1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
32 */
33
34 /*-
35 * Copyright (c) 1982, 1986, 1988, 1990, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
49 *
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
61 *
62 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
63 */
64
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #include "opt_kern_tls.h"
72 #include "opt_ratelimit.h"
73 #include "opt_route.h"
74 #include "opt_rss.h"
75 #include "opt_sctp.h"
76
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/ktls.h>
80 #include <sys/malloc.h>
81 #include <sys/mbuf.h>
82 #include <sys/errno.h>
83 #include <sys/priv.h>
84 #include <sys/proc.h>
85 #include <sys/protosw.h>
86 #include <sys/socket.h>
87 #include <sys/socketvar.h>
88 #include <sys/syslog.h>
89 #include <sys/ucred.h>
90
91 #include <machine/in_cksum.h>
92
93 #include <net/if.h>
94 #include <net/if_var.h>
95 #include <net/if_vlan_var.h>
96 #include <net/if_llatbl.h>
97 #include <net/ethernet.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/route/nhop.h>
101 #include <net/pfil.h>
102 #include <net/rss_config.h>
103 #include <net/vnet.h>
104
105 #include <netinet/in.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip_var.h>
108 #include <netinet6/in6_fib.h>
109 #include <netinet6/in6_var.h>
110 #include <netinet/ip6.h>
111 #include <netinet/icmp6.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet/in_pcb.h>
114 #include <netinet/tcp_var.h>
115 #include <netinet6/nd6.h>
116 #include <netinet6/in6_rss.h>
117
118 #include <netipsec/ipsec_support.h>
119 #if defined(SCTP) || defined(SCTP_SUPPORT)
120 #include <netinet/sctp.h>
121 #include <netinet/sctp_crc32.h>
122 #endif
123
124 #include <netinet6/ip6protosw.h>
125 #include <netinet6/scope6_var.h>
126
127 extern int in6_mcast_loop;
128
129 struct ip6_exthdrs {
130 struct mbuf *ip6e_ip6;
131 struct mbuf *ip6e_hbh;
132 struct mbuf *ip6e_dest1;
133 struct mbuf *ip6e_rthdr;
134 struct mbuf *ip6e_dest2;
135 };
136
137 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
138
139 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
140 struct ucred *, int);
141 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
142 struct socket *, struct sockopt *);
143 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
144 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
145 struct ucred *, int, int, int);
146
147 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
148 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
149 struct ip6_frag **);
150 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
151 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
152 static int ip6_getpmtu(struct route_in6 *, int,
153 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
154 u_int);
155 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
156 u_long *, int *, u_int);
157 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
158 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
159
160 /*
161 * Make an extension header from option data. hp is the source,
162 * mp is the destination, and _ol is the optlen.
163 */
164 #define MAKE_EXTHDR(hp, mp, _ol) \
165 do { \
166 if (hp) { \
167 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
168 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
169 ((eh)->ip6e_len + 1) << 3); \
170 if (error) \
171 goto freehdrs; \
172 (_ol) += (*(mp))->m_len; \
173 } \
174 } while (/*CONSTCOND*/ 0)
175
176 /*
177 * Form a chain of extension headers.
178 * m is the extension header mbuf
179 * mp is the previous mbuf in the chain
180 * p is the next header
181 * i is the type of option.
182 */
183 #define MAKE_CHAIN(m, mp, p, i)\
184 do {\
185 if (m) {\
186 if (!hdrsplit) \
187 panic("%s:%d: assumption failed: "\
188 "hdr not split: hdrsplit %d exthdrs %p",\
189 __func__, __LINE__, hdrsplit, &exthdrs);\
190 *mtod((m), u_char *) = *(p);\
191 *(p) = (i);\
192 p = mtod((m), u_char *);\
193 (m)->m_next = (mp)->m_next;\
194 (mp)->m_next = (m);\
195 (mp) = (m);\
196 }\
197 } while (/*CONSTCOND*/ 0)
198
199 void
200 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
201 {
202 u_short csum;
203
204 csum = in_cksum_skip(m, offset + plen, offset);
205 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
206 csum = 0xffff;
207 offset += m->m_pkthdr.csum_data; /* checksum offset */
208
209 if (offset + sizeof(csum) > m->m_len)
210 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
211 else
212 *(u_short *)mtodo(m, offset) = csum;
213 }
214
215 static void
216 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
217 int plen, int optlen)
218 {
219
220 KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
221 "csum_flags %#x",
222 __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
223
224 if (csum_flags & CSUM_DELAY_DATA_IPV6) {
225 in6_delayed_cksum(m, plen - optlen,
226 sizeof(struct ip6_hdr) + optlen);
227 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
228 }
229 #if defined(SCTP) || defined(SCTP_SUPPORT)
230 if (csum_flags & CSUM_SCTP_IPV6) {
231 sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
232 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
233 }
234 #endif
235 }
236
237 int
238 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
239 int fraglen , uint32_t id)
240 {
241 struct mbuf *m, **mnext, *m_frgpart;
242 struct ip6_hdr *ip6, *mhip6;
243 struct ip6_frag *ip6f;
244 int off;
245 int error;
246 int tlen = m0->m_pkthdr.len;
247
248 KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
249
250 m = m0;
251 ip6 = mtod(m, struct ip6_hdr *);
252 mnext = &m->m_nextpkt;
253
254 for (off = hlen; off < tlen; off += fraglen) {
255 m = m_gethdr(M_NOWAIT, MT_DATA);
256 if (!m) {
257 IP6STAT_INC(ip6s_odropped);
258 return (ENOBUFS);
259 }
260
261 /*
262 * Make sure the complete packet header gets copied
263 * from the originating mbuf to the newly created
264 * mbuf. This also ensures that existing firewall
265 * classification(s), VLAN tags and so on get copied
266 * to the resulting fragmented packet(s):
267 */
268 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
269 m_free(m);
270 IP6STAT_INC(ip6s_odropped);
271 return (ENOBUFS);
272 }
273
274 *mnext = m;
275 mnext = &m->m_nextpkt;
276 m->m_data += max_linkhdr;
277 mhip6 = mtod(m, struct ip6_hdr *);
278 *mhip6 = *ip6;
279 m->m_len = sizeof(*mhip6);
280 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
281 if (error) {
282 IP6STAT_INC(ip6s_odropped);
283 return (error);
284 }
285 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
286 if (off + fraglen >= tlen)
287 fraglen = tlen - off;
288 else
289 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
290 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
291 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
292 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
293 IP6STAT_INC(ip6s_odropped);
294 return (ENOBUFS);
295 }
296 m_cat(m, m_frgpart);
297 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
298 ip6f->ip6f_reserved = 0;
299 ip6f->ip6f_ident = id;
300 ip6f->ip6f_nxt = nextproto;
301 IP6STAT_INC(ip6s_ofragments);
302 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
303 }
304
305 return (0);
306 }
307
308 static int
309 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
310 struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
311 bool stamp_tag)
312 {
313 #ifdef KERN_TLS
314 struct ktls_session *tls = NULL;
315 #endif
316 struct m_snd_tag *mst;
317 int error;
318
319 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
320 mst = NULL;
321
322 #ifdef KERN_TLS
323 /*
324 * If this is an unencrypted TLS record, save a reference to
325 * the record. This local reference is used to call
326 * ktls_output_eagain after the mbuf has been freed (thus
327 * dropping the mbuf's reference) in if_output.
328 */
329 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
330 tls = ktls_hold(m->m_next->m_epg_tls);
331 mst = tls->snd_tag;
332
333 /*
334 * If a TLS session doesn't have a valid tag, it must
335 * have had an earlier ifp mismatch, so drop this
336 * packet.
337 */
338 if (mst == NULL) {
339 error = EAGAIN;
340 goto done;
341 }
342 /*
343 * Always stamp tags that include NIC ktls.
344 */
345 stamp_tag = true;
346 }
347 #endif
348 #ifdef RATELIMIT
349 if (inp != NULL && mst == NULL) {
350 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
351 (inp->inp_snd_tag != NULL &&
352 inp->inp_snd_tag->ifp != ifp))
353 in_pcboutput_txrtlmt(inp, ifp, m);
354
355 if (inp->inp_snd_tag != NULL)
356 mst = inp->inp_snd_tag;
357 }
358 #endif
359 if (stamp_tag && mst != NULL) {
360 KASSERT(m->m_pkthdr.rcvif == NULL,
361 ("trying to add a send tag to a forwarded packet"));
362 if (mst->ifp != ifp) {
363 error = EAGAIN;
364 goto done;
365 }
366
367 /* stamp send tag on mbuf */
368 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
369 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
370 }
371
372 error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
373
374 done:
375 /* Check for route change invalidating send tags. */
376 #ifdef KERN_TLS
377 if (tls != NULL) {
378 if (error == EAGAIN)
379 error = ktls_output_eagain(inp, tls);
380 ktls_free(tls);
381 }
382 #endif
383 #ifdef RATELIMIT
384 if (error == EAGAIN)
385 in_pcboutput_eagain(inp);
386 #endif
387 return (error);
388 }
389
390 /*
391 * IP6 output.
392 * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
393 * nxt, hlim, src, dst).
394 * This function may modify ver and hlim only.
395 * The mbuf chain containing the packet will be freed.
396 * The mbuf opt, if present, will not be freed.
397 * If route_in6 ro is present and has ro_nh initialized, route lookup would be
398 * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
399 * then result of route lookup is stored in ro->ro_nh.
400 *
401 * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
402 * is uint32_t. So we use u_long to hold largest one, which is rt_mtu.
403 *
404 * ifpp - XXX: just for statistics
405 */
406 int
407 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
408 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
409 struct ifnet **ifpp, struct inpcb *inp)
410 {
411 struct ip6_hdr *ip6;
412 struct ifnet *ifp, *origifp;
413 struct mbuf *m = m0;
414 struct mbuf *mprev;
415 struct route_in6 *ro_pmtu;
416 struct nhop_object *nh;
417 struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
418 struct in6_addr odst;
419 u_char *nexthdrp;
420 int tlen, len;
421 int error = 0;
422 int vlan_pcp = -1;
423 struct in6_ifaddr *ia = NULL;
424 u_long mtu;
425 int alwaysfrag, dontfrag;
426 u_int32_t optlen, plen = 0, unfragpartlen;
427 struct ip6_exthdrs exthdrs;
428 struct in6_addr src0, dst0;
429 u_int32_t zone;
430 bool hdrsplit;
431 int sw_csum, tso;
432 int needfiblookup;
433 uint32_t fibnum;
434 struct m_tag *fwd_tag = NULL;
435 uint32_t id;
436
437 NET_EPOCH_ASSERT();
438
439 if (inp != NULL) {
440 INP_LOCK_ASSERT(inp);
441 M_SETFIB(m, inp->inp_inc.inc_fibnum);
442 if ((flags & IP_NODEFAULTFLOWID) == 0) {
443 /* Unconditionally set flowid. */
444 m->m_pkthdr.flowid = inp->inp_flowid;
445 M_HASHTYPE_SET(m, inp->inp_flowtype);
446 }
447 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
448 vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
449 INP_2PCP_SHIFT;
450 #ifdef NUMA
451 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
452 #endif
453 }
454
455 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
456 /*
457 * IPSec checking which handles several cases.
458 * FAST IPSEC: We re-injected the packet.
459 * XXX: need scope argument.
460 */
461 if (IPSEC_ENABLED(ipv6)) {
462 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
463 if (error == EINPROGRESS)
464 error = 0;
465 goto done;
466 }
467 }
468 #endif /* IPSEC */
469
470 /* Source address validation. */
471 ip6 = mtod(m, struct ip6_hdr *);
472 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
473 (flags & IPV6_UNSPECSRC) == 0) {
474 error = EOPNOTSUPP;
475 IP6STAT_INC(ip6s_badscope);
476 goto bad;
477 }
478 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
479 error = EOPNOTSUPP;
480 IP6STAT_INC(ip6s_badscope);
481 goto bad;
482 }
483
484 /*
485 * If we are given packet options to add extension headers prepare them.
486 * Calculate the total length of the extension header chain.
487 * Keep the length of the unfragmentable part for fragmentation.
488 */
489 bzero(&exthdrs, sizeof(exthdrs));
490 optlen = 0;
491 unfragpartlen = sizeof(struct ip6_hdr);
492 if (opt) {
493 /* Hop-by-Hop options header. */
494 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
495
496 /* Destination options header (1st part). */
497 if (opt->ip6po_rthdr) {
498 #ifndef RTHDR_SUPPORT_IMPLEMENTED
499 /*
500 * If there is a routing header, discard the packet
501 * right away here. RH0/1 are obsolete and we do not
502 * currently support RH2/3/4.
503 * People trying to use RH253/254 may want to disable
504 * this check.
505 * The moment we do support any routing header (again)
506 * this block should check the routing type more
507 * selectively.
508 */
509 error = EINVAL;
510 goto bad;
511 #endif
512
513 /*
514 * Destination options header (1st part).
