]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet6/ip6_input.c
xnu-2422.90.20.tar.gz
[apple/xnu.git] / bsd / netinet6 / ip6_input.c
1 /*
2 * Copyright (c) 2003-2013 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29 /*
30 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
31 * All rights reserved.
32 *
33 * Redistribution and use in source and binary forms, with or without
34 * modification, are permitted provided that the following conditions
35 * are met:
36 * 1. Redistributions of source code must retain the above copyright
37 * notice, this list of conditions and the following disclaimer.
38 * 2. Redistributions in binary form must reproduce the above copyright
39 * notice, this list of conditions and the following disclaimer in the
40 * documentation and/or other materials provided with the distribution.
41 * 3. Neither the name of the project nor the names of its contributors
42 * may be used to endorse or promote products derived from this software
43 * without specific prior written permission.
44 *
45 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
46 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
49 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
55 * SUCH DAMAGE.
56 */
57
58 /*
59 * Copyright (c) 1982, 1986, 1988, 1993
60 * The Regents of the University of California. All rights reserved.
61 *
62 * Redistribution and use in source and binary forms, with or without
63 * modification, are permitted provided that the following conditions
64 * are met:
65 * 1. Redistributions of source code must retain the above copyright
66 * notice, this list of conditions and the following disclaimer.
67 * 2. Redistributions in binary form must reproduce the above copyright
68 * notice, this list of conditions and the following disclaimer in the
69 * documentation and/or other materials provided with the distribution.
70 * 3. All advertising materials mentioning features or use of this software
71 * must display the following acknowledgement:
72 * This product includes software developed by the University of
73 * California, Berkeley and its contributors.
74 * 4. Neither the name of the University nor the names of its contributors
75 * may be used to endorse or promote products derived from this software
76 * without specific prior written permission.
77 *
78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88 * SUCH DAMAGE.
89 *
90 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
91 */
92
93 #include <sys/param.h>
94 #include <sys/systm.h>
95 #include <sys/malloc.h>
96 #include <sys/mbuf.h>
97 #include <sys/domain.h>
98 #include <sys/protosw.h>
99 #include <sys/socket.h>
100 #include <sys/socketvar.h>
101 #include <sys/errno.h>
102 #include <sys/time.h>
103 #include <sys/kernel.h>
104 #include <sys/syslog.h>
105 #include <sys/sysctl.h>
106 #include <sys/proc.h>
107 #include <sys/kauth.h>
108 #include <sys/mcache.h>
109
110 #include <mach/mach_time.h>
111 #include <mach/sdt.h>
112 #include <pexpert/pexpert.h>
113 #include <dev/random/randomdev.h>
114
115 #include <net/if.h>
116 #include <net/if_var.h>
117 #include <net/if_types.h>
118 #include <net/if_dl.h>
119 #include <net/route.h>
120 #include <net/kpi_protocol.h>
121 #include <net/ntstat.h>
122 #include <net/init.h>
123 #include <net/net_osdep.h>
124
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #if INET
128 #include <netinet/ip.h>
129 #include <netinet/ip_icmp.h>
130 #endif /* INET */
131 #include <netinet/kpi_ipfilter_var.h>
132 #include <netinet/ip6.h>
133 #include <netinet6/in6_var.h>
134 #include <netinet6/ip6_var.h>
135 #include <netinet/in_pcb.h>
136 #include <netinet/icmp6.h>
137 #include <netinet6/in6_ifattach.h>
138 #include <netinet6/nd6.h>
139 #include <netinet6/scope6_var.h>
140 #include <netinet6/ip6protosw.h>
141
142 #if IPSEC
143 #include <netinet6/ipsec.h>
144 #include <netinet6/ipsec6.h>
145 extern int ipsec_bypass;
146 #endif /* IPSEC */
147
148 #if IPFW2
149 #include <netinet6/ip6_fw.h>
150 #endif /* IPFW2 */
151
152 #if DUMMYNET
153 #include <netinet/ip_fw.h>
154 #include <netinet/ip_dummynet.h>
155 #endif /* DUMMYNET */
156
157 /* we need it for NLOOP. */
158 #include "loop.h"
159
160 #if PF
161 #include <net/pfvar.h>
162 #endif /* PF */
163
164 struct ip6protosw *ip6_protox[IPPROTO_MAX];
165
166 static lck_grp_attr_t *in6_ifaddr_rwlock_grp_attr;
167 static lck_grp_t *in6_ifaddr_rwlock_grp;
168 static lck_attr_t *in6_ifaddr_rwlock_attr;
169 decl_lck_rw_data(, in6_ifaddr_rwlock);
170
171 /* Protected by in6_ifaddr_rwlock */
172 struct in6_ifaddr *in6_ifaddrs = NULL;
173
174 #define IN6_IFSTAT_REQUIRE_ALIGNED_64(f) \
175 _CASSERT(!(offsetof(struct in6_ifstat, f) % sizeof (uint64_t)))
176
177 #define ICMP6_IFSTAT_REQUIRE_ALIGNED_64(f) \
178 _CASSERT(!(offsetof(struct icmp6_ifstat, f) % sizeof (uint64_t)))
179
180 #if IPFW2
181 /* firewall hooks */
182 ip6_fw_chk_t *ip6_fw_chk_ptr;
183 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
184 int ip6_fw_enable = 1;
185 #endif /* IPFW2 */
186
187 struct ip6stat ip6stat;
188
189 decl_lck_mtx_data(, proxy6_lock);
190 decl_lck_mtx_data(static, dad6_mutex_data);
191 decl_lck_mtx_data(static, nd6_mutex_data);
192 decl_lck_mtx_data(static, prefix6_mutex_data);
193 lck_mtx_t *dad6_mutex = &dad6_mutex_data;
194 lck_mtx_t *nd6_mutex = &nd6_mutex_data;
195 lck_mtx_t *prefix6_mutex = &prefix6_mutex_data;
196 #ifdef ENABLE_ADDRSEL
197 decl_lck_mtx_data(static, addrsel_mutex_data);
198 lck_mtx_t *addrsel_mutex = &addrsel_mutex_data;
199 #endif
200 static lck_attr_t *ip6_mutex_attr;
201 static lck_grp_t *ip6_mutex_grp;
202 static lck_grp_attr_t *ip6_mutex_grp_attr;
203
204 extern int loopattach_done;
205 extern void addrsel_policy_init(void);
206
207 static void ip6_init_delayed(void);
208 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
209 #if PULLDOWN_TEST
210 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
211 #endif
212
213 #if NSTF
214 extern void stfattach(void);
215 #endif /* NSTF */
216
217 SYSCTL_DECL(_net_inet6_ip6);
218
219 int ip6_doscopedroute = 1;
220 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, scopedroute,
221 CTLFLAG_RD | CTLFLAG_LOCKED, &ip6_doscopedroute, 0,
222 "Enable IPv6 scoped routing");
223
224 static uint32_t ip6_adj_clear_hwcksum = 0;
225 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, adj_clear_hwcksum,
226 CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_adj_clear_hwcksum, 0,
227 "Invalidate hwcksum info when adjusting length");
228
229 /*
230 * On platforms which require strict alignment (currently for anything but
231 * i386 or x86_64), check if the IP header pointer is 32-bit aligned; if not,
232 * copy the contents of the mbuf chain into a new chain, and free the original
233 * one. Create some head room in the first mbuf of the new chain, in case
234 * it's needed later on.
235 *
236 * RFC 2460 says that IPv6 headers are 64-bit aligned, but network interfaces
237 * mostly align to 32-bit boundaries. Care should be taken never to use 64-bit
238 * load/store operations on the fields in IPv6 headers.
239 */
240 #if defined(__i386__) || defined(__x86_64__)
241 #define IP6_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { } while (0)
242 #else /* !__i386__ && !__x86_64__ */
243 #define IP6_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { \
244 if (!IP6_HDR_ALIGNED_P(mtod(_m, caddr_t))) { \
245 struct mbuf *_n; \
246 struct ifnet *__ifp = (_ifp); \
247 atomic_add_64(&(__ifp)->if_alignerrs, 1); \
248 if (((_m)->m_flags & M_PKTHDR) && \
249 (_m)->m_pkthdr.pkt_hdr != NULL) \
250 (_m)->m_pkthdr.pkt_hdr = NULL; \
251 _n = m_defrag_offset(_m, max_linkhdr, M_NOWAIT); \
252 if (_n == NULL) { \
253 ip6stat.ip6s_toosmall++; \
254 m_freem(_m); \
255 (_m) = NULL; \
256 _action; \
257 } else { \
258 VERIFY(_n != (_m)); \
259 (_m) = _n; \
260 } \
261 } \
262 } while (0)
263 #endif /* !__i386__ && !__x86_64__ */
264
265 static void
266 ip6_proto_input(protocol_family_t protocol, mbuf_t packet)
267 {
268 #pragma unused(protocol)
269 ip6_input(packet);
270 }
271
272 /*
273 * IP6 initialization: fill in IP6 protocol switch table.
274 * All protocols not implemented in kernel go to raw IP6 protocol handler.
