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 <sys/cdefs.h>
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 #include "opt_route.h"
68 #include "opt_rss.h"
69 #include "opt_sctp.h"
70
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/hhook.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/proc.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/sdt.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/lock.h>
86 #include <sys/rmlock.h>
87 #include <sys/syslog.h>
88 #include <sys/sysctl.h>
89 #include <sys/eventhandler.h>
90
91 #include <net/if.h>
92 #include <net/if_var.h>
93 #include <net/if_types.h>
94 #include <net/if_private.h>
95 #include <net/if_dl.h>
96 #include <net/route.h>
97 #include <net/netisr.h>
98 #include <net/rss_config.h>
99 #include <net/pfil.h>
100 #include <net/vnet.h>
101
102 #include <netinet/in.h>
103 #include <netinet/in_kdtrace.h>
104 #include <netinet/ip_var.h>
105 #include <netinet/in_systm.h>
106 #include <net/if_llatbl.h>
107 #ifdef INET
108 #include <netinet/ip.h>
109 #include <netinet/ip_icmp.h>
110 #endif /* INET */
111 #include <netinet/ip6.h>
112 #include <netinet6/in6_var.h>
113 #include <netinet6/ip6_var.h>
114 #include <netinet/ip_encap.h>
115 #include <netinet/in_pcb.h>
116 #include <netinet/icmp6.h>
117 #include <netinet6/scope6_var.h>
118 #include <netinet6/in6_ifattach.h>
119 #include <netinet6/mld6_var.h>
120 #include <netinet6/nd6.h>
121 #include <netinet6/in6_rss.h>
122 #ifdef SCTP
123 #include <netinet/sctp_pcb.h>
124 #include <netinet6/sctp6_var.h>
125 #endif
126
127 #include <netipsec/ipsec_support.h>
128
129 ip6proto_input_t *ip6_protox[IPPROTO_MAX] = {
130 [0 ... IPPROTO_MAX - 1] = rip6_input };
131 ip6proto_ctlinput_t *ip6_ctlprotox[IPPROTO_MAX] = {
132 [0 ... IPPROTO_MAX - 1] = rip6_ctlinput };
133
134 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
135 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
136 VNET_DEFINE(u_long, in6_ifaddrhmask);
137
138 static struct netisr_handler ip6_nh = {
139 .nh_name = "ip6",
140 .nh_handler = ip6_input,
141 .nh_proto = NETISR_IPV6,
142 #ifdef RSS
143 .nh_m2cpuid = rss_soft_m2cpuid_v6,
144 .nh_policy = NETISR_POLICY_CPU,
145 .nh_dispatch = NETISR_DISPATCH_HYBRID,
146 #else
147 .nh_policy = NETISR_POLICY_FLOW,
148 #endif
149 };
150
151 static int
sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)152 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
153 {
154 int error, qlimit;
155
156 netisr_getqlimit(&ip6_nh, &qlimit);
157 error = sysctl_handle_int(oidp, &qlimit, 0, req);
158 if (error || !req->newptr)
159 return (error);
160 if (qlimit < 1)
161 return (EINVAL);
162 return (netisr_setqlimit(&ip6_nh, qlimit));
163 }
164 SYSCTL_DECL(_net_inet6_ip6);
165 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen,
166 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
167 0, 0, sysctl_netinet6_intr_queue_maxlen, "I",
168 "Maximum size of the IPv6 input queue");
169
170 VNET_DEFINE_STATIC(bool, ip6_sav) = true;
171 #define V_ip6_sav VNET(ip6_sav)
172 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation,
173 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true,
174 "Drop incoming packets with source address that is a local address");
175
176 #ifdef RSS
177 static struct netisr_handler ip6_direct_nh = {
178 .nh_name = "ip6_direct",
179 .nh_handler = ip6_direct_input,
180 .nh_proto = NETISR_IPV6_DIRECT,
181 .nh_m2cpuid = rss_soft_m2cpuid_v6,
182 .nh_policy = NETISR_POLICY_CPU,
183 .nh_dispatch = NETISR_DISPATCH_HYBRID,
184 };
185
186 static int
sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)187 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
188 {
189 int error, qlimit;
190
191 netisr_getqlimit(&ip6_direct_nh, &qlimit);
192 error = sysctl_handle_int(oidp, &qlimit, 0, req);
193 if (error || !req->newptr)
194 return (error);
195 if (qlimit < 1)
196 return (EINVAL);
197 return (netisr_setqlimit(&ip6_direct_nh, qlimit));
198 }
199 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
200 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
201 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I",
202 "Maximum size of the IPv6 direct input queue");
203
204 #endif
205
206 VNET_DEFINE(pfil_head_t, inet6_pfil_head);
207 VNET_DEFINE(pfil_head_t, inet6_local_pfil_head);
208
209 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
210 VNET_PCPUSTAT_SYSINIT(ip6stat);
211 #ifdef VIMAGE
212 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
213 #endif /* VIMAGE */
214
215 struct rmlock in6_ifaddr_lock;
216 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
217
218 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
219
220 /*
221 * IP6 initialization: fill in IP6 protocol switch table.
222 * All protocols not implemented in kernel go to raw IP6 protocol handler.
223 */
224 static void
ip6_vnet_init(void * arg __unused)225 ip6_vnet_init(void *arg __unused)
226 {
227 struct pfil_head_args args;
228
229 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
230 &V_ip6_auto_linklocal);
231 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
232 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
233
234 CK_STAILQ_INIT(&V_in6_ifaddrhead);
235 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
236 &V_in6_ifaddrhmask);
237
238 /* Initialize packet filter hooks. */
239 args.pa_version = PFIL_VERSION;
240 args.pa_flags = PFIL_IN | PFIL_OUT;
241 args.pa_type = PFIL_TYPE_IP6;
242 args.pa_headname = PFIL_INET6_NAME;
243 V_inet6_pfil_head = pfil_head_register(&args);
244
245 args.pa_flags = PFIL_OUT;
246 args.pa_headname = PFIL_INET6_LOCAL_NAME;
247 V_inet6_local_pfil_head = pfil_head_register(&args);
248
249 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
250 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
251 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
252 printf("%s: WARNING: unable to register input helper hook\n",
253 __func__);
254 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
255 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
256 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
257 printf("%s: WARNING: unable to register output helper hook\n",
258 __func__);
259
260 scope6_init();
261 addrsel_policy_init();
262 nd6_init();
263 frag6_init();
264
265 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
266
267 /* Skip global initialization stuff for non-default instances. */
268 #ifdef VIMAGE
269 netisr_register_vnet(&ip6_nh);
270 #ifdef RSS
271 netisr_register_vnet(&ip6_direct_nh);
272 #endif
273 #endif
274 }
275 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
276 ip6_vnet_init, NULL);
277
278 static void
ip6_init(void * arg __unused)279 ip6_init(void *arg __unused)
280 {
281
282 /*
283 * Register statically those protocols that are unlikely to ever go
284 * dynamic.
