xref: /trueos/sys/netinet6/ip6_input.c (revision 17d83a70d11062ccf00ec19e142b61af05794ef2)
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_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_kdtrace.h"
71 #include "opt_route.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.h>
78 #include <sys/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/sdt.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/kernel.h>
86 #include <sys/syslog.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/if_dl.h>
91 #include <net/route.h>
92 #include <net/netisr.h>
93 #include <net/pfil.h>
94 #include <net/vnet.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_kdtrace.h>
98 #include <netinet/ip_var.h>
99 #include <netinet/in_systm.h>
100 #include <net/if_llatbl.h>
101 #ifdef INET
102 #include <netinet/ip.h>
103 #include <netinet/ip_icmp.h>
104 #endif /* INET */
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/icmp6.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/in6_ifattach.h>
112 #include <netinet6/nd6.h>
113 
114 #ifdef IPSEC
115 #include <netipsec/ipsec.h>
116 #include <netinet6/ip6_ipsec.h>
117 #include <netipsec/ipsec6.h>
118 #endif /* IPSEC */
119 
120 #include <netinet6/ip6protosw.h>
121 
122 extern struct domain inet6domain;
123 
124 u_char ip6_protox[IPPROTO_MAX];
125 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
126 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
127 VNET_DEFINE(u_long, in6_ifaddrhmask);
128 
129 static struct netisr_handler ip6_nh = {
130 	.nh_name = "ip6",
131 	.nh_handler = ip6_input,
132 	.nh_proto = NETISR_IPV6,
133 	.nh_policy = NETISR_POLICY_FLOW,
134 };
135 
136 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
137 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
138 
139 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
140 
141 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
142 VNET_PCPUSTAT_SYSINIT(ip6stat);
143 #ifdef VIMAGE
144 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
145 #endif /* VIMAGE */
146 
147 struct rwlock in6_ifaddr_lock;
148 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
149 
150 static void ip6_init2(void *);
151 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
152 static struct ip6aux *ip6_addaux(struct mbuf *);
153 static struct ip6aux *ip6_findaux(struct mbuf *m);
154 static void ip6_delaux (struct mbuf *);
155 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
156 #ifdef PULLDOWN_TEST
157 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
158 #endif
159 
160 /*
161  * IP6 initialization: fill in IP6 protocol switch table.
162  * All protocols not implemented in kernel go to raw IP6 protocol handler.
163  */
164 void
ip6_init(void)165 ip6_init(void)
166 {
167 	struct ip6protosw *pr;
168 	int i;
169 
170 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
171 	    &V_ip6_auto_linklocal);
172 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
173 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
174 
175 	TAILQ_INIT(&V_in6_ifaddrhead);
176 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
177 	    &V_in6_ifaddrhmask);
178 
179 	/* Initialize packet filter hooks. */
180 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
181 	V_inet6_pfil_hook.ph_af = AF_INET6;
182 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
183 		printf("%s: WARNING: unable to register pfil hook, "
184 			"error %d\n", __func__, i);
185 
186 	scope6_init();
187 	addrsel_policy_init();
188 	nd6_init();
189 	frag6_init();
190 
191 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
192 
193 	/* Skip global initialization stuff for non-default instances. */
194 	if (!IS_DEFAULT_VNET(curvnet))
195 		return;
196 
197 #ifdef DIAGNOSTIC
198 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
199 		panic("sizeof(protosw) != sizeof(ip6protosw)");
200 #endif
201 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
202 	if (pr == NULL)
203 		panic("ip6_init");
204 
205 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
206 	for (i = 0; i < IPPROTO_MAX; i++)
207 		ip6_protox[i] = pr - inet6sw;
208 	/*
209 	 * Cycle through IP protocols and put them into the appropriate place
210 	 * in ip6_protox[].
211 	 */
212 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
213 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
214 		if (pr->pr_domain->dom_family == PF_INET6 &&
215 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
216 			/* Be careful to only index valid IP protocols. */
217 			if (pr->pr_protocol < IPPROTO_MAX)
218 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
219 		}
220 
221 	netisr_register(&ip6_nh);
222 }
223 
224 /*
225  * The protocol to be inserted into ip6_protox[] must be already registered
226  * in inet6sw[], either statically or through pf_proto_register().
227  */
228 int
ip6proto_register(short ip6proto)229 ip6proto_register(short ip6proto)
230 {
231 	struct ip6protosw *pr;
232 
233 	/* Sanity checks. */
234 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
235 		return (EPROTONOSUPPORT);
236 
237 	/*
238 	 * The protocol slot must not be occupied by another protocol
239 	 * already.  An index pointing to IPPROTO_RAW is unused.
240 	 */
241 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
242 	if (pr == NULL)
243 		return (EPFNOSUPPORT);
244 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
245 		return (EEXIST);
246 
247 	/*
248 	 * Find the protocol position in inet6sw[] and set the index.
249 	 */
250 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
251 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
252 		if (pr->pr_domain->dom_family == PF_INET6 &&
253 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
254 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
255 			return (0);
256 		}
257 	}
258 	return (EPROTONOSUPPORT);
259 }
260 
261 int
ip6proto_unregister(short ip6proto)262 ip6proto_unregister(short ip6proto)
263 {
264 	struct ip6protosw *pr;
265 
266 	/* Sanity checks. */
267 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
268 		return (EPROTONOSUPPORT);
269 
270 	/* Check if the protocol was indeed registered. */
271 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
272 	if (pr == NULL)
273 		return (EPFNOSUPPORT);
274 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
275 		return (ENOENT);
276 
277 	/* Reset the protocol slot to IPPROTO_RAW. */
278 	ip6_protox[ip6proto] = pr - inet6sw;
279 	return (0);
280 }
281 
282 #ifdef VIMAGE
283 void
ip6_destroy()284 ip6_destroy()
285 {
286 	int i;
287 
288 	if ((i = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
289 		printf("%s: WARNING: unable to unregister pfil hook, "
290 		    "error %d\n", __func__, i);
291 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
292 	nd6_destroy();
293 	callout_drain(&V_in6_tmpaddrtimer_ch);
294 }
295 #endif
296 
297 static int
ip6_init2_vnet(const void * unused __unused)298 ip6_init2_vnet(const void *unused __unused)
299 {
300 
301 	/* nd6_timer_init */
302 	callout_init(&V_nd6_timer_ch, 0);
303 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
304 
305 	/* timer for regeneranation of temporary addresses randomize ID */
306 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
307 	callout_reset(&V_in6_tmpaddrtimer_ch,
308 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
309 		       V_ip6_temp_regen_advance) * hz,
310 		      in6_tmpaddrtimer, curvnet);
311 
312 	return (0);
313 }
314 
315 static void
ip6_init2(void * dummy)316 ip6_init2(void *dummy)
317 {
318 
319 	ip6_init2_vnet(NULL);
320 }
321 
322 /* cheat */
323 /* This must be after route_init(), which is now SI_ORDER_THIRD */
324 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
325 
326 static int
ip6_input_hbh(struct mbuf * m,uint32_t * plen,uint32_t * rtalert,int * off,int * nxt,int * ours)327 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
328     int *nxt, int *ours)
329 {
330 	struct ip6_hdr *ip6;
331 	struct ip6_hbh *hbh;
332 
333 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
334 #if 0	/*touches NULL pointer*/
335 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
336 #endif
337 		goto out;	/* m have already been freed */
338 	}
339 
340 	/* adjust pointer */
341 	ip6 = mtod(m, struct ip6_hdr *);
342 
343 	/*
344 	 * if the payload length field is 0 and the next header field
345 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
346 	 * option MUST be included.
