1 /*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $
30 */
31
32 /*-
33 * Copyright (c) 1982, 1986, 1988, 1993
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
61 */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD: releng/9.0/sys/netinet6/ip6_input.c 236953 2012-06-12 12:10:10Z bz $");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_route.h"
71
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/proc.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/syslog.h>
85
86 #include <net/if.h>
87 #include <net/if_types.h>
88 #include <net/if_dl.h>
89 #include <net/route.h>
90 #include <net/netisr.h>
91 #include <net/pfil.h>
92 #include <net/vnet.h>
93
94 #include <netinet/in.h>
95 #include <netinet/ip_var.h>
96 #include <netinet/in_systm.h>
97 #include <net/if_llatbl.h>
98 #ifdef INET
99 #include <netinet/ip.h>
100 #include <netinet/ip_icmp.h>
101 #endif /* INET */
102 #include <netinet/ip6.h>
103 #include <netinet6/in6_var.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/scope6_var.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/nd6.h>
110
111 #ifdef IPSEC
112 #include <netipsec/ipsec.h>
113 #include <netinet6/ip6_ipsec.h>
114 #include <netipsec/ipsec6.h>
115 #endif /* IPSEC */
116
117 #include <netinet6/ip6protosw.h>
118
119 #ifdef FLOWTABLE
120 #include <net/flowtable.h>
121 VNET_DECLARE(int, ip6_output_flowtable_size);
122 #define V_ip6_output_flowtable_size VNET(ip6_output_flowtable_size)
123 #endif
124
125 extern struct domain inet6domain;
126
127 u_char ip6_protox[IPPROTO_MAX];
128 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
129
130 static struct netisr_handler ip6_nh = {
131 .nh_name = "ip6",
132 .nh_handler = ip6_input,
133 .nh_proto = NETISR_IPV6,
134 .nh_policy = NETISR_POLICY_FLOW,
135 };
136
137 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
138 #define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch)
139
140 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
141
142 VNET_DEFINE(struct ip6stat, ip6stat);
143
144 struct rwlock in6_ifaddr_lock;
145 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
146
147 static void ip6_init2(void *);
148 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
149 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
150 #ifdef PULLDOWN_TEST
151 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
152 #endif
153
154 /*
155 * IP6 initialization: fill in IP6 protocol switch table.
156 * All protocols not implemented in kernel go to raw IP6 protocol handler.
157 */
158 void
159 ip6_init(void)
160 {
161 struct ip6protosw *pr;
162 int i;
163
164 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
165 &V_ip6_auto_linklocal);
166
167 TAILQ_INIT(&V_in6_ifaddrhead);
168
169 /* Initialize packet filter hooks. */
170 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
171 V_inet6_pfil_hook.ph_af = AF_INET6;
172 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
173 printf("%s: WARNING: unable to register pfil hook, "
174 "error %d\n", __func__, i);
175
176 scope6_init();
177 addrsel_policy_init();
178 nd6_init();
179 frag6_init();
180
181 #ifdef FLOWTABLE
182 if (TUNABLE_INT_FETCH("net.inet6.ip6.output_flowtable_size",
183 &V_ip6_output_flowtable_size)) {
184 if (V_ip6_output_flowtable_size < 256)
185 V_ip6_output_flowtable_size = 256;
186 if (!powerof2(V_ip6_output_flowtable_size)) {
187 printf("flowtable must be power of 2 size\n");
188 V_ip6_output_flowtable_size = 2048;
189 }
190 } else {
191 /*
192 * round up to the next power of 2
193 */
194 V_ip6_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1);
195 }
196 V_ip6_ft = flowtable_alloc("ipv6", V_ip6_output_flowtable_size, FL_IPV6|FL_PCPU);
197 #endif
198
199 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
200
201 /* Skip global initialization stuff for non-default instances. */
202 if (!IS_DEFAULT_VNET(curvnet))
203 return;
204
205 #ifdef DIAGNOSTIC
206 if (sizeof(struct protosw) != sizeof(struct ip6protosw))
207 panic("sizeof(protosw) != sizeof(ip6protosw)");
208 #endif
209 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
210 if (pr == NULL)
211 panic("ip6_init");
212
213 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
214 for (i = 0; i < IPPROTO_MAX; i++)
215 ip6_protox[i] = pr - inet6sw;
216 /*
217 * Cycle through IP protocols and put them into the appropriate place
218 * in ip6_protox[].
219 */
220 for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
221 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
222 if (pr->pr_domain->dom_family == PF_INET6 &&
223 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
224 /* Be careful to only index valid IP protocols. */
225 if (pr->pr_protocol < IPPROTO_MAX)
226 ip6_protox[pr->pr_protocol] = pr - inet6sw;
227 }
228
229 netisr_register(&ip6_nh);
230 }
231
232 /*
233 * The protocol to be inserted into ip6_protox[] must be already registered
234 * in inet6sw[], either statically or through pf_proto_register().
235 */
236 int
237 ip6proto_register(short ip6proto)
238 {
239 struct ip6protosw *pr;
240
241 /* Sanity checks. */
242 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
243 return (EPROTONOSUPPORT);
244
245 /*
246 * The protocol slot must not be occupied by another protocol
247 * already. An index pointing to IPPROTO_RAW is unused.
248 */
249 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
250 if (pr == NULL)
251 return (EPFNOSUPPORT);
252 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */
253 return (EEXIST);
254
255 /*
256 * Find the protocol position in inet6sw[] and set the index.
