
774 /*
775 * Fill the outgoing inteface to tell the upper layer
776 * to increment per-interface statistics.
777 */
778 if (ifpp)
779 *ifpp = ifp;
780
781 /* Determine path MTU. */
782 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
783 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
784 goto bad;
785
786 /*
787 * The caller of this function may specify to use the minimum MTU
788 * in some cases.
789 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
790 * setting. The logic is a bit complicated; by default, unicast
791 * packets will follow path MTU while multicast packets will be sent at
792 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
793 * including unicast ones will be sent at the minimum MTU. Multicast
794 * packets will always be sent at the minimum MTU unless
795 * IP6PO_MINMTU_DISABLE is explicitly specified.
796 * See RFC 3542 for more details.
797 */
798 if (mtu > IPV6_MMTU) {
799 if ((flags & IPV6_MINMTU))
800 mtu = IPV6_MMTU;
801 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
802 mtu = IPV6_MMTU;
803 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
804 (opt == NULL ||
805 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
806 mtu = IPV6_MMTU;
807 }
808 }
809
810 /*
811 * Clear embedded scope identifiers if necessary.
812 * in6_clearscope() will touch the addresses only when necessary.
813 */
814 in6_clearscope(&ip6->ip6_src);
815 in6_clearscope(&ip6->ip6_dst);
816
817 /*
818 * If the outgoing packet contains a hop-by-hop options header,
819 * it must be examined and processed even by the source node.
820 * (RFC 2460, section 4.)
821 */
822 if (exthdrs.ip6e_hbh) {
823 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
824 u_int32_t dummy; /* XXX unused */
825 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
826
827 #ifdef DIAGNOSTIC
828 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
829 panic("ip6e_hbh is not contiguous");
830 #endif
831 /*
832 * XXX: if we have to send an ICMPv6 error to the sender,
833 * we need the M_LOOP flag since icmp6_error() expects
834 * the IPv6 and the hop-by-hop options header are
835 * contiguous unless the flag is set.
836 */
837 m->m_flags |= M_LOOP;
838 m->m_pkthdr.rcvif = ifp;
839 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
840 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
841 &dummy, &plen) < 0) {
842 /* m was already freed at this point. */
843 error = EINVAL;/* better error? */
844 goto done;
845 }
846 m->m_flags &= ~M_LOOP; /* XXX */
847 m->m_pkthdr.rcvif = NULL;
848 }
849
850 /* Jump over all PFIL processing if hooks are not active. */
851 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
852 goto passout;
853
854 odst = ip6->ip6_dst;
855 /* Run through list of hooks for output packets. */
856 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, 0, inp);
857 if (error != 0 || m == NULL)
858 goto done;
859 /* adjust pointer */
860 ip6 = mtod(m, struct ip6_hdr *);
861
862 needfiblookup = 0;
863 /* See if destination IP address was changed by packet filter. */
864 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
865 m->m_flags |= M_SKIP_FIREWALL;
866 /* If destination is now ourself drop to ip6_input(). */
867 if (in6_localip(&ip6->ip6_dst)) {
868 m->m_flags |= M_FASTFWD_OURS;
869 if (m->m_pkthdr.rcvif == NULL)
870 m->m_pkthdr.rcvif = V_loif;
871 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
872 m->m_pkthdr.csum_flags |=
873 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
874 m->m_pkthdr.csum_data = 0xffff;
875 }
876 #ifdef SCTP
877 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
878 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
879 #endif
880 error = netisr_queue(NETISR_IPV6, m);
881 goto done;
882 } else {
883 RO_RTFREE(ro);
884 needfiblookup = 1; /* Redo the routing table lookup. */
885 if (ro->ro_lle)
886 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
887 ro->ro_lle = NULL;
888 }
889 }
890 /* See if fib was changed by packet filter. */
891 if (fibnum != M_GETFIB(m)) {
892 m->m_flags |= M_SKIP_FIREWALL;
893 fibnum = M_GETFIB(m);
894 RO_RTFREE(ro);
895 needfiblookup = 1;
896 if (ro->ro_lle)
897 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
898 ro->ro_lle = NULL;
899 }
900 if (needfiblookup)
901 goto again;
902
903 /* See if local, if yes, send it to netisr. */
904 if (m->m_flags & M_FASTFWD_OURS) {
905 if (m->m_pkthdr.rcvif == NULL)
906 m->m_pkthdr.rcvif = V_loif;
907 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
908 m->m_pkthdr.csum_flags |=
909 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
910 m->m_pkthdr.csum_data = 0xffff;
911 }
912 #ifdef SCTP
913 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
914 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
915 #endif
916 error = netisr_queue(NETISR_IPV6, m);
917 goto done;
918 }
919 /* Or forward to some other address? */
920 if ((m->m_flags & M_IP6_NEXTHOP) &&
921 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
922 dst = (struct sockaddr_in6 *)&ro->ro_dst;
923 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
924 m->m_flags |= M_SKIP_FIREWALL;
925 m->m_flags &= ~M_IP6_NEXTHOP;
926 m_tag_delete(m, fwd_tag);
927 goto again;
928 }
929
930 passout:
931 /*
932 * Send the packet to the outgoing interface.
933 * If necessary, do IPv6 fragmentation before sending.
934 *
935 * The logic here is rather complex:
936 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
937 * 1-a: send as is if tlen <= path mtu
938 * 1-b: fragment if tlen > path mtu
939 *
940 * 2: if user asks us not to fragment (dontfrag == 1)
941 * 2-a: send as is if tlen <= interface mtu
942 * 2-b: error if tlen > interface mtu
943 *
944 * 3: if we always need to attach fragment header (alwaysfrag == 1)
945 * always fragment
946 *
947 * 4: if dontfrag == 1 && alwaysfrag == 1
948 * error, as we cannot handle this conflicting request.
949 */
950 sw_csum = m->m_pkthdr.csum_flags;
951 if (!hdrsplit) {
952 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
953 sw_csum &= ~ifp->if_hwassist;
954 } else
955 tso = 0;
956 /*
957 * If we added extension headers, we will not do TSO and calculate the
958 * checksums ourselves for now.
959 * XXX-BZ Need a framework to know when the NIC can handle it, even
960 * with ext. hdrs.
961 */
962 error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
963 if (error != 0)
964 goto bad;
965 /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
966 tlen = m->m_pkthdr.len;
967
968 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
969 dontfrag = 1;
970 else
971 dontfrag = 0;
972 if (dontfrag && alwaysfrag) { /* Case 4. */
973 /* Conflicting request - can't transmit. */
974 error = EMSGSIZE;
975 goto bad;
976 }
977 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* Case 2-b. */
978 /*
979 * Even if the DONTFRAG option is specified, we cannot send the
980 * packet when the data length is larger than the MTU of the
981 * outgoing interface.
982 * Notify the error by sending IPV6_PATHMTU ancillary data if
983 * application wanted to know the MTU value. Also return an
984 * error code (this is not described in the API spec).
985 */
986 if (inp != NULL)
987 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
988 error = EMSGSIZE;
989 goto bad;
990 }
991
992 /* Transmit packet without fragmentation. */
993 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
994 struct in6_ifaddr *ia6;
995
996 ip6 = mtod(m, struct ip6_hdr *);
997 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
998 if (ia6) {
999 /* Record statistics for this interface address. */
1000 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1001 counter_u64_add(ia6->ia_ifa.ifa_obytes,
1002 m->m_pkthdr.len);
1003 ifa_free(&ia6->ia_ifa);
1004 }
1005 error = nd6_output_ifp(ifp, origifp, m, dst,
1006 (struct route *)ro);
1007 goto done;
1008 }
1009
1010 /* Try to fragment the packet. Cases 1-b and 3. */
1011 if (mtu < IPV6_MMTU) {
1012 /* Path MTU cannot be less than IPV6_MMTU. */
1013 error = EMSGSIZE;
1014 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1015 goto bad;
1016 } else if (ip6->ip6_plen == 0) {
1017 /* Jumbo payload cannot be fragmented. */
1018 error = EMSGSIZE;
1019 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1020 goto bad;
1021 } else {
1022 u_char nextproto;
1023
1024 /*
1025 * Too large for the destination or interface;
1026 * fragment if possible.
1027 * Must be able to put at least 8 bytes per fragment.
1028 */
1029 if (mtu > IPV6_MAXPACKET)
1030 mtu = IPV6_MAXPACKET;
1031
1032 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1033 if (len < 8) {
1034 error = EMSGSIZE;
1035 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1036 goto bad;
1037 }
1038
1039 /*
1040 * If the interface will not calculate checksums on
1041 * fragmented packets, then do it here.
1042 * XXX-BZ handle the hw offloading case. Need flags.
1043 */
1044 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1045 plen, optlen, true);
1046 if (error != 0)
1047 goto bad;
1048
1049 /*
1050 * Change the next header field of the last header in the
1051 * unfragmentable part.
1052 */
1053 if (exthdrs.ip6e_rthdr) {
1054 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1055 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1056 } else if (exthdrs.ip6e_dest1) {
1057 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1058 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1059 } else if (exthdrs.ip6e_hbh) {
1060 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1061 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1062 } else {
1063 ip6 = mtod(m, struct ip6_hdr *);
1064 nextproto = ip6->ip6_nxt;
1065 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1066 }
1067
1068 /*
1069 * Loop through length of segment after first fragment,
1070 * make new header and copy data of each part and link onto
1071 * chain.
