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 #include "opt_route.h"
70 #include "opt_rss.h"
71
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/hhook.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/lock.h>
87 #include <sys/rmlock.h>
88 #include <sys/syslog.h>
89 #include <sys/sysctl.h>
90
91 #include <net/if.h>
92 #include <net/if_var.h>
93 #include <net/if_types.h>
94 #include <net/if_dl.h>
95 #include <net/route.h>
96 #include <net/netisr.h>
97 #include <net/rss_config.h>
98 #include <net/pfil.h>
99 #include <net/vnet.h>
100
101 #include <netinet/in.h>
102 #include <netinet/in_kdtrace.h>
103 #include <netinet/ip_var.h>
104 #include <netinet/in_systm.h>
105 #include <net/if_llatbl.h>
106 #ifdef INET
107 #include <netinet/ip.h>
108 #include <netinet/ip_icmp.h>
109 #endif /* INET */
110 #include <netinet/ip6.h>
111 #include <netinet6/in6_var.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet/in_pcb.h>
114 #include <netinet/icmp6.h>
115 #include <netinet6/scope6_var.h>
116 #include <netinet6/in6_ifattach.h>
117 #include <netinet6/mld6_var.h>
118 #include <netinet6/nd6.h>
119 #include <netinet6/in6_rss.h>
120
121 #include <netipsec/ipsec_support.h>
122
123 #include <netinet6/ip6protosw.h>
124
125 extern struct domain inet6domain;
126
127 u_char ip6_protox[IPPROTO_MAX];
128 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
129 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
130 VNET_DEFINE(u_long, in6_ifaddrhmask);
131
132 static struct netisr_handler ip6_nh = {
133 .nh_name = "ip6",
134 .nh_handler = ip6_input,
135 .nh_proto = NETISR_IPV6,
136 #ifdef RSS
137 .nh_m2cpuid = rss_soft_m2cpuid_v6,
138 .nh_policy = NETISR_POLICY_CPU,
139 .nh_dispatch = NETISR_DISPATCH_HYBRID,
140 #else
141 .nh_policy = NETISR_POLICY_FLOW,
142 #endif
143 };
144
145 static int
146 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
147 {
148 int error, qlimit;
149
150 netisr_getqlimit(&ip6_nh, &qlimit);
151 error = sysctl_handle_int(oidp, &qlimit, 0, req);
152 if (error || !req->newptr)
153 return (error);
154 if (qlimit < 1)
155 return (EINVAL);
156 return (netisr_setqlimit(&ip6_nh, qlimit));
157 }
158 SYSCTL_DECL(_net_inet6_ip6);
159 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen,
160 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_queue_maxlen, "I",
161 "Maximum size of the IPv6 input queue");
162
163 #ifdef RSS
164 static struct netisr_handler ip6_direct_nh = {
165 .nh_name = "ip6_direct",
166 .nh_handler = ip6_direct_input,
167 .nh_proto = NETISR_IPV6_DIRECT,
168 .nh_m2cpuid = rss_soft_m2cpuid_v6,
169 .nh_policy = NETISR_POLICY_CPU,
170 .nh_dispatch = NETISR_DISPATCH_HYBRID,
171 };
172
173 static int
174 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
175 {
176 int error, qlimit;
177
178 netisr_getqlimit(&ip6_direct_nh, &qlimit);
179 error = sysctl_handle_int(oidp, &qlimit, 0, req);
180 if (error || !req->newptr)
181 return (error);
182 if (qlimit < 1)
183 return (EINVAL);
184 return (netisr_setqlimit(&ip6_direct_nh, qlimit));
185 }
186 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
187 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_direct_queue_maxlen,
188 "I", "Maximum size of the IPv6 direct input queue");
189
190 #endif
191
192 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
193
194 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
195 VNET_PCPUSTAT_SYSINIT(ip6stat);
196 #ifdef VIMAGE
197 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
198 #endif /* VIMAGE */
199
200 struct rmlock in6_ifaddr_lock;
201 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
202
203 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
204 #ifdef PULLDOWN_TEST
205 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
206 #endif
207
208 /*
209 * IP6 initialization: fill in IP6 protocol switch table.
210 * All protocols not implemented in kernel go to raw IP6 protocol handler.
211 */
212 void
213 ip6_init(void)
214 {
215 struct protosw *pr;
216 int i;
217
218 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
219 &V_ip6_auto_linklocal);
220 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
221 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
222
223 TAILQ_INIT(&V_in6_ifaddrhead);
224 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
225 &V_in6_ifaddrhmask);
226
227 /* Initialize packet filter hooks. */
228 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
229 V_inet6_pfil_hook.ph_af = AF_INET6;
230 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
231 printf("%s: WARNING: unable to register pfil hook, "
232 "error %d\n", __func__, i);
233
234 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
235 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
236 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
237 printf("%s: WARNING: unable to register input helper hook\n",
238 __func__);
239 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
240 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
241 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
242 printf("%s: WARNING: unable to register output helper hook\n",
243 __func__);
244
245 scope6_init();
246 addrsel_policy_init();
247 nd6_init();
248 frag6_init();
249
250 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
251
252 /* Skip global initialization stuff for non-default instances. */
253 #ifdef VIMAGE
254 if (!IS_DEFAULT_VNET(curvnet)) {
255 netisr_register_vnet(&ip6_nh);
256 #ifdef RSS
257 netisr_register_vnet(&ip6_direct_nh);
258 #endif
259 return;
260 }
261 #endif
262
263 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
264 if (pr == NULL)
265 panic("ip6_init");
266
267 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
268 for (i = 0; i < IPPROTO_MAX; i++)
269 ip6_protox[i] = pr - inet6sw;
270 /*
271 * Cycle through IP protocols and put them into the appropriate place
272 * in ip6_protox[].
273 */
274 for (pr = inet6domain.dom_protosw;
275 pr < inet6domain.dom_protoswNPROTOSW; pr++)
276 if (pr->pr_domain->dom_family == PF_INET6 &&
277 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
278 /* Be careful to only index valid IP protocols. */
279 if (pr->pr_protocol < IPPROTO_MAX)
280 ip6_protox[pr->pr_protocol] = pr - inet6sw;
281 }
282
283 netisr_register(&ip6_nh);
284 #ifdef RSS
285 netisr_register(&ip6_direct_nh);
286 #endif
287 }
288
289 /*
290 * The protocol to be inserted into ip6_protox[] must be already registered
291 * in inet6sw[], either statically or through pf_proto_register().
