1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * $KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $
32 */
33
34 /*-
35 * Copyright (c) 1982, 1986, 1988, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
49 *
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
61 */
62
63 #include "opt_inet.h"
64 #include "opt_inet6.h"
65 #include "opt_ipsec.h"
66 #include "opt_route.h"
67 #include "opt_rss.h"
68 #include "opt_sctp.h"
69
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/hhook.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/proc.h>
76 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/sdt.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/lock.h>
85 #include <sys/rmlock.h>
86 #include <sys/syslog.h>
87 #include <sys/sysctl.h>
88 #include <sys/eventhandler.h>
89
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_types.h>
93 #include <net/if_private.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/ip_encap.h>
114 #include <netinet/in_pcb.h>
115 #include <netinet/icmp6.h>
116 #include <netinet6/scope6_var.h>
117 #include <netinet6/in6_ifattach.h>
118 #include <netinet6/mld6_var.h>
119 #include <netinet6/nd6.h>
120 #include <netinet6/in6_rss.h>
121 #ifdef SCTP
122 #include <netinet/sctp_pcb.h>
123 #include <netinet6/sctp6_var.h>
124 #endif
125
126 #include <netipsec/ipsec_support.h>
127
128 ip6proto_input_t *ip6_protox[IPPROTO_MAX] = {
129 [0 ... IPPROTO_MAX - 1] = rip6_input };
130 ip6proto_ctlinput_t *ip6_ctlprotox[IPPROTO_MAX] = {
131 [0 ... IPPROTO_MAX - 1] = rip6_ctlinput };
132
133 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
134 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
135 VNET_DEFINE(u_long, in6_ifaddrhmask);
136
137 static struct netisr_handler ip6_nh = {
138 .nh_name = "ip6",
139 .nh_handler = ip6_input,
140 .nh_proto = NETISR_IPV6,
141 #ifdef RSS
142 .nh_m2cpuid = rss_soft_m2cpuid_v6,
143 .nh_policy = NETISR_POLICY_CPU,
144 .nh_dispatch = NETISR_DISPATCH_HYBRID,
145 #else
146 .nh_policy = NETISR_POLICY_FLOW,
147 #endif
148 };
149
150 static int
sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)151 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
152 {
153 int error, qlimit;
154
155 netisr_getqlimit(&ip6_nh, &qlimit);
156 error = sysctl_handle_int(oidp, &qlimit, 0, req);
157 if (error || !req->newptr)
158 return (error);
159 if (qlimit < 1)
160 return (EINVAL);
161 return (netisr_setqlimit(&ip6_nh, qlimit));
162 }
163 SYSCTL_DECL(_net_inet6_ip6);
164 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen,
165 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
166 0, 0, sysctl_netinet6_intr_queue_maxlen, "I",
167 "Maximum size of the IPv6 input queue");
168
169 VNET_DEFINE_STATIC(bool, ip6_sav) = true;
170 #define V_ip6_sav VNET(ip6_sav)
171 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation,
172 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true,
173 "Drop incoming packets with source address that is a local address");
174
175 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, temp_max_desync_factor,
176 CTLFLAG_RD | CTLFLAG_VNET,
177 &VNET_NAME(ip6_temp_max_desync_factor), 0,
178 "RFC 8981 max desync factor");
179
180 #ifdef RSS
181 static struct netisr_handler ip6_direct_nh = {
182 .nh_name = "ip6_direct",
183 .nh_handler = ip6_direct_input,
184 .nh_proto = NETISR_IPV6_DIRECT,
185 .nh_m2cpuid = rss_soft_m2cpuid_v6,
186 .nh_policy = NETISR_POLICY_CPU,
187 .nh_dispatch = NETISR_DISPATCH_HYBRID,
188 };
189
190 static int
sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)191 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
192 {
193 int error, qlimit;
194
195 netisr_getqlimit(&ip6_direct_nh, &qlimit);
196 error = sysctl_handle_int(oidp, &qlimit, 0, req);
197 if (error || !req->newptr)
198 return (error);
199 if (qlimit < 1)
200 return (EINVAL);
201 return (netisr_setqlimit(&ip6_direct_nh, qlimit));
202 }
203 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
204 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
205 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I",
206 "Maximum size of the IPv6 direct input queue");
207
208 #endif
209
210 VNET_DEFINE(pfil_head_t, inet6_pfil_head);
211 VNET_DEFINE(pfil_head_t, inet6_local_pfil_head);
212
213 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
214 VNET_PCPUSTAT_SYSINIT(ip6stat);
215 #ifdef VIMAGE
216 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
217 #endif /* VIMAGE */
218
219 struct rmlock in6_ifaddr_lock;
220 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
221
222 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
223
224 /*
225 * IP6 initialization: fill in IP6 protocol switch table.
226 * All protocols not implemented in kernel go to raw IP6 protocol handler.
227 */
228 static void
ip6_vnet_init(void * arg __unused)229 ip6_vnet_init(void *arg __unused)
230 {
231 struct pfil_head_args args;
232
233 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
234 &V_ip6_auto_linklocal);
235 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
236 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
237 TUNABLE_BOOL_FETCH("net.inet6.ip6.use_stableaddr", &V_ip6_use_stableaddr);
238
239 CK_STAILQ_INIT(&V_in6_ifaddrhead);
240 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
241 &V_in6_ifaddrhmask);
242
243 /* Initialize packet filter hooks. */
244 args.pa_version = PFIL_VERSION;
245 args.pa_flags = PFIL_IN | PFIL_OUT;
246 args.pa_type = PFIL_TYPE_IP6;
247 args.pa_headname = PFIL_INET6_NAME;
248 V_inet6_pfil_head = pfil_head_register(&args);
249
250 args.pa_flags = PFIL_OUT;
251 args.pa_headname = PFIL_INET6_LOCAL_NAME;
252 V_inet6_local_pfil_head = pfil_head_register(&args);
253
254 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
255 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
256 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
257 printf("%s: WARNING: unable to register input helper hook\n",
258 __func__);
259 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
260 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
261 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
262 printf("%s: WARNING: unable to register output helper hook\n",
263 __func__);
264
265 scope6_init();
266 addrsel_policy_init();
267 nd6_init();
268 frag6_init();
269
270 V_ip6_temp_max_desync_factor = TEMP_MAX_DESYNC_FACTOR_BASE +
271 (V_ip6_temp_preferred_lifetime >> 2) +
272 (V_ip6_temp_preferred_lifetime >> 3);
273 V_ip6_desync_factor = arc4random() % V_ip6_temp_max_desync_factor;
274
275 /* Skip global initialization stuff for non-default instances. */
276 #ifdef VIMAGE
277 netisr_register_vnet(&ip6_nh);
278 #ifdef RSS
279 netisr_register_vnet(&ip6_direct_nh);
280 #endif
281 #endif
282 }
283 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
284 ip6_vnet_init, NULL);
285
286 static void
ip6_init(void * arg __unused)287 ip6_init(void *arg __unused)
288 {
289 struct ifnet *ifp;
290
291 /*
292 * Register statically those protocols that are unlikely to ever go
293 * dynamic.