1072 */
1073 m0 = m;
1074 id = htonl(ip6_randomid());
1075 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1076 if (error != 0)
1077 goto sendorfree;
1078
1079 in6_ifstat_inc(ifp, ifs6_out_fragok);
1080 }
1081
1082 /* Remove leading garbage. */
1083 sendorfree:
1084 m = m0->m_nextpkt;
1085 m0->m_nextpkt = 0;
1086 m_freem(m0);
1087 for (; m; m = m0) {
1088 m0 = m->m_nextpkt;
1089 m->m_nextpkt = 0;
1090 if (error == 0) {
1091 /* Record statistics for this interface address. */
1092 if (ia) {
1093 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1094 counter_u64_add(ia->ia_ifa.ifa_obytes,
1095 m->m_pkthdr.len);
1096 }
1097 error = nd6_output_ifp(ifp, origifp, m, dst,
1098 (struct route *)ro);
1099 } else
1100 m_freem(m);
1101 }
1102
1103 if (error == 0)
1104 IP6STAT_INC(ip6s_fragmented);
1105
1106 done:
1107 /*
1108 * Release the route if using our private route, or if
1109 * (with flowtable) we don't have our own reference.
1110 */
1111 if (ro == &ip6route ||
1112 (ro != NULL && ro->ro_flags & RT_NORTREF))
1113 RO_RTFREE(ro);
1114 return (error);
1115
1116 freehdrs:
1117 m_freem(exthdrs.ip6e_hbh); /* m_freem() checks if mbuf is NULL. */
1118 m_freem(exthdrs.ip6e_dest1);
1119 m_freem(exthdrs.ip6e_rthdr);
1120 m_freem(exthdrs.ip6e_dest2);
1121 /* FALLTHROUGH */
1122 bad:
1123 if (m)
1124 m_freem(m);
1125 goto done;
1126 }
1127
1128 static int
1129 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1130 {
1131 struct mbuf *m;
1132
1133 if (hlen > MCLBYTES)
1134 return (ENOBUFS); /* XXX */
1135
1136 if (hlen > MLEN)
1137 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1138 else
1139 m = m_get(M_NOWAIT, MT_DATA);
1140 if (m == NULL)
1141 return (ENOBUFS);
1142 m->m_len = hlen;
1143 if (hdr)
1144 bcopy(hdr, mtod(m, caddr_t), hlen);
1145
1146 *mp = m;
1147 return (0);
1148 }
1149
1150 /*
1151 * Insert jumbo payload option.
1152 */
1153 static int
1154 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1155 {
1156 struct mbuf *mopt;
1157 u_char *optbuf;
1158 u_int32_t v;
1159
1160 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1161
1162 /*
1163 * If there is no hop-by-hop options header, allocate new one.
1164 * If there is one but it doesn't have enough space to store the
1165 * jumbo payload option, allocate a cluster to store the whole options.
1166 * Otherwise, use it to store the options.
1167 */
1168 if (exthdrs->ip6e_hbh == NULL) {
1169 mopt = m_get(M_NOWAIT, MT_DATA);
1170 if (mopt == NULL)
1171 return (ENOBUFS);
1172 mopt->m_len = JUMBOOPTLEN;
1173 optbuf = mtod(mopt, u_char *);
1174 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1175 exthdrs->ip6e_hbh = mopt;
1176 } else {
1177 struct ip6_hbh *hbh;
1178
1179 mopt = exthdrs->ip6e_hbh;
1180 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1181 /*
1182 * XXX assumption:
1183 * - exthdrs->ip6e_hbh is not referenced from places
1184 * other than exthdrs.
1185 * - exthdrs->ip6e_hbh is not an mbuf chain.
1186 */
1187 int oldoptlen = mopt->m_len;
1188 struct mbuf *n;
1189
1190 /*
1191 * XXX: give up if the whole (new) hbh header does
1192 * not fit even in an mbuf cluster.
1193 */
1194 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1195 return (ENOBUFS);
1196
1197 /*
1198 * As a consequence, we must always prepare a cluster
1199 * at this point.
1200 */
1201 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1202 if (n == NULL)
1203 return (ENOBUFS);
1204 n->m_len = oldoptlen + JUMBOOPTLEN;
1205 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1206 oldoptlen);
1207 optbuf = mtod(n, caddr_t) + oldoptlen;
1208 m_freem(mopt);
1209 mopt = exthdrs->ip6e_hbh = n;
1210 } else {
1211 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1212 mopt->m_len += JUMBOOPTLEN;
1213 }
1214 optbuf[0] = IP6OPT_PADN;
1215 optbuf[1] = 1;
1216
1217 /*
1218 * Adjust the header length according to the pad and
1219 * the jumbo payload option.
1220 */
1221 hbh = mtod(mopt, struct ip6_hbh *);
1222 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1223 }
1224
1225 /* fill in the option. */
1226 optbuf[2] = IP6OPT_JUMBO;
1227 optbuf[3] = 4;
1228 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1229 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1230
1231 /* finally, adjust the packet header length */
1232 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1233
1234 return (0);
1235 #undef JUMBOOPTLEN
1236 }
1237
1238 /*
1239 * Insert fragment header and copy unfragmentable header portions.
1240 */
1241 static int
1242 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1243 struct ip6_frag **frghdrp)
1244 {
1245 struct mbuf *n, *mlast;
1246
1247 if (hlen > sizeof(struct ip6_hdr)) {
1248 n = m_copym(m0, sizeof(struct ip6_hdr),
1249 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1250 if (n == NULL)
1251 return (ENOBUFS);
1252 m->m_next = n;
1253 } else
1254 n = m;
1255
1256 /* Search for the last mbuf of unfragmentable part. */
1257 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1258 ;
1259
1260 if (M_WRITABLE(mlast) &&
1261 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1262 /* use the trailing space of the last mbuf for the fragment hdr */
1263 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1264 mlast->m_len);
1265 mlast->m_len += sizeof(struct ip6_frag);
1266 m->m_pkthdr.len += sizeof(struct ip6_frag);
1267 } else {
1268 /* allocate a new mbuf for the fragment header */
1269 struct mbuf *mfrg;
1270
1271 mfrg = m_get(M_NOWAIT, MT_DATA);
1272 if (mfrg == NULL)
1273 return (ENOBUFS);
1274 mfrg->m_len = sizeof(struct ip6_frag);
1275 *frghdrp = mtod(mfrg, struct ip6_frag *);
1276 mlast->m_next = mfrg;
1277 }
1278
1279 return (0);
1280 }
1281
1282 /*
1283 * Calculates IPv6 path mtu for destination @dst.
1284 * Resulting MTU is stored in @mtup.
1285 *
1286 * Returns 0 on success.
1287 */
1288 static int
1289 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1290 {
1291 struct nhop6_extended nh6;
1292 struct in6_addr kdst;
1293 uint32_t scopeid;
1294 struct ifnet *ifp;
1295 u_long mtu;
1296 int error;
1297
1298 in6_splitscope(dst, &kdst, &scopeid);
1299 if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1300 return (EHOSTUNREACH);
1301
1302 ifp = nh6.nh_ifp;
1303 mtu = nh6.nh_mtu;
1304
1305 error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1306 fib6_free_nh_ext(fibnum, &nh6);
1307
1308 return (error);
1309 }
1310
1311 /*
1312 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1313 * and cached data in @ro_pmtu.
1314 * MTU from (successful) route lookup is saved (along with dst)
1315 * inside @ro_pmtu to avoid subsequent route lookups after packet
1316 * filter processing.
1317 *
1318 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1319 * Returns 0 on success.
1320 */
1321 static int
1322 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1323 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1324 int *alwaysfragp, u_int fibnum, u_int proto)
1325 {
1326 struct nhop6_basic nh6;
1327 struct in6_addr kdst;
1328 uint32_t scopeid;
1329 struct sockaddr_in6 *sa6_dst;
1330 u_long mtu;
1331
1332 mtu = 0;
1333 if (do_lookup) {
1334
1335 /*
1336 * Here ro_pmtu has final destination address, while
1337 * ro might represent immediate destination.
1338 * Use ro_pmtu destination since mtu might differ.
1339 */
1340 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1341 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1342 ro_pmtu->ro_mtu = 0;
1343
1344 if (ro_pmtu->ro_mtu == 0) {
1345 bzero(sa6_dst, sizeof(*sa6_dst));
1346 sa6_dst->sin6_family = AF_INET6;
1347 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1348 sa6_dst->sin6_addr = *dst;
1349
1350 in6_splitscope(dst, &kdst, &scopeid);
1351 if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1352 &nh6) == 0)
1353 ro_pmtu->ro_mtu = nh6.nh_mtu;
1354 }
1355
1356 mtu = ro_pmtu->ro_mtu;
1357 }
1358
1359 if (ro_pmtu->ro_rt)
1360 mtu = ro_pmtu->ro_rt->rt_mtu;
1361
1362 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1363 }
1364
1365 /*
1366 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1367 * hostcache data for @dst.
1368 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1369 *
1370 * Returns 0 on success.