257 */
258 for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
259 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
260 if (pr->pr_domain->dom_family == PF_INET6 &&
261 pr->pr_protocol && pr->pr_protocol == ip6proto) {
262 ip6_protox[pr->pr_protocol] = pr - inet6sw;
263 return (0);
264 }
265 }
266 return (EPROTONOSUPPORT);
267 }
268
269 int
270 ip6proto_unregister(short ip6proto)
271 {
272 struct ip6protosw *pr;
273
274 /* Sanity checks. */
275 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
276 return (EPROTONOSUPPORT);
277
278 /* Check if the protocol was indeed registered. */
279 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
280 if (pr == NULL)
281 return (EPFNOSUPPORT);
282 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */
283 return (ENOENT);
284
285 /* Reset the protocol slot to IPPROTO_RAW. */
286 ip6_protox[ip6proto] = pr - inet6sw;
287 return (0);
288 }
289
290 #ifdef VIMAGE
291 void
292 ip6_destroy()
293 {
294
295 nd6_destroy();
296 callout_drain(&V_in6_tmpaddrtimer_ch);
297 }
298 #endif
299
300 static int
301 ip6_init2_vnet(const void *unused __unused)
302 {
303
304 /* nd6_timer_init */
305 callout_init(&V_nd6_timer_ch, 0);
306 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
307
308 /* timer for regeneranation of temporary addresses randomize ID */
309 callout_init(&V_in6_tmpaddrtimer_ch, 0);
310 callout_reset(&V_in6_tmpaddrtimer_ch,
311 (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
312 V_ip6_temp_regen_advance) * hz,
313 in6_tmpaddrtimer, curvnet);
314
315 return (0);
316 }
317
318 static void
319 ip6_init2(void *dummy)
320 {
321
322 ip6_init2_vnet(NULL);
323 }
324
325 /* cheat */
326 /* This must be after route_init(), which is now SI_ORDER_THIRD */
327 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
328
329 void
330 ip6_input(struct mbuf *m)
331 {
332 struct ip6_hdr *ip6;
333 int off = sizeof(struct ip6_hdr), nest;
334 u_int32_t plen;
335 u_int32_t rtalert = ~0;
336 int nxt, ours = 0;
337 struct ifnet *deliverifp = NULL, *ifp = NULL;
338 struct in6_addr odst;
339 struct route_in6 rin6;
340 int srcrt = 0;
341 struct llentry *lle = NULL;
342 struct sockaddr_in6 dst6, *dst;
343
344 bzero(&rin6, sizeof(struct route_in6));
345 #ifdef IPSEC
346 /*
347 * should the inner packet be considered authentic?
348 * see comment in ah4_input().
349 * NB: m cannot be NULL when passed to the input routine
350 */
351
352 m->m_flags &= ~M_AUTHIPHDR;
353 m->m_flags &= ~M_AUTHIPDGM;
354
355 #endif /* IPSEC */
356
357 /*
358 * make sure we don't have onion peering information into m_tag.
359 */
360 ip6_delaux(m);
361
362 if (m->m_flags & M_FASTFWD_OURS) {
363 /*
364 * Firewall changed destination to local.
365 */
366 m->m_flags &= ~M_FASTFWD_OURS;
367 ours = 1;
368 deliverifp = m->m_pkthdr.rcvif;
369 ip6 = mtod(m, struct ip6_hdr *);
370 goto hbhcheck;
371 }
372
373 /*
374 * mbuf statistics
375 */
376 if (m->m_flags & M_EXT) {
377 if (m->m_next)
378 V_ip6stat.ip6s_mext2m++;
379 else
380 V_ip6stat.ip6s_mext1++;
381 } else {
382 #define M2MMAX (sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
383 if (m->m_next) {
384 if (m->m_flags & M_LOOP) {
385 V_ip6stat.ip6s_m2m[V_loif->if_index]++;
386 } else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
387 V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
388 else
389 V_ip6stat.ip6s_m2m[0]++;
390 } else
391 V_ip6stat.ip6s_m1++;
392 #undef M2MMAX
393 }
394
395 /* drop the packet if IPv6 operation is disabled on the IF */
396 if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
397 m_freem(m);
398 return;
399 }
400
401 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
402 V_ip6stat.ip6s_total++;
403
404 #ifndef PULLDOWN_TEST
405 /*
406 * L2 bridge code and some other code can return mbuf chain
407 * that does not conform to KAME requirement. too bad.
408 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
409 */
410 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
411 struct mbuf *n;
412
413 MGETHDR(n, M_DONTWAIT, MT_HEADER);
414 if (n)
415 M_MOVE_PKTHDR(n, m);
416 if (n && n->m_pkthdr.len > MHLEN) {
417 MCLGET(n, M_DONTWAIT);
418 if ((n->m_flags & M_EXT) == 0) {
419 m_freem(n);
420 n = NULL;
421 }
422 }
423 if (n == NULL) {
424 m_freem(m);
425 return; /* ENOBUFS */
426 }
427
428 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
429 n->m_len = n->m_pkthdr.len;
430 m_freem(m);
431 m = n;
432 }
433 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
434 #endif
435
436 if (m->m_len < sizeof(struct ip6_hdr)) {
437 struct ifnet *inifp;
438 inifp = m->m_pkthdr.rcvif;
439 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
440 V_ip6stat.ip6s_toosmall++;
441 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
442 return;
443 }
444 }
445
446 ip6 = mtod(m, struct ip6_hdr *);
447
448 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
449 V_ip6stat.ip6s_badvers++;
450 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
451 goto bad;
452 }
453
454 V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
455
456 /*
457 * Check against address spoofing/corruption.
458 */
459 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
460 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
461 /*
462 * XXX: "badscope" is not very suitable for a multicast source.
463 */
464 V_ip6stat.ip6s_badscope++;
465 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
466 goto bad;
467 }
468 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
469 !(m->m_flags & M_LOOP)) {
470 /*
471 * In this case, the packet should come from the loopback
472 * interface. However, we cannot just check the if_flags,
473 * because ip6_mloopback() passes the "actual" interface
474 * as the outgoing/incoming interface.
475 */
476 V_ip6stat.ip6s_badscope++;
477 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
478 goto bad;
479 }
480
481 #ifdef ALTQ
482 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
483 /* packet is dropped by traffic conditioner */
484 return;
485 }
486 #endif
487 /*
488 * The following check is not documented in specs. A malicious
489 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
490 * and bypass security checks (act as if it was from 127.0.0.1 by using
491 * IPv6 src ::ffff:127.0.0.1). Be cautious.
492 *
493 * This check chokes if we are in an SIIT cloud. As none of BSDs
494 * support IPv4-less kernel compilation, we cannot support SIIT
495 * environment at all. So, it makes more sense for us to reject any
496 * malicious packets for non-SIIT environment, than try to do a
497 * partial support for SIIT environment.
498 */
499 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
500 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
501 V_ip6stat.ip6s_badscope++;
502 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
503 goto bad;
504 }
505 #if 0
506 /*
507 * Reject packets with IPv4 compatible addresses (auto tunnel).
508 *
509 * The code forbids auto tunnel relay case in RFC1933 (the check is
510 * stronger than RFC1933). We may want to re-enable it if mech-xx
511 * is revised to forbid relaying case.
512 */
513 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
514 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
515 V_ip6stat.ip6s_badscope++;
516 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
517 goto bad;
518 }
519 #endif
520 #ifdef IPSEC
521 /*
522 * Bypass packet filtering for packets previously handled by IPsec.
523 */
524 if (ip6_ipsec_filtertunnel(m))
525 goto passin;
526 #endif /* IPSEC */
527
528 /*
529 * Run through list of hooks for input packets.
