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
63 #include "opt_inet.h"
64 #include "opt_inet6.h"
65 #include "opt_ipsec.h"
66 #include "opt_kern_tls.h"
67 #include "opt_ratelimit.h"
68 #include "opt_route.h"
69 #include "opt_rss.h"
70 #include "opt_sctp.h"
71
72 #include <sys/param.h>
73 #include <sys/kernel.h>
74 #include <sys/ktls.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_private.h>
91 #include <net/if_vlan_var.h>
92 #include <net/if_llatbl.h>
93 #include <net/ethernet.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/route/nhop.h>
97 #include <net/pfil.h>
98 #include <net/rss_config.h>
99 #include <net/vnet.h>
100
101 #include <netinet/in.h>
102 #include <netinet/in_var.h>
103 #include <netinet/ip_var.h>
104 #include <netinet6/in6_fib.h>
105 #include <netinet6/in6_var.h>
106 #include <netinet/ip6.h>
107 #include <netinet/icmp6.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet/in_pcb.h>
110 #include <netinet/tcp_var.h>
111 #include <netinet6/nd6.h>
112 #include <netinet6/in6_rss.h>
113 #include <netinet6/ip6_mroute.h>
114
115 #include <netipsec/ipsec_support.h>
116 #if defined(SCTP) || defined(SCTP_SUPPORT)
117 #include <netinet/sctp.h>
118 #include <netinet/sctp_crc32.h>
119 #endif
120
121 #include <netinet6/scope6_var.h>
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 inpcb *, 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_splithdr(struct mbuf *, struct ip6_exthdrs *);
147 static void ip6_getpmtu(struct route_in6 *, int,
148 struct ifnet *, const struct in6_addr *, u_long *, u_int, u_int);
149 static void ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
150 u_long *, u_int);
151 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
152 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
153
154 /*
155 * Make an extension header from option data. hp is the source,
156 * mp is the destination, and _ol is the optlen.
157 */
158 #define MAKE_EXTHDR(hp, mp, _ol) \
159 do { \
160 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
161 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
162 ((eh)->ip6e_len + 1) << 3); \
163 if (error) \
164 goto freehdrs; \
165 (_ol) += (*(mp))->m_len; \
166 } while (/*CONSTCOND*/ 0)
167
168 /*
169 * Form a chain of extension headers.
170 * m is the extension header mbuf
171 * mp is the previous mbuf in the chain
172 * p is the next header
173 * i is the type of option.
174 */
175 #define MAKE_CHAIN(m, mp, p, i)\
176 do {\
177 if (m) {\
178 if (!hdrsplit) \
179 panic("%s:%d: assumption failed: "\
180 "hdr not split: hdrsplit %d exthdrs %p",\
181 __func__, __LINE__, hdrsplit, &exthdrs);\
182 *mtod((m), u_char *) = *(p);\
183 *(p) = (i);\
184 p = mtod((m), u_char *);\
185 (m)->m_next = (mp)->m_next;\
186 (mp)->m_next = (m);\
187 (mp) = (m);\
188 }\
189 } while (/*CONSTCOND*/ 0)
190
191 void
in6_delayed_cksum(struct mbuf * m,uint32_t plen,u_short offset)192 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
193 {
194 u_short csum;
195
196 csum = in_cksum_skip(m, offset + plen, offset);
197 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
198 csum = 0xffff;
199 offset += m->m_pkthdr.csum_data; /* checksum offset */
200
201 if (offset + sizeof(csum) > m->m_len)
202 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
203 else
204 *(u_short *)mtodo(m, offset) = csum;
205 }
206
207 static void
ip6_output_delayed_csum(struct mbuf * m,struct ifnet * ifp,int csum_flags,int plen,int optlen)208 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
209 int plen, int optlen)
210 {
211
212 KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
213 "csum_flags %#x",
214 __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
215
216 if (csum_flags & CSUM_DELAY_DATA_IPV6) {
217 in6_delayed_cksum(m, plen - optlen,
218 sizeof(struct ip6_hdr) + optlen);
219 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
220 }
221 #if defined(SCTP) || defined(SCTP_SUPPORT)
222 if (csum_flags & CSUM_SCTP_IPV6) {
223 sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
224 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
225 }
226 #endif
227 }
228
229 int
ip6_fragment(struct ifnet * ifp,struct mbuf * m0,int hlen,u_char nextproto,int fraglen,uint32_t id)230 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
231 int fraglen , uint32_t id)
232 {
233 struct mbuf *m, **mnext, *m_frgpart;
234 struct ip6_hdr *ip6, *mhip6;
235 struct ip6_frag *ip6f;
236 int off;
237 int error;
238 int tlen = m0->m_pkthdr.len;
239
240 KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
241
242 m = m0;
243 ip6 = mtod(m, struct ip6_hdr *);
244 mnext = &m->m_nextpkt;
245
246 for (off = hlen; off < tlen; off += fraglen) {
247 m = m_gethdr(M_NOWAIT, MT_DATA);
248 if (!m) {
249 IP6STAT_INC(ip6s_odropped);
250 return (ENOBUFS);
251 }
252
253 /*
254 * Make sure the complete packet header gets copied
255 * from the originating mbuf to the newly created
256 * mbuf. This also ensures that existing firewall
257 * classification(s), VLAN tags and so on get copied
258 * to the resulting fragmented packet(s):
259 */
260 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
261 m_free(m);
262 IP6STAT_INC(ip6s_odropped);
263 return (ENOBUFS);
264 }
265
266 *mnext = m;
267 mnext = &m->m_nextpkt;
268 m->m_data += max_linkhdr;
269 mhip6 = mtod(m, struct ip6_hdr *);
270 *mhip6 = *ip6;
271 m->m_len = sizeof(*mhip6);
272 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
273 if (error) {
274 IP6STAT_INC(ip6s_odropped);
275 return (error);
276 }
277 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
278 if (off + fraglen >= tlen)
279 fraglen = tlen - off;
280 else
281 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
282 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
283 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
284 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
285 IP6STAT_INC(ip6s_odropped);
286 return (ENOBUFS);
287 }
288 m_cat(m, m_frgpart);
289 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
290 ip6f->ip6f_reserved = 0;
291 ip6f->ip6f_ident = id;
292 ip6f->ip6f_nxt = nextproto;
293 IP6STAT_INC(ip6s_ofragments);
294 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
295 }
296
297 return (0);
298 }
299
300 static int
ip6_output_send(struct inpcb * inp,struct ifnet * ifp,struct ifnet * origifp,struct mbuf * m,struct sockaddr_in6 * dst,struct route_in6 * ro,bool stamp_tag)301 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
302 struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
303 bool stamp_tag)
304 {
305 #ifdef KERN_TLS
306 struct ktls_session *tls = NULL;
307 #endif
308 struct m_snd_tag *mst;
309 int error;
310
311 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
312 mst = NULL;
313
314 #ifdef KERN_TLS
315 /*
316 * If this is an unencrypted TLS record, save a reference to
317 * the record. This local reference is used to call
318 * ktls_output_eagain after the mbuf has been freed (thus
319 * dropping the mbuf's reference) in if_output.
320 */
321 if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
322 tls = ktls_hold(m->m_next->m_epg_tls);
323 mst = tls->snd_tag;
324
325 /*
326 * If a TLS session doesn't have a valid tag, it must
327 * have had an earlier ifp mismatch, so drop this
328 * packet.
329 */
330 if (mst == NULL) {
331 m_freem(m);
332 error = EAGAIN;
333 goto done;
334 }
335 /*
336 * Always stamp tags that include NIC ktls.
337 */
338 stamp_tag = true;
339 }
340 #endif
341 #ifdef RATELIMIT
342 if (inp != NULL && mst == NULL) {
343 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
344 (inp->inp_snd_tag != NULL &&
345 inp->inp_snd_tag->ifp != ifp))
346 in_pcboutput_txrtlmt(inp, ifp, m);
347
348 if (inp->inp_snd_tag != NULL)
349 mst = inp->inp_snd_tag;
350 }
351 #endif
352 if (stamp_tag && mst != NULL) {
353 KASSERT(m->m_pkthdr.rcvif == NULL,
354 ("trying to add a send tag to a forwarded packet"));
355 if (mst->ifp != ifp) {
356 m_freem(m);
357 error = EAGAIN;
358 goto done;
359 }
360
361 /* stamp send tag on mbuf */
362 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
363 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
364 }
365
366 error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
367
368 done:
369 /* Check for route change invalidating send tags. */
370 #ifdef KERN_TLS
371 if (tls != NULL) {
372 if (error == EAGAIN)
373 error = ktls_output_eagain(inp, tls);
374 ktls_free(tls);
375 }
376 #endif
377 #ifdef RATELIMIT
378 if (error == EAGAIN)
379 in_pcboutput_eagain(inp);
380 #endif
381 return (error);
382 }
383
384 /*
385 * IP6 output.
386 * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
387 * nxt, hlim, src, dst).
388 * This function may modify ver and hlim only.
389 * The mbuf chain containing the packet will be freed.
390 * The mbuf opt, if present, will not be freed.
391 * If route_in6 ro is present and has ro_nh initialized, route lookup would be
392 * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
393 * then result of route lookup is stored in ro->ro_nh.
394 *
395 * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
396 * is uint32_t. So we use u_long to hold largest one, which is rt_mtu.
397 *
398 * ifpp - XXX: just for statistics
399 */
400 int
ip6_output(struct mbuf * m0,struct ip6_pktopts * opt,struct route_in6 * ro,int flags,struct ip6_moptions * im6o,struct ifnet ** ifpp,struct inpcb * inp)401 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
402 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
403 struct ifnet **ifpp, struct inpcb *inp)
404 {
405 struct ip6_hdr *ip6;
406 struct ifnet *ifp, *origifp;
407 struct mbuf *m = m0;
408 struct mbuf *mprev;
409 struct route_in6 *ro_pmtu;
410 struct nhop_object *nh;
411 struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
412 struct in6_addr odst;
413 u_char *nexthdrp;
414 int tlen, len;
415 int error = 0;
416 int vlan_pcp = -1;
417 struct in6_ifaddr *ia = NULL;
418 u_long mtu;
419 int dontfrag;
420 u_int32_t optlen, plen = 0, unfragpartlen;
421 struct ip6_exthdrs exthdrs;
422 struct in6_addr src0, dst0;
423 u_int32_t zone;
424 bool hdrsplit;
425 int sw_csum, tso;
426 int needfiblookup;
427 uint32_t fibnum;
428 struct m_tag *fwd_tag = NULL;
429 uint32_t id;
430 uint32_t optvalid;
431
432 NET_EPOCH_ASSERT();
433
434 if (inp != NULL) {
435 INP_LOCK_ASSERT(inp);
436 M_SETFIB(m, inp->inp_inc.inc_fibnum);
437 if ((flags & IP_NODEFAULTFLOWID) == 0) {
438 /* Unconditionally set flowid. */
439 m->m_pkthdr.flowid = inp->inp_flowid;
440 M_HASHTYPE_SET(m, inp->inp_flowtype);
441 }
442 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
443 vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
444 INP_2PCP_SHIFT;
445 #ifdef NUMA
446 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
447 #endif
448 }
449
450 /* Source address validation. */
451 ip6 = mtod(m, struct ip6_hdr *);
452 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
453 (flags & IPV6_UNSPECSRC) == 0) {
454 error = EOPNOTSUPP;
455 IP6STAT_INC(ip6s_badscope);
456 goto bad;
457 }
458 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
459 error = EOPNOTSUPP;
460 IP6STAT_INC(ip6s_badscope);
461 goto bad;
462 }
463
464 /*
465 * If we are given packet options to add extension headers prepare them.
466 * Calculate the total length of the extension header chain.
467 * Keep the length of the unfragmentable part for fragmentation.
468 */
469 bzero(&exthdrs, sizeof(exthdrs));
470 optlen = optvalid = 0;
471 unfragpartlen = sizeof(struct ip6_hdr);
472 if (opt) {
473 optvalid = opt->ip6po_valid;
474
475 /* Hop-by-Hop options header. */
476 if ((optvalid & IP6PO_VALID_HBH) != 0)
477 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
478
479 /* Destination options header (1st part). */
480 if ((optvalid & IP6PO_VALID_RHINFO) != 0) {
481 #ifndef RTHDR_SUPPORT_IMPLEMENTED
482 /*
483 * If there is a routing header, discard the packet
484 * right away here. RH0/1 are obsolete and we do not
485 * currently support RH2/3/4.
486 * People trying to use RH253/254 may want to disable
487 * this check.
488 * The moment we do support any routing header (again)
489 * this block should check the routing type more
490 * selectively.
491 */
492 error = EINVAL;
493 goto bad;
494 #endif
495
496 /*
497 * Destination options header (1st part).
498 * This only makes sense with a routing header.
499 * See Section 9.2 of RFC 3542.
