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_rss.h"
67 #include "opt_sctp.h"
68
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/hhook.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/proc.h>
75 #include <sys/domain.h>
76 #include <sys/protosw.h>
77 #include <sys/sdt.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/errno.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/lock.h>
84 #include <sys/rmlock.h>
85 #include <sys/syslog.h>
86 #include <sys/sysctl.h>
87 #include <sys/eventhandler.h>
88
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/if_types.h>
92 #include <net/if_private.h>
93 #include <net/if_dl.h>
94 #include <net/route.h>
95 #include <net/netisr.h>
96 #include <net/rss_config.h>
97 #include <net/pfil.h>
98 #include <net/vnet.h>
99
100 #include <netinet/in.h>
101 #include <netinet/in_kdtrace.h>
102 #include <netinet/ip_var.h>
103 #include <netinet/in_systm.h>
104 #include <net/if_llatbl.h>
105 #ifdef INET
106 #include <netinet/ip.h>
107 #include <netinet/ip_icmp.h>
108 #endif /* INET */
109 #include <netinet/ip6.h>
110 #include <netinet6/in6_var.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet/ip_encap.h>
113 #include <netinet/in_pcb.h>
114 #include <netinet/icmp6.h>
115 #include <netinet6/scope6_var.h>
116 #include <netinet6/in6_ifattach.h>
117 #include <netinet6/mld6_var.h>
118 #include <netinet6/nd6.h>
119 #include <netinet6/in6_rss.h>
120 #include <netinet6/ip6_mroute.h>
121 #ifdef SCTP
122 #include <netinet/sctp_pcb.h>
123 #include <netinet6/sctp6_var.h>
124 #endif
125
126 #include <netipsec/ipsec_support.h>
127
128 ip6proto_input_t *ip6_protox[IPPROTO_MAX] = {
129 [0 ... IPPROTO_MAX - 1] = rip6_input };
130 ip6proto_ctlinput_t *ip6_ctlprotox[IPPROTO_MAX] = {
131 [0 ... IPPROTO_MAX - 1] = rip6_ctlinput };
132
133 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
134 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
135 VNET_DEFINE(u_long, in6_ifaddrhmask);
136
137 static struct netisr_handler ip6_nh = {
138 .nh_name = "ip6",
139 .nh_handler = ip6_input,
140 .nh_proto = NETISR_IPV6,
141 #ifdef RSS
142 .nh_m2cpuid = rss_soft_m2cpuid_v6,
143 .nh_policy = NETISR_POLICY_CPU,
144 .nh_dispatch = NETISR_DISPATCH_HYBRID,
145 #else
146 .nh_policy = NETISR_POLICY_FLOW,
147 #endif
148 };
149
150 static int
sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)151 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
152 {
153 int error, qlimit;
154
155 netisr_getqlimit(&ip6_nh, &qlimit);
156 error = sysctl_handle_int(oidp, &qlimit, 0, req);
157 if (error || !req->newptr)
158 return (error);
159 if (qlimit < 1)
160 return (EINVAL);
161 return (netisr_setqlimit(&ip6_nh, qlimit));
162 }
163 SYSCTL_DECL(_net_inet6_ip6);
164 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen,
165 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
166 0, 0, sysctl_netinet6_intr_queue_maxlen, "I",
167 "Maximum size of the IPv6 input queue");
168
169 VNET_DEFINE_STATIC(bool, ip6_sav) = true;
170 #define V_ip6_sav VNET(ip6_sav)
171 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation,
172 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true,
173 "Drop incoming packets with source address that is a local address");
174
175 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, temp_max_desync_factor,
176 CTLFLAG_RD | CTLFLAG_VNET,
177 &VNET_NAME(ip6_temp_max_desync_factor), 0,
178 "RFC 8981 max desync factor");
179
180 #ifdef RSS
181 static struct netisr_handler ip6_direct_nh = {
182 .nh_name = "ip6_direct",
183 .nh_handler = ip6_direct_input,
184 .nh_proto = NETISR_IPV6_DIRECT,
185 .nh_m2cpuid = rss_soft_m2cpuid_v6,
186 .nh_policy = NETISR_POLICY_CPU,
187 .nh_dispatch = NETISR_DISPATCH_HYBRID,
188 };
189
190 static int
sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)191 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
192 {
193 int error, qlimit;
194
195 netisr_getqlimit(&ip6_direct_nh, &qlimit);
196 error = sysctl_handle_int(oidp, &qlimit, 0, req);
197 if (error || !req->newptr)
198 return (error);
199 if (qlimit < 1)
200 return (EINVAL);
201 return (netisr_setqlimit(&ip6_direct_nh, qlimit));
202 }
203 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
204 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
205 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I",
206 "Maximum size of the IPv6 direct input queue");
207
208 #endif
209
210 VNET_DEFINE(pfil_head_t, inet6_pfil_head);
211 VNET_DEFINE(pfil_head_t, inet6_local_pfil_head);
212
213 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
214 VNET_PCPUSTAT_SYSINIT(ip6stat);
215 #ifdef VIMAGE
216 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
217 #endif /* VIMAGE */
218
219 struct rmlock in6_ifaddr_lock;
220 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
221
222 static int ip6_hopopts_input(u_int32_t *, struct mbuf **, int *);
223
224 /*
225 * IP6 initialization: fill in IP6 protocol switch table.
226 * All protocols not implemented in kernel go to raw IP6 protocol handler.