285 */
286 IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput);
287 IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL);
288 IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL);
289 IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL);
290 IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL);
291 IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL);
292 IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL);
293 IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL);
294 IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL);
295 #ifdef SCTP /* XXX: has a loadable & static version */
296 IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput);
297 #endif
298
299 EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT);
300 EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL,
301 LOWMEM_PRI_DEFAULT);
302
303 netisr_register(&ip6_nh);
304 #ifdef RSS
305 netisr_register(&ip6_direct_nh);
306 #endif
307 }
308 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL);
309
310 int
ip6proto_register(uint8_t proto,ip6proto_input_t input,ip6proto_ctlinput_t ctl)311 ip6proto_register(uint8_t proto, ip6proto_input_t input,
312 ip6proto_ctlinput_t ctl)
313 {
314
315 MPASS(proto > 0);
316
317 if (ip6_protox[proto] == rip6_input) {
318 ip6_protox[proto] = input;
319 ip6_ctlprotox[proto] = ctl;
320 return (0);
321 } else
322 return (EEXIST);
323 }
324
325 int
ip6proto_unregister(uint8_t proto)326 ip6proto_unregister(uint8_t proto)
327 {
328
329 MPASS(proto > 0);
330
331 if (ip6_protox[proto] != rip6_input) {
332 ip6_protox[proto] = rip6_input;
333 ip6_ctlprotox[proto] = rip6_ctlinput;
334 return (0);
335 } else
336 return (ENOENT);
337 }
338
339 #ifdef VIMAGE
340 static void
ip6_destroy(void * unused __unused)341 ip6_destroy(void *unused __unused)
342 {
343 struct ifaddr *ifa, *nifa;
344 struct ifnet *ifp;
345 int error;
346
347 #ifdef RSS
348 netisr_unregister_vnet(&ip6_direct_nh);
349 #endif
350 netisr_unregister_vnet(&ip6_nh);
351
352 pfil_head_unregister(V_inet6_pfil_head);
353 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
354 if (error != 0) {
355 printf("%s: WARNING: unable to deregister input helper hook "
356 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
357 "error %d returned\n", __func__, error);
358 }
359 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
360 if (error != 0) {
361 printf("%s: WARNING: unable to deregister output helper hook "
362 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
363 "error %d returned\n", __func__, error);
364 }
365
366 /* Cleanup addresses. */
367 IFNET_RLOCK();
368 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
369 /* Cannot lock here - lock recursion. */
370 /* IF_ADDR_LOCK(ifp); */
371 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
372 if (ifa->ifa_addr->sa_family != AF_INET6)
373 continue;
374 in6_purgeaddr(ifa);
375 }
376 /* IF_ADDR_UNLOCK(ifp); */
377 in6_ifdetach_destroy(ifp);
378 mld_domifdetach(ifp);
379 }
380 IFNET_RUNLOCK();
381
382 /* Make sure any routes are gone as well. */
383 rib_flush_routes_family(AF_INET6);
384
385 frag6_destroy();
386 nd6_destroy();
387 in6_ifattach_destroy();
388
389 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
390 }
391
392 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL);
393 #endif
394
395 static int
ip6_input_hbh(struct mbuf ** mp,uint32_t * plen,uint32_t * rtalert,int * off,int * nxt,int * ours)396 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off,
397 int *nxt, int *ours)
398 {
399 struct mbuf *m;
400 struct ip6_hdr *ip6;
401 struct ip6_hbh *hbh;
402
403 if (ip6_hopopts_input(plen, rtalert, mp, off)) {
404 #if 0 /*touches NULL pointer*/
405 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard);
406 #endif
407 goto out; /* m have already been freed */
408 }
409
410 /* adjust pointer */
411 m = *mp;
412 ip6 = mtod(m, struct ip6_hdr *);
413
414 /*
415 * if the payload length field is 0 and the next header field
416 * indicates Hop-by-Hop Options header, then a Jumbo Payload
417 * option MUST be included.
418 */
419 if (ip6->ip6_plen == 0 && *plen == 0) {
420 /*
421 * Note that if a valid jumbo payload option is
422 * contained, ip6_hopopts_input() must set a valid
423 * (non-zero) payload length to the variable plen.
424 */
425 IP6STAT_INC(ip6s_badoptions);
426 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
427 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
428 icmp6_error(m, ICMP6_PARAM_PROB,
429 ICMP6_PARAMPROB_HEADER,
430 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
431 goto out;
432 }
433 /* ip6_hopopts_input() ensures that mbuf is contiguous */
434 hbh = (struct ip6_hbh *)(ip6 + 1);
435 *nxt = hbh->ip6h_nxt;
436
437 /*
438 * If we are acting as a router and the packet contains a
439 * router alert option, see if we know the option value.
440 * Currently, we only support the option value for MLD, in which
441 * case we should pass the packet to the multicast routing
442 * daemon.
443 */
444 if (*rtalert != ~0) {
445 switch (*rtalert) {
446 case IP6OPT_RTALERT_MLD:
447 if (V_ip6_forwarding)
448 *ours = 1;
449 break;
450 default:
451 /*
452 * RFC2711 requires unrecognized values must be
453 * silently ignored.
454 */
455 break;
456 }
457 }
458
459 return (0);
460
461 out:
462 return (1);
463 }
464
465 #ifdef RSS
466 /*
467 * IPv6 direct input routine.
468 *
469 * This is called when reinjecting completed fragments where
470 * all of the previous checking and book-keeping has been done.
471 */
472 void
ip6_direct_input(struct mbuf * m)473 ip6_direct_input(struct mbuf *m)
474 {
475 int off, nxt;
476 int nest;
477 struct m_tag *mtag;
478 struct ip6_direct_ctx *ip6dc;
479
480 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
481 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
482
483 ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
484 nxt = ip6dc->ip6dc_nxt;
485 off = ip6dc->ip6dc_off;
486
487 nest = 0;
488
489 m_tag_delete(m, mtag);
490
491 while (nxt != IPPROTO_DONE) {
492 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
493 IP6STAT_INC(ip6s_toomanyhdr);
494 goto bad;
495 }
496
497 /*
498 * protection against faulty packet - there should be
499 * more sanity checks in header chain processing.