500 * Disabling this part just for MIP6 convenience is
501 * a bad idea. We need to think carefully about a
502 * way to make the advanced API coexist with MIP6
503 * options, which might automatically be inserted in
504 * the kernel.
505 */
506 if ((optvalid & IP6PO_VALID_DEST1) != 0)
507 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
508 optlen);
509 }
510 /* Routing header. */
511 if ((optvalid & IP6PO_VALID_RHINFO) != 0)
512 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
513
514 unfragpartlen += optlen;
515
516 /*
517 * NOTE: we don't add AH/ESP length here (done in
518 * ip6_ipsec_output()).
519 */
520
521 /* Destination options header (2nd part). */
522 if ((optvalid & IP6PO_VALID_DEST2) != 0)
523 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
524 }
525
526 /*
527 * If there is at least one extension header,
528 * separate IP6 header from the payload.
529 */
530 hdrsplit = false;
531 if (optlen) {
532 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
533 m = NULL;
534 goto freehdrs;
535 }
536 m = exthdrs.ip6e_ip6;
537 ip6 = mtod(m, struct ip6_hdr *);
538 hdrsplit = true;
539 }
540
541 /* Adjust mbuf packet header length. */
542 m->m_pkthdr.len += optlen;
543 plen = m->m_pkthdr.len - sizeof(*ip6);
544
545 if (plen > IPV6_MAXPACKET) {
546 error = EMSGSIZE;
547 goto freehdrs;
548 } else
549 ip6->ip6_plen = htons(plen);
550 nexthdrp = &ip6->ip6_nxt;
551
552 if (optlen) {
553 /*
554 * Concatenate headers and fill in next header fields.
555 * Here we have, on "m"
556 * IPv6 payload
557 * and we insert headers accordingly.
558 * Finally, we should be getting:
559 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
560 *
561 * During the header composing process "m" points to IPv6
562 * header. "mprev" points to an extension header prior to esp.
563 */
564 mprev = m;
565
566 /*
567 * We treat dest2 specially. This makes IPsec processing
568 * much easier. The goal here is to make mprev point the
569 * mbuf prior to dest2.
570 *
571 * Result: IPv6 dest2 payload.
572 * m and mprev will point to IPv6 header.
573 */
574 if (exthdrs.ip6e_dest2) {
575 if (!hdrsplit)
576 panic("%s:%d: assumption failed: "
577 "hdr not split: hdrsplit %d exthdrs %p",
578 __func__, __LINE__, hdrsplit, &exthdrs);
579 exthdrs.ip6e_dest2->m_next = m->m_next;
580 m->m_next = exthdrs.ip6e_dest2;
581 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
582 ip6->ip6_nxt = IPPROTO_DSTOPTS;
583 }
584
585 /*
586 * Result: IPv6 hbh dest1 rthdr dest2 payload.
587 * m will point to IPv6 header. mprev will point to the
588 * extension header prior to dest2 (rthdr in the above case).
589 */
590 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
591 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
592 IPPROTO_DSTOPTS);
593 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
594 IPPROTO_ROUTING);
595 }
596
597 IP6STAT_INC(ip6s_localout);
598
599 /* Route packet. */
600 ro_pmtu = ro;
601 if ((optvalid & IP6PO_VALID_RHINFO) != 0)
602 ro = &opt->ip6po_route;
603 if (ro != NULL)
604 dst = (struct sockaddr_in6 *)&ro->ro_dst;
605 else
606 dst = &sin6;
607 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
608
609 again:
610 /*
611 * If specified, try to fill in the traffic class field.
612 * Do not override if a non-zero value is already set.
613 * We check the diffserv field and the ECN field separately.
614 */
615 if ((optvalid & IP6PO_VALID_TC) != 0){
616 int mask = 0;
617
618 if (IPV6_DSCP(ip6) == 0)
619 mask |= 0xfc;
620 if (IPV6_ECN(ip6) == 0)
621 mask |= 0x03;
622 if (mask != 0)
623 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
624 }
625
626 /* Fill in or override the hop limit field, if necessary. */
627 if ((optvalid & IP6PO_VALID_HLIM) != 0)
628 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
629 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
630 if (im6o != NULL)
631 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
632 else
633 ip6->ip6_hlim = V_ip6_defmcasthlim;
634 }
635
636 if (ro == NULL || ro->ro_nh == NULL) {
637 bzero(dst, sizeof(*dst));
638 dst->sin6_family = AF_INET6;
639 dst->sin6_len = sizeof(*dst);
640 dst->sin6_addr = ip6->ip6_dst;
641 }
642 /*
643 * Validate route against routing table changes.
644 * Make sure that the address family is set in route.
645 */
646 nh = NULL;
647 ifp = NULL;
648 mtu = 0;
649 if (ro != NULL) {
650 if (ro->ro_nh != NULL && inp != NULL) {
651 ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
652 NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
653 fibnum);
654 }
655 if (ro->ro_nh != NULL && fwd_tag == NULL &&
656 (!NH_IS_VALID(ro->ro_nh) ||
657 ro->ro_dst.sin6_family != AF_INET6 ||
658 !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
659 RO_INVALIDATE_CACHE(ro);
660
661 if (ro->ro_nh != NULL && fwd_tag == NULL &&
662 ro->ro_dst.sin6_family == AF_INET6 &&
663 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
664 /* Nexthop is valid and contains valid ifp */
665 nh = ro->ro_nh;
666 } else {
667 if (ro->ro_lle)
668 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
669 ro->ro_lle = NULL;
670 if (fwd_tag == NULL) {
671 bzero(&dst_sa, sizeof(dst_sa));
672 dst_sa.sin6_family = AF_INET6;
673 dst_sa.sin6_len = sizeof(dst_sa);
674 dst_sa.sin6_addr = ip6->ip6_dst;
675 }
676 error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
677 &nh, fibnum, m->m_pkthdr.flowid);
678 if (error != 0) {
679 IP6STAT_INC(ip6s_noroute);
680 if (ifp != NULL)
681 in6_ifstat_inc(ifp, ifs6_out_discard);
682 goto bad;
683 }
684 /*
685 * At this point at least @ifp is not NULL
686 * Can be the case when dst is multicast, link-local or
687 * interface is explicitly specificed by the caller.
688 */
689 }
690 if (nh == NULL) {
691 /*
692 * If in6_selectroute() does not return a nexthop
693 * dst may not have been updated.
694 */
695 *dst = dst_sa; /* XXX */
696 origifp = ifp;
697 mtu = ifp->if_mtu;
698 } else {
699 ifp = nh->nh_ifp;
700 origifp = nh->nh_aifp;
701 ia = (struct in6_ifaddr *)(nh->nh_ifa);
702 counter_u64_add(nh->nh_pksent, 1);
703 }
704 } else {
705 struct nhop_object *nh;
706 struct in6_addr kdst;
707 uint32_t scopeid;
708
709 if (fwd_tag == NULL) {
710 bzero(&dst_sa, sizeof(dst_sa));
711 dst_sa.sin6_family = AF_INET6;
712 dst_sa.sin6_len = sizeof(dst_sa);
713 dst_sa.sin6_addr = ip6->ip6_dst;
714 }
715
716 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
717 im6o != NULL &&
718 (ifp = im6o->im6o_multicast_ifp) != NULL) {
719 /* We do not need a route lookup. */
720 *dst = dst_sa; /* XXX */
721 origifp = ifp;
722 goto nonh6lookup;
723 }
724
725 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
726
727 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
728 IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
729 if (scopeid > 0) {
730 ifp = in6_getlinkifnet(scopeid);
731 if (ifp == NULL) {
732 error = EHOSTUNREACH;
733 goto bad;
734 }
735 *dst = dst_sa; /* XXX */
736 origifp = ifp;
737 goto nonh6lookup;
738 }
739 }
740
741 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
742 m->m_pkthdr.flowid);
743 if (nh == NULL) {
744 IP6STAT_INC(ip6s_noroute);
745 /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
746 error = EHOSTUNREACH;
747 goto bad;
748 }
749
750 ifp = nh->nh_ifp;
751 origifp = nh->nh_aifp;
752 ia = ifatoia6(nh->nh_ifa);
753 if (nh->nh_flags & NHF_GATEWAY)
754 dst->sin6_addr = nh->gw6_sa.sin6_addr;
755 else if (fwd_tag != NULL)
756 dst->sin6_addr = dst_sa.sin6_addr;
757 nonh6lookup:
758 ;
759 }
760 /*
761 * At this point ifp MUST be pointing to the valid transmit ifp.
762 * origifp MUST be valid and pointing to either the same ifp or,
763 * in case of loopback output, to the interface which ip6_src
764 * belongs to.
765 * Examples:
766 * fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
767 * fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
768 * ::1 -> ::1 -> ifp=lo0, origifp=lo0
769 *
770 * mtu can be 0 and will be refined later.
771 */
772 KASSERT((ifp != NULL), ("output interface must not be NULL"));
773 KASSERT((origifp != NULL), ("output address interface must not be NULL"));
774
775 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
776 /*
777 * IPSec checking which handles several cases.
778 * FAST IPSEC: We re-injected the packet.
779 * XXX: need scope argument.
780 */
781 if (IPSEC_ENABLED(ipv6)) {
782 if ((error = IPSEC_OUTPUT(ipv6, ifp, m, inp, mtu == 0 ?
783 ifp->if_mtu : mtu)) != 0) {
784 if (error == EINPROGRESS)
785 error = 0;
786 goto done;
787 }
788 }
789 #endif /* IPSEC */
790
791 if ((flags & IPV6_FORWARDING) == 0) {
792 /* XXX: the FORWARDING flag can be set for mrouting. */
793 in6_ifstat_inc(ifp, ifs6_out_request);
794 }
795
796 /* Setup data structures for scope ID checks. */
797 src0 = ip6->ip6_src;
798 bzero(&src_sa, sizeof(src_sa));
799 src_sa.sin6_family = AF_INET6;
800 src_sa.sin6_len = sizeof(src_sa);
801 src_sa.sin6_addr = ip6->ip6_src;
802
803 dst0 = ip6->ip6_dst;
804 /* Re-initialize to be sure. */
805 bzero(&dst_sa, sizeof(dst_sa));
806 dst_sa.sin6_family = AF_INET6;
807 dst_sa.sin6_len = sizeof(dst_sa);
808 dst_sa.sin6_addr = ip6->ip6_dst;
809
810 /* Check for valid scope ID. */
811 if (in6_setscope(&src0, origifp, &zone) == 0 &&
812 sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
813 in6_setscope(&dst0, origifp, &zone) == 0 &&
814 sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
815 /*
816 * The outgoing interface is in the zone of the source
817 * and destination addresses.
818 *
819 */
820 } else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
821 sa6_recoverscope(&src_sa) != 0 ||
822 sa6_recoverscope(&dst_sa) != 0 ||
823 dst_sa.sin6_scope_id == 0 ||
824 (src_sa.sin6_scope_id != 0 &&
825 src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
826 ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
827 /*
828 * If the destination network interface is not a
829 * loopback interface, or the destination network
830 * address has no scope ID, or the source address has
831 * a scope ID set which is different from the
832 * destination address one, or there is no network
833 * interface representing this scope ID, the address
834 * pair is considered invalid.
835 */
836 IP6STAT_INC(ip6s_badscope);
837 in6_ifstat_inc(origifp, ifs6_out_discard);
838 if (error == 0)
839 error = EHOSTUNREACH; /* XXX */
840 goto bad;
841 }
842 /* All scope ID checks are successful. */
843
844 if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
845 if ((optvalid & IP6PO_VALID_NHINFO) != 0) {
846 /*
847 * The nexthop is explicitly specified by the
848 * application. We assume the next hop is an IPv6
849 * address.
850 */
851 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
852 }
853 else if ((nh->nh_flags & NHF_GATEWAY))
854 dst = &nh->gw6_sa;
855 }
856
857 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
858 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
859 } else {
860 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
861 in6_ifstat_inc(ifp, ifs6_out_mcast);
862
863 /* Confirm that the outgoing interface supports multicast. */
864 if (!(ifp->if_flags & IFF_MULTICAST)) {
865 IP6STAT_INC(ip6s_noroute);
866 in6_ifstat_inc(ifp, ifs6_out_discard);
867 error = ENETUNREACH;
868 goto bad;
869 }
870 if ((im6o == NULL && in6_mcast_loop) ||
871 (im6o && im6o->im6o_multicast_loop)) {
872 /*
873 * Loop back multicast datagram if not expressly
874 * forbidden to do so, even if we have not joined
875 * the address; protocols will filter it later,
876 * thus deferring a hash lookup and lock acquisition
877 * at the expense of an m_copym().
878 */
879 ip6_mloopback(ifp, m);
880 } else {
881 /*
882 * If we are acting as a multicast router, perform
883 * multicast forwarding as if the packet had just
884 * arrived on the interface to which we are about
885 * to send. The multicast forwarding function
886 * recursively calls this function, using the
887 * IPV6_FORWARDING flag to prevent infinite recursion.