294 */
295 IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput);
296 IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL);
297 IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL);
298 IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL);
299 IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL);
300 IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL);
301 IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL);
302 IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL);
303 IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL);
304 #ifdef SCTP /* XXX: has a loadable & static version */
305 IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput);
306 #endif
307
308 EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT);
309 EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL,
310 LOWMEM_PRI_DEFAULT);
311
312 netisr_register(&ip6_nh);
313 #ifdef RSS
314 netisr_register(&ip6_direct_nh);
315 #endif
316 /*
317 * XXXGL: we use SYSINIT() here, but go over V_ifnet. See comment
318 * in sys/netinet/ip_input.c:ip_init().
319 */
320 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
321 in6_ifarrival(NULL, ifp);
322 }
323 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL);
324
325 int
ip6proto_register(uint8_t proto,ip6proto_input_t input,ip6proto_ctlinput_t ctl)326 ip6proto_register(uint8_t proto, ip6proto_input_t input,
327 ip6proto_ctlinput_t ctl)
328 {
329
330 MPASS(proto > 0);
331
332 if (ip6_protox[proto] == rip6_input) {
333 ip6_protox[proto] = input;
334 ip6_ctlprotox[proto] = ctl;
335 return (0);
336 } else
337 return (EEXIST);
338 }
339
340 int
ip6proto_unregister(uint8_t proto)341 ip6proto_unregister(uint8_t proto)
342 {
343
344 MPASS(proto > 0);
345
346 if (ip6_protox[proto] != rip6_input) {
347 ip6_protox[proto] = rip6_input;
348 ip6_ctlprotox[proto] = rip6_ctlinput;
349 return (0);
350 } else
351 return (ENOENT);
352 }
353
354 #ifdef VIMAGE
355 static void
ip6_destroy(void * unused __unused)356 ip6_destroy(void *unused __unused)
357 {
358 struct ifaddr *ifa, *nifa;
359 struct ifnet *ifp;
360 int error;
361
362 #ifdef RSS
363 netisr_unregister_vnet(&ip6_direct_nh);
364 #endif
365 netisr_unregister_vnet(&ip6_nh);
366
367 pfil_head_unregister(V_inet6_pfil_head);
368 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
369 if (error != 0) {
370 printf("%s: WARNING: unable to deregister input helper hook "
371 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
372 "error %d returned\n", __func__, error);
373 }
374 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
375 if (error != 0) {
376 printf("%s: WARNING: unable to deregister output helper hook "
377 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
378 "error %d returned\n", __func__, error);
379 }
380
381 /* Cleanup addresses. */
382 IFNET_RLOCK();
383 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
384 /* Cannot lock here - lock recursion. */
385 /* IF_ADDR_LOCK(ifp); */
386 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
387 if (ifa->ifa_addr->sa_family != AF_INET6)
388 continue;
389 in6_purgeaddr(ifa);
390 }
391 /* IF_ADDR_UNLOCK(ifp); */
392 in6_ifdetach_destroy(ifp);
393 mld_domifdetach(ifp);
394 }
395 IFNET_RUNLOCK();
396
397 /* Make sure any routes are gone as well. */
398 rib_flush_routes_family(AF_INET6);
399
400 frag6_destroy();
401 nd6_destroy();
402 in6_ifattach_destroy();
403
404 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
405 }
406
407 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL);
408 #endif
409
410 static int
ip6_input_hbh(struct mbuf ** mp,uint32_t * plen,uint32_t * rtalert,int * off,int * nxt,int * ours)411 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off,
412 int *nxt, int *ours)
413 {
414 struct mbuf *m;
415 struct ip6_hdr *ip6;
416 struct ip6_hbh *hbh;
417
418 if (ip6_hopopts_input(plen, rtalert, mp, off)) {
419 #if 0 /*touches NULL pointer*/
420 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard);
421 #endif
422 goto out; /* m have already been freed */
423 }
424
425 /* adjust pointer */
426 m = *mp;
427 ip6 = mtod(m, struct ip6_hdr *);
428
429 /*
430 * if the payload length field is 0 and the next header field
431 * indicates Hop-by-Hop Options header, then a Jumbo Payload
432 * option MUST be included.
433 */
434 if (ip6->ip6_plen == 0 && *plen == 0) {
435 /*
436 * Note that if a valid jumbo payload option is
437 * contained, ip6_hopopts_input() must set a valid
438 * (non-zero) payload length to the variable plen.
439 */
440 IP6STAT_INC(ip6s_badoptions);
441 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
442 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
443 icmp6_error(m, ICMP6_PARAM_PROB,
444 ICMP6_PARAMPROB_HEADER,
445 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
446 goto out;
447 }
448 /* ip6_hopopts_input() ensures that mbuf is contiguous */
449 hbh = (struct ip6_hbh *)(ip6 + 1);
450 *nxt = hbh->ip6h_nxt;
451
452 /*
453 * If we are acting as a router and the packet contains a
454 * router alert option, see if we know the option value.
455 * Currently, we only support the option value for MLD, in which
456 * case we should pass the packet to the multicast routing
457 * daemon.
458 */
459 if (*rtalert != ~0) {
460 switch (*rtalert) {
461 case IP6OPT_RTALERT_MLD:
462 if (V_ip6_forwarding)
463 *ours = 1;
464 break;
465 default:
466 /*
467 * RFC2711 requires unrecognized values must be
468 * silently ignored.
469 */
470 break;
471 }
472 }
473
474 return (0);
475
476 out:
477 return (1);
478 }
479
480 #ifdef RSS
481 /*
482 * IPv6 direct input routine.
483 *
484 * This is called when reinjecting completed fragments where
485 * all of the previous checking and book-keeping has been done.
486 */
487 void
ip6_direct_input(struct mbuf * m)488 ip6_direct_input(struct mbuf *m)
489 {
490 int off, nxt;
491 int nest;
492 struct m_tag *mtag;
493 struct ip6_direct_ctx *ip6dc;
494
495 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
496 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
497
498 ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
499 nxt = ip6dc->ip6dc_nxt;
500 off = ip6dc->ip6dc_off;
501
502 nest = 0;
503
504 m_tag_delete(m, mtag);
505
506 while (nxt != IPPROTO_DONE) {
507 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
508 IP6STAT_INC(ip6s_toomanyhdr);
509 goto bad;
510 }
511
512 /*
513 * protection against faulty packet - there should be
514 * more sanity checks in header chain processing.