500 */
501 if (m->m_pkthdr.len < off) {
502 IP6STAT_INC(ip6s_tooshort);
503 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
504 goto bad;
505 }
506
507 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
508 if (IPSEC_ENABLED(ipv6)) {
509 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
510 return;
511 }
512 #endif /* IPSEC */
513
514 nxt = ip6_protox[nxt](&m, &off, nxt);
515 }
516 return;
517 bad:
518 m_freem(m);
519 }
520 #endif
521
522 void
ip6_input(struct mbuf * m)523 ip6_input(struct mbuf *m)
524 {
525 struct in6_addr odst;
526 struct ip6_hdr *ip6;
527 struct in6_ifaddr *ia;
528 struct ifnet *rcvif;
529 u_int32_t plen;
530 u_int32_t rtalert = ~0;
531 int off = sizeof(struct ip6_hdr), nest;
532 int nxt, ours = 0;
533 int srcrt = 0;
534
535 /*
536 * Drop the packet if IPv6 operation is disabled on the interface.
537 */
538 rcvif = m->m_pkthdr.rcvif;
539 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED))
540 goto bad;
541
542 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
543 /*
544 * should the inner packet be considered authentic?
545 * see comment in ah4_input().
546 * NB: m cannot be NULL when passed to the input routine
547 */
548
549 m->m_flags &= ~M_AUTHIPHDR;
550 m->m_flags &= ~M_AUTHIPDGM;
551
552 #endif /* IPSEC */
553
554 if (m->m_flags & M_FASTFWD_OURS) {
555 /*
556 * Firewall changed destination to local.
557 */
558 ip6 = mtod(m, struct ip6_hdr *);
559 goto passin;
560 }
561
562 /*
563 * mbuf statistics
564 */
565 if (m->m_flags & M_EXT) {
566 if (m->m_next)
567 IP6STAT_INC(ip6s_mext2m);
568 else
569 IP6STAT_INC(ip6s_mext1);
570 } else {
571 if (m->m_next) {
572 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif;
573 int ifindex = ifp->if_index;
574 if (ifindex >= IP6S_M2MMAX)
575 ifindex = 0;
576 IP6STAT_INC2(ip6s_m2m, ifindex);
577 } else
578 IP6STAT_INC(ip6s_m1);
579 }
580
581 in6_ifstat_inc(rcvif, ifs6_in_receive);
582 IP6STAT_INC(ip6s_total);
583
584 /*
585 * L2 bridge code and some other code can return mbuf chain
586 * that does not conform to KAME requirement. too bad.
587 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
588 */
589 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
590 struct mbuf *n;
591
592 if (m->m_pkthdr.len > MHLEN)
593 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
594 else
595 n = m_gethdr(M_NOWAIT, MT_DATA);
596 if (n == NULL)
597 goto bad;
598
599 m_move_pkthdr(n, m);
600 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
601 n->m_len = n->m_pkthdr.len;
602 m_freem(m);
603 m = n;
604 }
605 if (m->m_len < sizeof(struct ip6_hdr)) {
606 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
607 IP6STAT_INC(ip6s_toosmall);
608 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
609 goto bad;
610 }
611 }
612
613 ip6 = mtod(m, struct ip6_hdr *);
614 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
615 IP6STAT_INC(ip6s_badvers);
616 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
617 goto bad;
618 }
619
620 IP6STAT_INC2(ip6s_nxthist, ip6->ip6_nxt);
621 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6);
622
623 /*
624 * Check against address spoofing/corruption.
625 */
626 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
627 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
628 /*
629 * XXX: "badscope" is not very suitable for a multicast source.
630 */
631 IP6STAT_INC(ip6s_badscope);
632 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
633 goto bad;
634 }
635 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
636 !(m->m_flags & M_LOOP)) {
637 /*
638 * In this case, the packet should come from the loopback
639 * interface. However, we cannot just check the if_flags,
640 * because ip6_mloopback() passes the "actual" interface
641 * as the outgoing/incoming interface.
642 */
643 IP6STAT_INC(ip6s_badscope);
644 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
645 goto bad;
646 }
647 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
648 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
649 /*
650 * RFC4291 2.7:
651 * Nodes must not originate a packet to a multicast address
652 * whose scop field contains the reserved value 0; if such
653 * a packet is received, it must be silently dropped.
654 */
655 IP6STAT_INC(ip6s_badscope);
656 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
657 goto bad;
658 }
659 /*
660 * The following check is not documented in specs. A malicious
661 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
662 * and bypass security checks (act as if it was from 127.0.0.1 by using
663 * IPv6 src ::ffff:127.0.0.1). Be cautious.
664 *
665 * We have supported IPv6-only kernels for a few years and this issue
666 * has not come up. The world seems to move mostly towards not using
667 * v4mapped on the wire, so it makes sense for us to keep rejecting
668 * any such packets.
669 */
670 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
671 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
672 IP6STAT_INC(ip6s_badscope);
673 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
674 goto bad;
675 }
676 #if 0
677 /*
678 * Reject packets with IPv4 compatible addresses (auto tunnel).
679 *
680 * The code forbids auto tunnel relay case in RFC1933 (the check is
681 * stronger than RFC1933). We may want to re-enable it if mech-xx
682 * is revised to forbid relaying case.
683 */
684 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
685 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
686 IP6STAT_INC(ip6s_badscope);
687 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
688 goto bad;
689 }
690 #endif
691 /*
692 * Try to forward the packet, but if we fail continue.
693 * ip6_tryforward() does not generate redirects, so fall
694 * through to normal processing if redirects are required.
695 * ip6_tryforward() does inbound and outbound packet firewall
696 * processing. If firewall has decided that destination becomes
697 * our local address, it sets M_FASTFWD_OURS flag. In this
698 * case skip another inbound firewall processing and update
699 * ip6 pointer.
700 */
701 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0
702 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
703 && (!IPSEC_ENABLED(ipv6) ||
704 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0)
705 #endif
706 ) {
707 if ((m = ip6_tryforward(m)) == NULL)
708 return;
709 if (m->m_flags & M_FASTFWD_OURS) {
710 ip6 = mtod(m, struct ip6_hdr *);
711 goto passin;
712 }
713 }
714 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
715 /*
716 * Bypass packet filtering for packets previously handled by IPsec.