530 *
531 * NB: Beware of the destination address changing
532 * (e.g. by NAT rewriting). When this happens,
533 * tell ip6_forward to do the right thing.
534 */
535 odst = ip6->ip6_dst;
536
537 /* Jump over all PFIL processing if hooks are not active. */
538 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
539 goto passin;
540
541 if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
542 m->m_pkthdr.rcvif, PFIL_IN, NULL))
543 return;
544 if (m == NULL) /* consumed by filter */
545 return;
546 ip6 = mtod(m, struct ip6_hdr *);
547 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
548
549 #ifdef IPFIREWALL_FORWARD
550 if (m->m_flags & M_FASTFWD_OURS) {
551 m->m_flags &= ~M_FASTFWD_OURS;
552 ours = 1;
553 deliverifp = m->m_pkthdr.rcvif;
554 goto hbhcheck;
555 }
556 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
557 /*
558 * Directly ship the packet on. This allows forwarding
559 * packets originally destined to us to some other directly
560 * connected host.
561 */
562 ip6_forward(m, 1);
563 goto out;
564 }
565 #endif /* IPFIREWALL_FORWARD */
566
567 passin:
568 /*
569 * Disambiguate address scope zones (if there is ambiguity).
570 * We first make sure that the original source or destination address
571 * is not in our internal form for scoped addresses. Such addresses
572 * are not necessarily invalid spec-wise, but we cannot accept them due
573 * to the usage conflict.
574 * in6_setscope() then also checks and rejects the cases where src or
575 * dst are the loopback address and the receiving interface
576 * is not loopback.
577 */
578 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
579 V_ip6stat.ip6s_badscope++; /* XXX */
580 goto bad;
581 }
582 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
583 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
584 V_ip6stat.ip6s_badscope++;
585 goto bad;
586 }
587
588 /*
589 * Multicast check. Assume packet is for us to avoid
590 * prematurely taking locks.
591 */
592 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
593 ours = 1;
594 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
595 deliverifp = m->m_pkthdr.rcvif;
596 goto hbhcheck;
597 }
598
599 /*
600 * Unicast check
601 */
602
603 bzero(&dst6, sizeof(dst6));
604 dst6.sin6_family = AF_INET6;
605 dst6.sin6_len = sizeof(struct sockaddr_in6);
606 dst6.sin6_addr = ip6->ip6_dst;
607 ifp = m->m_pkthdr.rcvif;
608 IF_AFDATA_LOCK(ifp);
609 lle = lla_lookup(LLTABLE6(ifp), 0,
610 (struct sockaddr *)&dst6);
611 IF_AFDATA_UNLOCK(ifp);
612 if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
613 struct ifaddr *ifa;
614 struct in6_ifaddr *ia6;
615 int bad;
616
617 bad = 1;
618 #define sa_equal(a1, a2) \
619 (bcmp((a1), (a2), ((a1))->sin6_len) == 0)
620 IF_ADDR_LOCK(ifp);
621 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
622 if (ifa->ifa_addr->sa_family != dst6.sin6_family)
623 continue;
624 if (sa_equal(&dst6, ifa->ifa_addr))
625 break;
626 }
627 KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
628 __func__, lle));
629 #undef sa_equal
630
631 ia6 = (struct in6_ifaddr *)ifa;
632 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
633 /* Count the packet in the ip address stats */
634 ia6->ia_ifa.if_ipackets++;
635 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
636
637 /*
638 * record address information into m_tag.
639 */
640 (void)ip6_setdstifaddr(m, ia6);
641
642 bad = 0;
643 } else {
644 char ip6bufs[INET6_ADDRSTRLEN];
645 char ip6bufd[INET6_ADDRSTRLEN];
646 /* address is not ready, so discard the packet. */
647 nd6log((LOG_INFO,
648 "ip6_input: packet to an unready address %s->%s\n",
649 ip6_sprintf(ip6bufs, &ip6->ip6_src),
650 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
651 }
652 IF_ADDR_UNLOCK(ifp);
653 LLE_RUNLOCK(lle);
654 if (bad)
655 goto bad;
656 else {
657 ours = 1;
658 deliverifp = ifp;
659 goto hbhcheck;
660 }
661 }
662 if (lle != NULL)
663 LLE_RUNLOCK(lle);
664
665 dst = &rin6.ro_dst;
666 dst->sin6_len = sizeof(struct sockaddr_in6);
667 dst->sin6_family = AF_INET6;
668 dst->sin6_addr = ip6->ip6_dst;
669 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
670 if (rin6.ro_rt)
671 RT_UNLOCK(rin6.ro_rt);
672
673 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
674
675 /*
676 * Accept the packet if the forwarding interface to the destination
677 * according to the routing table is the loopback interface,
678 * unless the associated route has a gateway.
679 * Note that this approach causes to accept a packet if there is a
680 * route to the loopback interface for the destination of the packet.
681 * But we think it's even useful in some situations, e.g. when using
682 * a special daemon which wants to intercept the packet.
683 *
684 * XXX: some OSes automatically make a cloned route for the destination
685 * of an outgoing packet. If the outgoing interface of the packet
686 * is a loopback one, the kernel would consider the packet to be
687 * accepted, even if we have no such address assinged on the interface.
688 * We check the cloned flag of the route entry to reject such cases,
689 * assuming that route entries for our own addresses are not made by
690 * cloning (it should be true because in6_addloop explicitly installs
691 * the host route). However, we might have to do an explicit check
692 * while it would be less efficient. Or, should we rather install a
693 * reject route for such a case?
694 */
695 if (rin6.ro_rt &&
696 (rin6.ro_rt->rt_flags &
697 (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
698 #ifdef RTF_WASCLONED
699 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
700 #endif
701 #ifdef RTF_CLONED
702 !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
703 #endif
704 #if 0
705 /*
706 * The check below is redundant since the comparison of
707 * the destination and the key of the rtentry has
708 * already done through looking up the routing table.
709 */
710 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
711 &rt6_key(rin6.ro_rt)->sin6_addr)
712 #endif
713 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
714 int free_ia6 = 0;
715 struct in6_ifaddr *ia6;
716
717 /*
718 * found the loopback route to the interface address
719 */
720 if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
721 struct sockaddr_in6 dest6;
722
723 bzero(&dest6, sizeof(dest6));
724 dest6.sin6_family = AF_INET6;
725 dest6.sin6_len = sizeof(dest6);
726 dest6.sin6_addr = ip6->ip6_dst;
727 ia6 = (struct in6_ifaddr *)
728 ifa_ifwithaddr((struct sockaddr *)&dest6);
729 if (ia6 == NULL)
730 goto bad;
731 free_ia6 = 1;
732 }
733 else
734 ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
735
736 /*
737 * record address information into m_tag.