1371 */
1372 static int
1373 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1374 u_long *mtup, int *alwaysfragp, u_int proto)
1375 {
1376 u_long mtu = 0;
1377 int alwaysfrag = 0;
1378 int error = 0;
1379
1380 if (rt_mtu > 0) {
1381 u_int32_t ifmtu;
1382 struct in_conninfo inc;
1383
1384 bzero(&inc, sizeof(inc));
1385 inc.inc_flags |= INC_ISIPV6;
1386 inc.inc6_faddr = *dst;
1387
1388 ifmtu = IN6_LINKMTU(ifp);
1389
1390 /* TCP is known to react to pmtu changes so skip hc */
1391 if (proto != IPPROTO_TCP)
1392 mtu = tcp_hc_getmtu(&inc);
1393
1394 if (mtu)
1395 mtu = min(mtu, rt_mtu);
1396 else
1397 mtu = rt_mtu;
1398 if (mtu == 0)
1399 mtu = ifmtu;
1400 else if (mtu < IPV6_MMTU) {
1401 /*
1402 * RFC2460 section 5, last paragraph:
1403 * if we record ICMPv6 too big message with
1404 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1405 * or smaller, with framgent header attached.
1406 * (fragment header is needed regardless from the
1407 * packet size, for translators to identify packets)
1408 */
1409 alwaysfrag = 1;
1410 mtu = IPV6_MMTU;
1411 }
1412 } else if (ifp) {
1413 mtu = IN6_LINKMTU(ifp);
1414 } else
1415 error = EHOSTUNREACH; /* XXX */
1416
1417 *mtup = mtu;
1418 if (alwaysfragp)
1419 *alwaysfragp = alwaysfrag;
1420 return (error);
1421 }
1422
1423 /*
1424 * IP6 socket option processing.
1425 */
1426 int
1427 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1428 {
1429 int optdatalen, uproto;
1430 void *optdata;
1431 struct inpcb *in6p = sotoinpcb(so);
1432 int error, optval;
1433 int level, op, optname;
1434 int optlen;
1435 struct thread *td;
1436 #ifdef RSS
1437 uint32_t rss_bucket;
1438 int retval;
1439 #endif
1440
1441 /*
1442 * Don't use more than a quarter of mbuf clusters. N.B.:
1443 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1444 * on LP64 architectures, so cast to u_long to avoid undefined
1445 * behavior. ILP32 architectures cannot have nmbclusters
1446 * large enough to overflow for other reasons.
1447 */
1448 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1449
1450 level = sopt->sopt_level;
1451 op = sopt->sopt_dir;
1452 optname = sopt->sopt_name;
1453 optlen = sopt->sopt_valsize;
1454 td = sopt->sopt_td;
1455 error = 0;
1456 optval = 0;
1457 uproto = (int)so->so_proto->pr_protocol;
1458
1459 if (level != IPPROTO_IPV6) {
1460 error = EINVAL;
1461
1462 if (sopt->sopt_level == SOL_SOCKET &&
1463 sopt->sopt_dir == SOPT_SET) {
1464 switch (sopt->sopt_name) {
1465 case SO_REUSEADDR:
1466 INP_WLOCK(in6p);
1467 if ((so->so_options & SO_REUSEADDR) != 0)
1468 in6p->inp_flags2 |= INP_REUSEADDR;
1469 else
1470 in6p->inp_flags2 &= ~INP_REUSEADDR;
1471 INP_WUNLOCK(in6p);
1472 error = 0;
1473 break;
1474 case SO_REUSEPORT:
1475 INP_WLOCK(in6p);
1476 if ((so->so_options & SO_REUSEPORT) != 0)
1477 in6p->inp_flags2 |= INP_REUSEPORT;
1478 else
1479 in6p->inp_flags2 &= ~INP_REUSEPORT;
1480 INP_WUNLOCK(in6p);
1481 error = 0;
1482 break;
1483 case SO_SETFIB:
1484 INP_WLOCK(in6p);
1485 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1486 INP_WUNLOCK(in6p);
1487 error = 0;
1488 break;
1489 default:
1490 break;
1491 }
1492 }
1493 } else { /* level == IPPROTO_IPV6 */
1494 switch (op) {
1495
1496 case SOPT_SET:
1497 switch (optname) {
1498 case IPV6_2292PKTOPTIONS:
1499 #ifdef IPV6_PKTOPTIONS
1500 case IPV6_PKTOPTIONS:
1501 #endif
1502 {
1503 struct mbuf *m;
1504
1505 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1506 printf("ip6_ctloutput: mbuf limit hit\n");
1507 error = ENOBUFS;
1508 break;
1509 }
1510
1511 error = soopt_getm(sopt, &m); /* XXX */
1512 if (error != 0)
1513 break;
1514 error = soopt_mcopyin(sopt, m); /* XXX */
1515 if (error != 0)
1516 break;
1517 INP_WLOCK(in6p);
1518 error = ip6_pcbopts(&in6p->in6p_outputopts,
1519 m, so, sopt);
1520 INP_WUNLOCK(in6p);
1521 m_freem(m); /* XXX */
1522 break;
1523 }
1524
1525 /*
1526 * Use of some Hop-by-Hop options or some
1527 * Destination options, might require special
1528 * privilege. That is, normal applications
1529 * (without special privilege) might be forbidden
1530 * from setting certain options in outgoing packets,
1531 * and might never see certain options in received
1532 * packets. [RFC 2292 Section 6]
1533 * KAME specific note:
1534 * KAME prevents non-privileged users from sending or
1535 * receiving ANY hbh/dst options in order to avoid
1536 * overhead of parsing options in the kernel.
1537 */
1538 case IPV6_RECVHOPOPTS:
1539 case IPV6_RECVDSTOPTS:
1540 case IPV6_RECVRTHDRDSTOPTS:
1541 if (td != NULL) {
1542 error = priv_check(td,
1543 PRIV_NETINET_SETHDROPTS);
1544 if (error)
1545 break;
1546 }
1547 /* FALLTHROUGH */
1548 case IPV6_UNICAST_HOPS:
1549 case IPV6_HOPLIMIT:
1550
1551 case IPV6_RECVPKTINFO:
1552 case IPV6_RECVHOPLIMIT:
1553 case IPV6_RECVRTHDR:
1554 case IPV6_RECVPATHMTU:
1555 case IPV6_RECVTCLASS:
1556 case IPV6_RECVFLOWID:
1557 #ifdef RSS
1558 case IPV6_RECVRSSBUCKETID:
1559 #endif
1560 case IPV6_V6ONLY:
1561 case IPV6_AUTOFLOWLABEL:
1562 case IPV6_BINDANY:
1563 case IPV6_BINDMULTI:
1564 #ifdef RSS
1565 case IPV6_RSS_LISTEN_BUCKET:
1566 #endif
1567 if (optname == IPV6_BINDANY && td != NULL) {
1568 error = priv_check(td,
1569 PRIV_NETINET_BINDANY);
1570 if (error)
1571 break;
1572 }
1573
1574 if (optlen != sizeof(int)) {
1575 error = EINVAL;
1576 break;
1577 }
1578 error = sooptcopyin(sopt, &optval,
1579 sizeof optval, sizeof optval);
1580 if (error)
1581 break;
1582 switch (optname) {
1583
1584 case IPV6_UNICAST_HOPS:
1585 if (optval < -1 || optval >= 256)
1586 error = EINVAL;
1587 else {
1588 /* -1 = kernel default */
1589 in6p->in6p_hops = optval;
1590 if ((in6p->inp_vflag &
1591 INP_IPV4) != 0)
1592 in6p->inp_ip_ttl = optval;
1593 }
1594 break;
1595 #define OPTSET(bit) \
1596 do { \
1597 INP_WLOCK(in6p); \
1598 if (optval) \
1599 in6p->inp_flags |= (bit); \
1600 else \
1601 in6p->inp_flags &= ~(bit); \
1602 INP_WUNLOCK(in6p); \
1603 } while (/*CONSTCOND*/ 0)
1604 #define OPTSET2292(bit) \
1605 do { \
1606 INP_WLOCK(in6p); \
1607 in6p->inp_flags |= IN6P_RFC2292; \
1608 if (optval) \
1609 in6p->inp_flags |= (bit); \
1610 else \
1611 in6p->inp_flags &= ~(bit); \
1612 INP_WUNLOCK(in6p); \
1613 } while (/*CONSTCOND*/ 0)
1614 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1615
1616 #define OPTSET2(bit, val) do { \
1617 INP_WLOCK(in6p); \
1618 if (val) \
1619 in6p->inp_flags2 |= bit; \
1620 else \
1621 in6p->inp_flags2 &= ~bit; \
1622 INP_WUNLOCK(in6p); \
1623 } while (0)
1624 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1625
1626 case IPV6_RECVPKTINFO:
1627 /* cannot mix with RFC2292 */
1628 if (OPTBIT(IN6P_RFC2292)) {
1629 error = EINVAL;
1630 break;
1631 }
1632 OPTSET(IN6P_PKTINFO);
1633 break;
1634
1635 case IPV6_HOPLIMIT:
1636 {
1637 struct ip6_pktopts **optp;
1638
1639 /* cannot mix with RFC2292 */
1640 if (OPTBIT(IN6P_RFC2292)) {
1641 error = EINVAL;
1642 break;
1643 }
1644 optp = &in6p->in6p_outputopts;
1645 error = ip6_pcbopt(IPV6_HOPLIMIT,
1646 (u_char *)&optval, sizeof(optval),
1647 optp, (td != NULL) ? td->td_ucred :
1648 NULL, uproto);
1649 break;
1650 }
1651
1652 case IPV6_RECVHOPLIMIT:
1653 /* cannot mix with RFC2292 */
1654 if (OPTBIT(IN6P_RFC2292)) {
1655 error = EINVAL;
1656 break;
1657 }
1658 OPTSET(IN6P_HOPLIMIT);
1659 break;
1660
1661 case IPV6_RECVHOPOPTS:
1662 /* cannot mix with RFC2292 */
1663 if (OPTBIT(IN6P_RFC2292)) {
1664 error = EINVAL;
1665 break;
1666 }
1667 OPTSET(IN6P_HOPOPTS);
1668 break;
1669
1670 case IPV6_RECVDSTOPTS:
1671 /* cannot mix with RFC2292 */
1672 if (OPTBIT(IN6P_RFC2292)) {
1673 error = EINVAL;
1674 break;
1675 }
1676 OPTSET(IN6P_DSTOPTS);
1677 break;
1678
1679 case IPV6_RECVRTHDRDSTOPTS:
1680 /* cannot mix with RFC2292 */
1681 if (OPTBIT(IN6P_RFC2292)) {
1682 error = EINVAL;
1683 break;
1684 }
1685 OPTSET(IN6P_RTHDRDSTOPTS);
1686 break;
1687
1688 case IPV6_RECVRTHDR:
1689 /* cannot mix with RFC2292 */
1690 if (OPTBIT(IN6P_RFC2292)) {
1691 error = EINVAL;
1692 break;
1693 }
1694 OPTSET(IN6P_RTHDR);
1695 break;
1696
1697 case IPV6_RECVPATHMTU:
1698 /*
1699 * We ignore this option for TCP
1700 * sockets.