515 */
516 if (m->m_pkthdr.len < off) {
517 IP6STAT_INC(ip6s_tooshort);
518 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
519 goto bad;
520 }
521
522 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
523 if (IPSEC_ENABLED(ipv6)) {
524 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
525 return;
526 }
527 #endif /* IPSEC */
528
529 nxt = ip6_protox[nxt](&m, &off, nxt);
530 }
531 return;
532 bad:
533 m_freem(m);
534 }
535 #endif
536
537 void
ip6_input(struct mbuf * m)538 ip6_input(struct mbuf *m)
539 {
540 struct in6_addr odst;
541 struct ip6_hdr *ip6;
542 struct in6_ifaddr *ia;
543 struct ifnet *rcvif;
544 u_int32_t plen;
545 u_int32_t rtalert = ~0;
546 int off = sizeof(struct ip6_hdr), nest;
547 int nxt, ours = 0;
548 int srcrt = 0;
549
550 /*
551 * Drop the packet if IPv6 operation is disabled on the interface.
552 */
553 rcvif = m->m_pkthdr.rcvif;
554 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED))
555 goto bad;
556
557 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
558 /*
559 * should the inner packet be considered authentic?
560 * see comment in ah4_input().
561 * NB: m cannot be NULL when passed to the input routine
562 */
563
564 m->m_flags &= ~M_AUTHIPHDR;
565 m->m_flags &= ~M_AUTHIPDGM;
566
567 #endif /* IPSEC */
568
569 if (m->m_flags & M_FASTFWD_OURS) {
570 /*
571 * Firewall changed destination to local.
572 */
573 ip6 = mtod(m, struct ip6_hdr *);
574 goto passin;
575 }
576
577 /*
578 * mbuf statistics
579 */
580 if (m->m_flags & M_EXT) {
581 if (m->m_next)
582 IP6STAT_INC(ip6s_mext2m);
583 else
584 IP6STAT_INC(ip6s_mext1);
585 } else {
586 if (m->m_next) {
587 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif;
588 int ifindex = ifp->if_index;
589 if (ifindex >= IP6S_M2MMAX)
590 ifindex = 0;
591 IP6STAT_INC2(ip6s_m2m, ifindex);
592 } else
593 IP6STAT_INC(ip6s_m1);
594 }
595
596 in6_ifstat_inc(rcvif, ifs6_in_receive);
597 IP6STAT_INC(ip6s_total);
598
599 /*
600 * L2 bridge code and some other code can return mbuf chain
601 * that does not conform to KAME requirement. too bad.
602 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
603 */
604 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
605 struct mbuf *n;
606
607 if (m->m_pkthdr.len > MHLEN)
608 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
609 else
610 n = m_gethdr(M_NOWAIT, MT_DATA);
611 if (n == NULL)
612 goto bad;
613
614 m_move_pkthdr(n, m);
615 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
616 n->m_len = n->m_pkthdr.len;
617 m_freem(m);
618 m = n;
619 }
620 if (m->m_len < sizeof(struct ip6_hdr)) {
621 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
622 IP6STAT_INC(ip6s_toosmall);
623 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
624 goto bad;
625 }
626 }
627
628 ip6 = mtod(m, struct ip6_hdr *);
629 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
630 IP6STAT_INC(ip6s_badvers);
631 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
632 goto bad;
633 }
634
635 IP6STAT_INC2(ip6s_nxthist, ip6->ip6_nxt);
636 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6);
637
638 /*
639 * Check against address spoofing/corruption. The unspecified address
640 * is checked further below.
641 */
642 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
643 /*
644 * XXX: "badscope" is not very suitable for a multicast source.
645 */
646 IP6STAT_INC(ip6s_badscope);
647 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
648 goto bad;
649 }
650 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
651 !(m->m_flags & M_LOOP)) {
652 /*
653 * In this case, the packet should come from the loopback
654 * interface. However, we cannot just check the if_flags,
655 * because ip6_mloopback() passes the "actual" interface
656 * as the outgoing/incoming interface.
657 */
658 IP6STAT_INC(ip6s_badscope);
659 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
660 goto bad;
661 }
662 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
663 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
664 /*
665 * RFC4291 2.7:
666 * Nodes must not originate a packet to a multicast address
667 * whose scop field contains the reserved value 0; if such
668 * a packet is received, it must be silently dropped.
669 */
670 IP6STAT_INC(ip6s_badscope);
671 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
672 goto bad;
673 }
674 /*
675 * The following check is not documented in specs. A malicious
676 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
677 * and bypass security checks (act as if it was from 127.0.0.1 by using
678 * IPv6 src ::ffff:127.0.0.1). Be cautious.
679 *
680 * We have supported IPv6-only kernels for a few years and this issue
681 * has not come up. The world seems to move mostly towards not using
682 * v4mapped on the wire, so it makes sense for us to keep rejecting
683 * any such packets.
684 */
685 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
686 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
687 IP6STAT_INC(ip6s_badscope);
688 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
689 goto bad;
690 }
691 #if 0
692 /*
693 * Reject packets with IPv4 compatible addresses (auto tunnel).
694 *
695 * The code forbids auto tunnel relay case in RFC1933 (the check is
696 * stronger than RFC1933). We may want to re-enable it if mech-xx
697 * is revised to forbid relaying case.
698 */
699 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
700 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
701 IP6STAT_INC(ip6s_badscope);
702 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
703 goto bad;
704 }
705 #endif
706 /*
707 * Try to forward the packet, but if we fail continue.
708 * ip6_tryforward() does not generate redirects, so fall
709 * through to normal processing if redirects are required.
710 * ip6_tryforward() does inbound and outbound packet firewall
711 * processing. If firewall has decided that destination becomes
712 * our local address, it sets M_FASTFWD_OURS flag. In this
713 * case skip another inbound firewall processing and update
714 * ip6 pointer.
715 */
716 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0
717 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
718 && (!IPSEC_ENABLED(ipv6) ||
719 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0)
720 #endif
721 ) {
722 if ((m = ip6_tryforward(m)) == NULL)
723 return;
724 if (m->m_flags & M_FASTFWD_OURS) {
725 ip6 = mtod(m, struct ip6_hdr *);
726 goto passin;
727 }
728 }
729 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
730 /*
731 * Bypass packet filtering for packets previously handled by IPsec.