292 */
293 int
294 ip6proto_register(short ip6proto)
295 {
296 struct protosw *pr;
297
298 /* Sanity checks. */
299 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
300 return (EPROTONOSUPPORT);
301
302 /*
303 * The protocol slot must not be occupied by another protocol
304 * already. An index pointing to IPPROTO_RAW is unused.
305 */
306 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
307 if (pr == NULL)
308 return (EPFNOSUPPORT);
309 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */
310 return (EEXIST);
311
312 /*
313 * Find the protocol position in inet6sw[] and set the index.
314 */
315 for (pr = inet6domain.dom_protosw;
316 pr < inet6domain.dom_protoswNPROTOSW; pr++) {
317 if (pr->pr_domain->dom_family == PF_INET6 &&
318 pr->pr_protocol && pr->pr_protocol == ip6proto) {
319 ip6_protox[pr->pr_protocol] = pr - inet6sw;
320 return (0);
321 }
322 }
323 return (EPROTONOSUPPORT);
324 }
325
326 int
327 ip6proto_unregister(short ip6proto)
328 {
329 struct protosw *pr;
330
331 /* Sanity checks. */
332 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
333 return (EPROTONOSUPPORT);
334
335 /* Check if the protocol was indeed registered. */
336 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
337 if (pr == NULL)
338 return (EPFNOSUPPORT);
339 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */
340 return (ENOENT);
341
342 /* Reset the protocol slot to IPPROTO_RAW. */
343 ip6_protox[ip6proto] = pr - inet6sw;
344 return (0);
345 }
346
347 #ifdef VIMAGE
348 static void
349 ip6_destroy(void *unused __unused)
350 {
351 struct ifaddr *ifa, *nifa;
352 struct ifnet *ifp;
353 int error;
354
355 #ifdef RSS
356 netisr_unregister_vnet(&ip6_direct_nh);
357 #endif
358 netisr_unregister_vnet(&ip6_nh);
359
360 if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
361 printf("%s: WARNING: unable to unregister pfil hook, "
362 "error %d\n", __func__, error);
363 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
364 if (error != 0) {
365 printf("%s: WARNING: unable to deregister input helper hook "
366 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
367 "error %d returned\n", __func__, error);
368 }
369 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
370 if (error != 0) {
371 printf("%s: WARNING: unable to deregister output helper hook "
372 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
373 "error %d returned\n", __func__, error);
374 }
375
376 /* Cleanup addresses. */
377 IFNET_RLOCK();
378 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
379 /* Cannot lock here - lock recursion. */
380 /* IF_ADDR_LOCK(ifp); */
381 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
382
383 if (ifa->ifa_addr->sa_family != AF_INET6)
384 continue;
385 in6_purgeaddr(ifa);
386 }
387 /* IF_ADDR_UNLOCK(ifp); */
388 in6_ifdetach_destroy(ifp);
389 mld_domifdetach(ifp);
390 /* Make sure any routes are gone as well. */
391 rt_flushifroutes_af(ifp, AF_INET6);
392 }
393 IFNET_RUNLOCK();
394
395 nd6_destroy();
396 in6_ifattach_destroy();
397
398 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
399 }
400
401 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL);
402 #endif
403
404 static int
405 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off,
406 int *nxt, int *ours)
407 {
408 struct mbuf *m;
409 struct ip6_hdr *ip6;
410 struct ip6_hbh *hbh;
411
412 if (ip6_hopopts_input(plen, rtalert, mp, off)) {
413 #if 0 /*touches NULL pointer*/
414 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard);
415 #endif
416 goto out; /* m have already been freed */
417 }
418
419 /* adjust pointer */
420 m = *mp;
421 ip6 = mtod(m, struct ip6_hdr *);
422
423 /*
424 * if the payload length field is 0 and the next header field
425 * indicates Hop-by-Hop Options header, then a Jumbo Payload
426 * option MUST be included.
427 */
428 if (ip6->ip6_plen == 0 && *plen == 0) {
429 /*
430 * Note that if a valid jumbo payload option is
431 * contained, ip6_hopopts_input() must set a valid
432 * (non-zero) payload length to the variable plen.
433 */
434 IP6STAT_INC(ip6s_badoptions);
435 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
436 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
437 icmp6_error(m, ICMP6_PARAM_PROB,
438 ICMP6_PARAMPROB_HEADER,
439 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
440 goto out;
441 }
442 #ifndef PULLDOWN_TEST
443 /* ip6_hopopts_input() ensures that mbuf is contiguous */
444 hbh = (struct ip6_hbh *)(ip6 + 1);
445 #else
446 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
447 sizeof(struct ip6_hbh));
448 if (hbh == NULL) {
449 IP6STAT_INC(ip6s_tooshort);
450 goto out;
451 }
452 #endif
453 *nxt = hbh->ip6h_nxt;
454
455 /*
456 * If we are acting as a router and the packet contains a
457 * router alert option, see if we know the option value.
458 * Currently, we only support the option value for MLD, in which
459 * case we should pass the packet to the multicast routing
460 * daemon.
461 */
462 if (*rtalert != ~0) {
463 switch (*rtalert) {
464 case IP6OPT_RTALERT_MLD:
465 if (V_ip6_forwarding)
466 *ours = 1;
467 break;
468 default:
469 /*
470 * RFC2711 requires unrecognized values must be
471 * silently ignored.
472 */
473 break;
474 }
475 }
476
477 return (0);
478
479 out:
480 return (1);
481 }
482
483 #ifdef RSS
484 /*
485 * IPv6 direct input routine.
486 *
487 * This is called when reinjecting completed fragments where
488 * all of the previous checking and book-keeping has been done.
489 */
490 void
491 ip6_direct_input(struct mbuf *m)
492 {
493 int off, nxt;
494 int nest;
495 struct m_tag *mtag;
496 struct ip6_direct_ctx *ip6dc;
497
498 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
499 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
500
501 ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
502 nxt = ip6dc->ip6dc_nxt;
503 off = ip6dc->ip6dc_off;
504
505 nest = 0;
506
507 m_tag_delete(m, mtag);
508
509 while (nxt != IPPROTO_DONE) {
510 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
511 IP6STAT_INC(ip6s_toomanyhdr);
512 goto bad;
513 }
514
515 /*
516 * protection against faulty packet - there should be
517 * more sanity checks in header chain processing.
