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("$FreeBSD$");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/proc.h>
75 #include <sys/domain.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/errno.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/if_dl.h>
87 #include <net/route.h>
88 #include <net/netisr.h>
89 #include <net/pfil.h>
90 #include <net/vnet.h>
91
92 #include <netinet/in.h>
93 #include <netinet/in_systm.h>
94 #include <net/if_llatbl.h>
95 #ifdef INET
96 #include <netinet/ip.h>
97 #include <netinet/ip_icmp.h>
98 #endif /* INET */
99 #include <netinet/ip6.h>
100 #include <netinet6/in6_var.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet/in_pcb.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/scope6_var.h>
105 #include <netinet6/in6_ifattach.h>
106 #include <netinet6/nd6.h>
107
108 #ifdef IPSEC
109 #include <netipsec/ipsec.h>
110 #include <netinet6/ip6_ipsec.h>
111 #include <netipsec/ipsec6.h>
112 #endif /* IPSEC */
113
114 #include <netinet6/ip6protosw.h>
115
116 extern struct domain inet6domain;
117
118 u_char ip6_protox[IPPROTO_MAX];
119 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
120
121 static struct netisr_handler ip6_nh = {
122 .nh_name = "ip6",
123 .nh_handler = ip6_input,
124 .nh_proto = NETISR_IPV6,
125 .nh_policy = NETISR_POLICY_FLOW,
126 };
127
128 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
129 #define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch)
130
131 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
132
133 VNET_DEFINE(struct ip6stat, ip6stat);
134
135 struct rwlock in6_ifaddr_lock;
136 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
137
138 static void ip6_init2(void *);
139 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
140 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
141 #ifdef PULLDOWN_TEST
142 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
143 #endif
144
145 /*
146 * IP6 initialization: fill in IP6 protocol switch table.
147 * All protocols not implemented in kernel go to raw IP6 protocol handler.
148 */
149 void
150 ip6_init(void)
151 {
152 struct ip6protosw *pr;
153 int i;
154
155 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
156 &V_ip6_auto_linklocal);
157
158 TAILQ_INIT(&V_in6_ifaddrhead);
159
160 /* Initialize packet filter hooks. */
161 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
162 V_inet6_pfil_hook.ph_af = AF_INET6;
163 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
164 printf("%s: WARNING: unable to register pfil hook, "
165 "error %d\n", __func__, i);
166
167 scope6_init();
168 addrsel_policy_init();
169 nd6_init();
170 frag6_init();
171
172 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
173
174 /* Skip global initialization stuff for non-default instances. */
175 if (!IS_DEFAULT_VNET(curvnet))
176 return;
177
178 #ifdef DIAGNOSTIC
179 if (sizeof(struct protosw) != sizeof(struct ip6protosw))
180 panic("sizeof(protosw) != sizeof(ip6protosw)");
181 #endif
182 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
183 if (pr == NULL)
184 panic("ip6_init");
185
186 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
187 for (i = 0; i < IPPROTO_MAX; i++)
188 ip6_protox[i] = pr - inet6sw;
189 /*
190 * Cycle through IP protocols and put them into the appropriate place
191 * in ip6_protox[].
192 */
193 for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
194 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
195 if (pr->pr_domain->dom_family == PF_INET6 &&
196 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
197 /* Be careful to only index valid IP protocols. */
198 if (pr->pr_protocol < IPPROTO_MAX)
199 ip6_protox[pr->pr_protocol] = pr - inet6sw;
200 }
201
202 netisr_register(&ip6_nh);
203 }
204
205 /*
206 * The protocol to be inserted into ip6_protox[] must be already registered
207 * in inet6sw[], either statically or through pf_proto_register().
208 */
209 int
210 ip6proto_register(short ip6proto)
211 {
212 struct ip6protosw *pr;
213
214 /* Sanity checks. */
215 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
216 return (EPROTONOSUPPORT);
217
218 /*
219 * The protocol slot must not be occupied by another protocol
220 * already. An index pointing to IPPROTO_RAW is unused.
221 */
222 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
223 if (pr == NULL)
224 return (EPFNOSUPPORT);
225 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */
226 return (EEXIST);
227
228 /*
229 * Find the protocol position in inet6sw[] and set the index.
230 */
231 for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
232 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
233 if (pr->pr_domain->dom_family == PF_INET6 &&
234 pr->pr_protocol && pr->pr_protocol == ip6proto) {
235 ip6_protox[pr->pr_protocol] = pr - inet6sw;
236 return (0);
237 }
238 }
239 return (EPROTONOSUPPORT);
240 }
241
242 int
243 ip6proto_unregister(short ip6proto)
244 {
245 struct ip6protosw *pr;
246
247 /* Sanity checks. */
248 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
249 return (EPROTONOSUPPORT);
250
251 /* Check if the protocol was indeed registered. */
252 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
253 if (pr == NULL)
254 return (EPFNOSUPPORT);
255 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */
256 return (ENOENT);
257
258 /* Reset the protocol slot to IPPROTO_RAW. */
259 ip6_protox[ip6proto] = pr - inet6sw;
260 return (0);
261 }
262
263 #ifdef VIMAGE
264 void
265 ip6_destroy()
266 {
267
268 nd6_destroy();
269 callout_drain(&V_in6_tmpaddrtimer_ch);
270 }
271 #endif
272
273 static int
274 ip6_init2_vnet(const void *unused __unused)
275 {
276
277 /* nd6_timer_init */
278 callout_init(&V_nd6_timer_ch, 0);
279 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
280
281 /* timer for regeneranation of temporary addresses randomize ID */
282 callout_init(&V_in6_tmpaddrtimer_ch, 0);
283 callout_reset(&V_in6_tmpaddrtimer_ch,
284 (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
285 V_ip6_temp_regen_advance) * hz,
286 in6_tmpaddrtimer, curvnet);
287
288 return (0);
289 }
290
291 static void
292 ip6_init2(void *dummy)
293 {
294
295 ip6_init2_vnet(NULL);
296 }
297
298 /* cheat */
299 /* This must be after route_init(), which is now SI_ORDER_THIRD */
300 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
301
302 void
303 ip6_input(struct mbuf *m)
304 {
305 struct ip6_hdr *ip6;
306 int off = sizeof(struct ip6_hdr), nest;
307 u_int32_t plen;
308 u_int32_t rtalert = ~0;
309 int nxt, ours = 0;
310 struct ifnet *deliverifp = NULL, *ifp = NULL;
311 struct in6_addr odst;
312 struct route_in6 rin6;
313 int srcrt = 0;
314 struct llentry *lle = NULL;
315 struct sockaddr_in6 dst6, *dst;
316
317 bzero(&rin6, sizeof(struct route_in6));
318 #ifdef IPSEC
319 /*
320 * should the inner packet be considered authentic?
