1 /*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30 */
31
32 /*-
33 * Copyright (c) 1982, 1986, 1988, 1990, 1993
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
61 */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD: releng/11.1/sys/netinet6/ip6_output.c 318846 2017-05-25 01:14:34Z np $");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 #include "opt_sctp.h"
70 #include "opt_route.h"
71 #include "opt_rss.h"
72
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85
86 #include <machine/in_cksum.h>
87
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_llatbl.h>
91 #include <net/netisr.h>
92 #include <net/route.h>
93 #include <net/pfil.h>
94 #include <net/rss_config.h>
95 #include <net/vnet.h>
96
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip_var.h>
100 #include <netinet6/in6_fib.h>
101 #include <netinet6/in6_var.h>
102 #include <netinet/ip6.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/tcp_var.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/in6_rss.h>
109
110 #include <netipsec/ipsec_support.h>
111 #ifdef SCTP
112 #include <netinet/sctp.h>
113 #include <netinet/sctp_crc32.h>
114 #endif
115
116 #include <netinet6/ip6protosw.h>
117 #include <netinet6/scope6_var.h>
118
119 #ifdef FLOWTABLE
120 #include <net/flowtable.h>
121 #endif
122
123 extern int in6_mcast_loop;
124
125 struct ip6_exthdrs {
126 struct mbuf *ip6e_ip6;
127 struct mbuf *ip6e_hbh;
128 struct mbuf *ip6e_dest1;
129 struct mbuf *ip6e_rthdr;
130 struct mbuf *ip6e_dest2;
131 };
132
133 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
134
135 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
136 struct ucred *, int);
137 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
138 struct socket *, struct sockopt *);
139 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
140 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
141 struct ucred *, int, int, int);
142
143 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
144 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
145 struct ip6_frag **);
146 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
147 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
148 static int ip6_getpmtu(struct route_in6 *, int,
149 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
150 u_int);
151 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
152 u_long *, int *, u_int);
153 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
154 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
155
156
157 /*
158 * Make an extension header from option data. hp is the source, and
159 * mp is the destination.
160 */
161 #define MAKE_EXTHDR(hp, mp) \
162 do { \
163 if (hp) { \
164 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
165 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
166 ((eh)->ip6e_len + 1) << 3); \
167 if (error) \
168 goto freehdrs; \
169 } \
170 } while (/*CONSTCOND*/ 0)
171
172 /*
173 * Form a chain of extension headers.
174 * m is the extension header mbuf
175 * mp is the previous mbuf in the chain
176 * p is the next header
177 * i is the type of option.
178 */
179 #define MAKE_CHAIN(m, mp, p, i)\
180 do {\
181 if (m) {\
182 if (!hdrsplit) \
183 panic("assumption failed: hdr not split"); \
184 *mtod((m), u_char *) = *(p);\
185 *(p) = (i);\
186 p = mtod((m), u_char *);\
187 (m)->m_next = (mp)->m_next;\
188 (mp)->m_next = (m);\
189 (mp) = (m);\
190 }\
191 } while (/*CONSTCOND*/ 0)
192
193 void
194 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
195 {
196 u_short csum;
197
198 csum = in_cksum_skip(m, offset + plen, offset);
199 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
200 csum = 0xffff;
201 offset += m->m_pkthdr.csum_data; /* checksum offset */
202
203 if (offset + sizeof(u_short) > m->m_len) {
204 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
205 "csum_flags=%b\n", __func__, m->m_len, plen, offset,
206 (int)m->m_pkthdr.csum_flags, CSUM_BITS);
207 /*
208 * XXX this should not happen, but if it does, the correct
209 * behavior may be to insert the checksum in the appropriate
210 * next mbuf in the chain.
211 */
212 return;
213 }
214 *(u_short *)(m->m_data + offset) = csum;
215 }
216
217 int
218 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
219 int mtu, uint32_t id)
220 {
221 struct mbuf *m, **mnext, *m_frgpart;
222 struct ip6_hdr *ip6, *mhip6;
223 struct ip6_frag *ip6f;
224 int off;
225 int error;
226 int tlen = m0->m_pkthdr.len;
227
228 KASSERT(( mtu % 8 == 0), ("Fragment length must be a multiple of 8"));
229
230 m = m0;
231 ip6 = mtod(m, struct ip6_hdr *);
232 mnext = &m->m_nextpkt;
233
234 for (off = hlen; off < tlen; off += mtu) {
235 m = m_gethdr(M_NOWAIT, MT_DATA);
236 if (!m) {
237 IP6STAT_INC(ip6s_odropped);
238 return (ENOBUFS);
239 }
240 m->m_flags = m0->m_flags & M_COPYFLAGS;
241 *mnext = m;
242 mnext = &m->m_nextpkt;
243 m->m_data += max_linkhdr;
244 mhip6 = mtod(m, struct ip6_hdr *);
245 *mhip6 = *ip6;
246 m->m_len = sizeof(*mhip6);
247 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
248 if (error) {
249 IP6STAT_INC(ip6s_odropped);
250 return (error);
251 }
252 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
253 if (off + mtu >= tlen)
254 mtu = tlen - off;
255 else
256 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
257 mhip6->ip6_plen = htons((u_short)(mtu + hlen +
258 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
259 if ((m_frgpart = m_copy(m0, off, mtu)) == NULL) {
260 IP6STAT_INC(ip6s_odropped);
261 return (ENOBUFS);
262 }
263 m_cat(m, m_frgpart);
264 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
265 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
266 m->m_pkthdr.rcvif = NULL;
267 ip6f->ip6f_reserved = 0;
268 ip6f->ip6f_ident = id;
269 ip6f->ip6f_nxt = nextproto;
270 IP6STAT_INC(ip6s_ofragments);
271 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
272 }
273
274 return (0);
275 }
276
277 /*
278 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
279 * header (with pri, len, nxt, hlim, src, dst).
280 * This function may modify ver and hlim only.
281 * The mbuf chain containing the packet will be freed.
282 * The mbuf opt, if present, will not be freed.
283 * If route_in6 ro is present and has ro_rt initialized, route lookup would be
284 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
285 * then result of route lookup is stored in ro->ro_rt.
286 *
287 * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
288 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
289 * which is rt_mtu.
290 *
291 * ifpp - XXX: just for statistics
292 */
293 /*
294 * XXX TODO: no flowid is assigned for outbound flows?
295 */
296 int
297 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
298 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
299 struct ifnet **ifpp, struct inpcb *inp)
300 {
301 struct ip6_hdr *ip6;
302 struct ifnet *ifp, *origifp;
303 struct mbuf *m = m0;
304 struct mbuf *mprev = NULL;
305 int hlen, tlen, len;
306 struct route_in6 ip6route;
307 struct rtentry *rt = NULL;
308 struct sockaddr_in6 *dst, src_sa, dst_sa;
309 struct in6_addr odst;
310 int error = 0;
311 struct in6_ifaddr *ia = NULL;
312 u_long mtu;
313 int alwaysfrag, dontfrag;
314 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
315 struct ip6_exthdrs exthdrs;
316 struct in6_addr src0, dst0;
317 u_int32_t zone;
318 struct route_in6 *ro_pmtu = NULL;
319 int hdrsplit = 0;
320 int sw_csum, tso;
321 int needfiblookup;
322 uint32_t fibnum;
323 struct m_tag *fwd_tag = NULL;
324 uint32_t id;
325
326 if (inp != NULL) {
327 INP_LOCK_ASSERT(inp);
328 M_SETFIB(m, inp->inp_inc.inc_fibnum);
329 if ((flags & IP_NODEFAULTFLOWID) == 0) {
330 /* unconditionally set flowid */
331 m->m_pkthdr.flowid = inp->inp_flowid;
332 M_HASHTYPE_SET(m, inp->inp_flowtype);
333 }
334 }
335
336 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
337 /*
338 * IPSec checking which handles several cases.
339 * FAST IPSEC: We re-injected the packet.
340 * XXX: need scope argument.
341 */
342 if (IPSEC_ENABLED(ipv6)) {
343 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
344 if (error == EINPROGRESS)
345 error = 0;
346 goto done;
347 }
348 }
349 #endif /* IPSEC */
350
351 bzero(&exthdrs, sizeof(exthdrs));
352 if (opt) {
353 /* Hop-by-Hop options header */
354 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
355 /* Destination options header(1st part) */
356 if (opt->ip6po_rthdr) {
357 /*
358 * Destination options header(1st part)
359 * This only makes sense with a routing header.
360 * See Section 9.2 of RFC 3542.
361 * Disabling this part just for MIP6 convenience is
362 * a bad idea. We need to think carefully about a
363 * way to make the advanced API coexist with MIP6
364 * options, which might automatically be inserted in
365 * the kernel.
366 */
367 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
368 }
369 /* Routing header */
370 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
371 /* Destination options header(2nd part) */
372 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
373 }
374
375 /*
376 * Calculate the total length of the extension header chain.
377 * Keep the length of the unfragmentable part for fragmentation.
378 */
379 optlen = 0;
380 if (exthdrs.ip6e_hbh)
381 optlen += exthdrs.ip6e_hbh->m_len;
382 if (exthdrs.ip6e_dest1)
383 optlen += exthdrs.ip6e_dest1->m_len;
384 if (exthdrs.ip6e_rthdr)
385 optlen += exthdrs.ip6e_rthdr->m_len;
386 unfragpartlen = optlen + sizeof(struct ip6_hdr);
387
388 /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
389 if (exthdrs.ip6e_dest2)
390 optlen += exthdrs.ip6e_dest2->m_len;
391
392 /*
393 * If there is at least one extension header,
394 * separate IP6 header from the payload.
395 */
396 if (optlen && !hdrsplit) {
397 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
398 m = NULL;
399 goto freehdrs;
400 }
401 m = exthdrs.ip6e_ip6;
402 hdrsplit++;
403 }
404
405 ip6 = mtod(m, struct ip6_hdr *);
406
407 /* adjust mbuf packet header length */
408 m->m_pkthdr.len += optlen;
409 plen = m->m_pkthdr.len - sizeof(*ip6);
410
411 /* If this is a jumbo payload, insert a jumbo payload option. */
412 if (plen > IPV6_MAXPACKET) {
413 if (!hdrsplit) {
414 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
415 m = NULL;
416 goto freehdrs;
417 }
418 m = exthdrs.ip6e_ip6;
419 hdrsplit++;
420 }
421 /* adjust pointer */
422 ip6 = mtod(m, struct ip6_hdr *);
423 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
424 goto freehdrs;
425 ip6->ip6_plen = 0;
426 } else
427 ip6->ip6_plen = htons(plen);
428
429 /*
430 * Concatenate headers and fill in next header fields.
431 * Here we have, on "m"
432 * IPv6 payload
433 * and we insert headers accordingly. Finally, we should be getting:
434 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
435 *
436 * during the header composing process, "m" points to IPv6 header.
437 * "mprev" points to an extension header prior to esp.
438 */
439 u_char *nexthdrp = &ip6->ip6_nxt;
440 mprev = m;
441
442 /*
443 * we treat dest2 specially. this makes IPsec processing
444 * much easier. the goal here is to make mprev point the
445 * mbuf prior to dest2.
446 *
447 * result: IPv6 dest2 payload
448 * m and mprev will point to IPv6 header.
449 */
450 if (exthdrs.ip6e_dest2) {
451 if (!hdrsplit)
452 panic("assumption failed: hdr not split");
453 exthdrs.ip6e_dest2->m_next = m->m_next;
454 m->m_next = exthdrs.ip6e_dest2;
455 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
456 ip6->ip6_nxt = IPPROTO_DSTOPTS;
457 }
458
459 /*
460 * result: IPv6 hbh dest1 rthdr dest2 payload
461 * m will point to IPv6 header. mprev will point to the
462 * extension header prior to dest2 (rthdr in the above case).