515 * This only makes sense with a routing header.
516 * See Section 9.2 of RFC 3542.
517 * Disabling this part just for MIP6 convenience is
518 * a bad idea. We need to think carefully about a
519 * way to make the advanced API coexist with MIP6
520 * options, which might automatically be inserted in
521 * the kernel.
522 */
523 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
524 optlen);
525 }
526 /* Routing header. */
527 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
528
529 unfragpartlen += optlen;
530
531 /*
532 * NOTE: we don't add AH/ESP length here (done in
533 * ip6_ipsec_output()).
534 */
535
536 /* Destination options header (2nd part). */
537 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
538 }
539
540 /*
541 * If there is at least one extension header,
542 * separate IP6 header from the payload.
543 */
544 hdrsplit = false;
545 if (optlen) {
546 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
547 m = NULL;
548 goto freehdrs;
549 }
550 m = exthdrs.ip6e_ip6;
551 ip6 = mtod(m, struct ip6_hdr *);
552 hdrsplit = true;
553 }
554
555 /* Adjust mbuf packet header length. */
556 m->m_pkthdr.len += optlen;
557 plen = m->m_pkthdr.len - sizeof(*ip6);
558
559 /* If this is a jumbo payload, insert a jumbo payload option. */
560 if (plen > IPV6_MAXPACKET) {
561 if (!hdrsplit) {
562 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
563 m = NULL;
564 goto freehdrs;
565 }
566 m = exthdrs.ip6e_ip6;
567 ip6 = mtod(m, struct ip6_hdr *);
568 hdrsplit = true;
569 }
570 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
571 goto freehdrs;
572 ip6->ip6_plen = 0;
573 } else
574 ip6->ip6_plen = htons(plen);
575 nexthdrp = &ip6->ip6_nxt;
576
577 if (optlen) {
578 /*
579 * Concatenate headers and fill in next header fields.
580 * Here we have, on "m"
581 * IPv6 payload
582 * and we insert headers accordingly.
583 * Finally, we should be getting:
584 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
585 *
586 * During the header composing process "m" points to IPv6
587 * header. "mprev" points to an extension header prior to esp.
588 */
589 mprev = m;
590
591 /*
592 * We treat dest2 specially. This makes IPsec processing
593 * much easier. The goal here is to make mprev point the
594 * mbuf prior to dest2.
595 *
596 * Result: IPv6 dest2 payload.
597 * m and mprev will point to IPv6 header.
598 */
599 if (exthdrs.ip6e_dest2) {
600 if (!hdrsplit)
601 panic("%s:%d: assumption failed: "
602 "hdr not split: hdrsplit %d exthdrs %p",
603 __func__, __LINE__, hdrsplit, &exthdrs);
604 exthdrs.ip6e_dest2->m_next = m->m_next;
605 m->m_next = exthdrs.ip6e_dest2;
606 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
607 ip6->ip6_nxt = IPPROTO_DSTOPTS;
608 }
609
610 /*
611 * Result: IPv6 hbh dest1 rthdr dest2 payload.
612 * m will point to IPv6 header. mprev will point to the
613 * extension header prior to dest2 (rthdr in the above case).
614 */
615 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
616 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
617 IPPROTO_DSTOPTS);
618 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
619 IPPROTO_ROUTING);
620 }
621
622 IP6STAT_INC(ip6s_localout);
623
624 /* Route packet. */
625 ro_pmtu = ro;
626 if (opt && opt->ip6po_rthdr)
627 ro = &opt->ip6po_route;
628 if (ro != NULL)
629 dst = (struct sockaddr_in6 *)&ro->ro_dst;
630 else
631 dst = &sin6;
632 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
633
634 again:
635 /*
636 * If specified, try to fill in the traffic class field.
637 * Do not override if a non-zero value is already set.
638 * We check the diffserv field and the ECN field separately.
639 */
640 if (opt && opt->ip6po_tclass >= 0) {
641 int mask = 0;
642
643 if (IPV6_DSCP(ip6) == 0)
644 mask |= 0xfc;
645 if (IPV6_ECN(ip6) == 0)
646 mask |= 0x03;
647 if (mask != 0)
648 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
649 }
650
651 /* Fill in or override the hop limit field, if necessary. */
652 if (opt && opt->ip6po_hlim != -1)
653 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
654 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
655 if (im6o != NULL)
656 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
657 else
658 ip6->ip6_hlim = V_ip6_defmcasthlim;
659 }
660
661 if (ro == NULL || ro->ro_nh == NULL) {
662 bzero(dst, sizeof(*dst));
663 dst->sin6_family = AF_INET6;
664 dst->sin6_len = sizeof(*dst);
665 dst->sin6_addr = ip6->ip6_dst;
666 }
667 /*
668 * Validate route against routing table changes.
669 * Make sure that the address family is set in route.
670 */
671 nh = NULL;
672 ifp = NULL;
673 mtu = 0;
674 if (ro != NULL) {
675 if (ro->ro_nh != NULL && inp != NULL) {
676 ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
677 NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
678 fibnum);
679 }
680 if (ro->ro_nh != NULL && fwd_tag == NULL &&
681 (!NH_IS_VALID(ro->ro_nh) ||
682 ro->ro_dst.sin6_family != AF_INET6 ||
683 !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
684 RO_INVALIDATE_CACHE(ro);
685
686 if (ro->ro_nh != NULL && fwd_tag == NULL &&
687 ro->ro_dst.sin6_family == AF_INET6 &&
688 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
689 nh = ro->ro_nh;
690 ifp = nh->nh_ifp;
691 } else {
692 if (ro->ro_lle)
693 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
694 ro->ro_lle = NULL;
695 if (fwd_tag == NULL) {
696 bzero(&dst_sa, sizeof(dst_sa));
697 dst_sa.sin6_family = AF_INET6;
698 dst_sa.sin6_len = sizeof(dst_sa);
699 dst_sa.sin6_addr = ip6->ip6_dst;
700 }
701 error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
702 &nh, fibnum, m->m_pkthdr.flowid);
703 if (error != 0) {
704 IP6STAT_INC(ip6s_noroute);
705 if (ifp != NULL)
706 in6_ifstat_inc(ifp, ifs6_out_discard);
707 goto bad;
708 }
709 if (ifp != NULL)
710 mtu = ifp->if_mtu;
711 }
712 if (nh == NULL) {
713 /*
714 * If in6_selectroute() does not return a nexthop
715 * dst may not have been updated.
716 */
717 *dst = dst_sa; /* XXX */
718 } else {
719 if (nh->nh_flags & NHF_HOST)
720 mtu = nh->nh_mtu;
721 ia = (struct in6_ifaddr *)(nh->nh_ifa);
722 counter_u64_add(nh->nh_pksent, 1);
723 }
724 } else {
725 struct nhop_object *nh;
726 struct in6_addr kdst;
727 uint32_t scopeid;
728
729 if (fwd_tag == NULL) {
730 bzero(&dst_sa, sizeof(dst_sa));
731 dst_sa.sin6_family = AF_INET6;
732 dst_sa.sin6_len = sizeof(dst_sa);
733 dst_sa.sin6_addr = ip6->ip6_dst;
734 }
735
736 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
737 im6o != NULL &&
738 (ifp = im6o->im6o_multicast_ifp) != NULL) {
739 /* We do not need a route lookup. */
740 *dst = dst_sa; /* XXX */
741 goto nonh6lookup;
742 }
743
744 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
745
746 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
747 IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
748 if (scopeid > 0) {
749 ifp = in6_getlinkifnet(scopeid);
750 if (ifp == NULL) {
751 error = EHOSTUNREACH;
752 goto bad;
753 }
754 *dst = dst_sa; /* XXX */
755 goto nonh6lookup;
756 }
757 }
758
759 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
760 m->m_pkthdr.flowid);
761 if (nh == NULL) {
762 IP6STAT_INC(ip6s_noroute);
763 /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
764 error = EHOSTUNREACH;;
765 goto bad;
766 }
767
768 ifp = nh->nh_ifp;
769 mtu = nh->nh_mtu;
770 ia = ifatoia6(nh->nh_ifa);
771 if (nh->nh_flags & NHF_GATEWAY)
772 dst->sin6_addr = nh->gw6_sa.sin6_addr;
773 nonh6lookup:
774 ;
775 }
776
777 /* Then nh (for unicast) and ifp must be non-NULL valid values. */
778 if ((flags & IPV6_FORWARDING) == 0) {
779 /* XXX: the FORWARDING flag can be set for mrouting. */
780 in6_ifstat_inc(ifp, ifs6_out_request);
781 }
782
783 /* Setup data structures for scope ID checks. */
784 src0 = ip6->ip6_src;
785 bzero(&src_sa, sizeof(src_sa));
786 src_sa.sin6_family = AF_INET6;
787 src_sa.sin6_len = sizeof(src_sa);
788 src_sa.sin6_addr = ip6->ip6_src;
789
790 dst0 = ip6->ip6_dst;
791 /* Re-initialize to be sure. */
792 bzero(&dst_sa, sizeof(dst_sa));
793 dst_sa.sin6_family = AF_INET6;
794 dst_sa.sin6_len = sizeof(dst_sa);
795 dst_sa.sin6_addr = ip6->ip6_dst;
796
797 /* Check for valid scope ID. */
798 if (in6_setscope(&src0, ifp, &zone) == 0 &&
799 sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
800 in6_setscope(&dst0, ifp, &zone) == 0 &&
801 sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
802 /*
803 * The outgoing interface is in the zone of the source
804 * and destination addresses.
805 *
806 * Because the loopback interface cannot receive
807 * packets with a different scope ID than its own,
808 * there is a trick to pretend the outgoing packet
809 * was received by the real network interface, by
810 * setting "origifp" different from "ifp". This is
811 * only allowed when "ifp" is a loopback network
812 * interface. Refer to code in nd6_output_ifp() for
813 * more details.
814 */
815 origifp = ifp;
816
817 /*
818 * We should use ia_ifp to support the case of sending
819 * packets to an address of our own.
820 */
821 if (ia != NULL && ia->ia_ifp)
822 ifp = ia->ia_ifp;
823
824 } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
825 sa6_recoverscope(&src_sa) != 0 ||
826 sa6_recoverscope(&dst_sa) != 0 ||
827 dst_sa.sin6_scope_id == 0 ||
828 (src_sa.sin6_scope_id != 0 &&
829 src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
830 (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
831 /*
832 * If the destination network interface is not a
833 * loopback interface, or the destination network
834 * address has no scope ID, or the source address has
835 * a scope ID set which is different from the
836 * destination address one, or there is no network
837 * interface representing this scope ID, the address
838 * pair is considered invalid.
839 */
840 IP6STAT_INC(ip6s_badscope);
841 in6_ifstat_inc(ifp, ifs6_out_discard);
842 if (error == 0)
843 error = EHOSTUNREACH; /* XXX */
844 goto bad;
845 }
846 /* All scope ID checks are successful. */
847
848 if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
849 if (opt && opt->ip6po_nextroute.ro_nh) {
850 /*
851 * The nexthop is explicitly specified by the
852 * application. We assume the next hop is an IPv6
853 * address.
854 */
855 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
856 }
857 else if ((nh->nh_flags & NHF_GATEWAY))
858 dst = &nh->gw6_sa;
859 }
860
861 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
862 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
863 } else {
864 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
865 in6_ifstat_inc(ifp, ifs6_out_mcast);
866
867 /* Confirm that the outgoing interface supports multicast. */
868 if (!(ifp->if_flags & IFF_MULTICAST)) {
869 IP6STAT_INC(ip6s_noroute);
870 in6_ifstat_inc(ifp, ifs6_out_discard);
871 error = ENETUNREACH;
872 goto bad;
873 }
874 if ((im6o == NULL && in6_mcast_loop) ||
875 (im6o && im6o->im6o_multicast_loop)) {
876 /*
877 * Loop back multicast datagram if not expressly
878 * forbidden to do so, even if we have not joined
879 * the address; protocols will filter it later,
880 * thus deferring a hash lookup and lock acquisition
881 * at the expense of an m_copym().
882 */
883 ip6_mloopback(ifp, m);
884 } else {
885 /*
886 * If we are acting as a multicast router, perform
887 * multicast forwarding as if the packet had just
888 * arrived on the interface to which we are about
889 * to send. The multicast forwarding function
890 * recursively calls this function, using the
891 * IPV6_FORWARDING flag to prevent infinite recursion.
892 *
893 * Multicasts that are looped back by ip6_mloopback(),
894 * above, will be forwarded by the ip6_input() routine,
895 * if necessary.
896 */
897 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
898 /*
899 * XXX: ip6_mforward expects that rcvif is NULL
900 * when it is called from the originating path.
901 * However, it may not always be the case.