518 */
519 if (m->m_pkthdr.len < off) {
520 IP6STAT_INC(ip6s_tooshort);
521 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
522 goto bad;
523 }
524
525 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
526 if (IPSEC_ENABLED(ipv6)) {
527 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
528 return;
529 }
530 #endif /* IPSEC */
531
532 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
533 }
534 return;
535 bad:
536 m_freem(m);
537 }
538 #endif
539
540 void
541 ip6_input(struct mbuf *m)
542 {
543 struct in6_addr odst;
544 struct ip6_hdr *ip6;
545 struct in6_ifaddr *ia;
546 struct ifnet *rcvif;
547 u_int32_t plen;
548 u_int32_t rtalert = ~0;
549 int off = sizeof(struct ip6_hdr), nest;
550 int nxt, ours = 0;
551 int srcrt = 0;
552
553 /*
554 * Drop the packet if IPv6 operation is disabled on the interface.
555 */
556 rcvif = m->m_pkthdr.rcvif;
557 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED))
558 goto bad;
559
560 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
561 /*
562 * should the inner packet be considered authentic?
563 * see comment in ah4_input().
564 * NB: m cannot be NULL when passed to the input routine
565 */
566
567 m->m_flags &= ~M_AUTHIPHDR;
568 m->m_flags &= ~M_AUTHIPDGM;
569
570 #endif /* IPSEC */
571
572 if (m->m_flags & M_FASTFWD_OURS) {
573 /*
574 * Firewall changed destination to local.
575 */
576 ip6 = mtod(m, struct ip6_hdr *);
577 goto passin;
578 }
579
580 /*
581 * mbuf statistics
582 */
583 if (m->m_flags & M_EXT) {
584 if (m->m_next)
585 IP6STAT_INC(ip6s_mext2m);
586 else
587 IP6STAT_INC(ip6s_mext1);
588 } else {
589 if (m->m_next) {
590 if (m->m_flags & M_LOOP) {
591 IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
592 } else if (rcvif->if_index < IP6S_M2MMAX)
593 IP6STAT_INC(ip6s_m2m[rcvif->if_index]);
594 else
595 IP6STAT_INC(ip6s_m2m[0]);
596 } else
597 IP6STAT_INC(ip6s_m1);
598 }
599
600 in6_ifstat_inc(rcvif, ifs6_in_receive);
601 IP6STAT_INC(ip6s_total);
602
603 #ifndef PULLDOWN_TEST
604 /*
605 * L2 bridge code and some other code can return mbuf chain
606 * that does not conform to KAME requirement. too bad.
607 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
608 */
609 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
610 struct mbuf *n;
611
612 if (m->m_pkthdr.len > MHLEN)
613 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
614 else
615 n = m_gethdr(M_NOWAIT, MT_DATA);
616 if (n == NULL)
617 goto bad;
618
619 m_move_pkthdr(n, m);
620 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
621 n->m_len = n->m_pkthdr.len;
622 m_freem(m);
623 m = n;
624 }
625 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
626 #endif
627
628 if (m->m_len < sizeof(struct ip6_hdr)) {
629 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
630 IP6STAT_INC(ip6s_toosmall);
631 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
632 goto bad;
633 }
634 }
635
636 ip6 = mtod(m, struct ip6_hdr *);
637 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
638 IP6STAT_INC(ip6s_badvers);
639 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
640 goto bad;
641 }
642
643 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
644 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6);
645
646 /*
647 * Check against address spoofing/corruption.
648 */
649 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
650 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
651 /*
652 * XXX: "badscope" is not very suitable for a multicast source.
653 */
654 IP6STAT_INC(ip6s_badscope);
655 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
656 goto bad;
657 }
658 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
659 !(m->m_flags & M_LOOP)) {
660 /*
661 * In this case, the packet should come from the loopback
662 * interface. However, we cannot just check the if_flags,
663 * because ip6_mloopback() passes the "actual" interface
664 * as the outgoing/incoming interface.
665 */
666 IP6STAT_INC(ip6s_badscope);
667 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
668 goto bad;
669 }
670 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
671 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
672 /*
673 * RFC4291 2.7:
674 * Nodes must not originate a packet to a multicast address
675 * whose scop field contains the reserved value 0; if such
676 * a packet is received, it must be silently dropped.
677 */
678 IP6STAT_INC(ip6s_badscope);
679 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
680 goto bad;
681 }
682 #ifdef ALTQ
683 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
684 /* packet is dropped by traffic conditioner */
685 return;
686 }
687 #endif
688 /*
689 * The following check is not documented in specs. A malicious
690 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
691 * and bypass security checks (act as if it was from 127.0.0.1 by using
692 * IPv6 src ::ffff:127.0.0.1). Be cautious.
693 *
694 * This check chokes if we are in an SIIT cloud. As none of BSDs
695 * support IPv4-less kernel compilation, we cannot support SIIT
696 * environment at all. So, it makes more sense for us to reject any
697 * malicious packets for non-SIIT environment, than try to do a
698 * partial support for SIIT environment.
699 */
700 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
701 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
702 IP6STAT_INC(ip6s_badscope);
703 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
704 goto bad;
705 }
706 #if 0
707 /*
708 * Reject packets with IPv4 compatible addresses (auto tunnel).
709 *
710 * The code forbids auto tunnel relay case in RFC1933 (the check is
711 * stronger than RFC1933). We may want to re-enable it if mech-xx
712 * is revised to forbid relaying case.
713 */
714 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
715 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
716 IP6STAT_INC(ip6s_badscope);
717 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
718 goto bad;
719 }
720 #endif
721 /*
722 * Try to forward the packet, but if we fail continue.
723 * ip6_tryforward() does not generate redirects, so fall
724 * through to normal processing if redirects are required.
725 * ip6_tryforward() does inbound and outbound packet firewall
726 * processing. If firewall has decided that destination becomes
727 * our local address, it sets M_FASTFWD_OURS flag. In this
728 * case skip another inbound firewall processing and update
729 * ip6 pointer.
730 */
731 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0
732 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
733 && (!IPSEC_ENABLED(ipv6) ||
734 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0)
735 #endif
736 ) {
737 if ((m = ip6_tryforward(m)) == NULL)
738 return;
739 if (m->m_flags & M_FASTFWD_OURS) {
740 ip6 = mtod(m, struct ip6_hdr *);
741 goto passin;
742 }
743 }
744 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
745 /*
746 * Bypass packet filtering for packets previously handled by IPsec.
747 */
748 if (IPSEC_ENABLED(ipv6) &&
749 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0)
750 goto passin;
751 #endif
752 /*
753 * Run through list of hooks for input packets.