888 *
889 * Multicasts that are looped back by ip6_mloopback(),
890 * above, will be forwarded by the ip6_input() routine,
891 * if necessary.
892 */
893 if (V_ip6_mrouting_enabled &&
894 (flags & IPV6_FORWARDING) == 0) {
895 /*
896 * XXX: ip6_mforward expects that rcvif is NULL
897 * when it is called from the originating path.
898 * However, it may not always be the case.
899 */
900 m->m_pkthdr.rcvif = NULL;
901 if (ip6_mforward(ip6, ifp, m) != 0) {
902 m_freem(m);
903 goto done;
904 }
905 }
906 }
907 /*
908 * Multicasts with a hoplimit of zero may be looped back,
909 * above, but must not be transmitted on a network.
910 * Also, multicasts addressed to the loopback interface
911 * are not sent -- the above call to ip6_mloopback() will
912 * loop back a copy if this host actually belongs to the
913 * destination group on the loopback interface.
914 */
915 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
916 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
917 m_freem(m);
918 goto done;
919 }
920 }
921
922 /*
923 * Fill the outgoing inteface to tell the upper layer
924 * to increment per-interface statistics.
925 */
926 if (ifpp)
927 *ifpp = ifp;
928
929 /* Determine path MTU. */
930 ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst, &mtu, fibnum,
931 *nexthdrp);
932 KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p fibnum %u",
933 __func__, __LINE__, mtu, ro_pmtu, ro, ifp, fibnum));
934
935 /*
936 * The caller of this function may specify to use the minimum MTU
937 * in some cases.
938 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
939 * setting. The logic is a bit complicated; by default, unicast
940 * packets will follow path MTU while multicast packets will be sent at
941 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
942 * including unicast ones will be sent at the minimum MTU. Multicast
943 * packets will always be sent at the minimum MTU unless
944 * IP6PO_MINMTU_DISABLE is explicitly specified.
945 * See RFC 3542 for more details.
946 */
947 if (mtu > IPV6_MMTU) {
948 if ((flags & IPV6_MINMTU))
949 mtu = IPV6_MMTU;
950 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
951 mtu = IPV6_MMTU;
952 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
953 (opt == NULL ||
954 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
955 mtu = IPV6_MMTU;
956 }
957 }
958
959 /*
960 * Clear embedded scope identifiers if necessary.
961 * in6_clearscope() will touch the addresses only when necessary.
962 */
963 in6_clearscope(&ip6->ip6_src);
964 in6_clearscope(&ip6->ip6_dst);
965
966 /*
967 * If the outgoing packet contains a hop-by-hop options header,
968 * it must be examined and processed even by the source node.
969 * (RFC 2460, section 4.)
970 */
971 if (exthdrs.ip6e_hbh) {
972 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
973 u_int32_t dummy; /* XXX unused */
974
975 #ifdef DIAGNOSTIC
976 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
977 panic("ip6e_hbh is not contiguous");
978 #endif
979 /*
980 * XXX: if we have to send an ICMPv6 error to the sender,
981 * we need the M_LOOP flag since icmp6_error() expects
982 * the IPv6 and the hop-by-hop options header are
983 * contiguous unless the flag is set.
984 */
985 m->m_flags |= M_LOOP;
986 m->m_pkthdr.rcvif = ifp;
987 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
988 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
989 &dummy) < 0) {
990 /* m was already freed at this point. */
991 error = EINVAL;/* better error? */
992 goto done;
993 }
994 m->m_flags &= ~M_LOOP; /* XXX */
995 m->m_pkthdr.rcvif = NULL;
996 }
997
998 /* Jump over all PFIL processing if hooks are not active. */
999 if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1000 goto passout;
1001
1002 odst = ip6->ip6_dst;
1003 /* Run through list of hooks for output packets. */
1004 switch (pfil_mbuf_out(V_inet6_pfil_head, &m, ifp, inp)) {
1005 case PFIL_PASS:
1006 ip6 = mtod(m, struct ip6_hdr *);
1007 break;
1008 case PFIL_DROPPED:
1009 error = EACCES;
1010 /* FALLTHROUGH */
1011 case PFIL_CONSUMED:
1012 goto done;
1013 }
1014
1015 needfiblookup = 0;
1016 /* See if destination IP address was changed by packet filter. */
1017 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1018 m->m_flags |= M_SKIP_FIREWALL;
1019 /* If destination is now ourself drop to ip6_input(). */
1020 if (in6_localip(&ip6->ip6_dst)) {
1021 m->m_flags |= M_FASTFWD_OURS;
1022 if (m->m_pkthdr.rcvif == NULL)
1023 m->m_pkthdr.rcvif = V_loif;
1024 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1025 m->m_pkthdr.csum_flags |=
1026 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1027 m->m_pkthdr.csum_data = 0xffff;
1028 }
1029 #if defined(SCTP) || defined(SCTP_SUPPORT)
1030 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1031 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1032 #endif
1033 error = netisr_queue(NETISR_IPV6, m);
1034 goto done;
1035 } else {
1036 if (ro != NULL)
1037 RO_INVALIDATE_CACHE(ro);
1038 needfiblookup = 1; /* Redo the routing table lookup. */
1039 }
1040 }
1041 /* See if fib was changed by packet filter. */
1042 if (fibnum != M_GETFIB(m)) {
1043 m->m_flags |= M_SKIP_FIREWALL;
1044 fibnum = M_GETFIB(m);
1045 if (ro != NULL)
1046 RO_INVALIDATE_CACHE(ro);
1047 needfiblookup = 1;
1048 }
1049 if (needfiblookup)
1050 goto again;
1051
1052 /* See if local, if yes, send it to netisr. */
1053 if (m->m_flags & M_FASTFWD_OURS) {
1054 if (m->m_pkthdr.rcvif == NULL)
1055 m->m_pkthdr.rcvif = V_loif;
1056 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1057 m->m_pkthdr.csum_flags |=
1058 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1059 m->m_pkthdr.csum_data = 0xffff;
1060 }
1061 #if defined(SCTP) || defined(SCTP_SUPPORT)
1062 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1063 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1064 #endif
1065 error = netisr_queue(NETISR_IPV6, m);
1066 goto done;
1067 }
1068 /* Or forward to some other address? */
1069 if ((m->m_flags & M_IP6_NEXTHOP) &&
1070 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1071 if (ro != NULL)
1072 dst = (struct sockaddr_in6 *)&ro->ro_dst;
1073 else
1074 dst = &sin6;
1075 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1076 m->m_flags |= M_SKIP_FIREWALL;
1077 m->m_flags &= ~M_IP6_NEXTHOP;
1078 m_tag_delete(m, fwd_tag);
1079 goto again;
1080 }
1081
1082 passout:
1083 if (vlan_pcp > -1)
1084 EVL_APPLY_PRI(m, vlan_pcp);
1085
1086 /* Ensure the packet data is mapped if the interface requires it. */
1087 if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1088 struct mbuf *m1;
1089
1090 error = mb_unmapped_to_ext(m, &m1);
1091 if (error != 0) {
1092 if (error == EINVAL) {
1093 if_printf(ifp, "TLS packet\n");
1094 /* XXXKIB */
1095 } else if (error == ENOMEM) {
1096 error = ENOBUFS;
1097 }
1098 IP6STAT_INC(ip6s_odropped);
1099 return (error);
1100 } else {
1101 m = m1;
1102 }
1103 }
1104
1105 /*
1106 * Send the packet to the outgoing interface.
1107 * If necessary, do IPv6 fragmentation before sending.
1108 *
1109 * 1: normal case (dontfrag == 0)
1110 * 1-a: send as is if tlen <= path mtu
1111 * 1-b: fragment if tlen > path mtu
1112 *
1113 * 2: if user asks us not to fragment (dontfrag == 1)
1114 * 2-a: send as is if tlen <= interface mtu
1115 * 2-b: error if tlen > interface mtu
1116 */
1117 sw_csum = m->m_pkthdr.csum_flags;
1118 if (!hdrsplit) {
1119 tso = ((sw_csum & ifp->if_hwassist &
1120 (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1121 sw_csum &= ~ifp->if_hwassist;
1122 } else
1123 tso = 0;
1124 /*
1125 * If we added extension headers, we will not do TSO and calculate the
1126 * checksums ourselves for now.
1127 * XXX-BZ Need a framework to know when the NIC can handle it, even
1128 * with ext. hdrs.
1129 */
1130 ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1131 /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1132 tlen = m->m_pkthdr.len;
1133
1134 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1135 dontfrag = 1;
1136 else
1137 dontfrag = 0;
1138 if (dontfrag && tlen > in6_ifmtu(ifp) && !tso) { /* Case 2-b. */
1139 /*
1140 * If the DONTFRAG option is specified, we cannot send the
1141 * packet when the data length is larger than the MTU of the
1142 * outgoing interface.
1143 * Notify the error by sending IPV6_PATHMTU ancillary data if
1144 * application wanted to know the MTU value. Also return an
1145 * error code (this is not described in the API spec).
1146 */
1147 if (inp != NULL)
1148 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1149 error = EMSGSIZE;
1150 goto bad;
1151 }
1152
1153 /* Transmit packet without fragmentation. */
1154 if (dontfrag || tlen <= mtu) { /* Cases 1-a and 2-a. */
1155 struct in6_ifaddr *ia6;
1156
1157 ip6 = mtod(m, struct ip6_hdr *);
1158 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1159 if (ia6) {
1160 /* Record statistics for this interface address. */
1161 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1162 counter_u64_add(ia6->ia_ifa.ifa_obytes,
1163 m->m_pkthdr.len);
1164 }
1165 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1166 (flags & IP_NO_SND_TAG_RL) ? false : true);
1167 goto done;
1168 }
1169
1170 /* Try to fragment the packet. Case 1-b. */
1171 if (mtu < IPV6_MMTU) {
1172 /* Path MTU cannot be less than IPV6_MMTU. */
1173 error = EMSGSIZE;
1174 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1175 goto bad;
1176 } else if (ip6->ip6_plen == 0) {
1177 /* We do not support jumbo payload. */
1178 error = EMSGSIZE;
1179 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1180 goto bad;
1181 } else {
1182 u_char nextproto;
1183
1184 /*
1185 * Too large for the destination or interface;
1186 * fragment if possible.
1187 * Must be able to put at least 8 bytes per fragment.
1188 */
1189 if (mtu > IPV6_MAXPACKET)
1190 mtu = IPV6_MAXPACKET;
1191
1192 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1193 if (len < 8) {
1194 error = EMSGSIZE;
1195 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1196 goto bad;
1197 }
1198
1199 /*
1200 * If the interface will not calculate checksums on
1201 * fragmented packets, then do it here.
1202 * XXX-BZ handle the hw offloading case. Need flags.
1203 */
1204 ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1205 optlen);
1206
1207 /*
1208 * Change the next header field of the last header in the
1209 * unfragmentable part.
1210 */
1211 if (exthdrs.ip6e_rthdr) {
1212 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1213 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1214 } else if (exthdrs.ip6e_dest1) {
1215 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1216 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1217 } else if (exthdrs.ip6e_hbh) {
1218 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1219 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1220 } else {
1221 ip6 = mtod(m, struct ip6_hdr *);
1222 nextproto = ip6->ip6_nxt;
1223 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1224 }
1225
1226 /*
1227 * Loop through length of segment after first fragment,
1228 * make new header and copy data of each part and link onto
1229 * chain.
1230 */
1231 m0 = m;
1232 id = htonl(ip6_randomid());
1233 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1234 if (error != 0)
1235 goto sendorfree;
1236
1237 in6_ifstat_inc(ifp, ifs6_out_fragok);
1238 }
1239
1240 /* Remove leading garbage. */
1241 sendorfree:
1242 m = m0->m_nextpkt;
1243 m0->m_nextpkt = 0;
1244 m_freem(m0);
1245 for (; m; m = m0) {
1246 m0 = m->m_nextpkt;
1247 m->m_nextpkt = 0;
1248 if (error == 0) {
1249 /* Record statistics for this interface address. */
1250 if (ia) {
1251 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1252 counter_u64_add(ia->ia_ifa.ifa_obytes,
1253 m->m_pkthdr.len);
1254 }
1255 if (vlan_pcp > -1)
1256 EVL_APPLY_PRI(m, vlan_pcp);
1257 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1258 true);
1259 } else
1260 m_freem(m);
1261 }
1262
1263 if (error == 0)
1264 IP6STAT_INC(ip6s_fragmented);
1265
1266 done:
1267 return (error);
1268
1269 freehdrs:
1270 m_freem(exthdrs.ip6e_hbh); /* m_freem() checks if mbuf is NULL. */
1271 m_freem(exthdrs.ip6e_dest1);
1272 m_freem(exthdrs.ip6e_rthdr);
1273 m_freem(exthdrs.ip6e_dest2);
1274 /* FALLTHROUGH */
1275 bad:
1276 if (m)
1277 m_freem(m);
1278 goto done;
1279 }
1280
1281 static int
ip6_copyexthdr(struct mbuf ** mp,caddr_t hdr,int hlen)1282 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1283 {
1284 struct mbuf *m;
1285
1286 if (hlen > MCLBYTES)
1287 return (ENOBUFS); /* XXX */
1288
1289 if (hlen > MLEN)
1290 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1291 else
1292 m = m_get(M_NOWAIT, MT_DATA);
1293 if (m == NULL)
1294 return (ENOBUFS);
1295 m->m_len = hlen;
1296 if (hdr)
1297 bcopy(hdr, mtod(m, caddr_t), hlen);
1298
1299 *mp = m;
1300 return (0);
1301 }
1302
1303 /*
1304 * Insert fragment header and copy unfragmentable header portions.