227 */
228 static void
ip6_vnet_init(void * arg __unused)229 ip6_vnet_init(void *arg __unused)
230 {
231 struct pfil_head_args args;
232
233 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
234 &V_ip6_auto_linklocal);
235 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
236 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
237 TUNABLE_BOOL_FETCH("net.inet6.ip6.use_stableaddr", &V_ip6_use_stableaddr);
238
239 CK_STAILQ_INIT(&V_in6_ifaddrhead);
240 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
241 &V_in6_ifaddrhmask);
242
243 /* Initialize packet filter hooks. */
244 args.pa_version = PFIL_VERSION;
245 args.pa_flags = PFIL_IN | PFIL_OUT;
246 args.pa_type = PFIL_TYPE_IP6;
247 args.pa_headname = PFIL_INET6_NAME;
248 V_inet6_pfil_head = pfil_head_register(&args);
249
250 args.pa_flags = PFIL_OUT;
251 args.pa_headname = PFIL_INET6_LOCAL_NAME;
252 V_inet6_local_pfil_head = pfil_head_register(&args);
253
254 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
255 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
256 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
257 printf("%s: WARNING: unable to register input helper hook\n",
258 __func__);
259 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
260 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
261 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
262 printf("%s: WARNING: unable to register output helper hook\n",
263 __func__);
264
265 scope6_init();
266 addrsel_policy_init();
267 nd6_init();
268 frag6_init();
269
270 V_ip6_temp_max_desync_factor = TEMP_MAX_DESYNC_FACTOR_BASE +
271 (V_ip6_temp_preferred_lifetime >> 2) +
272 (V_ip6_temp_preferred_lifetime >> 3);
273 V_ip6_desync_factor = arc4random() % V_ip6_temp_max_desync_factor;
274
275 /* Skip global initialization stuff for non-default instances. */
276 #ifdef VIMAGE
277 netisr_register_vnet(&ip6_nh);
278 #ifdef RSS
279 netisr_register_vnet(&ip6_direct_nh);
280 #endif
281 #endif
282 }
283 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
284 ip6_vnet_init, NULL);
285
286 static void
ip6_init(void * arg __unused)287 ip6_init(void *arg __unused)
288 {
289 struct ifnet *ifp;
290
291 /*
292 * Register statically those protocols that are unlikely to ever go
293 * dynamic.
294 */
295 IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput);
296 IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL);
297 IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL);
298 IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL);
299 IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL);
300 IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL);
301 IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL);
302 IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL);
303 IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL);
304 #ifdef SCTP /* XXX: has a loadable & static version */
305 IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput);
306 #endif
307
308 EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT);
309 EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL,
310 LOWMEM_PRI_DEFAULT);
311
312 netisr_register(&ip6_nh);
313 #ifdef RSS
314 netisr_register(&ip6_direct_nh);
315 #endif
316 /*
317 * XXXGL: we use SYSINIT() here, but go over V_ifnet. See comment
318 * in sys/netinet/ip_input.c:ip_init().
319 */
320 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
321 in6_ifarrival(NULL, ifp);
322 }
323 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL);
324
325 int
ip6proto_register(uint8_t proto,ip6proto_input_t input,ip6proto_ctlinput_t ctl)326 ip6proto_register(uint8_t proto, ip6proto_input_t input,
327 ip6proto_ctlinput_t ctl)
328 {
329
330 MPASS(proto > 0);
331
332 if (ip6_protox[proto] == rip6_input) {
333 ip6_protox[proto] = input;
334 ip6_ctlprotox[proto] = ctl;
335 return (0);
336 } else
337 return (EEXIST);
338 }
339
340 int
ip6proto_unregister(uint8_t proto)341 ip6proto_unregister(uint8_t proto)
342 {
343
344 MPASS(proto > 0);
345
346 if (ip6_protox[proto] != rip6_input) {
347 ip6_protox[proto] = rip6_input;
348 ip6_ctlprotox[proto] = rip6_ctlinput;
349 return (0);
350 } else
351 return (ENOENT);
352 }
353
354 #ifdef VIMAGE
355 static void
ip6_destroy(void * unused __unused)356 ip6_destroy(void *unused __unused)
357 {
358 struct ifaddr *ifa, *nifa;
359 struct ifnet *ifp;
360 int error;
361
362 #ifdef RSS
363 netisr_unregister_vnet(&ip6_direct_nh);
364 #endif
365 netisr_unregister_vnet(&ip6_nh);
366
367 pfil_head_unregister(V_inet6_pfil_head);
368 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
369 if (error != 0) {
370 printf("%s: WARNING: unable to deregister input helper hook "
371 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
372 "error %d returned\n", __func__, error);
373 }
374 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
375 if (error != 0) {
376 printf("%s: WARNING: unable to deregister output helper hook "
377 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
378 "error %d returned\n", __func__, error);
379 }
380
381 /* Cleanup addresses. */
382 IFNET_RLOCK();
383 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
384 /* Cannot lock here - lock recursion. */
385 /* IF_ADDR_LOCK(ifp); */
386 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
387 if (ifa->ifa_addr->sa_family != AF_INET6)
388 continue;
389 in6_purgeaddr(ifa);
390 }
391 /* IF_ADDR_UNLOCK(ifp); */
392 in6_ifdetach_destroy(ifp);
393 }
394 IFNET_RUNLOCK();
395
396 /* Make sure any routes are gone as well. */
397 rib_flush_routes_family(AF_INET6);
398
399 frag6_destroy();
400 nd6_destroy();
401 in6_ifattach_destroy();
402
403 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
404 }
405
406 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL);
407 #endif
408
409 static int
ip6_input_hbh(struct mbuf ** mp,uint32_t * rtalert,int * off,int * nxt,int * ours)410 ip6_input_hbh(struct mbuf **mp, uint32_t *rtalert, int *off,
411 int *nxt, int *ours)
412 {
413 struct mbuf *m;
414 struct ip6_hdr *ip6;
415 struct ip6_hbh *hbh;
416
417 if (ip6_hopopts_input(rtalert, mp, off)) {
418 #if 0 /*touches NULL pointer*/
419 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard);
420 #endif
421 goto out; /* m have already been freed */
422 }
423
424 /* adjust pointer */
425 m = *mp;
426 ip6 = mtod(m, struct ip6_hdr *);
427
428 /*
429 * If the payload length field is 0 and the next header field indicates
430 * Hop-by-Hop Options header, then a Jumbo Payload option MUST be
431 * included. We no not support Jumbo Payloads so report an error.
432 */
433 if (ip6->ip6_plen == 0) {
434 IP6STAT_INC(ip6s_badoptions);
435 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
436 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
437 icmp6_error(m, ICMP6_PARAM_PROB,
438 ICMP6_PARAMPROB_HEADER,
439 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
440 goto out;
441 }
442 /* ip6_hopopts_input() ensures that mbuf is contiguous */
443 hbh = (struct ip6_hbh *)(ip6 + 1);
444 *nxt = hbh->ip6h_nxt;
445
446 /*
447 * If we are acting as a router and the packet contains a
448 * router alert option, see if we know the option value.
449 * Currently, we only support the option value for MLD, in which
450 * case we should pass the packet to the multicast routing
451 * daemon.