275 */
276 void
277 ip6_init(struct ip6protosw *pp, struct domain *dp)
278 {
279 static int ip6_initialized = 0;
280 struct protosw *pr;
281 struct timeval tv;
282 int i;
283 domain_unguard_t unguard;
284
285 domain_proto_mtx_lock_assert_held();
286 VERIFY((pp->pr_flags & (PR_INITIALIZED|PR_ATTACHED)) == PR_ATTACHED);
287
288 _CASSERT((sizeof (struct ip6_hdr) +
289 sizeof (struct icmp6_hdr)) <= _MHLEN);
290
291 if (ip6_initialized)
292 return;
293 ip6_initialized = 1;
294
295 PE_parse_boot_argn("net.inet6.ip6.scopedroute", &ip6_doscopedroute,
296 sizeof (ip6_doscopedroute));
297
298 pr = pffindproto_locked(PF_INET6, IPPROTO_RAW, SOCK_RAW);
299 if (pr == NULL) {
300 panic("%s: Unable to find [PF_INET6,IPPROTO_RAW,SOCK_RAW]\n",
301 __func__);
302 /* NOTREACHED */
303 }
304
305 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
306 for (i = 0; i < IPPROTO_MAX; i++)
307 ip6_protox[i] = (struct ip6protosw *)pr;
308 /*
309 * Cycle through IP protocols and put them into the appropriate place
310 * in ip6_protox[], skipping protocols IPPROTO_{IP,RAW}.
311 */
312 VERIFY(dp == inet6domain && dp->dom_family == PF_INET6);
313 TAILQ_FOREACH(pr, &dp->dom_protosw, pr_entry) {
314 VERIFY(pr->pr_domain == dp);
315 if (pr->pr_protocol != 0 && pr->pr_protocol != IPPROTO_RAW) {
316 /* Be careful to only index valid IP protocols. */
317 if (pr->pr_protocol < IPPROTO_MAX)
318 ip6_protox[pr->pr_protocol] =
319 (struct ip6protosw *)pr;
320 }
321 }
322
323 ip6_mutex_grp_attr = lck_grp_attr_alloc_init();
324
325 ip6_mutex_grp = lck_grp_alloc_init("ip6", ip6_mutex_grp_attr);
326 ip6_mutex_attr = lck_attr_alloc_init();
327
328 lck_mtx_init(dad6_mutex, ip6_mutex_grp, ip6_mutex_attr);
329 lck_mtx_init(nd6_mutex, ip6_mutex_grp, ip6_mutex_attr);
330 lck_mtx_init(prefix6_mutex, ip6_mutex_grp, ip6_mutex_attr);
331 scope6_init(ip6_mutex_grp, ip6_mutex_attr);
332
333 #ifdef ENABLE_ADDRSEL
334 lck_mtx_init(addrsel_mutex, ip6_mutex_grp, ip6_mutex_attr);
335 #endif
336
337 lck_mtx_init(&proxy6_lock, ip6_mutex_grp, ip6_mutex_attr);
338
339 in6_ifaddr_rwlock_grp_attr = lck_grp_attr_alloc_init();
340 in6_ifaddr_rwlock_grp = lck_grp_alloc_init("in6_ifaddr_rwlock",
341 in6_ifaddr_rwlock_grp_attr);
342 in6_ifaddr_rwlock_attr = lck_attr_alloc_init();
343 lck_rw_init(&in6_ifaddr_rwlock, in6_ifaddr_rwlock_grp,
344 in6_ifaddr_rwlock_attr);
345
346 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_receive);
347 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_hdrerr);
348 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_toobig);
349 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_noroute);
350 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_addrerr);
351 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_protounknown);
352 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_truncated);
353 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_discard);
354 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_deliver);
355 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_forward);
356 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_request);
357 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_discard);
358 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragok);
359 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragfail);
360 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragcreat);
361 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_reqd);
362 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_ok);
363 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_fail);
364 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mcast);
365 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mcast);
366
367 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_msg);
368 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_error);
369 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_dstunreach);
370 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_adminprohib);
371 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_timeexceed);
372 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_paramprob);
373 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_pkttoobig);
374 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_echo);
375 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_echoreply);
376 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_routersolicit);
377 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_routeradvert);
378 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_neighborsolicit);
379 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_neighboradvert);
380 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_redirect);
381 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mldquery);
382 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mldreport);
383 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mlddone);
384
385 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_msg);
386 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_error);
387 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_dstunreach);
388 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_adminprohib);
389 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_timeexceed);
390 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_paramprob);
391 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_pkttoobig);
392 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_echo);
393 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_echoreply);
394 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_routersolicit);
395 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_routeradvert);
396 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_neighborsolicit);
397 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_neighboradvert);
398 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_redirect);
399 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mldquery);
400 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mldreport);
401 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mlddone);
402
403 getmicrotime(&tv);
404 ip6_desync_factor =
405 (RandomULong() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
406
407 in6_ifaddr_init();
408 ip6_moptions_init();
409 nd6_init();
410 frag6_init();
411 icmp6_init(NULL, dp);
412 addrsel_policy_init();
413
414 /*
415 * P2P interfaces often route the local address to the loopback
416 * interface. At this point, lo0 hasn't been initialized yet, which
417 * means that we need to delay the IPv6 configuration of lo0.
418 */
419 net_init_add(ip6_init_delayed);
420
421 unguard = domain_unguard_deploy();
422 i = proto_register_input(PF_INET6, ip6_proto_input, NULL, 0);
423 if (i != 0) {
424 panic("%s: failed to register PF_INET6 protocol: %d\n",
425 __func__, i);
426 /* NOTREACHED */
427 }
428 domain_unguard_release(unguard);
429 }
430
431 static void
432 ip6_init_delayed(void)
433 {
434 (void) in6_ifattach_prelim(lo_ifp);
435
436 /* timer for regeneranation of temporary addresses randomize ID */
437 timeout(in6_tmpaddrtimer, NULL,
438 (ip6_temp_preferred_lifetime - ip6_desync_factor -
439 ip6_temp_regen_advance) * hz);
440
441 #if NSTF
442 stfattach();
443 #endif /* NSTF */
444 }
445
446 void
447 ip6_input(struct mbuf *m)
448 {
449 struct ip6_hdr *ip6;
450 int off = sizeof (struct ip6_hdr), nest;
451 u_int32_t plen;
452 u_int32_t rtalert = ~0;
453 int nxt = 0, ours = 0;
454 struct ifnet *inifp, *deliverifp = NULL;
455 ipfilter_t inject_ipfref = NULL;
456 int seen;
457 struct in6_ifaddr *ia6 = NULL;
458 struct sockaddr_in6 *dst6;
459 #if DUMMYNET
460 struct m_tag *tag;
461 #endif /* DUMMYNET */
462 struct {
463 struct route_in6 rin6;
464 #if DUMMYNET
465 struct ip_fw_args args;
466 #endif /* DUMMYNET */
467 } ip6ibz;
468 #define rin6 ip6ibz.rin6
469 #define args ip6ibz.args
470
471 /* zero out {rin6, args} */
472 bzero(&ip6ibz, sizeof (ip6ibz));
473
474 /*
475 * Check if the packet we received is valid after interface filter
476 * processing
477 */
478 MBUF_INPUT_CHECK(m, m->m_pkthdr.rcvif);
479 inifp = m->m_pkthdr.rcvif;
480 VERIFY(inifp != NULL);
481
482 /* Perform IP header alignment fixup, if needed */
483 IP6_HDR_ALIGNMENT_FIXUP(m, inifp, return);
484
485 m->m_pkthdr.pkt_flags &= ~PKTF_FORWARDED;
486 #if IPSEC
487 /*
488 * should the inner packet be considered authentic?
489 * see comment in ah4_input().
490 */
491 m->m_flags &= ~M_AUTHIPHDR;
492 m->m_flags &= ~M_AUTHIPDGM;
493 #endif /* IPSEC */
494
495 /*
496 * make sure we don't have onion peering information into m_aux.
497 */
498 ip6_delaux(m);
499
500 #if DUMMYNET
501 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
502 KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) {
503 struct dn_pkt_tag *dn_tag;
504
505 dn_tag = (struct dn_pkt_tag *)(tag+1);
506
507 args.fwa_pf_rule = dn_tag->dn_pf_rule;
508
509 m_tag_delete(m, tag);
510 }
511
512 if (args.fwa_pf_rule) {
513 ip6 = mtod(m, struct ip6_hdr *); /* In case PF got disabled */
514
515 goto check_with_pf;
516 }
517 #endif /* DUMMYNET */
518
519 /*
520 * No need to proccess packet twice if we've already seen it.