738 */
739 (void)ip6_setdstifaddr(m, ia6);
740
741 /*
742 * packets to a tentative, duplicated, or somehow invalid
743 * address must not be accepted.
744 */
745 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
746 /* this address is ready */
747 ours = 1;
748 deliverifp = ia6->ia_ifp; /* correct? */
749 /* Count the packet in the ip address stats */
750 ia6->ia_ifa.if_ipackets++;
751 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
752 if (ia6 != NULL && free_ia6 != 0)
753 ifa_free(&ia6->ia_ifa);
754 goto hbhcheck;
755 } else {
756 char ip6bufs[INET6_ADDRSTRLEN];
757 char ip6bufd[INET6_ADDRSTRLEN];
758 /* address is not ready, so discard the packet. */
759 nd6log((LOG_INFO,
760 "ip6_input: packet to an unready address %s->%s\n",
761 ip6_sprintf(ip6bufs, &ip6->ip6_src),
762 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
763
764 if (ia6 != NULL && free_ia6 != 0)
765 ifa_free(&ia6->ia_ifa);
766 goto bad;
767 }
768 }
769
770 /*
771 * FAITH (Firewall Aided Internet Translator)
772 */
773 if (V_ip6_keepfaith) {
774 if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
775 rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
776 /* XXX do we need more sanity checks? */
777 ours = 1;
778 deliverifp = rin6.ro_rt->rt_ifp; /* faith */
779 goto hbhcheck;
780 }
781 }
782
783 /*
784 * Now there is no reason to process the packet if it's not our own
785 * and we're not a router.
786 */
787 if (!V_ip6_forwarding) {
788 V_ip6stat.ip6s_cantforward++;
789 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
790 goto bad;
791 }
792
793 hbhcheck:
794 /*
795 * record address information into m_tag, if we don't have one yet.
796 * note that we are unable to record it, if the address is not listed
797 * as our interface address (e.g. multicast addresses, addresses
798 * within FAITH prefixes and such).
799 */
800 if (deliverifp) {
801 struct in6_ifaddr *ia6;
802
803 if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
804 ifa_free(&ia6->ia_ifa);
805 } else {
806 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
807 if (ia6) {
808 if (!ip6_setdstifaddr(m, ia6)) {
809 /*
810 * XXX maybe we should drop the packet here,
811 * as we could not provide enough information
812 * to the upper layers.
813 */
814 }
815 ifa_free(&ia6->ia_ifa);
816 }
817 }
818 }
819
820 /*
821 * Process Hop-by-Hop options header if it's contained.
822 * m may be modified in ip6_hopopts_input().
823 * If a JumboPayload option is included, plen will also be modified.
824 */
825 plen = (u_int32_t)ntohs(ip6->ip6_plen);
826 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
827 struct ip6_hbh *hbh;
828
829 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
830 #if 0 /*touches NULL pointer*/
831 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
832 #endif
833 goto out; /* m have already been freed */
834 }
835
836 /* adjust pointer */
837 ip6 = mtod(m, struct ip6_hdr *);
838
839 /*
840 * if the payload length field is 0 and the next header field
841 * indicates Hop-by-Hop Options header, then a Jumbo Payload
842 * option MUST be included.
843 */
844 if (ip6->ip6_plen == 0 && plen == 0) {
845 /*
846 * Note that if a valid jumbo payload option is
847 * contained, ip6_hopopts_input() must set a valid
848 * (non-zero) payload length to the variable plen.
849 */
850 V_ip6stat.ip6s_badoptions++;
851 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
852 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
853 icmp6_error(m, ICMP6_PARAM_PROB,
854 ICMP6_PARAMPROB_HEADER,
855 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
856 goto out;
857 }
858 #ifndef PULLDOWN_TEST
859 /* ip6_hopopts_input() ensures that mbuf is contiguous */
860 hbh = (struct ip6_hbh *)(ip6 + 1);
861 #else
862 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
863 sizeof(struct ip6_hbh));
864 if (hbh == NULL) {
865 V_ip6stat.ip6s_tooshort++;
866 goto out;
867 }
868 #endif
869 nxt = hbh->ip6h_nxt;
870
871 /*
872 * If we are acting as a router and the packet contains a
873 * router alert option, see if we know the option value.
874 * Currently, we only support the option value for MLD, in which
875 * case we should pass the packet to the multicast routing
876 * daemon.
877 */
878 if (rtalert != ~0) {
879 switch (rtalert) {
880 case IP6OPT_RTALERT_MLD:
881 if (V_ip6_forwarding)
882 ours = 1;
883 break;
884 default:
885 /*
886 * RFC2711 requires unrecognized values must be
887 * silently ignored.
888 */
889 break;
890 }
891 }
892 } else
893 nxt = ip6->ip6_nxt;
894
895 /*
896 * Check that the amount of data in the buffers
897 * is as at least much as the IPv6 header would have us expect.
898 * Trim mbufs if longer than we expect.
899 * Drop packet if shorter than we expect.
900 */
901 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
902 V_ip6stat.ip6s_tooshort++;
903 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
904 goto bad;
905 }
906 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
907 if (m->m_len == m->m_pkthdr.len) {
908 m->m_len = sizeof(struct ip6_hdr) + plen;
909 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
910 } else
911 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
912 }
913
914 /*
915 * Forward if desirable.
916 */
917 if (V_ip6_mrouter &&
918 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
919 /*
920 * If we are acting as a multicast router, all
921 * incoming multicast packets are passed to the
922 * kernel-level multicast forwarding function.
923 * The packet is returned (relatively) intact; if
924 * ip6_mforward() returns a non-zero value, the packet
925 * must be discarded, else it may be accepted below.
926 *
927 * XXX TODO: Check hlim and multicast scope here to avoid
928 * unnecessarily calling into ip6_mforward().
929 */
930 if (ip6_mforward &&
931 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
932 IP6STAT_INC(ip6s_cantforward);
933 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
934 goto bad;
935 }
936 } else if (!ours) {
937 ip6_forward(m, srcrt);
938 goto out;
939 }
940
941 ip6 = mtod(m, struct ip6_hdr *);
942
943 /*
944 * Malicious party may be able to use IPv4 mapped addr to confuse
945 * tcp/udp stack and bypass security checks (act as if it was from
946 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious.
947 *
948 * For SIIT end node behavior, you may want to disable the check.