1701 * (RFC3542 leaves this case
1702 * unspecified.)
1703 */
1704 if (uproto != IPPROTO_TCP)
1705 OPTSET(IN6P_MTU);
1706 break;
1707
1708 case IPV6_RECVFLOWID:
1709 OPTSET2(INP_RECVFLOWID, optval);
1710 break;
1711
1712 #ifdef RSS
1713 case IPV6_RECVRSSBUCKETID:
1714 OPTSET2(INP_RECVRSSBUCKETID, optval);
1715 break;
1716 #endif
1717
1718 case IPV6_V6ONLY:
1719 /*
1720 * make setsockopt(IPV6_V6ONLY)
1721 * available only prior to bind(2).
1722 * see ipng mailing list, Jun 22 2001.
1723 */
1724 if (in6p->inp_lport ||
1725 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1726 error = EINVAL;
1727 break;
1728 }
1729 OPTSET(IN6P_IPV6_V6ONLY);
1730 if (optval)
1731 in6p->inp_vflag &= ~INP_IPV4;
1732 else
1733 in6p->inp_vflag |= INP_IPV4;
1734 break;
1735 case IPV6_RECVTCLASS:
1736 /* cannot mix with RFC2292 XXX */
1737 if (OPTBIT(IN6P_RFC2292)) {
1738 error = EINVAL;
1739 break;
1740 }
1741 OPTSET(IN6P_TCLASS);
1742 break;
1743 case IPV6_AUTOFLOWLABEL:
1744 OPTSET(IN6P_AUTOFLOWLABEL);
1745 break;
1746
1747 case IPV6_BINDANY:
1748 OPTSET(INP_BINDANY);
1749 break;
1750
1751 case IPV6_BINDMULTI:
1752 OPTSET2(INP_BINDMULTI, optval);
1753 break;
1754 #ifdef RSS
1755 case IPV6_RSS_LISTEN_BUCKET:
1756 if ((optval >= 0) &&
1757 (optval < rss_getnumbuckets())) {
1758 in6p->inp_rss_listen_bucket = optval;
1759 OPTSET2(INP_RSS_BUCKET_SET, 1);
1760 } else {
1761 error = EINVAL;
1762 }
1763 break;
1764 #endif
1765 }
1766 break;
1767
1768 case IPV6_TCLASS:
1769 case IPV6_DONTFRAG:
1770 case IPV6_USE_MIN_MTU:
1771 case IPV6_PREFER_TEMPADDR:
1772 if (optlen != sizeof(optval)) {
1773 error = EINVAL;
1774 break;
1775 }
1776 error = sooptcopyin(sopt, &optval,
1777 sizeof optval, sizeof optval);
1778 if (error)
1779 break;
1780 {
1781 struct ip6_pktopts **optp;
1782 optp = &in6p->in6p_outputopts;
1783 error = ip6_pcbopt(optname,
1784 (u_char *)&optval, sizeof(optval),
1785 optp, (td != NULL) ? td->td_ucred :
1786 NULL, uproto);
1787 break;
1788 }
1789
1790 case IPV6_2292PKTINFO:
1791 case IPV6_2292HOPLIMIT:
1792 case IPV6_2292HOPOPTS:
1793 case IPV6_2292DSTOPTS:
1794 case IPV6_2292RTHDR:
1795 /* RFC 2292 */
1796 if (optlen != sizeof(int)) {
1797 error = EINVAL;
1798 break;
1799 }
1800 error = sooptcopyin(sopt, &optval,
1801 sizeof optval, sizeof optval);
1802 if (error)
1803 break;
1804 switch (optname) {
1805 case IPV6_2292PKTINFO:
1806 OPTSET2292(IN6P_PKTINFO);
1807 break;
1808 case IPV6_2292HOPLIMIT:
1809 OPTSET2292(IN6P_HOPLIMIT);
1810 break;
1811 case IPV6_2292HOPOPTS:
1812 /*
1813 * Check super-user privilege.
1814 * See comments for IPV6_RECVHOPOPTS.
1815 */
1816 if (td != NULL) {
1817 error = priv_check(td,
1818 PRIV_NETINET_SETHDROPTS);
1819 if (error)
1820 return (error);
1821 }
1822 OPTSET2292(IN6P_HOPOPTS);
1823 break;
1824 case IPV6_2292DSTOPTS:
1825 if (td != NULL) {
1826 error = priv_check(td,
1827 PRIV_NETINET_SETHDROPTS);
1828 if (error)
1829 return (error);
1830 }
1831 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1832 break;
1833 case IPV6_2292RTHDR:
1834 OPTSET2292(IN6P_RTHDR);
1835 break;
1836 }
1837 break;
1838 case IPV6_PKTINFO:
1839 case IPV6_HOPOPTS:
1840 case IPV6_RTHDR:
1841 case IPV6_DSTOPTS:
1842 case IPV6_RTHDRDSTOPTS:
1843 case IPV6_NEXTHOP:
1844 {
1845 /* new advanced API (RFC3542) */
1846 u_char *optbuf;
1847 u_char optbuf_storage[MCLBYTES];
1848 int optlen;
1849 struct ip6_pktopts **optp;
1850
1851 /* cannot mix with RFC2292 */
1852 if (OPTBIT(IN6P_RFC2292)) {
1853 error = EINVAL;
1854 break;
1855 }
1856
1857 /*
1858 * We only ensure valsize is not too large
1859 * here. Further validation will be done
1860 * later.
1861 */
1862 error = sooptcopyin(sopt, optbuf_storage,
1863 sizeof(optbuf_storage), 0);
1864 if (error)
1865 break;
1866 optlen = sopt->sopt_valsize;
1867 optbuf = optbuf_storage;
1868 optp = &in6p->in6p_outputopts;
1869 error = ip6_pcbopt(optname, optbuf, optlen,
1870 optp, (td != NULL) ? td->td_ucred : NULL,
1871 uproto);
1872 break;
1873 }
1874 #undef OPTSET
1875
1876 case IPV6_MULTICAST_IF:
1877 case IPV6_MULTICAST_HOPS:
1878 case IPV6_MULTICAST_LOOP:
1879 case IPV6_JOIN_GROUP:
1880 case IPV6_LEAVE_GROUP:
1881 case IPV6_MSFILTER:
1882 case MCAST_BLOCK_SOURCE:
1883 case MCAST_UNBLOCK_SOURCE:
1884 case MCAST_JOIN_GROUP:
1885 case MCAST_LEAVE_GROUP:
1886 case MCAST_JOIN_SOURCE_GROUP:
1887 case MCAST_LEAVE_SOURCE_GROUP:
1888 error = ip6_setmoptions(in6p, sopt);
1889 break;
1890
1891 case IPV6_PORTRANGE:
1892 error = sooptcopyin(sopt, &optval,
1893 sizeof optval, sizeof optval);
1894 if (error)
1895 break;
1896
1897 INP_WLOCK(in6p);
1898 switch (optval) {
1899 case IPV6_PORTRANGE_DEFAULT:
1900 in6p->inp_flags &= ~(INP_LOWPORT);
1901 in6p->inp_flags &= ~(INP_HIGHPORT);
1902 break;
1903
1904 case IPV6_PORTRANGE_HIGH:
1905 in6p->inp_flags &= ~(INP_LOWPORT);
1906 in6p->inp_flags |= INP_HIGHPORT;
1907 break;
1908
1909 case IPV6_PORTRANGE_LOW:
1910 in6p->inp_flags &= ~(INP_HIGHPORT);
1911 in6p->inp_flags |= INP_LOWPORT;
1912 break;
1913
1914 default:
1915 error = EINVAL;
1916 break;
1917 }
1918 INP_WUNLOCK(in6p);
1919 break;
1920
1921 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1922 case IPV6_IPSEC_POLICY:
1923 if (IPSEC_ENABLED(ipv6)) {
1924 error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1925 break;
1926 }
1927 /* FALLTHROUGH */
1928 #endif /* IPSEC */
1929
1930 default:
1931 error = ENOPROTOOPT;
1932 break;
1933 }
1934 break;
1935
1936 case SOPT_GET:
1937 switch (optname) {
1938
1939 case IPV6_2292PKTOPTIONS:
1940 #ifdef IPV6_PKTOPTIONS
1941 case IPV6_PKTOPTIONS:
1942 #endif
1943 /*
1944 * RFC3542 (effectively) deprecated the
1945 * semantics of the 2292-style pktoptions.