452 */
453 if (*rtalert != ~0) {
454 switch (*rtalert) {
455 case IP6OPT_RTALERT_MLD:
456 if (V_ip6_forwarding)
457 *ours = 1;
458 break;
459 default:
460 /*
461 * RFC2711 requires unrecognized values must be
462 * silently ignored.
463 */
464 break;
465 }
466 }
467
468 return (0);
469
470 out:
471 return (1);
472 }
473
474 #ifdef RSS
475 /*
476 * IPv6 direct input routine.
477 *
478 * This is called when reinjecting completed fragments where
479 * all of the previous checking and book-keeping has been done.
480 */
481 void
ip6_direct_input(struct mbuf * m)482 ip6_direct_input(struct mbuf *m)
483 {
484 int off, nxt;
485 int nest;
486 struct m_tag *mtag;
487 struct ip6_direct_ctx *ip6dc;
488
489 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
490 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
491
492 ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
493 nxt = ip6dc->ip6dc_nxt;
494 off = ip6dc->ip6dc_off;
495
496 nest = 0;
497
498 m_tag_delete(m, mtag);
499
500 while (nxt != IPPROTO_DONE) {
501 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
502 IP6STAT_INC(ip6s_toomanyhdr);
503 goto bad;
504 }
505
506 /*
507 * protection against faulty packet - there should be
508 * more sanity checks in header chain processing.
509 */
510 if (m->m_pkthdr.len < off) {
511 IP6STAT_INC(ip6s_tooshort);
512 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
513 goto bad;
514 }
515
516 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
517 if (IPSEC_ENABLED(ipv6)) {
518 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
519 return;
520 }
521 #endif /* IPSEC */
522
523 nxt = ip6_protox[nxt](&m, &off, nxt);
524 }
525 return;
526 bad:
527 m_freem(m);
528 }
529 #endif
530
531 void
ip6_input(struct mbuf * m)532 ip6_input(struct mbuf *m)
533 {
534 struct in6_addr odst;
535 struct ip6_hdr *ip6;
536 struct in6_ifaddr *ia;
537 struct ifnet *rcvif;
538 u_int32_t plen;
539 u_int32_t rtalert = ~0;
540 int off = sizeof(struct ip6_hdr), nest;
541 int nxt, ours = 0;
542 int srcrt = 0;
543
544 /*
545 * Drop the packet if IPv6 operation is disabled on the interface.
546 */
547 rcvif = m->m_pkthdr.rcvif;
548 if ((rcvif->if_inet6->nd_flags & ND6_IFF_IFDISABLED))
549 goto bad;
550
551 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
552 /*
553 * should the inner packet be considered authentic?
554 * see comment in ah4_input().
555 * NB: m cannot be NULL when passed to the input routine
556 */
557
558 m->m_flags &= ~M_AUTHIPHDR;
559 m->m_flags &= ~M_AUTHIPDGM;
560
561 #endif /* IPSEC */
562
563 if (m->m_flags & M_FASTFWD_OURS) {
564 /*
565 * Firewall changed destination to local.
566 */
567 ip6 = mtod(m, struct ip6_hdr *);
568 goto passin;
569 }
570
571 /*
572 * mbuf statistics
573 */
574 if (m->m_flags & M_EXT) {
575 if (m->m_next)
576 IP6STAT_INC(ip6s_mext2m);
577 else
578 IP6STAT_INC(ip6s_mext1);
579 } else {
580 if (m->m_next) {
581 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif;
582 int ifindex = ifp->if_index;
583 if (ifindex >= IP6S_M2MMAX)
584 ifindex = 0;
585 IP6STAT_INC2(ip6s_m2m, ifindex);
586 } else
587 IP6STAT_INC(ip6s_m1);
588 }
589
590 in6_ifstat_inc(rcvif, ifs6_in_receive);
591 IP6STAT_INC(ip6s_total);
592
593 /*
594 * L2 bridge code and some other code can return mbuf chain
595 * that does not conform to KAME requirement. too bad.
596 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
597 */
598 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
599 struct mbuf *n;
600
601 if (m->m_pkthdr.len > MHLEN)
602 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
603 else
604 n = m_gethdr(M_NOWAIT, MT_DATA);
605 if (n == NULL)
606 goto bad;
607
608 m_move_pkthdr(n, m);
609 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
610 n->m_len = n->m_pkthdr.len;
611 m_freem(m);
612 m = n;
613 }
614 if (m->m_len < sizeof(struct ip6_hdr)) {
615 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
616 IP6STAT_INC(ip6s_toosmall);
617 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
618 goto bad;
619 }
620 }
621
622 ip6 = mtod(m, struct ip6_hdr *);
623 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
624 IP6STAT_INC(ip6s_badvers);
625 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
626 goto bad;
627 }
628
629 IP6STAT_INC2(ip6s_nxthist, ip6->ip6_nxt);
630 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6);
631
632 /*
633 * Check against address spoofing/corruption. The unspecified address
634 * is checked further below.
635 */
636 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
637 /*
638 * XXX: "badscope" is not very suitable for a multicast source.
639 */
640 IP6STAT_INC(ip6s_badscope);
641 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
642 goto bad;
643 }
644 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
645 !(m->m_flags & M_LOOP)) {
646 /*
647 * In this case, the packet should come from the loopback
648 * interface. However, we cannot just check the if_flags,
649 * because ip6_mloopback() passes the "actual" interface
650 * as the outgoing/incoming interface.
651 */
652 IP6STAT_INC(ip6s_badscope);
653 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
654 goto bad;
655 }
656 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
657 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
658 /*
659 * RFC4291 2.7:
660 * Nodes must not originate a packet to a multicast address
661 * whose scop field contains the reserved value 0; if such
662 * a packet is received, it must be silently dropped.
663 */
664 IP6STAT_INC(ip6s_badscope);
665 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
666 goto bad;
667 }
668 /*
669 * The following check is not documented in specs. A malicious
670 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
671 * and bypass security checks (act as if it was from 127.0.0.1 by using
672 * IPv6 src ::ffff:127.0.0.1). Be cautious.
673 *
674 * We have supported IPv6-only kernels for a few years and this issue
675 * has not come up. The world seems to move mostly towards not using
676 * v4mapped on the wire, so it makes sense for us to keep rejecting
677 * any such packets.
