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 #include <netinet6/ip6_mroute.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 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, temp_max_desync_factor,
177 CTLFLAG_RD | CTLFLAG_VNET,
178 &VNET_NAME(ip6_temp_max_desync_factor), 0,
179 "RFC 8981 max desync factor");
180
181 #ifdef RSS
182 static struct netisr_handler ip6_direct_nh = {
183 .nh_name = "ip6_direct",
184 .nh_handler = ip6_direct_input,
185 .nh_proto = NETISR_IPV6_DIRECT,
186 .nh_m2cpuid = rss_soft_m2cpuid_v6,
187 .nh_policy = NETISR_POLICY_CPU,
188 .nh_dispatch = NETISR_DISPATCH_HYBRID,
189 };
190
191 static int
sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)192 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
193 {
194 int error, qlimit;
195
196 netisr_getqlimit(&ip6_direct_nh, &qlimit);
197 error = sysctl_handle_int(oidp, &qlimit, 0, req);
198 if (error || !req->newptr)
199 return (error);
200 if (qlimit < 1)
201 return (EINVAL);
202 return (netisr_setqlimit(&ip6_direct_nh, qlimit));
203 }
204 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
205 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
206 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I",
207 "Maximum size of the IPv6 direct input queue");
208
209 #endif
210
211 VNET_DEFINE(pfil_head_t, inet6_pfil_head);
212 VNET_DEFINE(pfil_head_t, inet6_local_pfil_head);
213
214 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
215 VNET_PCPUSTAT_SYSINIT(ip6stat);
216 #ifdef VIMAGE
217 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
218 #endif /* VIMAGE */
219
220 struct rmlock in6_ifaddr_lock;
221 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
222
223 static int ip6_hopopts_input(u_int32_t *, struct mbuf **, int *);
224
225 /*
226 * IP6 initialization: fill in IP6 protocol switch table.
227 * All protocols not implemented in kernel go to raw IP6 protocol handler.
228 */
229 static void
ip6_vnet_init(void * arg __unused)230 ip6_vnet_init(void *arg __unused)
231 {
232 struct pfil_head_args args;
233
234 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
235 &V_ip6_auto_linklocal);
236 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
237 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
238 TUNABLE_BOOL_FETCH("net.inet6.ip6.use_stableaddr", &V_ip6_use_stableaddr);
239
240 CK_STAILQ_INIT(&V_in6_ifaddrhead);
241 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
242 &V_in6_ifaddrhmask);
243
244 /* Initialize packet filter hooks. */
245 args.pa_version = PFIL_VERSION;
246 args.pa_flags = PFIL_IN | PFIL_OUT;
247 args.pa_type = PFIL_TYPE_IP6;
248 args.pa_headname = PFIL_INET6_NAME;
249 V_inet6_pfil_head = pfil_head_register(&args);
250
251 args.pa_flags = PFIL_OUT;
252 args.pa_headname = PFIL_INET6_LOCAL_NAME;
253 V_inet6_local_pfil_head = pfil_head_register(&args);
254
255 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
256 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
257 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
258 printf("%s: WARNING: unable to register input helper hook\n",
259 __func__);
260 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
261 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
262 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
263 printf("%s: WARNING: unable to register output helper hook\n",
264 __func__);
265
266 scope6_init();
267 addrsel_policy_init();
268 nd6_init();
269 frag6_init();
270
271 V_ip6_temp_max_desync_factor = TEMP_MAX_DESYNC_FACTOR_BASE +
272 (V_ip6_temp_preferred_lifetime >> 2) +
273 (V_ip6_temp_preferred_lifetime >> 3);
274 V_ip6_desync_factor = arc4random() % V_ip6_temp_max_desync_factor;
275
276 /* Skip global initialization stuff for non-default instances. */
277 #ifdef VIMAGE
278 netisr_register_vnet(&ip6_nh);
279 #ifdef RSS
280 netisr_register_vnet(&ip6_direct_nh);
281 #endif
282 #endif
283 }
284 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
285 ip6_vnet_init, NULL);
286
287 static void
ip6_init(void * arg __unused)288 ip6_init(void *arg __unused)
289 {
290 struct ifnet *ifp;
291
292 /*
293 * Register statically those protocols that are unlikely to ever go
294 * dynamic.
295 */
296 IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput);
297 IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL);
298 IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL);
299 IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL);
300 IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL);
301 IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL);
302 IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL);
303 IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL);
304 IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL);
305 #ifdef SCTP /* XXX: has a loadable & static version */
306 IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput);
307 #endif
308
309 EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT);
310 EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL,
311 LOWMEM_PRI_DEFAULT);
312
313 netisr_register(&ip6_nh);
314 #ifdef RSS
315 netisr_register(&ip6_direct_nh);
316 #endif
317 /*
318 * XXXGL: we use SYSINIT() here, but go over V_ifnet. See comment
319 * in sys/netinet/ip_input.c:ip_init().