1946 * Since it was not reliable in nature (i.e.,
1947 * applications had to expect the lack of some
1948 * information after all), it would make sense
1949 * to simplify this part by always returning
1950 * empty data.
1951 */
1952 sopt->sopt_valsize = 0;
1953 break;
1954
1955 case IPV6_RECVHOPOPTS:
1956 case IPV6_RECVDSTOPTS:
1957 case IPV6_RECVRTHDRDSTOPTS:
1958 case IPV6_UNICAST_HOPS:
1959 case IPV6_RECVPKTINFO:
1960 case IPV6_RECVHOPLIMIT:
1961 case IPV6_RECVRTHDR:
1962 case IPV6_RECVPATHMTU:
1963
1964 case IPV6_V6ONLY:
1965 case IPV6_PORTRANGE:
1966 case IPV6_RECVTCLASS:
1967 case IPV6_AUTOFLOWLABEL:
1968 case IPV6_BINDANY:
1969 case IPV6_FLOWID:
1970 case IPV6_FLOWTYPE:
1971 case IPV6_RECVFLOWID:
1972 #ifdef RSS
1973 case IPV6_RSSBUCKETID:
1974 case IPV6_RECVRSSBUCKETID:
1975 #endif
1976 case IPV6_BINDMULTI:
1977 switch (optname) {
1978
1979 case IPV6_RECVHOPOPTS:
1980 optval = OPTBIT(IN6P_HOPOPTS);
1981 break;
1982
1983 case IPV6_RECVDSTOPTS:
1984 optval = OPTBIT(IN6P_DSTOPTS);
1985 break;
1986
1987 case IPV6_RECVRTHDRDSTOPTS:
1988 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1989 break;
1990
1991 case IPV6_UNICAST_HOPS:
1992 optval = in6p->in6p_hops;
1993 break;
1994
1995 case IPV6_RECVPKTINFO:
1996 optval = OPTBIT(IN6P_PKTINFO);
1997 break;
1998
1999 case IPV6_RECVHOPLIMIT:
2000 optval = OPTBIT(IN6P_HOPLIMIT);
2001 break;
2002
2003 case IPV6_RECVRTHDR:
2004 optval = OPTBIT(IN6P_RTHDR);
2005 break;
2006
2007 case IPV6_RECVPATHMTU:
2008 optval = OPTBIT(IN6P_MTU);
2009 break;
2010
2011 case IPV6_V6ONLY:
2012 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2013 break;
2014
2015 case IPV6_PORTRANGE:
2016 {
2017 int flags;
2018 flags = in6p->inp_flags;
2019 if (flags & INP_HIGHPORT)
2020 optval = IPV6_PORTRANGE_HIGH;
2021 else if (flags & INP_LOWPORT)
2022 optval = IPV6_PORTRANGE_LOW;
2023 else
2024 optval = 0;
2025 break;
2026 }
2027 case IPV6_RECVTCLASS:
2028 optval = OPTBIT(IN6P_TCLASS);
2029 break;
2030
2031 case IPV6_AUTOFLOWLABEL:
2032 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2033 break;
2034
2035 case IPV6_BINDANY:
2036 optval = OPTBIT(INP_BINDANY);
2037 break;
2038
2039 case IPV6_FLOWID:
2040 optval = in6p->inp_flowid;
2041 break;
2042
2043 case IPV6_FLOWTYPE:
2044 optval = in6p->inp_flowtype;
2045 break;
2046
2047 case IPV6_RECVFLOWID:
2048 optval = OPTBIT2(INP_RECVFLOWID);
2049 break;
2050 #ifdef RSS
2051 case IPV6_RSSBUCKETID:
2052 retval =
2053 rss_hash2bucket(in6p->inp_flowid,
2054 in6p->inp_flowtype,
2055 &rss_bucket);
2056 if (retval == 0)
2057 optval = rss_bucket;
2058 else
2059 error = EINVAL;
2060 break;
2061
2062 case IPV6_RECVRSSBUCKETID:
2063 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2064 break;
2065 #endif
2066
2067 case IPV6_BINDMULTI:
2068 optval = OPTBIT2(INP_BINDMULTI);
2069 break;
2070
2071 }
2072 if (error)
2073 break;
2074 error = sooptcopyout(sopt, &optval,
2075 sizeof optval);
2076 break;
2077
2078 case IPV6_PATHMTU:
2079 {
2080 u_long pmtu = 0;
2081 struct ip6_mtuinfo mtuinfo;
2082
2083 if (!(so->so_state & SS_ISCONNECTED))
2084 return (ENOTCONN);
2085 /*
2086 * XXX: we dot not consider the case of source
2087 * routing, or optional information to specify
2088 * the outgoing interface.
2089 */
2090 error = ip6_getpmtu_ctl(so->so_fibnum,
2091 &in6p->in6p_faddr, &pmtu);
2092 if (error)
2093 break;
2094 if (pmtu > IPV6_MAXPACKET)
2095 pmtu = IPV6_MAXPACKET;
2096
2097 bzero(&mtuinfo, sizeof(mtuinfo));
2098 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2099 optdata = (void *)&mtuinfo;
2100 optdatalen = sizeof(mtuinfo);
2101 error = sooptcopyout(sopt, optdata,
2102 optdatalen);
2103 break;
2104 }
2105
2106 case IPV6_2292PKTINFO:
2107 case IPV6_2292HOPLIMIT:
2108 case IPV6_2292HOPOPTS:
2109 case IPV6_2292RTHDR:
2110 case IPV6_2292DSTOPTS:
2111 switch (optname) {
2112 case IPV6_2292PKTINFO:
2113 optval = OPTBIT(IN6P_PKTINFO);
2114 break;
2115 case IPV6_2292HOPLIMIT:
2116 optval = OPTBIT(IN6P_HOPLIMIT);
2117 break;
2118 case IPV6_2292HOPOPTS:
2119 optval = OPTBIT(IN6P_HOPOPTS);
2120 break;
2121 case IPV6_2292RTHDR:
2122 optval = OPTBIT(IN6P_RTHDR);
2123 break;
2124 case IPV6_2292DSTOPTS:
2125 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2126 break;
2127 }
2128 error = sooptcopyout(sopt, &optval,
2129 sizeof optval);
2130 break;
2131 case IPV6_PKTINFO:
2132 case IPV6_HOPOPTS:
2133 case IPV6_RTHDR:
2134 case IPV6_DSTOPTS:
2135 case IPV6_RTHDRDSTOPTS:
2136 case IPV6_NEXTHOP:
2137 case IPV6_TCLASS:
2138 case IPV6_DONTFRAG:
2139 case IPV6_USE_MIN_MTU:
2140 case IPV6_PREFER_TEMPADDR:
2141 error = ip6_getpcbopt(in6p->in6p_outputopts,
2142 optname, sopt);
2143 break;
2144
2145 case IPV6_MULTICAST_IF:
2146 case IPV6_MULTICAST_HOPS:
2147 case IPV6_MULTICAST_LOOP:
2148 case IPV6_MSFILTER:
2149 error = ip6_getmoptions(in6p, sopt);
2150 break;
2151
2152 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2153 case IPV6_IPSEC_POLICY:
2154 if (IPSEC_ENABLED(ipv6)) {
2155 error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2156 break;
2157 }
2158 /* FALLTHROUGH */
2159 #endif /* IPSEC */
2160 default:
2161 error = ENOPROTOOPT;
2162 break;
2163 }
2164 break;
2165 }
2166 }
2167 return (error);
2168 }
2169
2170 int
2171 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2172 {
2173 int error = 0, optval, optlen;
2174 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2175 struct inpcb *in6p = sotoinpcb(so);
2176 int level, op, optname;
2177
2178 level = sopt->sopt_level;
2179 op = sopt->sopt_dir;
2180 optname = sopt->sopt_name;
2181 optlen = sopt->sopt_valsize;
2182
2183 if (level != IPPROTO_IPV6) {
2184 return (EINVAL);
2185 }
2186
2187 switch (optname) {
2188 case IPV6_CHECKSUM:
2189 /*
2190 * For ICMPv6 sockets, no modification allowed for checksum
2191 * offset, permit "no change" values to help existing apps.
2192 *
2193 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2194 * for an ICMPv6 socket will fail."
2195 * The current behavior does not meet RFC3542.
2196 */
2197 switch (op) {
2198 case SOPT_SET:
2199 if (optlen != sizeof(int)) {
2200 error = EINVAL;
2201 break;
2202 }
2203 error = sooptcopyin(sopt, &optval, sizeof(optval),
2204 sizeof(optval));
2205 if (error)
2206 break;
2207 if (optval < -1 || (optval % 2) != 0) {
2208 /*
2209 * The API assumes non-negative even offset
2210 * values or -1 as a special value.