347 	 */
348 	if (ip6->ip6_plen == 0 && *plen == 0) {
349 		/*
350 		 * Note that if a valid jumbo payload option is
351 		 * contained, ip6_hopopts_input() must set a valid
352 		 * (non-zero) payload length to the variable plen.
353 		 */
354 		IP6STAT_INC(ip6s_badoptions);
355 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
356 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
357 		icmp6_error(m, ICMP6_PARAM_PROB,
358 			    ICMP6_PARAMPROB_HEADER,
359 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
360 		goto out;
361 	}
362 #ifndef PULLDOWN_TEST
363 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
364 	hbh = (struct ip6_hbh *)(ip6 + 1);
365 #else
366 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
367 		sizeof(struct ip6_hbh));
368 	if (hbh == NULL) {
369 		IP6STAT_INC(ip6s_tooshort);
370 		goto out;
371 	}
372 #endif
373 	*nxt = hbh->ip6h_nxt;
374 
375 	/*
376 	 * If we are acting as a router and the packet contains a
377 	 * router alert option, see if we know the option value.
378 	 * Currently, we only support the option value for MLD, in which
379 	 * case we should pass the packet to the multicast routing
380 	 * daemon.
381 	 */
382 	if (*rtalert != ~0) {
383 		switch (*rtalert) {
384 		case IP6OPT_RTALERT_MLD:
385 			if (V_ip6_forwarding)
386 				*ours = 1;
387 			break;
388 		default:
389 			/*
390 			 * RFC2711 requires unrecognized values must be
391 			 * silently ignored.
392 			 */
393 			break;
394 		}
395 	}
396 
397 	return (0);
398 
399 out:
400 	return (1);
401 }
402 
403 void
ip6_input(struct mbuf * m)404 ip6_input(struct mbuf *m)
405 {
406 	struct ip6_hdr *ip6;
407 	int off = sizeof(struct ip6_hdr), nest;
408 	u_int32_t plen;
409 	u_int32_t rtalert = ~0;
410 	int nxt, ours = 0;
411 	struct ifnet *deliverifp = NULL, *ifp = NULL;
412 	struct in6_addr odst;
413 	struct route_in6 rin6;
414 	int srcrt = 0;
415 	struct llentry *lle = NULL;
416 	struct sockaddr_in6 dst6, *dst;
417 
418 	bzero(&rin6, sizeof(struct route_in6));
419 #ifdef IPSEC
420 	/*
421 	 * should the inner packet be considered authentic?
422 	 * see comment in ah4_input().
423 	 * NB: m cannot be NULL when passed to the input routine
424 	 */
425 
426 	m->m_flags &= ~M_AUTHIPHDR;
427 	m->m_flags &= ~M_AUTHIPDGM;
428 
429 #endif /* IPSEC */
430 
431 	/*
432 	 * make sure we don't have onion peering information into m_tag.
433 	 */
434 	ip6_delaux(m);
435 
436 	if (m->m_flags & M_FASTFWD_OURS) {
437 		/*
438 		 * Firewall changed destination to local.
439 		 */
440 		m->m_flags &= ~M_FASTFWD_OURS;
441 		ours = 1;
442 		deliverifp = m->m_pkthdr.rcvif;
443 		ip6 = mtod(m, struct ip6_hdr *);
444 		goto hbhcheck;
445 	}
446 
447 	/*
448 	 * mbuf statistics
449 	 */
450 	if (m->m_flags & M_EXT) {
451 		if (m->m_next)
452 			IP6STAT_INC(ip6s_mext2m);
453 		else
454 			IP6STAT_INC(ip6s_mext1);
455 	} else {
456 		if (m->m_next) {
457 			if (m->m_flags & M_LOOP) {
458 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
459 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
460 				IP6STAT_INC(
461 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
462 			else
463 				IP6STAT_INC(ip6s_m2m[0]);
464 		} else
465 			IP6STAT_INC(ip6s_m1);
466 	}
467 
468 	/* drop the packet if IPv6 operation is disabled on the IF */
469 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
470 		m_freem(m);
471 		return;
472 	}
473 
474 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
475 	IP6STAT_INC(ip6s_total);
476 
477 #ifndef PULLDOWN_TEST
478 	/*
479 	 * L2 bridge code and some other code can return mbuf chain
480 	 * that does not conform to KAME requirement.  too bad.
481 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
482 	 */
483 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
484 		struct mbuf *n;
485 
486 		if (m->m_pkthdr.len > MHLEN)
487 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
488 		else
489 			n = m_gethdr(M_NOWAIT, MT_DATA);
490 		if (n == NULL) {
491 			m_freem(m);
492 			return;	/* ENOBUFS */
493 		}
494 
495 		m_move_pkthdr(n, m);
496 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
497 		n->m_len = n->m_pkthdr.len;
498 		m_freem(m);
499 		m = n;
500 	}
501 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
502 #endif
503 
504 	if (m->m_len < sizeof(struct ip6_hdr)) {
505 		struct ifnet *inifp;
506 		inifp = m->m_pkthdr.rcvif;
507 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
508 			IP6STAT_INC(ip6s_toosmall);
509 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
510 			return;
511 		}
512 	}
513 
514 	ip6 = mtod(m, struct ip6_hdr *);
515 
516 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
517 		IP6STAT_INC(ip6s_badvers);
518 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
519 		goto bad;
520 	}
521 
522 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
523 
524 	IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6);
525 
526 	/*
527 	 * Check against address spoofing/corruption.
528 	 */
529 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
530 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
531 		/*
532 		 * XXX: "badscope" is not very suitable for a multicast source.
533 		 */
534 		IP6STAT_INC(ip6s_badscope);
535 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
536 		goto bad;
537 	}
538 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
539 	    !(m->m_flags & M_LOOP)) {
540 		/*
541 		 * In this case, the packet should come from the loopback
542 		 * interface.  However, we cannot just check the if_flags,
543 		 * because ip6_mloopback() passes the "actual" interface
544 		 * as the outgoing/incoming interface.
545 		 */
546 		IP6STAT_INC(ip6s_badscope);
547 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
548 		goto bad;
549 	}
550 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
551 	    IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
552 		/*
553 		 * RFC4291 2.7:
554 		 * Nodes must not originate a packet to a multicast address
555 		 * whose scop field contains the reserved value 0; if such
556 		 * a packet is received, it must be silently dropped.