320 */
321 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
322 in6_ifarrival(NULL, ifp);
323 }
324 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL);
325
326 int
ip6proto_register(uint8_t proto,ip6proto_input_t input,ip6proto_ctlinput_t ctl)327 ip6proto_register(uint8_t proto, ip6proto_input_t input,
328 ip6proto_ctlinput_t ctl)
329 {
330
331 MPASS(proto > 0);
332
333 if (ip6_protox[proto] == rip6_input) {
334 ip6_protox[proto] = input;
335 ip6_ctlprotox[proto] = ctl;
336 return (0);
337 } else
338 return (EEXIST);
339 }
340
341 int
ip6proto_unregister(uint8_t proto)342 ip6proto_unregister(uint8_t proto)
343 {
344
345 MPASS(proto > 0);
346
347 if (ip6_protox[proto] != rip6_input) {
348 ip6_protox[proto] = rip6_input;
349 ip6_ctlprotox[proto] = rip6_ctlinput;
350 return (0);
351 } else
352 return (ENOENT);
353 }
354
355 #ifdef VIMAGE
356 static void
ip6_destroy(void * unused __unused)357 ip6_destroy(void *unused __unused)
358 {
359 struct ifaddr *ifa, *nifa;
360 struct ifnet *ifp;
361 int error;
362
363 #ifdef RSS
364 netisr_unregister_vnet(&ip6_direct_nh);
365 #endif
366 netisr_unregister_vnet(&ip6_nh);
367
368 pfil_head_unregister(V_inet6_pfil_head);
369 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
370 if (error != 0) {
371 printf("%s: WARNING: unable to deregister input helper hook "
372 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
373 "error %d returned\n", __func__, error);
374 }
375 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
376 if (error != 0) {
377 printf("%s: WARNING: unable to deregister output helper hook "
378 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
379 "error %d returned\n", __func__, error);
380 }
381
382 /* Cleanup addresses. */
383 IFNET_RLOCK();
384 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
385 /* Cannot lock here - lock recursion. */
386 /* IF_ADDR_LOCK(ifp); */
387 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
388 if (ifa->ifa_addr->sa_family != AF_INET6)
389 continue;
390 in6_purgeaddr(ifa);
391 }
392 /* IF_ADDR_UNLOCK(ifp); */
393 in6_ifdetach_destroy(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 * rtalert,int * off,int * nxt,int * ours)411 ip6_input_hbh(struct mbuf **mp, 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(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 indicates
431 * Hop-by-Hop Options header, then a Jumbo Payload option MUST be
432 * included. We no not support Jumbo Payloads so report an error.
433 */
434 if (ip6->ip6_plen == 0) {
435 IP6STAT_INC(ip6s_badoptions);
436 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
437 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
438 icmp6_error(m, ICMP6_PARAM_PROB,
439 ICMP6_PARAMPROB_HEADER,
440 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
441 goto out;
442 }
443 /* ip6_hopopts_input() ensures that mbuf is contiguous */
444 hbh = (struct ip6_hbh *)(ip6 + 1);
445 *nxt = hbh->ip6h_nxt;
446
447 /*
448 * If we are acting as a router and the packet contains a
449 * router alert option, see if we know the option value.
450 * Currently, we only support the option value for MLD, in which
451 * case we should pass the packet to the multicast routing
452 * daemon.
453 */
454 if (*rtalert != ~0) {
455 switch (*rtalert) {
456 case IP6OPT_RTALERT_MLD:
457 if (V_ip6_forwarding)
458 *ours = 1;
459 break;
460 default:
461 /*
462 * RFC2711 requires unrecognized values must be
463 * silently ignored.
464 */
465 break;
466 }
467 }
468
469 return (0);
470
471 out:
472 return (1);
473 }
474
475 #ifdef RSS
476 /*
477 * IPv6 direct input routine.
478 *
479 * This is called when reinjecting completed fragments where
480 * all of the previous checking and book-keeping has been done.
481 */
482 void
ip6_direct_input(struct mbuf * m)483 ip6_direct_input(struct mbuf *m)
484 {
485 int off, nxt;
486 int nest;
487 struct m_tag *mtag;
488 struct ip6_direct_ctx *ip6dc;
489
490 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
491 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
492
493 ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
494 nxt = ip6dc->ip6dc_nxt;
495 off = ip6dc->ip6dc_off;
496
497 nest = 0;
498
499 m_tag_delete(m, mtag);
500
501 while (nxt != IPPROTO_DONE) {
502 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
503 IP6STAT_INC(ip6s_toomanyhdr);
504 goto bad;
505 }
506
507 /*
508 * protection against faulty packet - there should be
509 * more sanity checks in header chain processing.
510 */
511 if (m->m_pkthdr.len < off) {
512 IP6STAT_INC(ip6s_tooshort);
513 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
514 goto bad;
515 }
516
517 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
518 if (IPSEC_ENABLED(ipv6)) {
519 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
520 return;
521 }
522 #endif /* IPSEC */
523
524 nxt = ip6_protox[nxt](&m, &off, nxt);
525 }
526 return;
527 bad:
528 m_freem(m);
529 }
530 #endif
531
532 void
ip6_input(struct mbuf * m)533 ip6_input(struct mbuf *m)
534 {
535 struct in6_addr odst;
536 struct ip6_hdr *ip6;
537 struct in6_ifaddr *ia;
538 struct ifnet *rcvif;
539 u_int32_t plen;
540 u_int32_t rtalert = ~0;
541 int off = sizeof(struct ip6_hdr), nest;
542 int nxt, ours = 0;
543 int srcrt = 0;
544
545 /*
546 * Drop the packet if IPv6 operation is disabled on the interface.
547 */
548 rcvif = m->m_pkthdr.rcvif;
549 if ((rcvif->if_inet6->nd_flags & ND6_IFF_IFDISABLED))
550 goto bad;
551
552 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
553 /*
554 * should the inner packet be considered authentic?
555 * see comment in ah4_input().
556 * NB: m cannot be NULL when passed to the input routine
557 */
558
559 m->m_flags &= ~M_AUTHIPHDR;
560 m->m_flags &= ~M_AUTHIPDGM;
561
562 #endif /* IPSEC */
563
564 if (m->m_flags & M_FASTFWD_OURS) {
565 /*
566 * Firewall changed destination to local.