949 * However, you will become vulnerable to attacks using IPv4 mapped
950 * source.
951 */
952 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
953 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
954 V_ip6stat.ip6s_badscope++;
955 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
956 goto bad;
957 }
958
959 /*
960 * Tell launch routine the next header
961 */
962 V_ip6stat.ip6s_delivered++;
963 in6_ifstat_inc(deliverifp, ifs6_in_deliver);
964 nest = 0;
965
966 while (nxt != IPPROTO_DONE) {
967 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
968 V_ip6stat.ip6s_toomanyhdr++;
969 goto bad;
970 }
971
972 /*
973 * protection against faulty packet - there should be
974 * more sanity checks in header chain processing.
975 */
976 if (m->m_pkthdr.len < off) {
977 V_ip6stat.ip6s_tooshort++;
978 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
979 goto bad;
980 }
981
982 #ifdef IPSEC
983 /*
984 * enforce IPsec policy checking if we are seeing last header.
985 * note that we do not visit this with protocols with pcb layer
986 * code - like udp/tcp/raw ip.
987 */
988 if (ip6_ipsec_input(m, nxt))
989 goto bad;
990 #endif /* IPSEC */
991
992 /*
993 * Use mbuf flags to propagate Router Alert option to
994 * ICMPv6 layer, as hop-by-hop options have been stripped.
995 */
996 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
997 m->m_flags |= M_RTALERT_MLD;
998
999 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1000 }
1001 goto out;
1002 bad:
1003 m_freem(m);
1004 out:
1005 if (rin6.ro_rt)
1006 RTFREE(rin6.ro_rt);
1007 }
1008
1009 /*
1010 * set/grab in6_ifaddr correspond to IPv6 destination address.
1011 * XXX backward compatibility wrapper
1012 *
1013 * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1014 * and then bump it when the tag is copied, and release it when the tag is
1015 * freed. Unfortunately, m_tags don't support deep copies (yet), so instead
1016 * we just bump the ia refcount when we receive it. This should be fixed.
1017 */
1018 static struct ip6aux *
1019 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1020 {
1021 struct ip6aux *ip6a;
1022
1023 ip6a = ip6_addaux(m);
1024 if (ip6a)
1025 ip6a->ip6a_dstia6 = ia6;
1026 return ip6a; /* NULL if failed to set */
1027 }
1028
1029 struct in6_ifaddr *
1030 ip6_getdstifaddr(struct mbuf *m)
1031 {
1032 struct ip6aux *ip6a;
1033 struct in6_ifaddr *ia;
1034
1035 ip6a = ip6_findaux(m);
1036 if (ip6a) {
1037 ia = ip6a->ip6a_dstia6;
1038 ifa_ref(&ia->ia_ifa);
1039 return ia;
1040 } else
1041 return NULL;
1042 }
1043
1044 /*
1045 * Hop-by-Hop options header processing. If a valid jumbo payload option is
1046 * included, the real payload length will be stored in plenp.
1047 *
1048 * rtalertp - XXX: should be stored more smart way
1049 */
1050 static int
1051 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1052 struct mbuf **mp, int *offp)
1053 {
1054 struct mbuf *m = *mp;
1055 int off = *offp, hbhlen;
1056 struct ip6_hbh *hbh;
1057 u_int8_t *opt;
1058
1059 /* validation of the length of the header */
1060 #ifndef PULLDOWN_TEST
1061 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1062 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1063 hbhlen = (hbh->ip6h_len + 1) << 3;
1064
1065 IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1066 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1067 #else
1068 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1069 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1070 if (hbh == NULL) {
1071 V_ip6stat.ip6s_tooshort++;
1072 return -1;
1073 }
1074 hbhlen = (hbh->ip6h_len + 1) << 3;
1075 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1076 hbhlen);
1077 if (hbh == NULL) {
1078 V_ip6stat.ip6s_tooshort++;
1079 return -1;
1080 }
1081 #endif
1082 off += hbhlen;
1083 hbhlen -= sizeof(struct ip6_hbh);
1084 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1085
1086 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1087 hbhlen, rtalertp, plenp) < 0)
1088 return (-1);
1089
1090 *offp = off;
1091 *mp = m;
1092 return (0);
1093 }
1094
1095 /*
1096 * Search header for all Hop-by-hop options and process each option.
1097 * This function is separate from ip6_hopopts_input() in order to
1098 * handle a case where the sending node itself process its hop-by-hop
1099 * options header. In such a case, the function is called from ip6_output().
1100 *
1101 * The function assumes that hbh header is located right after the IPv6 header
1102 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1103 * opthead + hbhlen is located in contiguous memory region.
1104 */
1105 int
1106 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1107 u_int32_t *rtalertp, u_int32_t *plenp)
1108 {
1109 struct ip6_hdr *ip6;
1110 int optlen = 0;
1111 u_int8_t *opt = opthead;
1112 u_int16_t rtalert_val;
1113 u_int32_t jumboplen;
1114 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1115
1116 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1117 switch (*opt) {
1118 case IP6OPT_PAD1:
1119 optlen = 1;
1120 break;
1121 case IP6OPT_PADN:
1122 if (hbhlen < IP6OPT_MINLEN) {
1123 V_ip6stat.ip6s_toosmall++;
1124 goto bad;
1125 }
1126 optlen = *(opt + 1) + 2;
1127 break;
1128 case IP6OPT_ROUTER_ALERT:
1129 /* XXX may need check for alignment */
1130 if (hbhlen < IP6OPT_RTALERT_LEN) {
1131 V_ip6stat.ip6s_toosmall++;
1132 goto bad;
1133 }
1134 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1135 /* XXX stat */
1136 icmp6_error(m, ICMP6_PARAM_PROB,
1137 ICMP6_PARAMPROB_HEADER,
1138 erroff + opt + 1 - opthead);
1139 return (-1);
1140 }
1141 optlen = IP6OPT_RTALERT_LEN;
1142 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1143 *rtalertp = ntohs(rtalert_val);
1144 break;
1145 case IP6OPT_JUMBO:
1146 /* XXX may need check for alignment */
1147 if (hbhlen < IP6OPT_JUMBO_LEN) {
1148 V_ip6stat.ip6s_toosmall++;
1149 goto bad;
1150 }
1151 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1152 /* XXX stat */
1153 icmp6_error(m, ICMP6_PARAM_PROB,
1154 ICMP6_PARAMPROB_HEADER,
1155 erroff + opt + 1 - opthead);
1156 return (-1);
1157 }
1158 optlen = IP6OPT_JUMBO_LEN;
1159
1160 /*
1161 * IPv6 packets that have non 0 payload length
1162 * must not contain a jumbo payload option.