321 * see comment in ah4_input().
322 * NB: m cannot be NULL when passed to the input routine
323 */
324
325 m->m_flags &= ~M_AUTHIPHDR;
326 m->m_flags &= ~M_AUTHIPDGM;
327
328 #endif /* IPSEC */
329
330 /*
331 * make sure we don't have onion peering information into m_tag.
332 */
333 ip6_delaux(m);
334
335 /*
336 * mbuf statistics
337 */
338 if (m->m_flags & M_EXT) {
339 if (m->m_next)
340 V_ip6stat.ip6s_mext2m++;
341 else
342 V_ip6stat.ip6s_mext1++;
343 } else {
344 #define M2MMAX (sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
345 if (m->m_next) {
346 if (m->m_flags & M_LOOP) {
347 V_ip6stat.ip6s_m2m[V_loif->if_index]++;
348 } else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
349 V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
350 else
351 V_ip6stat.ip6s_m2m[0]++;
352 } else
353 V_ip6stat.ip6s_m1++;
354 #undef M2MMAX
355 }
356
357 /* drop the packet if IPv6 operation is disabled on the IF */
358 if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
359 m_freem(m);
360 return;
361 }
362
363 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
364 V_ip6stat.ip6s_total++;
365
366 #ifndef PULLDOWN_TEST
367 /*
368 * L2 bridge code and some other code can return mbuf chain
369 * that does not conform to KAME requirement. too bad.
370 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
371 */
372 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
373 struct mbuf *n;
374
375 MGETHDR(n, M_DONTWAIT, MT_HEADER);
376 if (n)
377 M_MOVE_PKTHDR(n, m);
378 if (n && n->m_pkthdr.len > MHLEN) {
379 MCLGET(n, M_DONTWAIT);
380 if ((n->m_flags & M_EXT) == 0) {
381 m_freem(n);
382 n = NULL;
383 }
384 }
385 if (n == NULL) {
386 m_freem(m);
387 return; /* ENOBUFS */
388 }
389
390 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
391 n->m_len = n->m_pkthdr.len;
392 m_freem(m);
393 m = n;
394 }
395 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
396 #endif
397
398 if (m->m_len < sizeof(struct ip6_hdr)) {
399 struct ifnet *inifp;
400 inifp = m->m_pkthdr.rcvif;
401 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
402 V_ip6stat.ip6s_toosmall++;
403 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
404 return;
405 }
406 }
407
408 ip6 = mtod(m, struct ip6_hdr *);
409
410 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
411 V_ip6stat.ip6s_badvers++;
412 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
413 goto bad;
414 }
415
416 V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
417
418 /*
419 * Check against address spoofing/corruption.
420 */
421 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
422 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
423 /*
424 * XXX: "badscope" is not very suitable for a multicast source.
425 */
426 V_ip6stat.ip6s_badscope++;
427 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
428 goto bad;
429 }
430 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
431 !(m->m_flags & M_LOOP)) {
432 /*
433 * In this case, the packet should come from the loopback
434 * interface. However, we cannot just check the if_flags,
435 * because ip6_mloopback() passes the "actual" interface
436 * as the outgoing/incoming interface.
437 */
438 V_ip6stat.ip6s_badscope++;
439 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
440 goto bad;
441 }
442
443 #ifdef ALTQ
444 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
445 /* packet is dropped by traffic conditioner */
446 return;
447 }
448 #endif
449 /*
450 * The following check is not documented in specs. A malicious
451 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
452 * and bypass security checks (act as if it was from 127.0.0.1 by using
453 * IPv6 src ::ffff:127.0.0.1). Be cautious.
454 *
455 * This check chokes if we are in an SIIT cloud. As none of BSDs
456 * support IPv4-less kernel compilation, we cannot support SIIT
457 * environment at all. So, it makes more sense for us to reject any
458 * malicious packets for non-SIIT environment, than try to do a
459 * partial support for SIIT environment.
460 */
461 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
462 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
463 V_ip6stat.ip6s_badscope++;
464 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
465 goto bad;
466 }
467 #if 0
468 /*
469 * Reject packets with IPv4 compatible addresses (auto tunnel).
470 *
471 * The code forbids auto tunnel relay case in RFC1933 (the check is
472 * stronger than RFC1933). We may want to re-enable it if mech-xx
473 * is revised to forbid relaying case.
474 */
475 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
476 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
477 V_ip6stat.ip6s_badscope++;
478 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
479 goto bad;
480 }
481 #endif
482
483 /*
484 * Run through list of hooks for input packets.
485 *
486 * NB: Beware of the destination address changing
487 * (e.g. by NAT rewriting). When this happens,
488 * tell ip6_forward to do the right thing.
489 */
490 odst = ip6->ip6_dst;
491
492 /* Jump over all PFIL processing if hooks are not active. */
493 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
494 goto passin;
495
496 if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
497 m->m_pkthdr.rcvif, PFIL_IN, NULL))
498 return;
499 if (m == NULL) /* consumed by filter */
500 return;
501 ip6 = mtod(m, struct ip6_hdr *);
502 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
503
504 passin:
505 /*
506 * Disambiguate address scope zones (if there is ambiguity).
507 * We first make sure that the original source or destination address
508 * is not in our internal form for scoped addresses. Such addresses
509 * are not necessarily invalid spec-wise, but we cannot accept them due
510 * to the usage conflict.
511 * in6_setscope() then also checks and rejects the cases where src or
512 * dst are the loopback address and the receiving interface
513 * is not loopback.
514 */
515 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
516 V_ip6stat.ip6s_badscope++; /* XXX */
517 goto bad;
518 }
519 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
520 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
521 V_ip6stat.ip6s_badscope++;
522 goto bad;
523 }
524
525 /*
526 * Multicast check. Assume packet is for us to avoid
527 * prematurely taking locks.