567 */
568 ip6 = mtod(m, struct ip6_hdr *);
569 goto passin;
570 }
571
572 /*
573 * mbuf statistics
574 */
575 if (m->m_flags & M_EXT) {
576 if (m->m_next)
577 IP6STAT_INC(ip6s_mext2m);
578 else
579 IP6STAT_INC(ip6s_mext1);
580 } else {
581 if (m->m_next) {
582 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif;
583 int ifindex = ifp->if_index;
584 if (ifindex >= IP6S_M2MMAX)
585 ifindex = 0;
586 IP6STAT_INC2(ip6s_m2m, ifindex);
587 } else
588 IP6STAT_INC(ip6s_m1);
589 }
590
591 in6_ifstat_inc(rcvif, ifs6_in_receive);
592 IP6STAT_INC(ip6s_total);
593
594 /*
595 * L2 bridge code and some other code can return mbuf chain
596 * that does not conform to KAME requirement. too bad.
597 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
598 */
599 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
600 struct mbuf *n;
601
602 if (m->m_pkthdr.len > MHLEN)
603 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
604 else
605 n = m_gethdr(M_NOWAIT, MT_DATA);
606 if (n == NULL)
607 goto bad;
608
609 m_move_pkthdr(n, m);
610 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
611 n->m_len = n->m_pkthdr.len;
612 m_freem(m);
613 m = n;
614 }
615 if (m->m_len < sizeof(struct ip6_hdr)) {
616 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
617 IP6STAT_INC(ip6s_toosmall);
618 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
619 goto bad;
620 }
621 }
622
623 ip6 = mtod(m, struct ip6_hdr *);
624 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
625 IP6STAT_INC(ip6s_badvers);
626 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
627 goto bad;
628 }
629
630 IP6STAT_INC2(ip6s_nxthist, ip6->ip6_nxt);
631 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6);
632
633 /*
634 * Check against address spoofing/corruption. The unspecified address
635 * is checked further below.
636 */
637 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
638 /*
639 * XXX: "badscope" is not very suitable for a multicast source.
640 */
641 IP6STAT_INC(ip6s_badscope);
642 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
643 goto bad;
644 }
645 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
646 !(m->m_flags & M_LOOP)) {
647 /*
648 * In this case, the packet should come from the loopback
649 * interface. However, we cannot just check the if_flags,
650 * because ip6_mloopback() passes the "actual" interface
651 * as the outgoing/incoming interface.
652 */
653 IP6STAT_INC(ip6s_badscope);
654 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
655 goto bad;
656 }
657 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
658 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
659 /*
660 * RFC4291 2.7:
661 * Nodes must not originate a packet to a multicast address
662 * whose scop field contains the reserved value 0; if such
663 * a packet is received, it must be silently dropped.
664 */
665 IP6STAT_INC(ip6s_badscope);
666 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
667 goto bad;
668 }
669 /*
670 * The following check is not documented in specs. A malicious
671 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
672 * and bypass security checks (act as if it was from 127.0.0.1 by using
673 * IPv6 src ::ffff:127.0.0.1). Be cautious.
674 *
675 * We have supported IPv6-only kernels for a few years and this issue
676 * has not come up. The world seems to move mostly towards not using
677 * v4mapped on the wire, so it makes sense for us to keep rejecting
678 * any such packets.
679 */
680 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
681 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
682 IP6STAT_INC(ip6s_badscope);
683 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
684 goto bad;
685 }
686 /*
687 * Try to forward the packet, but if we fail continue.
688 * ip6_tryforward() does not generate redirects, so fall
689 * through to normal processing if redirects are required.
690 * ip6_tryforward() does inbound and outbound packet firewall
691 * processing. If firewall has decided that destination becomes
692 * our local address, it sets M_FASTFWD_OURS flag. In this
693 * case skip another inbound firewall processing and update
694 * ip6 pointer.
695 */
696 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0
697 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
698 && (!IPSEC_ENABLED(ipv6) ||
699 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0)
700 #endif
701 ) {
702 if ((m = ip6_tryforward(m)) == NULL)
703 return;
704 if (m->m_flags & M_FASTFWD_OURS) {
705 ip6 = mtod(m, struct ip6_hdr *);
706 goto passin;
707 }
708 }
709 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
710 /*
711 * Bypass packet filtering for packets previously handled by IPsec.
712 */
713 if (IPSEC_ENABLED(ipv6) &&
714 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0)
715 goto passin;
716 #endif
717 /*
718 * Run through list of hooks for input packets.
719 *
720 * NB: Beware of the destination address changing
721 * (e.g. by NAT rewriting). When this happens,
722 * tell ip6_forward to do the right thing.
723 */
724
725 /* Jump over all PFIL processing if hooks are not active. */
726 if (!PFIL_HOOKED_IN(V_inet6_pfil_head))
727 goto passin;
728
729 odst = ip6->ip6_dst;
730 if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif,
731 NULL) != PFIL_PASS)
732 return;
733 ip6 = mtod(m, struct ip6_hdr *);
734 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
735 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP &&
736 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
737 /*
738 * Directly ship the packet on. This allows forwarding
739 * packets originally destined to us to some other directly
740 * connected host.
741 */
742 ip6_forward(m, 1);
743 return;
744 }
745
746 passin:
747 /*
748 * The check is deferred to here to give firewalls a chance to block
749 * (and log) such packets. ip6_tryforward() will not process such
750 * packets.
751 */
752 if (__predict_false(IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst))) {
753 IP6STAT_INC(ip6s_badscope);
754 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
755 goto bad;
756 }
757
758 plen = (uint32_t)ntohs(ip6->ip6_plen);
759
760 /*
761 * We don't support Jumbograms, reject packets with plen == 0 as early
762 * as we can.
763 */
764 if (plen == 0)
765 goto bad;
766
767 /*
768 * Disambiguate address scope zones (if there is ambiguity).
769 * We first make sure that the original source or destination address
770 * is not in our internal form for scoped addresses. Such addresses
771 * are not necessarily invalid spec-wise, but we cannot accept them due
772 * to the usage conflict.
773 * in6_setscope() then also checks and rejects the cases where src or
774 * dst are the loopback address and the receiving interface
775 * is not loopback.