521 */
522 inject_ipfref = ipf_get_inject_filter(m);
523 if (inject_ipfref != NULL) {
524 ip6 = mtod(m, struct ip6_hdr *);
525 nxt = ip6->ip6_nxt;
526 seen = 0;
527 goto injectit;
528 } else {
529 seen = 1;
530 }
531
532 /*
533 * mbuf statistics
534 */
535 if (m->m_flags & M_EXT) {
536 if (m->m_next != NULL)
537 ip6stat.ip6s_mext2m++;
538 else
539 ip6stat.ip6s_mext1++;
540 } else {
541 #define M2MMAX (sizeof (ip6stat.ip6s_m2m) / sizeof (ip6stat.ip6s_m2m[0]))
542 if (m->m_next != NULL) {
543 if (m->m_pkthdr.pkt_flags & PKTF_LOOP) {
544 /* XXX */
545 ip6stat.ip6s_m2m[ifnet_index(lo_ifp)]++;
546 } else if (inifp->if_index < M2MMAX) {
547 ip6stat.ip6s_m2m[inifp->if_index]++;
548 } else {
549 ip6stat.ip6s_m2m[0]++;
550 }
551 } else {
552 ip6stat.ip6s_m1++;
553 }
554 #undef M2MMAX
555 }
556
557 /*
558 * Drop the packet if IPv6 operation is disabled on the interface.
559 */
560 if (inifp->if_eflags & IFEF_IPV6_DISABLED)
561 goto bad;
562
563 in6_ifstat_inc_na(inifp, ifs6_in_receive);
564 ip6stat.ip6s_total++;
565
566 #ifndef PULLDOWN_TEST
567 /*
568 * L2 bridge code and some other code can return mbuf chain
569 * that does not conform to KAME requirement. too bad.
570 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
571 */
572 if (m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
573 struct mbuf *n;
574
575 MGETHDR(n, M_DONTWAIT, MT_HEADER); /* MAC-OK */
576 if (n)
577 M_COPY_PKTHDR(n, m);
578 if (n && m->m_pkthdr.len > MHLEN) {
579 MCLGET(n, M_DONTWAIT);
580 if ((n->m_flags & M_EXT) == 0) {
581 m_freem(n);
582 n = NULL;
583 }
584 }
585 if (n == NULL)
586 goto bad;
587
588 m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
589 n->m_len = m->m_pkthdr.len;
590 m_freem(m);
591 m = n;
592 }
593 IP6_EXTHDR_CHECK(m, 0, sizeof (struct ip6_hdr), { goto done; });
594 #endif
595
596 if (m->m_len < sizeof (struct ip6_hdr)) {
597 if ((m = m_pullup(m, sizeof (struct ip6_hdr))) == 0) {
598 ip6stat.ip6s_toosmall++;
599 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
600 goto done;
601 }
602 }
603
604 ip6 = mtod(m, struct ip6_hdr *);
605
606 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
607 ip6stat.ip6s_badvers++;
608 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
609 goto bad;
610 }
611
612 ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
613
614 #if IPFW2
615 /*
616 * Check with the firewall...
617 */
618 if (ip6_fw_enable && ip6_fw_chk_ptr) {
619 u_short port = 0;
620 /* If ipfw says divert, we have to just drop packet */
621 /* use port as a dummy argument */
622 if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
623 m_freem(m);
624 m = NULL;
625 }
626 if (!m)
627 goto done;
628 }
629 #endif /* IPFW2 */
630
631 /*
632 * Check against address spoofing/corruption.
633 */
634 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
635 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
636 /*
637 * XXX: "badscope" is not very suitable for a multicast source.
638 */
639 ip6stat.ip6s_badscope++;
640 in6_ifstat_inc(inifp, ifs6_in_addrerr);
641 goto bad;
642 }
643 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
644 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
645 /*
646 * In this case, the packet should come from the loopback
647 * interface. However, we cannot just check the if_flags,
648 * because ip6_mloopback() passes the "actual" interface
649 * as the outgoing/incoming interface.
650 */
651 ip6stat.ip6s_badscope++;
652 in6_ifstat_inc(inifp, ifs6_in_addrerr);
653 goto bad;
654 }
655
656 /*
657 * The following check is not documented in specs. A malicious
658 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
659 * and bypass security checks (act as if it was from 127.0.0.1 by using
660 * IPv6 src ::ffff:127.0.0.1). Be cautious.
661 *
662 * This check chokes if we are in an SIIT cloud. As none of BSDs
663 * support IPv4-less kernel compilation, we cannot support SIIT
664 * environment at all. So, it makes more sense for us to reject any
665 * malicious packets for non-SIIT environment, than try to do a
666 * partial support for SIIT environment.
667 */
668 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
669 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
670 ip6stat.ip6s_badscope++;
671 in6_ifstat_inc(inifp, ifs6_in_addrerr);
672 goto bad;
673 }
674 #if 0
675 /*
676 * Reject packets with IPv4 compatible addresses (auto tunnel).
677 *
678 * The code forbids auto tunnel relay case in RFC1933 (the check is
679 * stronger than RFC1933). We may want to re-enable it if mech-xx
680 * is revised to forbid relaying case.
681 */
682 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
683 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
684 ip6stat.ip6s_badscope++;
685 in6_ifstat_inc(inifp, ifs6_in_addrerr);
686 goto bad;
687 }
688 #endif
689
690 /*
691 * Naively assume we can attribute inbound data to the route we would
692 * use to send to this destination. Asymetric routing breaks this
693 * assumption, but it still allows us to account for traffic from
694 * a remote node in the routing table.
695 * this has a very significant performance impact so we bypass
696 * if nstat_collect is disabled. We may also bypass if the
697 * protocol is tcp in the future because tcp will have a route that
698 * we can use to attribute the data to. That does mean we would not
699 * account for forwarded tcp traffic.
700 */
701 if (nstat_collect) {
702 struct rtentry *rte =
703 ifnet_cached_rtlookup_inet6(inifp, &ip6->ip6_src);
704 if (rte != NULL) {
705 nstat_route_rx(rte, 1, m->m_pkthdr.len, 0);
706 rtfree(rte);
707 }
708 }
709
710 /* for consistency */
711 m->m_pkthdr.pkt_proto = ip6->ip6_nxt;
712
713 #if DUMMYNET
714 check_with_pf:
715 #endif /* DUMMYNET */
716 #if PF
717 /* Invoke inbound packet filter */
718 if (PF_IS_ENABLED) {
719 int error;
720 #if DUMMYNET
721 error = pf_af_hook(inifp, NULL, &m, AF_INET6, TRUE, &args);
722 #else /* !DUMMYNET */
723 error = pf_af_hook(inifp, NULL, &m, AF_INET6, TRUE, NULL);
724 #endif /* !DUMMYNET */
725 if (error != 0 || m == NULL) {
726 if (m != NULL) {
727 panic("%s: unexpected packet %p\n",
728 __func__, m);
729 /* NOTREACHED */
730 }
731 /* Already freed by callee */
732 goto done;
733 }
734 ip6 = mtod(m, struct ip6_hdr *);
735 }
736 #endif /* PF */
737
738 /* drop packets if interface ID portion is already filled */
739 if (!(inifp->if_flags & IFF_LOOPBACK) &&
740 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
741 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
742 ip6->ip6_src.s6_addr16[1]) {
743 ip6stat.ip6s_badscope++;
744 goto bad;
745 }
746 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
747 ip6->ip6_dst.s6_addr16[1]) {
748 ip6stat.ip6s_badscope++;
749 goto bad;
750 }
751 }
752
753 if (m->m_pkthdr.pkt_flags & PKTF_IFAINFO) {
754 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
755 ip6->ip6_src.s6_addr16[1] =
756 htons(m->m_pkthdr.src_ifindex);
757 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
758 ip6->ip6_dst.s6_addr16[1] =
759 htons(m->m_pkthdr.dst_ifindex);
760 } else {
761 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
762 ip6->ip6_src.s6_addr16[1] = htons(inifp->if_index);
763 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
764 ip6->ip6_dst.s6_addr16[1] = htons(inifp->if_index);
765 }
766
767 /*
768 * Multicast check
769 */
770 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
771 struct in6_multi *in6m = NULL;
772
773 in6_ifstat_inc_na(inifp, ifs6_in_mcast);
774 /*
775 * See if we belong to the destination multicast group on the
776 * arrival interface.
777 */
778 in6_multihead_lock_shared();
779 IN6_LOOKUP_MULTI(&ip6->ip6_dst, inifp, in6m);
780 in6_multihead_lock_done();
781 if (in6m != NULL) {
782 IN6M_REMREF(in6m);
783 ours = 1;
784 } else if (!nd6_prproxy
785 #if MROUTING
786 && !ip6_mrouter
787 #endif /* MROUTING */
788 ) {
789 ip6stat.ip6s_notmember++;
790 ip6stat.ip6s_cantforward++;
791 in6_ifstat_inc(inifp, ifs6_in_discard);
792 goto bad;
793 }
794 deliverifp = inifp;
795 VERIFY(ia6 == NULL);
796 goto hbhcheck;
797 }
798
799 /*
800 * Unicast check
801 *
802 * Fast path: see if the target is ourselves.
803 */
804 lck_rw_lock_shared(&in6_ifaddr_rwlock);
805 for (ia6 = in6_ifaddrs; ia6 != NULL; ia6 = ia6->ia_next) {
806 /*
807 * No reference is held on the address, as we just need
808 * to test for a few things while holding the RW lock.