528 */
529 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
530 ours = 1;
531 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
532 deliverifp = m->m_pkthdr.rcvif;
533 goto hbhcheck;
534 }
535
536 /*
537 * Unicast check
538 */
539
540 bzero(&dst6, sizeof(dst6));
541 dst6.sin6_family = AF_INET6;
542 dst6.sin6_len = sizeof(struct sockaddr_in6);
543 dst6.sin6_addr = ip6->ip6_dst;
544 ifp = m->m_pkthdr.rcvif;
545 IF_AFDATA_RLOCK(ifp);
546 lle = lla_lookup(LLTABLE6(ifp), 0,
547 (struct sockaddr *)&dst6);
548 IF_AFDATA_RUNLOCK(ifp);
549 if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
550 struct ifaddr *ifa;
551 struct in6_ifaddr *ia6;
552 int bad;
553
554 bad = 1;
555 #define sa_equal(a1, a2) \
556 (bcmp((a1), (a2), ((a1))->sin6_len) == 0)
557 IF_ADDR_RLOCK(ifp);
558 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
559 if (ifa->ifa_addr->sa_family != dst6.sin6_family)
560 continue;
561 if (sa_equal(&dst6, ifa->ifa_addr))
562 break;
563 }
564 KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
565 __func__, lle));
566 #undef sa_equal
567
568 ia6 = (struct in6_ifaddr *)ifa;
569 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
570 /* Count the packet in the ip address stats */
571 ia6->ia_ifa.if_ipackets++;
572 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
573
574 /*
575 * record address information into m_tag.
576 */
577 (void)ip6_setdstifaddr(m, ia6);
578
579 bad = 0;
580 } else {
581 char ip6bufs[INET6_ADDRSTRLEN];
582 char ip6bufd[INET6_ADDRSTRLEN];
583 /* address is not ready, so discard the packet. */
584 nd6log((LOG_INFO,
585 "ip6_input: packet to an unready address %s->%s\n",
586 ip6_sprintf(ip6bufs, &ip6->ip6_src),
587 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
588 }
589 IF_ADDR_RUNLOCK(ifp);
590 LLE_RUNLOCK(lle);
591 if (bad)
592 goto bad;
593 else {
594 ours = 1;
595 deliverifp = ifp;
596 goto hbhcheck;
597 }
598 }
599 if (lle != NULL)
600 LLE_RUNLOCK(lle);
601
602 dst = &rin6.ro_dst;
603 dst->sin6_len = sizeof(struct sockaddr_in6);
604 dst->sin6_family = AF_INET6;
605 dst->sin6_addr = ip6->ip6_dst;
606 rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
607 if (rin6.ro_rt)
608 RT_UNLOCK(rin6.ro_rt);
609
610 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
611
612 /*
613 * Accept the packet if the forwarding interface to the destination
614 * according to the routing table is the loopback interface,
615 * unless the associated route has a gateway.
616 * Note that this approach causes to accept a packet if there is a
617 * route to the loopback interface for the destination of the packet.
618 * But we think it's even useful in some situations, e.g. when using
619 * a special daemon which wants to intercept the packet.
620 *
621 * XXX: some OSes automatically make a cloned route for the destination
622 * of an outgoing packet. If the outgoing interface of the packet
623 * is a loopback one, the kernel would consider the packet to be
624 * accepted, even if we have no such address assinged on the interface.
625 * We check the cloned flag of the route entry to reject such cases,
626 * assuming that route entries for our own addresses are not made by
627 * cloning (it should be true because in6_addloop explicitly installs
628 * the host route). However, we might have to do an explicit check
629 * while it would be less efficient. Or, should we rather install a
630 * reject route for such a case?
631 */
632 if (rin6.ro_rt &&
633 (rin6.ro_rt->rt_flags &
634 (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
635 #ifdef RTF_WASCLONED
636 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
637 #endif
638 #ifdef RTF_CLONED
639 !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
640 #endif
641 #if 0
642 /*
643 * The check below is redundant since the comparison of
644 * the destination and the key of the rtentry has
645 * already done through looking up the routing table.
646 */
647 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
648 &rt6_key(rin6.ro_rt)->sin6_addr)
649 #endif
650 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
651 int free_ia6 = 0;
652 struct in6_ifaddr *ia6;
653
654 /*
655 * found the loopback route to the interface address
656 */
657 if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
658 struct sockaddr_in6 dest6;
659
660 bzero(&dest6, sizeof(dest6));
661 dest6.sin6_family = AF_INET6;
662 dest6.sin6_len = sizeof(dest6);
663 dest6.sin6_addr = ip6->ip6_dst;
664 ia6 = (struct in6_ifaddr *)
665 ifa_ifwithaddr((struct sockaddr *)&dest6);
666 if (ia6 == NULL)
667 goto bad;
668 free_ia6 = 1;
669 }
670 else
671 ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
672
673 /*
674 * record address information into m_tag.
675 */
676 (void)ip6_setdstifaddr(m, ia6);
677
678 /*
679 * packets to a tentative, duplicated, or somehow invalid
680 * address must not be accepted.
681 */
682 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
683 /* this address is ready */
684 ours = 1;
685 deliverifp = ia6->ia_ifp; /* correct? */
686 /* Count the packet in the ip address stats */
687 ia6->ia_ifa.if_ipackets++;
688 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
689 if (ia6 != NULL && free_ia6 != 0)
690 ifa_free(&ia6->ia_ifa);
691 goto hbhcheck;
692 } else {
693 char ip6bufs[INET6_ADDRSTRLEN];
694 char ip6bufd[INET6_ADDRSTRLEN];
695 /* address is not ready, so discard the packet. */
696 nd6log((LOG_INFO,
697 "ip6_input: packet to an unready address %s->%s\n",
698 ip6_sprintf(ip6bufs, &ip6->ip6_src),
699 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
700
701 if (ia6 != NULL && free_ia6 != 0)
702 ifa_free(&ia6->ia_ifa);
703 goto bad;
704 }
705 }
706
707 /*
708 * FAITH (Firewall Aided Internet Translator)
709 */
710 if (V_ip6_keepfaith) {
711 if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
712 rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
713 /* XXX do we need more sanity checks? */
714 ours = 1;
715 deliverifp = rin6.ro_rt->rt_ifp; /* faith */
716 goto hbhcheck;
717 }
718 }
719
720 /*
721 * Now there is no reason to process the packet if it's not our own
722 * and we're not a router.