557 		 */
558 		IP6STAT_INC(ip6s_badscope);
559 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
560 		goto bad;
561 	}
562 #ifdef ALTQ
563 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
564 		/* packet is dropped by traffic conditioner */
565 		return;
566 	}
567 #endif
568 	/*
569 	 * The following check is not documented in specs.  A malicious
570 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
571 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
572 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
573 	 *
574 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
575 	 * support IPv4-less kernel compilation, we cannot support SIIT
576 	 * environment at all.  So, it makes more sense for us to reject any
577 	 * malicious packets for non-SIIT environment, than try to do a
578 	 * partial support for SIIT environment.
579 	 */
580 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
581 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
582 		IP6STAT_INC(ip6s_badscope);
583 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
584 		goto bad;
585 	}
586 #if 0
587 	/*
588 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
589 	 *
590 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
591 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
592 	 * is revised to forbid relaying case.
593 	 */
594 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
595 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
596 		IP6STAT_INC(ip6s_badscope);
597 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
598 		goto bad;
599 	}
600 #endif
601 #ifdef IPSEC
602 	/*
603 	 * Bypass packet filtering for packets previously handled by IPsec.
604 	 */
605 	if (ip6_ipsec_filtertunnel(m))
606 		goto passin;
607 #endif /* IPSEC */
608 
609 	/*
610 	 * Run through list of hooks for input packets.
611 	 *
612 	 * NB: Beware of the destination address changing
613 	 *     (e.g. by NAT rewriting).  When this happens,
614 	 *     tell ip6_forward to do the right thing.
615 	 */
616 	odst = ip6->ip6_dst;
617 
618 	/* Jump over all PFIL processing if hooks are not active. */
619 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
620 		goto passin;
621 
622 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
623 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
624 		return;
625 	if (m == NULL)			/* consumed by filter */
626 		return;
627 	ip6 = mtod(m, struct ip6_hdr *);
628 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
629 
630 	if (m->m_flags & M_FASTFWD_OURS) {
631 		m->m_flags &= ~M_FASTFWD_OURS;
632 		ours = 1;
633 		deliverifp = m->m_pkthdr.rcvif;
634 		goto hbhcheck;
635 	}
636 	if ((m->m_flags & M_IP6_NEXTHOP) &&
637 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
638 		/*
639 		 * Directly ship the packet on.  This allows forwarding
640 		 * packets originally destined to us to some other directly
641 		 * connected host.
642 		 */
643 		ip6_forward(m, 1);
644 		goto out;
645 	}
646 
647 passin:
648 	/*
649 	 * Disambiguate address scope zones (if there is ambiguity).
650 	 * We first make sure that the original source or destination address
651 	 * is not in our internal form for scoped addresses.  Such addresses
652 	 * are not necessarily invalid spec-wise, but we cannot accept them due
653 	 * to the usage conflict.
654 	 * in6_setscope() then also checks and rejects the cases where src or
655 	 * dst are the loopback address and the receiving interface
656 	 * is not loopback.
657 	 */
658 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
659 		IP6STAT_INC(ip6s_badscope); /* XXX */
660 		goto bad;
661 	}
662 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
663 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
664 		IP6STAT_INC(ip6s_badscope);
665 		goto bad;
666 	}
667 
668 	/*
669 	 * Multicast check. Assume packet is for us to avoid
670 	 * prematurely taking locks.
671 	 */
672 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
673 		ours = 1;
674 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
675 		deliverifp = m->m_pkthdr.rcvif;
676 		goto hbhcheck;
677 	}
678 
679 	/*
680 	 *  Unicast check
681 	 */
682 
683 	bzero(&dst6, sizeof(dst6));
684 	dst6.sin6_family = AF_INET6;
685 	dst6.sin6_len = sizeof(struct sockaddr_in6);
686 	dst6.sin6_addr = ip6->ip6_dst;
687 	ifp = m->m_pkthdr.rcvif;
688 	IF_AFDATA_RLOCK(ifp);
689 	lle = lla_lookup(LLTABLE6(ifp), 0,
690 	     (struct sockaddr *)&dst6);
691 	IF_AFDATA_RUNLOCK(ifp);
692 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
693 		struct ifaddr *ifa;
694 		struct in6_ifaddr *ia6;
695 		int bad;
696 
697 		bad = 1;
698 #define	sa_equal(a1, a2)						\
699 	(bcmp((a1), (a2), ((a1))->sin6_len) == 0)
700 		IF_ADDR_RLOCK(ifp);
701 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
702 			if (ifa->ifa_addr->sa_family != dst6.sin6_family)
703 				continue;
704 			if (sa_equal(&dst6, ifa->ifa_addr))
705 				break;
706 		}
707 		KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
708 		    __func__, lle));
709 #undef sa_equal
710 
711 		ia6 = (struct in6_ifaddr *)ifa;
712 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
713 			/* Count the packet in the ip address stats */
714 			ia6->ia_ifa.if_ipackets++;
715 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
716 
717 			/*
718 			 * record address information into m_tag.
719 			 */
720 			(void)ip6_setdstifaddr(m, ia6);
721 
722 			bad = 0;
723 		} else {
724 			char ip6bufs[INET6_ADDRSTRLEN];
725 			char ip6bufd[INET6_ADDRSTRLEN];
726 			/* address is not ready, so discard the packet. */
727 			nd6log((LOG_INFO,
728 			    "ip6_input: packet to an unready address %s->%s\n",
729 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
730 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
731 		}
732 		IF_ADDR_RUNLOCK(ifp);
733 		LLE_RUNLOCK(lle);
734 		if (bad)
735 			goto bad;
736 		else {
737 			ours = 1;
738 			deliverifp = ifp;
739 			goto hbhcheck;
740 		}
741 	}
742 	if (lle != NULL)
743 		LLE_RUNLOCK(lle);
744 
745 	dst = &rin6.ro_dst;
746 	dst->sin6_len = sizeof(struct sockaddr_in6);
747 	dst->sin6_family = AF_INET6;
748 	dst->sin6_addr = ip6->ip6_dst;
749 	rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
750 	if (rin6.ro_rt)
751 		RT_UNLOCK(rin6.ro_rt);
752 
753 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
754 
755 	/*
756 	 * Accept the packet if the forwarding interface to the destination
757 	 * according to the routing table is the loopback interface,
758 	 * unless the associated route has a gateway.
759 	 * Note that this approach causes to accept a packet if there is a
760 	 * route to the loopback interface for the destination of the packet.
761 	 * But we think it's even useful in some situations, e.g. when using
762 	 * a special daemon which wants to intercept the packet.
763 	 *
764 	 * XXX: some OSes automatically make a cloned route for the destination
765 	 * of an outgoing packet.  If the outgoing interface of the packet
766 	 * is a loopback one, the kernel would consider the packet to be
767 	 * accepted, even if we have no such address assinged on the interface.