902 */
903 m->m_pkthdr.rcvif = NULL;
904 if (ip6_mforward(ip6, ifp, m) != 0) {
905 m_freem(m);
906 goto done;
907 }
908 }
909 }
910 /*
911 * Multicasts with a hoplimit of zero may be looped back,
912 * above, but must not be transmitted on a network.
913 * Also, multicasts addressed to the loopback interface
914 * are not sent -- the above call to ip6_mloopback() will
915 * loop back a copy if this host actually belongs to the
916 * destination group on the loopback interface.
917 */
918 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
919 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
920 m_freem(m);
921 goto done;
922 }
923 }
924
925 /*
926 * Fill the outgoing inteface to tell the upper layer
927 * to increment per-interface statistics.
928 */
929 if (ifpp)
930 *ifpp = ifp;
931
932 /* Determine path MTU. */
933 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
934 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
935 goto bad;
936 KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
937 "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
938 ifp, alwaysfrag, fibnum));
939
940 /*
941 * The caller of this function may specify to use the minimum MTU
942 * in some cases.
943 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
944 * setting. The logic is a bit complicated; by default, unicast
945 * packets will follow path MTU while multicast packets will be sent at
946 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
947 * including unicast ones will be sent at the minimum MTU. Multicast
948 * packets will always be sent at the minimum MTU unless
949 * IP6PO_MINMTU_DISABLE is explicitly specified.
950 * See RFC 3542 for more details.
951 */
952 if (mtu > IPV6_MMTU) {
953 if ((flags & IPV6_MINMTU))
954 mtu = IPV6_MMTU;
955 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
956 mtu = IPV6_MMTU;
957 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
958 (opt == NULL ||
959 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
960 mtu = IPV6_MMTU;
961 }
962 }
963
964 /*
965 * Clear embedded scope identifiers if necessary.
966 * in6_clearscope() will touch the addresses only when necessary.
967 */
968 in6_clearscope(&ip6->ip6_src);
969 in6_clearscope(&ip6->ip6_dst);
970
971 /*
972 * If the outgoing packet contains a hop-by-hop options header,
973 * it must be examined and processed even by the source node.
974 * (RFC 2460, section 4.)
975 */
976 if (exthdrs.ip6e_hbh) {
977 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
978 u_int32_t dummy; /* XXX unused */
979 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
980
981 #ifdef DIAGNOSTIC
982 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
983 panic("ip6e_hbh is not contiguous");
984 #endif
985 /*
986 * XXX: if we have to send an ICMPv6 error to the sender,
987 * we need the M_LOOP flag since icmp6_error() expects
988 * the IPv6 and the hop-by-hop options header are
989 * contiguous unless the flag is set.
990 */
991 m->m_flags |= M_LOOP;
992 m->m_pkthdr.rcvif = ifp;
993 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
994 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
995 &dummy, &plen) < 0) {
996 /* m was already freed at this point. */
997 error = EINVAL;/* better error? */
998 goto done;
999 }
1000 m->m_flags &= ~M_LOOP; /* XXX */
1001 m->m_pkthdr.rcvif = NULL;
1002 }
1003
1004 /* Jump over all PFIL processing if hooks are not active. */
1005 if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1006 goto passout;
1007
1008 odst = ip6->ip6_dst;
1009 /* Run through list of hooks for output packets. */
1010 switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1011 case PFIL_PASS:
1012 ip6 = mtod(m, struct ip6_hdr *);
1013 break;
1014 case PFIL_DROPPED:
1015 error = EACCES;
1016 /* FALLTHROUGH */
1017 case PFIL_CONSUMED:
1018 goto done;
1019 }
1020
1021 needfiblookup = 0;
1022 /* See if destination IP address was changed by packet filter. */
1023 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1024 m->m_flags |= M_SKIP_FIREWALL;
1025 /* If destination is now ourself drop to ip6_input(). */
1026 if (in6_localip(&ip6->ip6_dst)) {
1027 m->m_flags |= M_FASTFWD_OURS;
1028 if (m->m_pkthdr.rcvif == NULL)
1029 m->m_pkthdr.rcvif = V_loif;
1030 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1031 m->m_pkthdr.csum_flags |=
1032 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1033 m->m_pkthdr.csum_data = 0xffff;
1034 }
1035 #if defined(SCTP) || defined(SCTP_SUPPORT)
1036 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1037 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1038 #endif
1039 error = netisr_queue(NETISR_IPV6, m);
1040 goto done;
1041 } else {
1042 if (ro != NULL)
1043 RO_INVALIDATE_CACHE(ro);
1044 needfiblookup = 1; /* Redo the routing table lookup. */
1045 }
1046 }
1047 /* See if fib was changed by packet filter. */
1048 if (fibnum != M_GETFIB(m)) {
1049 m->m_flags |= M_SKIP_FIREWALL;
1050 fibnum = M_GETFIB(m);
1051 if (ro != NULL)
1052 RO_INVALIDATE_CACHE(ro);
1053 needfiblookup = 1;
1054 }
1055 if (needfiblookup)
1056 goto again;
1057
1058 /* See if local, if yes, send it to netisr. */
1059 if (m->m_flags & M_FASTFWD_OURS) {
1060 if (m->m_pkthdr.rcvif == NULL)
1061 m->m_pkthdr.rcvif = V_loif;
1062 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1063 m->m_pkthdr.csum_flags |=
1064 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1065 m->m_pkthdr.csum_data = 0xffff;
1066 }
1067 #if defined(SCTP) || defined(SCTP_SUPPORT)
1068 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1069 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1070 #endif
1071 error = netisr_queue(NETISR_IPV6, m);
1072 goto done;
1073 }
1074 /* Or forward to some other address? */
1075 if ((m->m_flags & M_IP6_NEXTHOP) &&
1076 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1077 if (ro != NULL)
1078 dst = (struct sockaddr_in6 *)&ro->ro_dst;
1079 else
1080 dst = &sin6;
1081 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1082 m->m_flags |= M_SKIP_FIREWALL;
1083 m->m_flags &= ~M_IP6_NEXTHOP;
1084 m_tag_delete(m, fwd_tag);
1085 goto again;
1086 }
1087
1088 passout:
1089 if (vlan_pcp > -1)
1090 EVL_APPLY_PRI(m, vlan_pcp);
1091
1092 /* Ensure the packet data is mapped if the interface requires it. */
1093 if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1094 m = mb_unmapped_to_ext(m);
1095 if (m == NULL) {
1096 IP6STAT_INC(ip6s_odropped);
1097 return (ENOBUFS);
1098 }
1099 }
1100
1101 /*
1102 * Send the packet to the outgoing interface.
1103 * If necessary, do IPv6 fragmentation before sending.
1104 *
1105 * The logic here is rather complex:
1106 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1107 * 1-a: send as is if tlen <= path mtu
1108 * 1-b: fragment if tlen > path mtu
1109 *
1110 * 2: if user asks us not to fragment (dontfrag == 1)
1111 * 2-a: send as is if tlen <= interface mtu
1112 * 2-b: error if tlen > interface mtu
1113 *
1114 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1115 * always fragment
1116 *
1117 * 4: if dontfrag == 1 && alwaysfrag == 1
1118 * error, as we cannot handle this conflicting request.
1119 */
1120 sw_csum = m->m_pkthdr.csum_flags;
1121 if (!hdrsplit) {
1122 tso = ((sw_csum & ifp->if_hwassist &
1123 (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1124 sw_csum &= ~ifp->if_hwassist;
1125 } else
1126 tso = 0;
1127 /*
1128 * If we added extension headers, we will not do TSO and calculate the
1129 * checksums ourselves for now.
1130 * XXX-BZ Need a framework to know when the NIC can handle it, even
1131 * with ext. hdrs.
1132 */
1133 ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1134 /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1135 tlen = m->m_pkthdr.len;
1136
1137 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1138 dontfrag = 1;
1139 else
1140 dontfrag = 0;
1141 if (dontfrag && alwaysfrag) { /* Case 4. */
1142 /* Conflicting request - can't transmit. */
1143 error = EMSGSIZE;
1144 goto bad;
1145 }
1146 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* Case 2-b. */
1147 /*
1148 * Even if the DONTFRAG option is specified, we cannot send the
1149 * packet when the data length is larger than the MTU of the
1150 * outgoing interface.
1151 * Notify the error by sending IPV6_PATHMTU ancillary data if
1152 * application wanted to know the MTU value. Also return an
1153 * error code (this is not described in the API spec).
1154 */
1155 if (inp != NULL)
1156 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1157 error = EMSGSIZE;
1158 goto bad;
1159 }
1160
1161 /* Transmit packet without fragmentation. */
1162 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1163 struct in6_ifaddr *ia6;
1164
1165 ip6 = mtod(m, struct ip6_hdr *);
1166 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1167 if (ia6) {
1168 /* Record statistics for this interface address. */
1169 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1170 counter_u64_add(ia6->ia_ifa.ifa_obytes,
1171 m->m_pkthdr.len);
1172 }
1173 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1174 (flags & IP_NO_SND_TAG_RL) ? false : true);
1175 goto done;
1176 }
1177
1178 /* Try to fragment the packet. Cases 1-b and 3. */
1179 if (mtu < IPV6_MMTU) {
1180 /* Path MTU cannot be less than IPV6_MMTU. */
1181 error = EMSGSIZE;
1182 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1183 goto bad;
1184 } else if (ip6->ip6_plen == 0) {
1185 /* Jumbo payload cannot be fragmented. */
1186 error = EMSGSIZE;
1187 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1188 goto bad;
1189 } else {
1190 u_char nextproto;
1191
1192 /*
1193 * Too large for the destination or interface;
1194 * fragment if possible.
1195 * Must be able to put at least 8 bytes per fragment.
1196 */
1197 if (mtu > IPV6_MAXPACKET)
1198 mtu = IPV6_MAXPACKET;
1199
1200 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1201 if (len < 8) {
1202 error = EMSGSIZE;
1203 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1204 goto bad;
1205 }
1206
1207 /*
1208 * If the interface will not calculate checksums on
1209 * fragmented packets, then do it here.
1210 * XXX-BZ handle the hw offloading case. Need flags.
1211 */
1212 ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1213 optlen);
1214
1215 /*
1216 * Change the next header field of the last header in the
1217 * unfragmentable part.
1218 */
1219 if (exthdrs.ip6e_rthdr) {
1220 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1221 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1222 } else if (exthdrs.ip6e_dest1) {
1223 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1224 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1225 } else if (exthdrs.ip6e_hbh) {
1226 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1227 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1228 } else {
1229 ip6 = mtod(m, struct ip6_hdr *);
1230 nextproto = ip6->ip6_nxt;
1231 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1232 }
1233
1234 /*
1235 * Loop through length of segment after first fragment,
1236 * make new header and copy data of each part and link onto
1237 * chain.
1238 */
1239 m0 = m;
1240 id = htonl(ip6_randomid());
1241 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1242 if (error != 0)
1243 goto sendorfree;
1244
1245 in6_ifstat_inc(ifp, ifs6_out_fragok);
1246 }
1247
1248 /* Remove leading garbage. */
1249 sendorfree:
1250 m = m0->m_nextpkt;
1251 m0->m_nextpkt = 0;
1252 m_freem(m0);
1253 for (; m; m = m0) {
1254 m0 = m->m_nextpkt;
1255 m->m_nextpkt = 0;
1256 if (error == 0) {
1257 /* Record statistics for this interface address. */
1258 if (ia) {
1259 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1260 counter_u64_add(ia->ia_ifa.ifa_obytes,
1261 m->m_pkthdr.len);
1262 }
1263 if (vlan_pcp > -1)
1264 EVL_APPLY_PRI(m, vlan_pcp);
1265 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1266 true);
1267 } else
1268 m_freem(m);
1269 }
1270
1271 if (error == 0)
1272 IP6STAT_INC(ip6s_fragmented);
1273
1274 done:
1275 return (error);
1276
1277 freehdrs:
1278 m_freem(exthdrs.ip6e_hbh); /* m_freem() checks if mbuf is NULL. */
1279 m_freem(exthdrs.ip6e_dest1);
1280 m_freem(exthdrs.ip6e_rthdr);
1281 m_freem(exthdrs.ip6e_dest2);
1282 /* FALLTHROUGH */
1283 bad:
1284 if (m)
1285 m_freem(m);
1286 goto done;
1287 }
1288
1289 static int
1290 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1291 {
1292 struct mbuf *m;
1293
1294 if (hlen > MCLBYTES)
1295 return (ENOBUFS); /* XXX */
1296
1297 if (hlen > MLEN)
1298 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1299 else
1300 m = m_get(M_NOWAIT, MT_DATA);
1301 if (m == NULL)
1302 return (ENOBUFS);
1303 m->m_len = hlen;
1304 if (hdr)
1305 bcopy(hdr, mtod(m, caddr_t), hlen);
1306
1307 *mp = m;
1308 return (0);
1309 }
1310
1311 /*
1312 * Insert jumbo payload option.