463 */
464 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
465 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
466 IPPROTO_DSTOPTS);
467 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
468 IPPROTO_ROUTING);
469
470 /*
471 * If there is a routing header, discard the packet.
472 */
473 if (exthdrs.ip6e_rthdr) {
474 error = EINVAL;
475 goto bad;
476 }
477
478 /* Source address validation */
479 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
480 (flags & IPV6_UNSPECSRC) == 0) {
481 error = EOPNOTSUPP;
482 IP6STAT_INC(ip6s_badscope);
483 goto bad;
484 }
485 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
486 error = EOPNOTSUPP;
487 IP6STAT_INC(ip6s_badscope);
488 goto bad;
489 }
490
491 IP6STAT_INC(ip6s_localout);
492
493 /*
494 * Route packet.
495 */
496 if (ro == NULL) {
497 ro = &ip6route;
498 bzero((caddr_t)ro, sizeof(*ro));
499 }
500 ro_pmtu = ro;
501 if (opt && opt->ip6po_rthdr)
502 ro = &opt->ip6po_route;
503 dst = (struct sockaddr_in6 *)&ro->ro_dst;
504 #ifdef FLOWTABLE
505 if (ro->ro_rt == NULL)
506 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
507 #endif
508 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
509 again:
510 /*
511 * if specified, try to fill in the traffic class field.
512 * do not override if a non-zero value is already set.
513 * we check the diffserv field and the ecn field separately.
514 */
515 if (opt && opt->ip6po_tclass >= 0) {
516 int mask = 0;
517
518 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
519 mask |= 0xfc;
520 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
521 mask |= 0x03;
522 if (mask != 0)
523 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
524 }
525
526 /* fill in or override the hop limit field, if necessary. */
527 if (opt && opt->ip6po_hlim != -1)
528 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
529 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
530 if (im6o != NULL)
531 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
532 else
533 ip6->ip6_hlim = V_ip6_defmcasthlim;
534 }
535 /*
536 * Validate route against routing table additions;
537 * a better/more specific route might have been added.
538 * Make sure address family is set in route.
539 */
540 if (inp) {
541 ro->ro_dst.sin6_family = AF_INET6;
542 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
543 }
544 if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
545 ro->ro_dst.sin6_family == AF_INET6 &&
546 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
547 rt = ro->ro_rt;
548 ifp = ro->ro_rt->rt_ifp;
549 } else {
550 if (ro->ro_lle)
551 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
552 ro->ro_lle = NULL;
553 if (fwd_tag == NULL) {
554 bzero(&dst_sa, sizeof(dst_sa));
555 dst_sa.sin6_family = AF_INET6;
556 dst_sa.sin6_len = sizeof(dst_sa);
557 dst_sa.sin6_addr = ip6->ip6_dst;
558 }
559 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
560 &rt, fibnum);
561 if (error != 0) {
562 if (ifp != NULL)
563 in6_ifstat_inc(ifp, ifs6_out_discard);
564 goto bad;
565 }
566 }
567 if (rt == NULL) {
568 /*
569 * If in6_selectroute() does not return a route entry,
570 * dst may not have been updated.
571 */
572 *dst = dst_sa; /* XXX */
573 }
574
575 /*
576 * then rt (for unicast) and ifp must be non-NULL valid values.
577 */
578 if ((flags & IPV6_FORWARDING) == 0) {
579 /* XXX: the FORWARDING flag can be set for mrouting. */
580 in6_ifstat_inc(ifp, ifs6_out_request);
581 }
582 if (rt != NULL) {
583 ia = (struct in6_ifaddr *)(rt->rt_ifa);
584 counter_u64_add(rt->rt_pksent, 1);
585 }
586
587
588 /*
589 * The outgoing interface must be in the zone of source and
590 * destination addresses.
591 */
592 origifp = ifp;
593
594 src0 = ip6->ip6_src;
595 if (in6_setscope(&src0, origifp, &zone))
596 goto badscope;
597 bzero(&src_sa, sizeof(src_sa));
598 src_sa.sin6_family = AF_INET6;
599 src_sa.sin6_len = sizeof(src_sa);
600 src_sa.sin6_addr = ip6->ip6_src;
601 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
602 goto badscope;
603
604 dst0 = ip6->ip6_dst;
605 if (in6_setscope(&dst0, origifp, &zone))
606 goto badscope;
607 /* re-initialize to be sure */
608 bzero(&dst_sa, sizeof(dst_sa));
609 dst_sa.sin6_family = AF_INET6;
610 dst_sa.sin6_len = sizeof(dst_sa);
611 dst_sa.sin6_addr = ip6->ip6_dst;
612 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
613 goto badscope;
614 }
615
616 /* We should use ia_ifp to support the case of
617 * sending packets to an address of our own.
618 */
619 if (ia != NULL && ia->ia_ifp)
620 ifp = ia->ia_ifp;
621
622 /* scope check is done. */
623 goto routefound;
624
625 badscope:
626 IP6STAT_INC(ip6s_badscope);
627 in6_ifstat_inc(origifp, ifs6_out_discard);
628 if (error == 0)
629 error = EHOSTUNREACH; /* XXX */
630 goto bad;
631
632 routefound:
633 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
634 if (opt && opt->ip6po_nextroute.ro_rt) {
635 /*
636 * The nexthop is explicitly specified by the
637 * application. We assume the next hop is an IPv6
638 * address.
639 */
640 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
641 }
642 else if ((rt->rt_flags & RTF_GATEWAY))
643 dst = (struct sockaddr_in6 *)rt->rt_gateway;
644 }
645
646 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
647 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
648 } else {
649 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
650 in6_ifstat_inc(ifp, ifs6_out_mcast);
651 /*
652 * Confirm that the outgoing interface supports multicast.
653 */
654 if (!(ifp->if_flags & IFF_MULTICAST)) {
655 IP6STAT_INC(ip6s_noroute);
656 in6_ifstat_inc(ifp, ifs6_out_discard);
657 error = ENETUNREACH;
658 goto bad;
659 }
660 if ((im6o == NULL && in6_mcast_loop) ||
661 (im6o && im6o->im6o_multicast_loop)) {
662 /*
663 * Loop back multicast datagram if not expressly
664 * forbidden to do so, even if we have not joined
665 * the address; protocols will filter it later,
666 * thus deferring a hash lookup and lock acquisition
667 * at the expense of an m_copym().
668 */
669 ip6_mloopback(ifp, m);
670 } else {
671 /*
672 * If we are acting as a multicast router, perform
673 * multicast forwarding as if the packet had just
674 * arrived on the interface to which we are about
675 * to send. The multicast forwarding function
676 * recursively calls this function, using the
677 * IPV6_FORWARDING flag to prevent infinite recursion.
678 *
679 * Multicasts that are looped back by ip6_mloopback(),
680 * above, will be forwarded by the ip6_input() routine,
681 * if necessary.
682 */
683 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
684 /*
685 * XXX: ip6_mforward expects that rcvif is NULL
686 * when it is called from the originating path.
687 * However, it may not always be the case.
688 */
689 m->m_pkthdr.rcvif = NULL;
690 if (ip6_mforward(ip6, ifp, m) != 0) {
691 m_freem(m);
692 goto done;
693 }
694 }
695 }
696 /*
697 * Multicasts with a hoplimit of zero may be looped back,
698 * above, but must not be transmitted on a network.
699 * Also, multicasts addressed to the loopback interface
700 * are not sent -- the above call to ip6_mloopback() will
701 * loop back a copy if this host actually belongs to the
702 * destination group on the loopback interface.
703 */
704 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
705 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
706 m_freem(m);
707 goto done;
708 }
709 }
710
711 /*
712 * Fill the outgoing inteface to tell the upper layer
713 * to increment per-interface statistics.
714 */
715 if (ifpp)
716 *ifpp = ifp;
717
718 /* Determine path MTU. */
719 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
720 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
721 goto bad;
722
723 /*
724 * The caller of this function may specify to use the minimum MTU
725 * in some cases.
726 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
727 * setting. The logic is a bit complicated; by default, unicast
728 * packets will follow path MTU while multicast packets will be sent at
729 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
730 * including unicast ones will be sent at the minimum MTU. Multicast
731 * packets will always be sent at the minimum MTU unless
732 * IP6PO_MINMTU_DISABLE is explicitly specified.
733 * See RFC 3542 for more details.
734 */
735 if (mtu > IPV6_MMTU) {
736 if ((flags & IPV6_MINMTU))
737 mtu = IPV6_MMTU;
738 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
739 mtu = IPV6_MMTU;
740 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
741 (opt == NULL ||
742 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
743 mtu = IPV6_MMTU;
744 }
745 }
746
747 /*
748 * clear embedded scope identifiers if necessary.
749 * in6_clearscope will touch the addresses only when necessary.
750 */
751 in6_clearscope(&ip6->ip6_src);
752 in6_clearscope(&ip6->ip6_dst);
753
754 /*
755 * If the outgoing packet contains a hop-by-hop options header,
756 * it must be examined and processed even by the source node.
757 * (RFC 2460, section 4.)
758 */
759 if (exthdrs.ip6e_hbh) {
760 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
761 u_int32_t dummy; /* XXX unused */
762 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
763
764 #ifdef DIAGNOSTIC
765 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
766 panic("ip6e_hbh is not contiguous");
767 #endif
768 /*
769 * XXX: if we have to send an ICMPv6 error to the sender,
770 * we need the M_LOOP flag since icmp6_error() expects
771 * the IPv6 and the hop-by-hop options header are
772 * contiguous unless the flag is set.
773 */
774 m->m_flags |= M_LOOP;
775 m->m_pkthdr.rcvif = ifp;
776 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
777 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
778 &dummy, &plen) < 0) {
779 /* m was already freed at this point */
780 error = EINVAL;/* better error? */
781 goto done;
782 }
783 m->m_flags &= ~M_LOOP; /* XXX */
784 m->m_pkthdr.rcvif = NULL;
785 }
786
787 /* Jump over all PFIL processing if hooks are not active. */
788 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
789 goto passout;
790
791 odst = ip6->ip6_dst;
792 /* Run through list of hooks for output packets. */
793 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
794 if (error != 0 || m == NULL)
795 goto done;
796 /* adjust pointer */
797 ip6 = mtod(m, struct ip6_hdr *);
798
799 needfiblookup = 0;
800 /* See if destination IP address was changed by packet filter. */
801 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
802 m->m_flags |= M_SKIP_FIREWALL;
803 /* If destination is now ourself drop to ip6_input(). */
804 if (in6_localip(&ip6->ip6_dst)) {
805 m->m_flags |= M_FASTFWD_OURS;
806 if (m->m_pkthdr.rcvif == NULL)
807 m->m_pkthdr.rcvif = V_loif;
808 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
809 m->m_pkthdr.csum_flags |=
810 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
811 m->m_pkthdr.csum_data = 0xffff;
812 }
813 #ifdef SCTP
814 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
815 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
816 #endif
817 error = netisr_queue(NETISR_IPV6, m);
818 goto done;
819 } else {
820 RO_RTFREE(ro);
821 needfiblookup = 1; /* Redo the routing table lookup. */
822 if (ro->ro_lle)
823 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
824 ro->ro_lle = NULL;
825 }
826 }
827 /* See if fib was changed by packet filter. */
828 if (fibnum != M_GETFIB(m)) {
829 m->m_flags |= M_SKIP_FIREWALL;
830 fibnum = M_GETFIB(m);
831 RO_RTFREE(ro);
832 needfiblookup = 1;
833 if (ro->ro_lle)
834 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
835 ro->ro_lle = NULL;
836 }
837 if (needfiblookup)
838 goto again;
839
840 /* See if local, if yes, send it to netisr. */
841 if (m->m_flags & M_FASTFWD_OURS) {
842 if (m->m_pkthdr.rcvif == NULL)
843 m->m_pkthdr.rcvif = V_loif;
844 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
845 m->m_pkthdr.csum_flags |=
846 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
847 m->m_pkthdr.csum_data = 0xffff;
848 }
849 #ifdef SCTP
850 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
851 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
852 #endif
853 error = netisr_queue(NETISR_IPV6, m);
854 goto done;
855 }
856 /* Or forward to some other address? */
857 if ((m->m_flags & M_IP6_NEXTHOP) &&
858 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
859 dst = (struct sockaddr_in6 *)&ro->ro_dst;
860 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
861 m->m_flags |= M_SKIP_FIREWALL;
862 m->m_flags &= ~M_IP6_NEXTHOP;
863 m_tag_delete(m, fwd_tag);
864 goto again;
865 }
866
867 passout:
868 /*
869 * Send the packet to the outgoing interface.