809 */
810 if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &ip6->ip6_dst))
811 break;
812 }
813
814 if (ia6 != NULL) {
815 /*
816 * For performance, test without acquiring the address lock;
817 * a lot of things in the address are set once and never
818 * changed (e.g. ia_ifp.)
819 */
820 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
821 /* this address is ready */
822 ours = 1;
823 deliverifp = ia6->ia_ifp;
824 /*
825 * record dst address information into mbuf.
826 */
827 (void) ip6_setdstifaddr_info(m, 0, ia6);
828 lck_rw_done(&in6_ifaddr_rwlock);
829 goto hbhcheck;
830 }
831 lck_rw_done(&in6_ifaddr_rwlock);
832 ia6 = NULL;
833 /* address is not ready, so discard the packet. */
834 nd6log((LOG_INFO, "%s: packet to an unready address %s->%s\n",
835 __func__, ip6_sprintf(&ip6->ip6_src),
836 ip6_sprintf(&ip6->ip6_dst)));
837 goto bad;
838 }
839 lck_rw_done(&in6_ifaddr_rwlock);
840
841 /*
842 * Slow path: route lookup.
843 */
844 dst6 = SIN6(&rin6.ro_dst);
845 dst6->sin6_len = sizeof (struct sockaddr_in6);
846 dst6->sin6_family = AF_INET6;
847 dst6->sin6_addr = ip6->ip6_dst;
848
849 rtalloc_scoped_ign((struct route *)&rin6,
850 RTF_PRCLONING, IFSCOPE_NONE);
851 if (rin6.ro_rt != NULL)
852 RT_LOCK_SPIN(rin6.ro_rt);
853
854 #define rt6_key(r) (SIN6((r)->rt_nodes->rn_key))
855
856 /*
857 * Accept the packet if the forwarding interface to the destination
858 * according to the routing table is the loopback interface,
859 * unless the associated route has a gateway.
860 * Note that this approach causes to accept a packet if there is a
861 * route to the loopback interface for the destination of the packet.
862 * But we think it's even useful in some situations, e.g. when using
863 * a special daemon which wants to intercept the packet.
864 *
865 * XXX: some OSes automatically make a cloned route for the destination
866 * of an outgoing packet. If the outgoing interface of the packet
867 * is a loopback one, the kernel would consider the packet to be
868 * accepted, even if we have no such address assinged on the interface.
869 * We check the cloned flag of the route entry to reject such cases,
870 * assuming that route entries for our own addresses are not made by
871 * cloning (it should be true because in6_addloop explicitly installs
872 * the host route). However, we might have to do an explicit check
873 * while it would be less efficient. Or, should we rather install a
874 * reject route for such a case?
875 */
876 if (rin6.ro_rt != NULL &&
877 (rin6.ro_rt->rt_flags & (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
878 #if RTF_WASCLONED
879 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
880 #endif
881 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
882 ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
883 /*
884 * Packets to a tentative, duplicated, or somehow invalid
885 * address must not be accepted.
886 *
887 * For performance, test without acquiring the address lock;
888 * a lot of things in the address are set once and never
889 * changed (e.g. ia_ifp.)
890 */
891 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
892 /* this address is ready */
893 ours = 1;
894 deliverifp = ia6->ia_ifp; /* correct? */
895 /*
896 * record dst address information into mbuf.
897 */
898 (void) ip6_setdstifaddr_info(m, 0, ia6);
899 RT_UNLOCK(rin6.ro_rt);
900 goto hbhcheck;
901 }
902 RT_UNLOCK(rin6.ro_rt);
903 ia6 = NULL;
904 /* address is not ready, so discard the packet. */
905 nd6log((LOG_INFO, "%s: packet to an unready address %s->%s\n",
906 __func__, ip6_sprintf(&ip6->ip6_src),
907 ip6_sprintf(&ip6->ip6_dst)));
908 goto bad;
909 }
910
911 if (rin6.ro_rt != NULL)
912 RT_UNLOCK(rin6.ro_rt);
913
914 /*
915 * Now there is no reason to process the packet if it's not our own
916 * and we're not a router.
917 */
918 if (!ip6_forwarding) {
919 ip6stat.ip6s_cantforward++;
920 in6_ifstat_inc(inifp, ifs6_in_discard);
921 goto bad;
922 }
923
924 hbhcheck:
925 /*
926 * record dst address information into mbuf, if we don't have one yet.
927 * note that we are unable to record it, if the address is not listed
928 * as our interface address (e.g. multicast addresses, etc.)
929 */
930 if (deliverifp != NULL && ia6 == NULL) {
931 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
932 if (ia6 != NULL) {
933 (void) ip6_setdstifaddr_info(m, 0, ia6);
934 IFA_REMREF(&ia6->ia_ifa);
935 }
936 }
937
938 /*
939 * Process Hop-by-Hop options header if it's contained.
940 * m may be modified in ip6_hopopts_input().
941 * If a JumboPayload option is included, plen will also be modified.
942 */
943 plen = (u_int32_t)ntohs(ip6->ip6_plen);
944 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
945 struct ip6_hbh *hbh;
946
947 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
948 #if 0 /* touches NULL pointer */
949 in6_ifstat_inc(inifp, ifs6_in_discard);
950 #endif
951 goto done; /* m have already been freed */
952 }
953
954 /* adjust pointer */
955 ip6 = mtod(m, struct ip6_hdr *);
956
957 /*
958 * if the payload length field is 0 and the next header field
959 * indicates Hop-by-Hop Options header, then a Jumbo Payload
960 * option MUST be included.
961 */
962 if (ip6->ip6_plen == 0 && plen == 0) {
963 /*
964 * Note that if a valid jumbo payload option is
965 * contained, ip6_hopopts_input() must set a valid
966 * (non-zero) payload length to the variable plen.
967 */
968 ip6stat.ip6s_badoptions++;
969 in6_ifstat_inc(inifp, ifs6_in_discard);
970 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
971 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
972 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
973 goto done;
974 }
975 #ifndef PULLDOWN_TEST
976 /* ip6_hopopts_input() ensures that mbuf is contiguous */
977 hbh = (struct ip6_hbh *)(ip6 + 1);
978 #else
979 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
980 sizeof (struct ip6_hdr), sizeof (struct ip6_hbh));
981 if (hbh == NULL) {
982 ip6stat.ip6s_tooshort++;
983 goto done;
984 }
985 #endif
986 nxt = hbh->ip6h_nxt;
987
988 /*
989 * If we are acting as a router and the packet contains a
990 * router alert option, see if we know the option value.
991 * Currently, we only support the option value for MLD, in which
992 * case we should pass the packet to the multicast routing
993 * daemon.
994 */
995 if (rtalert != ~0 && ip6_forwarding) {
996 switch (rtalert) {
997 case IP6OPT_RTALERT_MLD:
998 ours = 1;
999 break;
1000 default:
1001 /*
1002 * RFC2711 requires unrecognized values must be
1003 * silently ignored.
1004 */
1005 break;
1006 }
1007 }
1008 } else
1009 nxt = ip6->ip6_nxt;
1010
1011 /*
1012 * Check that the amount of data in the buffers
1013 * is as at least much as the IPv6 header would have us expect.
1014 * Trim mbufs if longer than we expect.
1015 * Drop packet if shorter than we expect.
1016 */
1017 if (m->m_pkthdr.len - sizeof (struct ip6_hdr) < plen) {
1018 ip6stat.ip6s_tooshort++;
1019 in6_ifstat_inc(inifp, ifs6_in_truncated);
1020 goto bad;
1021 }
1022 if (m->m_pkthdr.len > sizeof (struct ip6_hdr) + plen) {
1023 /*
1024 * Invalidate hardware checksum info if ip6_adj_clear_hwcksum
1025 * is set; useful to handle buggy drivers. Note that this
1026 * should not be enabled by default, as we may get here due
1027 * to link-layer padding.
1028 */
1029 if (ip6_adj_clear_hwcksum &&
1030 (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) &&
1031 !(inifp->if_flags & IFF_LOOPBACK) &&
1032 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
1033 m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID;
1034 m->m_pkthdr.csum_data = 0;
1035 ip6stat.ip6s_adj_hwcsum_clr++;
1036 }
1037
1038 ip6stat.ip6s_adj++;
1039 if (m->m_len == m->m_pkthdr.len) {
1040 m->m_len = sizeof (struct ip6_hdr) + plen;
1041 m->m_pkthdr.len = sizeof (struct ip6_hdr) + plen;
1042 } else {
1043 m_adj(m, sizeof (struct ip6_hdr) + plen -
1044 m->m_pkthdr.len);
1045 }
1046 }
1047
1048 /*
1049 * Forward if desirable.
1050 */
1051 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1052 /*
1053 * If we are acting as a multicast router, all
1054 * incoming multicast packets are passed to the
1055 * kernel-level multicast forwarding function.
1056 * The packet is returned (relatively) intact; if
1057 * ip6_mforward() returns a non-zero value, the packet
1058 * must be discarded, else it may be accepted below.