723 */
724 if (!V_ip6_forwarding) {
725 V_ip6stat.ip6s_cantforward++;
726 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
727 goto bad;
728 }
729
730 hbhcheck:
731 /*
732 * record address information into m_tag, if we don't have one yet.
733 * note that we are unable to record it, if the address is not listed
734 * as our interface address (e.g. multicast addresses, addresses
735 * within FAITH prefixes and such).
736 */
737 if (deliverifp) {
738 struct in6_ifaddr *ia6;
739
740 if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
741 ifa_free(&ia6->ia_ifa);
742 } else {
743 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
744 if (ia6) {
745 if (!ip6_setdstifaddr(m, ia6)) {
746 /*
747 * XXX maybe we should drop the packet here,
748 * as we could not provide enough information
749 * to the upper layers.
750 */
751 }
752 ifa_free(&ia6->ia_ifa);
753 }
754 }
755 }
756
757 /*
758 * Process Hop-by-Hop options header if it's contained.
759 * m may be modified in ip6_hopopts_input().
760 * If a JumboPayload option is included, plen will also be modified.
761 */
762 plen = (u_int32_t)ntohs(ip6->ip6_plen);
763 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
764 struct ip6_hbh *hbh;
765
766 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
767 #if 0 /*touches NULL pointer*/
768 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
769 #endif
770 goto out; /* m have already been freed */
771 }
772
773 /* adjust pointer */
774 ip6 = mtod(m, struct ip6_hdr *);
775
776 /*
777 * if the payload length field is 0 and the next header field
778 * indicates Hop-by-Hop Options header, then a Jumbo Payload
779 * option MUST be included.
780 */
781 if (ip6->ip6_plen == 0 && plen == 0) {
782 /*
783 * Note that if a valid jumbo payload option is
784 * contained, ip6_hopopts_input() must set a valid
785 * (non-zero) payload length to the variable plen.
786 */
787 V_ip6stat.ip6s_badoptions++;
788 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
789 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
790 icmp6_error(m, ICMP6_PARAM_PROB,
791 ICMP6_PARAMPROB_HEADER,
792 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
793 goto out;
794 }
795 #ifndef PULLDOWN_TEST
796 /* ip6_hopopts_input() ensures that mbuf is contiguous */
797 hbh = (struct ip6_hbh *)(ip6 + 1);
798 #else
799 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
800 sizeof(struct ip6_hbh));
801 if (hbh == NULL) {
802 V_ip6stat.ip6s_tooshort++;
803 goto out;
804 }
805 #endif
806 nxt = hbh->ip6h_nxt;
807
808 /*
809 * If we are acting as a router and the packet contains a
810 * router alert option, see if we know the option value.
811 * Currently, we only support the option value for MLD, in which
812 * case we should pass the packet to the multicast routing
813 * daemon.
814 */
815 if (rtalert != ~0) {
816 switch (rtalert) {
817 case IP6OPT_RTALERT_MLD:
818 if (V_ip6_forwarding)
819 ours = 1;
820 break;
821 default:
822 /*
823 * RFC2711 requires unrecognized values must be
824 * silently ignored.
825 */
826 break;
827 }
828 }
829 } else
830 nxt = ip6->ip6_nxt;
831
832 /*
833 * Check that the amount of data in the buffers
834 * is as at least much as the IPv6 header would have us expect.
835 * Trim mbufs if longer than we expect.
836 * Drop packet if shorter than we expect.
837 */
838 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
839 V_ip6stat.ip6s_tooshort++;
840 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
841 goto bad;
842 }
843 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
844 if (m->m_len == m->m_pkthdr.len) {
845 m->m_len = sizeof(struct ip6_hdr) + plen;
846 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
847 } else
848 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
849 }
850
851 /*
852 * Forward if desirable.
853 */
854 if (V_ip6_mrouter &&
855 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
856 /*
857 * If we are acting as a multicast router, all
858 * incoming multicast packets are passed to the
859 * kernel-level multicast forwarding function.
860 * The packet is returned (relatively) intact; if
861 * ip6_mforward() returns a non-zero value, the packet
862 * must be discarded, else it may be accepted below.
863 *
864 * XXX TODO: Check hlim and multicast scope here to avoid
865 * unnecessarily calling into ip6_mforward().
866 */
867 if (ip6_mforward &&
868 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
869 IP6STAT_INC(ip6s_cantforward);
870 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
871 goto bad;
872 }
873 } else if (!ours) {
874 ip6_forward(m, srcrt);
875 goto out;
876 }
877
878 ip6 = mtod(m, struct ip6_hdr *);
879
880 /*
881 * Malicious party may be able to use IPv4 mapped addr to confuse
882 * tcp/udp stack and bypass security checks (act as if it was from
883 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious.
884 *
885 * For SIIT end node behavior, you may want to disable the check.
886 * However, you will become vulnerable to attacks using IPv4 mapped
887 * source.
888 */
889 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
890 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
891 V_ip6stat.ip6s_badscope++;
892 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
893 goto bad;
894 }
895
896 /*
897 * Tell launch routine the next header
898 */
899 V_ip6stat.ip6s_delivered++;
900 in6_ifstat_inc(deliverifp, ifs6_in_deliver);
901 nest = 0;
902
903 while (nxt != IPPROTO_DONE) {
904 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
905 V_ip6stat.ip6s_toomanyhdr++;
906 goto bad;
907 }
908
909 /*
910 * protection against faulty packet - there should be
911 * more sanity checks in header chain processing.
912 */
913 if (m->m_pkthdr.len < off) {
914 V_ip6stat.ip6s_tooshort++;
915 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
916 goto bad;
917 }
918
919 #ifdef IPSEC
920 /*
921 * enforce IPsec policy checking if we are seeing last header.
922 * note that we do not visit this with protocols with pcb layer
923 * code - like udp/tcp/raw ip.
924 */
925 if (ip6_ipsec_input(m, nxt))
926 goto bad;
927 #endif /* IPSEC */
928
929 /*
930 * Use mbuf flags to propagate Router Alert option to
931 * ICMPv6 layer, as hop-by-hop options have been stripped.
932 */
933 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
934 m->m_flags |= M_RTALERT_MLD;
935
936 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
937 }
938 goto out;
939 bad:
940 m_freem(m);
941 out:
942 if (rin6.ro_rt)
943 RTFREE(rin6.ro_rt);
944 }
945
946 /*
947 * set/grab in6_ifaddr correspond to IPv6 destination address.