717 */
718 if (IPSEC_ENABLED(ipv6) &&
719 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0)
720 goto passin;
721 #endif
722 /*
723 * Run through list of hooks for input packets.
724 *
725 * NB: Beware of the destination address changing
726 * (e.g. by NAT rewriting). When this happens,
727 * tell ip6_forward to do the right thing.
728 */
729
730 /* Jump over all PFIL processing if hooks are not active. */
731 if (!PFIL_HOOKED_IN(V_inet6_pfil_head))
732 goto passin;
733
734 odst = ip6->ip6_dst;
735 if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif,
736 NULL) != PFIL_PASS)
737 return;
738 ip6 = mtod(m, struct ip6_hdr *);
739 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
740 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP &&
741 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
742 /*
743 * Directly ship the packet on. This allows forwarding
744 * packets originally destined to us to some other directly
745 * connected host.
746 */
747 ip6_forward(m, 1);
748 return;
749 }
750
751 passin:
752 /*
753 * Disambiguate address scope zones (if there is ambiguity).
754 * We first make sure that the original source or destination address
755 * is not in our internal form for scoped addresses. Such addresses
756 * are not necessarily invalid spec-wise, but we cannot accept them due
757 * to the usage conflict.
758 * in6_setscope() then also checks and rejects the cases where src or
759 * dst are the loopback address and the receiving interface
760 * is not loopback.
761 */
762 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
763 IP6STAT_INC(ip6s_badscope); /* XXX */
764 goto bad;
765 }
766 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) ||
767 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) {
768 IP6STAT_INC(ip6s_badscope);
769 goto bad;
770 }
771 if (m->m_flags & M_FASTFWD_OURS) {
772 m->m_flags &= ~M_FASTFWD_OURS;
773 ours = 1;
774 goto hbhcheck;
775 }
776 /*
777 * Multicast check. Assume packet is for us to avoid
778 * prematurely taking locks.
779 */
780 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
781 ours = 1;
782 in6_ifstat_inc(rcvif, ifs6_in_mcast);
783 goto hbhcheck;
784 }
785 /*
786 * Unicast check
787 * XXX: For now we keep link-local IPv6 addresses with embedded
788 * scope zone id, therefore we use zero zoneid here.
789 */
790 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
791 if (ia != NULL) {
792 if (ia->ia6_flags & IN6_IFF_NOTREADY) {
793 char ip6bufs[INET6_ADDRSTRLEN];
794 char ip6bufd[INET6_ADDRSTRLEN];
795 /* address is not ready, so discard the packet. */
796 nd6log((LOG_INFO,
797 "ip6_input: packet to an unready address %s->%s\n",
798 ip6_sprintf(ip6bufs, &ip6->ip6_src),
799 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
800 goto bad;
801 }
802 if (V_ip6_sav && !(m->m_flags & M_LOOP) &&
803 __predict_false(in6_localip_fib(&ip6->ip6_src,
804 rcvif->if_fib))) {
805 IP6STAT_INC(ip6s_badscope); /* XXX */
806 goto bad;
807 }
808 /* Count the packet in the ip address stats */
809 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
810 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
811 ours = 1;
812 goto hbhcheck;
813 }
814
815 /*
816 * Now there is no reason to process the packet if it's not our own
817 * and we're not a router.
818 */
819 if (!V_ip6_forwarding) {
820 IP6STAT_INC(ip6s_cantforward);
821 goto bad;
822 }
823
824 hbhcheck:
825 /*
826 * Process Hop-by-Hop options header if it's contained.
827 * m may be modified in ip6_hopopts_input().
828 * If a JumboPayload option is included, plen will also be modified.
829 */
830 plen = (u_int32_t)ntohs(ip6->ip6_plen);
831 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
832 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0)
833 return;
834 } else
835 nxt = ip6->ip6_nxt;
836
837 /*
838 * Use mbuf flags to propagate Router Alert option to
839 * ICMPv6 layer, as hop-by-hop options have been stripped.
840 */
841 if (rtalert != ~0)
842 m->m_flags |= M_RTALERT_MLD;
843
844 /*
845 * Check that the amount of data in the buffers
846 * is as at least much as the IPv6 header would have us expect.
847 * Trim mbufs if longer than we expect.
848 * Drop packet if shorter than we expect.
849 */
850 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
851 IP6STAT_INC(ip6s_tooshort);
852 in6_ifstat_inc(rcvif, ifs6_in_truncated);
853 goto bad;
854 }
855 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
856 if (m->m_len == m->m_pkthdr.len) {
857 m->m_len = sizeof(struct ip6_hdr) + plen;
858 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
859 } else
860 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
861 }
862
863 /*
864 * Forward if desirable.
865 */
866 if (V_ip6_mrouter &&
867 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
868 /*
869 * If we are acting as a multicast router, all
870 * incoming multicast packets are passed to the
871 * kernel-level multicast forwarding function.
872 * The packet is returned (relatively) intact; if
873 * ip6_mforward() returns a non-zero value, the packet
874 * must be discarded, else it may be accepted below.
875 *
876 * XXX TODO: Check hlim and multicast scope here to avoid
877 * unnecessarily calling into ip6_mforward().
878 */
879 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) {
880 IP6STAT_INC(ip6s_cantforward);
881 goto bad;
882 }
883 } else if (!ours) {
884 ip6_forward(m, srcrt);
885 return;
886 }
887
888 /*
889 * We are going to ship the packet to the local protocol stack. Call the
890 * filter again for this 'output' action, allowing redirect-like rules
891 * to adjust the source address.
892 */
893 if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) {
894 if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) !=
895 PFIL_PASS)
896 return;
897 ip6 = mtod(m, struct ip6_hdr *);
898 }
899
900 /*
901 * Tell launch routine the next header
902 */
903 IP6STAT_INC(ip6s_delivered);
904 in6_ifstat_inc(rcvif, ifs6_in_deliver);
905 nest = 0;
906
907 while (nxt != IPPROTO_DONE) {
908 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
909 IP6STAT_INC(ip6s_toomanyhdr);
910 goto bad;
911 }
912
913 /*
914 * protection against faulty packet - there should be
915 * more sanity checks in header chain processing.