1313 */
1314 static int
1315 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1316 {
1317 struct mbuf *mopt;
1318 u_char *optbuf;
1319 u_int32_t v;
1320
1321 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1322
1323 /*
1324 * If there is no hop-by-hop options header, allocate new one.
1325 * If there is one but it doesn't have enough space to store the
1326 * jumbo payload option, allocate a cluster to store the whole options.
1327 * Otherwise, use it to store the options.
1328 */
1329 if (exthdrs->ip6e_hbh == NULL) {
1330 mopt = m_get(M_NOWAIT, MT_DATA);
1331 if (mopt == NULL)
1332 return (ENOBUFS);
1333 mopt->m_len = JUMBOOPTLEN;
1334 optbuf = mtod(mopt, u_char *);
1335 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1336 exthdrs->ip6e_hbh = mopt;
1337 } else {
1338 struct ip6_hbh *hbh;
1339
1340 mopt = exthdrs->ip6e_hbh;
1341 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1342 /*
1343 * XXX assumption:
1344 * - exthdrs->ip6e_hbh is not referenced from places
1345 * other than exthdrs.
1346 * - exthdrs->ip6e_hbh is not an mbuf chain.
1347 */
1348 int oldoptlen = mopt->m_len;
1349 struct mbuf *n;
1350
1351 /*
1352 * XXX: give up if the whole (new) hbh header does
1353 * not fit even in an mbuf cluster.
1354 */
1355 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1356 return (ENOBUFS);
1357
1358 /*
1359 * As a consequence, we must always prepare a cluster
1360 * at this point.
1361 */
1362 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1363 if (n == NULL)
1364 return (ENOBUFS);
1365 n->m_len = oldoptlen + JUMBOOPTLEN;
1366 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1367 oldoptlen);
1368 optbuf = mtod(n, caddr_t) + oldoptlen;
1369 m_freem(mopt);
1370 mopt = exthdrs->ip6e_hbh = n;
1371 } else {
1372 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1373 mopt->m_len += JUMBOOPTLEN;
1374 }
1375 optbuf[0] = IP6OPT_PADN;
1376 optbuf[1] = 1;
1377
1378 /*
1379 * Adjust the header length according to the pad and
1380 * the jumbo payload option.
1381 */
1382 hbh = mtod(mopt, struct ip6_hbh *);
1383 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1384 }
1385
1386 /* fill in the option. */
1387 optbuf[2] = IP6OPT_JUMBO;
1388 optbuf[3] = 4;
1389 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1390 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1391
1392 /* finally, adjust the packet header length */
1393 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1394
1395 return (0);
1396 #undef JUMBOOPTLEN
1397 }
1398
1399 /*
1400 * Insert fragment header and copy unfragmentable header portions.
1401 */
1402 static int
1403 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1404 struct ip6_frag **frghdrp)
1405 {
1406 struct mbuf *n, *mlast;
1407
1408 if (hlen > sizeof(struct ip6_hdr)) {
1409 n = m_copym(m0, sizeof(struct ip6_hdr),
1410 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1411 if (n == NULL)
1412 return (ENOBUFS);
1413 m->m_next = n;
1414 } else
1415 n = m;
1416
1417 /* Search for the last mbuf of unfragmentable part. */
1418 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1419 ;
1420
1421 if (M_WRITABLE(mlast) &&
1422 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1423 /* use the trailing space of the last mbuf for the fragment hdr */
1424 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1425 mlast->m_len);
1426 mlast->m_len += sizeof(struct ip6_frag);
1427 m->m_pkthdr.len += sizeof(struct ip6_frag);
1428 } else {
1429 /* allocate a new mbuf for the fragment header */
1430 struct mbuf *mfrg;
1431
1432 mfrg = m_get(M_NOWAIT, MT_DATA);
1433 if (mfrg == NULL)
1434 return (ENOBUFS);
1435 mfrg->m_len = sizeof(struct ip6_frag);
1436 *frghdrp = mtod(mfrg, struct ip6_frag *);
1437 mlast->m_next = mfrg;
1438 }
1439
1440 return (0);
1441 }
1442
1443 /*
1444 * Calculates IPv6 path mtu for destination @dst.
1445 * Resulting MTU is stored in @mtup.
1446 *
1447 * Returns 0 on success.
1448 */
1449 static int
1450 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1451 {
1452 struct epoch_tracker et;
1453 struct nhop_object *nh;
1454 struct in6_addr kdst;
1455 uint32_t scopeid;
1456 int error;
1457
1458 in6_splitscope(dst, &kdst, &scopeid);
1459
1460 NET_EPOCH_ENTER(et);
1461 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1462 if (nh != NULL)
1463 error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1464 else
1465 error = EHOSTUNREACH;
1466 NET_EPOCH_EXIT(et);
1467
1468 return (error);
1469 }
1470
1471 /*
1472 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1473 * and cached data in @ro_pmtu.
1474 * MTU from (successful) route lookup is saved (along with dst)
1475 * inside @ro_pmtu to avoid subsequent route lookups after packet
1476 * filter processing.
1477 *
1478 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1479 * Returns 0 on success.
1480 */
1481 static int
1482 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1483 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1484 int *alwaysfragp, u_int fibnum, u_int proto)
1485 {
1486 struct nhop_object *nh;
1487 struct in6_addr kdst;
1488 uint32_t scopeid;
1489 struct sockaddr_in6 *sa6_dst, sin6;
1490 u_long mtu;
1491
1492 NET_EPOCH_ASSERT();
1493
1494 mtu = 0;
1495 if (ro_pmtu == NULL || do_lookup) {
1496 /*
1497 * Here ro_pmtu has final destination address, while
1498 * ro might represent immediate destination.
1499 * Use ro_pmtu destination since mtu might differ.
1500 */
1501 if (ro_pmtu != NULL) {
1502 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1503 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1504 ro_pmtu->ro_mtu = 0;
1505 } else
1506 sa6_dst = &sin6;
1507
1508 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1509 bzero(sa6_dst, sizeof(*sa6_dst));
1510 sa6_dst->sin6_family = AF_INET6;
1511 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1512 sa6_dst->sin6_addr = *dst;
1513
1514 in6_splitscope(dst, &kdst, &scopeid);
1515 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1516 if (nh != NULL) {
1517 mtu = nh->nh_mtu;
1518 if (ro_pmtu != NULL)
1519 ro_pmtu->ro_mtu = mtu;
1520 }
1521 } else
1522 mtu = ro_pmtu->ro_mtu;
1523 }
1524
1525 if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1526 mtu = ro_pmtu->ro_nh->nh_mtu;
1527
1528 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1529 }
1530
1531 /*
1532 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1533 * hostcache data for @dst.
1534 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1535 *
1536 * Returns 0 on success.
1537 */
1538 static int
1539 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1540 u_long *mtup, int *alwaysfragp, u_int proto)
1541 {
1542 u_long mtu = 0;
1543 int alwaysfrag = 0;
1544 int error = 0;
1545
1546 if (rt_mtu > 0) {
1547 u_int32_t ifmtu;
1548 struct in_conninfo inc;
1549
1550 bzero(&inc, sizeof(inc));
1551 inc.inc_flags |= INC_ISIPV6;
1552 inc.inc6_faddr = *dst;
1553
1554 ifmtu = IN6_LINKMTU(ifp);
1555
1556 /* TCP is known to react to pmtu changes so skip hc */
1557 if (proto != IPPROTO_TCP)
1558 mtu = tcp_hc_getmtu(&inc);
1559
1560 if (mtu)
1561 mtu = min(mtu, rt_mtu);
1562 else
1563 mtu = rt_mtu;
1564 if (mtu == 0)
1565 mtu = ifmtu;
1566 else if (mtu < IPV6_MMTU) {
1567 /*
1568 * RFC2460 section 5, last paragraph:
1569 * if we record ICMPv6 too big message with
1570 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1571 * or smaller, with framgent header attached.
1572 * (fragment header is needed regardless from the
1573 * packet size, for translators to identify packets)
1574 */
1575 alwaysfrag = 1;
1576 mtu = IPV6_MMTU;
1577 }
1578 } else if (ifp) {
1579 mtu = IN6_LINKMTU(ifp);
1580 } else
1581 error = EHOSTUNREACH; /* XXX */
1582
1583 *mtup = mtu;
1584 if (alwaysfragp)
1585 *alwaysfragp = alwaysfrag;
1586 return (error);
1587 }
1588
1589 /*
1590 * IP6 socket option processing.
1591 */
1592 int
1593 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1594 {
1595 int optdatalen, uproto;
1596 void *optdata;
1597 struct inpcb *inp = sotoinpcb(so);
1598 int error, optval;
1599 int level, op, optname;
1600 int optlen;
1601 struct thread *td;
1602 #ifdef RSS
1603 uint32_t rss_bucket;
1604 int retval;
1605 #endif
1606
1607 /*
1608 * Don't use more than a quarter of mbuf clusters. N.B.:
1609 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1610 * on LP64 architectures, so cast to u_long to avoid undefined
1611 * behavior. ILP32 architectures cannot have nmbclusters
1612 * large enough to overflow for other reasons.
1613 */
1614 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1615
1616 level = sopt->sopt_level;
1617 op = sopt->sopt_dir;
1618 optname = sopt->sopt_name;
1619 optlen = sopt->sopt_valsize;
1620 td = sopt->sopt_td;
1621 error = 0;
1622 optval = 0;
1623 uproto = (int)so->so_proto->pr_protocol;
1624
1625 if (level != IPPROTO_IPV6) {
1626 error = EINVAL;
1627
1628 if (sopt->sopt_level == SOL_SOCKET &&
1629 sopt->sopt_dir == SOPT_SET) {
1630 switch (sopt->sopt_name) {
1631 case SO_REUSEADDR:
1632 INP_WLOCK(inp);
1633 if ((so->so_options & SO_REUSEADDR) != 0)
1634 inp->inp_flags2 |= INP_REUSEADDR;
1635 else
1636 inp->inp_flags2 &= ~INP_REUSEADDR;
1637 INP_WUNLOCK(inp);
1638 error = 0;
1639 break;
1640 case SO_REUSEPORT:
1641 INP_WLOCK(inp);
1642 if ((so->so_options & SO_REUSEPORT) != 0)
1643 inp->inp_flags2 |= INP_REUSEPORT;
1644 else
1645 inp->inp_flags2 &= ~INP_REUSEPORT;
1646 INP_WUNLOCK(inp);
1647 error = 0;
1648 break;
1649 case SO_REUSEPORT_LB:
1650 INP_WLOCK(inp);
1651 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1652 inp->inp_flags2 |= INP_REUSEPORT_LB;
1653 else
1654 inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1655 INP_WUNLOCK(inp);
1656 error = 0;
1657 break;
1658 case SO_SETFIB:
1659 INP_WLOCK(inp);
1660 inp->inp_inc.inc_fibnum = so->so_fibnum;
1661 INP_WUNLOCK(inp);
1662 error = 0;
1663 break;
1664 case SO_MAX_PACING_RATE:
1665 #ifdef RATELIMIT
1666 INP_WLOCK(inp);
1667 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1668 INP_WUNLOCK(inp);
1669 error = 0;
1670 #else
1671 error = EOPNOTSUPP;
1672 #endif
1673 break;
1674 default:
1675 break;
1676 }
1677 }
1678 } else { /* level == IPPROTO_IPV6 */
1679 switch (op) {
1680 case SOPT_SET:
1681 switch (optname) {
1682 case IPV6_2292PKTOPTIONS:
1683 #ifdef IPV6_PKTOPTIONS
1684 case IPV6_PKTOPTIONS:
1685 #endif
1686 {
1687 struct mbuf *m;
1688
1689 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1690 printf("ip6_ctloutput: mbuf limit hit\n");
1691 error = ENOBUFS;
1692 break;
1693 }
1694
1695 error = soopt_getm(sopt, &m); /* XXX */
1696 if (error != 0)
1697 break;
1698 error = soopt_mcopyin(sopt, m); /* XXX */
1699 if (error != 0)
1700 break;
1701 INP_WLOCK(inp);
1702 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1703 so, sopt);
1704 INP_WUNLOCK(inp);
1705 m_freem(m); /* XXX */
1706 break;
1707 }
1708
1709 /*
1710 * Use of some Hop-by-Hop options or some
1711 * Destination options, might require special
1712 * privilege. That is, normal applications
1713 * (without special privilege) might be forbidden
1714 * from setting certain options in outgoing packets,
1715 * and might never see certain options in received
1716 * packets. [RFC 2292 Section 6]
1717 * KAME specific note:
1718 * KAME prevents non-privileged users from sending or
1719 * receiving ANY hbh/dst options in order to avoid
1720 * overhead of parsing options in the kernel.