948 * XXX backward compatibility wrapper
949 *
950 * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
951 * and then bump it when the tag is copied, and release it when the tag is
952 * freed. Unfortunately, m_tags don't support deep copies (yet), so instead
953 * we just bump the ia refcount when we receive it. This should be fixed.
954 */
955 static struct ip6aux *
956 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
957 {
958 struct ip6aux *ip6a;
959
960 ip6a = ip6_addaux(m);
961 if (ip6a)
962 ip6a->ip6a_dstia6 = ia6;
963 return ip6a; /* NULL if failed to set */
964 }
965
966 struct in6_ifaddr *
967 ip6_getdstifaddr(struct mbuf *m)
968 {
969 struct ip6aux *ip6a;
970 struct in6_ifaddr *ia;
971
972 ip6a = ip6_findaux(m);
973 if (ip6a) {
974 ia = ip6a->ip6a_dstia6;
975 ifa_ref(&ia->ia_ifa);
976 return ia;
977 } else
978 return NULL;
979 }
980
981 /*
982 * Hop-by-Hop options header processing. If a valid jumbo payload option is
983 * included, the real payload length will be stored in plenp.
984 *
985 * rtalertp - XXX: should be stored more smart way
986 */
987 static int
988 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
989 struct mbuf **mp, int *offp)
990 {
991 struct mbuf *m = *mp;
992 int off = *offp, hbhlen;
993 struct ip6_hbh *hbh;
994
995 /* validation of the length of the header */
996 #ifndef PULLDOWN_TEST
997 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
998 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
999 hbhlen = (hbh->ip6h_len + 1) << 3;
1000
1001 IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1002 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1003 #else
1004 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1005 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1006 if (hbh == NULL) {
1007 V_ip6stat.ip6s_tooshort++;
1008 return -1;
1009 }
1010 hbhlen = (hbh->ip6h_len + 1) << 3;
1011 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1012 hbhlen);
1013 if (hbh == NULL) {
1014 V_ip6stat.ip6s_tooshort++;
1015 return -1;
1016 }
1017 #endif
1018 off += hbhlen;
1019 hbhlen -= sizeof(struct ip6_hbh);
1020 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1021 hbhlen, rtalertp, plenp) < 0)
1022 return (-1);
1023
1024 *offp = off;
1025 *mp = m;
1026 return (0);
1027 }
1028
1029 /*
1030 * Search header for all Hop-by-hop options and process each option.
1031 * This function is separate from ip6_hopopts_input() in order to
1032 * handle a case where the sending node itself process its hop-by-hop
1033 * options header. In such a case, the function is called from ip6_output().
1034 *
1035 * The function assumes that hbh header is located right after the IPv6 header
1036 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1037 * opthead + hbhlen is located in continuous memory region.
1038 */
1039 int
1040 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1041 u_int32_t *rtalertp, u_int32_t *plenp)
1042 {
1043 struct ip6_hdr *ip6;
1044 int optlen = 0;
1045 u_int8_t *opt = opthead;
1046 u_int16_t rtalert_val;
1047 u_int32_t jumboplen;
1048 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1049
1050 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1051 switch (*opt) {
1052 case IP6OPT_PAD1:
1053 optlen = 1;
1054 break;
1055 case IP6OPT_PADN:
1056 if (hbhlen < IP6OPT_MINLEN) {
1057 V_ip6stat.ip6s_toosmall++;
1058 goto bad;
1059 }
1060 optlen = *(opt + 1) + 2;
1061 break;
1062 case IP6OPT_ROUTER_ALERT:
1063 /* XXX may need check for alignment */
1064 if (hbhlen < IP6OPT_RTALERT_LEN) {
1065 V_ip6stat.ip6s_toosmall++;
1066 goto bad;
1067 }
1068 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1069 /* XXX stat */
1070 icmp6_error(m, ICMP6_PARAM_PROB,
1071 ICMP6_PARAMPROB_HEADER,
1072 erroff + opt + 1 - opthead);
1073 return (-1);
1074 }
1075 optlen = IP6OPT_RTALERT_LEN;
1076 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1077 *rtalertp = ntohs(rtalert_val);
1078 break;
1079 case IP6OPT_JUMBO:
1080 /* XXX may need check for alignment */
1081 if (hbhlen < IP6OPT_JUMBO_LEN) {
1082 V_ip6stat.ip6s_toosmall++;
1083 goto bad;
1084 }
1085 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1086 /* XXX stat */
1087 icmp6_error(m, ICMP6_PARAM_PROB,
1088 ICMP6_PARAMPROB_HEADER,
1089 erroff + opt + 1 - opthead);
1090 return (-1);
1091 }
1092 optlen = IP6OPT_JUMBO_LEN;
1093
1094 /*
1095 * IPv6 packets that have non 0 payload length
1096 * must not contain a jumbo payload option.
1097 */
1098 ip6 = mtod(m, struct ip6_hdr *);
1099 if (ip6->ip6_plen) {
1100 V_ip6stat.ip6s_badoptions++;
1101 icmp6_error(m, ICMP6_PARAM_PROB,
1102 ICMP6_PARAMPROB_HEADER,
1103 erroff + opt - opthead);
1104 return (-1);
1105 }
1106
1107 /*
1108 * We may see jumbolen in unaligned location, so
1109 * we'd need to perform bcopy().
1110 */
1111 bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1112 jumboplen = (u_int32_t)htonl(jumboplen);
1113
1114 #if 1
1115 /*
1116 * if there are multiple jumbo payload options,
1117 * *plenp will be non-zero and the packet will be
1118 * rejected.
1119 * the behavior may need some debate in ipngwg -
1120 * multiple options does not make sense, however,
1121 * there's no explicit mention in specification.
1122 */
1123 if (*plenp != 0) {
1124 V_ip6stat.ip6s_badoptions++;
1125 icmp6_error(m, ICMP6_PARAM_PROB,
1126 ICMP6_PARAMPROB_HEADER,
1127 erroff + opt + 2 - opthead);
1128 return (-1);
1129 }
1130 #endif
1131
1132 /*
1133 * jumbo payload length must be larger than 65535.