754 *
755 * NB: Beware of the destination address changing
756 * (e.g. by NAT rewriting). When this happens,
757 * tell ip6_forward to do the right thing.
758 */
759
760 /* Jump over all PFIL processing if hooks are not active. */
761 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
762 goto passin;
763
764 odst = ip6->ip6_dst;
765 if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
766 m->m_pkthdr.rcvif, PFIL_IN, 0, NULL))
767 return;
768 if (m == NULL) /* consumed by filter */
769 return;
770 ip6 = mtod(m, struct ip6_hdr *);
771 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
772 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP &&
773 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
774 /*
775 * Directly ship the packet on. This allows forwarding
776 * packets originally destined to us to some other directly
777 * connected host.
778 */
779 ip6_forward(m, 1);
780 return;
781 }
782
783 passin:
784 /*
785 * Disambiguate address scope zones (if there is ambiguity).
786 * We first make sure that the original source or destination address
787 * is not in our internal form for scoped addresses. Such addresses
788 * are not necessarily invalid spec-wise, but we cannot accept them due
789 * to the usage conflict.
790 * in6_setscope() then also checks and rejects the cases where src or
791 * dst are the loopback address and the receiving interface
792 * is not loopback.
793 */
794 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
795 IP6STAT_INC(ip6s_badscope); /* XXX */
796 goto bad;
797 }
798 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) ||
799 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) {
800 IP6STAT_INC(ip6s_badscope);
801 goto bad;
802 }
803 if (m->m_flags & M_FASTFWD_OURS) {
804 m->m_flags &= ~M_FASTFWD_OURS;
805 ours = 1;
806 goto hbhcheck;
807 }
808 /*
809 * Multicast check. Assume packet is for us to avoid
810 * prematurely taking locks.
811 */
812 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
813 ours = 1;
814 in6_ifstat_inc(rcvif, ifs6_in_mcast);
815 goto hbhcheck;
816 }
817 /*
818 * Unicast check
819 * XXX: For now we keep link-local IPv6 addresses with embedded
820 * scope zone id, therefore we use zero zoneid here.
821 */
822 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
823 if (ia != NULL) {
824 if (ia->ia6_flags & IN6_IFF_NOTREADY) {
825 char ip6bufs[INET6_ADDRSTRLEN];
826 char ip6bufd[INET6_ADDRSTRLEN];
827 /* address is not ready, so discard the packet. */
828 nd6log((LOG_INFO,
829 "ip6_input: packet to an unready address %s->%s\n",
830 ip6_sprintf(ip6bufs, &ip6->ip6_src),
831 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
832 ifa_free(&ia->ia_ifa);
833 goto bad;
834 }
835 /* Count the packet in the ip address stats */
836 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
837 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
838 ifa_free(&ia->ia_ifa);
839 ours = 1;
840 goto hbhcheck;
841 }
842
843 /*
844 * Now there is no reason to process the packet if it's not our own
845 * and we're not a router.
846 */
847 if (!V_ip6_forwarding) {
848 IP6STAT_INC(ip6s_cantforward);
849 goto bad;
850 }
851
852 hbhcheck:
853 /*
854 * Process Hop-by-Hop options header if it's contained.
855 * m may be modified in ip6_hopopts_input().
856 * If a JumboPayload option is included, plen will also be modified.
857 */
858 plen = (u_int32_t)ntohs(ip6->ip6_plen);
859 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
860 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0)
861 return;
862 } else
863 nxt = ip6->ip6_nxt;
864
865 /*
866 * Use mbuf flags to propagate Router Alert option to
867 * ICMPv6 layer, as hop-by-hop options have been stripped.
868 */
869 if (rtalert != ~0)
870 m->m_flags |= M_RTALERT_MLD;
871
872 /*
873 * Check that the amount of data in the buffers
874 * is as at least much as the IPv6 header would have us expect.
875 * Trim mbufs if longer than we expect.
876 * Drop packet if shorter than we expect.
877 */
878 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
879 IP6STAT_INC(ip6s_tooshort);
880 in6_ifstat_inc(rcvif, ifs6_in_truncated);
881 goto bad;
882 }
883 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
884 if (m->m_len == m->m_pkthdr.len) {
885 m->m_len = sizeof(struct ip6_hdr) + plen;
886 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
887 } else
888 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
889 }
890
891 /*
892 * Forward if desirable.
893 */
894 if (V_ip6_mrouter &&
895 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
896 /*
897 * If we are acting as a multicast router, all
898 * incoming multicast packets are passed to the
899 * kernel-level multicast forwarding function.
900 * The packet is returned (relatively) intact; if
901 * ip6_mforward() returns a non-zero value, the packet
902 * must be discarded, else it may be accepted below.
903 *
904 * XXX TODO: Check hlim and multicast scope here to avoid
905 * unnecessarily calling into ip6_mforward().
906 */
907 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) {
908 IP6STAT_INC(ip6s_cantforward);
909 goto bad;
910 }
911 } else if (!ours) {
912 ip6_forward(m, srcrt);
913 return;
914 }
915
916 ip6 = mtod(m, struct ip6_hdr *);
917
918 /*
919 * Malicious party may be able to use IPv4 mapped addr to confuse
920 * tcp/udp stack and bypass security checks (act as if it was from
921 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious.
922 *
923 * For SIIT end node behavior, you may want to disable the check.
924 * However, you will become vulnerable to attacks using IPv4 mapped
925 * source.
926 */
927 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
928 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
929 IP6STAT_INC(ip6s_badscope);
930 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
931 goto bad;
932 }
933
934 /*
935 * Tell launch routine the next header
936 */
937 IP6STAT_INC(ip6s_delivered);
938 in6_ifstat_inc(rcvif, ifs6_in_deliver);
939 nest = 0;
940
941 while (nxt != IPPROTO_DONE) {
942 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
943 IP6STAT_INC(ip6s_toomanyhdr);
944 goto bad;
945 }
946
947 /*
948 * protection against faulty packet - there should be
949 * more sanity checks in header chain processing.
950 */
951 if (m->m_pkthdr.len < off) {
952 IP6STAT_INC(ip6s_tooshort);
953 in6_ifstat_inc(rcvif, ifs6_in_truncated);
954 goto bad;
955 }
956
957 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
958 if (IPSEC_ENABLED(ipv6)) {
959 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
960 return;
961 }
962 #endif /* IPSEC */
963
964 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
965 }
966 return;
967 bad:
968 in6_ifstat_inc(rcvif, ifs6_in_discard);
969 if (m != NULL)
970 m_freem(m);
971 }
972
973 /*
974 * Hop-by-Hop options header processing. If a valid jumbo payload option is
975 * included, the real payload length will be stored in plenp.