2211 */
2212 error = EINVAL;
2213 } else if (so->so_proto->pr_protocol ==
2214 IPPROTO_ICMPV6) {
2215 if (optval != icmp6off)
2216 error = EINVAL;
2217 } else
2218 in6p->in6p_cksum = optval;
2219 break;
2220
2221 case SOPT_GET:
2222 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2223 optval = icmp6off;
2224 else
2225 optval = in6p->in6p_cksum;
2226
2227 error = sooptcopyout(sopt, &optval, sizeof(optval));
2228 break;
2229
2230 default:
2231 error = EINVAL;
2232 break;
2233 }
2234 break;
2235
2236 default:
2237 error = ENOPROTOOPT;
2238 break;
2239 }
2240
2241 return (error);
2242 }
2243
2244 /*
2245 * Set up IP6 options in pcb for insertion in output packets or
2246 * specifying behavior of outgoing packets.
2247 */
2248 static int
2249 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2250 struct socket *so, struct sockopt *sopt)
2251 {
2252 struct ip6_pktopts *opt = *pktopt;
2253 int error = 0;
2254 struct thread *td = sopt->sopt_td;
2255
2256 /* turn off any old options. */
2257 if (opt) {
2258 #ifdef DIAGNOSTIC
2259 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2260 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2261 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2262 printf("ip6_pcbopts: all specified options are cleared.\n");
2263 #endif
2264 ip6_clearpktopts(opt, -1);
2265 } else {
2266 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2267 if (opt == NULL)
2268 return (ENOMEM);
2269 }
2270 *pktopt = NULL;
2271
2272 if (!m || m->m_len == 0) {
2273 /*
2274 * Only turning off any previous options, regardless of
2275 * whether the opt is just created or given.
2276 */
2277 free(opt, M_IP6OPT);
2278 return (0);
2279 }
2280
2281 /* set options specified by user. */
2282 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2283 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2284 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2285 free(opt, M_IP6OPT);
2286 return (error);
2287 }
2288 *pktopt = opt;
2289 return (0);
2290 }
2291
2292 /*
2293 * initialize ip6_pktopts. beware that there are non-zero default values in
2294 * the struct.
2295 */
2296 void
2297 ip6_initpktopts(struct ip6_pktopts *opt)
2298 {
2299
2300 bzero(opt, sizeof(*opt));
2301 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2302 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2303 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2304 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2305 }
2306
2307 static int
2308 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2309 struct ucred *cred, int uproto)
2310 {
2311 struct ip6_pktopts *opt;
2312
2313 if (*pktopt == NULL) {
2314 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2315 M_WAITOK);
2316 ip6_initpktopts(*pktopt);
2317 }
2318 opt = *pktopt;
2319
2320 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2321 }
2322
2323 static int
2324 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2325 {
2326 void *optdata = NULL;
2327 int optdatalen = 0;
2328 struct ip6_ext *ip6e;
2329 int error = 0;
2330 struct in6_pktinfo null_pktinfo;
2331 int deftclass = 0, on;
2332 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2333 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2334
2335 switch (optname) {
2336 case IPV6_PKTINFO:
2337 optdata = (void *)&null_pktinfo;
2338 if (pktopt && pktopt->ip6po_pktinfo) {
2339 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2340 sizeof(null_pktinfo));
2341 in6_clearscope(&null_pktinfo.ipi6_addr);
2342 } else {
2343 /* XXX: we don't have to do this every time... */
2344 bzero(&null_pktinfo, sizeof(null_pktinfo));
2345 }
2346 optdatalen = sizeof(struct in6_pktinfo);
2347 break;
2348 case IPV6_TCLASS:
2349 if (pktopt && pktopt->ip6po_tclass >= 0)
2350 optdata = (void *)&pktopt->ip6po_tclass;
2351 else
2352 optdata = (void *)&deftclass;
2353 optdatalen = sizeof(int);
2354 break;
2355 case IPV6_HOPOPTS:
2356 if (pktopt && pktopt->ip6po_hbh) {
2357 optdata = (void *)pktopt->ip6po_hbh;
2358 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2359 optdatalen = (ip6e->ip6e_len + 1) << 3;
2360 }
2361 break;
2362 case IPV6_RTHDR:
2363 if (pktopt && pktopt->ip6po_rthdr) {
2364 optdata = (void *)pktopt->ip6po_rthdr;
2365 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2366 optdatalen = (ip6e->ip6e_len + 1) << 3;
2367 }
2368 break;
2369 case IPV6_RTHDRDSTOPTS:
2370 if (pktopt && pktopt->ip6po_dest1) {
2371 optdata = (void *)pktopt->ip6po_dest1;
2372 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2373 optdatalen = (ip6e->ip6e_len + 1) << 3;
2374 }
2375 break;
2376 case IPV6_DSTOPTS:
2377 if (pktopt && pktopt->ip6po_dest2) {
2378 optdata = (void *)pktopt->ip6po_dest2;
2379 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2380 optdatalen = (ip6e->ip6e_len + 1) << 3;
2381 }
2382 break;
2383 case IPV6_NEXTHOP:
2384 if (pktopt && pktopt->ip6po_nexthop) {
2385 optdata = (void *)pktopt->ip6po_nexthop;
2386 optdatalen = pktopt->ip6po_nexthop->sa_len;
2387 }
2388 break;
2389 case IPV6_USE_MIN_MTU:
2390 if (pktopt)
2391 optdata = (void *)&pktopt->ip6po_minmtu;
2392 else
2393 optdata = (void *)&defminmtu;
2394 optdatalen = sizeof(int);
2395 break;
2396 case IPV6_DONTFRAG:
2397 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2398 on = 1;
2399 else
2400 on = 0;
2401 optdata = (void *)&on;
2402 optdatalen = sizeof(on);
2403 break;
2404 case IPV6_PREFER_TEMPADDR:
2405 if (pktopt)
2406 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2407 else
2408 optdata = (void *)&defpreftemp;
2409 optdatalen = sizeof(int);
2410 break;
2411 default: /* should not happen */
2412 #ifdef DIAGNOSTIC
2413 panic("ip6_getpcbopt: unexpected option\n");
2414 #endif
2415 return (ENOPROTOOPT);
2416 }
2417
2418 error = sooptcopyout(sopt, optdata, optdatalen);
2419
2420 return (error);
2421 }
2422
2423 void
2424 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2425 {
2426 if (pktopt == NULL)
2427 return;
2428
2429 if (optname == -1 || optname == IPV6_PKTINFO) {
2430 if (pktopt->ip6po_pktinfo)
2431 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2432 pktopt->ip6po_pktinfo = NULL;
2433 }
2434 if (optname == -1 || optname == IPV6_HOPLIMIT)
2435 pktopt->ip6po_hlim = -1;
2436 if (optname == -1 || optname == IPV6_TCLASS)
2437 pktopt->ip6po_tclass = -1;
2438 if (optname == -1 || optname == IPV6_NEXTHOP) {
2439 if (pktopt->ip6po_nextroute.ro_rt) {
2440 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2441 pktopt->ip6po_nextroute.ro_rt = NULL;
2442 }
2443 if (pktopt->ip6po_nexthop)
2444 free(pktopt->ip6po_nexthop, M_IP6OPT);
2445 pktopt->ip6po_nexthop = NULL;
2446 }
2447 if (optname == -1 || optname == IPV6_HOPOPTS) {
2448 if (pktopt->ip6po_hbh)
2449 free(pktopt->ip6po_hbh, M_IP6OPT);
2450 pktopt->ip6po_hbh = NULL;
2451 }
2452 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2453 if (pktopt->ip6po_dest1)
2454 free(pktopt->ip6po_dest1, M_IP6OPT);
2455 pktopt->ip6po_dest1 = NULL;
2456 }
2457 if (optname == -1 || optname == IPV6_RTHDR) {
2458 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2459 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2460 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2461 if (pktopt->ip6po_route.ro_rt) {
2462 RTFREE(pktopt->ip6po_route.ro_rt);
2463 pktopt->ip6po_route.ro_rt = NULL;
2464 }
2465 }
2466 if (optname == -1 || optname == IPV6_DSTOPTS) {
2467 if (pktopt->ip6po_dest2)
2468 free(pktopt->ip6po_dest2, M_IP6OPT);
2469 pktopt->ip6po_dest2 = NULL;
2470 }
2471 }
2472
2473 #define PKTOPT_EXTHDRCPY(type) \
2474 do {\
2475 if (src->type) {\
2476 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2477 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2478 if (dst->type == NULL)\
2479 goto bad;\
2480 bcopy(src->type, dst->type, hlen);\
2481 }\
2482 } while (/*CONSTCOND*/ 0)
2483
2484 static int
2485 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2486 {
2487 if (dst == NULL || src == NULL) {
2488 printf("ip6_clearpktopts: invalid argument\n");
2489 return (EINVAL);
2490 }
2491
2492 dst->ip6po_hlim = src->ip6po_hlim;
2493 dst->ip6po_tclass = src->ip6po_tclass;
2494 dst->ip6po_flags = src->ip6po_flags;
2495 dst->ip6po_minmtu = src->ip6po_minmtu;
2496 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2497 if (src->ip6po_pktinfo) {
2498 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2499 M_IP6OPT, canwait);
2500 if (dst->ip6po_pktinfo == NULL)
2501 goto bad;
2502 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2503 }
2504 if (src->ip6po_nexthop) {
2505 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2506 M_IP6OPT, canwait);
2507 if (dst->ip6po_nexthop == NULL)
2508 goto bad;
2509 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2510 src->ip6po_nexthop->sa_len);
2511 }
2512 PKTOPT_EXTHDRCPY(ip6po_hbh);
2513 PKTOPT_EXTHDRCPY(ip6po_dest1);
2514 PKTOPT_EXTHDRCPY(ip6po_dest2);
2515 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2516 return (0);
2517
2518 bad:
2519 ip6_clearpktopts(dst, -1);
2520 return (ENOBUFS);
2521 }
2522 #undef PKTOPT_EXTHDRCPY
2523
2524 struct ip6_pktopts *
2525 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2526 {
2527 int error;
2528 struct ip6_pktopts *dst;
2529
2530 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2531 if (dst == NULL)
2532 return (NULL);
2533 ip6_initpktopts(dst);
2534
2535 if ((error = copypktopts(dst, src, canwait)) != 0) {
2536 free(dst, M_IP6OPT);
2537 return (NULL);
2538 }
2539
2540 return (dst);
2541 }
2542
2543 void
2544 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2545 {
2546 if (pktopt == NULL)
2547 return;
2548
2549 ip6_clearpktopts(pktopt, -1);
2550
2551 free(pktopt, M_IP6OPT);
2552 }
2553
2554 /*
2555 * Set IPv6 outgoing packet options based on advanced API.