1059 */
1060 #if MROUTING
1061 if (ip6_mrouter && ip6_mforward(ip6, inifp, m)) {
1062 ip6stat.ip6s_cantforward++;
1063 goto bad;
1064 }
1065 #endif /* MROUTING */
1066 if (!ours && nd6_prproxy) {
1067 /*
1068 * If this isn't for us, this might be a Neighbor
1069 * Solicitation (dst is solicited-node multicast)
1070 * against an address in one of the proxied prefixes;
1071 * if so, claim the packet and let icmp6_input()
1072 * handle the rest.
1073 */
1074 ours = nd6_prproxy_isours(m, ip6, NULL, IFSCOPE_NONE);
1075 VERIFY(!ours ||
1076 (m->m_pkthdr.pkt_flags & PKTF_PROXY_DST));
1077 }
1078 if (!ours)
1079 goto bad;
1080 } else if (!ours) {
1081 /*
1082 * The unicast forwarding function might return the packet
1083 * if we are proxying prefix(es), and if the packet is an
1084 * ICMPv6 packet that has failed the zone checks, but is
1085 * targetted towards a proxied address (this is optimized by
1086 * way of RTF_PROXY test.) If so, claim the packet as ours
1087 * and let icmp6_input() handle the rest. The packet's hop
1088 * limit value is kept intact (it's not decremented). This
1089 * is for supporting Neighbor Unreachability Detection between
1090 * proxied nodes on different links (src is link-local, dst
1091 * is target address.)
1092 */
1093 if ((m = ip6_forward(m, &rin6, 0)) == NULL)
1094 goto done;
1095 VERIFY(rin6.ro_rt != NULL);
1096 VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
1097 deliverifp = rin6.ro_rt->rt_ifp;
1098 ours = 1;
1099 }
1100
1101 ip6 = mtod(m, struct ip6_hdr *);
1102
1103 /*
1104 * Malicious party may be able to use IPv4 mapped addr to confuse
1105 * tcp/udp stack and bypass security checks (act as if it was from
1106 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious.
1107 *
1108 * For SIIT end node behavior, you may want to disable the check.
1109 * However, you will become vulnerable to attacks using IPv4 mapped
1110 * source.
1111 */
1112 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
1113 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
1114 ip6stat.ip6s_badscope++;
1115 in6_ifstat_inc(inifp, ifs6_in_addrerr);
1116 goto bad;
1117 }
1118
1119 /*
1120 * Tell launch routine the next header
1121 */
1122 ip6stat.ip6s_delivered++;
1123 in6_ifstat_inc_na(deliverifp, ifs6_in_deliver);
1124
1125 injectit:
1126 nest = 0;
1127
1128 /*
1129 * Perform IP header alignment fixup again, if needed. Note that
1130 * we do it once for the outermost protocol, and we assume each
1131 * protocol handler wouldn't mess with the alignment afterwards.
1132 */
1133 IP6_HDR_ALIGNMENT_FIXUP(m, inifp, return);
1134
1135 while (nxt != IPPROTO_DONE) {
1136 struct ipfilter *filter;
1137 int (*pr_input)(struct mbuf **, int *, int);
1138
1139 if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
1140 ip6stat.ip6s_toomanyhdr++;
1141 goto bad;
1142 }
1143
1144 /*
1145 * protection against faulty packet - there should be
1146 * more sanity checks in header chain processing.
1147 */
1148 if (m->m_pkthdr.len < off) {
1149 ip6stat.ip6s_tooshort++;
1150 in6_ifstat_inc(inifp, ifs6_in_truncated);
1151 goto bad;
1152 }
1153
1154
1155 #if IPSEC
1156 /*
1157 * enforce IPsec policy checking if we are seeing last header.
1158 * note that we do not visit this with protocols with pcb layer
1159 * code - like udp/tcp/raw ip.
1160 */
1161 if ((ipsec_bypass == 0) &&
1162 (ip6_protox[nxt]->pr_flags & PR_LASTHDR) != 0) {
1163 if (ipsec6_in_reject(m, NULL)) {
1164 IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
1165 goto bad;
1166 }
1167 }
1168 #endif /* IPSEC */
1169
1170 /*
1171 * Call IP filter
1172 */
1173 if (!TAILQ_EMPTY(&ipv6_filters)) {
1174 ipf_ref();
1175 TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) {
1176 if (seen == 0) {
1177 if ((struct ipfilter *)inject_ipfref ==
1178 filter)
1179 seen = 1;
1180 } else if (filter->ipf_filter.ipf_input) {
1181 errno_t result;
1182
1183 result = filter->ipf_filter.ipf_input(
1184 filter->ipf_filter.cookie,
1185 (mbuf_t *)&m, off, nxt);
1186 if (result == EJUSTRETURN) {
1187 ipf_unref();
1188 goto done;
1189 }
1190 if (result != 0) {
1191 ipf_unref();
1192 goto bad;
1193 }
1194 }
1195 }
1196 ipf_unref();
1197 }
1198
1199 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
1200 struct ip6_hdr *, ip6, struct ifnet *, inifp,
1201 struct ip *, NULL, struct ip6_hdr *, ip6);
1202
1203 if ((pr_input = ip6_protox[nxt]->pr_input) == NULL) {
1204 m_freem(m);
1205 m = NULL;
1206 nxt = IPPROTO_DONE;
1207 } else if (!(ip6_protox[nxt]->pr_flags & PR_PROTOLOCK)) {
1208 lck_mtx_lock(inet6_domain_mutex);
1209 nxt = pr_input(&m, &off, nxt);
1210 lck_mtx_unlock(inet6_domain_mutex);
1211 } else {
1212 nxt = pr_input(&m, &off, nxt);
1213 }
1214 }
1215 done:
1216 ROUTE_RELEASE(&rin6);
1217 return;
1218 bad:
1219 m_freem(m);
1220 goto done;
1221 }
1222
1223 void
1224 ip6_setsrcifaddr_info(struct mbuf *m, uint32_t src_idx, struct in6_ifaddr *ia6)
1225 {
1226 VERIFY(m->m_flags & M_PKTHDR);
1227
1228 /*
1229 * If the source ifaddr is specified, pick up the information
1230 * from there; otherwise just grab the passed-in ifindex as the
1231 * caller may not have the ifaddr available.
1232 */
1233 if (ia6 != NULL) {
1234 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1235 m->m_pkthdr.src_ifindex = ia6->ia_ifp->if_index;
1236
1237 /* See IN6_IFF comments in in6_var.h */
1238 m->m_pkthdr.src_iff = (ia6->ia6_flags & 0xffff);
1239 } else {
1240 m->m_pkthdr.src_iff = 0;
1241 m->m_pkthdr.src_ifindex = src_idx;
1242 if (src_idx != 0)
1243 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1244 }
1245 }
1246
1247 void
1248 ip6_setdstifaddr_info(struct mbuf *m, uint32_t dst_idx, struct in6_ifaddr *ia6)
1249 {
1250 VERIFY(m->m_flags & M_PKTHDR);
1251
1252 /*
1253 * If the destination ifaddr is specified, pick up the information
1254 * from there; otherwise just grab the passed-in ifindex as the
1255 * caller may not have the ifaddr available.
1256 */
1257 if (ia6 != NULL) {
1258 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1259 m->m_pkthdr.dst_ifindex = ia6->ia_ifp->if_index;
1260
1261 /* See IN6_IFF comments in in6_var.h */
1262 m->m_pkthdr.dst_iff = (ia6->ia6_flags & 0xffff);
1263 } else {
1264 m->m_pkthdr.dst_iff = 0;
1265 m->m_pkthdr.dst_ifindex = dst_idx;
1266 if (dst_idx != 0)
1267 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1268 }
1269 }
1270
1271 int
1272 ip6_getsrcifaddr_info(struct mbuf *m, uint32_t *src_idx, uint32_t *ia6f)
1273 {
1274 VERIFY(m->m_flags & M_PKTHDR);
1275
1276 if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO))
1277 return (-1);
1278
1279 if (src_idx != NULL)
1280 *src_idx = m->m_pkthdr.src_ifindex;
1281
1282 if (ia6f != NULL)
1283 *ia6f = m->m_pkthdr.src_iff;
1284
1285 return (0);
1286 }
1287
1288 int
1289 ip6_getdstifaddr_info(struct mbuf *m, uint32_t *dst_idx, uint32_t *ia6f)
1290 {
1291 VERIFY(m->m_flags & M_PKTHDR);
1292
1293 if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO))
1294 return (-1);
1295
1296 if (dst_idx != NULL)
1297 *dst_idx = m->m_pkthdr.dst_ifindex;
1298
1299 if (ia6f != NULL)
1300 *ia6f = m->m_pkthdr.dst_iff;
1301
1302 return (0);
1303 }
1304
1305 /*
1306 * Hop-by-Hop options header processing. If a valid jumbo payload option is
1307 * included, the real payload length will be stored in plenp.