732 */
733 if (IPSEC_ENABLED(ipv6) &&
734 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0)
735 goto passin;
736 #endif
737 /*
738 * Run through list of hooks for input packets.
739 *
740 * NB: Beware of the destination address changing
741 * (e.g. by NAT rewriting). When this happens,
742 * tell ip6_forward to do the right thing.
743 */
744
745 /* Jump over all PFIL processing if hooks are not active. */
746 if (!PFIL_HOOKED_IN(V_inet6_pfil_head))
747 goto passin;
748
749 odst = ip6->ip6_dst;
750 if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif,
751 NULL) != PFIL_PASS)
752 return;
753 ip6 = mtod(m, struct ip6_hdr *);
754 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
755 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP &&
756 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
757 /*
758 * Directly ship the packet on. This allows forwarding
759 * packets originally destined to us to some other directly
760 * connected host.
761 */
762 ip6_forward(m, 1);
763 return;
764 }
765
766 passin:
767 /*
768 * The check is deferred to here to give firewalls a chance to block
769 * (and log) such packets. ip6_tryforward() will not process such
770 * packets.
771 */
772 if (__predict_false(IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst))) {
773 IP6STAT_INC(ip6s_badscope);
774 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
775 goto bad;
776 }
777
778 /*
779 * Disambiguate address scope zones (if there is ambiguity).
780 * We first make sure that the original source or destination address
781 * is not in our internal form for scoped addresses. Such addresses
782 * are not necessarily invalid spec-wise, but we cannot accept them due
783 * to the usage conflict.
784 * in6_setscope() then also checks and rejects the cases where src or
785 * dst are the loopback address and the receiving interface
786 * is not loopback.
787 */
788 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
789 IP6STAT_INC(ip6s_badscope); /* XXX */
790 goto bad;
791 }
792 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) ||
793 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) {
794 IP6STAT_INC(ip6s_badscope);
795 goto bad;
796 }
797 if (m->m_flags & M_FASTFWD_OURS) {
798 m->m_flags &= ~M_FASTFWD_OURS;
799 ours = 1;
800 goto hbhcheck;
801 }
802 /*
803 * Multicast check. Assume packet is for us to avoid
804 * prematurely taking locks.
805 */
806 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
807 ours = 1;
808 in6_ifstat_inc(rcvif, ifs6_in_mcast);
809 goto hbhcheck;
810 }
811 /*
812 * Unicast check
813 * XXX: For now we keep link-local IPv6 addresses with embedded
814 * scope zone id, therefore we use zero zoneid here.
815 */
816 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
817 if (ia != NULL) {
818 if (ia->ia6_flags & IN6_IFF_NOTREADY) {
819 char ip6bufs[INET6_ADDRSTRLEN];
820 char ip6bufd[INET6_ADDRSTRLEN];
821 /* address is not ready, so discard the packet. */
822 nd6log((LOG_INFO,
823 "ip6_input: packet to an unready address %s->%s\n",
824 ip6_sprintf(ip6bufs, &ip6->ip6_src),
825 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
826 goto bad;
827 }
828 if (V_ip6_sav && !(m->m_flags & M_LOOP) &&
829 __predict_false(in6_localip_fib(&ip6->ip6_src,
830 rcvif->if_fib))) {
831 IP6STAT_INC(ip6s_badscope); /* XXX */
832 goto bad;
833 }
834 /* Count the packet in the ip address stats */
835 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
836 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
837 ours = 1;
838 goto hbhcheck;
839 }
840
841 /*
842 * Now there is no reason to process the packet if it's not our own
843 * and we're not a router.
844 */
845 if (!V_ip6_forwarding) {
846 IP6STAT_INC(ip6s_cantforward);
847 goto bad;
848 }
849
850 hbhcheck:
851 /*
852 * Process Hop-by-Hop options header if it's contained.
853 * m may be modified in ip6_hopopts_input().
854 * If a JumboPayload option is included, plen will also be modified.
855 */
856 plen = (u_int32_t)ntohs(ip6->ip6_plen);
857 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
858 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0)
859 return;
860 } else
861 nxt = ip6->ip6_nxt;
862
863 /*
864 * Use mbuf flags to propagate Router Alert option to
865 * ICMPv6 layer, as hop-by-hop options have been stripped.
866 */
867 if (rtalert != ~0)
868 m->m_flags |= M_RTALERT_MLD;
869
870 /*
871 * Check that the amount of data in the buffers
872 * is as at least much as the IPv6 header would have us expect.
873 * Trim mbufs if longer than we expect.
874 * Drop packet if shorter than we expect.
875 */
876 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
877 IP6STAT_INC(ip6s_tooshort);
878 in6_ifstat_inc(rcvif, ifs6_in_truncated);
879 goto bad;
880 }
881 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
882 if (m->m_len == m->m_pkthdr.len) {
883 m->m_len = sizeof(struct ip6_hdr) + plen;
884 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
885 } else
886 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
887 }
888
889 /*
890 * Forward if desirable.
891 */
892 if (V_ip6_mrouter &&
893 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
894 /*
895 * If we are acting as a multicast router, all
896 * incoming multicast packets are passed to the
897 * kernel-level multicast forwarding function.
898 * The packet is returned (relatively) intact; if
899 * ip6_mforward() returns a non-zero value, the packet
900 * must be discarded, else it may be accepted below.
901 *
902 * XXX TODO: Check hlim and multicast scope here to avoid
903 * unnecessarily calling into ip6_mforward().
904 */
905 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) {
906 IP6STAT_INC(ip6s_cantforward);
907 goto bad;
908 }
909 } else if (!ours) {
910 ip6_forward(m, srcrt);
911 return;
912 }
913
914 /*
915 * We are going to ship the packet to the local protocol stack. Call the
916 * filter again for this 'output' action, allowing redirect-like rules
917 * to adjust the source address.
918 */
919 if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) {
920 if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) !=
921 PFIL_PASS)
922 return;
923 ip6 = mtod(m, struct ip6_hdr *);
924 }
925
926 /*
927 * Tell launch routine the next header
928 */
929 IP6STAT_INC(ip6s_delivered);
930 in6_ifstat_inc(rcvif, ifs6_in_deliver);
931 nest = 0;
932
933 while (nxt != IPPROTO_DONE) {
934 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
935 IP6STAT_INC(ip6s_toomanyhdr);
936 goto bad;
937 }
938
939 /*
940 * protection against faulty packet - there should be
941 * more sanity checks in header chain processing.