2556 */
2557 int
2558 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2559 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2560 {
2561 struct cmsghdr *cm = NULL;
2562
2563 if (control == NULL || opt == NULL)
2564 return (EINVAL);
2565
2566 ip6_initpktopts(opt);
2567 if (stickyopt) {
2568 int error;
2569
2570 /*
2571 * If stickyopt is provided, make a local copy of the options
2572 * for this particular packet, then override them by ancillary
2573 * objects.
2574 * XXX: copypktopts() does not copy the cached route to a next
2575 * hop (if any). This is not very good in terms of efficiency,
2576 * but we can allow this since this option should be rarely
2577 * used.
2578 */
2579 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2580 return (error);
2581 }
2582
2583 /*
2584 * XXX: Currently, we assume all the optional information is stored
2585 * in a single mbuf.
2586 */
2587 if (control->m_next)
2588 return (EINVAL);
2589
2590 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2591 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2592 int error;
2593
2594 if (control->m_len < CMSG_LEN(0))
2595 return (EINVAL);
2596
2597 cm = mtod(control, struct cmsghdr *);
2598 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2599 return (EINVAL);
2600 if (cm->cmsg_level != IPPROTO_IPV6)
2601 continue;
2602
2603 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2604 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2605 if (error)
2606 return (error);
2607 }
2608
2609 return (0);
2610 }
2611
2612 /*
2613 * Set a particular packet option, as a sticky option or an ancillary data
2614 * item. "len" can be 0 only when it's a sticky option.
2615 * We have 4 cases of combination of "sticky" and "cmsg":
2616 * "sticky=0, cmsg=0": impossible
2617 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2618 * "sticky=1, cmsg=0": RFC3542 socket option
2619 * "sticky=1, cmsg=1": RFC2292 socket option
2620 */
2621 static int
2622 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2623 struct ucred *cred, int sticky, int cmsg, int uproto)
2624 {
2625 int minmtupolicy, preftemp;
2626 int error;
2627
2628 if (!sticky && !cmsg) {
2629 #ifdef DIAGNOSTIC
2630 printf("ip6_setpktopt: impossible case\n");
2631 #endif
2632 return (EINVAL);
2633 }
2634
2635 /*
2636 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2637 * not be specified in the context of RFC3542. Conversely,
2638 * RFC3542 types should not be specified in the context of RFC2292.
2639 */
2640 if (!cmsg) {
2641 switch (optname) {
2642 case IPV6_2292PKTINFO:
2643 case IPV6_2292HOPLIMIT:
2644 case IPV6_2292NEXTHOP:
2645 case IPV6_2292HOPOPTS:
2646 case IPV6_2292DSTOPTS:
2647 case IPV6_2292RTHDR:
2648 case IPV6_2292PKTOPTIONS:
2649 return (ENOPROTOOPT);
2650 }
2651 }
2652 if (sticky && cmsg) {
2653 switch (optname) {
2654 case IPV6_PKTINFO:
2655 case IPV6_HOPLIMIT:
2656 case IPV6_NEXTHOP:
2657 case IPV6_HOPOPTS:
2658 case IPV6_DSTOPTS:
2659 case IPV6_RTHDRDSTOPTS:
2660 case IPV6_RTHDR:
2661 case IPV6_USE_MIN_MTU:
2662 case IPV6_DONTFRAG:
2663 case IPV6_TCLASS:
2664 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2665 return (ENOPROTOOPT);
2666 }
2667 }
2668
2669 switch (optname) {
2670 case IPV6_2292PKTINFO:
2671 case IPV6_PKTINFO:
2672 {
2673 struct ifnet *ifp = NULL;
2674 struct in6_pktinfo *pktinfo;
2675
2676 if (len != sizeof(struct in6_pktinfo))
2677 return (EINVAL);
2678
2679 pktinfo = (struct in6_pktinfo *)buf;
2680
2681 /*
2682 * An application can clear any sticky IPV6_PKTINFO option by
2683 * doing a "regular" setsockopt with ipi6_addr being
2684 * in6addr_any and ipi6_ifindex being zero.
2685 * [RFC 3542, Section 6]
2686 */
2687 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2688 pktinfo->ipi6_ifindex == 0 &&
2689 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2690 ip6_clearpktopts(opt, optname);
2691 break;
2692 }
2693
2694 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2695 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2696 return (EINVAL);
2697 }
2698 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2699 return (EINVAL);
2700 /* validate the interface index if specified. */
2701 if (pktinfo->ipi6_ifindex > V_if_index)
2702 return (ENXIO);
2703 if (pktinfo->ipi6_ifindex) {
2704 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2705 if (ifp == NULL)
2706 return (ENXIO);
2707 }
2708 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2709 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2710 return (ENETDOWN);
2711
2712 if (ifp != NULL &&
2713 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2714 struct in6_ifaddr *ia;
2715
2716 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2717 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2718 if (ia == NULL)
2719 return (EADDRNOTAVAIL);
2720 ifa_free(&ia->ia_ifa);
2721 }
2722 /*
2723 * We store the address anyway, and let in6_selectsrc()
2724 * validate the specified address. This is because ipi6_addr
2725 * may not have enough information about its scope zone, and
2726 * we may need additional information (such as outgoing
2727 * interface or the scope zone of a destination address) to
2728 * disambiguate the scope.
2729 * XXX: the delay of the validation may confuse the
2730 * application when it is used as a sticky option.
2731 */
2732 if (opt->ip6po_pktinfo == NULL) {
2733 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2734 M_IP6OPT, M_NOWAIT);
2735 if (opt->ip6po_pktinfo == NULL)
2736 return (ENOBUFS);
2737 }
2738 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2739 break;
2740 }
2741
2742 case IPV6_2292HOPLIMIT:
2743 case IPV6_HOPLIMIT:
2744 {
2745 int *hlimp;
2746
2747 /*
2748 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2749 * to simplify the ordering among hoplimit options.
2750 */
2751 if (optname == IPV6_HOPLIMIT && sticky)
2752 return (ENOPROTOOPT);
2753
2754 if (len != sizeof(int))
2755 return (EINVAL);
2756 hlimp = (int *)buf;
2757 if (*hlimp < -1 || *hlimp > 255)
2758 return (EINVAL);
2759
2760 opt->ip6po_hlim = *hlimp;
2761 break;
2762 }
2763
2764 case IPV6_TCLASS:
2765 {
2766 int tclass;
2767
2768 if (len != sizeof(int))
2769 return (EINVAL);
2770 tclass = *(int *)buf;
2771 if (tclass < -1 || tclass > 255)
2772 return (EINVAL);
2773
2774 opt->ip6po_tclass = tclass;
2775 break;
2776 }
2777
2778 case IPV6_2292NEXTHOP:
2779 case IPV6_NEXTHOP:
2780 if (cred != NULL) {
2781 error = priv_check_cred(cred,
2782 PRIV_NETINET_SETHDROPTS, 0);
2783 if (error)
2784 return (error);
2785 }
2786
2787 if (len == 0) { /* just remove the option */
2788 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2789 break;
2790 }
2791
2792 /* check if cmsg_len is large enough for sa_len */
2793 if (len < sizeof(struct sockaddr) || len < *buf)
2794 return (EINVAL);
2795
2796 switch (((struct sockaddr *)buf)->sa_family) {
2797 case AF_INET6:
2798 {
2799 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2800 int error;
2801
2802 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2803 return (EINVAL);
2804
2805 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2806 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2807 return (EINVAL);
2808 }
2809 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2810 != 0) {
2811 return (error);
2812 }
2813 break;
2814 }
2815 case AF_LINK: /* should eventually be supported */
2816 default:
2817 return (EAFNOSUPPORT);
2818 }
2819
2820 /* turn off the previous option, then set the new option. */
2821 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2822 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2823 if (opt->ip6po_nexthop == NULL)
2824 return (ENOBUFS);
2825 bcopy(buf, opt->ip6po_nexthop, *buf);
2826 break;
2827
2828 case IPV6_2292HOPOPTS:
2829 case IPV6_HOPOPTS:
2830 {
2831 struct ip6_hbh *hbh;
2832 int hbhlen;
2833
2834 /*
2835 * XXX: We don't allow a non-privileged user to set ANY HbH
2836 * options, since per-option restriction has too much
2837 * overhead.