870 * If necessary, do IPv6 fragmentation before sending.
871 *
872 * the logic here is rather complex:
873 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
874 * 1-a: send as is if tlen <= path mtu
875 * 1-b: fragment if tlen > path mtu
876 *
877 * 2: if user asks us not to fragment (dontfrag == 1)
878 * 2-a: send as is if tlen <= interface mtu
879 * 2-b: error if tlen > interface mtu
880 *
881 * 3: if we always need to attach fragment header (alwaysfrag == 1)
882 * always fragment
883 *
884 * 4: if dontfrag == 1 && alwaysfrag == 1
885 * error, as we cannot handle this conflicting request
886 */
887 sw_csum = m->m_pkthdr.csum_flags;
888 if (!hdrsplit) {
889 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
890 sw_csum &= ~ifp->if_hwassist;
891 } else
892 tso = 0;
893 /*
894 * If we added extension headers, we will not do TSO and calculate the
895 * checksums ourselves for now.
896 * XXX-BZ Need a framework to know when the NIC can handle it, even
897 * with ext. hdrs.
898 */
899 if (sw_csum & CSUM_DELAY_DATA_IPV6) {
900 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
901 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
902 }
903 #ifdef SCTP
904 if (sw_csum & CSUM_SCTP_IPV6) {
905 sw_csum &= ~CSUM_SCTP_IPV6;
906 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
907 }
908 #endif
909 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
910 tlen = m->m_pkthdr.len;
911
912 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
913 dontfrag = 1;
914 else
915 dontfrag = 0;
916 if (dontfrag && alwaysfrag) { /* case 4 */
917 /* conflicting request - can't transmit */
918 error = EMSGSIZE;
919 goto bad;
920 }
921 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */
922 /*
923 * Even if the DONTFRAG option is specified, we cannot send the
924 * packet when the data length is larger than the MTU of the
925 * outgoing interface.
926 * Notify the error by sending IPV6_PATHMTU ancillary data if
927 * application wanted to know the MTU value. Also return an
928 * error code (this is not described in the API spec).
929 */
930 if (inp != NULL)
931 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
932 error = EMSGSIZE;
933 goto bad;
934 }
935
936 /*
937 * transmit packet without fragmentation
938 */
939 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
940 struct in6_ifaddr *ia6;
941
942 ip6 = mtod(m, struct ip6_hdr *);
943 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
944 if (ia6) {
945 /* Record statistics for this interface address. */
946 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
947 counter_u64_add(ia6->ia_ifa.ifa_obytes,
948 m->m_pkthdr.len);
949 ifa_free(&ia6->ia_ifa);
950 }
951 error = nd6_output_ifp(ifp, origifp, m, dst,
952 (struct route *)ro);
953 goto done;
954 }
955
956 /*
957 * try to fragment the packet. case 1-b and 3
958 */
959 if (mtu < IPV6_MMTU) {
960 /* path MTU cannot be less than IPV6_MMTU */
961 error = EMSGSIZE;
962 in6_ifstat_inc(ifp, ifs6_out_fragfail);
963 goto bad;
964 } else if (ip6->ip6_plen == 0) {
965 /* jumbo payload cannot be fragmented */
966 error = EMSGSIZE;
967 in6_ifstat_inc(ifp, ifs6_out_fragfail);
968 goto bad;
969 } else {
970 u_char nextproto;
971
972 /*
973 * Too large for the destination or interface;
974 * fragment if possible.
975 * Must be able to put at least 8 bytes per fragment.
976 */
977 hlen = unfragpartlen;
978 if (mtu > IPV6_MAXPACKET)
979 mtu = IPV6_MAXPACKET;
980
981 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
982 if (len < 8) {
983 error = EMSGSIZE;
984 in6_ifstat_inc(ifp, ifs6_out_fragfail);
985 goto bad;
986 }
987
988 /*
989 * If the interface will not calculate checksums on
990 * fragmented packets, then do it here.
991 * XXX-BZ handle the hw offloading case. Need flags.
992 */
993 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
994 in6_delayed_cksum(m, plen, hlen);
995 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
996 }
997 #ifdef SCTP
998 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
999 sctp_delayed_cksum(m, hlen);
1000 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1001 }
1002 #endif
1003 /*
1004 * Change the next header field of the last header in the
1005 * unfragmentable part.
1006 */
1007 if (exthdrs.ip6e_rthdr) {
1008 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1009 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1010 } else if (exthdrs.ip6e_dest1) {
1011 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1012 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1013 } else if (exthdrs.ip6e_hbh) {
1014 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1015 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1016 } else {
1017 nextproto = ip6->ip6_nxt;
1018 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1019 }
1020
1021 /*
1022 * Loop through length of segment after first fragment,
1023 * make new header and copy data of each part and link onto
1024 * chain.
1025 */
1026 m0 = m;
1027 id = htonl(ip6_randomid());
1028 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1029 goto sendorfree;
1030
1031 in6_ifstat_inc(ifp, ifs6_out_fragok);
1032 }
1033
1034 /*
1035 * Remove leading garbages.
1036 */
1037 sendorfree:
1038 m = m0->m_nextpkt;
1039 m0->m_nextpkt = 0;
1040 m_freem(m0);
1041 for (m0 = m; m; m = m0) {
1042 m0 = m->m_nextpkt;
1043 m->m_nextpkt = 0;
1044 if (error == 0) {
1045 /* Record statistics for this interface address. */
1046 if (ia) {
1047 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1048 counter_u64_add(ia->ia_ifa.ifa_obytes,
1049 m->m_pkthdr.len);
1050 }
1051 error = nd6_output_ifp(ifp, origifp, m, dst,
1052 (struct route *)ro);
1053 } else
1054 m_freem(m);
1055 }
1056
1057 if (error == 0)
1058 IP6STAT_INC(ip6s_fragmented);
1059
1060 done:
1061 /*
1062 * Release the route if using our private route, or if
1063 * (with flowtable) we don't have our own reference.
1064 */
1065 if (ro == &ip6route ||
1066 (ro != NULL && ro->ro_flags & RT_NORTREF))
1067 RO_RTFREE(ro);
1068 return (error);
1069
1070 freehdrs:
1071 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1072 m_freem(exthdrs.ip6e_dest1);
1073 m_freem(exthdrs.ip6e_rthdr);
1074 m_freem(exthdrs.ip6e_dest2);
1075 /* FALLTHROUGH */
1076 bad:
1077 if (m)
1078 m_freem(m);
1079 goto done;
1080 }
1081
1082 static int
1083 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1084 {
1085 struct mbuf *m;
1086
1087 if (hlen > MCLBYTES)
1088 return (ENOBUFS); /* XXX */
1089
1090 if (hlen > MLEN)
1091 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1092 else
1093 m = m_get(M_NOWAIT, MT_DATA);
1094 if (m == NULL)
1095 return (ENOBUFS);
1096 m->m_len = hlen;
1097 if (hdr)
1098 bcopy(hdr, mtod(m, caddr_t), hlen);
1099
1100 *mp = m;
1101 return (0);
1102 }
1103
1104 /*
1105 * Insert jumbo payload option.
1106 */
1107 static int
1108 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1109 {
1110 struct mbuf *mopt;
1111 u_char *optbuf;
1112 u_int32_t v;
1113
1114 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1115
1116 /*
1117 * If there is no hop-by-hop options header, allocate new one.
1118 * If there is one but it doesn't have enough space to store the
1119 * jumbo payload option, allocate a cluster to store the whole options.
1120 * Otherwise, use it to store the options.
1121 */
1122 if (exthdrs->ip6e_hbh == NULL) {
1123 mopt = m_get(M_NOWAIT, MT_DATA);
1124 if (mopt == NULL)
1125 return (ENOBUFS);
1126 mopt->m_len = JUMBOOPTLEN;
1127 optbuf = mtod(mopt, u_char *);
1128 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1129 exthdrs->ip6e_hbh = mopt;
1130 } else {
1131 struct ip6_hbh *hbh;
1132
1133 mopt = exthdrs->ip6e_hbh;
1134 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1135 /*
1136 * XXX assumption:
1137 * - exthdrs->ip6e_hbh is not referenced from places
1138 * other than exthdrs.
1139 * - exthdrs->ip6e_hbh is not an mbuf chain.
1140 */
1141 int oldoptlen = mopt->m_len;
1142 struct mbuf *n;
1143
1144 /*
1145 * XXX: give up if the whole (new) hbh header does
1146 * not fit even in an mbuf cluster.
1147 */
1148 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1149 return (ENOBUFS);
1150
1151 /*
1152 * As a consequence, we must always prepare a cluster
1153 * at this point.
1154 */
1155 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1156 if (n == NULL)
1157 return (ENOBUFS);
1158 n->m_len = oldoptlen + JUMBOOPTLEN;
1159 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1160 oldoptlen);
1161 optbuf = mtod(n, caddr_t) + oldoptlen;
1162 m_freem(mopt);
1163 mopt = exthdrs->ip6e_hbh = n;
1164 } else {
1165 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1166 mopt->m_len += JUMBOOPTLEN;
1167 }
1168 optbuf[0] = IP6OPT_PADN;
1169 optbuf[1] = 1;
1170
1171 /*
1172 * Adjust the header length according to the pad and
1173 * the jumbo payload option.
1174 */
1175 hbh = mtod(mopt, struct ip6_hbh *);
1176 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1177 }
1178
1179 /* fill in the option. */
1180 optbuf[2] = IP6OPT_JUMBO;
1181 optbuf[3] = 4;
1182 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1183 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1184
1185 /* finally, adjust the packet header length */
1186 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1187
1188 return (0);
1189 #undef JUMBOOPTLEN
1190 }
1191
1192 /*
1193 * Insert fragment header and copy unfragmentable header portions.
1194 */
1195 static int
1196 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1197 struct ip6_frag **frghdrp)
1198 {
1199 struct mbuf *n, *mlast;
1200
1201 if (hlen > sizeof(struct ip6_hdr)) {
1202 n = m_copym(m0, sizeof(struct ip6_hdr),
1203 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1204 if (n == NULL)
1205 return (ENOBUFS);
1206 m->m_next = n;
1207 } else
1208 n = m;
1209
1210 /* Search for the last mbuf of unfragmentable part. */
1211 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1212 ;
1213
1214 if (M_WRITABLE(mlast) &&
1215 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1216 /* use the trailing space of the last mbuf for the fragment hdr */
1217 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1218 mlast->m_len);
1219 mlast->m_len += sizeof(struct ip6_frag);
1220 m->m_pkthdr.len += sizeof(struct ip6_frag);
1221 } else {
1222 /* allocate a new mbuf for the fragment header */
1223 struct mbuf *mfrg;
1224
1225 mfrg = m_get(M_NOWAIT, MT_DATA);
1226 if (mfrg == NULL)
1227 return (ENOBUFS);
1228 mfrg->m_len = sizeof(struct ip6_frag);
1229 *frghdrp = mtod(mfrg, struct ip6_frag *);
1230 mlast->m_next = mfrg;
1231 }
1232
1233 return (0);
1234 }
1235
1236 /*
1237 * Calculates IPv6 path mtu for destination @dst.