776 */
777 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
778 IP6STAT_INC(ip6s_badscope); /* XXX */
779 goto bad;
780 }
781 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) ||
782 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) {
783 IP6STAT_INC(ip6s_badscope);
784 goto bad;
785 }
786 if (m->m_flags & M_FASTFWD_OURS) {
787 m->m_flags &= ~M_FASTFWD_OURS;
788 ours = 1;
789 goto hbhcheck;
790 }
791 /*
792 * Multicast check. Assume packet is for us to avoid
793 * prematurely taking locks.
794 */
795 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
796 ours = 1;
797 in6_ifstat_inc(rcvif, ifs6_in_mcast);
798 goto hbhcheck;
799 }
800 /*
801 * Unicast check
802 * XXX: For now we keep link-local IPv6 addresses with embedded
803 * scope zone id, therefore we use zero zoneid here.
804 */
805 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
806 if (ia != NULL) {
807 if (ia->ia6_flags & IN6_IFF_NOTREADY) {
808 char ip6bufs[INET6_ADDRSTRLEN];
809 char ip6bufd[INET6_ADDRSTRLEN];
810 /* address is not ready, so discard the packet. */
811 nd6log((LOG_INFO,
812 "ip6_input: packet to an unready address %s->%s\n",
813 ip6_sprintf(ip6bufs, &ip6->ip6_src),
814 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
815 goto bad;
816 }
817 if (V_ip6_sav && !(m->m_flags & M_LOOP) &&
818 __predict_false(in6_localip_fib(&ip6->ip6_src,
819 rcvif->if_fib))) {
820 IP6STAT_INC(ip6s_badscope); /* XXX */
821 goto bad;
822 }
823 /* Count the packet in the ip address stats */
824 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
825 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
826 ours = 1;
827 goto hbhcheck;
828 }
829
830 /*
831 * Now there is no reason to process the packet if it's not our own
832 * and we're not a router.
833 */
834 if (!V_ip6_forwarding) {
835 IP6STAT_INC(ip6s_cantforward);
836 goto bad;
837 }
838
839 hbhcheck:
840 /*
841 * Process Hop-by-Hop options header if it's contained.
842 * m may be modified in ip6_hopopts_input().
843 */
844 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
845 if (ip6_input_hbh(&m, &rtalert, &off, &nxt, &ours) != 0)
846 return;
847 } else
848 nxt = ip6->ip6_nxt;
849
850 /*
851 * Use mbuf flags to propagate Router Alert option to
852 * ICMPv6 layer, as hop-by-hop options have been stripped.
853 */
854 if (rtalert != ~0)
855 m->m_flags |= M_RTALERT_MLD;
856
857 /*
858 * Check that the amount of data in the buffers
859 * is as at least much as the IPv6 header would have us expect.
860 * Trim mbufs if longer than we expect.
861 * Drop packet if shorter than we expect.
862 */
863 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
864 IP6STAT_INC(ip6s_tooshort);
865 in6_ifstat_inc(rcvif, ifs6_in_truncated);
866 goto bad;
867 }
868 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
869 if (m->m_len == m->m_pkthdr.len) {
870 m->m_len = sizeof(struct ip6_hdr) + plen;
871 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
872 } else
873 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
874 }
875
876 /*
877 * Forward if desirable.
878 */
879 if (V_ip6_mrouting_enabled &&
880 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
881 /*
882 * If we are acting as a multicast router, all
883 * incoming multicast packets are passed to the
884 * kernel-level multicast forwarding function.
885 * The packet is returned (relatively) intact; if
886 * ip6_mforward() returns a non-zero value, the packet
887 * must be discarded, else it may be accepted below.
888 *
889 * XXX TODO: Check hlim and multicast scope here to avoid
890 * unnecessarily calling into ip6_mforward().
891 */
892 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) {
893 IP6STAT_INC(ip6s_cantforward);
894 goto bad;
895 }
896 } else if (!ours) {
897 ip6_forward(m, srcrt);
898 return;
899 }
900
901 /*
902 * We are going to ship the packet to the local protocol stack. Call the
903 * filter again for this 'output' action, allowing redirect-like rules
904 * to adjust the source address.
905 */
906 if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) {
907 if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) !=
908 PFIL_PASS)
909 return;
910 ip6 = mtod(m, struct ip6_hdr *);
911 }
912
913 /*
914 * Tell launch routine the next header
915 */
916 IP6STAT_INC(ip6s_delivered);
917 in6_ifstat_inc(rcvif, ifs6_in_deliver);
918 nest = 0;
919
920 while (nxt != IPPROTO_DONE) {
921 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
922 IP6STAT_INC(ip6s_toomanyhdr);
923 goto bad;
924 }
925
926 /*
927 * protection against faulty packet - there should be
928 * more sanity checks in header chain processing.
929 */
930 if (m->m_pkthdr.len < off) {
931 IP6STAT_INC(ip6s_tooshort);
932 in6_ifstat_inc(rcvif, ifs6_in_truncated);
933 goto bad;
934 }
935
936 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
937 if (IPSEC_ENABLED(ipv6)) {
938 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
939 return;
940 }
941 #endif /* IPSEC */
942
943 nxt = ip6_protox[nxt](&m, &off, nxt);
944 }
945 return;
946 bad:
947 in6_ifstat_inc(rcvif, ifs6_in_discard);
948 if (m != NULL)
949 m_freem(m);
950 }
951
952 /*
953 * Hop-by-Hop options header processing. If a valid jumbo payload option is
954 * included report an error.