916 */
917 if (m->m_pkthdr.len < off) {
918 IP6STAT_INC(ip6s_tooshort);
919 in6_ifstat_inc(rcvif, ifs6_in_truncated);
920 goto bad;
921 }
922
923 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
924 if (IPSEC_ENABLED(ipv6)) {
925 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
926 return;
927 }
928 #endif /* IPSEC */
929
930 nxt = ip6_protox[nxt](&m, &off, nxt);
931 }
932 return;
933 bad:
934 in6_ifstat_inc(rcvif, ifs6_in_discard);
935 if (m != NULL)
936 m_freem(m);
937 }
938
939 /*
940 * Hop-by-Hop options header processing. If a valid jumbo payload option is
941 * included, the real payload length will be stored in plenp.
942 *
943 * rtalertp - XXX: should be stored more smart way
944 */
945 static int
ip6_hopopts_input(u_int32_t * plenp,u_int32_t * rtalertp,struct mbuf ** mp,int * offp)946 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
947 struct mbuf **mp, int *offp)
948 {
949 struct mbuf *m = *mp;
950 int off = *offp, hbhlen;
951 struct ip6_hbh *hbh;
952
953 /* validation of the length of the header */
954 if (m->m_len < off + sizeof(*hbh)) {
955 m = m_pullup(m, off + sizeof(*hbh));
956 if (m == NULL) {
957 IP6STAT_INC(ip6s_exthdrtoolong);
958 *mp = NULL;
959 return (-1);
960 }
961 }
962 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
963 hbhlen = (hbh->ip6h_len + 1) << 3;
964
965 if (m->m_len < off + hbhlen) {
966 m = m_pullup(m, off + hbhlen);
967 if (m == NULL) {
968 IP6STAT_INC(ip6s_exthdrtoolong);
969 *mp = NULL;
970 return (-1);
971 }
972 }
973 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
974 off += hbhlen;
975 hbhlen -= sizeof(struct ip6_hbh);
976 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
977 hbhlen, rtalertp, plenp) < 0) {
978 *mp = NULL;
979 return (-1);
980 }
981
982 *offp = off;
983 *mp = m;
984 return (0);
985 }
986
987 /*
988 * Search header for all Hop-by-hop options and process each option.
989 * This function is separate from ip6_hopopts_input() in order to
990 * handle a case where the sending node itself process its hop-by-hop
991 * options header. In such a case, the function is called from ip6_output().
992 *
993 * The function assumes that hbh header is located right after the IPv6 header
994 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
995 * opthead + hbhlen is located in contiguous memory region.
996 */
997 int
ip6_process_hopopts(struct mbuf * m,u_int8_t * opthead,int hbhlen,u_int32_t * rtalertp,u_int32_t * plenp)998 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
999 u_int32_t *rtalertp, u_int32_t *plenp)
1000 {
1001 struct ip6_hdr *ip6;
1002 int optlen = 0;
1003 u_int8_t *opt = opthead;
1004 u_int16_t rtalert_val;
1005 u_int32_t jumboplen;
1006 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1007
1008 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1009 switch (*opt) {
1010 case IP6OPT_PAD1:
1011 optlen = 1;
1012 break;
1013 case IP6OPT_PADN:
1014 if (hbhlen < IP6OPT_MINLEN) {
1015 IP6STAT_INC(ip6s_toosmall);
1016 goto bad;
1017 }
1018 optlen = *(opt + 1) + 2;
1019 break;
1020 case IP6OPT_ROUTER_ALERT:
1021 /* XXX may need check for alignment */
1022 if (hbhlen < IP6OPT_RTALERT_LEN) {
1023 IP6STAT_INC(ip6s_toosmall);
1024 goto bad;
1025 }
1026 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1027 /* XXX stat */
1028 icmp6_error(m, ICMP6_PARAM_PROB,
1029 ICMP6_PARAMPROB_HEADER,
1030 erroff + opt + 1 - opthead);
1031 return (-1);
1032 }
1033 optlen = IP6OPT_RTALERT_LEN;
1034 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1035 *rtalertp = ntohs(rtalert_val);
1036 break;
1037 case IP6OPT_JUMBO:
1038 /* XXX may need check for alignment */
1039 if (hbhlen < IP6OPT_JUMBO_LEN) {
1040 IP6STAT_INC(ip6s_toosmall);
1041 goto bad;
1042 }
1043 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1044 /* XXX stat */
1045 icmp6_error(m, ICMP6_PARAM_PROB,
1046 ICMP6_PARAMPROB_HEADER,
1047 erroff + opt + 1 - opthead);
1048 return (-1);
1049 }
1050 optlen = IP6OPT_JUMBO_LEN;
1051
1052 /*
1053 * IPv6 packets that have non 0 payload length
1054 * must not contain a jumbo payload option.
1055 */
1056 ip6 = mtod(m, struct ip6_hdr *);
1057 if (ip6->ip6_plen) {
1058 IP6STAT_INC(ip6s_badoptions);
1059 icmp6_error(m, ICMP6_PARAM_PROB,
1060 ICMP6_PARAMPROB_HEADER,
1061 erroff + opt - opthead);
1062 return (-1);
1063 }
1064
1065 /*
1066 * We may see jumbolen in unaligned location, so
1067 * we'd need to perform bcopy().
1068 */
1069 bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1070 jumboplen = (u_int32_t)htonl(jumboplen);
1071
1072 #if 1
1073 /*
1074 * if there are multiple jumbo payload options,
1075 * *plenp will be non-zero and the packet will be
1076 * rejected.
1077 * the behavior may need some debate in ipngwg -
1078 * multiple options does not make sense, however,
1079 * there's no explicit mention in specification.
1080 */
1081 if (*plenp != 0) {
1082 IP6STAT_INC(ip6s_badoptions);
1083 icmp6_error(m, ICMP6_PARAM_PROB,
1084 ICMP6_PARAMPROB_HEADER,
1085 erroff + opt + 2 - opthead);
1086 return (-1);
1087 }
1088 #endif
1089
1090 /*
1091 * jumbo payload length must be larger than 65535.