768 	 * We check the cloned flag of the route entry to reject such cases,
769 	 * assuming that route entries for our own addresses are not made by
770 	 * cloning (it should be true because in6_addloop explicitly installs
771 	 * the host route).  However, we might have to do an explicit check
772 	 * while it would be less efficient.  Or, should we rather install a
773 	 * reject route for such a case?
774 	 */
775 	if (rin6.ro_rt &&
776 	    (rin6.ro_rt->rt_flags &
777 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
778 #ifdef RTF_WASCLONED
779 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
780 #endif
781 #ifdef RTF_CLONED
782 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
783 #endif
784 #if 0
785 	    /*
786 	     * The check below is redundant since the comparison of
787 	     * the destination and the key of the rtentry has
788 	     * already done through looking up the routing table.
789 	     */
790 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
791 	    &rt6_key(rin6.ro_rt)->sin6_addr)
792 #endif
793 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
794 		int free_ia6 = 0;
795 		struct in6_ifaddr *ia6;
796 
797 		/*
798 		 * found the loopback route to the interface address
799 		 */
800 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
801 			struct sockaddr_in6 dest6;
802 
803 			bzero(&dest6, sizeof(dest6));
804 			dest6.sin6_family = AF_INET6;
805 			dest6.sin6_len = sizeof(dest6);
806 			dest6.sin6_addr = ip6->ip6_dst;
807 			ia6 = (struct in6_ifaddr *)
808 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
809 			if (ia6 == NULL)
810 				goto bad;
811 			free_ia6 = 1;
812 		}
813 		else
814 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
815 
816 		/*
817 		 * record address information into m_tag.
818 		 */
819 		(void)ip6_setdstifaddr(m, ia6);
820 
821 		/*
822 		 * packets to a tentative, duplicated, or somehow invalid
823 		 * address must not be accepted.
824 		 */
825 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
826 			/* this address is ready */
827 			ours = 1;
828 			deliverifp = ia6->ia_ifp;	/* correct? */
829 			/* Count the packet in the ip address stats */
830 			ia6->ia_ifa.if_ipackets++;
831 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
832 			if (ia6 != NULL && free_ia6 != 0)
833 				ifa_free(&ia6->ia_ifa);
834 			goto hbhcheck;
835 		} else {
836 			char ip6bufs[INET6_ADDRSTRLEN];
837 			char ip6bufd[INET6_ADDRSTRLEN];
838 			/* address is not ready, so discard the packet. */
839 			nd6log((LOG_INFO,
840 			    "ip6_input: packet to an unready address %s->%s\n",
841 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
842 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
843 
844 			if (ia6 != NULL && free_ia6 != 0)
845 				ifa_free(&ia6->ia_ifa);
846 			goto bad;
847 		}
848 	}
849 
850 	/*
851 	 * FAITH (Firewall Aided Internet Translator)
852 	 */
853 	if (V_ip6_keepfaith) {
854 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
855 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
856 			/* XXX do we need more sanity checks? */
857 			ours = 1;
858 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
859 			goto hbhcheck;
860 		}
861 	}
862 
863 	/*
864 	 * Now there is no reason to process the packet if it's not our own
865 	 * and we're not a router.
866 	 */
867 	if (!V_ip6_forwarding) {
868 		IP6STAT_INC(ip6s_cantforward);
869 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
870 		goto bad;
871 	}
872 
873   hbhcheck:
874 	/*
875 	 * record address information into m_tag, if we don't have one yet.
876 	 * note that we are unable to record it, if the address is not listed
877 	 * as our interface address (e.g. multicast addresses, addresses
878 	 * within FAITH prefixes and such).
879 	 */
880 	if (deliverifp) {
881 		struct in6_ifaddr *ia6;
882 
883  		if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
884 			ifa_free(&ia6->ia_ifa);
885 		} else {
886 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
887 			if (ia6) {
888 				if (!ip6_setdstifaddr(m, ia6)) {
889 					/*
890 					 * XXX maybe we should drop the packet here,
891 					 * as we could not provide enough information
892 					 * to the upper layers.
893 					 */
894 				}
895 				ifa_free(&ia6->ia_ifa);
896 			}
897 		}
898 	}
899 
900 	/*
901 	 * Process Hop-by-Hop options header if it's contained.
902 	 * m may be modified in ip6_hopopts_input().
903 	 * If a JumboPayload option is included, plen will also be modified.
904 	 */
905 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
906 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
907 		int error;
908 
909 		error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours);
910 		if (error != 0)
911 			goto out;
912 	} else
913 		nxt = ip6->ip6_nxt;
914 
915 	/*
916 	 * Check that the amount of data in the buffers
917 	 * is as at least much as the IPv6 header would have us expect.
918 	 * Trim mbufs if longer than we expect.
919 	 * Drop packet if shorter than we expect.
920 	 */
921 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
922 		IP6STAT_INC(ip6s_tooshort);
923 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
924 		goto bad;
925 	}
926 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
927 		if (m->m_len == m->m_pkthdr.len) {
928 			m->m_len = sizeof(struct ip6_hdr) + plen;
929 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
930 		} else
931 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
932 	}
933 
934 	/*
935 	 * Forward if desirable.
936 	 */
937 	if (V_ip6_mrouter &&
938 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
939 		/*
940 		 * If we are acting as a multicast router, all
941 		 * incoming multicast packets are passed to the
942 		 * kernel-level multicast forwarding function.
943 		 * The packet is returned (relatively) intact; if
944 		 * ip6_mforward() returns a non-zero value, the packet
945 		 * must be discarded, else it may be accepted below.
946 		 *
947 		 * XXX TODO: Check hlim and multicast scope here to avoid
948 		 * unnecessarily calling into ip6_mforward().
949 		 */
950 		if (ip6_mforward &&
951 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
952 			IP6STAT_INC(ip6s_cantforward);
953 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
954 			goto bad;
955 		}
956 	} else if (!ours) {
957 		ip6_forward(m, srcrt);
958 		goto out;
959 	}
960 
961 	ip6 = mtod(m, struct ip6_hdr *);
962 
963 	/*
964 	 * Malicious party may be able to use IPv4 mapped addr to confuse
965 	 * tcp/udp stack and bypass security checks (act as if it was from
966 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
967 	 *
968 	 * For SIIT end node behavior, you may want to disable the check.
969 	 * However, you will  become vulnerable to attacks using IPv4 mapped
970 	 * source.
971 	 */
972 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
973 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
974 		IP6STAT_INC(ip6s_badscope);
975 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
976 		goto bad;
977 	}
978 
979 	/*
980 	 * Tell launch routine the next header
981 	 */
982 	IP6STAT_INC(ip6s_delivered);
983 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
984 	nest = 0;
985 
986 	while (nxt != IPPROTO_DONE) {
987 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
988 			IP6STAT_INC(ip6s_toomanyhdr);
989 			goto bad;
990 		}
991 
992 		/*
993 		 * protection against faulty packet - there should be
994 		 * more sanity checks in header chain processing.