1238 * Resulting MTU is stored in @mtup.
1239 *
1240 * Returns 0 on success.
1241 */
1242 static int
1243 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1244 {
1245 struct nhop6_extended nh6;
1246 struct in6_addr kdst;
1247 uint32_t scopeid;
1248 struct ifnet *ifp;
1249 u_long mtu;
1250 int error;
1251
1252 in6_splitscope(dst, &kdst, &scopeid);
1253 if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1254 return (EHOSTUNREACH);
1255
1256 ifp = nh6.nh_ifp;
1257 mtu = nh6.nh_mtu;
1258
1259 error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1260 fib6_free_nh_ext(fibnum, &nh6);
1261
1262 return (error);
1263 }
1264
1265 /*
1266 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1267 * and cached data in @ro_pmtu.
1268 * MTU from (successful) route lookup is saved (along with dst)
1269 * inside @ro_pmtu to avoid subsequent route lookups after packet
1270 * filter processing.
1271 *
1272 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1273 * Returns 0 on success.
1274 */
1275 static int
1276 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1277 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1278 int *alwaysfragp, u_int fibnum, u_int proto)
1279 {
1280 struct nhop6_basic nh6;
1281 struct in6_addr kdst;
1282 uint32_t scopeid;
1283 struct sockaddr_in6 *sa6_dst;
1284 u_long mtu;
1285
1286 mtu = 0;
1287 if (do_lookup) {
1288
1289 /*
1290 * Here ro_pmtu has final destination address, while
1291 * ro might represent immediate destination.
1292 * Use ro_pmtu destination since mtu might differ.
1293 */
1294 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1295 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1296 ro_pmtu->ro_mtu = 0;
1297
1298 if (ro_pmtu->ro_mtu == 0) {
1299 bzero(sa6_dst, sizeof(*sa6_dst));
1300 sa6_dst->sin6_family = AF_INET6;
1301 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1302 sa6_dst->sin6_addr = *dst;
1303
1304 in6_splitscope(dst, &kdst, &scopeid);
1305 if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1306 &nh6) == 0)
1307 ro_pmtu->ro_mtu = nh6.nh_mtu;
1308 }
1309
1310 mtu = ro_pmtu->ro_mtu;
1311 }
1312
1313 if (ro_pmtu->ro_rt)
1314 mtu = ro_pmtu->ro_rt->rt_mtu;
1315
1316 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1317 }
1318
1319 /*
1320 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1321 * hostcache data for @dst.
1322 * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1323 *
1324 * Returns 0 on success.
1325 */
1326 static int
1327 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1328 u_long *mtup, int *alwaysfragp, u_int proto)
1329 {
1330 u_long mtu = 0;
1331 int alwaysfrag = 0;
1332 int error = 0;
1333
1334 if (rt_mtu > 0) {
1335 u_int32_t ifmtu;
1336 struct in_conninfo inc;
1337
1338 bzero(&inc, sizeof(inc));
1339 inc.inc_flags |= INC_ISIPV6;
1340 inc.inc6_faddr = *dst;
1341
1342 ifmtu = IN6_LINKMTU(ifp);
1343
1344 /* TCP is known to react to pmtu changes so skip hc */
1345 if (proto != IPPROTO_TCP)
1346 mtu = tcp_hc_getmtu(&inc);
1347
1348 if (mtu)
1349 mtu = min(mtu, rt_mtu);
1350 else
1351 mtu = rt_mtu;
1352 if (mtu == 0)
1353 mtu = ifmtu;
1354 else if (mtu < IPV6_MMTU) {
1355 /*
1356 * RFC2460 section 5, last paragraph:
1357 * if we record ICMPv6 too big message with
1358 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1359 * or smaller, with framgent header attached.
1360 * (fragment header is needed regardless from the
1361 * packet size, for translators to identify packets)
1362 */
1363 alwaysfrag = 1;
1364 mtu = IPV6_MMTU;
1365 }
1366 } else if (ifp) {
1367 mtu = IN6_LINKMTU(ifp);
1368 } else
1369 error = EHOSTUNREACH; /* XXX */
1370
1371 *mtup = mtu;
1372 if (alwaysfragp)
1373 *alwaysfragp = alwaysfrag;
1374 return (error);
1375 }
1376
1377 /*
1378 * IP6 socket option processing.
1379 */
1380 int
1381 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1382 {
1383 int optdatalen, uproto;
1384 void *optdata;
1385 struct inpcb *in6p = sotoinpcb(so);
1386 int error, optval;
1387 int level, op, optname;
1388 int optlen;
1389 struct thread *td;
1390 #ifdef RSS
1391 uint32_t rss_bucket;
1392 int retval;
1393 #endif
1394
1395 /*
1396 * Don't use more than a quarter of mbuf clusters. N.B.:
1397 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1398 * on LP64 architectures, so cast to u_long to avoid undefined
1399 * behavior. ILP32 architectures cannot have nmbclusters
1400 * large enough to overflow for other reasons.
1401 */
1402 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1403
1404 level = sopt->sopt_level;
1405 op = sopt->sopt_dir;
1406 optname = sopt->sopt_name;
1407 optlen = sopt->sopt_valsize;
1408 td = sopt->sopt_td;
1409 error = 0;
1410 optval = 0;
1411 uproto = (int)so->so_proto->pr_protocol;
1412
1413 if (level != IPPROTO_IPV6) {
1414 error = EINVAL;
1415
1416 if (sopt->sopt_level == SOL_SOCKET &&
1417 sopt->sopt_dir == SOPT_SET) {
1418 switch (sopt->sopt_name) {
1419 case SO_REUSEADDR:
1420 INP_WLOCK(in6p);
1421 if ((so->so_options & SO_REUSEADDR) != 0)
1422 in6p->inp_flags2 |= INP_REUSEADDR;
1423 else
1424 in6p->inp_flags2 &= ~INP_REUSEADDR;
1425 INP_WUNLOCK(in6p);
1426 error = 0;
1427 break;
1428 case SO_REUSEPORT:
1429 INP_WLOCK(in6p);
1430 if ((so->so_options & SO_REUSEPORT) != 0)
1431 in6p->inp_flags2 |= INP_REUSEPORT;
1432 else
1433 in6p->inp_flags2 &= ~INP_REUSEPORT;
1434 INP_WUNLOCK(in6p);
1435 error = 0;
1436 break;
1437 case SO_SETFIB:
1438 INP_WLOCK(in6p);
1439 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1440 INP_WUNLOCK(in6p);
1441 error = 0;
1442 break;
1443 default:
1444 break;
1445 }
1446 }
1447 } else { /* level == IPPROTO_IPV6 */
1448 switch (op) {
1449
1450 case SOPT_SET:
1451 switch (optname) {
1452 case IPV6_2292PKTOPTIONS:
1453 #ifdef IPV6_PKTOPTIONS
1454 case IPV6_PKTOPTIONS:
1455 #endif
1456 {
1457 struct mbuf *m;
1458
1459 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1460 printf("ip6_ctloutput: mbuf limit hit\n");
1461 error = ENOBUFS;
1462 break;
1463 }
1464
1465 error = soopt_getm(sopt, &m); /* XXX */
1466 if (error != 0)
1467 break;
1468 error = soopt_mcopyin(sopt, m); /* XXX */
1469 if (error != 0)
1470 break;
1471 error = ip6_pcbopts(&in6p->in6p_outputopts,
1472 m, so, sopt);
1473 m_freem(m); /* XXX */
1474 break;
1475 }
1476
1477 /*
1478 * Use of some Hop-by-Hop options or some
1479 * Destination options, might require special
1480 * privilege. That is, normal applications
1481 * (without special privilege) might be forbidden
1482 * from setting certain options in outgoing packets,
1483 * and might never see certain options in received
1484 * packets. [RFC 2292 Section 6]
1485 * KAME specific note:
1486 * KAME prevents non-privileged users from sending or
1487 * receiving ANY hbh/dst options in order to avoid
1488 * overhead of parsing options in the kernel.
1489 */
1490 case IPV6_RECVHOPOPTS:
1491 case IPV6_RECVDSTOPTS:
1492 case IPV6_RECVRTHDRDSTOPTS:
1493 if (td != NULL) {
1494 error = priv_check(td,
1495 PRIV_NETINET_SETHDROPTS);
1496 if (error)
1497 break;
1498 }
1499 /* FALLTHROUGH */
1500 case IPV6_UNICAST_HOPS:
1501 case IPV6_HOPLIMIT:
1502
1503 case IPV6_RECVPKTINFO:
1504 case IPV6_RECVHOPLIMIT:
1505 case IPV6_RECVRTHDR:
1506 case IPV6_RECVPATHMTU:
1507 case IPV6_RECVTCLASS:
1508 case IPV6_RECVFLOWID:
1509 #ifdef RSS
1510 case IPV6_RECVRSSBUCKETID:
1511 #endif
1512 case IPV6_V6ONLY:
1513 case IPV6_AUTOFLOWLABEL:
1514 case IPV6_BINDANY:
1515 case IPV6_BINDMULTI:
1516 #ifdef RSS
1517 case IPV6_RSS_LISTEN_BUCKET:
1518 #endif
1519 if (optname == IPV6_BINDANY && td != NULL) {
1520 error = priv_check(td,
1521 PRIV_NETINET_BINDANY);
1522 if (error)
1523 break;
1524 }
1525
1526 if (optlen != sizeof(int)) {
1527 error = EINVAL;
1528 break;
1529 }
1530 error = sooptcopyin(sopt, &optval,
1531 sizeof optval, sizeof optval);
1532 if (error)
1533 break;
1534 switch (optname) {
1535
1536 case IPV6_UNICAST_HOPS:
1537 if (optval < -1 || optval >= 256)
1538 error = EINVAL;
1539 else {
1540 /* -1 = kernel default */
1541 in6p->in6p_hops = optval;
1542 if ((in6p->inp_vflag &
1543 INP_IPV4) != 0)
1544 in6p->inp_ip_ttl = optval;
1545 }
1546 break;
1547 #define OPTSET(bit) \
1548 do { \
1549 INP_WLOCK(in6p); \
1550 if (optval) \
1551 in6p->inp_flags |= (bit); \
1552 else \
1553 in6p->inp_flags &= ~(bit); \
1554 INP_WUNLOCK(in6p); \
1555 } while (/*CONSTCOND*/ 0)
1556 #define OPTSET2292(bit) \
1557 do { \
1558 INP_WLOCK(in6p); \
1559 in6p->inp_flags |= IN6P_RFC2292; \
1560 if (optval) \
1561 in6p->inp_flags |= (bit); \
1562 else \
1563 in6p->inp_flags &= ~(bit); \
1564 INP_WUNLOCK(in6p); \
1565 } while (/*CONSTCOND*/ 0)
1566 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1567
1568 #define OPTSET2(bit, val) do { \
1569 INP_WLOCK(in6p); \
1570 if (val) \
1571 in6p->inp_flags2 |= bit; \
1572 else \
1573 in6p->inp_flags2 &= ~bit; \
1574 INP_WUNLOCK(in6p); \
1575 } while (0)
1576 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1577
1578 case IPV6_RECVPKTINFO:
1579 /* cannot mix with RFC2292 */
1580 if (OPTBIT(IN6P_RFC2292)) {
1581 error = EINVAL;
1582 break;
1583 }
1584 OPTSET(IN6P_PKTINFO);
1585 break;
1586
1587 case IPV6_HOPLIMIT:
1588 {
1589 struct ip6_pktopts **optp;
1590
1591 /* cannot mix with RFC2292 */
1592 if (OPTBIT(IN6P_RFC2292)) {
1593 error = EINVAL;
1594 break;
1595 }
1596 optp = &in6p->in6p_outputopts;
1597 error = ip6_pcbopt(IPV6_HOPLIMIT,
1598 (u_char *)&optval, sizeof(optval),
1599 optp, (td != NULL) ? td->td_ucred :
1600 NULL, uproto);
1601 break;
1602 }
1603
1604 case IPV6_RECVHOPLIMIT:
1605 /* cannot mix with RFC2292 */
1606 if (OPTBIT(IN6P_RFC2292)) {
1607 error = EINVAL;
1608 break;
1609 }
1610 OPTSET(IN6P_HOPLIMIT);
1611 break;
1612
1613 case IPV6_RECVHOPOPTS:
1614 /* cannot mix with RFC2292 */
1615 if (OPTBIT(IN6P_RFC2292)) {
1616 error = EINVAL;
1617 break;
1618 }
1619 OPTSET(IN6P_HOPOPTS);
1620 break;
1621
1622 case IPV6_RECVDSTOPTS:
1623 /* cannot mix with RFC2292 */
1624 if (OPTBIT(IN6P_RFC2292)) {
1625 error = EINVAL;
1626 break;
1627 }
1628 OPTSET(IN6P_DSTOPTS);
1629 break;
1630
1631 case IPV6_RECVRTHDRDSTOPTS:
1632 /* cannot mix with RFC2292 */
1633 if (OPTBIT(IN6P_RFC2292)) {
1634 error = EINVAL;
1635 break;
1636 }
1637 OPTSET(IN6P_RTHDRDSTOPTS);
1638 break;
1639
1640 case IPV6_RECVRTHDR:
1641 /* cannot mix with RFC2292 */
1642 if (OPTBIT(IN6P_RFC2292)) {
1643 error = EINVAL;
1644 break;
1645 }
1646 OPTSET(IN6P_RTHDR);
1647 break;
1648
1649 case IPV6_RECVPATHMTU:
1650 /*
1651 * We ignore this option for TCP
1652 * sockets.