1163 */
1164 ip6 = mtod(m, struct ip6_hdr *);
1165 if (ip6->ip6_plen) {
1166 V_ip6stat.ip6s_badoptions++;
1167 icmp6_error(m, ICMP6_PARAM_PROB,
1168 ICMP6_PARAMPROB_HEADER,
1169 erroff + opt - opthead);
1170 return (-1);
1171 }
1172
1173 /*
1174 * We may see jumbolen in unaligned location, so
1175 * we'd need to perform bcopy().
1176 */
1177 bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1178 jumboplen = (u_int32_t)htonl(jumboplen);
1179
1180 #if 1
1181 /*
1182 * if there are multiple jumbo payload options,
1183 * *plenp will be non-zero and the packet will be
1184 * rejected.
1185 * the behavior may need some debate in ipngwg -
1186 * multiple options does not make sense, however,
1187 * there's no explicit mention in specification.
1188 */
1189 if (*plenp != 0) {
1190 V_ip6stat.ip6s_badoptions++;
1191 icmp6_error(m, ICMP6_PARAM_PROB,
1192 ICMP6_PARAMPROB_HEADER,
1193 erroff + opt + 2 - opthead);
1194 return (-1);
1195 }
1196 #endif
1197
1198 /*
1199 * jumbo payload length must be larger than 65535.
1200 */
1201 if (jumboplen <= IPV6_MAXPACKET) {
1202 V_ip6stat.ip6s_badoptions++;
1203 icmp6_error(m, ICMP6_PARAM_PROB,
1204 ICMP6_PARAMPROB_HEADER,
1205 erroff + opt + 2 - opthead);
1206 return (-1);
1207 }
1208 *plenp = jumboplen;
1209
1210 break;
1211 default: /* unknown option */
1212 if (hbhlen < IP6OPT_MINLEN) {
1213 V_ip6stat.ip6s_toosmall++;
1214 goto bad;
1215 }
1216 optlen = ip6_unknown_opt(opt, m,
1217 erroff + opt - opthead);
1218 if (optlen == -1)
1219 return (-1);
1220 optlen += 2;
1221 break;
1222 }
1223 }
1224
1225 return (0);
1226
1227 bad:
1228 m_freem(m);
1229 return (-1);
1230 }
1231
1232 /*
1233 * Unknown option processing.
1234 * The third argument `off' is the offset from the IPv6 header to the option,
1235 * which is necessary if the IPv6 header the and option header and IPv6 header
1236 * is not contiguous in order to return an ICMPv6 error.
1237 */
1238 int
1239 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1240 {
1241 struct ip6_hdr *ip6;
1242
1243 switch (IP6OPT_TYPE(*optp)) {
1244 case IP6OPT_TYPE_SKIP: /* ignore the option */
1245 return ((int)*(optp + 1));
1246 case IP6OPT_TYPE_DISCARD: /* silently discard */
1247 m_freem(m);
1248 return (-1);
1249 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1250 V_ip6stat.ip6s_badoptions++;
1251 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1252 return (-1);
1253 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1254 V_ip6stat.ip6s_badoptions++;
1255 ip6 = mtod(m, struct ip6_hdr *);
1256 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1257 (m->m_flags & (M_BCAST|M_MCAST)))
1258 m_freem(m);
1259 else
1260 icmp6_error(m, ICMP6_PARAM_PROB,
1261 ICMP6_PARAMPROB_OPTION, off);
1262 return (-1);
1263 }
1264
1265 m_freem(m); /* XXX: NOTREACHED */
1266 return (-1);
1267 }
1268
1269 /*
1270 * Create the "control" list for this pcb.
1271 * These functions will not modify mbuf chain at all.
1272 *
1273 * With KAME mbuf chain restriction:
1274 * The routine will be called from upper layer handlers like tcp6_input().
1275 * Thus the routine assumes that the caller (tcp6_input) have already
1276 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1277 * very first mbuf on the mbuf chain.
1278 *
1279 * ip6_savecontrol_v4 will handle those options that are possible to be
1280 * set on a v4-mapped socket.
1281 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1282 * options and handle the v6-only ones itself.
1283 */
1284 struct mbuf **
1285 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1286 int *v4only)
1287 {
1288 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1289
1290 #ifdef SO_TIMESTAMP
1291 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1292 struct timeval tv;
1293
1294 microtime(&tv);
1295 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1296 SCM_TIMESTAMP, SOL_SOCKET);
1297 if (*mp)
1298 mp = &(*mp)->m_next;
1299 }
1300 #endif
1301
1302 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1303 if (v4only != NULL)
1304 *v4only = 1;
1305 return (mp);
1306 }
1307
1308 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1309 /* RFC 2292 sec. 5 */
1310 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1311 struct in6_pktinfo pi6;
1312
1313 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1314 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1315 pi6.ipi6_ifindex =
1316 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1317
1318 *mp = sbcreatecontrol((caddr_t) &pi6,
1319 sizeof(struct in6_pktinfo),
1320 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1321 if (*mp)
1322 mp = &(*mp)->m_next;
1323 }
1324
1325 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1326 int hlim = ip6->ip6_hlim & 0xff;
1327
1328 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1329 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1330 IPPROTO_IPV6);
1331 if (*mp)
1332 mp = &(*mp)->m_next;
1333 }
1334
1335 if (v4only != NULL)
1336 *v4only = 0;
1337 return (mp);
1338 }
1339
1340 void
1341 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1342 {
1343 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1344 int v4only = 0;
1345
1346 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1347 if (v4only)
1348 return;
1349
1350 if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1351 u_int32_t flowinfo;
1352 int tclass;
1353
1354 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1355 flowinfo >>= 20;
1356
1357 tclass = flowinfo & 0xff;
1358 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1359 IPV6_TCLASS, IPPROTO_IPV6);
1360 if (*mp)
1361 mp = &(*mp)->m_next;
1362 }
1363
1364 /*
1365 * IPV6_HOPOPTS socket option. Recall that we required super-user
1366 * privilege for the option (see ip6_ctloutput), but it might be too
1367 * strict, since there might be some hop-by-hop options which can be
1368 * returned to normal user.
1369 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1370 */
1371 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1372 /*
1373 * Check if a hop-by-hop options header is contatined in the
1374 * received packet, and if so, store the options as ancillary
1375 * data. Note that a hop-by-hop options header must be
1376 * just after the IPv6 header, which is assured through the
1377 * IPv6 input processing.