1305 */
1306 static int
ip6_insertfraghdr(struct mbuf * m0,struct mbuf * m,int hlen,struct ip6_frag ** frghdrp)1307 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1308 struct ip6_frag **frghdrp)
1309 {
1310 struct mbuf *n, *mlast;
1311
1312 if (hlen > sizeof(struct ip6_hdr)) {
1313 n = m_copym(m0, sizeof(struct ip6_hdr),
1314 hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1315 if (n == NULL)
1316 return (ENOBUFS);
1317 m->m_next = n;
1318 } else
1319 n = m;
1320
1321 /* Search for the last mbuf of unfragmentable part. */
1322 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1323 ;
1324
1325 if (M_WRITABLE(mlast) &&
1326 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1327 /* use the trailing space of the last mbuf for the fragment hdr */
1328 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1329 mlast->m_len);
1330 mlast->m_len += sizeof(struct ip6_frag);
1331 m->m_pkthdr.len += sizeof(struct ip6_frag);
1332 } else {
1333 /* allocate a new mbuf for the fragment header */
1334 struct mbuf *mfrg;
1335
1336 mfrg = m_get(M_NOWAIT, MT_DATA);
1337 if (mfrg == NULL)
1338 return (ENOBUFS);
1339 mfrg->m_len = sizeof(struct ip6_frag);
1340 *frghdrp = mtod(mfrg, struct ip6_frag *);
1341 mlast->m_next = mfrg;
1342 }
1343
1344 return (0);
1345 }
1346
1347 /*
1348 * Calculates IPv6 path mtu for destination @dst.
1349 * Resulting MTU is stored in @mtup.
1350 *
1351 * Returns 0 on success.
1352 */
1353 static int
ip6_getpmtu_ctl(u_int fibnum,const struct in6_addr * dst,u_long * mtup)1354 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1355 {
1356 struct epoch_tracker et;
1357 struct nhop_object *nh;
1358 struct in6_addr kdst;
1359 uint32_t scopeid;
1360 int error;
1361
1362 in6_splitscope(dst, &kdst, &scopeid);
1363
1364 NET_EPOCH_ENTER(et);
1365 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1366 if (nh != NULL) {
1367 ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, 0);
1368 error = 0;
1369 } else
1370 error = EHOSTUNREACH;
1371 NET_EPOCH_EXIT(et);
1372
1373 return (error);
1374 }
1375
1376 /*
1377 * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1378 * and cached data in @ro_pmtu.
1379 * MTU from (successful) route lookup is saved (along with dst)
1380 * inside @ro_pmtu to avoid subsequent route lookups after packet
1381 * filter processing.
1382 *
1383 * Stores mtu into @mtup.
1384 */
1385 static void
ip6_getpmtu(struct route_in6 * ro_pmtu,int do_lookup,struct ifnet * ifp,const struct in6_addr * dst,u_long * mtup,u_int fibnum,u_int proto)1386 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1387 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1388 u_int fibnum, u_int proto)
1389 {
1390 struct nhop_object *nh;
1391 struct in6_addr kdst;
1392 uint32_t scopeid;
1393 struct sockaddr_in6 *sa6_dst, sin6;
1394 u_long mtu;
1395
1396 NET_EPOCH_ASSERT();
1397
1398 mtu = 0;
1399 if (ro_pmtu == NULL || do_lookup) {
1400 /*
1401 * Here ro_pmtu has final destination address, while
1402 * ro might represent immediate destination.
1403 * Use ro_pmtu destination since mtu might differ.
1404 */
1405 if (ro_pmtu != NULL) {
1406 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1407 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1408 ro_pmtu->ro_mtu = 0;
1409 } else
1410 sa6_dst = &sin6;
1411
1412 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1413 bzero(sa6_dst, sizeof(*sa6_dst));
1414 sa6_dst->sin6_family = AF_INET6;
1415 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1416 sa6_dst->sin6_addr = *dst;
1417
1418 in6_splitscope(dst, &kdst, &scopeid);
1419 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1420 if (nh != NULL) {
1421 mtu = nh->nh_mtu;
1422 if (ro_pmtu != NULL)
1423 ro_pmtu->ro_mtu = mtu;
1424 }
1425 } else
1426 mtu = ro_pmtu->ro_mtu;
1427 }
1428
1429 if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1430 mtu = ro_pmtu->ro_nh->nh_mtu;
1431
1432 ip6_calcmtu(ifp, dst, mtu, mtup, proto);
1433 }
1434
1435 /*
1436 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1437 * hostcache data for @dst.
1438 * Stores mtu into @mtup.
1439 */
1440 static void
ip6_calcmtu(struct ifnet * ifp,const struct in6_addr * dst,u_long rt_mtu,u_long * mtup,u_int proto)1441 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1442 u_long *mtup, u_int proto)
1443 {
1444 u_long mtu = 0;
1445
1446 if (rt_mtu > 0) {
1447 /* Skip the hostcache if the protocol handles PMTU changes. */
1448 if (proto != IPPROTO_TCP && proto != IPPROTO_SCTP) {
1449 struct in_conninfo inc = {
1450 .inc_flags = INC_ISIPV6,
1451 .inc6_faddr = *dst,
1452 };
1453
1454 mtu = tcp_hc_getmtu(&inc);
1455 }
1456
1457 if (mtu)
1458 mtu = min(mtu, rt_mtu);
1459 else
1460 mtu = rt_mtu;
1461 }
1462
1463 if (mtu == 0)
1464 mtu = in6_ifmtu(ifp);
1465
1466 *mtup = mtu;
1467 }
1468
1469 /*
1470 * IP6 socket option processing.
1471 */
1472 int
ip6_ctloutput(struct socket * so,struct sockopt * sopt)1473 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1474 {
1475 int optdatalen, uproto;
1476 void *optdata;
1477 struct inpcb *inp = sotoinpcb(so);
1478 int error, optval;
1479 int level, op, optname;
1480 int optlen;
1481 struct thread *td;
1482 #ifdef RSS
1483 uint32_t rss_bucket;
1484 int retval;
1485 #endif
1486
1487 /*
1488 * Don't use more than a quarter of mbuf clusters. N.B.:
1489 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1490 * on LP64 architectures, so cast to u_long to avoid undefined
1491 * behavior. ILP32 architectures cannot have nmbclusters
1492 * large enough to overflow for other reasons.
1493 */
1494 #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4)
1495
1496 level = sopt->sopt_level;
1497 op = sopt->sopt_dir;
1498 optname = sopt->sopt_name;
1499 optlen = sopt->sopt_valsize;
1500 td = sopt->sopt_td;
1501 error = 0;
1502 optval = 0;
1503 uproto = (int)so->so_proto->pr_protocol;
1504
1505 if (level != IPPROTO_IPV6) {
1506 error = EINVAL;
1507
1508 if (sopt->sopt_level == SOL_SOCKET &&
1509 sopt->sopt_dir == SOPT_SET) {
1510 switch (sopt->sopt_name) {
1511 case SO_SETFIB:
1512 error = sooptcopyin(sopt, &optval,
1513 sizeof(optval), sizeof(optval));
1514 if (error != 0)
1515 break;
1516
1517 INP_WLOCK(inp);
1518 if ((inp->inp_flags & INP_BOUNDFIB) != 0 &&
1519 optval != so->so_fibnum) {
1520 INP_WUNLOCK(inp);
1521 error = EISCONN;
1522 break;
1523 }
1524 error = sosetfib(inp->inp_socket, optval);
1525 if (error == 0)
1526 inp->inp_inc.inc_fibnum = optval;
1527 INP_WUNLOCK(inp);
1528 break;
1529 case SO_MAX_PACING_RATE:
1530 #ifdef RATELIMIT
1531 INP_WLOCK(inp);
1532 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1533 INP_WUNLOCK(inp);
1534 error = 0;
1535 #else
1536 error = EOPNOTSUPP;
1537 #endif
1538 break;
1539 default:
1540 break;
1541 }
1542 }
1543 } else { /* level == IPPROTO_IPV6 */
1544 switch (op) {
1545 case SOPT_SET:
1546 switch (optname) {
1547 case IPV6_2292PKTOPTIONS:
1548 #ifdef IPV6_PKTOPTIONS
1549 case IPV6_PKTOPTIONS:
1550 #endif
1551 {
1552 struct mbuf *m;
1553
1554 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1555 printf("ip6_ctloutput: mbuf limit hit\n");
1556 error = ENOBUFS;
1557 break;
1558 }
1559
1560 error = soopt_getm(sopt, &m); /* XXX */
1561 if (error != 0)
1562 break;
1563 error = soopt_mcopyin(sopt, m); /* XXX */
1564 if (error != 0)
1565 break;
1566 INP_WLOCK(inp);
1567 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1568 so, sopt);
1569 INP_WUNLOCK(inp);
1570 m_freem(m); /* XXX */
1571 break;
1572 }
1573
1574 /*
1575 * Use of some Hop-by-Hop options or some
1576 * Destination options, might require special
1577 * privilege. That is, normal applications
1578 * (without special privilege) might be forbidden
1579 * from setting certain options in outgoing packets,
1580 * and might never see certain options in received
1581 * packets. [RFC 2292 Section 6]
1582 * KAME specific note:
1583 * KAME prevents non-privileged users from sending or
1584 * receiving ANY hbh/dst options in order to avoid
1585 * overhead of parsing options in the kernel.
1586 */
1587 case IPV6_RECVHOPOPTS:
1588 case IPV6_RECVDSTOPTS:
1589 case IPV6_RECVRTHDRDSTOPTS:
1590 if (td != NULL) {
1591 error = priv_check(td,
1592 PRIV_NETINET_SETHDROPTS);
1593 if (error)
1594 break;
1595 }
1596 /* FALLTHROUGH */
1597 case IPV6_UNICAST_HOPS:
1598 case IPV6_HOPLIMIT:
1599
1600 case IPV6_RECVPKTINFO:
1601 case IPV6_RECVHOPLIMIT:
1602 case IPV6_RECVRTHDR:
1603 case IPV6_RECVPATHMTU:
1604 case IPV6_RECVTCLASS:
1605 case IPV6_RECVFLOWID:
1606 #ifdef RSS
1607 case IPV6_RECVRSSBUCKETID:
1608 #endif
1609 case IPV6_V6ONLY:
1610 case IPV6_AUTOFLOWLABEL:
1611 case IPV6_ORIGDSTADDR:
1612 case IPV6_BINDANY:
1613 case IPV6_VLAN_PCP:
1614 if (optname == IPV6_BINDANY && td != NULL) {
1615 error = priv_check(td,
1616 PRIV_NETINET_BINDANY);
1617 if (error)
1618 break;
1619 }
1620
1621 if (optlen != sizeof(int)) {
1622 error = EINVAL;
1623 break;
1624 }
1625 error = sooptcopyin(sopt, &optval,
1626 sizeof optval, sizeof optval);
1627 if (error)
1628 break;
1629 switch (optname) {
1630 case IPV6_UNICAST_HOPS:
1631 if (optval < -1 || optval >= 256)
1632 error = EINVAL;
1633 else {
1634 /* -1 = kernel default */
1635 inp->in6p_hops = optval;
1636 if ((inp->inp_vflag &
1637 INP_IPV4) != 0)
1638 inp->inp_ip_ttl = optval;
1639 }
1640 break;
1641 #define OPTSET(bit) \
1642 do { \
1643 INP_WLOCK(inp); \
1644 if (optval) \
1645 inp->inp_flags |= (bit); \
1646 else \
1647 inp->inp_flags &= ~(bit); \
1648 INP_WUNLOCK(inp); \
1649 } while (/*CONSTCOND*/ 0)
1650 #define OPTSET2292(bit) \
1651 do { \
1652 INP_WLOCK(inp); \
1653 inp->inp_flags |= IN6P_RFC2292; \
1654 if (optval) \
1655 inp->inp_flags |= (bit); \
1656 else \
1657 inp->inp_flags &= ~(bit); \
1658 INP_WUNLOCK(inp); \
1659 } while (/*CONSTCOND*/ 0)
1660 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1661
1662 #define OPTSET2_N(bit, val) do { \
1663 if (val) \
1664 inp->inp_flags2 |= bit; \
1665 else \
1666 inp->inp_flags2 &= ~bit; \
1667 } while (0)
1668 #define OPTSET2(bit, val) do { \
1669 INP_WLOCK(inp); \
1670 OPTSET2_N(bit, val); \
1671 INP_WUNLOCK(inp); \
1672 } while (0)
1673 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1674 #define OPTSET2292_EXCLUSIVE(bit) \
1675 do { \
1676 INP_WLOCK(inp); \
1677 if (OPTBIT(IN6P_RFC2292)) { \
1678 error = EINVAL; \
1679 } else { \
1680 if (optval) \
1681 inp->inp_flags |= (bit); \
1682 else \
1683 inp->inp_flags &= ~(bit); \
1684 } \
1685 INP_WUNLOCK(inp); \
1686 } while (/*CONSTCOND*/ 0)
1687
1688 case IPV6_RECVPKTINFO:
1689 OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1690 break;
1691
1692 case IPV6_HOPLIMIT:
1693 {
1694 struct ip6_pktopts **optp;
1695
1696 /* cannot mix with RFC2292 */
1697 if (OPTBIT(IN6P_RFC2292)) {
1698 error = EINVAL;
1699 break;
1700 }
1701 INP_WLOCK(inp);
1702 if (inp->inp_flags & INP_DROPPED) {
1703 INP_WUNLOCK(inp);
1704 return (ECONNRESET);
1705 }
1706 optp = &inp->in6p_outputopts;
1707 error = ip6_pcbopt(IPV6_HOPLIMIT,
1708 (u_char *)&optval, sizeof(optval),
1709 optp, (td != NULL) ? td->td_ucred :
1710 NULL, uproto);
1711 INP_WUNLOCK(inp);
1712 break;
1713 }
1714
1715 case IPV6_RECVHOPLIMIT:
1716 OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1717 break;
1718
1719 case IPV6_RECVHOPOPTS:
1720 OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1721 break;
1722
1723 case IPV6_RECVDSTOPTS:
1724 OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1725 break;
1726
1727 case IPV6_RECVRTHDRDSTOPTS:
1728 OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1729 break;
1730
1731 case IPV6_RECVRTHDR:
1732 OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1733 break;
1734
1735 case IPV6_RECVPATHMTU:
1736 /*
1737 * We ignore this option for TCP
1738 * sockets.