955 *
956 * rtalertp - XXX: should be stored more smart way
957 */
958 static int
ip6_hopopts_input(u_int32_t * rtalertp,struct mbuf ** mp,int * offp)959 ip6_hopopts_input(u_int32_t *rtalertp, struct mbuf **mp, int *offp)
960 {
961 struct mbuf *m = *mp;
962 int off = *offp, hbhlen;
963 struct ip6_hbh *hbh;
964
965 /* validation of the length of the header */
966 if (m->m_len < off + sizeof(*hbh)) {
967 m = m_pullup(m, off + sizeof(*hbh));
968 if (m == NULL) {
969 IP6STAT_INC(ip6s_exthdrtoolong);
970 *mp = NULL;
971 return (-1);
972 }
973 }
974 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
975 hbhlen = (hbh->ip6h_len + 1) << 3;
976
977 if (m->m_len < off + hbhlen) {
978 m = m_pullup(m, off + hbhlen);
979 if (m == NULL) {
980 IP6STAT_INC(ip6s_exthdrtoolong);
981 *mp = NULL;
982 return (-1);
983 }
984 }
985 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
986 off += hbhlen;
987 hbhlen -= sizeof(struct ip6_hbh);
988 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
989 hbhlen, rtalertp) < 0) {
990 *mp = NULL;
991 return (-1);
992 }
993
994 *offp = off;
995 *mp = m;
996 return (0);
997 }
998
999 /*
1000 * Search header for all Hop-by-hop options and process each option.
1001 * This function is separate from ip6_hopopts_input() in order to
1002 * handle a case where the sending node itself process its hop-by-hop
1003 * options header. In such a case, the function is called from ip6_output().
1004 *
1005 * The function assumes that hbh header is located right after the IPv6 header
1006 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1007 * opthead + hbhlen is located in contiguous memory region.
1008 */
1009 int
ip6_process_hopopts(struct mbuf * m,u_int8_t * opthead,int hbhlen,u_int32_t * rtalertp)1010 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1011 u_int32_t *rtalertp)
1012 {
1013 int optlen = 0;
1014 u_int8_t *opt = opthead;
1015 u_int16_t rtalert_val;
1016 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1017
1018 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1019 switch (*opt) {
1020 case IP6OPT_PAD1:
1021 optlen = 1;
1022 break;
1023 case IP6OPT_PADN:
1024 if (hbhlen < IP6OPT_MINLEN) {
1025 IP6STAT_INC(ip6s_toosmall);
1026 goto bad;
1027 }
1028 optlen = *(opt + 1) + 2;
1029 break;
1030 case IP6OPT_ROUTER_ALERT:
1031 /* XXX may need check for alignment */
1032 if (hbhlen < IP6OPT_RTALERT_LEN) {
1033 IP6STAT_INC(ip6s_toosmall);
1034 goto bad;
1035 }
1036 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1037 /* XXX stat */
1038 icmp6_error(m, ICMP6_PARAM_PROB,
1039 ICMP6_PARAMPROB_HEADER,
1040 erroff + opt + 1 - opthead);
1041 return (-1);
1042 }
1043 optlen = IP6OPT_RTALERT_LEN;
1044 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1045 *rtalertp = ntohs(rtalert_val);
1046 break;
1047 case IP6OPT_JUMBO:
1048 /* We do not support the Jumbo Payload option. */
1049 goto bad;
1050 default: /* unknown option */
1051 if (hbhlen < IP6OPT_MINLEN) {
1052 IP6STAT_INC(ip6s_toosmall);
1053 goto bad;
1054 }
1055 optlen = ip6_unknown_opt(opt, m,
1056 erroff + opt - opthead);
1057 if (optlen == -1)
1058 return (-1);
1059 optlen += 2;
1060 break;
1061 }
1062 }
1063
1064 return (0);
1065
1066 bad:
1067 m_freem(m);
1068 return (-1);
1069 }
1070
1071 /*
1072 * Unknown option processing.
1073 * The third argument `off' is the offset from the IPv6 header to the option,
1074 * which is necessary if the IPv6 header the and option header and IPv6 header
1075 * is not contiguous in order to return an ICMPv6 error.
1076 */
1077 int
ip6_unknown_opt(u_int8_t * optp,struct mbuf * m,int off)1078 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1079 {
1080 struct ip6_hdr *ip6;
1081
1082 switch (IP6OPT_TYPE(*optp)) {
1083 case IP6OPT_TYPE_SKIP: /* ignore the option */
1084 return ((int)*(optp + 1));
1085 case IP6OPT_TYPE_DISCARD: /* silently discard */
1086 m_freem(m);
1087 return (-1);
1088 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1089 IP6STAT_INC(ip6s_badoptions);
1090 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1091 return (-1);
1092 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1093 IP6STAT_INC(ip6s_badoptions);
1094 ip6 = mtod(m, struct ip6_hdr *);
1095 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1096 (m->m_flags & (M_BCAST|M_MCAST)))
1097 m_freem(m);
1098 else
1099 icmp6_error(m, ICMP6_PARAM_PROB,
1100 ICMP6_PARAMPROB_OPTION, off);
1101 return (-1);
1102 }
1103
1104 m_freem(m); /* XXX: NOTREACHED */
1105 return (-1);
1106 }
1107
1108 /*
1109 * Create the "control" list for this pcb.
1110 * These functions will not modify mbuf chain at all.
1111 *
1112 * The routine will be called from upper layer handlers like tcp6_input().
1113 * Thus the routine assumes that the caller (tcp6_input) have already
1114 * called m_pullup() and all the extension headers are located in the
1115 * very first mbuf on the mbuf chain.
1116 *
1117 * ip6_savecontrol_v4 will handle those options that are possible to be
1118 * set on a v4-mapped socket.
1119 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1120 * options and handle the v6-only ones itself.