976 *
977 * rtalertp - XXX: should be stored more smart way
978 */
979 static int
980 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
981 struct mbuf **mp, int *offp)
982 {
983 struct mbuf *m = *mp;
984 int off = *offp, hbhlen;
985 struct ip6_hbh *hbh;
986
987 /* validation of the length of the header */
988 #ifndef PULLDOWN_TEST
989 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
990 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
991 hbhlen = (hbh->ip6h_len + 1) << 3;
992
993 IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
994 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
995 #else
996 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
997 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
998 if (hbh == NULL) {
999 IP6STAT_INC(ip6s_tooshort);
1000 return -1;
1001 }
1002 hbhlen = (hbh->ip6h_len + 1) << 3;
1003 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1004 hbhlen);
1005 if (hbh == NULL) {
1006 IP6STAT_INC(ip6s_tooshort);
1007 return -1;
1008 }
1009 #endif
1010 off += hbhlen;
1011 hbhlen -= sizeof(struct ip6_hbh);
1012 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1013 hbhlen, rtalertp, plenp) < 0)
1014 return (-1);
1015
1016 *offp = off;
1017 *mp = m;
1018 return (0);
1019 }
1020
1021 /*
1022 * Search header for all Hop-by-hop options and process each option.
1023 * This function is separate from ip6_hopopts_input() in order to
1024 * handle a case where the sending node itself process its hop-by-hop
1025 * options header. In such a case, the function is called from ip6_output().
1026 *
1027 * The function assumes that hbh header is located right after the IPv6 header
1028 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1029 * opthead + hbhlen is located in contiguous memory region.
1030 */
1031 int
1032 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1033 u_int32_t *rtalertp, u_int32_t *plenp)
1034 {
1035 struct ip6_hdr *ip6;
1036 int optlen = 0;
1037 u_int8_t *opt = opthead;
1038 u_int16_t rtalert_val;
1039 u_int32_t jumboplen;
1040 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1041
1042 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1043 switch (*opt) {
1044 case IP6OPT_PAD1:
1045 optlen = 1;
1046 break;
1047 case IP6OPT_PADN:
1048 if (hbhlen < IP6OPT_MINLEN) {
1049 IP6STAT_INC(ip6s_toosmall);
1050 goto bad;
1051 }
1052 optlen = *(opt + 1) + 2;
1053 break;
1054 case IP6OPT_ROUTER_ALERT:
1055 /* XXX may need check for alignment */
1056 if (hbhlen < IP6OPT_RTALERT_LEN) {
1057 IP6STAT_INC(ip6s_toosmall);
1058 goto bad;
1059 }
1060 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1061 /* XXX stat */
1062 icmp6_error(m, ICMP6_PARAM_PROB,
1063 ICMP6_PARAMPROB_HEADER,
1064 erroff + opt + 1 - opthead);
1065 return (-1);
1066 }
1067 optlen = IP6OPT_RTALERT_LEN;
1068 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1069 *rtalertp = ntohs(rtalert_val);
1070 break;
1071 case IP6OPT_JUMBO:
1072 /* XXX may need check for alignment */
1073 if (hbhlen < IP6OPT_JUMBO_LEN) {
1074 IP6STAT_INC(ip6s_toosmall);
1075 goto bad;
1076 }
1077 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1078 /* XXX stat */
1079 icmp6_error(m, ICMP6_PARAM_PROB,
1080 ICMP6_PARAMPROB_HEADER,
1081 erroff + opt + 1 - opthead);
1082 return (-1);
1083 }
1084 optlen = IP6OPT_JUMBO_LEN;
1085
1086 /*
1087 * IPv6 packets that have non 0 payload length
1088 * must not contain a jumbo payload option.
1089 */
1090 ip6 = mtod(m, struct ip6_hdr *);
1091 if (ip6->ip6_plen) {
1092 IP6STAT_INC(ip6s_badoptions);
1093 icmp6_error(m, ICMP6_PARAM_PROB,
1094 ICMP6_PARAMPROB_HEADER,
1095 erroff + opt - opthead);
1096 return (-1);
1097 }
1098
1099 /*
1100 * We may see jumbolen in unaligned location, so
1101 * we'd need to perform bcopy().
1102 */
1103 bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1104 jumboplen = (u_int32_t)htonl(jumboplen);
1105
1106 #if 1
1107 /*
1108 * if there are multiple jumbo payload options,
1109 * *plenp will be non-zero and the packet will be
1110 * rejected.
1111 * the behavior may need some debate in ipngwg -
1112 * multiple options does not make sense, however,
1113 * there's no explicit mention in specification.
1114 */
1115 if (*plenp != 0) {
1116 IP6STAT_INC(ip6s_badoptions);
1117 icmp6_error(m, ICMP6_PARAM_PROB,
1118 ICMP6_PARAMPROB_HEADER,
1119 erroff + opt + 2 - opthead);
1120 return (-1);
1121 }
1122 #endif
1123
1124 /*
1125 * jumbo payload length must be larger than 65535.
1126 */
1127 if (jumboplen <= IPV6_MAXPACKET) {
1128 IP6STAT_INC(ip6s_badoptions);
1129 icmp6_error(m, ICMP6_PARAM_PROB,
1130 ICMP6_PARAMPROB_HEADER,
1131 erroff + opt + 2 - opthead);
1132 return (-1);
1133 }
1134 *plenp = jumboplen;
1135
1136 break;
1137 default: /* unknown option */
1138 if (hbhlen < IP6OPT_MINLEN) {
1139 IP6STAT_INC(ip6s_toosmall);
1140 goto bad;
1141 }
1142 optlen = ip6_unknown_opt(opt, m,
1143 erroff + opt - opthead);
1144 if (optlen == -1)
1145 return (-1);
1146 optlen += 2;
1147 break;
1148 }
1149 }
1150
1151 return (0);
1152
1153 bad:
1154 m_freem(m);
1155 return (-1);
1156 }
1157
1158 /*
1159 * Unknown option processing.
1160 * The third argument `off' is the offset from the IPv6 header to the option,
1161 * which is necessary if the IPv6 header the and option header and IPv6 header
1162 * is not contiguous in order to return an ICMPv6 error.