1739 * (RFC3542 leaves this case
1740 * unspecified.)
1741 */
1742 if (uproto != IPPROTO_TCP)
1743 OPTSET(IN6P_MTU);
1744 break;
1745
1746 case IPV6_RECVFLOWID:
1747 OPTSET2(INP_RECVFLOWID, optval);
1748 break;
1749
1750 #ifdef RSS
1751 case IPV6_RECVRSSBUCKETID:
1752 OPTSET2(INP_RECVRSSBUCKETID, optval);
1753 break;
1754 #endif
1755
1756 case IPV6_V6ONLY:
1757 INP_WLOCK(inp);
1758 if (inp->inp_lport ||
1759 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1760 /*
1761 * The socket is already bound.
1762 */
1763 INP_WUNLOCK(inp);
1764 error = EINVAL;
1765 break;
1766 }
1767 if (optval) {
1768 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1769 inp->inp_vflag &= ~INP_IPV4;
1770 } else {
1771 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1772 inp->inp_vflag |= INP_IPV4;
1773 }
1774 INP_WUNLOCK(inp);
1775 break;
1776 case IPV6_RECVTCLASS:
1777 /* cannot mix with RFC2292 XXX */
1778 OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1779 break;
1780 case IPV6_AUTOFLOWLABEL:
1781 OPTSET(IN6P_AUTOFLOWLABEL);
1782 break;
1783
1784 case IPV6_ORIGDSTADDR:
1785 OPTSET2(INP_ORIGDSTADDR, optval);
1786 break;
1787 case IPV6_BINDANY:
1788 OPTSET(INP_BINDANY);
1789 break;
1790 case IPV6_VLAN_PCP:
1791 if ((optval >= -1) && (optval <=
1792 (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1793 if (optval == -1) {
1794 INP_WLOCK(inp);
1795 inp->inp_flags2 &=
1796 ~(INP_2PCP_SET |
1797 INP_2PCP_MASK);
1798 INP_WUNLOCK(inp);
1799 } else {
1800 INP_WLOCK(inp);
1801 inp->inp_flags2 |=
1802 INP_2PCP_SET;
1803 inp->inp_flags2 &=
1804 ~INP_2PCP_MASK;
1805 inp->inp_flags2 |=
1806 optval <<
1807 INP_2PCP_SHIFT;
1808 INP_WUNLOCK(inp);
1809 }
1810 } else
1811 error = EINVAL;
1812 break;
1813 }
1814 break;
1815
1816 case IPV6_TCLASS:
1817 case IPV6_DONTFRAG:
1818 case IPV6_USE_MIN_MTU:
1819 case IPV6_PREFER_TEMPADDR:
1820 if (optlen != sizeof(optval)) {
1821 error = EINVAL;
1822 break;
1823 }
1824 error = sooptcopyin(sopt, &optval,
1825 sizeof optval, sizeof optval);
1826 if (error)
1827 break;
1828 {
1829 struct ip6_pktopts **optp;
1830 INP_WLOCK(inp);
1831 if (inp->inp_flags & INP_DROPPED) {
1832 INP_WUNLOCK(inp);
1833 return (ECONNRESET);
1834 }
1835 optp = &inp->in6p_outputopts;
1836 error = ip6_pcbopt(optname,
1837 (u_char *)&optval, sizeof(optval),
1838 optp, (td != NULL) ? td->td_ucred :
1839 NULL, uproto);
1840 INP_WUNLOCK(inp);
1841 break;
1842 }
1843
1844 case IPV6_2292PKTINFO:
1845 case IPV6_2292HOPLIMIT:
1846 case IPV6_2292HOPOPTS:
1847 case IPV6_2292DSTOPTS:
1848 case IPV6_2292RTHDR:
1849 /* RFC 2292 */
1850 if (optlen != sizeof(int)) {
1851 error = EINVAL;
1852 break;
1853 }
1854 error = sooptcopyin(sopt, &optval,
1855 sizeof optval, sizeof optval);
1856 if (error)
1857 break;
1858 switch (optname) {
1859 case IPV6_2292PKTINFO:
1860 OPTSET2292(IN6P_PKTINFO);
1861 break;
1862 case IPV6_2292HOPLIMIT:
1863 OPTSET2292(IN6P_HOPLIMIT);
1864 break;
1865 case IPV6_2292HOPOPTS:
1866 /*
1867 * Check super-user privilege.
1868 * See comments for IPV6_RECVHOPOPTS.
1869 */
1870 if (td != NULL) {
1871 error = priv_check(td,
1872 PRIV_NETINET_SETHDROPTS);
1873 if (error)
1874 return (error);
1875 }
1876 OPTSET2292(IN6P_HOPOPTS);
1877 break;
1878 case IPV6_2292DSTOPTS:
1879 if (td != NULL) {
1880 error = priv_check(td,
1881 PRIV_NETINET_SETHDROPTS);
1882 if (error)
1883 return (error);
1884 }
1885 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1886 break;
1887 case IPV6_2292RTHDR:
1888 OPTSET2292(IN6P_RTHDR);
1889 break;
1890 }
1891 break;
1892 case IPV6_PKTINFO:
1893 case IPV6_HOPOPTS:
1894 case IPV6_RTHDR:
1895 case IPV6_DSTOPTS:
1896 case IPV6_RTHDRDSTOPTS:
1897 case IPV6_NEXTHOP:
1898 {
1899 /* new advanced API (RFC3542) */
1900 u_char *optbuf;
1901 u_char optbuf_storage[MCLBYTES];
1902 int optlen;
1903 struct ip6_pktopts **optp;
1904
1905 /* cannot mix with RFC2292 */
1906 if (OPTBIT(IN6P_RFC2292)) {
1907 error = EINVAL;
1908 break;
1909 }
1910
1911 /*
1912 * We only ensure valsize is not too large
1913 * here. Further validation will be done
1914 * later.
1915 */
1916 error = sooptcopyin(sopt, optbuf_storage,
1917 sizeof(optbuf_storage), 0);
1918 if (error)
1919 break;
1920 optlen = sopt->sopt_valsize;
1921 optbuf = optbuf_storage;
1922 INP_WLOCK(inp);
1923 if (inp->inp_flags & INP_DROPPED) {
1924 INP_WUNLOCK(inp);
1925 return (ECONNRESET);
1926 }
1927 optp = &inp->in6p_outputopts;
1928 error = ip6_pcbopt(optname, optbuf, optlen,
1929 optp, (td != NULL) ? td->td_ucred : NULL,
1930 uproto);
1931 INP_WUNLOCK(inp);
1932 break;
1933 }
1934 #undef OPTSET
1935
1936 case IPV6_MULTICAST_IF:
1937 case IPV6_MULTICAST_HOPS:
1938 case IPV6_MULTICAST_LOOP:
1939 case IPV6_JOIN_GROUP:
1940 case IPV6_LEAVE_GROUP:
1941 case IPV6_MSFILTER:
1942 case MCAST_BLOCK_SOURCE:
1943 case MCAST_UNBLOCK_SOURCE:
1944 case MCAST_JOIN_GROUP:
1945 case MCAST_LEAVE_GROUP:
1946 case MCAST_JOIN_SOURCE_GROUP:
1947 case MCAST_LEAVE_SOURCE_GROUP:
1948 error = ip6_setmoptions(inp, sopt);
1949 break;
1950
1951 case IPV6_PORTRANGE:
1952 error = sooptcopyin(sopt, &optval,
1953 sizeof optval, sizeof optval);
1954 if (error)
1955 break;
1956
1957 INP_WLOCK(inp);
1958 switch (optval) {
1959 case IPV6_PORTRANGE_DEFAULT:
1960 inp->inp_flags &= ~(INP_LOWPORT);
1961 inp->inp_flags &= ~(INP_HIGHPORT);
1962 break;
1963
1964 case IPV6_PORTRANGE_HIGH:
1965 inp->inp_flags &= ~(INP_LOWPORT);
1966 inp->inp_flags |= INP_HIGHPORT;
1967 break;
1968
1969 case IPV6_PORTRANGE_LOW:
1970 inp->inp_flags &= ~(INP_HIGHPORT);
1971 inp->inp_flags |= INP_LOWPORT;
1972 break;
1973
1974 default:
1975 error = EINVAL;
1976 break;
1977 }
1978 INP_WUNLOCK(inp);
1979 break;
1980
1981 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1982 case IPV6_IPSEC_POLICY:
1983 if (IPSEC_ENABLED(ipv6)) {
1984 error = IPSEC_PCBCTL(ipv6, inp, sopt);
1985 break;
1986 }
1987 /* FALLTHROUGH */
1988 #endif /* IPSEC */
1989
1990 default:
1991 error = ENOPROTOOPT;
1992 break;
1993 }
1994 break;
1995
1996 case SOPT_GET:
1997 switch (optname) {
1998 case IPV6_2292PKTOPTIONS:
1999 #ifdef IPV6_PKTOPTIONS
2000 case IPV6_PKTOPTIONS:
2001 #endif
2002 /*
2003 * RFC3542 (effectively) deprecated the
2004 * semantics of the 2292-style pktoptions.
2005 * Since it was not reliable in nature (i.e.,
2006 * applications had to expect the lack of some
2007 * information after all), it would make sense
2008 * to simplify this part by always returning
2009 * empty data.