678 */
679 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
680 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
681 IP6STAT_INC(ip6s_badscope);
682 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
683 goto bad;
684 }
685 /*
686 * Try to forward the packet, but if we fail continue.
687 * ip6_tryforward() does not generate redirects, so fall
688 * through to normal processing if redirects are required.
689 * ip6_tryforward() does inbound and outbound packet firewall
690 * processing. If firewall has decided that destination becomes
691 * our local address, it sets M_FASTFWD_OURS flag. In this
692 * case skip another inbound firewall processing and update
693 * ip6 pointer.
694 */
695 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0
696 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
697 && (!IPSEC_ENABLED(ipv6) ||
698 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0)
699 #endif
700 ) {
701 if ((m = ip6_tryforward(m)) == NULL)
702 return;
703 if (m->m_flags & M_FASTFWD_OURS) {
704 ip6 = mtod(m, struct ip6_hdr *);
705 goto passin;
706 }
707 }
708 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
709 /*
710 * Bypass packet filtering for packets previously handled by IPsec.
711 */
712 if (IPSEC_ENABLED(ipv6) &&
713 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0)
714 goto passin;
715 #endif
716 /*
717 * Run through list of hooks for input packets.
718 *
719 * NB: Beware of the destination address changing
720 * (e.g. by NAT rewriting). When this happens,
721 * tell ip6_forward to do the right thing.
722 */
723
724 /* Jump over all PFIL processing if hooks are not active. */
725 if (!PFIL_HOOKED_IN(V_inet6_pfil_head))
726 goto passin;
727
728 odst = ip6->ip6_dst;
729 if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif,
730 NULL) != PFIL_PASS)
731 return;
732 ip6 = mtod(m, struct ip6_hdr *);
733 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
734 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP &&
735 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
736 /*
737 * Directly ship the packet on. This allows forwarding
738 * packets originally destined to us to some other directly
739 * connected host.
740 */
741 ip6_forward(m, 1);
742 return;
743 }
744
745 passin:
746 /*
747 * The check is deferred to here to give firewalls a chance to block
748 * (and log) such packets. ip6_tryforward() will not process such
749 * packets.
750 */
751 if (__predict_false(IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst))) {
752 IP6STAT_INC(ip6s_badscope);
753 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
754 goto bad;
755 }
756
757 plen = (uint32_t)ntohs(ip6->ip6_plen);
758
759 /*
760 * We don't support Jumbograms, reject packets with plen == 0 as early
761 * as we can.
762 */
763 if (plen == 0)
764 goto bad;
765
766 /*
767 * Disambiguate address scope zones (if there is ambiguity).
768 * We first make sure that the original source or destination address
769 * is not in our internal form for scoped addresses. Such addresses
770 * are not necessarily invalid spec-wise, but we cannot accept them due
771 * to the usage conflict.
772 * in6_setscope() then also checks and rejects the cases where src or
773 * dst are the loopback address and the receiving interface
774 * is not loopback.
775 */
776 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
777 IP6STAT_INC(ip6s_badscope); /* XXX */
778 goto bad;
779 }
780 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) ||
781 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) {
782 IP6STAT_INC(ip6s_badscope);
783 goto bad;
784 }
785 if (m->m_flags & M_FASTFWD_OURS) {
786 m->m_flags &= ~M_FASTFWD_OURS;
787 ours = 1;
788 goto hbhcheck;
789 }
790 /*
791 * Multicast check. Assume packet is for us to avoid
792 * prematurely taking locks.
793 */
794 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
795 ours = 1;
796 in6_ifstat_inc(rcvif, ifs6_in_mcast);
797 goto hbhcheck;
798 }
799 /*
800 * Unicast check
801 * XXX: For now we keep link-local IPv6 addresses with embedded
802 * scope zone id, therefore we use zero zoneid here.
803 */
804 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
805 if (ia != NULL) {
806 if (ia->ia6_flags & IN6_IFF_NOTREADY) {
807 char ip6bufs[INET6_ADDRSTRLEN];
808 char ip6bufd[INET6_ADDRSTRLEN];
809 /* address is not ready, so discard the packet. */
810 nd6log((LOG_INFO,
811 "ip6_input: packet to an unready address %s->%s\n",
812 ip6_sprintf(ip6bufs, &ip6->ip6_src),
813 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
814 goto bad;
815 }
816 if (V_ip6_sav && !(m->m_flags & M_LOOP) &&
817 __predict_false(in6_localip_fib(&ip6->ip6_src,
818 rcvif->if_fib))) {
819 IP6STAT_INC(ip6s_badscope); /* XXX */
820 goto bad;
821 }
822 /* Count the packet in the ip address stats */
823 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
824 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
825 ours = 1;
826 goto hbhcheck;
827 }
828
829 /*
830 * Now there is no reason to process the packet if it's not our own
831 * and we're not a router.
832 */
833 if (!V_ip6_forwarding) {
834 IP6STAT_INC(ip6s_cantforward);
835 goto bad;
836 }
837
838 hbhcheck:
839 /*
840 * Process Hop-by-Hop options header if it's contained.
841 * m may be modified in ip6_hopopts_input().
842 */
843 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
844 if (ip6_input_hbh(&m, &rtalert, &off, &nxt, &ours) != 0)
845 return;
846 } else
847 nxt = ip6->ip6_nxt;
848
849 /*
850 * Use mbuf flags to propagate Router Alert option to
851 * ICMPv6 layer, as hop-by-hop options have been stripped.
852 */
853 if (rtalert != ~0)
854 m->m_flags |= M_RTALERT_MLD;
855
856 /*
857 * Check that the amount of data in the buffers
858 * is as at least much as the IPv6 header would have us expect.
859 * Trim mbufs if longer than we expect.
860 * Drop packet if shorter than we expect.
861 */
862 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
863 IP6STAT_INC(ip6s_tooshort);
864 in6_ifstat_inc(rcvif, ifs6_in_truncated);
865 goto bad;
866 }
867 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
868 if (m->m_len == m->m_pkthdr.len) {
869 m->m_len = sizeof(struct ip6_hdr) + plen;
870 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
871 } else
872 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
873 }
874
875 /*
876 * Forward if desirable.
877 */
878 if (V_ip6_mrouting_enabled &&
879 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
880 /*
881 * If we are acting as a multicast router, all
882 * incoming multicast packets are passed to the
883 * kernel-level multicast forwarding function.