995 		 */
996 		if (m->m_pkthdr.len < off) {
997 			IP6STAT_INC(ip6s_tooshort);
998 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
999 			goto bad;
1000 		}
1001 
1002 #ifdef IPSEC
1003 		/*
1004 		 * enforce IPsec policy checking if we are seeing last header.
1005 		 * note that we do not visit this with protocols with pcb layer
1006 		 * code - like udp/tcp/raw ip.
1007 		 */
1008 		if (ip6_ipsec_input(m, nxt))
1009 			goto bad;
1010 #endif /* IPSEC */
1011 
1012 		/*
1013 		 * Use mbuf flags to propagate Router Alert option to
1014 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
1015 		 */
1016 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
1017 			m->m_flags |= M_RTALERT_MLD;
1018 
1019 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1020 	}
1021 	goto out;
1022 bad:
1023 	m_freem(m);
1024 out:
1025 	if (rin6.ro_rt)
1026 		RTFREE(rin6.ro_rt);
1027 }
1028 
1029 /*
1030  * set/grab in6_ifaddr correspond to IPv6 destination address.
1031  * XXX backward compatibility wrapper
1032  *
1033  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1034  * and then bump it when the tag is copied, and release it when the tag is
1035  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
1036  * we just bump the ia refcount when we receive it.  This should be fixed.
1037  */
1038 static struct ip6aux *
ip6_setdstifaddr(struct mbuf * m,struct in6_ifaddr * ia6)1039 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1040 {
1041 	struct ip6aux *ip6a;
1042 
1043 	ip6a = ip6_addaux(m);
1044 	if (ip6a)
1045 		ip6a->ip6a_dstia6 = ia6;
1046 	return ip6a;	/* NULL if failed to set */
1047 }
1048 
1049 struct in6_ifaddr *
ip6_getdstifaddr(struct mbuf * m)1050 ip6_getdstifaddr(struct mbuf *m)
1051 {
1052 	struct ip6aux *ip6a;
1053 	struct in6_ifaddr *ia;
1054 
1055 	ip6a = ip6_findaux(m);
1056 	if (ip6a) {
1057 		ia = ip6a->ip6a_dstia6;
1058 		ifa_ref(&ia->ia_ifa);
1059 		return ia;
1060 	} else
1061 		return NULL;
1062 }
1063 
1064 /*
1065  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1066  * included, the real payload length will be stored in plenp.
1067  *
1068  * rtalertp - XXX: should be stored more smart way
1069  */
1070 static int
ip6_hopopts_input(u_int32_t * plenp,u_int32_t * rtalertp,struct mbuf ** mp,int * offp)1071 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1072     struct mbuf **mp, int *offp)
1073 {
1074 	struct mbuf *m = *mp;
1075 	int off = *offp, hbhlen;
1076 	struct ip6_hbh *hbh;
1077 
1078 	/* validation of the length of the header */
1079 #ifndef PULLDOWN_TEST
1080 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1081 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1082 	hbhlen = (hbh->ip6h_len + 1) << 3;
1083 
1084 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1085 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1086 #else
1087 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1088 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1089 	if (hbh == NULL) {
1090 		IP6STAT_INC(ip6s_tooshort);
1091 		return -1;
1092 	}
1093 	hbhlen = (hbh->ip6h_len + 1) << 3;
1094 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1095 		hbhlen);
1096 	if (hbh == NULL) {
1097 		IP6STAT_INC(ip6s_tooshort);
1098 		return -1;
1099 	}
1100 #endif
1101 	off += hbhlen;
1102 	hbhlen -= sizeof(struct ip6_hbh);
1103 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1104 				hbhlen, rtalertp, plenp) < 0)
1105 		return (-1);
1106 
1107 	*offp = off;
1108 	*mp = m;
1109 	return (0);
1110 }
1111 
1112 /*
1113  * Search header for all Hop-by-hop options and process each option.
1114  * This function is separate from ip6_hopopts_input() in order to
1115  * handle a case where the sending node itself process its hop-by-hop
1116  * options header. In such a case, the function is called from ip6_output().
1117  *
1118  * The function assumes that hbh header is located right after the IPv6 header
1119  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1120  * opthead + hbhlen is located in contiguous memory region.
1121  */
1122 int
ip6_process_hopopts(struct mbuf * m,u_int8_t * opthead,int hbhlen,u_int32_t * rtalertp,u_int32_t * plenp)1123 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1124     u_int32_t *rtalertp, u_int32_t *plenp)
1125 {
1126 	struct ip6_hdr *ip6;
1127 	int optlen = 0;
1128 	u_int8_t *opt = opthead;
1129 	u_int16_t rtalert_val;
1130 	u_int32_t jumboplen;
1131 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1132 
1133 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1134 		switch (*opt) {
1135 		case IP6OPT_PAD1:
1136 			optlen = 1;
1137 			break;
1138 		case IP6OPT_PADN:
1139 			if (hbhlen < IP6OPT_MINLEN) {
1140 				IP6STAT_INC(ip6s_toosmall);
1141 				goto bad;
1142 			}
1143 			optlen = *(opt + 1) + 2;
1144 			break;
1145 		case IP6OPT_ROUTER_ALERT:
1146 			/* XXX may need check for alignment */
1147 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1148 				IP6STAT_INC(ip6s_toosmall);
1149 				goto bad;
1150 			}
1151 			if (*(opt + 1) != IP6OPT_RTALERT_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_RTALERT_LEN;
1159 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1160 			*rtalertp = ntohs(rtalert_val);
1161 			break;
1162 		case IP6OPT_JUMBO:
1163 			/* XXX may need check for alignment */
1164 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1165 				IP6STAT_INC(ip6s_toosmall);
1166 				goto bad;
1167 			}
1168 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1169 				/* XXX stat */
1170 				icmp6_error(m, ICMP6_PARAM_PROB,
1171 				    ICMP6_PARAMPROB_HEADER,
1172 				    erroff + opt + 1 - opthead);
1173 				return (-1);
1174 			}
1175 			optlen = IP6OPT_JUMBO_LEN;
1176 
1177 			/*
1178 			 * IPv6 packets that have non 0 payload length
1179 			 * must not contain a jumbo payload option.
1180 			 */
1181 			ip6 = mtod(m, struct ip6_hdr *);
1182 			if (ip6->ip6_plen) {
1183 				IP6STAT_INC(ip6s_badoptions);
1184 				icmp6_error(m, ICMP6_PARAM_PROB,
1185 				    ICMP6_PARAMPROB_HEADER,
1186 				    erroff + opt - opthead);
1187 				return (-1);
1188 			}
1189 
1190 			/*
1191 			 * We may see jumbolen in unaligned location, so
1192 			 * we'd need to perform bcopy().
1193 			 */
1194 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1195 			jumboplen = (u_int32_t)htonl(jumboplen);
1196 
1197 #if 1
1198 			/*
1199 			 * if there are multiple jumbo payload options,
1200 			 * *plenp will be non-zero and the packet will be
1201 			 * rejected.