1092 */
1093 if (jumboplen <= IPV6_MAXPACKET) {
1094 IP6STAT_INC(ip6s_badoptions);
1095 icmp6_error(m, ICMP6_PARAM_PROB,
1096 ICMP6_PARAMPROB_HEADER,
1097 erroff + opt + 2 - opthead);
1098 return (-1);
1099 }
1100 *plenp = jumboplen;
1101
1102 break;
1103 default: /* unknown option */
1104 if (hbhlen < IP6OPT_MINLEN) {
1105 IP6STAT_INC(ip6s_toosmall);
1106 goto bad;
1107 }
1108 optlen = ip6_unknown_opt(opt, m,
1109 erroff + opt - opthead);
1110 if (optlen == -1)
1111 return (-1);
1112 optlen += 2;
1113 break;
1114 }
1115 }
1116
1117 return (0);
1118
1119 bad:
1120 m_freem(m);
1121 return (-1);
1122 }
1123
1124 /*
1125 * Unknown option processing.
1126 * The third argument `off' is the offset from the IPv6 header to the option,
1127 * which is necessary if the IPv6 header the and option header and IPv6 header
1128 * is not contiguous in order to return an ICMPv6 error.
1129 */
1130 int
ip6_unknown_opt(u_int8_t * optp,struct mbuf * m,int off)1131 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1132 {
1133 struct ip6_hdr *ip6;
1134
1135 switch (IP6OPT_TYPE(*optp)) {
1136 case IP6OPT_TYPE_SKIP: /* ignore the option */
1137 return ((int)*(optp + 1));
1138 case IP6OPT_TYPE_DISCARD: /* silently discard */
1139 m_freem(m);
1140 return (-1);
1141 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1142 IP6STAT_INC(ip6s_badoptions);
1143 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1144 return (-1);
1145 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1146 IP6STAT_INC(ip6s_badoptions);
1147 ip6 = mtod(m, struct ip6_hdr *);
1148 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1149 (m->m_flags & (M_BCAST|M_MCAST)))
1150 m_freem(m);
1151 else
1152 icmp6_error(m, ICMP6_PARAM_PROB,
1153 ICMP6_PARAMPROB_OPTION, off);
1154 return (-1);
1155 }
1156
1157 m_freem(m); /* XXX: NOTREACHED */
1158 return (-1);
1159 }
1160
1161 /*
1162 * Create the "control" list for this pcb.
1163 * These functions will not modify mbuf chain at all.
1164 *
1165 * The routine will be called from upper layer handlers like tcp6_input().
1166 * Thus the routine assumes that the caller (tcp6_input) have already
1167 * called m_pullup() and all the extension headers are located in the
1168 * very first mbuf on the mbuf chain.
1169 *
1170 * ip6_savecontrol_v4 will handle those options that are possible to be
1171 * set on a v4-mapped socket.
1172 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1173 * options and handle the v6-only ones itself.
1174 */
1175 struct mbuf **
ip6_savecontrol_v4(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp,int * v4only)1176 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1177 int *v4only)
1178 {
1179 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1180
1181 #ifdef SO_TIMESTAMP
1182 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1183 union {
1184 struct timeval tv;
1185 struct bintime bt;
1186 struct timespec ts;
1187 } t;
1188 struct bintime boottimebin, bt1;
1189 struct timespec ts1;
1190 bool stamped;
1191
1192 stamped = false;
1193 switch (inp->inp_socket->so_ts_clock) {
1194 case SO_TS_REALTIME_MICRO:
1195 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1196 M_TSTMP)) {
1197 mbuf_tstmp2timespec(m, &ts1);
1198 timespec2bintime(&ts1, &bt1);
1199 getboottimebin(&boottimebin);
1200 bintime_add(&bt1, &boottimebin);
1201 bintime2timeval(&bt1, &t.tv);
1202 } else {
1203 microtime(&t.tv);
1204 }
1205 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv),
1206 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT);
1207 if (*mp != NULL) {
1208 mp = &(*mp)->m_next;
1209 stamped = true;
1210 }
1211 break;
1212
1213 case SO_TS_BINTIME:
1214 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1215 M_TSTMP)) {
1216 mbuf_tstmp2timespec(m, &ts1);
1217 timespec2bintime(&ts1, &t.bt);
1218 getboottimebin(&boottimebin);
1219 bintime_add(&t.bt, &boottimebin);
1220 } else {
1221 bintime(&t.bt);
1222 }
1223 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME,
1224 SOL_SOCKET, M_NOWAIT);
1225 if (*mp != NULL) {
1226 mp = &(*mp)->m_next;
1227 stamped = true;
1228 }
1229 break;
1230
1231 case SO_TS_REALTIME:
1232 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1233 M_TSTMP)) {
1234 mbuf_tstmp2timespec(m, &t.ts);
1235 getboottimebin(&boottimebin);
1236 bintime2timespec(&boottimebin, &ts1);
1237 timespecadd(&t.ts, &ts1, &t.ts);
1238 } else {
1239 nanotime(&t.ts);
1240 }
1241 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1242 SCM_REALTIME, SOL_SOCKET, M_NOWAIT);
1243 if (*mp != NULL) {
1244 mp = &(*mp)->m_next;
1245 stamped = true;
1246 }
1247 break;
1248
1249 case SO_TS_MONOTONIC:
1250 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1251 M_TSTMP))
1252 mbuf_tstmp2timespec(m, &t.ts);
1253 else
1254 nanouptime(&t.ts);
1255 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1256 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT);
1257 if (*mp != NULL) {
1258 mp = &(*mp)->m_next;
1259 stamped = true;
1260 }
1261 break;
1262
1263 default:
1264 panic("unknown (corrupted) so_ts_clock");
1265 }
1266 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) ==
1267 (M_PKTHDR | M_TSTMP)) {
1268 struct sock_timestamp_info sti;
1269
1270 bzero(&sti, sizeof(sti));
1271 sti.st_info_flags = ST_INFO_HW;
1272 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1273 sti.st_info_flags |= ST_INFO_HW_HPREC;
1274 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1275 SOL_SOCKET, M_NOWAIT);
1276 if (*mp != NULL)
1277 mp = &(*mp)->m_next;
1278 }
1279 }
1280 #endif
1281
1282 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1283 /* RFC 2292 sec. 5 */
1284 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1285 struct in6_pktinfo pi6;
1286
1287 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1288 #ifdef INET
1289 struct ip *ip;
1290
1291 ip = mtod(m, struct ip *);
1292 pi6.ipi6_addr.s6_addr32[0] = 0;
1293 pi6.ipi6_addr.s6_addr32[1] = 0;