2010 */
2011 sopt->sopt_valsize = 0;
2012 break;
2013
2014 case IPV6_RECVHOPOPTS:
2015 case IPV6_RECVDSTOPTS:
2016 case IPV6_RECVRTHDRDSTOPTS:
2017 case IPV6_UNICAST_HOPS:
2018 case IPV6_RECVPKTINFO:
2019 case IPV6_RECVHOPLIMIT:
2020 case IPV6_RECVRTHDR:
2021 case IPV6_RECVPATHMTU:
2022
2023 case IPV6_V6ONLY:
2024 case IPV6_PORTRANGE:
2025 case IPV6_RECVTCLASS:
2026 case IPV6_AUTOFLOWLABEL:
2027 case IPV6_BINDANY:
2028 case IPV6_FLOWID:
2029 case IPV6_FLOWTYPE:
2030 case IPV6_RECVFLOWID:
2031 #ifdef RSS
2032 case IPV6_RSSBUCKETID:
2033 case IPV6_RECVRSSBUCKETID:
2034 #endif
2035 case IPV6_VLAN_PCP:
2036 switch (optname) {
2037 case IPV6_RECVHOPOPTS:
2038 optval = OPTBIT(IN6P_HOPOPTS);
2039 break;
2040
2041 case IPV6_RECVDSTOPTS:
2042 optval = OPTBIT(IN6P_DSTOPTS);
2043 break;
2044
2045 case IPV6_RECVRTHDRDSTOPTS:
2046 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2047 break;
2048
2049 case IPV6_UNICAST_HOPS:
2050 optval = inp->in6p_hops;
2051 break;
2052
2053 case IPV6_RECVPKTINFO:
2054 optval = OPTBIT(IN6P_PKTINFO);
2055 break;
2056
2057 case IPV6_RECVHOPLIMIT:
2058 optval = OPTBIT(IN6P_HOPLIMIT);
2059 break;
2060
2061 case IPV6_RECVRTHDR:
2062 optval = OPTBIT(IN6P_RTHDR);
2063 break;
2064
2065 case IPV6_RECVPATHMTU:
2066 optval = OPTBIT(IN6P_MTU);
2067 break;
2068
2069 case IPV6_V6ONLY:
2070 optval = OPTBIT(IN6P_IPV6_V6ONLY);
2071 break;
2072
2073 case IPV6_PORTRANGE:
2074 {
2075 int flags;
2076 flags = inp->inp_flags;
2077 if (flags & INP_HIGHPORT)
2078 optval = IPV6_PORTRANGE_HIGH;
2079 else if (flags & INP_LOWPORT)
2080 optval = IPV6_PORTRANGE_LOW;
2081 else
2082 optval = 0;
2083 break;
2084 }
2085 case IPV6_RECVTCLASS:
2086 optval = OPTBIT(IN6P_TCLASS);
2087 break;
2088
2089 case IPV6_AUTOFLOWLABEL:
2090 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2091 break;
2092
2093 case IPV6_ORIGDSTADDR:
2094 optval = OPTBIT2(INP_ORIGDSTADDR);
2095 break;
2096
2097 case IPV6_BINDANY:
2098 optval = OPTBIT(INP_BINDANY);
2099 break;
2100
2101 case IPV6_FLOWID:
2102 optval = inp->inp_flowid;
2103 break;
2104
2105 case IPV6_FLOWTYPE:
2106 optval = inp->inp_flowtype;
2107 break;
2108
2109 case IPV6_RECVFLOWID:
2110 optval = OPTBIT2(INP_RECVFLOWID);
2111 break;
2112 #ifdef RSS
2113 case IPV6_RSSBUCKETID:
2114 retval =
2115 rss_hash2bucket(inp->inp_flowid,
2116 inp->inp_flowtype,
2117 &rss_bucket);
2118 if (retval == 0)
2119 optval = rss_bucket;
2120 else
2121 error = EINVAL;
2122 break;
2123
2124 case IPV6_RECVRSSBUCKETID:
2125 optval = OPTBIT2(INP_RECVRSSBUCKETID);
2126 break;
2127 #endif
2128
2129
2130 case IPV6_VLAN_PCP:
2131 if (OPTBIT2(INP_2PCP_SET)) {
2132 optval = (inp->inp_flags2 &
2133 INP_2PCP_MASK) >>
2134 INP_2PCP_SHIFT;
2135 } else {
2136 optval = -1;
2137 }
2138 break;
2139 }
2140
2141 if (error)
2142 break;
2143 error = sooptcopyout(sopt, &optval,
2144 sizeof optval);
2145 break;
2146
2147 case IPV6_PATHMTU:
2148 {
2149 u_long pmtu = 0;
2150 struct ip6_mtuinfo mtuinfo;
2151 struct in6_addr addr;
2152
2153 if (!(so->so_state & SS_ISCONNECTED))
2154 return (ENOTCONN);
2155 /*
2156 * XXX: we dot not consider the case of source
2157 * routing, or optional information to specify
2158 * the outgoing interface.
2159 * Copy faddr out of inp to avoid holding lock
2160 * on inp during route lookup.
2161 */
2162 INP_RLOCK(inp);
2163 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2164 INP_RUNLOCK(inp);
2165 error = ip6_getpmtu_ctl(so->so_fibnum,
2166 &addr, &pmtu);
2167 if (error)
2168 break;
2169 if (pmtu > IPV6_MAXPACKET)
2170 pmtu = IPV6_MAXPACKET;
2171
2172 bzero(&mtuinfo, sizeof(mtuinfo));
2173 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2174 optdata = (void *)&mtuinfo;
2175 optdatalen = sizeof(mtuinfo);
2176 error = sooptcopyout(sopt, optdata,
2177 optdatalen);
2178 break;
2179 }
2180
2181 case IPV6_2292PKTINFO:
2182 case IPV6_2292HOPLIMIT:
2183 case IPV6_2292HOPOPTS:
2184 case IPV6_2292RTHDR:
2185 case IPV6_2292DSTOPTS:
2186 switch (optname) {
2187 case IPV6_2292PKTINFO:
2188 optval = OPTBIT(IN6P_PKTINFO);
2189 break;
2190 case IPV6_2292HOPLIMIT:
2191 optval = OPTBIT(IN6P_HOPLIMIT);
2192 break;
2193 case IPV6_2292HOPOPTS:
2194 optval = OPTBIT(IN6P_HOPOPTS);
2195 break;
2196 case IPV6_2292RTHDR:
2197 optval = OPTBIT(IN6P_RTHDR);
2198 break;
2199 case IPV6_2292DSTOPTS:
2200 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2201 break;
2202 }
2203 error = sooptcopyout(sopt, &optval,
2204 sizeof optval);
2205 break;
2206 case IPV6_PKTINFO:
2207 case IPV6_HOPOPTS:
2208 case IPV6_RTHDR:
2209 case IPV6_DSTOPTS:
2210 case IPV6_RTHDRDSTOPTS:
2211 case IPV6_NEXTHOP:
2212 case IPV6_TCLASS:
2213 case IPV6_DONTFRAG:
2214 case IPV6_USE_MIN_MTU:
2215 case IPV6_PREFER_TEMPADDR:
2216 error = ip6_getpcbopt(inp, optname, sopt);
2217 break;
2218
2219 case IPV6_MULTICAST_IF:
2220 case IPV6_MULTICAST_HOPS:
2221 case IPV6_MULTICAST_LOOP:
2222 case IPV6_MSFILTER:
2223 error = ip6_getmoptions(inp, sopt);
2224 break;
2225
2226 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2227 case IPV6_IPSEC_POLICY:
2228 if (IPSEC_ENABLED(ipv6)) {
2229 error = IPSEC_PCBCTL(ipv6, inp, sopt);
2230 break;
2231 }
2232 /* FALLTHROUGH */
2233 #endif /* IPSEC */
2234 default:
2235 error = ENOPROTOOPT;
2236 break;
2237 }
2238 break;
2239 }
2240 }
2241 return (error);
2242 }
2243
2244 int
ip6_raw_ctloutput(struct socket * so,struct sockopt * sopt)2245 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2246 {
2247 int error = 0, optval, optlen;
2248 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2249 struct inpcb *inp = sotoinpcb(so);
2250 int level, op, optname;
2251
2252 level = sopt->sopt_level;
2253 op = sopt->sopt_dir;
2254 optname = sopt->sopt_name;
2255 optlen = sopt->sopt_valsize;
2256
2257 if (level != IPPROTO_IPV6) {
2258 return (EINVAL);
2259 }
2260
2261 switch (optname) {
2262 case IPV6_CHECKSUM:
2263 /*
2264 * For ICMPv6 sockets, no modification allowed for checksum
2265 * offset, permit "no change" values to help existing apps.
2266 *
2267 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2268 * for an ICMPv6 socket will fail."
2269 * The current behavior does not meet RFC3542.
2270 */
2271 switch (op) {
2272 case SOPT_SET:
2273 if (optlen != sizeof(int)) {
2274 error = EINVAL;
2275 break;
2276 }
2277 error = sooptcopyin(sopt, &optval, sizeof(optval),
2278 sizeof(optval));
2279 if (error)
2280 break;
2281 if (optval < -1 || (optval % 2) != 0) {
2282 /*
2283 * The API assumes non-negative even offset
2284 * values or -1 as a special value.
2285 */
2286 error = EINVAL;
2287 } else if (inp->inp_ip_p == IPPROTO_ICMPV6) {
2288 if (optval != icmp6off)
2289 error = EINVAL;
2290 } else
2291 inp->in6p_cksum = optval;
2292 break;
2293
2294 case SOPT_GET:
2295 if (inp->inp_ip_p == IPPROTO_ICMPV6)
2296 optval = icmp6off;
2297 else
2298 optval = inp->in6p_cksum;
2299
2300 error = sooptcopyout(sopt, &optval, sizeof(optval));
2301 break;
2302
2303 default:
2304 error = EINVAL;
2305 break;
2306 }
2307 break;
2308
2309 default:
2310 error = ENOPROTOOPT;
2311 break;
2312 }
2313
2314 return (error);
2315 }
2316
2317 /*
2318 * Set up IP6 options in pcb for insertion in output packets or
2319 * specifying behavior of outgoing packets.
2320 */
2321 static int
ip6_pcbopts(struct ip6_pktopts ** pktopt,struct mbuf * m,struct socket * so,struct sockopt * sopt)2322 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2323 struct socket *so, struct sockopt *sopt)
2324 {
2325 struct ip6_pktopts *opt = *pktopt;
2326 int error = 0;
2327 struct thread *td = sopt->sopt_td;
2328 struct epoch_tracker et;
2329
2330 /* turn off any old options. */
2331 if (opt) {
2332 #ifdef DIAGNOSTIC
2333 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2334 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2335 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2336 printf("ip6_pcbopts: all specified options are cleared.\n");
2337 #endif
2338 ip6_clearpktopts(opt, -1);
2339 } else {
2340 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2341 if (opt == NULL)
2342 return (ENOMEM);
2343 }
2344 *pktopt = NULL;
2345
2346 if (!m || m->m_len == 0) {
2347 /*
2348 * Only turning off any previous options, regardless of
2349 * whether the opt is just created or given.
2350 */
2351 free(opt, M_IP6OPT);
2352 return (0);
2353 }
2354
2355 /* set options specified by user. */
2356 NET_EPOCH_ENTER(et);
2357 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2358 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2359 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2360 free(opt, M_IP6OPT);
2361 NET_EPOCH_EXIT(et);
2362 return (error);
2363 }
2364 NET_EPOCH_EXIT(et);
2365 *pktopt = opt;
2366 return (0);
2367 }
2368
2369 /*
2370 * initialize ip6_pktopts. beware that there are non-zero default values in
2371 * the struct.