2838 */
2839 if (cred != NULL) {
2840 error = priv_check_cred(cred,
2841 PRIV_NETINET_SETHDROPTS, 0);
2842 if (error)
2843 return (error);
2844 }
2845
2846 if (len == 0) {
2847 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2848 break; /* just remove the option */
2849 }
2850
2851 /* message length validation */
2852 if (len < sizeof(struct ip6_hbh))
2853 return (EINVAL);
2854 hbh = (struct ip6_hbh *)buf;
2855 hbhlen = (hbh->ip6h_len + 1) << 3;
2856 if (len != hbhlen)
2857 return (EINVAL);
2858
2859 /* turn off the previous option, then set the new option. */
2860 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2861 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2862 if (opt->ip6po_hbh == NULL)
2863 return (ENOBUFS);
2864 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2865
2866 break;
2867 }
2868
2869 case IPV6_2292DSTOPTS:
2870 case IPV6_DSTOPTS:
2871 case IPV6_RTHDRDSTOPTS:
2872 {
2873 struct ip6_dest *dest, **newdest = NULL;
2874 int destlen;
2875
2876 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2877 error = priv_check_cred(cred,
2878 PRIV_NETINET_SETHDROPTS, 0);
2879 if (error)
2880 return (error);
2881 }
2882
2883 if (len == 0) {
2884 ip6_clearpktopts(opt, optname);
2885 break; /* just remove the option */
2886 }
2887
2888 /* message length validation */
2889 if (len < sizeof(struct ip6_dest))
2890 return (EINVAL);
2891 dest = (struct ip6_dest *)buf;
2892 destlen = (dest->ip6d_len + 1) << 3;
2893 if (len != destlen)
2894 return (EINVAL);
2895
2896 /*
2897 * Determine the position that the destination options header
2898 * should be inserted; before or after the routing header.
2899 */
2900 switch (optname) {
2901 case IPV6_2292DSTOPTS:
2902 /*
2903 * The old advacned API is ambiguous on this point.
2904 * Our approach is to determine the position based
2905 * according to the existence of a routing header.
2906 * Note, however, that this depends on the order of the
2907 * extension headers in the ancillary data; the 1st
2908 * part of the destination options header must appear
2909 * before the routing header in the ancillary data,
2910 * too.
2911 * RFC3542 solved the ambiguity by introducing
2912 * separate ancillary data or option types.
2913 */
2914 if (opt->ip6po_rthdr == NULL)
2915 newdest = &opt->ip6po_dest1;
2916 else
2917 newdest = &opt->ip6po_dest2;
2918 break;
2919 case IPV6_RTHDRDSTOPTS:
2920 newdest = &opt->ip6po_dest1;
2921 break;
2922 case IPV6_DSTOPTS:
2923 newdest = &opt->ip6po_dest2;
2924 break;
2925 }
2926
2927 /* turn off the previous option, then set the new option. */
2928 ip6_clearpktopts(opt, optname);
2929 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2930 if (*newdest == NULL)
2931 return (ENOBUFS);
2932 bcopy(dest, *newdest, destlen);
2933
2934 break;
2935 }
2936
2937 case IPV6_2292RTHDR:
2938 case IPV6_RTHDR:
2939 {
2940 struct ip6_rthdr *rth;
2941 int rthlen;
2942
2943 if (len == 0) {
2944 ip6_clearpktopts(opt, IPV6_RTHDR);
2945 break; /* just remove the option */
2946 }
2947
2948 /* message length validation */
2949 if (len < sizeof(struct ip6_rthdr))
2950 return (EINVAL);
2951 rth = (struct ip6_rthdr *)buf;
2952 rthlen = (rth->ip6r_len + 1) << 3;
2953 if (len != rthlen)
2954 return (EINVAL);
2955
2956 switch (rth->ip6r_type) {
2957 case IPV6_RTHDR_TYPE_0:
2958 if (rth->ip6r_len == 0) /* must contain one addr */
2959 return (EINVAL);
2960 if (rth->ip6r_len % 2) /* length must be even */
2961 return (EINVAL);
2962 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2963 return (EINVAL);
2964 break;
2965 default:
2966 return (EINVAL); /* not supported */
2967 }
2968
2969 /* turn off the previous option */
2970 ip6_clearpktopts(opt, IPV6_RTHDR);
2971 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2972 if (opt->ip6po_rthdr == NULL)
2973 return (ENOBUFS);
2974 bcopy(rth, opt->ip6po_rthdr, rthlen);
2975
2976 break;
2977 }
2978
2979 case IPV6_USE_MIN_MTU:
2980 if (len != sizeof(int))
2981 return (EINVAL);
2982 minmtupolicy = *(int *)buf;
2983 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2984 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2985 minmtupolicy != IP6PO_MINMTU_ALL) {
2986 return (EINVAL);
2987 }
2988 opt->ip6po_minmtu = minmtupolicy;
2989 break;
2990
2991 case IPV6_DONTFRAG:
2992 if (len != sizeof(int))
2993 return (EINVAL);
2994
2995 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2996 /*
2997 * we ignore this option for TCP sockets.
2998 * (RFC3542 leaves this case unspecified.)
2999 */
3000 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3001 } else
3002 opt->ip6po_flags |= IP6PO_DONTFRAG;
3003 break;
3004
3005 case IPV6_PREFER_TEMPADDR:
3006 if (len != sizeof(int))
3007 return (EINVAL);
3008 preftemp = *(int *)buf;
3009 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3010 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3011 preftemp != IP6PO_TEMPADDR_PREFER) {
3012 return (EINVAL);
3013 }
3014 opt->ip6po_prefer_tempaddr = preftemp;
3015 break;
3016
3017 default:
3018 return (ENOPROTOOPT);
3019 } /* end of switch */
3020
3021 return (0);
3022 }
3023
3024 /*
3025 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3026 * packet to the input queue of a specified interface. Note that this
3027 * calls the output routine of the loopback "driver", but with an interface
3028 * pointer that might NOT be &loif -- easier than replicating that code here.
3029 */
3030 void
3031 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3032 {
3033 struct mbuf *copym;
3034 struct ip6_hdr *ip6;
3035
3036 copym = m_copy(m, 0, M_COPYALL);
3037 if (copym == NULL)
3038 return;
3039
3040 /*
3041 * Make sure to deep-copy IPv6 header portion in case the data
3042 * is in an mbuf cluster, so that we can safely override the IPv6
3043 * header portion later.
3044 */
3045 if (!M_WRITABLE(copym) ||
3046 copym->m_len < sizeof(struct ip6_hdr)) {
3047 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3048 if (copym == NULL)
3049 return;
3050 }
3051 ip6 = mtod(copym, struct ip6_hdr *);
3052 /*
3053 * clear embedded scope identifiers if necessary.
3054 * in6_clearscope will touch the addresses only when necessary.
3055 */
3056 in6_clearscope(&ip6->ip6_src);
3057 in6_clearscope(&ip6->ip6_dst);
3058 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3059 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3060 CSUM_PSEUDO_HDR;
3061 copym->m_pkthdr.csum_data = 0xffff;
3062 }
3063 if_simloop(ifp, copym, AF_INET6, 0);
3064 }
3065
3066 /*
3067 * Chop IPv6 header off from the payload.
3068 */
3069 static int
3070 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3071 {
3072 struct mbuf *mh;
3073 struct ip6_hdr *ip6;
3074
3075 ip6 = mtod(m, struct ip6_hdr *);
3076 if (m->m_len > sizeof(*ip6)) {
3077 mh = m_gethdr(M_NOWAIT, MT_DATA);
3078 if (mh == NULL) {
3079 m_freem(m);
3080 return ENOBUFS;
3081 }
3082 m_move_pkthdr(mh, m);
3083 M_ALIGN(mh, sizeof(*ip6));
3084 m->m_len -= sizeof(*ip6);
3085 m->m_data += sizeof(*ip6);
3086 mh->m_next = m;
3087 m = mh;
3088 m->m_len = sizeof(*ip6);
3089 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3090 }
3091 exthdrs->ip6e_ip6 = m;
3092 return 0;
3093 }
3094
3095 /*
3096 * Compute IPv6 extension header length.
3097 */
3098 int
3099 ip6_optlen(struct inpcb *in6p)
3100 {
3101 int len;
3102
3103 if (!in6p->in6p_outputopts)
3104 return 0;
3105
3106 len = 0;
3107 #define elen(x) \
3108 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3109
3110 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3111 if (in6p->in6p_outputopts->ip6po_rthdr)
3112 /* dest1 is valid with rthdr only */
3113 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3114 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3115 len += elen(in6p->in6p_outputopts->ip6po_dest2);
3116 return len;
3117 #undef elen
3118 }
Cache object: e0dc77617cb5830e9299a1a7d9e05237
|