1202 			 * the behavior may need some debate in ipngwg -
1203 			 * multiple options does not make sense, however,
1204 			 * there's no explicit mention in specification.
1205 			 */
1206 			if (*plenp != 0) {
1207 				IP6STAT_INC(ip6s_badoptions);
1208 				icmp6_error(m, ICMP6_PARAM_PROB,
1209 				    ICMP6_PARAMPROB_HEADER,
1210 				    erroff + opt + 2 - opthead);
1211 				return (-1);
1212 			}
1213 #endif
1214 
1215 			/*
1216 			 * jumbo payload length must be larger than 65535.
1217 			 */
1218 			if (jumboplen <= IPV6_MAXPACKET) {
1219 				IP6STAT_INC(ip6s_badoptions);
1220 				icmp6_error(m, ICMP6_PARAM_PROB,
1221 				    ICMP6_PARAMPROB_HEADER,
1222 				    erroff + opt + 2 - opthead);
1223 				return (-1);
1224 			}
1225 			*plenp = jumboplen;
1226 
1227 			break;
1228 		default:		/* unknown option */
1229 			if (hbhlen < IP6OPT_MINLEN) {
1230 				IP6STAT_INC(ip6s_toosmall);
1231 				goto bad;
1232 			}
1233 			optlen = ip6_unknown_opt(opt, m,
1234 			    erroff + opt - opthead);
1235 			if (optlen == -1)
1236 				return (-1);
1237 			optlen += 2;
1238 			break;
1239 		}
1240 	}
1241 
1242 	return (0);
1243 
1244   bad:
1245 	m_freem(m);
1246 	return (-1);
1247 }
1248 
1249 /*
1250  * Unknown option processing.
1251  * The third argument `off' is the offset from the IPv6 header to the option,
1252  * which is necessary if the IPv6 header the and option header and IPv6 header
1253  * is not contiguous in order to return an ICMPv6 error.
1254  */
1255 int
ip6_unknown_opt(u_int8_t * optp,struct mbuf * m,int off)1256 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1257 {
1258 	struct ip6_hdr *ip6;
1259 
1260 	switch (IP6OPT_TYPE(*optp)) {
1261 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1262 		return ((int)*(optp + 1));
1263 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1264 		m_freem(m);
1265 		return (-1);
1266 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1267 		IP6STAT_INC(ip6s_badoptions);
1268 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1269 		return (-1);
1270 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1271 		IP6STAT_INC(ip6s_badoptions);
1272 		ip6 = mtod(m, struct ip6_hdr *);
1273 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1274 		    (m->m_flags & (M_BCAST|M_MCAST)))
1275 			m_freem(m);
1276 		else
1277 			icmp6_error(m, ICMP6_PARAM_PROB,
1278 				    ICMP6_PARAMPROB_OPTION, off);
1279 		return (-1);
1280 	}
1281 
1282 	m_freem(m);		/* XXX: NOTREACHED */
1283 	return (-1);
1284 }
1285 
1286 /*
1287  * Create the "control" list for this pcb.
1288  * These functions will not modify mbuf chain at all.
1289  *
1290  * With KAME mbuf chain restriction:
1291  * The routine will be called from upper layer handlers like tcp6_input().
1292  * Thus the routine assumes that the caller (tcp6_input) have already
1293  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1294  * very first mbuf on the mbuf chain.
1295  *
1296  * ip6_savecontrol_v4 will handle those options that are possible to be
1297  * set on a v4-mapped socket.
1298  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1299  * options and handle the v6-only ones itself.
1300  */
1301 struct mbuf **
ip6_savecontrol_v4(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp,int * v4only)1302 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1303     int *v4only)
1304 {
1305 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1306 
1307 #ifdef SO_TIMESTAMP
1308 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1309 		struct timeval tv;
1310 
1311 		microtime(&tv);
1312 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1313 		    SCM_TIMESTAMP, SOL_SOCKET);
1314 		if (*mp)
1315 			mp = &(*mp)->m_next;
1316 	}
1317 #endif
1318 
1319 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1320 	/* RFC 2292 sec. 5 */
1321 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1322 		struct in6_pktinfo pi6;
1323 
1324 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1325 #ifdef INET
1326 			struct ip *ip;
1327 
1328 			ip = mtod(m, struct ip *);
1329 			pi6.ipi6_addr.s6_addr32[0] = 0;
1330 			pi6.ipi6_addr.s6_addr32[1] = 0;
1331 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1332 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1333 #else
1334 			/* We won't hit this code */
1335 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1336 #endif
1337 		} else {
1338 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1339 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1340 		}
1341 		pi6.ipi6_ifindex =
1342 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1343 
1344 		*mp = sbcreatecontrol((caddr_t) &pi6,
1345 		    sizeof(struct in6_pktinfo),
1346 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1347 		if (*mp)
1348 			mp = &(*mp)->m_next;
1349 	}
1350 
1351 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1352 		int hlim;
1353 
1354 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1355 #ifdef INET
1356 			struct ip *ip;
1357 
1358 			ip = mtod(m, struct ip *);
1359 			hlim = ip->ip_ttl;
1360 #else
1361 			/* We won't hit this code */
1362 			hlim = 0;
1363 #endif
1364 		} else {
1365 			hlim = ip6->ip6_hlim & 0xff;
1366 		}
1367 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1368 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1369 		    IPPROTO_IPV6);
1370 		if (*mp)
1371 			mp = &(*mp)->m_next;
1372 	}
1373 
1374 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1375 		int tclass;
1376 
1377 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1378 #ifdef INET
1379 			struct ip *ip;
1380 
1381 			ip = mtod(m, struct ip *);
1382 			tclass = ip->ip_tos;
1383 #else
1384 			/* We won't hit this code */
1385 			tclass = 0;
1386 #endif
1387 		} else {
1388 			u_int32_t flowinfo;
1389 
1390 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1391 			flowinfo >>= 20;
1392 			tclass = flowinfo & 0xff;
1393 		}
1394 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1395 		    IPV6_TCLASS, IPPROTO_IPV6);
1396 		if (*mp)
1397 			mp = &(*mp)->m_next;
1398 	}
1399 
1400 	if (v4only != NULL) {
1401 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1402 			*v4only = 1;
1403 		} else {
1404 			*v4only = 0;
1405 		}
1406 	}
1407 
1408 	return (mp);
1409 }
1410 
1411 void
ip6_savecontrol(struct inpcb * in6p,struct mbuf * m,struct mbuf ** mp)1412 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1413 {
1414 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1415 	int v4only = 0;
1416 
1417 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1418 	if (v4only)
1419 		return;
1420 
1421 	/*
1422 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1423 	 * privilege for the option (see ip6_ctloutput), but it might be too
1424 	 * strict, since there might be some hop-by-hop options which can be
1425 	 * returned to normal user.
1426 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1427 	 */
1428 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1429 		/*
1430 		 * Check if a hop-by-hop options header is contatined in the
1431 		 * received packet, and if so, store the options as ancillary
1432 		 * data. Note that a hop-by-hop options header must be
1433 		 * just after the IPv6 header, which is assured through the
1434 		 * IPv6 input processing.