1653 * (RFC3542 leaves this case
1654 * unspecified.)
1655 */
1656 if (uproto != IPPROTO_TCP)
1657 OPTSET(IN6P_MTU);
1658 break;
1659
1660 case IPV6_RECVFLOWID:
1661 OPTSET2(INP_RECVFLOWID, optval);
1662 break;
1663
1664 #ifdef RSS
1665 case IPV6_RECVRSSBUCKETID:
1666 OPTSET2(INP_RECVRSSBUCKETID, optval);
1667 break;
1668 #endif
1669
1670 case IPV6_V6ONLY:
1671 /*
1672 * make setsockopt(IPV6_V6ONLY)
1673 * available only prior to bind(2).
1674 * see ipng mailing list, Jun 22 2001.
1675 */
1676 if (in6p->inp_lport ||
1677 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1678 error = EINVAL;
1679 break;
1680 }
1681 OPTSET(IN6P_IPV6_V6ONLY);
1682 if (optval)
1683 in6p->inp_vflag &= ~INP_IPV4;
1684 else
1685 in6p->inp_vflag |= INP_IPV4;
1686 break;
1687 case IPV6_RECVTCLASS:
1688 /* cannot mix with RFC2292 XXX */
1689 if (OPTBIT(IN6P_RFC2292)) {
1690 error = EINVAL;
1691 break;
1692 }
1693 OPTSET(IN6P_TCLASS);
1694 break;
1695 case IPV6_AUTOFLOWLABEL:
1696 OPTSET(IN6P_AUTOFLOWLABEL);
1697 break;
1698
1699 case IPV6_BINDANY:
1700 OPTSET(INP_BINDANY);
1701 break;
1702
1703 case IPV6_BINDMULTI:
1704 OPTSET2(INP_BINDMULTI, optval);
1705 break;
1706 #ifdef RSS
1707 case IPV6_RSS_LISTEN_BUCKET:
1708 if ((optval >= 0) &&
1709 (optval < rss_getnumbuckets())) {
1710 in6p->inp_rss_listen_bucket = optval;
1711 OPTSET2(INP_RSS_BUCKET_SET, 1);
1712 } else {
1713 error = EINVAL;
1714 }
1715 break;
1716 #endif
1717 }
1718 break;
1719
1720 case IPV6_TCLASS:
1721 case IPV6_DONTFRAG:
1722 case IPV6_USE_MIN_MTU:
1723 case IPV6_PREFER_TEMPADDR:
1724 if (optlen != sizeof(optval)) {
1725 error = EINVAL;
1726 break;
1727 }
1728 error = sooptcopyin(sopt, &optval,
1729 sizeof optval, sizeof optval);
1730 if (error)
1731 break;
1732 {
1733 struct ip6_pktopts **optp;
1734 optp = &in6p->in6p_outputopts;
1735 error = ip6_pcbopt(optname,
1736 (u_char *)&optval, sizeof(optval),
1737 optp, (td != NULL) ? td->td_ucred :
1738 NULL, uproto);
1739 break;
1740 }
1741
1742 case IPV6_2292PKTINFO:
1743 case IPV6_2292HOPLIMIT:
1744 case IPV6_2292HOPOPTS:
1745 case IPV6_2292DSTOPTS:
1746 case IPV6_2292RTHDR:
1747 /* RFC 2292 */
1748 if (optlen != sizeof(int)) {
1749 error = EINVAL;
1750 break;
1751 }
1752 error = sooptcopyin(sopt, &optval,
1753 sizeof optval, sizeof optval);
1754 if (error)
1755 break;
1756 switch (optname) {
1757 case IPV6_2292PKTINFO:
1758 OPTSET2292(IN6P_PKTINFO);
1759 break;
1760 case IPV6_2292HOPLIMIT:
1761 OPTSET2292(IN6P_HOPLIMIT);
1762 break;
1763 case IPV6_2292HOPOPTS:
1764 /*
1765 * Check super-user privilege.
1766 * See comments for IPV6_RECVHOPOPTS.
1767 */
1768 if (td != NULL) {
1769 error = priv_check(td,
1770 PRIV_NETINET_SETHDROPTS);
1771 if (error)
1772 return (error);
1773 }
1774 OPTSET2292(IN6P_HOPOPTS);
1775 break;
1776 case IPV6_2292DSTOPTS:
1777 if (td != NULL) {
1778 error = priv_check(td,
1779 PRIV_NETINET_SETHDROPTS);
1780 if (error)
1781 return (error);
1782 }
1783 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1784 break;
1785 case IPV6_2292RTHDR:
1786 OPTSET2292(IN6P_RTHDR);
1787 break;
1788 }
1789 break;
1790 case IPV6_PKTINFO:
1791 case IPV6_HOPOPTS:
1792 case IPV6_RTHDR:
1793 case IPV6_DSTOPTS:
1794 case IPV6_RTHDRDSTOPTS:
1795 case IPV6_NEXTHOP:
1796 {
1797 /* new advanced API (RFC3542) */
1798 u_char *optbuf;
1799 u_char optbuf_storage[MCLBYTES];
1800 int optlen;
1801 struct ip6_pktopts **optp;
1802
1803 /* cannot mix with RFC2292 */
1804 if (OPTBIT(IN6P_RFC2292)) {
1805 error = EINVAL;
1806 break;
1807 }
1808
1809 /*
1810 * We only ensure valsize is not too large
1811 * here. Further validation will be done
1812 * later.
1813 */
1814 error = sooptcopyin(sopt, optbuf_storage,
1815 sizeof(optbuf_storage), 0);
1816 if (error)
1817 break;
1818 optlen = sopt->sopt_valsize;
1819 optbuf = optbuf_storage;
1820 optp = &in6p->in6p_outputopts;
1821 error = ip6_pcbopt(optname, optbuf, optlen,
1822 optp, (td != NULL) ? td->td_ucred : NULL,
1823 uproto);
1824 break;
1825 }
1826 #undef OPTSET
1827
1828 case IPV6_MULTICAST_IF:
1829 case IPV6_MULTICAST_HOPS:
1830 case IPV6_MULTICAST_LOOP:
1831 case IPV6_JOIN_GROUP:
1832 case IPV6_LEAVE_GROUP:
1833 case IPV6_MSFILTER:
1834 case MCAST_BLOCK_SOURCE:
1835 case MCAST_UNBLOCK_SOURCE:
1836 case MCAST_JOIN_GROUP:
1837 case MCAST_LEAVE_GROUP:
1838 case MCAST_JOIN_SOURCE_GROUP:
1839 case MCAST_LEAVE_SOURCE_GROUP:
1840 error = ip6_setmoptions(in6p, sopt);
1841 break;
1842
1843 case IPV6_PORTRANGE:
1844 error = sooptcopyin(sopt, &optval,
1845 sizeof optval, sizeof optval);
1846 if (error)
1847 break;
1848
1849 INP_WLOCK(in6p);
1850 switch (optval) {
1851 case IPV6_PORTRANGE_DEFAULT:
1852 in6p->inp_flags &= ~(INP_LOWPORT);
1853 in6p->inp_flags &= ~(INP_HIGHPORT);
1854 break;
1855
1856 case IPV6_PORTRANGE_HIGH:
1857 in6p->inp_flags &= ~(INP_LOWPORT);
1858 in6p->inp_flags |= INP_HIGHPORT;
1859 break;
1860
1861 case IPV6_PORTRANGE_LOW:
1862 in6p->inp_flags &= ~(INP_HIGHPORT);
1863 in6p->inp_flags |= INP_LOWPORT;
1864 break;
1865
1866 default:
1867 error = EINVAL;
1868 break;
1869 }
1870 INP_WUNLOCK(in6p);
1871 break;
1872
1873 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1874 case IPV6_IPSEC_POLICY:
1875 if (IPSEC_ENABLED(ipv6)) {
1876 error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1877 break;
1878 }
1879 /* FALLTHROUGH */
1880 #endif /* IPSEC */
1881
1882 default:
1883 error = ENOPROTOOPT;
1884 break;
1885 }
1886 break;
1887
1888 case SOPT_GET:
1889 switch (optname) {
1890
1891 case IPV6_2292PKTOPTIONS:
1892 #ifdef IPV6_PKTOPTIONS
1893 case IPV6_PKTOPTIONS:
1894 #endif
1895 /*
1896 * RFC3542 (effectively) deprecated the
1897 * semantics of the 2292-style pktoptions.
1898 * Since it was not reliable in nature (i.e.,
1899 * applications had to expect the lack of some
1900 * information after all), it would make sense
1901 * to simplify this part by always returning
1902 * empty data.