1134 */
1135 if (jumboplen <= IPV6_MAXPACKET) {
1136 V_ip6stat.ip6s_badoptions++;
1137 icmp6_error(m, ICMP6_PARAM_PROB,
1138 ICMP6_PARAMPROB_HEADER,
1139 erroff + opt + 2 - opthead);
1140 return (-1);
1141 }
1142 *plenp = jumboplen;
1143
1144 break;
1145 default: /* unknown option */
1146 if (hbhlen < IP6OPT_MINLEN) {
1147 V_ip6stat.ip6s_toosmall++;
1148 goto bad;
1149 }
1150 optlen = ip6_unknown_opt(opt, m,
1151 erroff + opt - opthead);
1152 if (optlen == -1)
1153 return (-1);
1154 optlen += 2;
1155 break;
1156 }
1157 }
1158
1159 return (0);
1160
1161 bad:
1162 m_freem(m);
1163 return (-1);
1164 }
1165
1166 /*
1167 * Unknown option processing.
1168 * The third argument `off' is the offset from the IPv6 header to the option,
1169 * which is necessary if the IPv6 header the and option header and IPv6 header
1170 * is not continuous in order to return an ICMPv6 error.
1171 */
1172 int
1173 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1174 {
1175 struct ip6_hdr *ip6;
1176
1177 switch (IP6OPT_TYPE(*optp)) {
1178 case IP6OPT_TYPE_SKIP: /* ignore the option */
1179 return ((int)*(optp + 1));
1180 case IP6OPT_TYPE_DISCARD: /* silently discard */
1181 m_freem(m);
1182 return (-1);
1183 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1184 V_ip6stat.ip6s_badoptions++;
1185 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1186 return (-1);
1187 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1188 V_ip6stat.ip6s_badoptions++;
1189 ip6 = mtod(m, struct ip6_hdr *);
1190 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1191 (m->m_flags & (M_BCAST|M_MCAST)))
1192 m_freem(m);
1193 else
1194 icmp6_error(m, ICMP6_PARAM_PROB,
1195 ICMP6_PARAMPROB_OPTION, off);
1196 return (-1);
1197 }
1198
1199 m_freem(m); /* XXX: NOTREACHED */
1200 return (-1);
1201 }
1202
1203 /*
1204 * Create the "control" list for this pcb.
1205 * These functions will not modify mbuf chain at all.
1206 *
1207 * With KAME mbuf chain restriction:
1208 * The routine will be called from upper layer handlers like tcp6_input().
1209 * Thus the routine assumes that the caller (tcp6_input) have already
1210 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1211 * very first mbuf on the mbuf chain.
1212 *
1213 * ip6_savecontrol_v4 will handle those options that are possible to be
1214 * set on a v4-mapped socket.
1215 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1216 * options and handle the v6-only ones itself.
1217 */
1218 struct mbuf **
1219 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1220 int *v4only)
1221 {
1222 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1223
1224 #ifdef SO_TIMESTAMP
1225 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1226 struct timeval tv;
1227
1228 microtime(&tv);
1229 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1230 SCM_TIMESTAMP, SOL_SOCKET);
1231 if (*mp)
1232 mp = &(*mp)->m_next;
1233 }
1234 #endif
1235
1236 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1237 /* RFC 2292 sec. 5 */
1238 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1239 struct in6_pktinfo pi6;
1240
1241 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1242 #ifdef INET
1243 struct ip *ip;
1244
1245 ip = mtod(m, struct ip *);
1246 pi6.ipi6_addr.s6_addr32[0] = 0;
1247 pi6.ipi6_addr.s6_addr32[1] = 0;
1248 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1249 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1250 #else
1251 /* We won't hit this code */
1252 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1253 #endif
1254 } else {
1255 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1256 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1257 }
1258 pi6.ipi6_ifindex =
1259 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1260
1261 *mp = sbcreatecontrol((caddr_t) &pi6,
1262 sizeof(struct in6_pktinfo),
1263 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1264 if (*mp)
1265 mp = &(*mp)->m_next;
1266 }
1267
1268 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1269 int hlim;
1270
1271 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1272 #ifdef INET
1273 struct ip *ip;
1274
1275 ip = mtod(m, struct ip *);
1276 hlim = ip->ip_ttl;
1277 #else
1278 /* We won't hit this code */
1279 hlim = 0;
1280 #endif
1281 } else {
1282 hlim = ip6->ip6_hlim & 0xff;
1283 }
1284 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1285 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1286 IPPROTO_IPV6);
1287 if (*mp)
1288 mp = &(*mp)->m_next;
1289 }
1290
1291 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1292 int tclass;
1293
1294 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1295 #ifdef INET
1296 struct ip *ip;
1297
1298 ip = mtod(m, struct ip *);
1299 tclass = ip->ip_tos;
1300 #else
1301 /* We won't hit this code */
1302 tclass = 0;
1303 #endif
1304 } else {
1305 u_int32_t flowinfo;
1306
1307 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1308 flowinfo >>= 20;
1309 tclass = flowinfo & 0xff;
1310 }
1311 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1312 IPV6_TCLASS, IPPROTO_IPV6);
1313 if (*mp)
1314 mp = &(*mp)->m_next;
1315 }
1316
1317 if (v4only != NULL) {
1318 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1319 *v4only = 1;
1320 } else {
1321 *v4only = 0;
1322 }
1323 }
1324
1325 return (mp);
1326 }
1327
1328 void
1329 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1330 {
1331 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1332 int v4only = 0;
1333
1334 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1335 if (v4only)
1336 return;
1337
1338 /*
1339 * IPV6_HOPOPTS socket option. Recall that we required super-user
1340 * privilege for the option (see ip6_ctloutput), but it might be too
1341 * strict, since there might be some hop-by-hop options which can be
1342 * returned to normal user.
1343 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1344 */
1345 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1346 /*
1347 * Check if a hop-by-hop options header is contatined in the
1348 * received packet, and if so, store the options as ancillary
1349 * data. Note that a hop-by-hop options header must be
1350 * just after the IPv6 header, which is assured through the
1351 * IPv6 input processing.