884 * The packet is returned (relatively) intact; if
885 * ip6_mforward() returns a non-zero value, the packet
886 * must be discarded, else it may be accepted below.
887 *
888 * XXX TODO: Check hlim and multicast scope here to avoid
889 * unnecessarily calling into ip6_mforward().
890 */
891 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) {
892 IP6STAT_INC(ip6s_cantforward);
893 goto bad;
894 }
895 } else if (!ours) {
896 ip6_forward(m, srcrt);
897 return;
898 }
899
900 /*
901 * We are going to ship the packet to the local protocol stack. Call the
902 * filter again for this 'output' action, allowing redirect-like rules
903 * to adjust the source address.
904 */
905 if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) {
906 if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) !=
907 PFIL_PASS)
908 return;
909 ip6 = mtod(m, struct ip6_hdr *);
910 }
911
912 /*
913 * Tell launch routine the next header
914 */
915 IP6STAT_INC(ip6s_delivered);
916 in6_ifstat_inc(rcvif, ifs6_in_deliver);
917 nest = 0;
918
919 while (nxt != IPPROTO_DONE) {
920 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
921 IP6STAT_INC(ip6s_toomanyhdr);
922 goto bad;
923 }
924
925 /*
926 * protection against faulty packet - there should be
927 * more sanity checks in header chain processing.
928 */
929 if (m->m_pkthdr.len < off) {
930 IP6STAT_INC(ip6s_tooshort);
931 in6_ifstat_inc(rcvif, ifs6_in_truncated);
932 goto bad;
933 }
934
935 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
936 if (IPSEC_ENABLED(ipv6)) {
937 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
938 return;
939 }
940 #endif /* IPSEC */
941
942 nxt = ip6_protox[nxt](&m, &off, nxt);
943 }
944 return;
945 bad:
946 in6_ifstat_inc(rcvif, ifs6_in_discard);
947 if (m != NULL)
948 m_freem(m);
949 }
950
951 /*
952 * Hop-by-Hop options header processing. If a valid jumbo payload option is
953 * included report an error.
954 *
955 * rtalertp - XXX: should be stored more smart way
956 */
957 static int
ip6_hopopts_input(u_int32_t * rtalertp,struct mbuf ** mp,int * offp)958 ip6_hopopts_input(u_int32_t *rtalertp, struct mbuf **mp, int *offp)
959 {
960 struct mbuf *m = *mp;
961 int off = *offp, hbhlen;
962 struct ip6_hbh *hbh;
963
964 /* validation of the length of the header */
965 if (m->m_len < off + sizeof(*hbh)) {
966 m = m_pullup(m, off + sizeof(*hbh));
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 hbhlen = (hbh->ip6h_len + 1) << 3;
975
976 if (m->m_len < off + hbhlen) {
977 m = m_pullup(m, off + hbhlen);
978 if (m == NULL) {
979 IP6STAT_INC(ip6s_exthdrtoolong);
980 *mp = NULL;
981 return (-1);
982 }
983 }
984 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
985 off += hbhlen;
986 hbhlen -= sizeof(struct ip6_hbh);
987 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
988 hbhlen, rtalertp) < 0) {
989 *mp = NULL;
990 return (-1);
991 }
992
993 *offp = off;
994 *mp = m;
995 return (0);
996 }
997
998 /*
999 * Search header for all Hop-by-hop options and process each option.
1000 * This function is separate from ip6_hopopts_input() in order to
1001 * handle a case where the sending node itself process its hop-by-hop
1002 * options header. In such a case, the function is called from ip6_output().
1003 *
1004 * The function assumes that hbh header is located right after the IPv6 header
1005 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1006 * opthead + hbhlen is located in contiguous memory region.
1007 */
1008 int
ip6_process_hopopts(struct mbuf * m,u_int8_t * opthead,int hbhlen,u_int32_t * rtalertp)1009 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1010 u_int32_t *rtalertp)
1011 {
1012 int optlen = 0;
1013 u_int8_t *opt = opthead;
1014 u_int16_t rtalert_val;
1015 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1016
1017 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1018 switch (*opt) {
1019 case IP6OPT_PAD1:
1020 optlen = 1;
1021 break;
1022 case IP6OPT_PADN:
1023 if (hbhlen < IP6OPT_MINLEN) {
1024 IP6STAT_INC(ip6s_toosmall);
1025 goto bad;
1026 }
1027 optlen = *(opt + 1) + 2;
1028 break;
1029 case IP6OPT_ROUTER_ALERT:
1030 /* XXX may need check for alignment */
1031 if (hbhlen < IP6OPT_RTALERT_LEN) {
1032 IP6STAT_INC(ip6s_toosmall);
1033 goto bad;
1034 }
1035 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1036 /* XXX stat */
1037 icmp6_error(m, ICMP6_PARAM_PROB,
1038 ICMP6_PARAMPROB_HEADER,
1039 erroff + opt + 1 - opthead);
1040 return (-1);
1041 }
1042 optlen = IP6OPT_RTALERT_LEN;
1043 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1044 *rtalertp = ntohs(rtalert_val);
1045 break;
1046 case IP6OPT_JUMBO:
1047 /* We do not support the Jumbo Payload option. */
1048 goto bad;
1049 default: /* unknown option */
1050 if (hbhlen < IP6OPT_MINLEN) {
1051 IP6STAT_INC(ip6s_toosmall);
1052 goto bad;
1053 }
1054 optlen = ip6_unknown_opt(opt, m,
1055 erroff + opt - opthead);
1056 if (optlen == -1)
1057 return (-1);
1058 optlen += 2;
1059 break;
1060 }
1061 }
1062
1063 return (0);
1064
1065 bad:
1066 m_freem(m);
1067 return (-1);
1068 }
1069
1070 /*
1071 * Unknown option processing.
1072 * The third argument `off' is the offset from the IPv6 header to the option,
1073 * which is necessary if the IPv6 header the and option header and IPv6 header
1074 * is not contiguous in order to return an ICMPv6 error.