1294 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1295 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1296 #else
1297 /* We won't hit this code */
1298 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1299 #endif
1300 } else {
1301 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1302 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1303 }
1304 pi6.ipi6_ifindex =
1305 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1306
1307 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo),
1308 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6,
1309 M_NOWAIT);
1310 if (*mp)
1311 mp = &(*mp)->m_next;
1312 }
1313
1314 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1315 int hlim;
1316
1317 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1318 #ifdef INET
1319 struct ip *ip;
1320
1321 ip = mtod(m, struct ip *);
1322 hlim = ip->ip_ttl;
1323 #else
1324 /* We won't hit this code */
1325 hlim = 0;
1326 #endif
1327 } else {
1328 hlim = ip6->ip6_hlim & 0xff;
1329 }
1330 *mp = sbcreatecontrol(&hlim, sizeof(int),
1331 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1332 IPPROTO_IPV6, M_NOWAIT);
1333 if (*mp)
1334 mp = &(*mp)->m_next;
1335 }
1336
1337 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1338 int tclass;
1339
1340 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1341 #ifdef INET
1342 struct ip *ip;
1343
1344 ip = mtod(m, struct ip *);
1345 tclass = ip->ip_tos;
1346 #else
1347 /* We won't hit this code */
1348 tclass = 0;
1349 #endif
1350 } else {
1351 u_int32_t flowinfo;
1352
1353 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1354 flowinfo >>= 20;
1355 tclass = flowinfo & 0xff;
1356 }
1357 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS,
1358 IPPROTO_IPV6, M_NOWAIT);
1359 if (*mp)
1360 mp = &(*mp)->m_next;
1361 }
1362
1363 if (v4only != NULL) {
1364 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1365 *v4only = 1;
1366 } else {
1367 *v4only = 0;
1368 }
1369 }
1370
1371 return (mp);
1372 }
1373
1374 void
ip6_savecontrol(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp)1375 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp)
1376 {
1377 struct ip6_hdr *ip6;
1378 int v4only = 0;
1379
1380 mp = ip6_savecontrol_v4(inp, m, mp, &v4only);
1381 if (v4only)
1382 return;
1383
1384 ip6 = mtod(m, struct ip6_hdr *);
1385 /*
1386 * IPV6_HOPOPTS socket option. Recall that we required super-user
1387 * privilege for the option (see ip6_ctloutput), but it might be too
1388 * strict, since there might be some hop-by-hop options which can be
1389 * returned to normal user.
1390 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1391 */
1392 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) {
1393 /*
1394 * Check if a hop-by-hop options header is contatined in the
1395 * received packet, and if so, store the options as ancillary
1396 * data. Note that a hop-by-hop options header must be
1397 * just after the IPv6 header, which is assured through the
1398 * IPv6 input processing.
1399 */
1400 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1401 struct ip6_hbh *hbh;
1402 u_int hbhlen;
1403
1404 hbh = (struct ip6_hbh *)(ip6 + 1);
1405 hbhlen = (hbh->ip6h_len + 1) << 3;
1406
1407 /*
1408 * XXX: We copy the whole header even if a
1409 * jumbo payload option is included, the option which
1410 * is to be removed before returning according to
1411 * RFC2292.
1412 * Note: this constraint is removed in RFC3542
1413 */
1414 *mp = sbcreatecontrol(hbh, hbhlen,
1415 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1416 IPPROTO_IPV6, M_NOWAIT);
1417 if (*mp)
1418 mp = &(*mp)->m_next;
1419 }
1420 }
1421
1422 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1423 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1424
1425 /*
1426 * Search for destination options headers or routing
1427 * header(s) through the header chain, and stores each
1428 * header as ancillary data.
1429 * Note that the order of the headers remains in
1430 * the chain of ancillary data.
1431 */
1432 while (1) { /* is explicit loop prevention necessary? */
1433 struct ip6_ext *ip6e = NULL;
1434 u_int elen;
1435
1436 /*
1437 * if it is not an extension header, don't try to
1438 * pull it from the chain.
1439 */
1440 switch (nxt) {
1441 case IPPROTO_DSTOPTS:
1442 case IPPROTO_ROUTING:
1443 case IPPROTO_HOPOPTS:
1444 case IPPROTO_AH: /* is it possible? */
1445 break;
1446 default:
1447 goto loopend;
1448 }
1449
1450 if (off + sizeof(*ip6e) > m->m_len)
1451 goto loopend;
1452 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1453 if (nxt == IPPROTO_AH)
1454 elen = (ip6e->ip6e_len + 2) << 2;
1455 else
1456 elen = (ip6e->ip6e_len + 1) << 3;
1457 if (off + elen > m->m_len)
1458 goto loopend;
1459
1460 switch (nxt) {
1461 case IPPROTO_DSTOPTS:
1462 if (!(inp->inp_flags & IN6P_DSTOPTS))
1463 break;
1464
1465 *mp = sbcreatecontrol(ip6e, elen,
1466 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1467 IPPROTO_IPV6, M_NOWAIT);
1468 if (*mp)
1469 mp = &(*mp)->m_next;
1470 break;
1471 case IPPROTO_ROUTING:
1472 if (!(inp->inp_flags & IN6P_RTHDR))
1473 break;
1474
1475 *mp = sbcreatecontrol(ip6e, elen,
1476 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR),
1477 IPPROTO_IPV6, M_NOWAIT);
1478 if (*mp)
1479 mp = &(*mp)->m_next;
1480 break;
1481 case IPPROTO_HOPOPTS:
1482 case IPPROTO_AH: /* is it possible? */
1483 break;
1484
1485 default:
1486 /*
1487 * other cases have been filtered in the above.
1488 * none will visit this case. here we supply
1489 * the code just in case (nxt overwritten or
1490 * other cases).
1491 */
1492 goto loopend;
1493 }
1494
1495 /* proceed with the next header. */
1496 off += elen;
1497 nxt = ip6e->ip6e_nxt;
1498 ip6e = NULL;
1499 }
1500 loopend:
1501 ;
1502 }
1503
1504 if (inp->inp_flags2 & INP_RECVFLOWID) {
1505 uint32_t flowid, flow_type;
1506
1507 flowid = m->m_pkthdr.flowid;
1508 flow_type = M_HASHTYPE_GET(m);
1509
1510 /*
1511 * XXX should handle the failure of one or the
1512 * other - don't populate both?