1163 */
1164 int
1165 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1166 {
1167 struct ip6_hdr *ip6;
1168
1169 switch (IP6OPT_TYPE(*optp)) {
1170 case IP6OPT_TYPE_SKIP: /* ignore the option */
1171 return ((int)*(optp + 1));
1172 case IP6OPT_TYPE_DISCARD: /* silently discard */
1173 m_freem(m);
1174 return (-1);
1175 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1176 IP6STAT_INC(ip6s_badoptions);
1177 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1178 return (-1);
1179 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1180 IP6STAT_INC(ip6s_badoptions);
1181 ip6 = mtod(m, struct ip6_hdr *);
1182 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1183 (m->m_flags & (M_BCAST|M_MCAST)))
1184 m_freem(m);
1185 else
1186 icmp6_error(m, ICMP6_PARAM_PROB,
1187 ICMP6_PARAMPROB_OPTION, off);
1188 return (-1);
1189 }
1190
1191 m_freem(m); /* XXX: NOTREACHED */
1192 return (-1);
1193 }
1194
1195 /*
1196 * Create the "control" list for this pcb.
1197 * These functions will not modify mbuf chain at all.
1198 *
1199 * With KAME mbuf chain restriction:
1200 * The routine will be called from upper layer handlers like tcp6_input().
1201 * Thus the routine assumes that the caller (tcp6_input) have already
1202 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1203 * very first mbuf on the mbuf chain.
1204 *
1205 * ip6_savecontrol_v4 will handle those options that are possible to be
1206 * set on a v4-mapped socket.
1207 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1208 * options and handle the v6-only ones itself.
1209 */
1210 struct mbuf **
1211 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1212 int *v4only)
1213 {
1214 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1215
1216 #ifdef SO_TIMESTAMP
1217 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1218 union {
1219 struct timeval tv;
1220 struct bintime bt;
1221 struct timespec ts;
1222 } t;
1223
1224 switch (inp->inp_socket->so_ts_clock) {
1225 case SO_TS_REALTIME_MICRO:
1226 microtime(&t.tv);
1227 *mp = sbcreatecontrol((caddr_t) &t.tv, sizeof(t.tv),
1228 SCM_TIMESTAMP, SOL_SOCKET);
1229 if (*mp)
1230 mp = &(*mp)->m_next;
1231 break;
1232
1233 case SO_TS_BINTIME:
1234 bintime(&t.bt);
1235 *mp = sbcreatecontrol((caddr_t)&t.bt, sizeof(t.bt),
1236 SCM_BINTIME, SOL_SOCKET);
1237 if (*mp)
1238 mp = &(*mp)->m_next;
1239 break;
1240
1241 case SO_TS_REALTIME:
1242 nanotime(&t.ts);
1243 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts),
1244 SCM_REALTIME, SOL_SOCKET);
1245 if (*mp)
1246 mp = &(*mp)->m_next;
1247 break;
1248
1249 case SO_TS_MONOTONIC:
1250 nanouptime(&t.ts);
1251 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts),
1252 SCM_MONOTONIC, SOL_SOCKET);
1253 if (*mp)
1254 mp = &(*mp)->m_next;
1255 break;
1256
1257 default:
1258 panic("unknown (corrupted) so_ts_clock");
1259 }
1260 }
1261 #endif
1262
1263 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1264 /* RFC 2292 sec. 5 */
1265 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1266 struct in6_pktinfo pi6;
1267
1268 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1269 #ifdef INET
1270 struct ip *ip;
1271
1272 ip = mtod(m, struct ip *);
1273 pi6.ipi6_addr.s6_addr32[0] = 0;
1274 pi6.ipi6_addr.s6_addr32[1] = 0;
1275 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1276 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1277 #else
1278 /* We won't hit this code */
1279 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1280 #endif
1281 } else {
1282 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1283 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1284 }
1285 pi6.ipi6_ifindex =
1286 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1287
1288 *mp = sbcreatecontrol((caddr_t) &pi6,
1289 sizeof(struct in6_pktinfo),
1290 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1291 if (*mp)
1292 mp = &(*mp)->m_next;
1293 }
1294
1295 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1296 int hlim;
1297
1298 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1299 #ifdef INET
1300 struct ip *ip;
1301
1302 ip = mtod(m, struct ip *);
1303 hlim = ip->ip_ttl;
1304 #else
1305 /* We won't hit this code */
1306 hlim = 0;
1307 #endif
1308 } else {
1309 hlim = ip6->ip6_hlim & 0xff;
1310 }
1311 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1312 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1313 IPPROTO_IPV6);
1314 if (*mp)
1315 mp = &(*mp)->m_next;
1316 }
1317
1318 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1319 int tclass;
1320
1321 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1322 #ifdef INET
1323 struct ip *ip;
1324
1325 ip = mtod(m, struct ip *);
1326 tclass = ip->ip_tos;
1327 #else
1328 /* We won't hit this code */
1329 tclass = 0;
1330 #endif
1331 } else {
1332 u_int32_t flowinfo;
1333
1334 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1335 flowinfo >>= 20;
1336 tclass = flowinfo & 0xff;
1337 }
1338 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1339 IPV6_TCLASS, IPPROTO_IPV6);
1340 if (*mp)
1341 mp = &(*mp)->m_next;
1342 }
1343
1344 if (v4only != NULL) {
1345 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1346 *v4only = 1;
1347 } else {
1348 *v4only = 0;
1349 }
1350 }
1351
1352 return (mp);
1353 }
1354
1355 void
1356 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1357 {
1358 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1359 int v4only = 0;
1360
1361 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1362 if (v4only)
1363 return;
1364
1365 /*
1366 * IPV6_HOPOPTS socket option. Recall that we required super-user
1367 * privilege for the option (see ip6_ctloutput), but it might be too
1368 * strict, since there might be some hop-by-hop options which can be
1369 * returned to normal user.
1370 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1371 */
1372 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1373 /*
1374 * Check if a hop-by-hop options header is contatined in the
1375 * received packet, and if so, store the options as ancillary
1376 * data. Note that a hop-by-hop options header must be
1377 * just after the IPv6 header, which is assured through the
1378 * IPv6 input processing.