942 */
943 if (m->m_pkthdr.len < off) {
944 IP6STAT_INC(ip6s_tooshort);
945 in6_ifstat_inc(rcvif, ifs6_in_truncated);
946 goto bad;
947 }
948
949 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
950 if (IPSEC_ENABLED(ipv6)) {
951 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
952 return;
953 }
954 #endif /* IPSEC */
955
956 nxt = ip6_protox[nxt](&m, &off, nxt);
957 }
958 return;
959 bad:
960 in6_ifstat_inc(rcvif, ifs6_in_discard);
961 if (m != NULL)
962 m_freem(m);
963 }
964
965 /*
966 * Hop-by-Hop options header processing. If a valid jumbo payload option is
967 * included, the real payload length will be stored in plenp.
968 *
969 * rtalertp - XXX: should be stored more smart way
970 */
971 static int
ip6_hopopts_input(u_int32_t * plenp,u_int32_t * rtalertp,struct mbuf ** mp,int * offp)972 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
973 struct mbuf **mp, int *offp)
974 {
975 struct mbuf *m = *mp;
976 int off = *offp, hbhlen;
977 struct ip6_hbh *hbh;
978
979 /* validation of the length of the header */
980 if (m->m_len < off + sizeof(*hbh)) {
981 m = m_pullup(m, off + sizeof(*hbh));
982 if (m == NULL) {
983 IP6STAT_INC(ip6s_exthdrtoolong);
984 *mp = NULL;
985 return (-1);
986 }
987 }
988 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
989 hbhlen = (hbh->ip6h_len + 1) << 3;
990
991 if (m->m_len < off + hbhlen) {
992 m = m_pullup(m, off + hbhlen);
993 if (m == NULL) {
994 IP6STAT_INC(ip6s_exthdrtoolong);
995 *mp = NULL;
996 return (-1);
997 }
998 }
999 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1000 off += hbhlen;
1001 hbhlen -= sizeof(struct ip6_hbh);
1002 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1003 hbhlen, rtalertp, plenp) < 0) {
1004 *mp = NULL;
1005 return (-1);
1006 }
1007
1008 *offp = off;
1009 *mp = m;
1010 return (0);
1011 }
1012
1013 /*
1014 * Search header for all Hop-by-hop options and process each option.
1015 * This function is separate from ip6_hopopts_input() in order to
1016 * handle a case where the sending node itself process its hop-by-hop
1017 * options header. In such a case, the function is called from ip6_output().
1018 *
1019 * The function assumes that hbh header is located right after the IPv6 header
1020 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1021 * opthead + hbhlen is located in contiguous memory region.
1022 */
1023 int
ip6_process_hopopts(struct mbuf * m,u_int8_t * opthead,int hbhlen,u_int32_t * rtalertp,u_int32_t * plenp)1024 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1025 u_int32_t *rtalertp, u_int32_t *plenp)
1026 {
1027 struct ip6_hdr *ip6;
1028 int optlen = 0;
1029 u_int8_t *opt = opthead;
1030 u_int16_t rtalert_val;
1031 u_int32_t jumboplen;
1032 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1033
1034 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1035 switch (*opt) {
1036 case IP6OPT_PAD1:
1037 optlen = 1;
1038 break;
1039 case IP6OPT_PADN:
1040 if (hbhlen < IP6OPT_MINLEN) {
1041 IP6STAT_INC(ip6s_toosmall);
1042 goto bad;
1043 }
1044 optlen = *(opt + 1) + 2;
1045 break;
1046 case IP6OPT_ROUTER_ALERT:
1047 /* XXX may need check for alignment */
1048 if (hbhlen < IP6OPT_RTALERT_LEN) {
1049 IP6STAT_INC(ip6s_toosmall);
1050 goto bad;
1051 }
1052 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1053 /* XXX stat */
1054 icmp6_error(m, ICMP6_PARAM_PROB,
1055 ICMP6_PARAMPROB_HEADER,
1056 erroff + opt + 1 - opthead);
1057 return (-1);
1058 }
1059 optlen = IP6OPT_RTALERT_LEN;
1060 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1061 *rtalertp = ntohs(rtalert_val);
1062 break;
1063 case IP6OPT_JUMBO:
1064 /* XXX may need check for alignment */
1065 if (hbhlen < IP6OPT_JUMBO_LEN) {
1066 IP6STAT_INC(ip6s_toosmall);
1067 goto bad;
1068 }
1069 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1070 /* XXX stat */
1071 icmp6_error(m, ICMP6_PARAM_PROB,
1072 ICMP6_PARAMPROB_HEADER,
1073 erroff + opt + 1 - opthead);
1074 return (-1);
1075 }
1076 optlen = IP6OPT_JUMBO_LEN;
1077
1078 /*
1079 * IPv6 packets that have non 0 payload length
1080 * must not contain a jumbo payload option.
1081 */
1082 ip6 = mtod(m, struct ip6_hdr *);
1083 if (ip6->ip6_plen) {
1084 IP6STAT_INC(ip6s_badoptions);
1085 icmp6_error(m, ICMP6_PARAM_PROB,
1086 ICMP6_PARAMPROB_HEADER,
1087 erroff + opt - opthead);
1088 return (-1);
1089 }
1090
1091 /*
1092 * We may see jumbolen in unaligned location, so
1093 * we'd need to perform bcopy().
1094 */
1095 bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1096 jumboplen = (u_int32_t)htonl(jumboplen);
1097
1098 #if 1
1099 /*
1100 * if there are multiple jumbo payload options,
1101 * *plenp will be non-zero and the packet will be
1102 * rejected.
1103 * the behavior may need some debate in ipngwg -
1104 * multiple options does not make sense, however,
1105 * there's no explicit mention in specification.
1106 */
1107 if (*plenp != 0) {
1108 IP6STAT_INC(ip6s_badoptions);
1109 icmp6_error(m, ICMP6_PARAM_PROB,
1110 ICMP6_PARAMPROB_HEADER,
1111 erroff + opt + 2 - opthead);
1112 return (-1);
1113 }
1114 #endif
1115
1116 /*
1117 * jumbo payload length must be larger than 65535.
1118 */
1119 if (jumboplen <= IPV6_MAXPACKET) {
1120 IP6STAT_INC(ip6s_badoptions);
1121 icmp6_error(m, ICMP6_PARAM_PROB,
1122 ICMP6_PARAMPROB_HEADER,
1123 erroff + opt + 2 - opthead);
1124 return (-1);
1125 }
1126 *plenp = jumboplen;
1127
1128 break;
1129 default: /* unknown option */
1130 if (hbhlen < IP6OPT_MINLEN) {
1131 IP6STAT_INC(ip6s_toosmall);
1132 goto bad;
1133 }
1134 optlen = ip6_unknown_opt(opt, m,
1135 erroff + opt - opthead);
1136 if (optlen == -1)
1137 return (-1);
1138 optlen += 2;
1139 break;
1140 }
1141 }
1142
1143 return (0);
1144
1145 bad:
1146 m_freem(m);
1147 return (-1);
1148 }
1149
1150 /*
1151 * Unknown option processing.