1721 */
1722 case IPV6_RECVHOPOPTS:
1723 case IPV6_RECVDSTOPTS:
1724 case IPV6_RECVRTHDRDSTOPTS:
1725 if (td != NULL) {
1726 error = priv_check(td,
1727 PRIV_NETINET_SETHDROPTS);
1728 if (error)
1729 break;
1730 }
1731 /* FALLTHROUGH */
1732 case IPV6_UNICAST_HOPS:
1733 case IPV6_HOPLIMIT:
1734
1735 case IPV6_RECVPKTINFO:
1736 case IPV6_RECVHOPLIMIT:
1737 case IPV6_RECVRTHDR:
1738 case IPV6_RECVPATHMTU:
1739 case IPV6_RECVTCLASS:
1740 case IPV6_RECVFLOWID:
1741 #ifdef RSS
1742 case IPV6_RECVRSSBUCKETID:
1743 #endif
1744 case IPV6_V6ONLY:
1745 case IPV6_AUTOFLOWLABEL:
1746 case IPV6_ORIGDSTADDR:
1747 case IPV6_BINDANY:
1748 case IPV6_BINDMULTI:
1749 #ifdef RSS
1750 case IPV6_RSS_LISTEN_BUCKET:
1751 #endif
1752 case IPV6_VLAN_PCP:
1753 if (optname == IPV6_BINDANY && td != NULL) {
1754 error = priv_check(td,
1755 PRIV_NETINET_BINDANY);
1756 if (error)
1757 break;
1758 }
1759
1760 if (optlen != sizeof(int)) {
1761 error = EINVAL;
1762 break;
1763 }
1764 error = sooptcopyin(sopt, &optval,
1765 sizeof optval, sizeof optval);
1766 if (error)
1767 break;
1768 switch (optname) {
1769 case IPV6_UNICAST_HOPS:
1770 if (optval < -1 || optval >= 256)
1771 error = EINVAL;
1772 else {
1773 /* -1 = kernel default */
1774 inp->in6p_hops = optval;
1775 if ((inp->inp_vflag &
1776 INP_IPV4) != 0)
1777 inp->inp_ip_ttl = optval;
1778 }
1779 break;
1780 #define OPTSET(bit) \
1781 do { \
1782 INP_WLOCK(inp); \
1783 if (optval) \
1784 inp->inp_flags |= (bit); \
1785 else \
1786 inp->inp_flags &= ~(bit); \
1787 INP_WUNLOCK(inp); \
1788 } while (/*CONSTCOND*/ 0)
1789 #define OPTSET2292(bit) \
1790 do { \
1791 INP_WLOCK(inp); \
1792 inp->inp_flags |= IN6P_RFC2292; \
1793 if (optval) \
1794 inp->inp_flags |= (bit); \
1795 else \
1796 inp->inp_flags &= ~(bit); \
1797 INP_WUNLOCK(inp); \
1798 } while (/*CONSTCOND*/ 0)
1799 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1800
1801 #define OPTSET2_N(bit, val) do { \
1802 if (val) \
1803 inp->inp_flags2 |= bit; \
1804 else \
1805 inp->inp_flags2 &= ~bit; \
1806 } while (0)
1807 #define OPTSET2(bit, val) do { \
1808 INP_WLOCK(inp); \
1809 OPTSET2_N(bit, val); \
1810 INP_WUNLOCK(inp); \
1811 } while (0)
1812 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1813 #define OPTSET2292_EXCLUSIVE(bit) \
1814 do { \
1815 INP_WLOCK(inp); \
1816 if (OPTBIT(IN6P_RFC2292)) { \
1817 error = EINVAL; \
1818 } else { \
1819 if (optval) \
1820 inp->inp_flags |= (bit); \
1821 else \
1822 inp->inp_flags &= ~(bit); \
1823 } \
1824 INP_WUNLOCK(inp); \
1825 } while (/*CONSTCOND*/ 0)
1826
1827 case IPV6_RECVPKTINFO:
1828 OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1829 break;
1830
1831 case IPV6_HOPLIMIT:
1832 {
1833 struct ip6_pktopts **optp;
1834
1835 /* cannot mix with RFC2292 */
1836 if (OPTBIT(IN6P_RFC2292)) {
1837 error = EINVAL;
1838 break;
1839 }
1840 INP_WLOCK(inp);
1841 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1842 INP_WUNLOCK(inp);
1843 return (ECONNRESET);
1844 }
1845 optp = &inp->in6p_outputopts;
1846 error = ip6_pcbopt(IPV6_HOPLIMIT,
1847 (u_char *)&optval, sizeof(optval),
1848 optp, (td != NULL) ? td->td_ucred :
1849 NULL, uproto);
1850 INP_WUNLOCK(inp);
1851 break;
1852 }
1853
1854 case IPV6_RECVHOPLIMIT:
1855 OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1856 break;
1857
1858 case IPV6_RECVHOPOPTS:
1859 OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1860 break;
1861
1862 case IPV6_RECVDSTOPTS:
1863 OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1864 break;
1865
1866 case IPV6_RECVRTHDRDSTOPTS:
1867 OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1868 break;
1869
1870 case IPV6_RECVRTHDR:
1871 OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1872 break;
1873
1874 case IPV6_RECVPATHMTU:
1875 /*
1876 * We ignore this option for TCP
1877 * sockets.
1878 * (RFC3542 leaves this case
1879 * unspecified.)
1880 */
1881 if (uproto != IPPROTO_TCP)
1882 OPTSET(IN6P_MTU);
1883 break;
1884
1885 case IPV6_RECVFLOWID:
1886 OPTSET2(INP_RECVFLOWID, optval);
1887 break;
1888
1889 #ifdef RSS
1890 case IPV6_RECVRSSBUCKETID:
1891 OPTSET2(INP_RECVRSSBUCKETID, optval);
1892 break;
1893 #endif
1894
1895 case IPV6_V6ONLY:
1896 INP_WLOCK(inp);
1897 if (inp->inp_lport ||
1898 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1899 /*
1900 * The socket is already bound.
1901 */
1902 INP_WUNLOCK(inp);
1903 error = EINVAL;
1904 break;
1905 }
1906 if (optval) {
1907 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1908 inp->inp_vflag &= ~INP_IPV4;
1909 } else {
1910 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1911 inp->inp_vflag |= INP_IPV4;
1912 }
1913 INP_WUNLOCK(inp);
1914 break;
1915 case IPV6_RECVTCLASS:
1916 /* cannot mix with RFC2292 XXX */
1917 OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1918 break;
1919 case IPV6_AUTOFLOWLABEL:
1920 OPTSET(IN6P_AUTOFLOWLABEL);
1921 break;
1922
1923 case IPV6_ORIGDSTADDR:
1924 OPTSET2(INP_ORIGDSTADDR, optval);
1925 break;
1926 case IPV6_BINDANY:
1927 OPTSET(INP_BINDANY);
1928 break;
1929
1930 case IPV6_BINDMULTI:
1931 OPTSET2(INP_BINDMULTI, optval);
1932 break;
1933 #ifdef RSS
1934 case IPV6_RSS_LISTEN_BUCKET:
1935 if ((optval >= 0) &&
1936 (optval < rss_getnumbuckets())) {
1937 INP_WLOCK(inp);
1938 inp->inp_rss_listen_bucket = optval;
1939 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1940 INP_WUNLOCK(inp);
1941 } else {
1942 error = EINVAL;
1943 }
1944 break;
1945 #endif
1946 case IPV6_VLAN_PCP:
1947 if ((optval >= -1) && (optval <=
1948 (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1949 if (optval == -1) {
1950 INP_WLOCK(inp);
1951 inp->inp_flags2 &=
1952 ~(INP_2PCP_SET |
1953 INP_2PCP_MASK);
1954 INP_WUNLOCK(inp);
1955 } else {
1956 INP_WLOCK(inp);
1957 inp->inp_flags2 |=
1958 INP_2PCP_SET;
1959 inp->inp_flags2 &=
1960 ~INP_2PCP_MASK;
1961 inp->inp_flags2 |=
1962 optval <<
1963 INP_2PCP_SHIFT;
1964 INP_WUNLOCK(inp);
1965 }
1966 } else
1967 error = EINVAL;
1968 break;
1969 }
1970 break;
1971
1972 case IPV6_TCLASS:
1973 case IPV6_DONTFRAG:
1974 case IPV6_USE_MIN_MTU:
1975 case IPV6_PREFER_TEMPADDR:
1976 if (optlen != sizeof(optval)) {
1977 error = EINVAL;
1978 break;
1979 }
1980 error = sooptcopyin(sopt, &optval,
1981 sizeof optval, sizeof optval);
1982 if (error)
1983 break;
1984 {
1985 struct ip6_pktopts **optp;
1986 INP_WLOCK(inp);
1987 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1988 INP_WUNLOCK(inp);
1989 return (ECONNRESET);
1990 }
1991 optp = &inp->in6p_outputopts;
1992 error = ip6_pcbopt(optname,
1993 (u_char *)&optval, sizeof(optval),
1994 optp, (td != NULL) ? td->td_ucred :
1995 NULL, uproto);
1996 INP_WUNLOCK(inp);
1997 break;
1998 }
1999
2000 case IPV6_2292PKTINFO:
2001 case IPV6_2292HOPLIMIT:
2002 case IPV6_2292HOPOPTS:
2003 case IPV6_2292DSTOPTS:
2004 case IPV6_2292RTHDR:
2005 /* RFC 2292 */
2006 if (optlen != sizeof(int)) {
2007 error = EINVAL;
2008 break;
2009 }
2010 error = sooptcopyin(sopt, &optval,
2011 sizeof optval, sizeof optval);
2012 if (error)
2013 break;
2014 switch (optname) {
2015 case IPV6_2292PKTINFO:
2016 OPTSET2292(IN6P_PKTINFO);
2017 break;
2018 case IPV6_2292HOPLIMIT:
2019 OPTSET2292(IN6P_HOPLIMIT);
2020 break;
2021 case IPV6_2292HOPOPTS:
2022 /*
2023 * Check super-user privilege.
2024 * See comments for IPV6_RECVHOPOPTS.
2025 */
2026 if (td != NULL) {
2027 error = priv_check(td,
2028 PRIV_NETINET_SETHDROPTS);
2029 if (error)
2030 return (error);
2031 }
2032 OPTSET2292(IN6P_HOPOPTS);
2033 break;
2034 case IPV6_2292DSTOPTS:
2035 if (td != NULL) {
2036 error = priv_check(td,
2037 PRIV_NETINET_SETHDROPTS);
2038 if (error)
2039 return (error);
2040 }
2041 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2042 break;
2043 case IPV6_2292RTHDR:
2044 OPTSET2292(IN6P_RTHDR);
2045 break;
2046 }
2047 break;
2048 case IPV6_PKTINFO:
2049 case IPV6_HOPOPTS:
2050 case IPV6_RTHDR:
2051 case IPV6_DSTOPTS:
2052 case IPV6_RTHDRDSTOPTS:
2053 case IPV6_NEXTHOP:
2054 {
2055 /* new advanced API (RFC3542) */
2056 u_char *optbuf;
2057 u_char optbuf_storage[MCLBYTES];
2058 int optlen;
2059 struct ip6_pktopts **optp;
2060
2061 /* cannot mix with RFC2292 */
2062 if (OPTBIT(IN6P_RFC2292)) {
2063 error = EINVAL;
2064 break;
2065 }
2066
2067 /*
2068 * We only ensure valsize is not too large
2069 * here. Further validation will be done
2070 * later.
2071 */
2072 error = sooptcopyin(sopt, optbuf_storage,
2073 sizeof(optbuf_storage), 0);
2074 if (error)
2075 break;
2076 optlen = sopt->sopt_valsize;
2077 optbuf = optbuf_storage;
2078 INP_WLOCK(inp);
2079 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2080 INP_WUNLOCK(inp);
2081 return (ECONNRESET);
2082 }
2083 optp = &inp->in6p_outputopts;
2084 error = ip6_pcbopt(optname, optbuf, optlen,
2085 optp, (td != NULL) ? td->td_ucred : NULL,
2086 uproto);
2087 INP_WUNLOCK(inp);
2088 break;
2089 }
2090 #undef OPTSET
2091
2092 case IPV6_MULTICAST_IF:
2093 case IPV6_MULTICAST_HOPS:
2094 case IPV6_MULTICAST_LOOP:
2095 case IPV6_JOIN_GROUP:
2096 case IPV6_LEAVE_GROUP:
2097 case IPV6_MSFILTER:
2098 case MCAST_BLOCK_SOURCE:
2099 case MCAST_UNBLOCK_SOURCE:
2100 case MCAST_JOIN_GROUP:
2101 case MCAST_LEAVE_GROUP:
2102 case MCAST_JOIN_SOURCE_GROUP:
2103 case MCAST_LEAVE_SOURCE_GROUP:
2104 error = ip6_setmoptions(inp, sopt);
2105 break;
2106
2107 case IPV6_PORTRANGE:
2108 error = sooptcopyin(sopt, &optval,
2109 sizeof optval, sizeof optval);
2110 if (error)
2111 break;
2112
2113 INP_WLOCK(inp);
2114 switch (optval) {
2115 case IPV6_PORTRANGE_DEFAULT:
2116 inp->inp_flags &= ~(INP_LOWPORT);
2117 inp->inp_flags &= ~(INP_HIGHPORT);
2118 break;
2119
2120 case IPV6_PORTRANGE_HIGH:
2121 inp->inp_flags &= ~(INP_LOWPORT);
2122 inp->inp_flags |= INP_HIGHPORT;
2123 break;
2124
2125 case IPV6_PORTRANGE_LOW:
2126 inp->inp_flags &= ~(INP_HIGHPORT);
2127 inp->inp_flags |= INP_LOWPORT;
2128 break;
2129
2130 default:
2131 error = EINVAL;
2132 break;
2133 }
2134 INP_WUNLOCK(inp);
2135 break;
2136
2137 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2138 case IPV6_IPSEC_POLICY:
2139 if (IPSEC_ENABLED(ipv6)) {
2140 error = IPSEC_PCBCTL(ipv6, inp, sopt);
2141 break;
2142 }
2143 /* FALLTHROUGH */
2144 #endif /* IPSEC */
2145
2146 default:
2147 error = ENOPROTOOPT;
2148 break;
2149 }
2150 break;
2151
2152 case SOPT_GET:
2153 switch (optname) {
2154 case IPV6_2292PKTOPTIONS:
2155 #ifdef IPV6_PKTOPTIONS
2156 case IPV6_PKTOPTIONS:
2157 #endif
2158 /*
2159 * RFC3542 (effectively) deprecated the
2160 * semantics of the 2292-style pktoptions.