1379 */
1380 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1381 struct ip6_hbh *hbh;
1382 int hbhlen = 0;
1383 #ifdef PULLDOWN_TEST
1384 struct mbuf *ext;
1385 #endif
1386
1387 #ifndef PULLDOWN_TEST
1388 hbh = (struct ip6_hbh *)(ip6 + 1);
1389 hbhlen = (hbh->ip6h_len + 1) << 3;
1390 #else
1391 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1392 ip6->ip6_nxt);
1393 if (ext == NULL) {
1394 IP6STAT_INC(ip6s_tooshort);
1395 return;
1396 }
1397 hbh = mtod(ext, struct ip6_hbh *);
1398 hbhlen = (hbh->ip6h_len + 1) << 3;
1399 if (hbhlen != ext->m_len) {
1400 m_freem(ext);
1401 IP6STAT_INC(ip6s_tooshort);
1402 return;
1403 }
1404 #endif
1405
1406 /*
1407 * XXX: We copy the whole header even if a
1408 * jumbo payload option is included, the option which
1409 * is to be removed before returning according to
1410 * RFC2292.
1411 * Note: this constraint is removed in RFC3542
1412 */
1413 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1414 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1415 IPPROTO_IPV6);
1416 if (*mp)
1417 mp = &(*mp)->m_next;
1418 #ifdef PULLDOWN_TEST
1419 m_freem(ext);
1420 #endif
1421 }
1422 }
1423
1424 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1425 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1426
1427 /*
1428 * Search for destination options headers or routing
1429 * header(s) through the header chain, and stores each
1430 * header as ancillary data.
1431 * Note that the order of the headers remains in
1432 * the chain of ancillary data.
1433 */
1434 while (1) { /* is explicit loop prevention necessary? */
1435 struct ip6_ext *ip6e = NULL;
1436 int elen;
1437 #ifdef PULLDOWN_TEST
1438 struct mbuf *ext = NULL;
1439 #endif
1440
1441 /*
1442 * if it is not an extension header, don't try to
1443 * pull it from the chain.
1444 */
1445 switch (nxt) {
1446 case IPPROTO_DSTOPTS:
1447 case IPPROTO_ROUTING:
1448 case IPPROTO_HOPOPTS:
1449 case IPPROTO_AH: /* is it possible? */
1450 break;
1451 default:
1452 goto loopend;
1453 }
1454
1455 #ifndef PULLDOWN_TEST
1456 if (off + sizeof(*ip6e) > m->m_len)
1457 goto loopend;
1458 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1459 if (nxt == IPPROTO_AH)
1460 elen = (ip6e->ip6e_len + 2) << 2;
1461 else
1462 elen = (ip6e->ip6e_len + 1) << 3;
1463 if (off + elen > m->m_len)
1464 goto loopend;
1465 #else
1466 ext = ip6_pullexthdr(m, off, nxt);
1467 if (ext == NULL) {
1468 IP6STAT_INC(ip6s_tooshort);
1469 return;
1470 }
1471 ip6e = mtod(ext, struct ip6_ext *);
1472 if (nxt == IPPROTO_AH)
1473 elen = (ip6e->ip6e_len + 2) << 2;
1474 else
1475 elen = (ip6e->ip6e_len + 1) << 3;
1476 if (elen != ext->m_len) {
1477 m_freem(ext);
1478 IP6STAT_INC(ip6s_tooshort);
1479 return;
1480 }
1481 #endif
1482
1483 switch (nxt) {
1484 case IPPROTO_DSTOPTS:
1485 if (!(in6p->inp_flags & IN6P_DSTOPTS))
1486 break;
1487
1488 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1489 IS2292(in6p,
1490 IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1491 IPPROTO_IPV6);
1492 if (*mp)
1493 mp = &(*mp)->m_next;
1494 break;
1495 case IPPROTO_ROUTING:
1496 if (!(in6p->inp_flags & IN6P_RTHDR))
1497 break;
1498
1499 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1500 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1501 IPPROTO_IPV6);
1502 if (*mp)
1503 mp = &(*mp)->m_next;
1504 break;
1505 case IPPROTO_HOPOPTS:
1506 case IPPROTO_AH: /* is it possible? */
1507 break;
1508
1509 default:
1510 /*
1511 * other cases have been filtered in the above.
1512 * none will visit this case. here we supply
1513 * the code just in case (nxt overwritten or
1514 * other cases).
1515 */
1516 #ifdef PULLDOWN_TEST
1517 m_freem(ext);
1518 #endif
1519 goto loopend;
1520
1521 }
1522
1523 /* proceed with the next header. */
1524 off += elen;
1525 nxt = ip6e->ip6e_nxt;
1526 ip6e = NULL;
1527 #ifdef PULLDOWN_TEST
1528 m_freem(ext);
1529 ext = NULL;
1530 #endif
1531 }
1532 loopend:
1533 ;
1534 }
1535
1536 if (in6p->inp_flags2 & INP_RECVFLOWID) {
1537 uint32_t flowid, flow_type;
1538
1539 flowid = m->m_pkthdr.flowid;
1540 flow_type = M_HASHTYPE_GET(m);
1541
1542 /*
1543 * XXX should handle the failure of one or the
1544 * other - don't populate both?
1545 */
1546 *mp = sbcreatecontrol((caddr_t) &flowid,
1547 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6);
1548 if (*mp)
1549 mp = &(*mp)->m_next;
1550 *mp = sbcreatecontrol((caddr_t) &flow_type,
1551 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6);
1552 if (*mp)
1553 mp = &(*mp)->m_next;
1554 }
1555
1556 #ifdef RSS
1557 if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) {
1558 uint32_t flowid, flow_type;
1559 uint32_t rss_bucketid;
1560
1561 flowid = m->m_pkthdr.flowid;
1562 flow_type = M_HASHTYPE_GET(m);
1563
1564 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1565 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1566 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6);
1567 if (*mp)
1568 mp = &(*mp)->m_next;
1569 }
1570 }
1571 #endif
1572
1573 }
1574 #undef IS2292
1575
1576 void
1577 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1578 {
1579 struct socket *so;
1580 struct mbuf *m_mtu;
1581 struct ip6_mtuinfo mtuctl;
1582
1583 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1584 /*
1585 * Notify the error by sending IPV6_PATHMTU ancillary data if
1586 * application wanted to know the MTU value.
1587 * NOTE: we notify disconnected sockets, because some udp
1588 * applications keep sending sockets disconnected.
1589 * NOTE: our implementation doesn't notify connected sockets that has
1590 * foreign address that is different than given destination addresses
1591 * (this is permitted by RFC 3542).