1152 * The third argument `off' is the offset from the IPv6 header to the option,
1153 * which is necessary if the IPv6 header the and option header and IPv6 header
1154 * is not contiguous in order to return an ICMPv6 error.
1155 */
1156 int
ip6_unknown_opt(u_int8_t * optp,struct mbuf * m,int off)1157 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1158 {
1159 struct ip6_hdr *ip6;
1160
1161 switch (IP6OPT_TYPE(*optp)) {
1162 case IP6OPT_TYPE_SKIP: /* ignore the option */
1163 return ((int)*(optp + 1));
1164 case IP6OPT_TYPE_DISCARD: /* silently discard */
1165 m_freem(m);
1166 return (-1);
1167 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1168 IP6STAT_INC(ip6s_badoptions);
1169 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1170 return (-1);
1171 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1172 IP6STAT_INC(ip6s_badoptions);
1173 ip6 = mtod(m, struct ip6_hdr *);
1174 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1175 (m->m_flags & (M_BCAST|M_MCAST)))
1176 m_freem(m);
1177 else
1178 icmp6_error(m, ICMP6_PARAM_PROB,
1179 ICMP6_PARAMPROB_OPTION, off);
1180 return (-1);
1181 }
1182
1183 m_freem(m); /* XXX: NOTREACHED */
1184 return (-1);
1185 }
1186
1187 /*
1188 * Create the "control" list for this pcb.
1189 * These functions will not modify mbuf chain at all.
1190 *
1191 * The routine will be called from upper layer handlers like tcp6_input().
1192 * Thus the routine assumes that the caller (tcp6_input) have already
1193 * called m_pullup() and all the extension headers are located in the
1194 * very first mbuf on the mbuf chain.
1195 *
1196 * ip6_savecontrol_v4 will handle those options that are possible to be
1197 * set on a v4-mapped socket.
1198 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1199 * options and handle the v6-only ones itself.
1200 */
1201 struct mbuf **
ip6_savecontrol_v4(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp,int * v4only)1202 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1203 int *v4only)
1204 {
1205 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1206
1207 #if defined(SO_TIMESTAMP) && defined(SO_BINTIME)
1208 if ((inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) != 0) {
1209 union {
1210 struct timeval tv;
1211 struct bintime bt;
1212 struct timespec ts;
1213 } t;
1214 struct bintime boottimebin, bt1;
1215 struct timespec ts1;
1216 int ts_clock;
1217 bool stamped;
1218
1219 ts_clock = inp->inp_socket->so_ts_clock;
1220 stamped = false;
1221
1222 /*
1223 * Handle BINTIME first. We create the same output options
1224 * for both SO_BINTIME and the case where SO_TIMESTAMP is
1225 * set with the timestamp clock set to SO_TS_BINTIME.
1226 */
1227 if ((inp->inp_socket->so_options & SO_BINTIME) != 0 ||
1228 ts_clock == SO_TS_BINTIME) {
1229 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1230 M_TSTMP)) {
1231 mbuf_tstmp2timespec(m, &ts1);
1232 timespec2bintime(&ts1, &t.bt);
1233 getboottimebin(&boottimebin);
1234 bintime_add(&t.bt, &boottimebin);
1235 } else {
1236 bintime(&t.bt);
1237 }
1238 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME,
1239 SOL_SOCKET, M_NOWAIT);
1240 if (*mp != NULL) {
1241 mp = &(*mp)->m_next;
1242 stamped = true;
1243 }
1244
1245 /*
1246 * Suppress other timestamps if SO_TIMESTAMP is not
1247 * set.
1248 */
1249 if ((inp->inp_socket->so_options & SO_TIMESTAMP) == 0)
1250 ts_clock = SO_TS_BINTIME;
1251 }
1252
1253 switch (ts_clock) {
1254 case SO_TS_REALTIME_MICRO:
1255 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1256 M_TSTMP)) {
1257 mbuf_tstmp2timespec(m, &ts1);
1258 timespec2bintime(&ts1, &bt1);
1259 getboottimebin(&boottimebin);
1260 bintime_add(&bt1, &boottimebin);
1261 bintime2timeval(&bt1, &t.tv);
1262 } else {
1263 microtime(&t.tv);
1264 }
1265 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv),
1266 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT);
1267 if (*mp != NULL) {
1268 mp = &(*mp)->m_next;
1269 stamped = true;
1270 }
1271 break;
1272
1273 case SO_TS_BINTIME:
1274 break;
1275
1276 case SO_TS_REALTIME:
1277 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1278 M_TSTMP)) {
1279 mbuf_tstmp2timespec(m, &t.ts);
1280 getboottimebin(&boottimebin);
1281 bintime2timespec(&boottimebin, &ts1);
1282 timespecadd(&t.ts, &ts1, &t.ts);
1283 } else {
1284 nanotime(&t.ts);
1285 }
1286 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1287 SCM_REALTIME, SOL_SOCKET, M_NOWAIT);
1288 if (*mp != NULL) {
1289 mp = &(*mp)->m_next;
1290 stamped = true;
1291 }
1292 break;
1293
1294 case SO_TS_MONOTONIC:
1295 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1296 M_TSTMP))
1297 mbuf_tstmp2timespec(m, &t.ts);
1298 else
1299 nanouptime(&t.ts);
1300 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1301 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT);
1302 if (*mp != NULL) {
1303 mp = &(*mp)->m_next;
1304 stamped = true;
1305 }
1306 break;
1307
1308 default:
1309 panic("unknown (corrupted) so_ts_clock");
1310 }
1311 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) ==
1312 (M_PKTHDR | M_TSTMP)) {
1313 struct sock_timestamp_info sti;
1314
1315 bzero(&sti, sizeof(sti));
1316 sti.st_info_flags = ST_INFO_HW;
1317 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1318 sti.st_info_flags |= ST_INFO_HW_HPREC;
1319 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1320 SOL_SOCKET, M_NOWAIT);
1321 if (*mp != NULL)
1322 mp = &(*mp)->m_next;
1323 }
1324 }
1325 #endif
1326
1327 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1328 /* RFC 2292 sec. 5 */
1329 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1330 struct in6_pktinfo pi6;
1331
1332 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1333 #ifdef INET
1334 struct ip *ip;
1335
1336 ip = mtod(m, struct ip *);
1337 pi6.ipi6_addr.s6_addr32[0] = 0;
1338 pi6.ipi6_addr.s6_addr32[1] = 0;
1339 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1340 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1341 #else