1308 */
1309 static int
1310 ip6_hopopts_input(uint32_t *plenp, uint32_t *rtalertp, struct mbuf **mp,
1311 int *offp)
1312 {
1313 struct mbuf *m = *mp;
1314 int off = *offp, hbhlen;
1315 struct ip6_hbh *hbh;
1316 u_int8_t *opt;
1317
1318 /* validation of the length of the header */
1319 #ifndef PULLDOWN_TEST
1320 IP6_EXTHDR_CHECK(m, off, sizeof (*hbh), return (-1));
1321 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1322 hbhlen = (hbh->ip6h_len + 1) << 3;
1323
1324 IP6_EXTHDR_CHECK(m, off, hbhlen, return (-1));
1325 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1326 #else
1327 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof (struct ip6_hdr),
1328 sizeof (struct ip6_hbh));
1329 if (hbh == NULL) {
1330 ip6stat.ip6s_tooshort++;
1331 return (-1);
1332 }
1333 hbhlen = (hbh->ip6h_len + 1) << 3;
1334 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof (struct ip6_hdr),
1335 hbhlen);
1336 if (hbh == NULL) {
1337 ip6stat.ip6s_tooshort++;
1338 return (-1);
1339 }
1340 #endif
1341 off += hbhlen;
1342 hbhlen -= sizeof (struct ip6_hbh);
1343 opt = (u_int8_t *)hbh + sizeof (struct ip6_hbh);
1344
1345 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof (struct ip6_hbh),
1346 hbhlen, rtalertp, plenp) < 0)
1347 return (-1);
1348
1349 *offp = off;
1350 *mp = m;
1351 return (0);
1352 }
1353
1354 /*
1355 * Search header for all Hop-by-hop options and process each option.
1356 * This function is separate from ip6_hopopts_input() in order to
1357 * handle a case where the sending node itself process its hop-by-hop
1358 * options header. In such a case, the function is called from ip6_output().
1359 *
1360 * The function assumes that hbh header is located right after the IPv6 header
1361 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1362 * opthead + hbhlen is located in continuous memory region.
1363 */
1364 int
1365 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
1366 struct mbuf *m;
1367 u_int8_t *opthead;
1368 int hbhlen;
1369 u_int32_t *rtalertp;
1370 u_int32_t *plenp;
1371 {
1372 struct ip6_hdr *ip6;
1373 int optlen = 0;
1374 u_int8_t *opt = opthead;
1375 u_int16_t rtalert_val;
1376 u_int32_t jumboplen;
1377 const int erroff = sizeof (struct ip6_hdr) + sizeof (struct ip6_hbh);
1378
1379 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1380 switch (*opt) {
1381 case IP6OPT_PAD1:
1382 optlen = 1;
1383 break;
1384 case IP6OPT_PADN:
1385 if (hbhlen < IP6OPT_MINLEN) {
1386 ip6stat.ip6s_toosmall++;
1387 goto bad;
1388 }
1389 optlen = *(opt + 1) + 2;
1390 break;
1391 case IP6OPT_ROUTER_ALERT:
1392 /* XXX may need check for alignment */
1393 if (hbhlen < IP6OPT_RTALERT_LEN) {
1394 ip6stat.ip6s_toosmall++;
1395 goto bad;
1396 }
1397 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1398 /* XXX stat */
1399 icmp6_error(m, ICMP6_PARAM_PROB,
1400 ICMP6_PARAMPROB_HEADER,
1401 erroff + opt + 1 - opthead);
1402 return (-1);
1403 }
1404 optlen = IP6OPT_RTALERT_LEN;
1405 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1406 *rtalertp = ntohs(rtalert_val);
1407 break;
1408 case IP6OPT_JUMBO:
1409 /* XXX may need check for alignment */
1410 if (hbhlen < IP6OPT_JUMBO_LEN) {
1411 ip6stat.ip6s_toosmall++;
1412 goto bad;
1413 }
1414 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1415 /* XXX stat */
1416 icmp6_error(m, ICMP6_PARAM_PROB,
1417 ICMP6_PARAMPROB_HEADER,
1418 erroff + opt + 1 - opthead);
1419 return (-1);
1420 }
1421 optlen = IP6OPT_JUMBO_LEN;
1422
1423 /*
1424 * IPv6 packets that have non 0 payload length
1425 * must not contain a jumbo payload option.
1426 */
1427 ip6 = mtod(m, struct ip6_hdr *);
1428 if (ip6->ip6_plen) {
1429 ip6stat.ip6s_badoptions++;
1430 icmp6_error(m, ICMP6_PARAM_PROB,
1431 ICMP6_PARAMPROB_HEADER,
1432 erroff + opt - opthead);
1433 return (-1);
1434 }
1435
1436 /*
1437 * We may see jumbolen in unaligned location, so
1438 * we'd need to perform bcopy().
1439 */
1440 bcopy(opt + 2, &jumboplen, sizeof (jumboplen));
1441 jumboplen = (u_int32_t)htonl(jumboplen);
1442
1443 #if 1
1444 /*
1445 * if there are multiple jumbo payload options,
1446 * *plenp will be non-zero and the packet will be
1447 * rejected.
1448 * the behavior may need some debate in ipngwg -
1449 * multiple options does not make sense, however,
1450 * there's no explicit mention in specification.
1451 */
1452 if (*plenp != 0) {
1453 ip6stat.ip6s_badoptions++;
1454 icmp6_error(m, ICMP6_PARAM_PROB,
1455 ICMP6_PARAMPROB_HEADER,
1456 erroff + opt + 2 - opthead);
1457 return (-1);
1458 }
1459 #endif
1460
1461 /*
1462 * jumbo payload length must be larger than 65535.
1463 */
1464 if (jumboplen <= IPV6_MAXPACKET) {
1465 ip6stat.ip6s_badoptions++;
1466 icmp6_error(m, ICMP6_PARAM_PROB,
1467 ICMP6_PARAMPROB_HEADER,
1468 erroff + opt + 2 - opthead);
1469 return (-1);
1470 }
1471 *plenp = jumboplen;
1472
1473 break;
1474 default: /* unknown option */
1475 if (hbhlen < IP6OPT_MINLEN) {
1476 ip6stat.ip6s_toosmall++;
1477 goto bad;
1478 }
1479 optlen = ip6_unknown_opt(opt, m,
1480 erroff + opt - opthead);
1481 if (optlen == -1) {
1482 return (-1);
1483 }
1484 optlen += 2;
1485 break;
1486 }
1487 }
1488
1489 return (0);
1490
1491 bad:
1492 m_freem(m);
1493 return (-1);
1494 }
1495
1496 /*
1497 * Unknown option processing.
1498 * The third argument `off' is the offset from the IPv6 header to the option,
1499 * which is necessary if the IPv6 header the and option header and IPv6 header
1500 * is not continuous in order to return an ICMPv6 error.
1501 */
1502 int
1503 ip6_unknown_opt(uint8_t *optp, struct mbuf *m, int off)
1504 {
1505 struct ip6_hdr *ip6;
1506
1507 switch (IP6OPT_TYPE(*optp)) {
1508 case IP6OPT_TYPE_SKIP: /* ignore the option */
1509 return ((int)*(optp + 1));
1510
1511 case IP6OPT_TYPE_DISCARD: /* silently discard */
1512 m_freem(m);
1513 return (-1);
1514
1515 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1516 ip6stat.ip6s_badoptions++;
1517 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1518 return (-1);
1519
1520 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1521 ip6stat.ip6s_badoptions++;
1522 ip6 = mtod(m, struct ip6_hdr *);
1523 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1524 (m->m_flags & (M_BCAST|M_MCAST))) {
1525 m_freem(m);
1526 } else {
1527 icmp6_error(m, ICMP6_PARAM_PROB,
1528 ICMP6_PARAMPROB_OPTION, off);
1529 }
1530 return (-1);
1531 }
1532
1533 m_freem(m); /* XXX: NOTREACHED */
1534 return (-1);
1535 }
1536
1537 /*
1538 * Create the "control" list for this pcb.
1539 * These functions will not modify mbuf chain at all.
1540 *
1541 * With KAME mbuf chain restriction:
1542 * The routine will be called from upper layer handlers like tcp6_input().
1543 * Thus the routine assumes that the caller (tcp6_input) have already
1544 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1545 * very first mbuf on the mbuf chain.
1546 *
1547 * ip6_savecontrol_v4 will handle those options that are possible to be
1548 * set on a v4-mapped socket.
1549 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1550 * options and handle the v6-only ones itself.