1513 */
1514 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID,
1515 IPPROTO_IPV6, M_NOWAIT);
1516 if (*mp)
1517 mp = &(*mp)->m_next;
1518 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1519 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT);
1520 if (*mp)
1521 mp = &(*mp)->m_next;
1522 }
1523
1524 #ifdef RSS
1525 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1526 uint32_t flowid, flow_type;
1527 uint32_t rss_bucketid;
1528
1529 flowid = m->m_pkthdr.flowid;
1530 flow_type = M_HASHTYPE_GET(m);
1531
1532 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1533 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1534 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT);
1535 if (*mp)
1536 mp = &(*mp)->m_next;
1537 }
1538 }
1539 #endif
1540
1541 }
1542 #undef IS2292
1543
1544 void
ip6_notify_pmtu(struct inpcb * inp,struct sockaddr_in6 * dst,u_int32_t mtu)1545 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1546 {
1547 struct socket *so;
1548 struct mbuf *m_mtu;
1549 struct ip6_mtuinfo mtuctl;
1550
1551 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1552 /*
1553 * Notify the error by sending IPV6_PATHMTU ancillary data if
1554 * application wanted to know the MTU value.
1555 * NOTE: we notify disconnected sockets, because some udp
1556 * applications keep sending sockets disconnected.
1557 * NOTE: our implementation doesn't notify connected sockets that has
1558 * foreign address that is different than given destination addresses
1559 * (this is permitted by RFC 3542).
1560 */
1561 if ((inp->inp_flags & IN6P_MTU) == 0 || (
1562 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1563 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1564 return;
1565
1566 mtuctl.ip6m_mtu = mtu;
1567 mtuctl.ip6m_addr = *dst;
1568 if (sa6_recoverscope(&mtuctl.ip6m_addr))
1569 return;
1570
1571 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU,
1572 IPPROTO_IPV6, M_NOWAIT)) == NULL)
1573 return;
1574
1575 so = inp->inp_socket;
1576 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1577 == 0) {
1578 soroverflow(so);
1579 m_freem(m_mtu);
1580 /* XXX: should count statistics */
1581 } else
1582 sorwakeup(so);
1583 }
1584
1585 /*
1586 * Get pointer to the previous header followed by the header
1587 * currently processed.
1588 */
1589 int
ip6_get_prevhdr(const struct mbuf * m,int off)1590 ip6_get_prevhdr(const struct mbuf *m, int off)
1591 {
1592 struct ip6_ext ip6e;
1593 struct ip6_hdr *ip6;
1594 int len, nlen, nxt;
1595
1596 if (off == sizeof(struct ip6_hdr))
1597 return (offsetof(struct ip6_hdr, ip6_nxt));
1598 if (off < sizeof(struct ip6_hdr))
1599 panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1600
1601 ip6 = mtod(m, struct ip6_hdr *);
1602 nxt = ip6->ip6_nxt;
1603 len = sizeof(struct ip6_hdr);
1604 nlen = 0;
1605 while (len < off) {
1606 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1607 switch (nxt) {
1608 case IPPROTO_FRAGMENT:
1609 nlen = sizeof(struct ip6_frag);
1610 break;
1611 case IPPROTO_AH:
1612 nlen = (ip6e.ip6e_len + 2) << 2;
1613 break;
1614 default:
1615 nlen = (ip6e.ip6e_len + 1) << 3;
1616 }
1617 len += nlen;
1618 nxt = ip6e.ip6e_nxt;
1619 }
1620 return (len - nlen);
1621 }
1622
1623 /*
1624 * get next header offset. m will be retained.
1625 */
1626 int
ip6_nexthdr(const struct mbuf * m,int off,int proto,int * nxtp)1627 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1628 {
1629 struct ip6_hdr ip6;
1630 struct ip6_ext ip6e;
1631 struct ip6_frag fh;
1632
1633 /* just in case */
1634 if (m == NULL)
1635 panic("ip6_nexthdr: m == NULL");
1636 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1637 return -1;
1638
1639 switch (proto) {
1640 case IPPROTO_IPV6:
1641 if (m->m_pkthdr.len < off + sizeof(ip6))
1642 return -1;
1643 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1644 if (nxtp)
1645 *nxtp = ip6.ip6_nxt;
1646 off += sizeof(ip6);
1647 return off;
1648
1649 case IPPROTO_FRAGMENT:
1650 /*
1651 * terminate parsing if it is not the first fragment,
1652 * it does not make sense to parse through it.
1653 */
1654 if (m->m_pkthdr.len < off + sizeof(fh))
1655 return -1;
1656 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1657 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1658 if (fh.ip6f_offlg & IP6F_OFF_MASK)
1659 return -1;
1660 if (nxtp)
1661 *nxtp = fh.ip6f_nxt;
1662 off += sizeof(struct ip6_frag);
1663 return off;
1664
1665 case IPPROTO_AH:
1666 if (m->m_pkthdr.len < off + sizeof(ip6e))
1667 return -1;
1668 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1669 if (nxtp)
1670 *nxtp = ip6e.ip6e_nxt;
1671 off += (ip6e.ip6e_len + 2) << 2;
1672 return off;
1673
1674 case IPPROTO_HOPOPTS:
1675 case IPPROTO_ROUTING:
1676 case IPPROTO_DSTOPTS:
1677 if (m->m_pkthdr.len < off + sizeof(ip6e))
1678 return -1;
1679 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1680 if (nxtp)
1681 *nxtp = ip6e.ip6e_nxt;
1682 off += (ip6e.ip6e_len + 1) << 3;
1683 return off;
1684
1685 case IPPROTO_NONE:
1686 case IPPROTO_ESP:
1687 case IPPROTO_IPCOMP:
1688 /* give up */
1689 return -1;
1690
1691 default:
1692 return -1;
1693 }
1694
1695 /* NOTREACHED */
1696 }
1697
1698 /*
1699 * get offset for the last header in the chain. m will be kept untainted.
1700 */
1701 int
ip6_lasthdr(const struct mbuf * m,int off,int proto,int * nxtp)1702 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1703 {
1704 int newoff;
1705 int nxt;
1706
1707 if (!nxtp) {
1708 nxt = -1;
1709 nxtp = &nxt;
1710 }
1711 while (1) {
1712 newoff = ip6_nexthdr(m, off, proto, nxtp);
1713 if (newoff < 0)
1714 return off;
1715 else if (newoff < off)
1716 return -1; /* invalid */
1717 else if (newoff == off)
1718 return newoff;
1719
1720 off = newoff;
1721 proto = *nxtp;
1722 }
1723 }
1724