1435 		 */
1436 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1437 			struct ip6_hbh *hbh;
1438 			int hbhlen = 0;
1439 #ifdef PULLDOWN_TEST
1440 			struct mbuf *ext;
1441 #endif
1442 
1443 #ifndef PULLDOWN_TEST
1444 			hbh = (struct ip6_hbh *)(ip6 + 1);
1445 			hbhlen = (hbh->ip6h_len + 1) << 3;
1446 #else
1447 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1448 			    ip6->ip6_nxt);
1449 			if (ext == NULL) {
1450 				IP6STAT_INC(ip6s_tooshort);
1451 				return;
1452 			}
1453 			hbh = mtod(ext, struct ip6_hbh *);
1454 			hbhlen = (hbh->ip6h_len + 1) << 3;
1455 			if (hbhlen != ext->m_len) {
1456 				m_freem(ext);
1457 				IP6STAT_INC(ip6s_tooshort);
1458 				return;
1459 			}
1460 #endif
1461 
1462 			/*
1463 			 * XXX: We copy the whole header even if a
1464 			 * jumbo payload option is included, the option which
1465 			 * is to be removed before returning according to
1466 			 * RFC2292.
1467 			 * Note: this constraint is removed in RFC3542
1468 			 */
1469 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1470 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1471 			    IPPROTO_IPV6);
1472 			if (*mp)
1473 				mp = &(*mp)->m_next;
1474 #ifdef PULLDOWN_TEST
1475 			m_freem(ext);
1476 #endif
1477 		}
1478 	}
1479 
1480 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1481 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1482 
1483 		/*
1484 		 * Search for destination options headers or routing
1485 		 * header(s) through the header chain, and stores each
1486 		 * header as ancillary data.
1487 		 * Note that the order of the headers remains in
1488 		 * the chain of ancillary data.
1489 		 */
1490 		while (1) {	/* is explicit loop prevention necessary? */
1491 			struct ip6_ext *ip6e = NULL;
1492 			int elen;
1493 #ifdef PULLDOWN_TEST
1494 			struct mbuf *ext = NULL;
1495 #endif
1496 
1497 			/*
1498 			 * if it is not an extension header, don't try to
1499 			 * pull it from the chain.
1500 			 */
1501 			switch (nxt) {
1502 			case IPPROTO_DSTOPTS:
1503 			case IPPROTO_ROUTING:
1504 			case IPPROTO_HOPOPTS:
1505 			case IPPROTO_AH: /* is it possible? */
1506 				break;
1507 			default:
1508 				goto loopend;
1509 			}
1510 
1511 #ifndef PULLDOWN_TEST
1512 			if (off + sizeof(*ip6e) > m->m_len)
1513 				goto loopend;
1514 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1515 			if (nxt == IPPROTO_AH)
1516 				elen = (ip6e->ip6e_len + 2) << 2;
1517 			else
1518 				elen = (ip6e->ip6e_len + 1) << 3;
1519 			if (off + elen > m->m_len)
1520 				goto loopend;
1521 #else
1522 			ext = ip6_pullexthdr(m, off, nxt);
1523 			if (ext == NULL) {
1524 				IP6STAT_INC(ip6s_tooshort);
1525 				return;
1526 			}
1527 			ip6e = mtod(ext, struct ip6_ext *);
1528 			if (nxt == IPPROTO_AH)
1529 				elen = (ip6e->ip6e_len + 2) << 2;
1530 			else
1531 				elen = (ip6e->ip6e_len + 1) << 3;
1532 			if (elen != ext->m_len) {
1533 				m_freem(ext);
1534 				IP6STAT_INC(ip6s_tooshort);
1535 				return;
1536 			}
1537 #endif
1538 
1539 			switch (nxt) {
1540 			case IPPROTO_DSTOPTS:
1541 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1542 					break;
1543 
1544 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1545 				    IS2292(in6p,
1546 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1547 				    IPPROTO_IPV6);
1548 				if (*mp)
1549 					mp = &(*mp)->m_next;
1550 				break;
1551 			case IPPROTO_ROUTING:
1552 				if (!(in6p->inp_flags & IN6P_RTHDR))
1553 					break;
1554 
1555 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1556 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1557 				    IPPROTO_IPV6);
1558 				if (*mp)
1559 					mp = &(*mp)->m_next;
1560 				break;
1561 			case IPPROTO_HOPOPTS:
1562 			case IPPROTO_AH: /* is it possible? */
1563 				break;
1564 
1565 			default:
1566 				/*
1567 				 * other cases have been filtered in the above.
1568 				 * none will visit this case.  here we supply
1569 				 * the code just in case (nxt overwritten or
1570 				 * other cases).
1571 				 */
1572 #ifdef PULLDOWN_TEST
1573 				m_freem(ext);
1574 #endif
1575 				goto loopend;
1576 
1577 			}
1578 
1579 			/* proceed with the next header. */
1580 			off += elen;
1581 			nxt = ip6e->ip6e_nxt;
1582 			ip6e = NULL;
1583 #ifdef PULLDOWN_TEST
1584 			m_freem(ext);
1585 			ext = NULL;
1586 #endif
1587 		}
1588 	  loopend:
1589 		;
1590 	}
1591 }
1592 #undef IS2292
1593 
1594 void
ip6_notify_pmtu(struct inpcb * inp,struct sockaddr_in6 * dst,u_int32_t mtu)1595 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1596 {
1597 	struct socket *so;
1598 	struct mbuf *m_mtu;
1599 	struct ip6_mtuinfo mtuctl;
1600 
1601 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1602 	/*
1603 	 * Notify the error by sending IPV6_PATHMTU ancillary data if
1604 	 * application wanted to know the MTU value.
1605 	 * NOTE: we notify disconnected sockets, because some udp
1606 	 * applications keep sending sockets disconnected.
1607 	 * NOTE: our implementation doesn't notify connected sockets that has
1608 	 * foreign address that is different than given destination addresses
1609 	 * (this is permitted by RFC 3542).
1610 	 */
1611 	if ((inp->inp_flags & IN6P_MTU) == 0 || (
1612 	    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1613 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1614 		return;
1615 
1616 	mtuctl.ip6m_mtu = mtu;
1617 	mtuctl.ip6m_addr = *dst;
1618 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1619 		return;
1620 
1621 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1622 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1623 		return;
1624 
1625 	so =  inp->inp_socket;
1626 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1627 	    == 0) {
1628 		m_freem(m_mtu);
1629 		/* XXX: should count statistics */
1630 	} else
1631 		sorwakeup(so);
1632 }
1633 
1634 #ifdef PULLDOWN_TEST
1635 /*
1636  * pull single extension header from mbuf chain.  returns single mbuf that
1637  * contains the result, or NULL on error.