1378 */
1379 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1380 struct ip6_hbh *hbh;
1381 int hbhlen = 0;
1382 #ifdef PULLDOWN_TEST
1383 struct mbuf *ext;
1384 #endif
1385
1386 #ifndef PULLDOWN_TEST
1387 hbh = (struct ip6_hbh *)(ip6 + 1);
1388 hbhlen = (hbh->ip6h_len + 1) << 3;
1389 #else
1390 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1391 ip6->ip6_nxt);
1392 if (ext == NULL) {
1393 V_ip6stat.ip6s_tooshort++;
1394 return;
1395 }
1396 hbh = mtod(ext, struct ip6_hbh *);
1397 hbhlen = (hbh->ip6h_len + 1) << 3;
1398 if (hbhlen != ext->m_len) {
1399 m_freem(ext);
1400 V_ip6stat.ip6s_tooshort++;
1401 return;
1402 }
1403 #endif
1404
1405 /*
1406 * XXX: We copy the whole header even if a
1407 * jumbo payload option is included, the option which
1408 * is to be removed before returning according to
1409 * RFC2292.
1410 * Note: this constraint is removed in RFC3542
1411 */
1412 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1413 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1414 IPPROTO_IPV6);
1415 if (*mp)
1416 mp = &(*mp)->m_next;
1417 #ifdef PULLDOWN_TEST
1418 m_freem(ext);
1419 #endif
1420 }
1421 }
1422
1423 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1424 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1425
1426 /*
1427 * Search for destination options headers or routing
1428 * header(s) through the header chain, and stores each
1429 * header as ancillary data.
1430 * Note that the order of the headers remains in
1431 * the chain of ancillary data.
1432 */
1433 while (1) { /* is explicit loop prevention necessary? */
1434 struct ip6_ext *ip6e = NULL;
1435 int elen;
1436 #ifdef PULLDOWN_TEST
1437 struct mbuf *ext = NULL;
1438 #endif
1439
1440 /*
1441 * if it is not an extension header, don't try to
1442 * pull it from the chain.
1443 */
1444 switch (nxt) {
1445 case IPPROTO_DSTOPTS:
1446 case IPPROTO_ROUTING:
1447 case IPPROTO_HOPOPTS:
1448 case IPPROTO_AH: /* is it possible? */
1449 break;
1450 default:
1451 goto loopend;
1452 }
1453
1454 #ifndef PULLDOWN_TEST
1455 if (off + sizeof(*ip6e) > m->m_len)
1456 goto loopend;
1457 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1458 if (nxt == IPPROTO_AH)
1459 elen = (ip6e->ip6e_len + 2) << 2;
1460 else
1461 elen = (ip6e->ip6e_len + 1) << 3;
1462 if (off + elen > m->m_len)
1463 goto loopend;
1464 #else
1465 ext = ip6_pullexthdr(m, off, nxt);
1466 if (ext == NULL) {
1467 V_ip6stat.ip6s_tooshort++;
1468 return;
1469 }
1470 ip6e = mtod(ext, struct ip6_ext *);
1471 if (nxt == IPPROTO_AH)
1472 elen = (ip6e->ip6e_len + 2) << 2;
1473 else
1474 elen = (ip6e->ip6e_len + 1) << 3;
1475 if (elen != ext->m_len) {
1476 m_freem(ext);
1477 V_ip6stat.ip6s_tooshort++;
1478 return;
1479 }
1480 #endif
1481
1482 switch (nxt) {
1483 case IPPROTO_DSTOPTS:
1484 if (!(in6p->inp_flags & IN6P_DSTOPTS))
1485 break;
1486
1487 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1488 IS2292(in6p,
1489 IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1490 IPPROTO_IPV6);
1491 if (*mp)
1492 mp = &(*mp)->m_next;
1493 break;
1494 case IPPROTO_ROUTING:
1495 if (!(in6p->inp_flags & IN6P_RTHDR))
1496 break;
1497
1498 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1499 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1500 IPPROTO_IPV6);
1501 if (*mp)
1502 mp = &(*mp)->m_next;
1503 break;
1504 case IPPROTO_HOPOPTS:
1505 case IPPROTO_AH: /* is it possible? */
1506 break;
1507
1508 default:
1509 /*
1510 * other cases have been filtered in the above.
1511 * none will visit this case. here we supply
1512 * the code just in case (nxt overwritten or
1513 * other cases).
1514 */
1515 #ifdef PULLDOWN_TEST
1516 m_freem(ext);
1517 #endif
1518 goto loopend;
1519
1520 }
1521
1522 /* proceed with the next header. */
1523 off += elen;
1524 nxt = ip6e->ip6e_nxt;
1525 ip6e = NULL;
1526 #ifdef PULLDOWN_TEST
1527 m_freem(ext);
1528 ext = NULL;
1529 #endif
1530 }
1531 loopend:
1532 ;
1533 }
1534 }
1535 #undef IS2292
1536
1537 void
1538 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1539 {
1540 struct socket *so;
1541 struct mbuf *m_mtu;
1542 struct ip6_mtuinfo mtuctl;
1543
1544 so = in6p->inp_socket;
1545
1546 if (mtu == NULL)
1547 return;
1548
1549 #ifdef DIAGNOSTIC
1550 if (so == NULL) /* I believe this is impossible */
1551 panic("ip6_notify_pmtu: socket is NULL");
1552 #endif
1553
1554 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */
1555 mtuctl.ip6m_mtu = *mtu;
1556 mtuctl.ip6m_addr = *dst;
1557 if (sa6_recoverscope(&mtuctl.ip6m_addr))
1558 return;
1559
1560 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1561 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1562 return;
1563
1564 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1565 == 0) {
1566 m_freem(m_mtu);
1567 /* XXX: should count statistics */
1568 } else
1569 sorwakeup(so);
1570
1571 return;
1572 }
1573
1574 #ifdef PULLDOWN_TEST
1575 /*
1576 * pull single extension header from mbuf chain. returns single mbuf that
1577 * contains the result, or NULL on error.