1121 */
1122 struct mbuf **
ip6_savecontrol_v4(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp,int * v4only)1123 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1124 int *v4only)
1125 {
1126 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1127
1128 #if defined(SO_TIMESTAMP) && defined(SO_BINTIME)
1129 if ((inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) != 0) {
1130 union {
1131 struct timeval tv;
1132 struct bintime bt;
1133 struct timespec ts;
1134 } t;
1135 struct bintime boottimebin, bt1;
1136 struct timespec ts1;
1137 int ts_clock;
1138 bool stamped;
1139
1140 ts_clock = inp->inp_socket->so_ts_clock;
1141 stamped = false;
1142
1143 /*
1144 * Handle BINTIME first. We create the same output options
1145 * for both SO_BINTIME and the case where SO_TIMESTAMP is
1146 * set with the timestamp clock set to SO_TS_BINTIME.
1147 */
1148 if ((inp->inp_socket->so_options & SO_BINTIME) != 0 ||
1149 ts_clock == SO_TS_BINTIME) {
1150 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1151 M_TSTMP)) {
1152 mbuf_tstmp2timespec(m, &ts1);
1153 timespec2bintime(&ts1, &t.bt);
1154 getboottimebin(&boottimebin);
1155 bintime_add(&t.bt, &boottimebin);
1156 } else {
1157 bintime(&t.bt);
1158 }
1159 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME,
1160 SOL_SOCKET, M_NOWAIT);
1161 if (*mp != NULL) {
1162 mp = &(*mp)->m_next;
1163 stamped = true;
1164 }
1165
1166 /*
1167 * Suppress other timestamps if SO_TIMESTAMP is not
1168 * set.
1169 */
1170 if ((inp->inp_socket->so_options & SO_TIMESTAMP) == 0)
1171 ts_clock = SO_TS_BINTIME;
1172 }
1173
1174 switch (ts_clock) {
1175 case SO_TS_REALTIME_MICRO:
1176 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1177 M_TSTMP)) {
1178 mbuf_tstmp2timespec(m, &ts1);
1179 timespec2bintime(&ts1, &bt1);
1180 getboottimebin(&boottimebin);
1181 bintime_add(&bt1, &boottimebin);
1182 bintime2timeval(&bt1, &t.tv);
1183 } else {
1184 microtime(&t.tv);
1185 }
1186 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv),
1187 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT);
1188 if (*mp != NULL) {
1189 mp = &(*mp)->m_next;
1190 stamped = true;
1191 }
1192 break;
1193
1194 case SO_TS_BINTIME:
1195 break;
1196
1197 case SO_TS_REALTIME:
1198 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1199 M_TSTMP)) {
1200 mbuf_tstmp2timespec(m, &t.ts);
1201 getboottimebin(&boottimebin);
1202 bintime2timespec(&boottimebin, &ts1);
1203 timespecadd(&t.ts, &ts1, &t.ts);
1204 } else {
1205 nanotime(&t.ts);
1206 }
1207 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1208 SCM_REALTIME, SOL_SOCKET, M_NOWAIT);
1209 if (*mp != NULL) {
1210 mp = &(*mp)->m_next;
1211 stamped = true;
1212 }
1213 break;
1214
1215 case SO_TS_MONOTONIC:
1216 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1217 M_TSTMP))
1218 mbuf_tstmp2timespec(m, &t.ts);
1219 else
1220 nanouptime(&t.ts);
1221 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1222 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT);
1223 if (*mp != NULL) {
1224 mp = &(*mp)->m_next;
1225 stamped = true;
1226 }
1227 break;
1228
1229 default:
1230 panic("unknown (corrupted) so_ts_clock");
1231 }
1232 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) ==
1233 (M_PKTHDR | M_TSTMP)) {
1234 struct sock_timestamp_info sti;
1235
1236 bzero(&sti, sizeof(sti));
1237 sti.st_info_flags = ST_INFO_HW;
1238 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1239 sti.st_info_flags |= ST_INFO_HW_HPREC;
1240 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1241 SOL_SOCKET, M_NOWAIT);
1242 if (*mp != NULL)
1243 mp = &(*mp)->m_next;
1244 }
1245 }
1246 #endif
1247
1248 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1249 /* RFC 2292 sec. 5 */
1250 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1251 struct in6_pktinfo pi6;
1252
1253 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1254 #ifdef INET
1255 struct ip *ip;
1256
1257 ip = mtod(m, struct ip *);
1258 pi6.ipi6_addr.s6_addr32[0] = 0;
1259 pi6.ipi6_addr.s6_addr32[1] = 0;
1260 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1261 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1262 #else
1263 /* We won't hit this code */
1264 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1265 #endif
1266 } else {
1267 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1268 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1269 }
1270 pi6.ipi6_ifindex =
1271 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1272
1273 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo),
1274 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6,
1275 M_NOWAIT);
1276 if (*mp)
1277 mp = &(*mp)->m_next;
1278 }
1279
1280 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1281 int hlim;
1282
1283 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1284 #ifdef INET
1285 struct ip *ip;
1286
1287 ip = mtod(m, struct ip *);
1288 hlim = ip->ip_ttl;
1289 #else
1290 /* We won't hit this code */
1291 hlim = 0;
1292 #endif
1293 } else {
1294 hlim = ip6->ip6_hlim & 0xff;
1295 }
1296 *mp = sbcreatecontrol(&hlim, sizeof(int),
1297 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1298 IPPROTO_IPV6, M_NOWAIT);
1299 if (*mp)
1300 mp = &(*mp)->m_next;
1301 }
1302
1303 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1304 int tclass;
1305
1306 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1307 #ifdef INET
1308 struct ip *ip;
1309
1310 ip = mtod(m, struct ip *);
1311 tclass = ip->ip_tos;
1312 #else
1313 /* We won't hit this code */
1314 tclass = 0;
1315 #endif
1316 } else {
1317 u_int32_t flowinfo;
1318
1319 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1320 flowinfo >>= 20;
1321 tclass = flowinfo & 0xff;
1322 }
1323 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS,
1324 IPPROTO_IPV6, M_NOWAIT);
1325 if (*mp)
1326 mp = &(*mp)->m_next;
1327 }
1328
1329 if (v4only != NULL) {
1330 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1331 *v4only = 1;
1332 } else {
1333 *v4only = 0;
1334 }
1335 }
1336
1337 return (mp);
1338 }
1339
1340 void
ip6_savecontrol(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp)1341 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp)
1342 {
1343 struct ip6_hdr *ip6;
1344 int v4only = 0;
1345
1346 mp = ip6_savecontrol_v4(inp, m, mp, &v4only);
1347 if (v4only)
1348 return;
1349
1350 ip6 = mtod(m, struct ip6_hdr *);
1351 /*
1352 * IPV6_HOPOPTS socket option. Recall that we required super-user
1353 * privilege for the option (see ip6_ctloutput), but it might be too
1354 * strict, since there might be some hop-by-hop options which can be
1355 * returned to normal user.