2372 */
2373 void
ip6_initpktopts(struct ip6_pktopts * opt)2374 ip6_initpktopts(struct ip6_pktopts *opt)
2375 {
2376
2377 bzero(opt, sizeof(*opt));
2378 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2379 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2380 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2381 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2382 }
2383
2384 static int
ip6_pcbopt(int optname,u_char * buf,int len,struct ip6_pktopts ** pktopt,struct ucred * cred,int uproto)2385 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2386 struct ucred *cred, int uproto)
2387 {
2388 struct epoch_tracker et;
2389 struct ip6_pktopts *opt;
2390 int ret;
2391
2392 if (*pktopt == NULL) {
2393 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2394 M_NOWAIT);
2395 if (*pktopt == NULL)
2396 return (ENOBUFS);
2397 ip6_initpktopts(*pktopt);
2398 }
2399 opt = *pktopt;
2400
2401 NET_EPOCH_ENTER(et);
2402 ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2403 NET_EPOCH_EXIT(et);
2404
2405 return (ret);
2406 }
2407
2408 #define GET_PKTOPT_VAR(field, lenexpr) do { \
2409 if (pktopt && pktopt->field) { \
2410 INP_RUNLOCK(inp); \
2411 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \
2412 malloc_optdata = true; \
2413 INP_RLOCK(inp); \
2414 if (inp->inp_flags & INP_DROPPED) { \
2415 INP_RUNLOCK(inp); \
2416 free(optdata, M_TEMP); \
2417 return (ECONNRESET); \
2418 } \
2419 pktopt = inp->in6p_outputopts; \
2420 if (pktopt && pktopt->field) { \
2421 optdatalen = min(lenexpr, sopt->sopt_valsize); \
2422 bcopy(pktopt->field, optdata, optdatalen); \
2423 } else { \
2424 free(optdata, M_TEMP); \
2425 optdata = NULL; \
2426 malloc_optdata = false; \
2427 } \
2428 } \
2429 } while(0)
2430
2431 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field, \
2432 (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2433
2434 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field, \
2435 pktopt->field->sa_len)
2436
2437 static int
ip6_getpcbopt(struct inpcb * inp,int optname,struct sockopt * sopt)2438 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2439 {
2440 void *optdata = NULL;
2441 bool malloc_optdata = false;
2442 int optdatalen = 0;
2443 int error = 0;
2444 struct in6_pktinfo null_pktinfo;
2445 int deftclass = 0, on;
2446 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2447 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2448 struct ip6_pktopts *pktopt;
2449
2450 INP_RLOCK(inp);
2451 pktopt = inp->in6p_outputopts;
2452
2453 switch (optname) {
2454 case IPV6_PKTINFO:
2455 optdata = (void *)&null_pktinfo;
2456 if (pktopt && pktopt->ip6po_pktinfo) {
2457 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2458 sizeof(null_pktinfo));
2459 in6_clearscope(&null_pktinfo.ipi6_addr);
2460 } else {
2461 /* XXX: we don't have to do this every time... */
2462 bzero(&null_pktinfo, sizeof(null_pktinfo));
2463 }
2464 optdatalen = sizeof(struct in6_pktinfo);
2465 break;
2466 case IPV6_TCLASS:
2467 if (pktopt && pktopt->ip6po_tclass >= 0)
2468 deftclass = pktopt->ip6po_tclass;
2469 optdata = (void *)&deftclass;
2470 optdatalen = sizeof(int);
2471 break;
2472 case IPV6_HOPOPTS:
2473 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2474 break;
2475 case IPV6_RTHDR:
2476 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2477 break;
2478 case IPV6_RTHDRDSTOPTS:
2479 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2480 break;
2481 case IPV6_DSTOPTS:
2482 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2483 break;
2484 case IPV6_NEXTHOP:
2485 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2486 break;
2487 case IPV6_USE_MIN_MTU:
2488 if (pktopt)
2489 defminmtu = pktopt->ip6po_minmtu;
2490 optdata = (void *)&defminmtu;
2491 optdatalen = sizeof(int);
2492 break;
2493 case IPV6_DONTFRAG:
2494 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2495 on = 1;
2496 else
2497 on = 0;
2498 optdata = (void *)&on;
2499 optdatalen = sizeof(on);
2500 break;
2501 case IPV6_PREFER_TEMPADDR:
2502 if (pktopt)
2503 defpreftemp = pktopt->ip6po_prefer_tempaddr;
2504 optdata = (void *)&defpreftemp;
2505 optdatalen = sizeof(int);
2506 break;
2507 default: /* should not happen */
2508 #ifdef DIAGNOSTIC
2509 panic("ip6_getpcbopt: unexpected option\n");
2510 #endif
2511 INP_RUNLOCK(inp);
2512 return (ENOPROTOOPT);
2513 }
2514 INP_RUNLOCK(inp);
2515
2516 error = sooptcopyout(sopt, optdata, optdatalen);
2517 if (malloc_optdata)
2518 free(optdata, M_TEMP);
2519
2520 return (error);
2521 }
2522
2523 void
ip6_clearpktopts(struct ip6_pktopts * pktopt,int optname)2524 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2525 {
2526 if (pktopt == NULL)
2527 return;
2528
2529 if (optname == -1 || optname == IPV6_PKTINFO) {
2530 if (pktopt->ip6po_pktinfo)
2531 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2532 pktopt->ip6po_pktinfo = NULL;
2533 }
2534 if (optname == -1 || optname == IPV6_HOPLIMIT) {
2535 pktopt->ip6po_hlim = -1;
2536 pktopt->ip6po_valid &= ~IP6PO_VALID_HLIM;
2537 }
2538 if (optname == -1 || optname == IPV6_TCLASS) {
2539 pktopt->ip6po_tclass = -1;
2540 pktopt->ip6po_valid &= ~IP6PO_VALID_TC;
2541 }
2542 if (optname == -1 || optname == IPV6_NEXTHOP) {
2543 if (pktopt->ip6po_nextroute.ro_nh) {
2544 NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2545 pktopt->ip6po_nextroute.ro_nh = NULL;
2546 }
2547 if (pktopt->ip6po_nexthop)
2548 free(pktopt->ip6po_nexthop, M_IP6OPT);
2549 pktopt->ip6po_nexthop = NULL;
2550 pktopt->ip6po_valid &= ~IP6PO_VALID_NHINFO;
2551 }
2552 if (optname == -1 || optname == IPV6_HOPOPTS) {
2553 if (pktopt->ip6po_hbh)
2554 free(pktopt->ip6po_hbh, M_IP6OPT);
2555 pktopt->ip6po_hbh = NULL;
2556 pktopt->ip6po_valid &= ~IP6PO_VALID_HBH;
2557 }
2558 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2559 if (pktopt->ip6po_dest1)
2560 free(pktopt->ip6po_dest1, M_IP6OPT);
2561 pktopt->ip6po_dest1 = NULL;
2562 pktopt->ip6po_valid &= ~IP6PO_VALID_DEST1;
2563 }
2564 if (optname == -1 || optname == IPV6_RTHDR) {
2565 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2566 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2567 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2568 if (pktopt->ip6po_route.ro_nh) {
2569 NH_FREE(pktopt->ip6po_route.ro_nh);
2570 pktopt->ip6po_route.ro_nh = NULL;
2571 }
2572 pktopt->ip6po_valid &= ~IP6PO_VALID_RHINFO;
2573 }
2574 if (optname == -1 || optname == IPV6_DSTOPTS) {
2575 if (pktopt->ip6po_dest2)
2576 free(pktopt->ip6po_dest2, M_IP6OPT);
2577 pktopt->ip6po_dest2 = NULL;
2578 pktopt->ip6po_valid &= ~IP6PO_VALID_DEST2;
2579 }
2580 }
2581
2582 #define PKTOPT_EXTHDRCPY(type) \
2583 do {\
2584 if (src->type) {\
2585 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2586 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2587 if (dst->type == NULL)\
2588 goto bad;\
2589 bcopy(src->type, dst->type, hlen);\
2590 }\
2591 } while (/*CONSTCOND*/ 0)
2592
2593 static int
copypktopts(struct ip6_pktopts * dst,struct ip6_pktopts * src,int canwait)2594 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2595 {
2596 if (dst == NULL || src == NULL) {
2597 printf("ip6_clearpktopts: invalid argument\n");
2598 return (EINVAL);
2599 }
2600
2601 dst->ip6po_hlim = src->ip6po_hlim;
2602 dst->ip6po_tclass = src->ip6po_tclass;
2603 dst->ip6po_flags = src->ip6po_flags;
2604 dst->ip6po_minmtu = src->ip6po_minmtu;
2605 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2606 if (src->ip6po_pktinfo) {
2607 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2608 M_IP6OPT, canwait);
2609 if (dst->ip6po_pktinfo == NULL)
2610 goto bad;
2611 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2612 }
2613 if (src->ip6po_nexthop) {
2614 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2615 M_IP6OPT, canwait);
2616 if (dst->ip6po_nexthop == NULL)
2617 goto bad;
2618 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2619 src->ip6po_nexthop->sa_len);
2620 }
2621 PKTOPT_EXTHDRCPY(ip6po_hbh);
2622 PKTOPT_EXTHDRCPY(ip6po_dest1);
2623 PKTOPT_EXTHDRCPY(ip6po_dest2);
2624 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2625 dst->ip6po_valid = src->ip6po_valid;
2626 return (0);
2627
2628 bad:
2629 ip6_clearpktopts(dst, -1);
2630 return (ENOBUFS);
2631 }
2632 #undef PKTOPT_EXTHDRCPY
2633
2634 struct ip6_pktopts *
ip6_copypktopts(struct ip6_pktopts * src,int canwait)2635 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2636 {
2637 int error;
2638 struct ip6_pktopts *dst;
2639
2640 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2641 if (dst == NULL)
2642 return (NULL);
2643 ip6_initpktopts(dst);
2644
2645 if ((error = copypktopts(dst, src, canwait)) != 0) {
2646 free(dst, M_IP6OPT);
2647 return (NULL);
2648 }
2649
2650 return (dst);
2651 }
2652
2653 void
ip6_freepcbopts(struct ip6_pktopts * pktopt)2654 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2655 {
2656 if (pktopt == NULL)
2657 return;
2658
2659 ip6_clearpktopts(pktopt, -1);
2660
2661 free(pktopt, M_IP6OPT);
2662 }
2663
2664 /*
2665 * Set IPv6 outgoing packet options based on advanced API.
2666 */
2667 int
ip6_setpktopts(struct mbuf * control,struct ip6_pktopts * opt,struct ip6_pktopts * stickyopt,struct ucred * cred,int uproto)2668 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2669 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2670 {
2671 struct cmsghdr *cm = NULL;
2672
2673 if (control == NULL || opt == NULL)
2674 return (EINVAL);
2675
2676 /*
2677 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2678 * are in the network epoch here.
2679 */
2680 NET_EPOCH_ASSERT();
2681
2682 ip6_initpktopts(opt);
2683 if (stickyopt) {
2684 int error;
2685
2686 /*
2687 * If stickyopt is provided, make a local copy of the options
2688 * for this particular packet, then override them by ancillary
2689 * objects.
2690 * XXX: copypktopts() does not copy the cached route to a next
2691 * hop (if any). This is not very good in terms of efficiency,
2692 * but we can allow this since this option should be rarely
2693 * used.
2694 */
2695 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2696 return (error);
2697 }
2698
2699 /*
2700 * XXX: Currently, we assume all the optional information is stored
2701 * in a single mbuf.
2702 */
2703 if (control->m_next)
2704 return (EINVAL);
2705
2706 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2707 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2708 int error;
2709
2710 if (control->m_len < CMSG_LEN(0))
2711 return (EINVAL);
2712
2713 cm = mtod(control, struct cmsghdr *);
2714 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2715 return (EINVAL);
2716 if (cm->cmsg_level != IPPROTO_IPV6)
2717 continue;
2718
2719 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2720 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2721 if (error)
2722 return (error);
2723 }
2724
2725 return (0);
2726 }
2727
2728 /*
2729 * Set a particular packet option, as a sticky option or an ancillary data
2730 * item. "len" can be 0 only when it's a sticky option.
2731 * We have 4 cases of combination of "sticky" and "cmsg":
2732 * "sticky=0, cmsg=0": impossible
2733 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2734 * "sticky=1, cmsg=0": RFC3542 socket option
2735 * "sticky=1, cmsg=1": RFC2292 socket option
2736 */
2737 static int
ip6_setpktopt(int optname,u_char * buf,int len,struct ip6_pktopts * opt,struct ucred * cred,int sticky,int cmsg,int uproto)2738 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2739 struct ucred *cred, int sticky, int cmsg, int uproto)
2740 {
2741 int minmtupolicy, preftemp;
2742 int error;
2743
2744 NET_EPOCH_ASSERT();
2745
2746 if (!sticky && !cmsg) {
2747 #ifdef DIAGNOSTIC
2748 printf("ip6_setpktopt: impossible case\n");
2749 #endif
2750 return (EINVAL);
2751 }
2752
2753 /*
2754 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2755 * not be specified in the context of RFC3542. Conversely,
2756 * RFC3542 types should not be specified in the context of RFC2292.
2757 */
2758 if (!cmsg) {
2759 switch (optname) {
2760 case IPV6_2292PKTINFO:
2761 case IPV6_2292HOPLIMIT:
2762 case IPV6_2292NEXTHOP:
2763 case IPV6_2292HOPOPTS:
2764 case IPV6_2292DSTOPTS:
2765 case IPV6_2292RTHDR:
2766 case IPV6_2292PKTOPTIONS:
2767 return (ENOPROTOOPT);
2768 }
2769 }
2770 if (sticky && cmsg) {
2771 switch (optname) {
2772 case IPV6_PKTINFO:
2773 case IPV6_HOPLIMIT:
2774 case IPV6_NEXTHOP:
2775 case IPV6_HOPOPTS:
2776 case IPV6_DSTOPTS:
2777 case IPV6_RTHDRDSTOPTS:
2778 case IPV6_RTHDR:
2779 case IPV6_USE_MIN_MTU:
2780 case IPV6_DONTFRAG:
2781 case IPV6_TCLASS:
2782 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2783 return (ENOPROTOOPT);
2784 }
2785 }
2786
2787 switch (optname) {
2788 case IPV6_2292PKTINFO:
2789 case IPV6_PKTINFO:
2790 {
2791 struct ifnet *ifp = NULL;
2792 struct in6_pktinfo *pktinfo;
2793
2794 if (len != sizeof(struct in6_pktinfo))
2795 return (EINVAL);
2796
2797 pktinfo = (struct in6_pktinfo *)buf;
2798
2799 /*
2800 * An application can clear any sticky IPV6_PKTINFO option by
2801 * doing a "regular" setsockopt with ipi6_addr being
2802 * in6addr_any and ipi6_ifindex being zero.
2803 * [RFC 3542, Section 6]
2804 */
2805 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2806 pktinfo->ipi6_ifindex == 0 &&
2807 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2808 ip6_clearpktopts(opt, optname);
2809 break;
2810 }
2811
2812 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2813 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2814 return (EINVAL);
2815 }
2816 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2817 return (EINVAL);
2818 /* validate the interface index if specified. */
2819 if (pktinfo->ipi6_ifindex) {
2820 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2821 if (ifp == NULL)
2822 return (ENXIO);
2823 }
2824 if (ifp != NULL && (ifp->if_inet6 == NULL ||
2825 (ifp->if_inet6->nd_flags & ND6_IFF_IFDISABLED) != 0))
2826 return (ENETDOWN);
2827
2828 if (ifp != NULL &&
2829 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2830 struct in6_ifaddr *ia;
2831
2832 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2833 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2834 if (ia == NULL)
2835 return (EADDRNOTAVAIL);
2836 ifa_free(&ia->ia_ifa);
2837 }
2838 /*
2839 * We store the address anyway, and let in6_selectsrc()
2840 * validate the specified address. This is because ipi6_addr
2841 * may not have enough information about its scope zone, and
2842 * we may need additional information (such as outgoing
2843 * interface or the scope zone of a destination address) to
2844 * disambiguate the scope.