1352 */
1353 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1354 struct ip6_hbh *hbh;
1355 int hbhlen = 0;
1356 #ifdef PULLDOWN_TEST
1357 struct mbuf *ext;
1358 #endif
1359
1360 #ifndef PULLDOWN_TEST
1361 hbh = (struct ip6_hbh *)(ip6 + 1);
1362 hbhlen = (hbh->ip6h_len + 1) << 3;
1363 #else
1364 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1365 ip6->ip6_nxt);
1366 if (ext == NULL) {
1367 V_ip6stat.ip6s_tooshort++;
1368 return;
1369 }
1370 hbh = mtod(ext, struct ip6_hbh *);
1371 hbhlen = (hbh->ip6h_len + 1) << 3;
1372 if (hbhlen != ext->m_len) {
1373 m_freem(ext);
1374 V_ip6stat.ip6s_tooshort++;
1375 return;
1376 }
1377 #endif
1378
1379 /*
1380 * XXX: We copy the whole header even if a
1381 * jumbo payload option is included, the option which
1382 * is to be removed before returning according to
1383 * RFC2292.
1384 * Note: this constraint is removed in RFC3542
1385 */
1386 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1387 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1388 IPPROTO_IPV6);
1389 if (*mp)
1390 mp = &(*mp)->m_next;
1391 #ifdef PULLDOWN_TEST
1392 m_freem(ext);
1393 #endif
1394 }
1395 }
1396
1397 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1398 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1399
1400 /*
1401 * Search for destination options headers or routing
1402 * header(s) through the header chain, and stores each
1403 * header as ancillary data.
1404 * Note that the order of the headers remains in
1405 * the chain of ancillary data.
1406 */
1407 while (1) { /* is explicit loop prevention necessary? */
1408 struct ip6_ext *ip6e = NULL;
1409 int elen;
1410 #ifdef PULLDOWN_TEST
1411 struct mbuf *ext = NULL;
1412 #endif
1413
1414 /*
1415 * if it is not an extension header, don't try to
1416 * pull it from the chain.
1417 */
1418 switch (nxt) {
1419 case IPPROTO_DSTOPTS:
1420 case IPPROTO_ROUTING:
1421 case IPPROTO_HOPOPTS:
1422 case IPPROTO_AH: /* is it possible? */
1423 break;
1424 default:
1425 goto loopend;
1426 }
1427
1428 #ifndef PULLDOWN_TEST
1429 if (off + sizeof(*ip6e) > m->m_len)
1430 goto loopend;
1431 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1432 if (nxt == IPPROTO_AH)
1433 elen = (ip6e->ip6e_len + 2) << 2;
1434 else
1435 elen = (ip6e->ip6e_len + 1) << 3;
1436 if (off + elen > m->m_len)
1437 goto loopend;
1438 #else
1439 ext = ip6_pullexthdr(m, off, nxt);
1440 if (ext == NULL) {
1441 V_ip6stat.ip6s_tooshort++;
1442 return;
1443 }
1444 ip6e = mtod(ext, struct ip6_ext *);
1445 if (nxt == IPPROTO_AH)
1446 elen = (ip6e->ip6e_len + 2) << 2;
1447 else
1448 elen = (ip6e->ip6e_len + 1) << 3;
1449 if (elen != ext->m_len) {
1450 m_freem(ext);
1451 V_ip6stat.ip6s_tooshort++;
1452 return;
1453 }
1454 #endif
1455
1456 switch (nxt) {
1457 case IPPROTO_DSTOPTS:
1458 if (!(in6p->inp_flags & IN6P_DSTOPTS))
1459 break;
1460
1461 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1462 IS2292(in6p,
1463 IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1464 IPPROTO_IPV6);
1465 if (*mp)
1466 mp = &(*mp)->m_next;
1467 break;
1468 case IPPROTO_ROUTING:
1469 if (!in6p->inp_flags & IN6P_RTHDR)
1470 break;
1471
1472 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1473 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1474 IPPROTO_IPV6);
1475 if (*mp)
1476 mp = &(*mp)->m_next;
1477 break;
1478 case IPPROTO_HOPOPTS:
1479 case IPPROTO_AH: /* is it possible? */
1480 break;
1481
1482 default:
1483 /*
1484 * other cases have been filtered in the above.
1485 * none will visit this case. here we supply
1486 * the code just in case (nxt overwritten or
1487 * other cases).
1488 */
1489 #ifdef PULLDOWN_TEST
1490 m_freem(ext);
1491 #endif
1492 goto loopend;
1493
1494 }
1495
1496 /* proceed with the next header. */
1497 off += elen;
1498 nxt = ip6e->ip6e_nxt;
1499 ip6e = NULL;
1500 #ifdef PULLDOWN_TEST
1501 m_freem(ext);
1502 ext = NULL;
1503 #endif
1504 }
1505 loopend:
1506 ;
1507 }
1508 }
1509 #undef IS2292
1510
1511 void
1512 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1513 {
1514 struct socket *so;
1515 struct mbuf *m_mtu;
1516 struct ip6_mtuinfo mtuctl;
1517
1518 so = in6p->inp_socket;
1519
1520 if (mtu == NULL)
1521 return;
1522
1523 #ifdef DIAGNOSTIC
1524 if (so == NULL) /* I believe this is impossible */
1525 panic("ip6_notify_pmtu: socket is NULL");
1526 #endif
1527
1528 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */
1529 mtuctl.ip6m_mtu = *mtu;
1530 mtuctl.ip6m_addr = *dst;
1531 if (sa6_recoverscope(&mtuctl.ip6m_addr))
1532 return;
1533
1534 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1535 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1536 return;
1537
1538 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1539 == 0) {
1540 m_freem(m_mtu);
1541 /* XXX: should count statistics */
1542 } else
1543 sorwakeup(so);
1544
1545 return;
1546 }
1547
1548 #ifdef PULLDOWN_TEST
1549 /*
1550 * pull single extension header from mbuf chain. returns single mbuf that
1551 * contains the result, or NULL on error.