1356 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1357 */
1358 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) {
1359 /*
1360 * Check if a hop-by-hop options header is contatined in the
1361 * received packet, and if so, store the options as ancillary
1362 * data. Note that a hop-by-hop options header must be
1363 * just after the IPv6 header, which is assured through the
1364 * IPv6 input processing.
1365 */
1366 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1367 struct ip6_hbh *hbh;
1368 u_int hbhlen;
1369
1370 hbh = (struct ip6_hbh *)(ip6 + 1);
1371 hbhlen = (hbh->ip6h_len + 1) << 3;
1372
1373 /*
1374 * XXX: We copy the whole header even if a
1375 * jumbo payload option is included, the option which
1376 * is to be removed before returning according to
1377 * RFC2292.
1378 * Note: this constraint is removed in RFC3542
1379 */
1380 *mp = sbcreatecontrol(hbh, hbhlen,
1381 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1382 IPPROTO_IPV6, M_NOWAIT);
1383 if (*mp)
1384 mp = &(*mp)->m_next;
1385 }
1386 }
1387
1388 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1389 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1390
1391 /*
1392 * Search for destination options headers or routing
1393 * header(s) through the header chain, and stores each
1394 * header as ancillary data.
1395 * Note that the order of the headers remains in
1396 * the chain of ancillary data.
1397 */
1398 while (1) { /* is explicit loop prevention necessary? */
1399 struct ip6_ext *ip6e = NULL;
1400 u_int elen;
1401
1402 /*
1403 * if it is not an extension header, don't try to
1404 * pull it from the chain.
1405 */
1406 switch (nxt) {
1407 case IPPROTO_DSTOPTS:
1408 case IPPROTO_ROUTING:
1409 case IPPROTO_HOPOPTS:
1410 case IPPROTO_AH: /* is it possible? */
1411 break;
1412 default:
1413 goto loopend;
1414 }
1415
1416 if (off + sizeof(*ip6e) > m->m_len)
1417 goto loopend;
1418 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1419 if (nxt == IPPROTO_AH)
1420 elen = (ip6e->ip6e_len + 2) << 2;
1421 else
1422 elen = (ip6e->ip6e_len + 1) << 3;
1423 if (off + elen > m->m_len)
1424 goto loopend;
1425
1426 switch (nxt) {
1427 case IPPROTO_DSTOPTS:
1428 if (!(inp->inp_flags & IN6P_DSTOPTS))
1429 break;
1430
1431 *mp = sbcreatecontrol(ip6e, elen,
1432 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1433 IPPROTO_IPV6, M_NOWAIT);
1434 if (*mp)
1435 mp = &(*mp)->m_next;
1436 break;
1437 case IPPROTO_ROUTING:
1438 if (!(inp->inp_flags & IN6P_RTHDR))
1439 break;
1440
1441 *mp = sbcreatecontrol(ip6e, elen,
1442 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR),
1443 IPPROTO_IPV6, M_NOWAIT);
1444 if (*mp)
1445 mp = &(*mp)->m_next;
1446 break;
1447 case IPPROTO_HOPOPTS:
1448 case IPPROTO_AH: /* is it possible? */
1449 break;
1450
1451 default:
1452 /*
1453 * other cases have been filtered in the above.
1454 * none will visit this case. here we supply
1455 * the code just in case (nxt overwritten or
1456 * other cases).
1457 */
1458 goto loopend;
1459 }
1460
1461 /* proceed with the next header. */
1462 off += elen;
1463 nxt = ip6e->ip6e_nxt;
1464 ip6e = NULL;
1465 }
1466 loopend:
1467 ;
1468 }
1469
1470 if (inp->inp_flags2 & INP_RECVFLOWID) {
1471 uint32_t flowid, flow_type;
1472
1473 flowid = m->m_pkthdr.flowid;
1474 flow_type = M_HASHTYPE_GET(m);
1475
1476 /*
1477 * XXX should handle the failure of one or the
1478 * other - don't populate both?
1479 */
1480 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID,
1481 IPPROTO_IPV6, M_NOWAIT);
1482 if (*mp)
1483 mp = &(*mp)->m_next;
1484 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1485 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT);
1486 if (*mp)
1487 mp = &(*mp)->m_next;
1488 }
1489
1490 #ifdef RSS
1491 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1492 uint32_t flowid, flow_type;
1493 uint32_t rss_bucketid;
1494
1495 flowid = m->m_pkthdr.flowid;
1496 flow_type = M_HASHTYPE_GET(m);
1497
1498 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1499 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1500 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT);
1501 if (*mp)
1502 mp = &(*mp)->m_next;
1503 }
1504 }
1505 #endif
1506
1507 }
1508 #undef IS2292
1509
1510 void
ip6_notify_pmtu(struct inpcb * inp,struct sockaddr_in6 * dst,u_int32_t mtu)1511 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1512 {
1513 struct socket *so;
1514 struct mbuf *m_mtu;
1515 struct ip6_mtuinfo mtuctl;
1516
1517 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1518 /*
1519 * Notify the error by sending IPV6_PATHMTU ancillary data if
1520 * application wanted to know the MTU value.