2161 * Since it was not reliable in nature (i.e.,
2162 * applications had to expect the lack of some
2163 * information after all), it would make sense
2164 * to simplify this part by always returning
2165 * empty data.
2166 */
2167 sopt->sopt_valsize = 0;
2168 break;
2169
2170 case IPV6_RECVHOPOPTS:
2171 case IPV6_RECVDSTOPTS:
2172 case IPV6_RECVRTHDRDSTOPTS:
2173 case IPV6_UNICAST_HOPS:
2174 case IPV6_RECVPKTINFO:
2175 case IPV6_RECVHOPLIMIT:
2176 case IPV6_RECVRTHDR:
2177 case IPV6_RECVPATHMTU:
2178
2179 case IPV6_V6ONLY:
2180 case IPV6_PORTRANGE:
2181 case IPV6_RECVTCLASS:
2182 case IPV6_AUTOFLOWLABEL:
2183 case IPV6_BINDANY:
2184 case IPV6_FLOWID:
2185 case IPV6_FLOWTYPE:
2186 case IPV6_RECVFLOWID:
2187 #ifdef RSS
2188 case IPV6_RSSBUCKETID:
2189 case IPV6_RECVRSSBUCKETID:
2190 #endif
2191 case IPV6_BINDMULTI:
2192 case IPV6_VLAN_PCP:
2193 switch (optname) {
2194 case IPV6_RECVHOPOPTS:
2195 optval = OPTBIT(IN6P_HOPOPTS);
2196 break;
2197
2198 case IPV6_RECVDSTOPTS:
2199 optval = OPTBIT(IN6P_DSTOPTS);
2200 break;
2201
2202 case IPV6_RECVRTHDRDSTOPTS:
2203 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2204 break;
2205
2206 case IPV6_UNICAST_HOPS:
2207 optval = inp->in6p_hops;
2208 break;
2209
2210 case IPV6_RECVPKTINFO:
2211 optval = OPTBIT(IN6P_PKTINFO);
2212 break;
2213
2214 case IPV6_RECVHOPLIMIT:
2215 optval = OPTBIT(IN6P_HOPLIMIT);
2216 break;
2217
2218 case IPV6_RECVRTHDR:
2219 optval = OPTBIT(IN6P_RTHDR);
2220 break;
2221
2222 case IPV6_RECVPATHMTU:
2223 optval = OPTBIT(IN6P_MTU);
2224 break;
2225
2226 case IPV6_V6ONLY:
2227 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2228 break;
2229
2230 case IPV6_PORTRANGE:
2231 {
2232 int flags;
2233 flags = inp->inp_flags;
2234 if (flags & INP_HIGHPORT)
2235 optval = IPV6_PORTRANGE_HIGH;
2236 else if (flags & INP_LOWPORT)
2237 optval = IPV6_PORTRANGE_LOW;
2238 else
2239 optval = 0;
2240 break;
2241 }
2242 case IPV6_RECVTCLASS:
2243 optval = OPTBIT(IN6P_TCLASS);
2244 break;
2245
2246 case IPV6_AUTOFLOWLABEL:
2247 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2248 break;
2249
2250 case IPV6_ORIGDSTADDR:
2251 optval = OPTBIT2(INP_ORIGDSTADDR);
2252 break;
2253
2254 case IPV6_BINDANY:
2255 optval = OPTBIT(INP_BINDANY);
2256 break;
2257
2258 case IPV6_FLOWID:
2259 optval = inp->inp_flowid;
2260 break;
2261
2262 case IPV6_FLOWTYPE:
2263 optval = inp->inp_flowtype;
2264 break;
2265
2266 case IPV6_RECVFLOWID:
2267 optval = OPTBIT2(INP_RECVFLOWID);
2268 break;
2269 #ifdef RSS
2270 case IPV6_RSSBUCKETID:
2271 retval =
2272 rss_hash2bucket(inp->inp_flowid,
2273 inp->inp_flowtype,
2274 &rss_bucket);
2275 if (retval == 0)
2276 optval = rss_bucket;
2277 else
2278 error = EINVAL;
2279 break;
2280
2281 case IPV6_RECVRSSBUCKETID:
2282 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2283 break;
2284 #endif
2285
2286 case IPV6_BINDMULTI:
2287 optval = OPTBIT2(INP_BINDMULTI);
2288 break;
2289
2290 case IPV6_VLAN_PCP:
2291 if (OPTBIT2(INP_2PCP_SET)) {
2292 optval = (inp->inp_flags2 &
2293 INP_2PCP_MASK) >>
2294 INP_2PCP_SHIFT;
2295 } else {
2296 optval = -1;
2297 }
2298 break;
2299 }
2300
2301 if (error)
2302 break;
2303 error = sooptcopyout(sopt, &optval,
2304 sizeof optval);
2305 break;
2306
2307 case IPV6_PATHMTU:
2308 {
2309 u_long pmtu = 0;
2310 struct ip6_mtuinfo mtuinfo;
2311 struct in6_addr addr;
2312
2313 if (!(so->so_state & SS_ISCONNECTED))
2314 return (ENOTCONN);
2315 /*
2316 * XXX: we dot not consider the case of source
2317 * routing, or optional information to specify
2318 * the outgoing interface.
2319 * Copy faddr out of inp to avoid holding lock
2320 * on inp during route lookup.
2321 */
2322 INP_RLOCK(inp);
2323 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2324 INP_RUNLOCK(inp);
2325 error = ip6_getpmtu_ctl(so->so_fibnum,
2326 &addr, &pmtu);
2327 if (error)
2328 break;
2329 if (pmtu > IPV6_MAXPACKET)
2330 pmtu = IPV6_MAXPACKET;
2331
2332 bzero(&mtuinfo, sizeof(mtuinfo));
2333 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2334 optdata = (void *)&mtuinfo;
2335 optdatalen = sizeof(mtuinfo);
2336 error = sooptcopyout(sopt, optdata,
2337 optdatalen);
2338 break;
2339 }
2340
2341 case IPV6_2292PKTINFO:
2342 case IPV6_2292HOPLIMIT:
2343 case IPV6_2292HOPOPTS:
2344 case IPV6_2292RTHDR:
2345 case IPV6_2292DSTOPTS:
2346 switch (optname) {
2347 case IPV6_2292PKTINFO:
2348 optval = OPTBIT(IN6P_PKTINFO);
2349 break;
2350 case IPV6_2292HOPLIMIT:
2351 optval = OPTBIT(IN6P_HOPLIMIT);
2352 break;
2353 case IPV6_2292HOPOPTS:
2354 optval = OPTBIT(IN6P_HOPOPTS);
2355 break;
2356 case IPV6_2292RTHDR:
2357 optval = OPTBIT(IN6P_RTHDR);
2358 break;
2359 case IPV6_2292DSTOPTS:
2360 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2361 break;
2362 }
2363 error = sooptcopyout(sopt, &optval,
2364 sizeof optval);
2365 break;
2366 case IPV6_PKTINFO:
2367 case IPV6_HOPOPTS:
2368 case IPV6_RTHDR:
2369 case IPV6_DSTOPTS:
2370 case IPV6_RTHDRDSTOPTS:
2371 case IPV6_NEXTHOP:
2372 case IPV6_TCLASS:
2373 case IPV6_DONTFRAG:
2374 case IPV6_USE_MIN_MTU:
2375 case IPV6_PREFER_TEMPADDR:
2376 error = ip6_getpcbopt(inp, optname, sopt);
2377 break;
2378
2379 case IPV6_MULTICAST_IF:
2380 case IPV6_MULTICAST_HOPS:
2381 case IPV6_MULTICAST_LOOP:
2382 case IPV6_MSFILTER:
2383 error = ip6_getmoptions(inp, sopt);
2384 break;
2385
2386 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2387 case IPV6_IPSEC_POLICY:
2388 if (IPSEC_ENABLED(ipv6)) {
2389 error = IPSEC_PCBCTL(ipv6, inp, sopt);
2390 break;
2391 }
2392 /* FALLTHROUGH */
2393 #endif /* IPSEC */
2394 default:
2395 error = ENOPROTOOPT;
2396 break;
2397 }
2398 break;
2399 }
2400 }
2401 return (error);
2402 }
2403
2404 int
2405 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2406 {
2407 int error = 0, optval, optlen;
2408 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2409 struct inpcb *inp = sotoinpcb(so);
2410 int level, op, optname;
2411
2412 level = sopt->sopt_level;
2413 op = sopt->sopt_dir;
2414 optname = sopt->sopt_name;
2415 optlen = sopt->sopt_valsize;
2416
2417 if (level != IPPROTO_IPV6) {
2418 return (EINVAL);
2419 }
2420
2421 switch (optname) {
2422 case IPV6_CHECKSUM:
2423 /*
2424 * For ICMPv6 sockets, no modification allowed for checksum
2425 * offset, permit "no change" values to help existing apps.
2426 *
2427 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2428 * for an ICMPv6 socket will fail."
2429 * The current behavior does not meet RFC3542.
2430 */
2431 switch (op) {
2432 case SOPT_SET:
2433 if (optlen != sizeof(int)) {
2434 error = EINVAL;
2435 break;
2436 }
2437 error = sooptcopyin(sopt, &optval, sizeof(optval),
2438 sizeof(optval));
2439 if (error)
2440 break;
2441 if (optval < -1 || (optval % 2) != 0) {
2442 /*
2443 * The API assumes non-negative even offset
2444 * values or -1 as a special value.
2445 */
2446 error = EINVAL;
2447 } else if (so->so_proto->pr_protocol ==
2448 IPPROTO_ICMPV6) {
2449 if (optval != icmp6off)
2450 error = EINVAL;
2451 } else
2452 inp->in6p_cksum = optval;
2453 break;
2454
2455 case SOPT_GET:
2456 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2457 optval = icmp6off;
2458 else
2459 optval = inp->in6p_cksum;
2460
2461 error = sooptcopyout(sopt, &optval, sizeof(optval));
2462 break;
2463
2464 default:
2465 error = EINVAL;
2466 break;
2467 }
2468 break;
2469
2470 default:
2471 error = ENOPROTOOPT;
2472 break;
2473 }
2474
2475 return (error);
2476 }
2477
2478 /*
2479 * Set up IP6 options in pcb for insertion in output packets or
2480 * specifying behavior of outgoing packets.
2481 */
2482 static int
2483 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2484 struct socket *so, struct sockopt *sopt)
2485 {
2486 struct ip6_pktopts *opt = *pktopt;
2487 int error = 0;
2488 struct thread *td = sopt->sopt_td;
2489
2490 /* turn off any old options. */
2491 if (opt) {
2492 #ifdef DIAGNOSTIC
2493 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2494 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2495 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2496 printf("ip6_pcbopts: all specified options are cleared.\n");
2497 #endif
2498 ip6_clearpktopts(opt, -1);
2499 } else {
2500 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2501 if (opt == NULL)
2502 return (ENOMEM);
2503 }
2504 *pktopt = NULL;
2505
2506 if (!m || m->m_len == 0) {
2507 /*
2508 * Only turning off any previous options, regardless of
2509 * whether the opt is just created or given.
2510 */
2511 free(opt, M_IP6OPT);
2512 return (0);
2513 }
2514
2515 /* set options specified by user. */
2516 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2517 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2518 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2519 free(opt, M_IP6OPT);
2520 return (error);
2521 }
2522 *pktopt = opt;
2523 return (0);
2524 }
2525
2526 /*
2527 * initialize ip6_pktopts. beware that there are non-zero default values in
2528 * the struct.