1075 */
1076 int
ip6_unknown_opt(u_int8_t * optp,struct mbuf * m,int off)1077 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1078 {
1079 struct ip6_hdr *ip6;
1080
1081 switch (IP6OPT_TYPE(*optp)) {
1082 case IP6OPT_TYPE_SKIP: /* ignore the option */
1083 return ((int)*(optp + 1));
1084 case IP6OPT_TYPE_DISCARD: /* silently discard */
1085 m_freem(m);
1086 return (-1);
1087 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1088 IP6STAT_INC(ip6s_badoptions);
1089 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1090 return (-1);
1091 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1092 IP6STAT_INC(ip6s_badoptions);
1093 ip6 = mtod(m, struct ip6_hdr *);
1094 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1095 (m->m_flags & (M_BCAST|M_MCAST)))
1096 m_freem(m);
1097 else
1098 icmp6_error(m, ICMP6_PARAM_PROB,
1099 ICMP6_PARAMPROB_OPTION, off);
1100 return (-1);
1101 }
1102
1103 m_freem(m); /* XXX: NOTREACHED */
1104 return (-1);
1105 }
1106
1107 /*
1108 * Create the "control" list for this pcb.
1109 * These functions will not modify mbuf chain at all.
1110 *
1111 * The routine will be called from upper layer handlers like tcp6_input().
1112 * Thus the routine assumes that the caller (tcp6_input) have already
1113 * called m_pullup() and all the extension headers are located in the
1114 * very first mbuf on the mbuf chain.
1115 *
1116 * ip6_savecontrol_v4 will handle those options that are possible to be
1117 * set on a v4-mapped socket.
1118 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1119 * options and handle the v6-only ones itself.
1120 */
1121 struct mbuf **
ip6_savecontrol_v4(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp,int * v4only)1122 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1123 int *v4only)
1124 {
1125 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1126
1127 #if defined(SO_TIMESTAMP) && defined(SO_BINTIME)
1128 if ((inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) != 0) {
1129 union {
1130 struct timeval tv;
1131 struct bintime bt;
1132 struct timespec ts;
1133 } t;
1134 struct bintime boottimebin, bt1;
1135 struct timespec ts1;
1136 int ts_clock;
1137 bool stamped;
1138
1139 ts_clock = inp->inp_socket->so_ts_clock;
1140 stamped = false;
1141
1142 /*
1143 * Handle BINTIME first. We create the same output options
1144 * for both SO_BINTIME and the case where SO_TIMESTAMP is
1145 * set with the timestamp clock set to SO_TS_BINTIME.
1146 */
1147 if ((inp->inp_socket->so_options & SO_BINTIME) != 0 ||
1148 ts_clock == SO_TS_BINTIME) {
1149 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1150 M_TSTMP)) {
1151 mbuf_tstmp2timespec(m, &ts1);
1152 timespec2bintime(&ts1, &t.bt);
1153 getboottimebin(&boottimebin);
1154 bintime_add(&t.bt, &boottimebin);
1155 } else {
1156 bintime(&t.bt);
1157 }
1158 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME,
1159 SOL_SOCKET, M_NOWAIT);
1160 if (*mp != NULL) {
1161 mp = &(*mp)->m_next;
1162 stamped = true;
1163 }
1164
1165 /*
1166 * Suppress other timestamps if SO_TIMESTAMP is not
1167 * set.
1168 */
1169 if ((inp->inp_socket->so_options & SO_TIMESTAMP) == 0)
1170 ts_clock = SO_TS_BINTIME;
1171 }
1172
1173 switch (ts_clock) {
1174 case SO_TS_REALTIME_MICRO:
1175 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1176 M_TSTMP)) {
1177 mbuf_tstmp2timespec(m, &ts1);
1178 timespec2bintime(&ts1, &bt1);
1179 getboottimebin(&boottimebin);
1180 bintime_add(&bt1, &boottimebin);
1181 bintime2timeval(&bt1, &t.tv);
1182 } else {
1183 microtime(&t.tv);
1184 }
1185 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv),
1186 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT);
1187 if (*mp != NULL) {
1188 mp = &(*mp)->m_next;
1189 stamped = true;
1190 }
1191 break;
1192
1193 case SO_TS_BINTIME:
1194 break;
1195
1196 case SO_TS_REALTIME:
1197 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1198 M_TSTMP)) {
1199 mbuf_tstmp2timespec(m, &t.ts);
1200 getboottimebin(&boottimebin);
1201 bintime2timespec(&boottimebin, &ts1);
1202 timespecadd(&t.ts, &ts1, &t.ts);
1203 } else {
1204 nanotime(&t.ts);
1205 }
1206 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1207 SCM_REALTIME, SOL_SOCKET, M_NOWAIT);
1208 if (*mp != NULL) {
1209 mp = &(*mp)->m_next;
1210 stamped = true;
1211 }
1212 break;
1213
1214 case SO_TS_MONOTONIC:
1215 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1216 M_TSTMP))
1217 mbuf_tstmp2timespec(m, &t.ts);
1218 else
1219 nanouptime(&t.ts);
1220 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1221 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT);
1222 if (*mp != NULL) {
1223 mp = &(*mp)->m_next;
1224 stamped = true;
1225 }
1226 break;
1227
1228 default:
1229 panic("unknown (corrupted) so_ts_clock");
1230 }
1231 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) ==
1232 (M_PKTHDR | M_TSTMP)) {
1233 struct sock_timestamp_info sti;
1234
1235 bzero(&sti, sizeof(sti));
1236 sti.st_info_flags = ST_INFO_HW;
1237 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1238 sti.st_info_flags |= ST_INFO_HW_HPREC;
1239 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1240 SOL_SOCKET, M_NOWAIT);
1241 if (*mp != NULL)
1242 mp = &(*mp)->m_next;
1243 }
1244 }
1245 #endif
1246
1247 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1248 /* RFC 2292 sec. 5 */
1249 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1250 struct in6_pktinfo pi6;
1251
1252 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1253 #ifdef INET
1254 struct ip *ip;
1255
1256 ip = mtod(m, struct ip *);
1257 pi6.ipi6_addr.s6_addr32[0] = 0;
1258 pi6.ipi6_addr.s6_addr32[1] = 0;
1259 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1260 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1261 #else
1262 /* We won't hit this code */
1263 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1264 #endif