1521 * NOTE: we notify disconnected sockets, because some udp
1522 * applications keep sending sockets disconnected.
1523 * NOTE: our implementation doesn't notify connected sockets that has
1524 * foreign address that is different than given destination addresses
1525 * (this is permitted by RFC 3542).
1526 */
1527 if ((inp->inp_flags & IN6P_MTU) == 0 || (
1528 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1529 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1530 return;
1531
1532 mtuctl.ip6m_mtu = mtu;
1533 mtuctl.ip6m_addr = *dst;
1534 if (sa6_recoverscope(&mtuctl.ip6m_addr))
1535 return;
1536
1537 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU,
1538 IPPROTO_IPV6, M_NOWAIT)) == NULL)
1539 return;
1540
1541 so = inp->inp_socket;
1542 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1543 == 0) {
1544 soroverflow(so);
1545 m_freem(m_mtu);
1546 /* XXX: should count statistics */
1547 } else
1548 sorwakeup(so);
1549 }
1550
1551 /*
1552 * Get pointer to the previous header followed by the header
1553 * currently processed.
1554 */
1555 int
ip6_get_prevhdr(const struct mbuf * m,int off)1556 ip6_get_prevhdr(const struct mbuf *m, int off)
1557 {
1558 struct ip6_ext ip6e;
1559 struct ip6_hdr *ip6;
1560 int len, nlen, nxt;
1561
1562 if (off == sizeof(struct ip6_hdr))
1563 return (offsetof(struct ip6_hdr, ip6_nxt));
1564 if (off < sizeof(struct ip6_hdr))
1565 panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1566
1567 ip6 = mtod(m, struct ip6_hdr *);
1568 nxt = ip6->ip6_nxt;
1569 len = sizeof(struct ip6_hdr);
1570 nlen = 0;
1571 while (len < off) {
1572 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1573 switch (nxt) {
1574 case IPPROTO_FRAGMENT:
1575 nlen = sizeof(struct ip6_frag);
1576 break;
1577 case IPPROTO_AH:
1578 nlen = (ip6e.ip6e_len + 2) << 2;
1579 break;
1580 default:
1581 nlen = (ip6e.ip6e_len + 1) << 3;
1582 }
1583 len += nlen;
1584 nxt = ip6e.ip6e_nxt;
1585 }
1586 return (len - nlen);
1587 }
1588
1589 /*
1590 * get next header offset. m will be retained.
1591 */
1592 int
ip6_nexthdr(const struct mbuf * m,int off,int proto,int * nxtp)1593 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1594 {
1595 struct ip6_hdr ip6;
1596 struct ip6_ext ip6e;
1597 struct ip6_frag fh;
1598
1599 /* just in case */
1600 if (m == NULL)
1601 panic("ip6_nexthdr: m == NULL");
1602 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1603 return -1;
1604
1605 switch (proto) {
1606 case IPPROTO_IPV6:
1607 if (m->m_pkthdr.len < off + sizeof(ip6))
1608 return -1;
1609 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1610 if (nxtp)
1611 *nxtp = ip6.ip6_nxt;
1612 off += sizeof(ip6);
1613 return off;
1614
1615 case IPPROTO_FRAGMENT:
1616 /*
1617 * terminate parsing if it is not the first fragment,
1618 * it does not make sense to parse through it.
1619 */
1620 if (m->m_pkthdr.len < off + sizeof(fh))
1621 return -1;
1622 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1623 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1624 if (fh.ip6f_offlg & IP6F_OFF_MASK)
1625 return -1;
1626 if (nxtp)
1627 *nxtp = fh.ip6f_nxt;
1628 off += sizeof(struct ip6_frag);
1629 return off;
1630
1631 case IPPROTO_AH:
1632 if (m->m_pkthdr.len < off + sizeof(ip6e))
1633 return -1;
1634 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1635 if (nxtp)
1636 *nxtp = ip6e.ip6e_nxt;
1637 off += (ip6e.ip6e_len + 2) << 2;
1638 return off;
1639
1640 case IPPROTO_HOPOPTS:
1641 case IPPROTO_ROUTING:
1642 case IPPROTO_DSTOPTS:
1643 if (m->m_pkthdr.len < off + sizeof(ip6e))
1644 return -1;
1645 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1646 if (nxtp)
1647 *nxtp = ip6e.ip6e_nxt;
1648 off += (ip6e.ip6e_len + 1) << 3;
1649 return off;
1650
1651 case IPPROTO_NONE:
1652 case IPPROTO_ESP:
1653 case IPPROTO_IPCOMP:
1654 /* give up */
1655 return -1;
1656
1657 default:
1658 return -1;
1659 }
1660
1661 /* NOTREACHED */
1662 }
1663
1664 /*
1665 * get offset for the last header in the chain. m will be kept untainted.
1666 */
1667 int
ip6_lasthdr(const struct mbuf * m,int off,int proto,int * nxtp)1668 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1669 {
1670 int newoff;
1671 int nxt;
1672
1673 if (!nxtp) {
1674 nxt = -1;
1675 nxtp = &nxt;
1676 }
1677 while (1) {
1678 newoff = ip6_nexthdr(m, off, proto, nxtp);
1679 if (newoff < 0)
1680 return off;
1681 else if (newoff < off)
1682 return -1; /* invalid */
1683 else if (newoff == off)
1684 return newoff;
1685
1686 off = newoff;
1687 proto = *nxtp;
1688 }
1689 }
1690