2529 */
2530 void
2531 ip6_initpktopts(struct ip6_pktopts *opt)
2532 {
2533
2534 bzero(opt, sizeof(*opt));
2535 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2536 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2537 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2538 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2539 }
2540
2541 static int
2542 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2543 struct ucred *cred, int uproto)
2544 {
2545 struct ip6_pktopts *opt;
2546
2547 if (*pktopt == NULL) {
2548 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2549 M_NOWAIT);
2550 if (*pktopt == NULL)
2551 return (ENOBUFS);
2552 ip6_initpktopts(*pktopt);
2553 }
2554 opt = *pktopt;
2555
2556 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2557 }
2558
2559 #define GET_PKTOPT_VAR(field, lenexpr) do { \
2560 if (pktopt && pktopt->field) { \
2561 INP_RUNLOCK(inp); \
2562 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \
2563 malloc_optdata = true; \
2564 INP_RLOCK(inp); \
2565 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
2566 INP_RUNLOCK(inp); \
2567 free(optdata, M_TEMP); \
2568 return (ECONNRESET); \
2569 } \
2570 pktopt = inp->in6p_outputopts; \
2571 if (pktopt && pktopt->field) { \
2572 optdatalen = min(lenexpr, sopt->sopt_valsize); \
2573 bcopy(&pktopt->field, optdata, optdatalen); \
2574 } else { \
2575 free(optdata, M_TEMP); \
2576 optdata = NULL; \
2577 malloc_optdata = false; \
2578 } \
2579 } \
2580 } while(0)
2581
2582 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field, \
2583 (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2584
2585 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field, \
2586 pktopt->field->sa_len)
2587
2588 static int
2589 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2590 {
2591 void *optdata = NULL;
2592 bool malloc_optdata = false;
2593 int optdatalen = 0;
2594 int error = 0;
2595 struct in6_pktinfo null_pktinfo;
2596 int deftclass = 0, on;
2597 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2598 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2599 struct ip6_pktopts *pktopt;
2600
2601 INP_RLOCK(inp);
2602 pktopt = inp->in6p_outputopts;
2603
2604 switch (optname) {
2605 case IPV6_PKTINFO:
2606 optdata = (void *)&null_pktinfo;
2607 if (pktopt && pktopt->ip6po_pktinfo) {
2608 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2609 sizeof(null_pktinfo));
2610 in6_clearscope(&null_pktinfo.ipi6_addr);
2611 } else {
2612 /* XXX: we don't have to do this every time... */
2613 bzero(&null_pktinfo, sizeof(null_pktinfo));
2614 }
2615 optdatalen = sizeof(struct in6_pktinfo);
2616 break;
2617 case IPV6_TCLASS:
2618 if (pktopt && pktopt->ip6po_tclass >= 0)
2619 deftclass = pktopt->ip6po_tclass;
2620 optdata = (void *)&deftclass;
2621 optdatalen = sizeof(int);
2622 break;
2623 case IPV6_HOPOPTS:
2624 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2625 break;
2626 case IPV6_RTHDR:
2627 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2628 break;
2629 case IPV6_RTHDRDSTOPTS:
2630 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2631 break;
2632 case IPV6_DSTOPTS:
2633 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2634 break;
2635 case IPV6_NEXTHOP:
2636 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2637 break;
2638 case IPV6_USE_MIN_MTU:
2639 if (pktopt)
2640 defminmtu = pktopt->ip6po_minmtu;
2641 optdata = (void *)&defminmtu;
2642 optdatalen = sizeof(int);
2643 break;
2644 case IPV6_DONTFRAG:
2645 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2646 on = 1;
2647 else
2648 on = 0;
2649 optdata = (void *)&on;
2650 optdatalen = sizeof(on);
2651 break;
2652 case IPV6_PREFER_TEMPADDR:
2653 if (pktopt)
2654 defpreftemp = pktopt->ip6po_prefer_tempaddr;
2655 optdata = (void *)&defpreftemp;
2656 optdatalen = sizeof(int);
2657 break;
2658 default: /* should not happen */
2659 #ifdef DIAGNOSTIC
2660 panic("ip6_getpcbopt: unexpected option\n");
2661 #endif
2662 INP_RUNLOCK(inp);
2663 return (ENOPROTOOPT);
2664 }
2665 INP_RUNLOCK(inp);
2666
2667 error = sooptcopyout(sopt, optdata, optdatalen);
2668 if (malloc_optdata)
2669 free(optdata, M_TEMP);
2670
2671 return (error);
2672 }
2673
2674 void
2675 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2676 {
2677 if (pktopt == NULL)
2678 return;
2679
2680 if (optname == -1 || optname == IPV6_PKTINFO) {
2681 if (pktopt->ip6po_pktinfo)
2682 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2683 pktopt->ip6po_pktinfo = NULL;
2684 }
2685 if (optname == -1 || optname == IPV6_HOPLIMIT)
2686 pktopt->ip6po_hlim = -1;
2687 if (optname == -1 || optname == IPV6_TCLASS)
2688 pktopt->ip6po_tclass = -1;
2689 if (optname == -1 || optname == IPV6_NEXTHOP) {
2690 if (pktopt->ip6po_nextroute.ro_nh) {
2691 NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2692 pktopt->ip6po_nextroute.ro_nh = NULL;
2693 }
2694 if (pktopt->ip6po_nexthop)
2695 free(pktopt->ip6po_nexthop, M_IP6OPT);
2696 pktopt->ip6po_nexthop = NULL;
2697 }
2698 if (optname == -1 || optname == IPV6_HOPOPTS) {
2699 if (pktopt->ip6po_hbh)
2700 free(pktopt->ip6po_hbh, M_IP6OPT);
2701 pktopt->ip6po_hbh = NULL;
2702 }
2703 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2704 if (pktopt->ip6po_dest1)
2705 free(pktopt->ip6po_dest1, M_IP6OPT);
2706 pktopt->ip6po_dest1 = NULL;
2707 }
2708 if (optname == -1 || optname == IPV6_RTHDR) {
2709 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2710 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2711 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2712 if (pktopt->ip6po_route.ro_nh) {
2713 NH_FREE(pktopt->ip6po_route.ro_nh);
2714 pktopt->ip6po_route.ro_nh = NULL;
2715 }
2716 }
2717 if (optname == -1 || optname == IPV6_DSTOPTS) {
2718 if (pktopt->ip6po_dest2)
2719 free(pktopt->ip6po_dest2, M_IP6OPT);
2720 pktopt->ip6po_dest2 = NULL;
2721 }
2722 }
2723
2724 #define PKTOPT_EXTHDRCPY(type) \
2725 do {\
2726 if (src->type) {\
2727 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2728 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2729 if (dst->type == NULL)\
2730 goto bad;\
2731 bcopy(src->type, dst->type, hlen);\
2732 }\
2733 } while (/*CONSTCOND*/ 0)
2734
2735 static int
2736 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2737 {
2738 if (dst == NULL || src == NULL) {
2739 printf("ip6_clearpktopts: invalid argument\n");
2740 return (EINVAL);
2741 }
2742
2743 dst->ip6po_hlim = src->ip6po_hlim;
2744 dst->ip6po_tclass = src->ip6po_tclass;
2745 dst->ip6po_flags = src->ip6po_flags;
2746 dst->ip6po_minmtu = src->ip6po_minmtu;
2747 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2748 if (src->ip6po_pktinfo) {
2749 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2750 M_IP6OPT, canwait);
2751 if (dst->ip6po_pktinfo == NULL)
2752 goto bad;
2753 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2754 }
2755 if (src->ip6po_nexthop) {
2756 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2757 M_IP6OPT, canwait);
2758 if (dst->ip6po_nexthop == NULL)
2759 goto bad;
2760 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2761 src->ip6po_nexthop->sa_len);
2762 }
2763 PKTOPT_EXTHDRCPY(ip6po_hbh);
2764 PKTOPT_EXTHDRCPY(ip6po_dest1);
2765 PKTOPT_EXTHDRCPY(ip6po_dest2);
2766 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2767 return (0);
2768
2769 bad:
2770 ip6_clearpktopts(dst, -1);
2771 return (ENOBUFS);
2772 }
2773 #undef PKTOPT_EXTHDRCPY
2774
2775 struct ip6_pktopts *
2776 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2777 {
2778 int error;
2779 struct ip6_pktopts *dst;
2780
2781 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2782 if (dst == NULL)
2783 return (NULL);
2784 ip6_initpktopts(dst);
2785
2786 if ((error = copypktopts(dst, src, canwait)) != 0) {
2787 free(dst, M_IP6OPT);
2788 return (NULL);
2789 }
2790
2791 return (dst);
2792 }
2793
2794 void
2795 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2796 {
2797 if (pktopt == NULL)
2798 return;
2799
2800 ip6_clearpktopts(pktopt, -1);
2801
2802 free(pktopt, M_IP6OPT);
2803 }
2804
2805 /*
2806 * Set IPv6 outgoing packet options based on advanced API.
2807 */
2808 int
2809 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2810 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2811 {
2812 struct cmsghdr *cm = NULL;
2813
2814 if (control == NULL || opt == NULL)
2815 return (EINVAL);
2816
2817 ip6_initpktopts(opt);
2818 if (stickyopt) {
2819 int error;
2820
2821 /*
2822 * If stickyopt is provided, make a local copy of the options
2823 * for this particular packet, then override them by ancillary
2824 * objects.
2825 * XXX: copypktopts() does not copy the cached route to a next
2826 * hop (if any). This is not very good in terms of efficiency,
2827 * but we can allow this since this option should be rarely
2828 * used.
2829 */
2830 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2831 return (error);
2832 }
2833
2834 /*
2835 * XXX: Currently, we assume all the optional information is stored
2836 * in a single mbuf.
2837 */
2838 if (control->m_next)
2839 return (EINVAL);
2840
2841 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2842 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2843 int error;
2844
2845 if (control->m_len < CMSG_LEN(0))
2846 return (EINVAL);
2847
2848 cm = mtod(control, struct cmsghdr *);
2849 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2850 return (EINVAL);
2851 if (cm->cmsg_level != IPPROTO_IPV6)
2852 continue;
2853
2854 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2855 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2856 if (error)
2857 return (error);
2858 }
2859
2860 return (0);
2861 }
2862
2863 /*
2864 * Set a particular packet option, as a sticky option or an ancillary data
2865 * item. "len" can be 0 only when it's a sticky option.
2866 * We have 4 cases of combination of "sticky" and "cmsg":
2867 * "sticky=0, cmsg=0": impossible
2868 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2869 * "sticky=1, cmsg=0": RFC3542 socket option
2870 * "sticky=1, cmsg=1": RFC2292 socket option
2871 */
2872 static int
2873 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2874 struct ucred *cred, int sticky, int cmsg, int uproto)
2875 {
2876 int minmtupolicy, preftemp;
2877 int error;
2878
2879 if (!sticky && !cmsg) {
2880 #ifdef DIAGNOSTIC
2881 printf("ip6_setpktopt: impossible case\n");
2882 #endif
2883 return (EINVAL);
2884 }
2885
2886 /*
2887 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2888 * not be specified in the context of RFC3542. Conversely,
2889 * RFC3542 types should not be specified in the context of RFC2292.
2890 */
2891 if (!cmsg) {
2892 switch (optname) {
2893 case IPV6_2292PKTINFO:
2894 case IPV6_2292HOPLIMIT:
2895 case IPV6_2292NEXTHOP:
2896 case IPV6_2292HOPOPTS:
2897 case IPV6_2292DSTOPTS:
2898 case IPV6_2292RTHDR:
2899 case IPV6_2292PKTOPTIONS:
2900 return (ENOPROTOOPT);
2901 }
2902 }
2903 if (sticky && cmsg) {
2904 switch (optname) {
2905 case IPV6_PKTINFO:
2906 case IPV6_HOPLIMIT:
2907 case IPV6_NEXTHOP:
2908 case IPV6_HOPOPTS:
2909 case IPV6_DSTOPTS:
2910 case IPV6_RTHDRDSTOPTS:
2911 case IPV6_RTHDR:
2912 case IPV6_USE_MIN_MTU:
2913 case IPV6_DONTFRAG:
2914 case IPV6_TCLASS:
2915 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2916 return (ENOPROTOOPT);
2917 }
2918 }
2919
2920 switch (optname) {
2921 case IPV6_2292PKTINFO:
2922 case IPV6_PKTINFO:
2923 {
2924 struct ifnet *ifp = NULL;
2925 struct in6_pktinfo *pktinfo;
2926
2927 if (len != sizeof(struct in6_pktinfo))
2928 return (EINVAL);
2929
2930 pktinfo = (struct in6_pktinfo *)buf;
2931
2932 /*
2933 * An application can clear any sticky IPV6_PKTINFO option by
2934 * doing a "regular" setsockopt with ipi6_addr being
2935 * in6addr_any and ipi6_ifindex being zero.