1903 */
1904 sopt->sopt_valsize = 0;
1905 break;
1906
1907 case IPV6_RECVHOPOPTS:
1908 case IPV6_RECVDSTOPTS:
1909 case IPV6_RECVRTHDRDSTOPTS:
1910 case IPV6_UNICAST_HOPS:
1911 case IPV6_RECVPKTINFO:
1912 case IPV6_RECVHOPLIMIT:
1913 case IPV6_RECVRTHDR:
1914 case IPV6_RECVPATHMTU:
1915
1916 case IPV6_V6ONLY:
1917 case IPV6_PORTRANGE:
1918 case IPV6_RECVTCLASS:
1919 case IPV6_AUTOFLOWLABEL:
1920 case IPV6_BINDANY:
1921 case IPV6_FLOWID:
1922 case IPV6_FLOWTYPE:
1923 case IPV6_RECVFLOWID:
1924 #ifdef RSS
1925 case IPV6_RSSBUCKETID:
1926 case IPV6_RECVRSSBUCKETID:
1927 #endif
1928 case IPV6_BINDMULTI:
1929 switch (optname) {
1930
1931 case IPV6_RECVHOPOPTS:
1932 optval = OPTBIT(IN6P_HOPOPTS);
1933 break;
1934
1935 case IPV6_RECVDSTOPTS:
1936 optval = OPTBIT(IN6P_DSTOPTS);
1937 break;
1938
1939 case IPV6_RECVRTHDRDSTOPTS:
1940 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1941 break;
1942
1943 case IPV6_UNICAST_HOPS:
1944 optval = in6p->in6p_hops;
1945 break;
1946
1947 case IPV6_RECVPKTINFO:
1948 optval = OPTBIT(IN6P_PKTINFO);
1949 break;
1950
1951 case IPV6_RECVHOPLIMIT:
1952 optval = OPTBIT(IN6P_HOPLIMIT);
1953 break;
1954
1955 case IPV6_RECVRTHDR:
1956 optval = OPTBIT(IN6P_RTHDR);
1957 break;
1958
1959 case IPV6_RECVPATHMTU:
1960 optval = OPTBIT(IN6P_MTU);
1961 break;
1962
1963 case IPV6_V6ONLY:
1964 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1965 break;
1966
1967 case IPV6_PORTRANGE:
1968 {
1969 int flags;
1970 flags = in6p->inp_flags;
1971 if (flags & INP_HIGHPORT)
1972 optval = IPV6_PORTRANGE_HIGH;
1973 else if (flags & INP_LOWPORT)
1974 optval = IPV6_PORTRANGE_LOW;
1975 else
1976 optval = 0;
1977 break;
1978 }
1979 case IPV6_RECVTCLASS:
1980 optval = OPTBIT(IN6P_TCLASS);
1981 break;
1982
1983 case IPV6_AUTOFLOWLABEL:
1984 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1985 break;
1986
1987 case IPV6_BINDANY:
1988 optval = OPTBIT(INP_BINDANY);
1989 break;
1990
1991 case IPV6_FLOWID:
1992 optval = in6p->inp_flowid;
1993 break;
1994
1995 case IPV6_FLOWTYPE:
1996 optval = in6p->inp_flowtype;
1997 break;
1998
1999 case IPV6_RECVFLOWID:
2000 optval = OPTBIT2(INP_RECVFLOWID);
2001 break;
2002 #ifdef RSS
2003 case IPV6_RSSBUCKETID:
2004 retval =
2005 rss_hash2bucket(in6p->inp_flowid,
2006 in6p->inp_flowtype,
2007 &rss_bucket);
2008 if (retval == 0)
2009 optval = rss_bucket;
2010 else
2011 error = EINVAL;
2012 break;
2013
2014 case IPV6_RECVRSSBUCKETID:
2015 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2016 break;
2017 #endif
2018
2019 case IPV6_BINDMULTI:
2020 optval = OPTBIT2(INP_BINDMULTI);
2021 break;
2022
2023 }
2024 if (error)
2025 break;
2026 error = sooptcopyout(sopt, &optval,
2027 sizeof optval);
2028 break;
2029
2030 case IPV6_PATHMTU:
2031 {
2032 u_long pmtu = 0;
2033 struct ip6_mtuinfo mtuinfo;
2034
2035 if (!(so->so_state & SS_ISCONNECTED))
2036 return (ENOTCONN);
2037 /*
2038 * XXX: we dot not consider the case of source
2039 * routing, or optional information to specify
2040 * the outgoing interface.
2041 */
2042 error = ip6_getpmtu_ctl(so->so_fibnum,
2043 &in6p->in6p_faddr, &pmtu);
2044 if (error)
2045 break;
2046 if (pmtu > IPV6_MAXPACKET)
2047 pmtu = IPV6_MAXPACKET;
2048
2049 bzero(&mtuinfo, sizeof(mtuinfo));
2050 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2051 optdata = (void *)&mtuinfo;
2052 optdatalen = sizeof(mtuinfo);
2053 error = sooptcopyout(sopt, optdata,
2054 optdatalen);
2055 break;
2056 }
2057
2058 case IPV6_2292PKTINFO:
2059 case IPV6_2292HOPLIMIT:
2060 case IPV6_2292HOPOPTS:
2061 case IPV6_2292RTHDR:
2062 case IPV6_2292DSTOPTS:
2063 switch (optname) {
2064 case IPV6_2292PKTINFO:
2065 optval = OPTBIT(IN6P_PKTINFO);
2066 break;
2067 case IPV6_2292HOPLIMIT:
2068 optval = OPTBIT(IN6P_HOPLIMIT);
2069 break;
2070 case IPV6_2292HOPOPTS:
2071 optval = OPTBIT(IN6P_HOPOPTS);
2072 break;
2073 case IPV6_2292RTHDR:
2074 optval = OPTBIT(IN6P_RTHDR);
2075 break;
2076 case IPV6_2292DSTOPTS:
2077 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2078 break;
2079 }
2080 error = sooptcopyout(sopt, &optval,
2081 sizeof optval);
2082 break;
2083 case IPV6_PKTINFO:
2084 case IPV6_HOPOPTS:
2085 case IPV6_RTHDR:
2086 case IPV6_DSTOPTS:
2087 case IPV6_RTHDRDSTOPTS:
2088 case IPV6_NEXTHOP:
2089 case IPV6_TCLASS:
2090 case IPV6_DONTFRAG:
2091 case IPV6_USE_MIN_MTU:
2092 case IPV6_PREFER_TEMPADDR:
2093 error = ip6_getpcbopt(in6p->in6p_outputopts,
2094 optname, sopt);
2095 break;
2096
2097 case IPV6_MULTICAST_IF:
2098 case IPV6_MULTICAST_HOPS:
2099 case IPV6_MULTICAST_LOOP:
2100 case IPV6_MSFILTER:
2101 error = ip6_getmoptions(in6p, sopt);
2102 break;
2103
2104 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2105 case IPV6_IPSEC_POLICY:
2106 if (IPSEC_ENABLED(ipv6)) {
2107 error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2108 break;
2109 }
2110 /* FALLTHROUGH */
2111 #endif /* IPSEC */
2112 default:
2113 error = ENOPROTOOPT;
2114 break;
2115 }
2116 break;
2117 }
2118 }
2119 return (error);
2120 }
2121
2122 int
2123 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2124 {
2125 int error = 0, optval, optlen;
2126 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2127 struct inpcb *in6p = sotoinpcb(so);
2128 int level, op, optname;
2129
2130 level = sopt->sopt_level;
2131 op = sopt->sopt_dir;
2132 optname = sopt->sopt_name;
2133 optlen = sopt->sopt_valsize;
2134
2135 if (level != IPPROTO_IPV6) {
2136 return (EINVAL);
2137 }
2138
2139 switch (optname) {
2140 case IPV6_CHECKSUM:
2141 /*
2142 * For ICMPv6 sockets, no modification allowed for checksum
2143 * offset, permit "no change" values to help existing apps.
2144 *
2145 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2146 * for an ICMPv6 socket will fail."
2147 * The current behavior does not meet RFC3542.
2148 */
2149 switch (op) {
2150 case SOPT_SET:
2151 if (optlen != sizeof(int)) {
2152 error = EINVAL;
2153 break;
2154 }
2155 error = sooptcopyin(sopt, &optval, sizeof(optval),
2156 sizeof(optval));
2157 if (error)
2158 break;
2159 if ((optval % 2) != 0) {
2160 /* the API assumes even offset values */
2161 error = EINVAL;
2162 } else if (so->so_proto->pr_protocol ==
2163 IPPROTO_ICMPV6) {
2164 if (optval != icmp6off)
2165 error = EINVAL;
2166 } else
2167 in6p->in6p_cksum = optval;
2168 break;
2169
2170 case SOPT_GET:
2171 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2172 optval = icmp6off;
2173 else
2174 optval = in6p->in6p_cksum;
2175
2176 error = sooptcopyout(sopt, &optval, sizeof(optval));
2177 break;
2178
2179 default:
2180 error = EINVAL;
2181 break;
2182 }
2183 break;
2184
2185 default:
2186 error = ENOPROTOOPT;
2187 break;
2188 }
2189
2190 return (error);
2191 }
2192
2193 /*
2194 * Set up IP6 options in pcb for insertion in output packets or
2195 * specifying behavior of outgoing packets.
2196 */
2197 static int
2198 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2199 struct socket *so, struct sockopt *sopt)
2200 {
2201 struct ip6_pktopts *opt = *pktopt;
2202 int error = 0;
2203 struct thread *td = sopt->sopt_td;
2204
2205 /* turn off any old options. */
2206 if (opt) {
2207 #ifdef DIAGNOSTIC
2208 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2209 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2210 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2211 printf("ip6_pcbopts: all specified options are cleared.\n");
2212 #endif
2213 ip6_clearpktopts(opt, -1);
2214 } else
2215 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2216 *pktopt = NULL;
2217
2218 if (!m || m->m_len == 0) {
2219 /*
2220 * Only turning off any previous options, regardless of
2221 * whether the opt is just created or given.
2222 */
2223 free(opt, M_IP6OPT);
2224 return (0);
2225 }
2226
2227 /* set options specified by user. */
2228 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2229 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2230 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2231 free(opt, M_IP6OPT);
2232 return (error);
2233 }
2234 *pktopt = opt;
2235 return (0);
2236 }
2237
2238 /*
2239 * initialize ip6_pktopts. beware that there are non-zero default values in
2240 * the struct.