1342 /* We won't hit this code */
1343 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1344 #endif
1345 } else {
1346 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1347 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1348 }
1349 pi6.ipi6_ifindex =
1350 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1351
1352 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo),
1353 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6,
1354 M_NOWAIT);
1355 if (*mp)
1356 mp = &(*mp)->m_next;
1357 }
1358
1359 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1360 int hlim;
1361
1362 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1363 #ifdef INET
1364 struct ip *ip;
1365
1366 ip = mtod(m, struct ip *);
1367 hlim = ip->ip_ttl;
1368 #else
1369 /* We won't hit this code */
1370 hlim = 0;
1371 #endif
1372 } else {
1373 hlim = ip6->ip6_hlim & 0xff;
1374 }
1375 *mp = sbcreatecontrol(&hlim, sizeof(int),
1376 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1377 IPPROTO_IPV6, M_NOWAIT);
1378 if (*mp)
1379 mp = &(*mp)->m_next;
1380 }
1381
1382 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1383 int tclass;
1384
1385 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1386 #ifdef INET
1387 struct ip *ip;
1388
1389 ip = mtod(m, struct ip *);
1390 tclass = ip->ip_tos;
1391 #else
1392 /* We won't hit this code */
1393 tclass = 0;
1394 #endif
1395 } else {
1396 u_int32_t flowinfo;
1397
1398 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1399 flowinfo >>= 20;
1400 tclass = flowinfo & 0xff;
1401 }
1402 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS,
1403 IPPROTO_IPV6, M_NOWAIT);
1404 if (*mp)
1405 mp = &(*mp)->m_next;
1406 }
1407
1408 if (v4only != NULL) {
1409 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1410 *v4only = 1;
1411 } else {
1412 *v4only = 0;
1413 }
1414 }
1415
1416 return (mp);
1417 }
1418
1419 void
ip6_savecontrol(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp)1420 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp)
1421 {
1422 struct ip6_hdr *ip6;
1423 int v4only = 0;
1424
1425 mp = ip6_savecontrol_v4(inp, m, mp, &v4only);
1426 if (v4only)
1427 return;
1428
1429 ip6 = mtod(m, struct ip6_hdr *);
1430 /*
1431 * IPV6_HOPOPTS socket option. Recall that we required super-user
1432 * privilege for the option (see ip6_ctloutput), but it might be too
1433 * strict, since there might be some hop-by-hop options which can be
1434 * returned to normal user.
1435 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1436 */
1437 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) {
1438 /*
1439 * Check if a hop-by-hop options header is contatined in the
1440 * received packet, and if so, store the options as ancillary
1441 * data. Note that a hop-by-hop options header must be
1442 * just after the IPv6 header, which is assured through the
1443 * IPv6 input processing.
1444 */
1445 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1446 struct ip6_hbh *hbh;
1447 u_int hbhlen;
1448
1449 hbh = (struct ip6_hbh *)(ip6 + 1);
1450 hbhlen = (hbh->ip6h_len + 1) << 3;
1451
1452 /*
1453 * XXX: We copy the whole header even if a
1454 * jumbo payload option is included, the option which
1455 * is to be removed before returning according to
1456 * RFC2292.
1457 * Note: this constraint is removed in RFC3542
1458 */
1459 *mp = sbcreatecontrol(hbh, hbhlen,
1460 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1461 IPPROTO_IPV6, M_NOWAIT);
1462 if (*mp)
1463 mp = &(*mp)->m_next;
1464 }
1465 }
1466
1467 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1468 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1469
1470 /*
1471 * Search for destination options headers or routing
1472 * header(s) through the header chain, and stores each
1473 * header as ancillary data.
1474 * Note that the order of the headers remains in
1475 * the chain of ancillary data.
1476 */
1477 while (1) { /* is explicit loop prevention necessary? */
1478 struct ip6_ext *ip6e = NULL;
1479 u_int elen;
1480
1481 /*
1482 * if it is not an extension header, don't try to
1483 * pull it from the chain.
1484 */
1485 switch (nxt) {
1486 case IPPROTO_DSTOPTS:
1487 case IPPROTO_ROUTING:
1488 case IPPROTO_HOPOPTS:
1489 case IPPROTO_AH: /* is it possible? */
1490 break;
1491 default:
1492 goto loopend;
1493 }
1494
1495 if (off + sizeof(*ip6e) > m->m_len)
1496 goto loopend;
1497 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1498 if (nxt == IPPROTO_AH)
1499 elen = (ip6e->ip6e_len + 2) << 2;
1500 else
1501 elen = (ip6e->ip6e_len + 1) << 3;
1502 if (off + elen > m->m_len)
1503 goto loopend;
1504
1505 switch (nxt) {
1506 case IPPROTO_DSTOPTS:
1507 if (!(inp->inp_flags & IN6P_DSTOPTS))
1508 break;
1509
1510 *mp = sbcreatecontrol(ip6e, elen,
1511 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1512 IPPROTO_IPV6, M_NOWAIT);
1513 if (*mp)
1514 mp = &(*mp)->m_next;
1515 break;
1516 case IPPROTO_ROUTING:
1517 if (!(inp->inp_flags & IN6P_RTHDR))
1518 break;
1519
1520 *mp = sbcreatecontrol(ip6e, elen,
1521 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR),
1522 IPPROTO_IPV6, M_NOWAIT);
1523 if (*mp)
1524 mp = &(*mp)->m_next;
1525 break;
1526 case IPPROTO_HOPOPTS:
1527 case IPPROTO_AH: /* is it possible? */
1528 break;
1529
1530 default:
1531 /*
1532 * other cases have been filtered in the above.
1533 * none will visit this case. here we supply
1534 * the code just in case (nxt overwritten or
1535 * other cases).
1536 */
1537 goto loopend;
1538 }
1539
1540 /* proceed with the next header. */
1541 off += elen;
1542 nxt = ip6e->ip6e_nxt;
1543 ip6e = NULL;
1544 }
1545 loopend:
1546 ;
1547 }
1548
1549 if (inp->inp_flags2 & INP_RECVFLOWID) {
1550 uint32_t flowid, flow_type;
1551
1552 flowid = m->m_pkthdr.flowid;
1553 flow_type = M_HASHTYPE_GET(m);
1554
1555 /*
1556 * XXX should handle the failure of one or the
1557 * other - don't populate both?