2936 * [RFC 3542, Section 6]
2937 */
2938 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2939 pktinfo->ipi6_ifindex == 0 &&
2940 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2941 ip6_clearpktopts(opt, optname);
2942 break;
2943 }
2944
2945 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2946 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2947 return (EINVAL);
2948 }
2949 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2950 return (EINVAL);
2951 /* validate the interface index if specified. */
2952 if (pktinfo->ipi6_ifindex > V_if_index)
2953 return (ENXIO);
2954 if (pktinfo->ipi6_ifindex) {
2955 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2956 if (ifp == NULL)
2957 return (ENXIO);
2958 }
2959 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2960 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2961 return (ENETDOWN);
2962
2963 if (ifp != NULL &&
2964 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2965 struct in6_ifaddr *ia;
2966
2967 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2968 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2969 if (ia == NULL)
2970 return (EADDRNOTAVAIL);
2971 ifa_free(&ia->ia_ifa);
2972 }
2973 /*
2974 * We store the address anyway, and let in6_selectsrc()
2975 * validate the specified address. This is because ipi6_addr
2976 * may not have enough information about its scope zone, and
2977 * we may need additional information (such as outgoing
2978 * interface or the scope zone of a destination address) to
2979 * disambiguate the scope.
2980 * XXX: the delay of the validation may confuse the
2981 * application when it is used as a sticky option.
2982 */
2983 if (opt->ip6po_pktinfo == NULL) {
2984 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2985 M_IP6OPT, M_NOWAIT);
2986 if (opt->ip6po_pktinfo == NULL)
2987 return (ENOBUFS);
2988 }
2989 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2990 break;
2991 }
2992
2993 case IPV6_2292HOPLIMIT:
2994 case IPV6_HOPLIMIT:
2995 {
2996 int *hlimp;
2997
2998 /*
2999 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3000 * to simplify the ordering among hoplimit options.
3001 */
3002 if (optname == IPV6_HOPLIMIT && sticky)
3003 return (ENOPROTOOPT);
3004
3005 if (len != sizeof(int))
3006 return (EINVAL);
3007 hlimp = (int *)buf;
3008 if (*hlimp < -1 || *hlimp > 255)
3009 return (EINVAL);
3010
3011 opt->ip6po_hlim = *hlimp;
3012 break;
3013 }
3014
3015 case IPV6_TCLASS:
3016 {
3017 int tclass;
3018
3019 if (len != sizeof(int))
3020 return (EINVAL);
3021 tclass = *(int *)buf;
3022 if (tclass < -1 || tclass > 255)
3023 return (EINVAL);
3024
3025 opt->ip6po_tclass = tclass;
3026 break;
3027 }
3028
3029 case IPV6_2292NEXTHOP:
3030 case IPV6_NEXTHOP:
3031 if (cred != NULL) {
3032 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3033 if (error)
3034 return (error);
3035 }
3036
3037 if (len == 0) { /* just remove the option */
3038 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3039 break;
3040 }
3041
3042 /* check if cmsg_len is large enough for sa_len */
3043 if (len < sizeof(struct sockaddr) || len < *buf)
3044 return (EINVAL);
3045
3046 switch (((struct sockaddr *)buf)->sa_family) {
3047 case AF_INET6:
3048 {
3049 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3050 int error;
3051
3052 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3053 return (EINVAL);
3054
3055 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3056 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3057 return (EINVAL);
3058 }
3059 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3060 != 0) {
3061 return (error);
3062 }
3063 break;
3064 }
3065 case AF_LINK: /* should eventually be supported */
3066 default:
3067 return (EAFNOSUPPORT);
3068 }
3069
3070 /* turn off the previous option, then set the new option. */
3071 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3072 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3073 if (opt->ip6po_nexthop == NULL)
3074 return (ENOBUFS);
3075 bcopy(buf, opt->ip6po_nexthop, *buf);
3076 break;
3077
3078 case IPV6_2292HOPOPTS:
3079 case IPV6_HOPOPTS:
3080 {
3081 struct ip6_hbh *hbh;
3082 int hbhlen;
3083
3084 /*
3085 * XXX: We don't allow a non-privileged user to set ANY HbH
3086 * options, since per-option restriction has too much
3087 * overhead.
3088 */
3089 if (cred != NULL) {
3090 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3091 if (error)
3092 return (error);
3093 }
3094
3095 if (len == 0) {
3096 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3097 break; /* just remove the option */
3098 }
3099
3100 /* message length validation */
3101 if (len < sizeof(struct ip6_hbh))
3102 return (EINVAL);
3103 hbh = (struct ip6_hbh *)buf;
3104 hbhlen = (hbh->ip6h_len + 1) << 3;
3105 if (len != hbhlen)
3106 return (EINVAL);
3107
3108 /* turn off the previous option, then set the new option. */
3109 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3110 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3111 if (opt->ip6po_hbh == NULL)
3112 return (ENOBUFS);
3113 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3114
3115 break;
3116 }
3117
3118 case IPV6_2292DSTOPTS:
3119 case IPV6_DSTOPTS:
3120 case IPV6_RTHDRDSTOPTS:
3121 {
3122 struct ip6_dest *dest, **newdest = NULL;
3123 int destlen;
3124
3125 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3126 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3127 if (error)
3128 return (error);
3129 }
3130
3131 if (len == 0) {
3132 ip6_clearpktopts(opt, optname);
3133 break; /* just remove the option */
3134 }
3135
3136 /* message length validation */
3137 if (len < sizeof(struct ip6_dest))
3138 return (EINVAL);
3139 dest = (struct ip6_dest *)buf;
3140 destlen = (dest->ip6d_len + 1) << 3;
3141 if (len != destlen)
3142 return (EINVAL);
3143
3144 /*
3145 * Determine the position that the destination options header
3146 * should be inserted; before or after the routing header.
3147 */
3148 switch (optname) {
3149 case IPV6_2292DSTOPTS:
3150 /*
3151 * The old advacned API is ambiguous on this point.
3152 * Our approach is to determine the position based
3153 * according to the existence of a routing header.
3154 * Note, however, that this depends on the order of the
3155 * extension headers in the ancillary data; the 1st
3156 * part of the destination options header must appear
3157 * before the routing header in the ancillary data,
3158 * too.
3159 * RFC3542 solved the ambiguity by introducing
3160 * separate ancillary data or option types.
3161 */
3162 if (opt->ip6po_rthdr == NULL)
3163 newdest = &opt->ip6po_dest1;
3164 else
3165 newdest = &opt->ip6po_dest2;
3166 break;
3167 case IPV6_RTHDRDSTOPTS:
3168 newdest = &opt->ip6po_dest1;
3169 break;
3170 case IPV6_DSTOPTS:
3171 newdest = &opt->ip6po_dest2;
3172 break;
3173 }
3174
3175 /* turn off the previous option, then set the new option. */
3176 ip6_clearpktopts(opt, optname);
3177 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3178 if (*newdest == NULL)
3179 return (ENOBUFS);
3180 bcopy(dest, *newdest, destlen);
3181
3182 break;
3183 }
3184
3185 case IPV6_2292RTHDR:
3186 case IPV6_RTHDR:
3187 {
3188 struct ip6_rthdr *rth;
3189 int rthlen;
3190
3191 if (len == 0) {
3192 ip6_clearpktopts(opt, IPV6_RTHDR);
3193 break; /* just remove the option */
3194 }
3195
3196 /* message length validation */
3197 if (len < sizeof(struct ip6_rthdr))
3198 return (EINVAL);
3199 rth = (struct ip6_rthdr *)buf;
3200 rthlen = (rth->ip6r_len + 1) << 3;
3201 if (len != rthlen)
3202 return (EINVAL);
3203
3204 switch (rth->ip6r_type) {
3205 case IPV6_RTHDR_TYPE_0:
3206 if (rth->ip6r_len == 0) /* must contain one addr */
3207 return (EINVAL);
3208 if (rth->ip6r_len % 2) /* length must be even */
3209 return (EINVAL);
3210 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3211 return (EINVAL);
3212 break;
3213 default:
3214 return (EINVAL); /* not supported */
3215 }
3216
3217 /* turn off the previous option */
3218 ip6_clearpktopts(opt, IPV6_RTHDR);
3219 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3220 if (opt->ip6po_rthdr == NULL)
3221 return (ENOBUFS);
3222 bcopy(rth, opt->ip6po_rthdr, rthlen);
3223
3224 break;
3225 }
3226
3227 case IPV6_USE_MIN_MTU:
3228 if (len != sizeof(int))
3229 return (EINVAL);
3230 minmtupolicy = *(int *)buf;
3231 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3232 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3233 minmtupolicy != IP6PO_MINMTU_ALL) {
3234 return (EINVAL);
3235 }
3236 opt->ip6po_minmtu = minmtupolicy;
3237 break;
3238
3239 case IPV6_DONTFRAG:
3240 if (len != sizeof(int))
3241 return (EINVAL);
3242
3243 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3244 /*
3245 * we ignore this option for TCP sockets.
3246 * (RFC3542 leaves this case unspecified.)
3247 */
3248 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3249 } else
3250 opt->ip6po_flags |= IP6PO_DONTFRAG;
3251 break;
3252
3253 case IPV6_PREFER_TEMPADDR:
3254 if (len != sizeof(int))
3255 return (EINVAL);
3256 preftemp = *(int *)buf;
3257 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3258 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3259 preftemp != IP6PO_TEMPADDR_PREFER) {
3260 return (EINVAL);
3261 }
3262 opt->ip6po_prefer_tempaddr = preftemp;
3263 break;
3264
3265 default:
3266 return (ENOPROTOOPT);
3267 } /* end of switch */
3268
3269 return (0);
3270 }
3271
3272 /*
3273 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3274 * packet to the input queue of a specified interface. Note that this
3275 * calls the output routine of the loopback "driver", but with an interface
3276 * pointer that might NOT be &loif -- easier than replicating that code here.
3277 */
3278 void
3279 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3280 {
3281 struct mbuf *copym;
3282 struct ip6_hdr *ip6;
3283
3284 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3285 if (copym == NULL)
3286 return;
3287
3288 /*
3289 * Make sure to deep-copy IPv6 header portion in case the data
3290 * is in an mbuf cluster, so that we can safely override the IPv6
3291 * header portion later.
3292 */
3293 if (!M_WRITABLE(copym) ||
3294 copym->m_len < sizeof(struct ip6_hdr)) {
3295 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3296 if (copym == NULL)
3297 return;
3298 }
3299 ip6 = mtod(copym, struct ip6_hdr *);
3300 /*
3301 * clear embedded scope identifiers if necessary.
3302 * in6_clearscope will touch the addresses only when necessary.
3303 */
3304 in6_clearscope(&ip6->ip6_src);
3305 in6_clearscope(&ip6->ip6_dst);
3306 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3307 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3308 CSUM_PSEUDO_HDR;
3309 copym->m_pkthdr.csum_data = 0xffff;
3310 }
3311 if_simloop(ifp, copym, AF_INET6, 0);
3312 }
3313
3314 /*
3315 * Chop IPv6 header off from the payload.
3316 */
3317 static int
3318 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3319 {
3320 struct mbuf *mh;
3321 struct ip6_hdr *ip6;
3322
3323 ip6 = mtod(m, struct ip6_hdr *);
3324 if (m->m_len > sizeof(*ip6)) {
3325 mh = m_gethdr(M_NOWAIT, MT_DATA);
3326 if (mh == NULL) {
3327 m_freem(m);
3328 return ENOBUFS;
3329 }
3330 m_move_pkthdr(mh, m);
3331 M_ALIGN(mh, sizeof(*ip6));
3332 m->m_len -= sizeof(*ip6);
3333 m->m_data += sizeof(*ip6);
3334 mh->m_next = m;
3335 m = mh;
3336 m->m_len = sizeof(*ip6);
3337 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3338 }
3339 exthdrs->ip6e_ip6 = m;
3340 return 0;
3341 }
3342
3343 /*
3344 * Compute IPv6 extension header length.
3345 */
3346 int
3347 ip6_optlen(struct inpcb *inp)
3348 {
3349 int len;
3350
3351 if (!inp->in6p_outputopts)
3352 return 0;
3353
3354 len = 0;
3355 #define elen(x) \
3356 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3357
3358 len += elen(inp->in6p_outputopts->ip6po_hbh);
3359 if (inp->in6p_outputopts->ip6po_rthdr)
3360 /* dest1 is valid with rthdr only */
3361 len += elen(inp->in6p_outputopts->ip6po_dest1);
3362 len += elen(inp->in6p_outputopts->ip6po_rthdr);
3363 len += elen(inp->in6p_outputopts->ip6po_dest2);
3364 return len;
3365 #undef elen
3366 }
Cache object: e58a34baef4945427e3092f6d3c35610
|