1552 */
1553 static struct mbuf *
1554 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1555 {
1556 struct ip6_ext ip6e;
1557 size_t elen;
1558 struct mbuf *n;
1559
1560 #ifdef DIAGNOSTIC
1561 switch (nxt) {
1562 case IPPROTO_DSTOPTS:
1563 case IPPROTO_ROUTING:
1564 case IPPROTO_HOPOPTS:
1565 case IPPROTO_AH: /* is it possible? */
1566 break;
1567 default:
1568 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1569 }
1570 #endif
1571
1572 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1573 if (nxt == IPPROTO_AH)
1574 elen = (ip6e.ip6e_len + 2) << 2;
1575 else
1576 elen = (ip6e.ip6e_len + 1) << 3;
1577
1578 MGET(n, M_DONTWAIT, MT_DATA);
1579 if (n && elen >= MLEN) {
1580 MCLGET(n, M_DONTWAIT);
1581 if ((n->m_flags & M_EXT) == 0) {
1582 m_free(n);
1583 n = NULL;
1584 }
1585 }
1586 if (!n)
1587 return NULL;
1588
1589 n->m_len = 0;
1590 if (elen >= M_TRAILINGSPACE(n)) {
1591 m_free(n);
1592 return NULL;
1593 }
1594
1595 m_copydata(m, off, elen, mtod(n, caddr_t));
1596 n->m_len = elen;
1597 return n;
1598 }
1599 #endif
1600
1601 /*
1602 * Get pointer to the previous header followed by the header
1603 * currently processed.
1604 * XXX: This function supposes that
1605 * M includes all headers,
1606 * the next header field and the header length field of each header
1607 * are valid, and
1608 * the sum of each header length equals to OFF.
1609 * Because of these assumptions, this function must be called very
1610 * carefully. Moreover, it will not be used in the near future when
1611 * we develop `neater' mechanism to process extension headers.
1612 */
1613 char *
1614 ip6_get_prevhdr(struct mbuf *m, int off)
1615 {
1616 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1617
1618 if (off == sizeof(struct ip6_hdr))
1619 return (&ip6->ip6_nxt);
1620 else {
1621 int len, nxt;
1622 struct ip6_ext *ip6e = NULL;
1623
1624 nxt = ip6->ip6_nxt;
1625 len = sizeof(struct ip6_hdr);
1626 while (len < off) {
1627 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1628
1629 switch (nxt) {
1630 case IPPROTO_FRAGMENT:
1631 len += sizeof(struct ip6_frag);
1632 break;
1633 case IPPROTO_AH:
1634 len += (ip6e->ip6e_len + 2) << 2;
1635 break;
1636 default:
1637 len += (ip6e->ip6e_len + 1) << 3;
1638 break;
1639 }
1640 nxt = ip6e->ip6e_nxt;
1641 }
1642 if (ip6e)
1643 return (&ip6e->ip6e_nxt);
1644 else
1645 return NULL;
1646 }
1647 }
1648
1649 /*
1650 * get next header offset. m will be retained.
1651 */
1652 int
1653 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1654 {
1655 struct ip6_hdr ip6;
1656 struct ip6_ext ip6e;
1657 struct ip6_frag fh;
1658
1659 /* just in case */
1660 if (m == NULL)
1661 panic("ip6_nexthdr: m == NULL");
1662 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1663 return -1;
1664
1665 switch (proto) {
1666 case IPPROTO_IPV6:
1667 if (m->m_pkthdr.len < off + sizeof(ip6))
1668 return -1;
1669 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1670 if (nxtp)
1671 *nxtp = ip6.ip6_nxt;
1672 off += sizeof(ip6);
1673 return off;
1674
1675 case IPPROTO_FRAGMENT:
1676 /*
1677 * terminate parsing if it is not the first fragment,
1678 * it does not make sense to parse through it.
1679 */
1680 if (m->m_pkthdr.len < off + sizeof(fh))
1681 return -1;
1682 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1683 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1684 if (fh.ip6f_offlg & IP6F_OFF_MASK)
1685 return -1;
1686 if (nxtp)
1687 *nxtp = fh.ip6f_nxt;
1688 off += sizeof(struct ip6_frag);
1689 return off;
1690
1691 case IPPROTO_AH:
1692 if (m->m_pkthdr.len < off + sizeof(ip6e))
1693 return -1;
1694 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1695 if (nxtp)
1696 *nxtp = ip6e.ip6e_nxt;
1697 off += (ip6e.ip6e_len + 2) << 2;
1698 return off;
1699
1700 case IPPROTO_HOPOPTS:
1701 case IPPROTO_ROUTING:
1702 case IPPROTO_DSTOPTS:
1703 if (m->m_pkthdr.len < off + sizeof(ip6e))
1704 return -1;
1705 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1706 if (nxtp)
1707 *nxtp = ip6e.ip6e_nxt;
1708 off += (ip6e.ip6e_len + 1) << 3;
1709 return off;
1710
1711 case IPPROTO_NONE:
1712 case IPPROTO_ESP:
1713 case IPPROTO_IPCOMP:
1714 /* give up */
1715 return -1;
1716
1717 default:
1718 return -1;
1719 }
1720
1721 return -1;
1722 }
1723
1724 /*
1725 * get offset for the last header in the chain. m will be kept untainted.
1726 */
1727 int
1728 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1729 {
1730 int newoff;
1731 int nxt;
1732
1733 if (!nxtp) {
1734 nxt = -1;
1735 nxtp = &nxt;
1736 }
1737 while (1) {
1738 newoff = ip6_nexthdr(m, off, proto, nxtp);
1739 if (newoff < 0)
1740 return off;
1741 else if (newoff < off)
1742 return -1; /* invalid */
1743 else if (newoff == off)
1744 return newoff;
1745
1746 off = newoff;
1747 proto = *nxtp;
1748 }
1749 }
1750
1751 struct ip6aux *
1752 ip6_addaux(struct mbuf *m)
1753 {
1754 struct m_tag *mtag;
1755
1756 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1757 if (!mtag) {
1758 mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1759 M_NOWAIT);
1760 if (mtag) {
1761 m_tag_prepend(m, mtag);
1762 bzero(mtag + 1, sizeof(struct ip6aux));
1763 }
1764 }
1765 return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1766 }
1767
1768 struct ip6aux *
1769 ip6_findaux(struct mbuf *m)
1770 {
1771 struct m_tag *mtag;
1772
1773 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1774 return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1775 }
1776
1777 void
1778 ip6_delaux(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 m_tag_delete(m, mtag);
1785 }
1786
1787 /*
1788 * System control for IP6
1789 */
1790
1791 u_char inet6ctlerrmap[PRC_NCMDS] = {
1792 0, 0, 0, 0,
1793 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1794 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1795 EMSGSIZE, EHOSTUNREACH, 0, 0,
1796 0, 0, 0, 0,
1797 ENOPROTOOPT
1798 };
Cache object: 20a19d51eddbbbb863b8a0cd5fa5322c
|