1638  */
1639 static struct mbuf *
ip6_pullexthdr(struct mbuf * m,size_t off,int nxt)1640 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1641 {
1642 	struct ip6_ext ip6e;
1643 	size_t elen;
1644 	struct mbuf *n;
1645 
1646 #ifdef DIAGNOSTIC
1647 	switch (nxt) {
1648 	case IPPROTO_DSTOPTS:
1649 	case IPPROTO_ROUTING:
1650 	case IPPROTO_HOPOPTS:
1651 	case IPPROTO_AH: /* is it possible? */
1652 		break;
1653 	default:
1654 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1655 	}
1656 #endif
1657 
1658 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1659 	if (nxt == IPPROTO_AH)
1660 		elen = (ip6e.ip6e_len + 2) << 2;
1661 	else
1662 		elen = (ip6e.ip6e_len + 1) << 3;
1663 
1664 	if (elen > MLEN)
1665 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1666 	else
1667 		n = m_get(M_NOWAIT, MT_DATA);
1668 	if (n == NULL)
1669 		return NULL;
1670 
1671 	m_copydata(m, off, elen, mtod(n, caddr_t));
1672 	n->m_len = elen;
1673 	return n;
1674 }
1675 #endif
1676 
1677 /*
1678  * Get pointer to the previous header followed by the header
1679  * currently processed.
1680  * XXX: This function supposes that
1681  *	M includes all headers,
1682  *	the next header field and the header length field of each header
1683  *	are valid, and
1684  *	the sum of each header length equals to OFF.
1685  * Because of these assumptions, this function must be called very
1686  * carefully. Moreover, it will not be used in the near future when
1687  * we develop `neater' mechanism to process extension headers.
1688  */
1689 char *
ip6_get_prevhdr(struct mbuf * m,int off)1690 ip6_get_prevhdr(struct mbuf *m, int off)
1691 {
1692 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1693 
1694 	if (off == sizeof(struct ip6_hdr))
1695 		return (&ip6->ip6_nxt);
1696 	else {
1697 		int len, nxt;
1698 		struct ip6_ext *ip6e = NULL;
1699 
1700 		nxt = ip6->ip6_nxt;
1701 		len = sizeof(struct ip6_hdr);
1702 		while (len < off) {
1703 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1704 
1705 			switch (nxt) {
1706 			case IPPROTO_FRAGMENT:
1707 				len += sizeof(struct ip6_frag);
1708 				break;
1709 			case IPPROTO_AH:
1710 				len += (ip6e->ip6e_len + 2) << 2;
1711 				break;
1712 			default:
1713 				len += (ip6e->ip6e_len + 1) << 3;
1714 				break;
1715 			}
1716 			nxt = ip6e->ip6e_nxt;
1717 		}
1718 		if (ip6e)
1719 			return (&ip6e->ip6e_nxt);
1720 		else
1721 			return NULL;
1722 	}
1723 }
1724 
1725 /*
1726  * get next header offset.  m will be retained.
1727  */
1728 int
ip6_nexthdr(struct mbuf * m,int off,int proto,int * nxtp)1729 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1730 {
1731 	struct ip6_hdr ip6;
1732 	struct ip6_ext ip6e;
1733 	struct ip6_frag fh;
1734 
1735 	/* just in case */
1736 	if (m == NULL)
1737 		panic("ip6_nexthdr: m == NULL");
1738 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1739 		return -1;
1740 
1741 	switch (proto) {
1742 	case IPPROTO_IPV6:
1743 		if (m->m_pkthdr.len < off + sizeof(ip6))
1744 			return -1;
1745 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1746 		if (nxtp)
1747 			*nxtp = ip6.ip6_nxt;
1748 		off += sizeof(ip6);
1749 		return off;
1750 
1751 	case IPPROTO_FRAGMENT:
1752 		/*
1753 		 * terminate parsing if it is not the first fragment,
1754 		 * it does not make sense to parse through it.
1755 		 */
1756 		if (m->m_pkthdr.len < off + sizeof(fh))
1757 			return -1;
1758 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1759 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1760 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1761 			return -1;
1762 		if (nxtp)
1763 			*nxtp = fh.ip6f_nxt;
1764 		off += sizeof(struct ip6_frag);
1765 		return off;
1766 
1767 	case IPPROTO_AH:
1768 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1769 			return -1;
1770 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1771 		if (nxtp)
1772 			*nxtp = ip6e.ip6e_nxt;
1773 		off += (ip6e.ip6e_len + 2) << 2;
1774 		return off;
1775 
1776 	case IPPROTO_HOPOPTS:
1777 	case IPPROTO_ROUTING:
1778 	case IPPROTO_DSTOPTS:
1779 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1780 			return -1;
1781 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1782 		if (nxtp)
1783 			*nxtp = ip6e.ip6e_nxt;
1784 		off += (ip6e.ip6e_len + 1) << 3;
1785 		return off;
1786 
1787 	case IPPROTO_NONE:
1788 	case IPPROTO_ESP:
1789 	case IPPROTO_IPCOMP:
1790 		/* give up */
1791 		return -1;
1792 
1793 	default:
1794 		return -1;
1795 	}
1796 
1797 	return -1;
1798 }
1799 
1800 /*
1801  * get offset for the last header in the chain.  m will be kept untainted.
1802  */
1803 int
ip6_lasthdr(struct mbuf * m,int off,int proto,int * nxtp)1804 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1805 {
1806 	int newoff;
1807 	int nxt;
1808 
1809 	if (!nxtp) {
1810 		nxt = -1;
1811 		nxtp = &nxt;
1812 	}
1813 	while (1) {
1814 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1815 		if (newoff < 0)
1816 			return off;
1817 		else if (newoff < off)
1818 			return -1;	/* invalid */
1819 		else if (newoff == off)
1820 			return newoff;
1821 
1822 		off = newoff;
1823 		proto = *nxtp;
1824 	}
1825 }
1826 
1827 static struct ip6aux *
ip6_addaux(struct mbuf * m)1828 ip6_addaux(struct mbuf *m)
1829 {
1830 	struct m_tag *mtag;
1831 
1832 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1833 	if (!mtag) {
1834 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1835 		    M_NOWAIT);
1836 		if (mtag) {
1837 			m_tag_prepend(m, mtag);
1838 			bzero(mtag + 1, sizeof(struct ip6aux));
1839 		}
1840 	}
1841 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1842 }
1843 
1844 static struct ip6aux *
ip6_findaux(struct mbuf * m)1845 ip6_findaux(struct mbuf *m)
1846 {
1847 	struct m_tag *mtag;
1848 
1849 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1850 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1851 }
1852 
1853 static void
ip6_delaux(struct mbuf * m)1854 ip6_delaux(struct mbuf *m)
1855 {
1856 	struct m_tag *mtag;
1857 
1858 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1859 	if (mtag)
1860 		m_tag_delete(m, mtag);
1861 }
1862 
1863 /*
1864  * System control for IP6
1865  */
1866 
1867 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1868 	0,		0,		0,		0,
1869 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1870 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1871 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1872 	0,		0,		0,		0,
1873 	ENOPROTOOPT
1874 };
1875