2241 */
2242 void
2243 ip6_initpktopts(struct ip6_pktopts *opt)
2244 {
2245
2246 bzero(opt, sizeof(*opt));
2247 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2248 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2249 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2250 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2251 }
2252
2253 static int
2254 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2255 struct ucred *cred, int uproto)
2256 {
2257 struct ip6_pktopts *opt;
2258
2259 if (*pktopt == NULL) {
2260 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2261 M_WAITOK);
2262 ip6_initpktopts(*pktopt);
2263 }
2264 opt = *pktopt;
2265
2266 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2267 }
2268
2269 static int
2270 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2271 {
2272 void *optdata = NULL;
2273 int optdatalen = 0;
2274 struct ip6_ext *ip6e;
2275 int error = 0;
2276 struct in6_pktinfo null_pktinfo;
2277 int deftclass = 0, on;
2278 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2279 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2280
2281 switch (optname) {
2282 case IPV6_PKTINFO:
2283 optdata = (void *)&null_pktinfo;
2284 if (pktopt && pktopt->ip6po_pktinfo) {
2285 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2286 sizeof(null_pktinfo));
2287 in6_clearscope(&null_pktinfo.ipi6_addr);
2288 } else {
2289 /* XXX: we don't have to do this every time... */
2290 bzero(&null_pktinfo, sizeof(null_pktinfo));
2291 }
2292 optdatalen = sizeof(struct in6_pktinfo);
2293 break;
2294 case IPV6_TCLASS:
2295 if (pktopt && pktopt->ip6po_tclass >= 0)
2296 optdata = (void *)&pktopt->ip6po_tclass;
2297 else
2298 optdata = (void *)&deftclass;
2299 optdatalen = sizeof(int);
2300 break;
2301 case IPV6_HOPOPTS:
2302 if (pktopt && pktopt->ip6po_hbh) {
2303 optdata = (void *)pktopt->ip6po_hbh;
2304 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2305 optdatalen = (ip6e->ip6e_len + 1) << 3;
2306 }
2307 break;
2308 case IPV6_RTHDR:
2309 if (pktopt && pktopt->ip6po_rthdr) {
2310 optdata = (void *)pktopt->ip6po_rthdr;
2311 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2312 optdatalen = (ip6e->ip6e_len + 1) << 3;
2313 }
2314 break;
2315 case IPV6_RTHDRDSTOPTS:
2316 if (pktopt && pktopt->ip6po_dest1) {
2317 optdata = (void *)pktopt->ip6po_dest1;
2318 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2319 optdatalen = (ip6e->ip6e_len + 1) << 3;
2320 }
2321 break;
2322 case IPV6_DSTOPTS:
2323 if (pktopt && pktopt->ip6po_dest2) {
2324 optdata = (void *)pktopt->ip6po_dest2;
2325 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2326 optdatalen = (ip6e->ip6e_len + 1) << 3;
2327 }
2328 break;
2329 case IPV6_NEXTHOP:
2330 if (pktopt && pktopt->ip6po_nexthop) {
2331 optdata = (void *)pktopt->ip6po_nexthop;
2332 optdatalen = pktopt->ip6po_nexthop->sa_len;
2333 }
2334 break;
2335 case IPV6_USE_MIN_MTU:
2336 if (pktopt)
2337 optdata = (void *)&pktopt->ip6po_minmtu;
2338 else
2339 optdata = (void *)&defminmtu;
2340 optdatalen = sizeof(int);
2341 break;
2342 case IPV6_DONTFRAG:
2343 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2344 on = 1;
2345 else
2346 on = 0;
2347 optdata = (void *)&on;
2348 optdatalen = sizeof(on);
2349 break;
2350 case IPV6_PREFER_TEMPADDR:
2351 if (pktopt)
2352 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2353 else
2354 optdata = (void *)&defpreftemp;
2355 optdatalen = sizeof(int);
2356 break;
2357 default: /* should not happen */
2358 #ifdef DIAGNOSTIC
2359 panic("ip6_getpcbopt: unexpected option\n");
2360 #endif
2361 return (ENOPROTOOPT);
2362 }
2363
2364 error = sooptcopyout(sopt, optdata, optdatalen);
2365
2366 return (error);
2367 }
2368
2369 void
2370 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2371 {
2372 if (pktopt == NULL)
2373 return;
2374
2375 if (optname == -1 || optname == IPV6_PKTINFO) {
2376 if (pktopt->ip6po_pktinfo)
2377 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2378 pktopt->ip6po_pktinfo = NULL;
2379 }
2380 if (optname == -1 || optname == IPV6_HOPLIMIT)
2381 pktopt->ip6po_hlim = -1;
2382 if (optname == -1 || optname == IPV6_TCLASS)
2383 pktopt->ip6po_tclass = -1;
2384 if (optname == -1 || optname == IPV6_NEXTHOP) {
2385 if (pktopt->ip6po_nextroute.ro_rt) {
2386 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2387 pktopt->ip6po_nextroute.ro_rt = NULL;
2388 }
2389 if (pktopt->ip6po_nexthop)
2390 free(pktopt->ip6po_nexthop, M_IP6OPT);
2391 pktopt->ip6po_nexthop = NULL;
2392 }
2393 if (optname == -1 || optname == IPV6_HOPOPTS) {
2394 if (pktopt->ip6po_hbh)
2395 free(pktopt->ip6po_hbh, M_IP6OPT);
2396 pktopt->ip6po_hbh = NULL;
2397 }
2398 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2399 if (pktopt->ip6po_dest1)
2400 free(pktopt->ip6po_dest1, M_IP6OPT);
2401 pktopt->ip6po_dest1 = NULL;
2402 }
2403 if (optname == -1 || optname == IPV6_RTHDR) {
2404 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2405 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2406 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2407 if (pktopt->ip6po_route.ro_rt) {
2408 RTFREE(pktopt->ip6po_route.ro_rt);
2409 pktopt->ip6po_route.ro_rt = NULL;
2410 }
2411 }
2412 if (optname == -1 || optname == IPV6_DSTOPTS) {
2413 if (pktopt->ip6po_dest2)
2414 free(pktopt->ip6po_dest2, M_IP6OPT);
2415 pktopt->ip6po_dest2 = NULL;
2416 }
2417 }
2418
2419 #define PKTOPT_EXTHDRCPY(type) \
2420 do {\
2421 if (src->type) {\
2422 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2423 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2424 if (dst->type == NULL && canwait == M_NOWAIT)\
2425 goto bad;\
2426 bcopy(src->type, dst->type, hlen);\
2427 }\
2428 } while (/*CONSTCOND*/ 0)
2429
2430 static int
2431 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2432 {
2433 if (dst == NULL || src == NULL) {
2434 printf("ip6_clearpktopts: invalid argument\n");
2435 return (EINVAL);
2436 }
2437
2438 dst->ip6po_hlim = src->ip6po_hlim;
2439 dst->ip6po_tclass = src->ip6po_tclass;
2440 dst->ip6po_flags = src->ip6po_flags;
2441 dst->ip6po_minmtu = src->ip6po_minmtu;
2442 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2443 if (src->ip6po_pktinfo) {
2444 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2445 M_IP6OPT, canwait);
2446 if (dst->ip6po_pktinfo == NULL)
2447 goto bad;
2448 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2449 }
2450 if (src->ip6po_nexthop) {
2451 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2452 M_IP6OPT, canwait);
2453 if (dst->ip6po_nexthop == NULL)
2454 goto bad;
2455 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2456 src->ip6po_nexthop->sa_len);
2457 }
2458 PKTOPT_EXTHDRCPY(ip6po_hbh);
2459 PKTOPT_EXTHDRCPY(ip6po_dest1);
2460 PKTOPT_EXTHDRCPY(ip6po_dest2);
2461 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2462 return (0);
2463
2464 bad:
2465 ip6_clearpktopts(dst, -1);
2466 return (ENOBUFS);
2467 }
2468 #undef PKTOPT_EXTHDRCPY
2469
2470 struct ip6_pktopts *
2471 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2472 {
2473 int error;
2474 struct ip6_pktopts *dst;
2475
2476 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2477 if (dst == NULL)
2478 return (NULL);
2479 ip6_initpktopts(dst);
2480
2481 if ((error = copypktopts(dst, src, canwait)) != 0) {
2482 free(dst, M_IP6OPT);
2483 return (NULL);
2484 }
2485
2486 return (dst);
2487 }
2488
2489 void
2490 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2491 {
2492 if (pktopt == NULL)
2493 return;
2494
2495 ip6_clearpktopts(pktopt, -1);
2496
2497 free(pktopt, M_IP6OPT);
2498 }
2499
2500 /*
2501 * Set IPv6 outgoing packet options based on advanced API.
2502 */
2503 int
2504 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2505 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2506 {
2507 struct cmsghdr *cm = NULL;
2508
2509 if (control == NULL || opt == NULL)
2510 return (EINVAL);
2511
2512 ip6_initpktopts(opt);
2513 if (stickyopt) {
2514 int error;
2515
2516 /*
2517 * If stickyopt is provided, make a local copy of the options
2518 * for this particular packet, then override them by ancillary
2519 * objects.
2520 * XXX: copypktopts() does not copy the cached route to a next
2521 * hop (if any). This is not very good in terms of efficiency,
2522 * but we can allow this since this option should be rarely
2523 * used.
2524 */
2525 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2526 return (error);
2527 }
2528
2529 /*
2530 * XXX: Currently, we assume all the optional information is stored
2531 * in a single mbuf.
2532 */
2533 if (control->m_next)
2534 return (EINVAL);
2535
2536 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2537 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2538 int error;
2539
2540 if (control->m_len < CMSG_LEN(0))
2541 return (EINVAL);
2542
2543 cm = mtod(control, struct cmsghdr *);
2544 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2545 return (EINVAL);
2546 if (cm->cmsg_level != IPPROTO_IPV6)
2547 continue;
2548
2549 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2550 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2551 if (error)
2552 return (error);
2553 }
2554
2555 return (0);
2556 }
2557
2558 /*
2559 * Set a particular packet option, as a sticky option or an ancillary data
2560 * item. "len" can be 0 only when it's a sticky option.
2561 * We have 4 cases of combination of "sticky" and "cmsg":
2562 * "sticky=0, cmsg=0": impossible
2563 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2564 * "sticky=1, cmsg=0": RFC3542 socket option
2565 * "sticky=1, cmsg=1": RFC2292 socket option
2566 */
2567 static int
2568 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2569 struct ucred *cred, int sticky, int cmsg, int uproto)
2570 {
2571 int minmtupolicy, preftemp;
2572 int error;
2573
2574 if (!sticky && !cmsg) {
2575 #ifdef DIAGNOSTIC
2576 printf("ip6_setpktopt: impossible case\n");
2577 #endif
2578 return (EINVAL);
2579 }
2580
2581 /*
2582 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2583 * not be specified in the context of RFC3542. Conversely,
2584 * RFC3542 types should not be specified in the context of RFC2292.
2585 */
2586 if (!cmsg) {
2587 switch (optname) {
2588 case IPV6_2292PKTINFO:
2589 case IPV6_2292HOPLIMIT:
2590 case IPV6_2292NEXTHOP:
2591 case IPV6_2292HOPOPTS:
2592 case IPV6_2292DSTOPTS:
2593 case IPV6_2292RTHDR:
2594 case IPV6_2292PKTOPTIONS:
2595 return (ENOPROTOOPT);
2596 }
2597 }
2598 if (sticky && cmsg) {
2599 switch (optname) {
2600 case IPV6_PKTINFO:
2601 case IPV6_HOPLIMIT:
2602 case IPV6_NEXTHOP:
2603 case IPV6_HOPOPTS:
2604 case IPV6_DSTOPTS:
2605 case IPV6_RTHDRDSTOPTS:
2606 case IPV6_RTHDR:
2607 case IPV6_USE_MIN_MTU:
2608 case IPV6_DONTFRAG:
2609 case IPV6_TCLASS:
2610 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2611 return (ENOPROTOOPT);
2612 }
2613 }
2614
2615 switch (optname) {
2616 case IPV6_2292PKTINFO:
2617 case IPV6_PKTINFO:
2618 {
2619 struct ifnet *ifp = NULL;
2620 struct in6_pktinfo *pktinfo;
2621
2622 if (len != sizeof(struct in6_pktinfo))
2623 return (EINVAL);
2624
2625 pktinfo = (struct in6_pktinfo *)buf;
2626
2627 /*
2628 * An application can clear any sticky IPV6_PKTINFO option by
2629 * doing a "regular" setsockopt with ipi6_addr being
2630 * in6addr_any and ipi6_ifindex being zero.
2631 * [RFC 3542, Section 6]
2632 */
2633 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2634 pktinfo->ipi6_ifindex == 0 &&
2635 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2636 ip6_clearpktopts(opt, optname);
2637 break;
2638 }
2639
2640 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2641 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2642 return (EINVAL);
2643 }
2644 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2645 return (EINVAL);
2646 /* validate the interface index if specified. */
2647 if (pktinfo->ipi6_ifindex > V_if_index)
2648 return (ENXIO);
2649 if (pktinfo->ipi6_ifindex) {
2650 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2651 if (ifp == NULL)
2652 return (ENXIO);
2653 }
2654 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2655 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2656 return (ENETDOWN);
2657
2658 if (ifp != NULL &&
2659 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2660 struct in6_ifaddr *ia;
2661
2662 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2663 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2664 if (ia == NULL)
2665 return (EADDRNOTAVAIL);
2666 ifa_free(&ia->ia_ifa);
2667 }
2668 /*
2669 * We store the address anyway, and let in6_selectsrc()
2670 * validate the specified address. This is because ipi6_addr
2671 * may not have enough information about its scope zone, and
2672 * we may need additional information (such as outgoing
2673 * interface or the scope zone of a destination address) to
2674 * disambiguate the scope.
2675 * XXX: the delay of the validation may confuse the
2676 * application when it is used as a sticky option.