1265 } else {
1266 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1267 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1268 }
1269 pi6.ipi6_ifindex =
1270 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1271
1272 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo),
1273 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6,
1274 M_NOWAIT);
1275 if (*mp)
1276 mp = &(*mp)->m_next;
1277 }
1278
1279 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1280 int hlim;
1281
1282 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1283 #ifdef INET
1284 struct ip *ip;
1285
1286 ip = mtod(m, struct ip *);
1287 hlim = ip->ip_ttl;
1288 #else
1289 /* We won't hit this code */
1290 hlim = 0;
1291 #endif
1292 } else {
1293 hlim = ip6->ip6_hlim & 0xff;
1294 }
1295 *mp = sbcreatecontrol(&hlim, sizeof(int),
1296 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1297 IPPROTO_IPV6, M_NOWAIT);
1298 if (*mp)
1299 mp = &(*mp)->m_next;
1300 }
1301
1302 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1303 int tclass;
1304
1305 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1306 #ifdef INET
1307 struct ip *ip;
1308
1309 ip = mtod(m, struct ip *);
1310 tclass = ip->ip_tos;
1311 #else
1312 /* We won't hit this code */
1313 tclass = 0;
1314 #endif
1315 } else {
1316 u_int32_t flowinfo;
1317
1318 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1319 flowinfo >>= 20;
1320 tclass = flowinfo & 0xff;
1321 }
1322 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS,
1323 IPPROTO_IPV6, M_NOWAIT);
1324 if (*mp)
1325 mp = &(*mp)->m_next;
1326 }
1327
1328 if (v4only != NULL) {
1329 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1330 *v4only = 1;
1331 } else {
1332 *v4only = 0;
1333 }
1334 }
1335
1336 return (mp);
1337 }
1338
1339 void
ip6_savecontrol(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp)1340 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp)
1341 {
1342 struct ip6_hdr *ip6;
1343 int v4only = 0;
1344
1345 mp = ip6_savecontrol_v4(inp, m, mp, &v4only);
1346 if (v4only)
1347 return;
1348
1349 ip6 = mtod(m, struct ip6_hdr *);
1350 /*
1351 * IPV6_HOPOPTS socket option. Recall that we required super-user
1352 * privilege for the option (see ip6_ctloutput), but it might be too
1353 * strict, since there might be some hop-by-hop options which can be
1354 * returned to normal user.
1355 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1356 */
1357 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) {
1358 /*
1359 * Check if a hop-by-hop options header is contatined in the
1360 * received packet, and if so, store the options as ancillary
1361 * data. Note that a hop-by-hop options header must be
1362 * just after the IPv6 header, which is assured through the
1363 * IPv6 input processing.
1364 */
1365 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1366 struct ip6_hbh *hbh;
1367 u_int hbhlen;
1368
1369 hbh = (struct ip6_hbh *)(ip6 + 1);
1370 hbhlen = (hbh->ip6h_len + 1) << 3;
1371
1372 /*
1373 * XXX: We copy the whole header even if a
1374 * jumbo payload option is included, the option which
1375 * is to be removed before returning according to
1376 * RFC2292.
1377 * Note: this constraint is removed in RFC3542
1378 */
1379 *mp = sbcreatecontrol(hbh, hbhlen,
1380 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1381 IPPROTO_IPV6, M_NOWAIT);
1382 if (*mp)
1383 mp = &(*mp)->m_next;
1384 }
1385 }
1386
1387 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1388 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1389
1390 /*
1391 * Search for destination options headers or routing
1392 * header(s) through the header chain, and stores each
1393 * header as ancillary data.
1394 * Note that the order of the headers remains in
1395 * the chain of ancillary data.
1396 */
1397 while (1) { /* is explicit loop prevention necessary? */
1398 struct ip6_ext *ip6e = NULL;
1399 u_int elen;
1400
1401 /*
1402 * if it is not an extension header, don't try to
1403 * pull it from the chain.
1404 */
1405 switch (nxt) {
1406 case IPPROTO_DSTOPTS:
1407 case IPPROTO_ROUTING:
1408 case IPPROTO_HOPOPTS:
1409 case IPPROTO_AH: /* is it possible? */
1410 break;
1411 default:
1412 goto loopend;
1413 }
1414
1415 if (off + sizeof(*ip6e) > m->m_len)
1416 goto loopend;
1417 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1418 if (nxt == IPPROTO_AH)
1419 elen = (ip6e->ip6e_len + 2) << 2;
1420 else
1421 elen = (ip6e->ip6e_len + 1) << 3;
1422 if (off + elen > m->m_len)
1423 goto loopend;
1424
1425 switch (nxt) {
1426 case IPPROTO_DSTOPTS:
1427 if (!(inp->inp_flags & IN6P_DSTOPTS))
1428 break;
1429
1430 *mp = sbcreatecontrol(ip6e, elen,
1431 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1432 IPPROTO_IPV6, M_NOWAIT);
1433 if (*mp)
1434 mp = &(*mp)->m_next;
1435 break;
1436 case IPPROTO_ROUTING:
1437 if (!(inp->inp_flags & IN6P_RTHDR))
1438 break;
1439
1440 *mp = sbcreatecontrol(ip6e, elen,
1441 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR),
1442 IPPROTO_IPV6, M_NOWAIT);
1443 if (*mp)
1444 mp = &(*mp)->m_next;
1445 break;
1446 case IPPROTO_HOPOPTS:
1447 case IPPROTO_AH: /* is it possible? */
1448 break;
1449
1450 default:
1451 /*
1452 * other cases have been filtered in the above.
1453 * none will visit this case. here we supply
1454 * the code just in case (nxt overwritten or
1455 * other cases).
1456 */
1457 goto loopend;
1458 }
1459
1460 /* proceed with the next header. */
1461 off += elen;
1462 nxt = ip6e->ip6e_nxt;
1463 ip6e = NULL;
1464 }
1465 loopend:
1466 ;
1467 }
1468
1469 if (inp->inp_flags2 & INP_RECVFLOWID) {
1470 uint32_t flowid, flow_type;
1471
1472 flowid = m->m_pkthdr.flowid;
1473 flow_type = M_HASHTYPE_GET(m);
1474
1475 /*
1476 * XXX should handle the failure of one or the
1477 * other - don't populate both?