1592 */
1593 if ((inp->inp_flags & IN6P_MTU) == 0 || (
1594 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1595 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1596 return;
1597
1598 mtuctl.ip6m_mtu = mtu;
1599 mtuctl.ip6m_addr = *dst;
1600 if (sa6_recoverscope(&mtuctl.ip6m_addr))
1601 return;
1602
1603 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1604 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1605 return;
1606
1607 so = inp->inp_socket;
1608 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1609 == 0) {
1610 m_freem(m_mtu);
1611 /* XXX: should count statistics */
1612 } else
1613 sorwakeup(so);
1614 }
1615
1616 #ifdef PULLDOWN_TEST
1617 /*
1618 * pull single extension header from mbuf chain. returns single mbuf that
1619 * contains the result, or NULL on error.
1620 */
1621 static struct mbuf *
1622 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1623 {
1624 struct ip6_ext ip6e;
1625 size_t elen;
1626 struct mbuf *n;
1627
1628 #ifdef DIAGNOSTIC
1629 switch (nxt) {
1630 case IPPROTO_DSTOPTS:
1631 case IPPROTO_ROUTING:
1632 case IPPROTO_HOPOPTS:
1633 case IPPROTO_AH: /* is it possible? */
1634 break;
1635 default:
1636 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1637 }
1638 #endif
1639
1640 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1641 if (nxt == IPPROTO_AH)
1642 elen = (ip6e.ip6e_len + 2) << 2;
1643 else
1644 elen = (ip6e.ip6e_len + 1) << 3;
1645
1646 if (elen > MLEN)
1647 n = m_getcl(M_NOWAIT, MT_DATA, 0);
1648 else
1649 n = m_get(M_NOWAIT, MT_DATA);
1650 if (n == NULL)
1651 return NULL;
1652
1653 m_copydata(m, off, elen, mtod(n, caddr_t));
1654 n->m_len = elen;
1655 return n;
1656 }
1657 #endif
1658
1659 /*
1660 * Get pointer to the previous header followed by the header
1661 * currently processed.
1662 */
1663 int
1664 ip6_get_prevhdr(const struct mbuf *m, int off)
1665 {
1666 struct ip6_ext ip6e;
1667 struct ip6_hdr *ip6;
1668 int len, nlen, nxt;
1669
1670 if (off == sizeof(struct ip6_hdr))
1671 return (offsetof(struct ip6_hdr, ip6_nxt));
1672 if (off < sizeof(struct ip6_hdr))
1673 panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1674
1675 ip6 = mtod(m, struct ip6_hdr *);
1676 nxt = ip6->ip6_nxt;
1677 len = sizeof(struct ip6_hdr);
1678 nlen = 0;
1679 while (len < off) {
1680 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1681 switch (nxt) {
1682 case IPPROTO_FRAGMENT:
1683 nlen = sizeof(struct ip6_frag);
1684 break;
1685 case IPPROTO_AH:
1686 nlen = (ip6e.ip6e_len + 2) << 2;
1687 break;
1688 default:
1689 nlen = (ip6e.ip6e_len + 1) << 3;
1690 }
1691 len += nlen;
1692 nxt = ip6e.ip6e_nxt;
1693 }
1694 return (len - nlen);
1695 }
1696
1697 /*
1698 * get next header offset. m will be retained.
1699 */
1700 int
1701 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1702 {
1703 struct ip6_hdr ip6;
1704 struct ip6_ext ip6e;
1705 struct ip6_frag fh;
1706
1707 /* just in case */
1708 if (m == NULL)
1709 panic("ip6_nexthdr: m == NULL");
1710 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1711 return -1;
1712
1713 switch (proto) {
1714 case IPPROTO_IPV6:
1715 if (m->m_pkthdr.len < off + sizeof(ip6))
1716 return -1;
1717 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1718 if (nxtp)
1719 *nxtp = ip6.ip6_nxt;
1720 off += sizeof(ip6);
1721 return off;
1722
1723 case IPPROTO_FRAGMENT:
1724 /*
1725 * terminate parsing if it is not the first fragment,
1726 * it does not make sense to parse through it.
1727 */
1728 if (m->m_pkthdr.len < off + sizeof(fh))
1729 return -1;
1730 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1731 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1732 if (fh.ip6f_offlg & IP6F_OFF_MASK)
1733 return -1;
1734 if (nxtp)
1735 *nxtp = fh.ip6f_nxt;
1736 off += sizeof(struct ip6_frag);
1737 return off;
1738
1739 case IPPROTO_AH:
1740 if (m->m_pkthdr.len < off + sizeof(ip6e))
1741 return -1;
1742 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1743 if (nxtp)
1744 *nxtp = ip6e.ip6e_nxt;
1745 off += (ip6e.ip6e_len + 2) << 2;
1746 return off;
1747
1748 case IPPROTO_HOPOPTS:
1749 case IPPROTO_ROUTING:
1750 case IPPROTO_DSTOPTS:
1751 if (m->m_pkthdr.len < off + sizeof(ip6e))
1752 return -1;
1753 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1754 if (nxtp)
1755 *nxtp = ip6e.ip6e_nxt;
1756 off += (ip6e.ip6e_len + 1) << 3;
1757 return off;
1758
1759 case IPPROTO_NONE:
1760 case IPPROTO_ESP:
1761 case IPPROTO_IPCOMP:
1762 /* give up */
1763 return -1;
1764
1765 default:
1766 return -1;
1767 }
1768
1769 /* NOTREACHED */
1770 }
1771
1772 /*
1773 * get offset for the last header in the chain. m will be kept untainted.
1774 */
1775 int
1776 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1777 {
1778 int newoff;
1779 int nxt;
1780
1781 if (!nxtp) {
1782 nxt = -1;
1783 nxtp = &nxt;
1784 }
1785 while (1) {
1786 newoff = ip6_nexthdr(m, off, proto, nxtp);
1787 if (newoff < 0)
1788 return off;
1789 else if (newoff < off)
1790 return -1; /* invalid */
1791 else if (newoff == off)
1792 return newoff;
1793
1794 off = newoff;
1795 proto = *nxtp;
1796 }
1797 }
1798
1799 /*
1800 * System control for IP6
1801 */
1802
1803 u_char inet6ctlerrmap[PRC_NCMDS] = {
1804 0, 0, 0, 0,
1805 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1806 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1807 EMSGSIZE, EHOSTUNREACH, 0, 0,
1808 0, 0, EHOSTUNREACH, 0,
1809 ENOPROTOOPT, ECONNREFUSED
1810 };
Cache object: 8d95b3985638895b420acb91fa075b71
|