2677 */
2678 if (opt->ip6po_pktinfo == NULL) {
2679 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2680 M_IP6OPT, M_NOWAIT);
2681 if (opt->ip6po_pktinfo == NULL)
2682 return (ENOBUFS);
2683 }
2684 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2685 break;
2686 }
2687
2688 case IPV6_2292HOPLIMIT:
2689 case IPV6_HOPLIMIT:
2690 {
2691 int *hlimp;
2692
2693 /*
2694 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2695 * to simplify the ordering among hoplimit options.
2696 */
2697 if (optname == IPV6_HOPLIMIT && sticky)
2698 return (ENOPROTOOPT);
2699
2700 if (len != sizeof(int))
2701 return (EINVAL);
2702 hlimp = (int *)buf;
2703 if (*hlimp < -1 || *hlimp > 255)
2704 return (EINVAL);
2705
2706 opt->ip6po_hlim = *hlimp;
2707 break;
2708 }
2709
2710 case IPV6_TCLASS:
2711 {
2712 int tclass;
2713
2714 if (len != sizeof(int))
2715 return (EINVAL);
2716 tclass = *(int *)buf;
2717 if (tclass < -1 || tclass > 255)
2718 return (EINVAL);
2719
2720 opt->ip6po_tclass = tclass;
2721 break;
2722 }
2723
2724 case IPV6_2292NEXTHOP:
2725 case IPV6_NEXTHOP:
2726 if (cred != NULL) {
2727 error = priv_check_cred(cred,
2728 PRIV_NETINET_SETHDROPTS, 0);
2729 if (error)
2730 return (error);
2731 }
2732
2733 if (len == 0) { /* just remove the option */
2734 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2735 break;
2736 }
2737
2738 /* check if cmsg_len is large enough for sa_len */
2739 if (len < sizeof(struct sockaddr) || len < *buf)
2740 return (EINVAL);
2741
2742 switch (((struct sockaddr *)buf)->sa_family) {
2743 case AF_INET6:
2744 {
2745 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2746 int error;
2747
2748 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2749 return (EINVAL);
2750
2751 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2752 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2753 return (EINVAL);
2754 }
2755 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2756 != 0) {
2757 return (error);
2758 }
2759 break;
2760 }
2761 case AF_LINK: /* should eventually be supported */
2762 default:
2763 return (EAFNOSUPPORT);
2764 }
2765
2766 /* turn off the previous option, then set the new option. */
2767 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2768 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2769 if (opt->ip6po_nexthop == NULL)
2770 return (ENOBUFS);
2771 bcopy(buf, opt->ip6po_nexthop, *buf);
2772 break;
2773
2774 case IPV6_2292HOPOPTS:
2775 case IPV6_HOPOPTS:
2776 {
2777 struct ip6_hbh *hbh;
2778 int hbhlen;
2779
2780 /*
2781 * XXX: We don't allow a non-privileged user to set ANY HbH
2782 * options, since per-option restriction has too much
2783 * overhead.
2784 */
2785 if (cred != NULL) {
2786 error = priv_check_cred(cred,
2787 PRIV_NETINET_SETHDROPTS, 0);
2788 if (error)
2789 return (error);
2790 }
2791
2792 if (len == 0) {
2793 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2794 break; /* just remove the option */
2795 }
2796
2797 /* message length validation */
2798 if (len < sizeof(struct ip6_hbh))
2799 return (EINVAL);
2800 hbh = (struct ip6_hbh *)buf;
2801 hbhlen = (hbh->ip6h_len + 1) << 3;
2802 if (len != hbhlen)
2803 return (EINVAL);
2804
2805 /* turn off the previous option, then set the new option. */
2806 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2807 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2808 if (opt->ip6po_hbh == NULL)
2809 return (ENOBUFS);
2810 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2811
2812 break;
2813 }
2814
2815 case IPV6_2292DSTOPTS:
2816 case IPV6_DSTOPTS:
2817 case IPV6_RTHDRDSTOPTS:
2818 {
2819 struct ip6_dest *dest, **newdest = NULL;
2820 int destlen;
2821
2822 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2823 error = priv_check_cred(cred,
2824 PRIV_NETINET_SETHDROPTS, 0);
2825 if (error)
2826 return (error);
2827 }
2828
2829 if (len == 0) {
2830 ip6_clearpktopts(opt, optname);
2831 break; /* just remove the option */
2832 }
2833
2834 /* message length validation */
2835 if (len < sizeof(struct ip6_dest))
2836 return (EINVAL);
2837 dest = (struct ip6_dest *)buf;
2838 destlen = (dest->ip6d_len + 1) << 3;
2839 if (len != destlen)
2840 return (EINVAL);
2841
2842 /*
2843 * Determine the position that the destination options header
2844 * should be inserted; before or after the routing header.
2845 */
2846 switch (optname) {
2847 case IPV6_2292DSTOPTS:
2848 /*
2849 * The old advacned API is ambiguous on this point.
2850 * Our approach is to determine the position based
2851 * according to the existence of a routing header.
2852 * Note, however, that this depends on the order of the
2853 * extension headers in the ancillary data; the 1st
2854 * part of the destination options header must appear
2855 * before the routing header in the ancillary data,
2856 * too.
2857 * RFC3542 solved the ambiguity by introducing
2858 * separate ancillary data or option types.
2859 */
2860 if (opt->ip6po_rthdr == NULL)
2861 newdest = &opt->ip6po_dest1;
2862 else
2863 newdest = &opt->ip6po_dest2;
2864 break;
2865 case IPV6_RTHDRDSTOPTS:
2866 newdest = &opt->ip6po_dest1;
2867 break;
2868 case IPV6_DSTOPTS:
2869 newdest = &opt->ip6po_dest2;
2870 break;
2871 }
2872
2873 /* turn off the previous option, then set the new option. */
2874 ip6_clearpktopts(opt, optname);
2875 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2876 if (*newdest == NULL)
2877 return (ENOBUFS);
2878 bcopy(dest, *newdest, destlen);
2879
2880 break;
2881 }
2882
2883 case IPV6_2292RTHDR:
2884 case IPV6_RTHDR:
2885 {
2886 struct ip6_rthdr *rth;
2887 int rthlen;
2888
2889 if (len == 0) {
2890 ip6_clearpktopts(opt, IPV6_RTHDR);
2891 break; /* just remove the option */
2892 }
2893
2894 /* message length validation */
2895 if (len < sizeof(struct ip6_rthdr))
2896 return (EINVAL);
2897 rth = (struct ip6_rthdr *)buf;
2898 rthlen = (rth->ip6r_len + 1) << 3;
2899 if (len != rthlen)
2900 return (EINVAL);
2901
2902 switch (rth->ip6r_type) {
2903 case IPV6_RTHDR_TYPE_0:
2904 if (rth->ip6r_len == 0) /* must contain one addr */
2905 return (EINVAL);
2906 if (rth->ip6r_len % 2) /* length must be even */
2907 return (EINVAL);
2908 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2909 return (EINVAL);
2910 break;
2911 default:
2912 return (EINVAL); /* not supported */
2913 }
2914
2915 /* turn off the previous option */
2916 ip6_clearpktopts(opt, IPV6_RTHDR);
2917 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2918 if (opt->ip6po_rthdr == NULL)
2919 return (ENOBUFS);
2920 bcopy(rth, opt->ip6po_rthdr, rthlen);
2921
2922 break;
2923 }
2924
2925 case IPV6_USE_MIN_MTU:
2926 if (len != sizeof(int))
2927 return (EINVAL);
2928 minmtupolicy = *(int *)buf;
2929 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2930 minmtupolicy != IP6PO_MINMTU_DISABLE &&
2931 minmtupolicy != IP6PO_MINMTU_ALL) {
2932 return (EINVAL);
2933 }
2934 opt->ip6po_minmtu = minmtupolicy;
2935 break;
2936
2937 case IPV6_DONTFRAG:
2938 if (len != sizeof(int))
2939 return (EINVAL);
2940
2941 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2942 /*
2943 * we ignore this option for TCP sockets.
2944 * (RFC3542 leaves this case unspecified.)
2945 */
2946 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2947 } else
2948 opt->ip6po_flags |= IP6PO_DONTFRAG;
2949 break;
2950
2951 case IPV6_PREFER_TEMPADDR:
2952 if (len != sizeof(int))
2953 return (EINVAL);
2954 preftemp = *(int *)buf;
2955 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2956 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2957 preftemp != IP6PO_TEMPADDR_PREFER) {
2958 return (EINVAL);
2959 }
2960 opt->ip6po_prefer_tempaddr = preftemp;
2961 break;
2962
2963 default:
2964 return (ENOPROTOOPT);
2965 } /* end of switch */
2966
2967 return (0);
2968 }
2969
2970 /*
2971 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2972 * packet to the input queue of a specified interface. Note that this
2973 * calls the output routine of the loopback "driver", but with an interface
2974 * pointer that might NOT be &loif -- easier than replicating that code here.
2975 */
2976 void
2977 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
2978 {
2979 struct mbuf *copym;
2980 struct ip6_hdr *ip6;
2981
2982 copym = m_copy(m, 0, M_COPYALL);
2983 if (copym == NULL)
2984 return;
2985
2986 /*
2987 * Make sure to deep-copy IPv6 header portion in case the data
2988 * is in an mbuf cluster, so that we can safely override the IPv6
2989 * header portion later.
2990 */
2991 if (!M_WRITABLE(copym) ||
2992 copym->m_len < sizeof(struct ip6_hdr)) {
2993 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2994 if (copym == NULL)
2995 return;
2996 }
2997 ip6 = mtod(copym, struct ip6_hdr *);
2998 /*
2999 * clear embedded scope identifiers if necessary.
3000 * in6_clearscope will touch the addresses only when necessary.
3001 */
3002 in6_clearscope(&ip6->ip6_src);
3003 in6_clearscope(&ip6->ip6_dst);
3004 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3005 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3006 CSUM_PSEUDO_HDR;
3007 copym->m_pkthdr.csum_data = 0xffff;
3008 }
3009 if_simloop(ifp, copym, AF_INET6, 0);
3010 }
3011
3012 /*
3013 * Chop IPv6 header off from the payload.
3014 */
3015 static int
3016 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3017 {
3018 struct mbuf *mh;
3019 struct ip6_hdr *ip6;
3020
3021 ip6 = mtod(m, struct ip6_hdr *);
3022 if (m->m_len > sizeof(*ip6)) {
3023 mh = m_gethdr(M_NOWAIT, MT_DATA);
3024 if (mh == NULL) {
3025 m_freem(m);
3026 return ENOBUFS;
3027 }
3028 m_move_pkthdr(mh, m);
3029 M_ALIGN(mh, sizeof(*ip6));
3030 m->m_len -= sizeof(*ip6);
3031 m->m_data += sizeof(*ip6);
3032 mh->m_next = m;
3033 m = mh;
3034 m->m_len = sizeof(*ip6);
3035 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3036 }
3037 exthdrs->ip6e_ip6 = m;
3038 return 0;
3039 }
3040
3041 /*
3042 * Compute IPv6 extension header length.
3043 */
3044 int
3045 ip6_optlen(struct inpcb *in6p)
3046 {
3047 int len;
3048
3049 if (!in6p->in6p_outputopts)
3050 return 0;
3051
3052 len = 0;
3053 #define elen(x) \
3054 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3055
3056 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3057 if (in6p->in6p_outputopts->ip6po_rthdr)
3058 /* dest1 is valid with rthdr only */
3059 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3060 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3061 len += elen(in6p->in6p_outputopts->ip6po_dest2);
3062 return len;
3063 #undef elen
3064 }
Cache object: bd3c88265ea7b8ff67a7f3d7db62e939
|