1578 */
1579 static struct mbuf *
1580 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1581 {
1582 struct ip6_ext ip6e;
1583 size_t elen;
1584 struct mbuf *n;
1585
1586 #ifdef DIAGNOSTIC
1587 switch (nxt) {
1588 case IPPROTO_DSTOPTS:
1589 case IPPROTO_ROUTING:
1590 case IPPROTO_HOPOPTS:
1591 case IPPROTO_AH: /* is it possible? */
1592 break;
1593 default:
1594 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1595 }
1596 #endif
1597
1598 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1599 if (nxt == IPPROTO_AH)
1600 elen = (ip6e.ip6e_len + 2) << 2;
1601 else
1602 elen = (ip6e.ip6e_len + 1) << 3;
1603
1604 MGET(n, M_DONTWAIT, MT_DATA);
1605 if (n && elen >= MLEN) {
1606 MCLGET(n, M_DONTWAIT);
1607 if ((n->m_flags & M_EXT) == 0) {
1608 m_free(n);
1609 n = NULL;
1610 }
1611 }
1612 if (!n)
1613 return NULL;
1614
1615 n->m_len = 0;
1616 if (elen >= M_TRAILINGSPACE(n)) {
1617 m_free(n);
1618 return NULL;
1619 }
1620
1621 m_copydata(m, off, elen, mtod(n, caddr_t));
1622 n->m_len = elen;
1623 return n;
1624 }
1625 #endif
1626
1627 /*
1628 * Get pointer to the previous header followed by the header
1629 * currently processed.
1630 * XXX: This function supposes that
1631 * M includes all headers,
1632 * the next header field and the header length field of each header
1633 * are valid, and
1634 * the sum of each header length equals to OFF.
1635 * Because of these assumptions, this function must be called very
1636 * carefully. Moreover, it will not be used in the near future when
1637 * we develop `neater' mechanism to process extension headers.
1638 */
1639 char *
1640 ip6_get_prevhdr(struct mbuf *m, int off)
1641 {
1642 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1643
1644 if (off == sizeof(struct ip6_hdr))
1645 return (&ip6->ip6_nxt);
1646 else {
1647 int len, nxt;
1648 struct ip6_ext *ip6e = NULL;
1649
1650 nxt = ip6->ip6_nxt;
1651 len = sizeof(struct ip6_hdr);
1652 while (len < off) {
1653 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1654
1655 switch (nxt) {
1656 case IPPROTO_FRAGMENT:
1657 len += sizeof(struct ip6_frag);
1658 break;
1659 case IPPROTO_AH:
1660 len += (ip6e->ip6e_len + 2) << 2;
1661 break;
1662 default:
1663 len += (ip6e->ip6e_len + 1) << 3;
1664 break;
1665 }
1666 nxt = ip6e->ip6e_nxt;
1667 }
1668 if (ip6e)
1669 return (&ip6e->ip6e_nxt);
1670 else
1671 return NULL;
1672 }
1673 }
1674
1675 /*
1676 * get next header offset. m will be retained.
1677 */
1678 int
1679 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1680 {
1681 struct ip6_hdr ip6;
1682 struct ip6_ext ip6e;
1683 struct ip6_frag fh;
1684
1685 /* just in case */
1686 if (m == NULL)
1687 panic("ip6_nexthdr: m == NULL");
1688 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1689 return -1;
1690
1691 switch (proto) {
1692 case IPPROTO_IPV6:
1693 if (m->m_pkthdr.len < off + sizeof(ip6))
1694 return -1;
1695 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1696 if (nxtp)
1697 *nxtp = ip6.ip6_nxt;
1698 off += sizeof(ip6);
1699 return off;
1700
1701 case IPPROTO_FRAGMENT:
1702 /*
1703 * terminate parsing if it is not the first fragment,
1704 * it does not make sense to parse through it.
1705 */
1706 if (m->m_pkthdr.len < off + sizeof(fh))
1707 return -1;
1708 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1709 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1710 if (fh.ip6f_offlg & IP6F_OFF_MASK)
1711 return -1;
1712 if (nxtp)
1713 *nxtp = fh.ip6f_nxt;
1714 off += sizeof(struct ip6_frag);
1715 return off;
1716
1717 case IPPROTO_AH:
1718 if (m->m_pkthdr.len < off + sizeof(ip6e))
1719 return -1;
1720 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1721 if (nxtp)
1722 *nxtp = ip6e.ip6e_nxt;
1723 off += (ip6e.ip6e_len + 2) << 2;
1724 return off;
1725
1726 case IPPROTO_HOPOPTS:
1727 case IPPROTO_ROUTING:
1728 case IPPROTO_DSTOPTS:
1729 if (m->m_pkthdr.len < off + sizeof(ip6e))
1730 return -1;
1731 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1732 if (nxtp)
1733 *nxtp = ip6e.ip6e_nxt;
1734 off += (ip6e.ip6e_len + 1) << 3;
1735 return off;
1736
1737 case IPPROTO_NONE:
1738 case IPPROTO_ESP:
1739 case IPPROTO_IPCOMP:
1740 /* give up */
1741 return -1;
1742
1743 default:
1744 return -1;
1745 }
1746
1747 return -1;
1748 }
1749
1750 /*
1751 * get offset for the last header in the chain. m will be kept untainted.
1752 */
1753 int
1754 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1755 {
1756 int newoff;
1757 int nxt;
1758
1759 if (!nxtp) {
1760 nxt = -1;
1761 nxtp = &nxt;
1762 }
1763 while (1) {
1764 newoff = ip6_nexthdr(m, off, proto, nxtp);
1765 if (newoff < 0)
1766 return off;
1767 else if (newoff < off)
1768 return -1; /* invalid */
1769 else if (newoff == off)
1770 return newoff;
1771
1772 off = newoff;
1773 proto = *nxtp;
1774 }
1775 }
1776
1777 struct ip6aux *
1778 ip6_addaux(struct mbuf *m)
1779 {
1780 struct m_tag *mtag;
1781
1782 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1783 if (!mtag) {
1784 mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1785 M_NOWAIT);
1786 if (mtag) {
1787 m_tag_prepend(m, mtag);
1788 bzero(mtag + 1, sizeof(struct ip6aux));
1789 }
1790 }
1791 return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1792 }
1793
1794 struct ip6aux *
1795 ip6_findaux(struct mbuf *m)
1796 {
1797 struct m_tag *mtag;
1798
1799 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1800 return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1801 }
1802
1803 void
1804 ip6_delaux(struct mbuf *m)
1805 {
1806 struct m_tag *mtag;
1807
1808 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1809 if (mtag)
1810 m_tag_delete(m, mtag);
1811 }
1812
1813 /*
1814 * System control for IP6
1815 */
1816
1817 u_char inet6ctlerrmap[PRC_NCMDS] = {
1818 0, 0, 0, 0,
1819 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1820 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1821 EMSGSIZE, EHOSTUNREACH, 0, 0,
1822 0, 0, 0, 0,
1823 ENOPROTOOPT
1824 };
Cache object: 104733822a1bae0dbc8f9c59ddd5a061
|