1558 */
1559 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID,
1560 IPPROTO_IPV6, M_NOWAIT);
1561 if (*mp)
1562 mp = &(*mp)->m_next;
1563 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1564 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT);
1565 if (*mp)
1566 mp = &(*mp)->m_next;
1567 }
1568
1569 #ifdef RSS
1570 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1571 uint32_t flowid, flow_type;
1572 uint32_t rss_bucketid;
1573
1574 flowid = m->m_pkthdr.flowid;
1575 flow_type = M_HASHTYPE_GET(m);
1576
1577 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1578 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1579 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT);
1580 if (*mp)
1581 mp = &(*mp)->m_next;
1582 }
1583 }
1584 #endif
1585
1586 }
1587 #undef IS2292
1588
1589 void
ip6_notify_pmtu(struct inpcb * inp,struct sockaddr_in6 * dst,u_int32_t mtu)1590 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1591 {
1592 struct socket *so;
1593 struct mbuf *m_mtu;
1594 struct ip6_mtuinfo mtuctl;
1595
1596 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1597 /*
1598 * Notify the error by sending IPV6_PATHMTU ancillary data if
1599 * application wanted to know the MTU value.
1600 * NOTE: we notify disconnected sockets, because some udp
1601 * applications keep sending sockets disconnected.
1602 * NOTE: our implementation doesn't notify connected sockets that has
1603 * foreign address that is different than given destination addresses
1604 * (this is permitted by RFC 3542).
1605 */
1606 if ((inp->inp_flags & IN6P_MTU) == 0 || (
1607 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1608 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1609 return;
1610
1611 mtuctl.ip6m_mtu = mtu;
1612 mtuctl.ip6m_addr = *dst;
1613 if (sa6_recoverscope(&mtuctl.ip6m_addr))
1614 return;
1615
1616 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU,
1617 IPPROTO_IPV6, M_NOWAIT)) == NULL)
1618 return;
1619
1620 so = inp->inp_socket;
1621 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1622 == 0) {
1623 soroverflow(so);
1624 m_freem(m_mtu);
1625 /* XXX: should count statistics */
1626 } else
1627 sorwakeup(so);
1628 }
1629
1630 /*
1631 * Get pointer to the previous header followed by the header
1632 * currently processed.
1633 */
1634 int
ip6_get_prevhdr(const struct mbuf * m,int off)1635 ip6_get_prevhdr(const struct mbuf *m, int off)
1636 {
1637 struct ip6_ext ip6e;
1638 struct ip6_hdr *ip6;
1639 int len, nlen, nxt;
1640
1641 if (off == sizeof(struct ip6_hdr))
1642 return (offsetof(struct ip6_hdr, ip6_nxt));
1643 if (off < sizeof(struct ip6_hdr))
1644 panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1645
1646 ip6 = mtod(m, struct ip6_hdr *);
1647 nxt = ip6->ip6_nxt;
1648 len = sizeof(struct ip6_hdr);
1649 nlen = 0;
1650 while (len < off) {
1651 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1652 switch (nxt) {
1653 case IPPROTO_FRAGMENT:
1654 nlen = sizeof(struct ip6_frag);
1655 break;
1656 case IPPROTO_AH:
1657 nlen = (ip6e.ip6e_len + 2) << 2;
1658 break;
1659 default:
1660 nlen = (ip6e.ip6e_len + 1) << 3;
1661 }
1662 len += nlen;
1663 nxt = ip6e.ip6e_nxt;
1664 }
1665 return (len - nlen);
1666 }
1667
1668 /*
1669 * get next header offset. m will be retained.
1670 */
1671 int
ip6_nexthdr(const struct mbuf * m,int off,int proto,int * nxtp)1672 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1673 {
1674 struct ip6_hdr ip6;
1675 struct ip6_ext ip6e;
1676 struct ip6_frag fh;
1677
1678 /* just in case */
1679 if (m == NULL)
1680 panic("ip6_nexthdr: m == NULL");
1681 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1682 return -1;
1683
1684 switch (proto) {
1685 case IPPROTO_IPV6:
1686 if (m->m_pkthdr.len < off + sizeof(ip6))
1687 return -1;
1688 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1689 if (nxtp)
1690 *nxtp = ip6.ip6_nxt;
1691 off += sizeof(ip6);
1692 return off;
1693
1694 case IPPROTO_FRAGMENT:
1695 /*
1696 * terminate parsing if it is not the first fragment,
1697 * it does not make sense to parse through it.
1698 */
1699 if (m->m_pkthdr.len < off + sizeof(fh))
1700 return -1;
1701 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1702 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1703 if (fh.ip6f_offlg & IP6F_OFF_MASK)
1704 return -1;
1705 if (nxtp)
1706 *nxtp = fh.ip6f_nxt;
1707 off += sizeof(struct ip6_frag);
1708 return off;
1709
1710 case IPPROTO_AH:
1711 if (m->m_pkthdr.len < off + sizeof(ip6e))
1712 return -1;
1713 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1714 if (nxtp)
1715 *nxtp = ip6e.ip6e_nxt;
1716 off += (ip6e.ip6e_len + 2) << 2;
1717 return off;
1718
1719 case IPPROTO_HOPOPTS:
1720 case IPPROTO_ROUTING:
1721 case IPPROTO_DSTOPTS:
1722 if (m->m_pkthdr.len < off + sizeof(ip6e))
1723 return -1;
1724 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1725 if (nxtp)
1726 *nxtp = ip6e.ip6e_nxt;
1727 off += (ip6e.ip6e_len + 1) << 3;
1728 return off;
1729
1730 case IPPROTO_NONE:
1731 case IPPROTO_ESP:
1732 case IPPROTO_IPCOMP:
1733 /* give up */
1734 return -1;
1735
1736 default:
1737 return -1;
1738 }
1739
1740 /* NOTREACHED */
1741 }
1742
1743 /*
1744 * get offset for the last header in the chain. m will be kept untainted.
1745 */
1746 int
ip6_lasthdr(const struct mbuf * m,int off,int proto,int * nxtp)1747 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1748 {
1749 int newoff;
1750 int nxt;
1751
1752 if (!nxtp) {
1753 nxt = -1;
1754 nxtp = &nxt;
1755 }
1756 while (1) {
1757 newoff = ip6_nexthdr(m, off, proto, nxtp);
1758 if (newoff < 0)
1759 return off;
1760 else if (newoff < off)
1761 return -1; /* invalid */
1762 else if (newoff == off)
1763 return newoff;
1764
1765 off = newoff;
1766 proto = *nxtp;
1767 }
1768 }
1769