1478 */
1479 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID,
1480 IPPROTO_IPV6, M_NOWAIT);
1481 if (*mp)
1482 mp = &(*mp)->m_next;
1483 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1484 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT);
1485 if (*mp)
1486 mp = &(*mp)->m_next;
1487 }
1488
1489 #ifdef RSS
1490 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1491 uint32_t flowid, flow_type;
1492 uint32_t rss_bucketid;
1493
1494 flowid = m->m_pkthdr.flowid;
1495 flow_type = M_HASHTYPE_GET(m);
1496
1497 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1498 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1499 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT);
1500 if (*mp)
1501 mp = &(*mp)->m_next;
1502 }
1503 }
1504 #endif
1505
1506 }
1507 #undef IS2292
1508
1509 void
ip6_notify_pmtu(struct inpcb * inp,struct sockaddr_in6 * dst,u_int32_t mtu)1510 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1511 {
1512 struct socket *so;
1513 struct mbuf *m_mtu;
1514 struct ip6_mtuinfo mtuctl;
1515
1516 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1517 /*
1518 * Notify the error by sending IPV6_PATHMTU ancillary data if
1519 * application wanted to know the MTU value.
1520 * NOTE: we notify disconnected sockets, because some udp
1521 * applications keep sending sockets disconnected.
1522 * NOTE: our implementation doesn't notify connected sockets that has
1523 * foreign address that is different than given destination addresses
1524 * (this is permitted by RFC 3542).
1525 */
1526 if ((inp->inp_flags & IN6P_MTU) == 0 || (
1527 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1528 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1529 return;
1530
1531 mtuctl.ip6m_mtu = mtu;
1532 mtuctl.ip6m_addr = *dst;
1533 if (sa6_recoverscope(&mtuctl.ip6m_addr))
1534 return;
1535
1536 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU,
1537 IPPROTO_IPV6, M_NOWAIT)) == NULL)
1538 return;
1539
1540 so = inp->inp_socket;
1541 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1542 == 0) {
1543 soroverflow(so);
1544 m_freem(m_mtu);
1545 /* XXX: should count statistics */
1546 } else
1547 sorwakeup(so);
1548 }
1549
1550 /*
1551 * Get pointer to the previous header followed by the header
1552 * currently processed.
1553 */
1554 int
ip6_get_prevhdr(const struct mbuf * m,int off)1555 ip6_get_prevhdr(const struct mbuf *m, int off)
1556 {
1557 struct ip6_ext ip6e;
1558 struct ip6_hdr *ip6;
1559 int len, nlen, nxt;
1560
1561 if (off == sizeof(struct ip6_hdr))
1562 return (offsetof(struct ip6_hdr, ip6_nxt));
1563 if (off < sizeof(struct ip6_hdr))
1564 panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1565
1566 ip6 = mtod(m, struct ip6_hdr *);
1567 nxt = ip6->ip6_nxt;
1568 len = sizeof(struct ip6_hdr);
1569 nlen = 0;
1570 while (len < off) {
1571 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1572 switch (nxt) {
1573 case IPPROTO_FRAGMENT:
1574 nlen = sizeof(struct ip6_frag);
1575 break;
1576 case IPPROTO_AH:
1577 nlen = (ip6e.ip6e_len + 2) << 2;
1578 break;
1579 default:
1580 nlen = (ip6e.ip6e_len + 1) << 3;
1581 }
1582 len += nlen;
1583 nxt = ip6e.ip6e_nxt;
1584 }
1585 return (len - nlen);
1586 }
1587
1588 /*
1589 * get next header offset. m will be retained.
1590 */
1591 int
ip6_nexthdr(const struct mbuf * m,int off,int proto,int * nxtp)1592 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1593 {
1594 struct ip6_hdr ip6;
1595 struct ip6_ext ip6e;
1596 struct ip6_frag fh;
1597
1598 /* just in case */
1599 if (m == NULL)
1600 panic("ip6_nexthdr: m == NULL");
1601 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1602 return -1;
1603
1604 switch (proto) {
1605 case IPPROTO_IPV6:
1606 if (m->m_pkthdr.len < off + sizeof(ip6))
1607 return -1;
1608 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1609 if (nxtp)
1610 *nxtp = ip6.ip6_nxt;
1611 off += sizeof(ip6);
1612 return off;
1613
1614 case IPPROTO_FRAGMENT:
1615 /*
1616 * terminate parsing if it is not the first fragment,
1617 * it does not make sense to parse through it.
1618 */
1619 if (m->m_pkthdr.len < off + sizeof(fh))
1620 return -1;
1621 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1622 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1623 if (fh.ip6f_offlg & IP6F_OFF_MASK)
1624 return -1;
1625 if (nxtp)
1626 *nxtp = fh.ip6f_nxt;
1627 off += sizeof(struct ip6_frag);
1628 return off;
1629
1630 case IPPROTO_AH:
1631 if (m->m_pkthdr.len < off + sizeof(ip6e))
1632 return -1;
1633 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1634 if (nxtp)
1635 *nxtp = ip6e.ip6e_nxt;
1636 off += (ip6e.ip6e_len + 2) << 2;
1637 return off;
1638
1639 case IPPROTO_HOPOPTS:
1640 case IPPROTO_ROUTING:
1641 case IPPROTO_DSTOPTS:
1642 if (m->m_pkthdr.len < off + sizeof(ip6e))
1643 return -1;
1644 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1645 if (nxtp)
1646 *nxtp = ip6e.ip6e_nxt;
1647 off += (ip6e.ip6e_len + 1) << 3;
1648 return off;
1649
1650 case IPPROTO_NONE:
1651 case IPPROTO_ESP:
1652 case IPPROTO_IPCOMP:
1653 /* give up */
1654 return -1;
1655
1656 default:
1657 return -1;
1658 }
1659
1660 /* NOTREACHED */
1661 }
1662
1663 /*
1664 * get offset for the last header in the chain. m will be kept untainted.
1665 */
1666 int
ip6_lasthdr(const struct mbuf * m,int off,int proto,int * nxtp)1667 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1668 {
1669 int newoff;
1670 int nxt;
1671
1672 if (!nxtp) {
1673 nxt = -1;
1674 nxtp = &nxt;
1675 }
1676 while (1) {
1677 newoff = ip6_nexthdr(m, off, proto, nxtp);
1678 if (newoff < 0)
1679 return off;
1680 else if (newoff < off)
1681 return -1; /* invalid */
1682 else if (newoff == off)
1683 return newoff;
1684
1685 off = newoff;
1686 proto = *nxtp;
1687 }
1688 }
1689