2845 * XXX: the delay of the validation may confuse the
2846 * application when it is used as a sticky option.
2847 */
2848 if (opt->ip6po_pktinfo == NULL) {
2849 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2850 M_IP6OPT, M_NOWAIT);
2851 if (opt->ip6po_pktinfo == NULL)
2852 return (ENOBUFS);
2853 }
2854 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2855 opt->ip6po_valid |= IP6PO_VALID_PKTINFO;
2856 break;
2857 }
2858
2859 case IPV6_2292HOPLIMIT:
2860 case IPV6_HOPLIMIT:
2861 {
2862 int *hlimp;
2863
2864 /*
2865 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2866 * to simplify the ordering among hoplimit options.
2867 */
2868 if (optname == IPV6_HOPLIMIT && sticky)
2869 return (ENOPROTOOPT);
2870
2871 if (len != sizeof(int))
2872 return (EINVAL);
2873 hlimp = (int *)buf;
2874 if (*hlimp < -1 || *hlimp > 255)
2875 return (EINVAL);
2876
2877 opt->ip6po_hlim = *hlimp;
2878 opt->ip6po_valid |= IP6PO_VALID_HLIM;
2879 break;
2880 }
2881
2882 case IPV6_TCLASS:
2883 {
2884 int tclass;
2885
2886 if (len != sizeof(int))
2887 return (EINVAL);
2888 tclass = *(int *)buf;
2889 if (tclass < -1 || tclass > 255)
2890 return (EINVAL);
2891
2892 opt->ip6po_tclass = tclass;
2893 opt->ip6po_valid |= IP6PO_VALID_TC;
2894 break;
2895 }
2896
2897 case IPV6_2292NEXTHOP:
2898 case IPV6_NEXTHOP:
2899 if (cred != NULL) {
2900 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2901 if (error)
2902 return (error);
2903 }
2904
2905 if (len == 0) { /* just remove the option */
2906 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2907 break;
2908 }
2909
2910 /* check if cmsg_len is large enough for sa_len */
2911 if (len < sizeof(struct sockaddr) || len < *buf)
2912 return (EINVAL);
2913
2914 switch (((struct sockaddr *)buf)->sa_family) {
2915 case AF_INET6:
2916 {
2917 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2918 int error;
2919
2920 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2921 return (EINVAL);
2922
2923 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2924 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2925 return (EINVAL);
2926 }
2927 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2928 != 0) {
2929 return (error);
2930 }
2931 break;
2932 }
2933 case AF_LINK: /* should eventually be supported */
2934 default:
2935 return (EAFNOSUPPORT);
2936 }
2937
2938 /* turn off the previous option, then set the new option. */
2939 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2940 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2941 if (opt->ip6po_nexthop == NULL)
2942 return (ENOBUFS);
2943 bcopy(buf, opt->ip6po_nexthop, *buf);
2944 opt->ip6po_valid |= IP6PO_VALID_NHINFO;
2945 break;
2946
2947 case IPV6_2292HOPOPTS:
2948 case IPV6_HOPOPTS:
2949 {
2950 struct ip6_hbh *hbh;
2951 int hbhlen;
2952
2953 /*
2954 * XXX: We don't allow a non-privileged user to set ANY HbH
2955 * options, since per-option restriction has too much
2956 * overhead.
2957 */
2958 if (cred != NULL) {
2959 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2960 if (error)
2961 return (error);
2962 }
2963
2964 if (len == 0) {
2965 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2966 break; /* just remove the option */
2967 }
2968
2969 /* message length validation */
2970 if (len < sizeof(struct ip6_hbh))
2971 return (EINVAL);
2972 hbh = (struct ip6_hbh *)buf;
2973 hbhlen = (hbh->ip6h_len + 1) << 3;
2974 if (len != hbhlen)
2975 return (EINVAL);
2976
2977 /* turn off the previous option, then set the new option. */
2978 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2979 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2980 if (opt->ip6po_hbh == NULL)
2981 return (ENOBUFS);
2982 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2983 opt->ip6po_valid |= IP6PO_VALID_HBH;
2984
2985 break;
2986 }
2987
2988 case IPV6_2292DSTOPTS:
2989 case IPV6_DSTOPTS:
2990 case IPV6_RTHDRDSTOPTS:
2991 {
2992 struct ip6_dest *dest, **newdest = NULL;
2993 int destlen;
2994
2995 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2996 error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2997 if (error)
2998 return (error);
2999 }
3000
3001 if (len == 0) {
3002 ip6_clearpktopts(opt, optname);
3003 break; /* just remove the option */
3004 }
3005
3006 /* message length validation */
3007 if (len < sizeof(struct ip6_dest))
3008 return (EINVAL);
3009 dest = (struct ip6_dest *)buf;
3010 destlen = (dest->ip6d_len + 1) << 3;
3011 if (len != destlen)
3012 return (EINVAL);
3013
3014 /*
3015 * Determine the position that the destination options header
3016 * should be inserted; before or after the routing header.
3017 */
3018 switch (optname) {
3019 case IPV6_2292DSTOPTS:
3020 /*
3021 * The old advacned API is ambiguous on this point.
3022 * Our approach is to determine the position based
3023 * according to the existence of a routing header.
3024 * Note, however, that this depends on the order of the
3025 * extension headers in the ancillary data; the 1st
3026 * part of the destination options header must appear
3027 * before the routing header in the ancillary data,
3028 * too.
3029 * RFC3542 solved the ambiguity by introducing
3030 * separate ancillary data or option types.
3031 */
3032 if (opt->ip6po_rthdr == NULL)
3033 newdest = &opt->ip6po_dest1;
3034 else
3035 newdest = &opt->ip6po_dest2;
3036 break;
3037 case IPV6_RTHDRDSTOPTS:
3038 newdest = &opt->ip6po_dest1;
3039 break;
3040 case IPV6_DSTOPTS:
3041 newdest = &opt->ip6po_dest2;
3042 break;
3043 }
3044
3045 /* turn off the previous option, then set the new option. */
3046 ip6_clearpktopts(opt, optname);
3047 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3048 if (*newdest == NULL)
3049 return (ENOBUFS);
3050 bcopy(dest, *newdest, destlen);
3051 if (newdest == &opt->ip6po_dest1)
3052 opt->ip6po_valid |= IP6PO_VALID_DEST1;
3053 else
3054 opt->ip6po_valid |= IP6PO_VALID_DEST2;
3055
3056 break;
3057 }
3058
3059 case IPV6_2292RTHDR:
3060 case IPV6_RTHDR:
3061 {
3062 struct ip6_rthdr *rth;
3063 int rthlen;
3064
3065 if (len == 0) {
3066 ip6_clearpktopts(opt, IPV6_RTHDR);
3067 break; /* just remove the option */
3068 }
3069
3070 /* message length validation */
3071 if (len < sizeof(struct ip6_rthdr))
3072 return (EINVAL);
3073 rth = (struct ip6_rthdr *)buf;
3074 rthlen = (rth->ip6r_len + 1) << 3;
3075 if (len != rthlen)
3076 return (EINVAL);
3077
3078 switch (rth->ip6r_type) {
3079 case IPV6_RTHDR_TYPE_0:
3080 if (rth->ip6r_len == 0) /* must contain one addr */
3081 return (EINVAL);
3082 if (rth->ip6r_len % 2) /* length must be even */
3083 return (EINVAL);
3084 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3085 return (EINVAL);
3086 break;
3087 default:
3088 return (EINVAL); /* not supported */
3089 }
3090
3091 /* turn off the previous option */
3092 ip6_clearpktopts(opt, IPV6_RTHDR);
3093 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3094 if (opt->ip6po_rthdr == NULL)
3095 return (ENOBUFS);
3096 bcopy(rth, opt->ip6po_rthdr, rthlen);
3097 opt->ip6po_valid |= IP6PO_VALID_RHINFO;
3098
3099 break;
3100 }
3101
3102 case IPV6_USE_MIN_MTU:
3103 if (len != sizeof(int))
3104 return (EINVAL);
3105 minmtupolicy = *(int *)buf;
3106 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3107 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3108 minmtupolicy != IP6PO_MINMTU_ALL) {
3109 return (EINVAL);
3110 }
3111 opt->ip6po_minmtu = minmtupolicy;
3112 break;
3113
3114 case IPV6_DONTFRAG:
3115 if (len != sizeof(int))
3116 return (EINVAL);
3117
3118 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3119 /*
3120 * we ignore this option for TCP sockets.
3121 * (RFC3542 leaves this case unspecified.)
3122 */
3123 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3124 } else
3125 opt->ip6po_flags |= IP6PO_DONTFRAG;
3126 break;
3127
3128 case IPV6_PREFER_TEMPADDR:
3129 if (len != sizeof(int))
3130 return (EINVAL);
3131 preftemp = *(int *)buf;
3132 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3133 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3134 preftemp != IP6PO_TEMPADDR_PREFER) {
3135 return (EINVAL);
3136 }
3137 opt->ip6po_prefer_tempaddr = preftemp;
3138 break;
3139
3140 default:
3141 return (ENOPROTOOPT);
3142 } /* end of switch */
3143
3144 return (0);
3145 }
3146
3147 /*
3148 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3149 * packet to the input queue of a specified interface. Note that this
3150 * calls the output routine of the loopback "driver", but with an interface
3151 * pointer that might NOT be &loif -- easier than replicating that code here.
3152 */
3153 void
ip6_mloopback(struct ifnet * ifp,struct mbuf * m)3154 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3155 {
3156 struct mbuf *copym;
3157 struct ip6_hdr *ip6;
3158
3159 copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3160 if (copym == NULL)
3161 return;
3162
3163 /*
3164 * Make sure to deep-copy IPv6 header portion in case the data
3165 * is in an mbuf cluster, so that we can safely override the IPv6
3166 * header portion later.
3167 */
3168 if (!M_WRITABLE(copym) ||
3169 copym->m_len < sizeof(struct ip6_hdr)) {
3170 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3171 if (copym == NULL)
3172 return;
3173 }
3174 ip6 = mtod(copym, struct ip6_hdr *);
3175 /*
3176 * clear embedded scope identifiers if necessary.
3177 * in6_clearscope will touch the addresses only when necessary.
3178 */
3179 in6_clearscope(&ip6->ip6_src);
3180 in6_clearscope(&ip6->ip6_dst);
3181 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3182 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3183 CSUM_PSEUDO_HDR;
3184 copym->m_pkthdr.csum_data = 0xffff;
3185 }
3186 if_simloop(ifp, copym, AF_INET6, 0);
3187 }
3188
3189 /*
3190 * Chop IPv6 header off from the payload.
3191 */
3192 static int
ip6_splithdr(struct mbuf * m,struct ip6_exthdrs * exthdrs)3193 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3194 {
3195 struct mbuf *mh;
3196 struct ip6_hdr *ip6;
3197
3198 ip6 = mtod(m, struct ip6_hdr *);
3199 if (m->m_len > sizeof(*ip6)) {
3200 mh = m_gethdr(M_NOWAIT, MT_DATA);
3201 if (mh == NULL) {
3202 m_freem(m);
3203 return ENOBUFS;
3204 }
3205 m_move_pkthdr(mh, m);
3206 M_ALIGN(mh, sizeof(*ip6));
3207 m->m_len -= sizeof(*ip6);
3208 m->m_data += sizeof(*ip6);
3209 mh->m_next = m;
3210 m = mh;
3211 m->m_len = sizeof(*ip6);
3212 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3213 }
3214 exthdrs->ip6e_ip6 = m;
3215 return 0;
3216 }
3217
3218 /*
3219 * Compute IPv6 extension header length.
3220 */
3221 int
ip6_optlen(struct inpcb * inp)3222 ip6_optlen(struct inpcb *inp)
3223 {
3224 int len;
3225
3226 if (!inp->in6p_outputopts)
3227 return 0;
3228
3229 len = 0;
3230 #define elen(x) \
3231 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3232
3233 len += elen(inp->in6p_outputopts->ip6po_hbh);
3234 if (inp->in6p_outputopts->ip6po_rthdr)
3235 /* dest1 is valid with rthdr only */
3236 len += elen(inp->in6p_outputopts->ip6po_dest1);
3237 len += elen(inp->in6p_outputopts->ip6po_rthdr);
3238 len += elen(inp->in6p_outputopts->ip6po_dest2);
3239 return len;
3240 #undef elen
3241 }
3242