1551 */
1552 struct mbuf **
1553 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1554 int *v4only)
1555 {
1556 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1557
1558 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1559 struct timeval tv;
1560
1561 getmicrotime(&tv);
1562 mp = sbcreatecontrol_mbuf((caddr_t)&tv, sizeof (tv),
1563 SCM_TIMESTAMP, SOL_SOCKET, mp);
1564 if (*mp == NULL)
1565 return (NULL);
1566 }
1567 if ((inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) {
1568 uint64_t time;
1569
1570 time = mach_absolute_time();
1571 mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof (time),
1572 SCM_TIMESTAMP_MONOTONIC, SOL_SOCKET, mp);
1573 if (*mp == NULL)
1574 return (NULL);
1575 }
1576 if ((inp->inp_socket->so_flags & SOF_RECV_TRAFFIC_CLASS) != 0) {
1577 int tc = m_get_traffic_class(m);
1578
1579 mp = sbcreatecontrol_mbuf((caddr_t)&tc, sizeof (tc),
1580 SO_TRAFFIC_CLASS, SOL_SOCKET, mp);
1581 if (*mp == NULL)
1582 return (NULL);
1583 }
1584
1585 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1586 if (v4only != NULL)
1587 *v4only = 1;
1588 return (mp);
1589 }
1590
1591 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1592 /* RFC 2292 sec. 5 */
1593 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1594 struct in6_pktinfo pi6;
1595
1596 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof (struct in6_addr));
1597 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1598 pi6.ipi6_ifindex =
1599 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1600
1601 mp = sbcreatecontrol_mbuf((caddr_t)&pi6,
1602 sizeof (struct in6_pktinfo),
1603 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO),
1604 IPPROTO_IPV6, mp);
1605 if (*mp == NULL)
1606 return (NULL);
1607 }
1608
1609 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1610 int hlim = ip6->ip6_hlim & 0xff;
1611
1612 mp = sbcreatecontrol_mbuf((caddr_t)&hlim, sizeof (int),
1613 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1614 IPPROTO_IPV6, mp);
1615 if (*mp == NULL)
1616 return (NULL);
1617 }
1618
1619 if (v4only != NULL)
1620 *v4only = 0;
1621 return (mp);
1622 }
1623
1624 int
1625 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1626 {
1627 struct mbuf **np;
1628 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1629 int v4only = 0;
1630
1631 *mp = NULL;
1632 np = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1633 if (np == NULL)
1634 goto no_mbufs;
1635
1636 mp = np;
1637 if (v4only)
1638 return (0);
1639
1640 if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1641 u_int32_t flowinfo;
1642 int tclass;
1643
1644 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1645 flowinfo >>= 20;
1646
1647 tclass = flowinfo & 0xff;
1648 mp = sbcreatecontrol_mbuf((caddr_t)&tclass, sizeof (tclass),
1649 IPV6_TCLASS, IPPROTO_IPV6, mp);
1650 if (*mp == NULL)
1651 goto no_mbufs;
1652 }
1653
1654 /*
1655 * IPV6_HOPOPTS socket option. Recall that we required super-user
1656 * privilege for the option (see ip6_ctloutput), but it might be too
1657 * strict, since there might be some hop-by-hop options which can be
1658 * returned to normal user.
1659 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1660 */
1661 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1662 /*
1663 * Check if a hop-by-hop options header is contatined in the
1664 * received packet, and if so, store the options as ancillary
1665 * data. Note that a hop-by-hop options header must be
1666 * just after the IPv6 header, which is assured through the
1667 * IPv6 input processing.
1668 */
1669 ip6 = mtod(m, struct ip6_hdr *);
1670 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1671 struct ip6_hbh *hbh;
1672 int hbhlen = 0;
1673 #if PULLDOWN_TEST
1674 struct mbuf *ext;
1675 #endif
1676
1677 #ifndef PULLDOWN_TEST
1678 hbh = (struct ip6_hbh *)(ip6 + 1);
1679 hbhlen = (hbh->ip6h_len + 1) << 3;
1680 #else
1681 ext = ip6_pullexthdr(m, sizeof (struct ip6_hdr),
1682 ip6->ip6_nxt);
1683 if (ext == NULL) {
1684 ip6stat.ip6s_tooshort++;
1685 return (0);
1686 }
1687 hbh = mtod(ext, struct ip6_hbh *);
1688 hbhlen = (hbh->ip6h_len + 1) << 3;
1689 if (hbhlen != ext->m_len) {
1690 m_freem(ext);
1691 ip6stat.ip6s_tooshort++;
1692 return (0);
1693 }
1694 #endif
1695
1696 /*
1697 * XXX: We copy the whole header even if a
1698 * jumbo payload option is included, the option which
1699 * is to be removed before returning according to
1700 * RFC2292.
1701 * Note: this constraint is removed in RFC3542
1702 */
1703 mp = sbcreatecontrol_mbuf((caddr_t)hbh, hbhlen,
1704 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1705 IPPROTO_IPV6, mp);
1706
1707 #if PULLDOWN_TEST
1708 m_freem(ext);
1709 #endif
1710 if (*mp == NULL) {
1711 goto no_mbufs;
1712 }
1713 }
1714 }
1715
1716 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1717 int nxt = ip6->ip6_nxt, off = sizeof (struct ip6_hdr);
1718
1719 /*
1720 * Search for destination options headers or routing
1721 * header(s) through the header chain, and stores each
1722 * header as ancillary data.
1723 * Note that the order of the headers remains in
1724 * the chain of ancillary data.
1725 */
1726 while (1) { /* is explicit loop prevention necessary? */
1727 struct ip6_ext *ip6e = NULL;
1728 int elen;
1729 #if PULLDOWN_TEST
1730 struct mbuf *ext = NULL;
1731 #endif
1732
1733 /*
1734 * if it is not an extension header, don't try to
1735 * pull it from the chain.
1736 */
1737 switch (nxt) {
1738 case IPPROTO_DSTOPTS:
1739 case IPPROTO_ROUTING:
1740 case IPPROTO_HOPOPTS:
1741 case IPPROTO_AH: /* is it possible? */
1742 break;
1743 default:
1744 goto loopend;
1745 }
1746
1747 #ifndef PULLDOWN_TEST
1748 if (off + sizeof (*ip6e) > m->m_len)
1749 goto loopend;
1750 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1751 if (nxt == IPPROTO_AH)
1752 elen = (ip6e->ip6e_len + 2) << 2;
1753 else
1754 elen = (ip6e->ip6e_len + 1) << 3;
1755 if (off + elen > m->m_len)
1756 goto loopend;
1757 #else
1758 ext = ip6_pullexthdr(m, off, nxt);
1759 if (ext == NULL) {
1760 ip6stat.ip6s_tooshort++;
1761 return (0);
1762 }
1763 ip6e = mtod(ext, struct ip6_ext *);
1764 if (nxt == IPPROTO_AH)
1765 elen = (ip6e->ip6e_len + 2) << 2;
1766 else
1767 elen = (ip6e->ip6e_len + 1) << 3;
1768 if (elen != ext->m_len) {
1769 m_freem(ext);
1770 ip6stat.ip6s_tooshort++;
1771 return (0);
1772 }
1773 #endif
1774
1775 switch (nxt) {
1776 case IPPROTO_DSTOPTS:
1777 if (!(in6p->inp_flags & IN6P_DSTOPTS))
1778 break;
1779
1780 mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
1781 IS2292(in6p, IPV6_2292DSTOPTS,
1782 IPV6_DSTOPTS), IPPROTO_IPV6, mp);
1783 if (*mp == NULL) {
1784 #if PULLDOWN_TEST
1785 m_freem(ext);
1786 #endif
1787 goto no_mbufs;
1788 }
1789 break;
1790 case IPPROTO_ROUTING:
1791 if (!in6p->inp_flags & IN6P_RTHDR)
1792 break;
1793
1794 mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
1795 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1796 IPPROTO_IPV6, mp);
1797 if (*mp == NULL) {
1798 #if PULLDOWN_TEST
1799 m_freem(ext);
1800 #endif
1801 goto no_mbufs;
1802 }
1803 break;
1804 case IPPROTO_HOPOPTS:
1805 case IPPROTO_AH: /* is it possible? */
1806 break;
1807
1808 default:
1809 /*
1810 * other cases have been filtered in the above.
1811 * none will visit this case. here we supply
1812 * the code just in case (nxt overwritten or
1813 * other cases).
1814 */
1815 #if PULLDOWN_TEST
1816 m_freem(ext);
1817 #endif
1818 goto loopend;
1819
1820 }
1821
1822 /* proceed with the next header. */
1823 off += elen;
1824 nxt = ip6e->ip6e_nxt;
1825 ip6e = NULL;
1826 #if PULLDOWN_TEST
1827 m_freem(ext);
1828 ext = NULL;
1829 #endif
1830 }
1831 loopend:
1832 ;
1833 }
1834 return (0);
1835 no_mbufs:
1836 ip6stat.ip6s_pktdropcntrl++;
1837 /* XXX increment a stat to show the failure */
1838 return (ENOBUFS);
1839 }
1840 #undef IS2292
1841
1842 void
1843 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1844 {
1845 struct socket *so;
1846 struct mbuf *m_mtu;
1847 struct ip6_mtuinfo mtuctl;
1848
1849 so = in6p->inp_socket;
1850
1851 if (mtu == NULL)
1852 return;
1853
1854 #ifdef DIAGNOSTIC
1855 if (so == NULL) { /* I believe this is impossible */
1856 panic("ip6_notify_pmtu: socket is NULL");
1857 /* NOTREACHED */
1858 }
1859 #endif
1860
1861 bzero(&mtuctl, sizeof (mtuctl)); /* zero-clear for safety */
1862 mtuctl.ip6m_mtu = *mtu;
1863 mtuctl.ip6m_addr = *dst;
1864 if (sa6_recoverscope(&mtuctl.ip6m_addr, TRUE))
1865 return;
1866
1867 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof (mtuctl),
1868 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1869 return;
1870
1871 if (sbappendaddr(&so->so_rcv, SA(dst), NULL, m_mtu, NULL) == 0) {
1872 m_freem(m_mtu);
1873 /* XXX: should count statistics */
1874 } else {
1875 sorwakeup(so);
1876 }
1877 }
1878
1879 #if PULLDOWN_TEST
1880 /*
1881 * pull single extension header from mbuf chain. returns single mbuf that
1882 * contains the result, or NULL on error.
1883 */
1884 static struct mbuf *
1885 ip6_pullexthdr(m, off, nxt)
1886 struct mbuf *m;
1887 size_t off;
1888 int nxt;
1889 {
1890 struct ip6_ext ip6e;
1891 size_t elen;
1892 struct mbuf *n;
1893
1894 #if DIAGNOSTIC
1895 switch (nxt) {
1896 case IPPROTO_DSTOPTS:
1897 case IPPROTO_ROUTING:
1898 case IPPROTO_HOPOPTS:
1899 case IPPROTO_AH: /* is it possible? */
1900 break;
1901 default:
1902 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1903 }
1904 #endif
1905
1906 m_copydata(m, off, sizeof (ip6e), (caddr_t)&ip6e);
1907 if (nxt == IPPROTO_AH)
1908 elen = (ip6e.ip6e_len + 2) << 2;
1909 else
1910 elen = (ip6e.ip6e_len + 1) << 3;
1911
1912 MGET(n, M_DONTWAIT, MT_DATA);
1913 if (n && elen >= MLEN) {
1914 MCLGET(n, M_DONTWAIT);
1915 if ((n->m_flags & M_EXT) == 0) {
1916 m_free(n);
1917 n = NULL;
1918 }
1919 }
1920 if (!n)
1921 return (NULL);
1922
1923 n->m_len = 0;
1924 if (elen >= M_TRAILINGSPACE(n)) {
1925 m_free(n);
1926 return (NULL);
1927 }
1928
1929 m_copydata(m, off, elen, mtod(n, caddr_t));
1930 n->m_len = elen;
1931 return (n);
1932 }
1933 #endif
1934
1935 /*
1936 * Get pointer to the previous header followed by the header
1937 * currently processed.
1938 * XXX: This function supposes that
1939 * M includes all headers,
1940 * the next header field and the header length field of each header
1941 * are valid, and
1942 * the sum of each header length equals to OFF.
1943 * Because of these assumptions, this function must be called very
1944 * carefully. Moreover, it will not be used in the near future when
1945 * we develop `neater' mechanism to process extension headers.
1946 */
1947 char *
1948 ip6_get_prevhdr(m, off)
1949 struct mbuf *m;
1950 int off;
1951 {
1952 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1953
1954 if (off == sizeof (struct ip6_hdr)) {
1955 return ((char *)&ip6->ip6_nxt);
1956 } else {
1957 int len, nxt;
1958 struct ip6_ext *ip6e = NULL;
1959
1960 nxt = ip6->ip6_nxt;
1961 len = sizeof (struct ip6_hdr);
1962 while (len < off) {
1963 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1964
1965 switch (nxt) {
1966 case IPPROTO_FRAGMENT:
1967 len += sizeof (struct ip6_frag);
1968 break;
1969 case IPPROTO_AH:
1970 len += (ip6e->ip6e_len + 2) << 2;
1971 break;
1972 default:
1973 len += (ip6e->ip6e_len + 1) << 3;
1974 break;
1975 }
1976 nxt = ip6e->ip6e_nxt;
1977 }
1978 if (ip6e)
1979 return ((char *)&ip6e->ip6e_nxt);
1980 else
1981 return (NULL);
1982 }
1983 }
1984
1985 /*
1986 * get next header offset. m will be retained.
1987 */
1988 int
1989 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1990 {
1991 struct ip6_hdr ip6;
1992 struct ip6_ext ip6e;
1993 struct ip6_frag fh;
1994
1995 /* just in case */
1996 VERIFY(m != NULL);
1997 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1998 return (-1);
1999
2000 switch (proto) {
2001 case IPPROTO_IPV6:
2002 if (m->m_pkthdr.len < off + sizeof (ip6))
2003 return (-1);
2004 m_copydata(m, off, sizeof (ip6), (caddr_t)&ip6);
2005 if (nxtp)
2006 *nxtp = ip6.ip6_nxt;
2007 off += sizeof (ip6);
2008 return (off);
2009
2010 case IPPROTO_FRAGMENT:
2011 /*
2012 * terminate parsing if it is not the first fragment,
2013 * it does not make sense to parse through it.
2014 */
2015 if (m->m_pkthdr.len < off + sizeof (fh))
2016 return (-1);
2017 m_copydata(m, off, sizeof (fh), (caddr_t)&fh);
2018 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
2019 if (fh.ip6f_offlg & IP6F_OFF_MASK)
2020 return (-1);
2021 if (nxtp)
2022 *nxtp = fh.ip6f_nxt;
2023 off += sizeof (struct ip6_frag);
2024 return (off);
2025
2026 case IPPROTO_AH:
2027 if (m->m_pkthdr.len < off + sizeof (ip6e))
2028 return (-1);
2029 m_copydata(m, off, sizeof (ip6e), (caddr_t)&ip6e);
2030 if (nxtp)
2031 *nxtp = ip6e.ip6e_nxt;
2032 off += (ip6e.ip6e_len + 2) << 2;
2033 return (off);
2034
2035 case IPPROTO_HOPOPTS:
2036 case IPPROTO_ROUTING:
2037 case IPPROTO_DSTOPTS:
2038 if (m->m_pkthdr.len < off + sizeof (ip6e))
2039 return (-1);
2040 m_copydata(m, off, sizeof (ip6e), (caddr_t)&ip6e);
2041 if (nxtp)
2042 *nxtp = ip6e.ip6e_nxt;
2043 off += (ip6e.ip6e_len + 1) << 3;
2044 return (off);
2045
2046 case IPPROTO_NONE:
2047 case IPPROTO_ESP:
2048 case IPPROTO_IPCOMP:
2049 /* give up */
2050 return (-1);
2051
2052 default:
2053 return (-1);
2054 }
2055
2056 return (-1);
2057 }
2058
2059 /*
2060 * get offset for the last header in the chain. m will be kept untainted.
2061 */
2062 int
2063 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
2064 {
2065 int newoff;
2066 int nxt;
2067
2068 if (!nxtp) {
2069 nxt = -1;
2070 nxtp = &nxt;
2071 }
2072 while (1) {
2073 newoff = ip6_nexthdr(m, off, proto, nxtp);
2074 if (newoff < 0)
2075 return (off);
2076 else if (newoff < off)
2077 return (-1); /* invalid */
2078 else if (newoff == off)
2079 return (newoff);
2080
2081 off = newoff;
2082 proto = *nxtp;
2083 }
2084 }
2085
2086 struct ip6aux *
2087 ip6_addaux(struct mbuf *m)
2088 {
2089 struct m_tag *tag;
2090
2091 /* Check if one is already allocated */
2092 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2093 KERNEL_TAG_TYPE_INET6, NULL);
2094 if (tag == NULL) {
2095 /* Allocate a tag */
2096 tag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6,
2097 sizeof (struct ip6aux), M_DONTWAIT, m);
2098
2099 /* Attach it to the mbuf */
2100 if (tag) {
2101 m_tag_prepend(m, tag);
2102 }
2103 }
2104
2105 return (tag ? (struct ip6aux *)(tag + 1) : NULL);
2106 }
2107
2108 struct ip6aux *
2109 ip6_findaux(struct mbuf *m)
2110 {
2111 struct m_tag *tag;
2112
2113 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2114 KERNEL_TAG_TYPE_INET6, NULL);
2115
2116 return (tag ? (struct ip6aux *)(tag + 1) : NULL);
2117 }
2118
2119 void
2120 ip6_delaux(struct mbuf *m)
2121 {
2122 struct m_tag *tag;
2123
2124 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2125 KERNEL_TAG_TYPE_INET6, NULL);
2126 if (tag) {
2127 m_tag_delete(m, tag);
2128 }
2129 }
2130
2131 /*
2132 * Drain callback
2133 */
2134 void
2135 ip6_drain(void)
2136 {
2137 frag6_drain(); /* fragments */
2138 in6_rtqdrain(); /* protocol cloned routes */
2139 nd6_drain(NULL); /* cloned routes: ND6 */
2140 }
2141
2142 /*
2143 * System control for IP6
2144 */
2145
2146 u_char inet6ctlerrmap[PRC_NCMDS] = {
2147 0, 0, 0, 0,
2148 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
2149 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
2150 EMSGSIZE, EHOSTUNREACH, 0, 0,
2151 0, 0, 0, 0,
2152 ENOPROTOOPT
2153 };