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 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
63 */
64
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #include "opt_route.h"
72 #include "opt_rss.h"
73
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/hhook.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/proc.h>
80 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/sdt.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/errno.h>
86 #include <sys/time.h>
87 #include <sys/kernel.h>
88 #include <sys/lock.h>
89 #include <sys/rmlock.h>
90 #include <sys/syslog.h>
91 #include <sys/sysctl.h>
92
93 #include <net/if.h>
94 #include <net/if_var.h>
95 #include <net/if_types.h>
96 #include <net/if_dl.h>
97 #include <net/route.h>
98 #include <net/netisr.h>
99 #include <net/rss_config.h>
100 #include <net/pfil.h>
101 #include <net/vnet.h>
102
103 #include <netinet/in.h>
104 #include <netinet/in_kdtrace.h>
105 #include <netinet/ip_var.h>
106 #include <netinet/in_systm.h>
107 #include <net/if_llatbl.h>
108 #ifdef INET
109 #include <netinet/ip.h>
110 #include <netinet/ip_icmp.h>
111 #endif /* INET */
112 #include <netinet/ip6.h>
113 #include <netinet6/in6_var.h>
114 #include <netinet6/ip6_var.h>
115 #include <netinet/in_pcb.h>
116 #include <netinet/icmp6.h>
117 #include <netinet6/scope6_var.h>
118 #include <netinet6/in6_ifattach.h>
119 #include <netinet6/mld6_var.h>
120 #include <netinet6/nd6.h>
121 #include <netinet6/in6_rss.h>
122
123 #include <netipsec/ipsec_support.h>
124
125 #include <netinet6/ip6protosw.h>
126
127 extern struct domain inet6domain;
128
129 u_char ip6_protox[IPPROTO_MAX];
130 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
131 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
132 VNET_DEFINE(u_long, in6_ifaddrhmask);
133
134 static struct netisr_handler ip6_nh = {
135 .nh_name = "ip6",
136 .nh_handler = ip6_input,
137 .nh_proto = NETISR_IPV6,
138 #ifdef RSS
139 .nh_m2cpuid = rss_soft_m2cpuid_v6,
140 .nh_policy = NETISR_POLICY_CPU,
141 .nh_dispatch = NETISR_DISPATCH_HYBRID,
142 #else
143 .nh_policy = NETISR_POLICY_FLOW,
144 #endif
145 };
146
147 static int
148 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
149 {
150 int error, qlimit;
151
152 netisr_getqlimit(&ip6_nh, &qlimit);
153 error = sysctl_handle_int(oidp, &qlimit, 0, req);
154 if (error || !req->newptr)
155 return (error);
156 if (qlimit < 1)
157 return (EINVAL);
158 return (netisr_setqlimit(&ip6_nh, qlimit));
159 }
160 SYSCTL_DECL(_net_inet6_ip6);
161 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen,
162 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
163 0, 0, sysctl_netinet6_intr_queue_maxlen, "I",
164 "Maximum size of the IPv6 input queue");
165
166 #ifdef RSS
167 static struct netisr_handler ip6_direct_nh = {
168 .nh_name = "ip6_direct",
169 .nh_handler = ip6_direct_input,
170 .nh_proto = NETISR_IPV6_DIRECT,
171 .nh_m2cpuid = rss_soft_m2cpuid_v6,
172 .nh_policy = NETISR_POLICY_CPU,
173 .nh_dispatch = NETISR_DISPATCH_HYBRID,
174 };
175
176 static int
177 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
178 {
179 int error, qlimit;
180
181 netisr_getqlimit(&ip6_direct_nh, &qlimit);
182 error = sysctl_handle_int(oidp, &qlimit, 0, req);
183 if (error || !req->newptr)
184 return (error);
185 if (qlimit < 1)
186 return (EINVAL);
187 return (netisr_setqlimit(&ip6_direct_nh, qlimit));
188 }
189 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
190 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
191 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I",
192 "Maximum size of the IPv6 direct input queue");
193
194 #endif
195
196 VNET_DEFINE(pfil_head_t, inet6_pfil_head);
197
198 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
199 VNET_PCPUSTAT_SYSINIT(ip6stat);
200 #ifdef VIMAGE
201 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
202 #endif /* VIMAGE */
203
204 struct rmlock in6_ifaddr_lock;
205 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
206
207 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
208
209 /*
210 * IP6 initialization: fill in IP6 protocol switch table.
211 * All protocols not implemented in kernel go to raw IP6 protocol handler.
212 */
213 void
214 ip6_init(void)
215 {
216 struct pfil_head_args args;
217 struct protosw *pr;
218 int i;
219
220 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
221 &V_ip6_auto_linklocal);
222 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
223 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
224
225 CK_STAILQ_INIT(&V_in6_ifaddrhead);
226 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
227 &V_in6_ifaddrhmask);
228
229 /* Initialize packet filter hooks. */
230 args.pa_version = PFIL_VERSION;
231 args.pa_flags = PFIL_IN | PFIL_OUT;
232 args.pa_type = PFIL_TYPE_IP6;
233 args.pa_headname = PFIL_INET6_NAME;
234 V_inet6_pfil_head = pfil_head_register(&args);
235
236 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
237 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
238 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
239 printf("%s: WARNING: unable to register input helper hook\n",
240 __func__);
241 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
242 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
243 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
244 printf("%s: WARNING: unable to register output helper hook\n",
245 __func__);
246
247 scope6_init();
248 addrsel_policy_init();
249 nd6_init();
250 frag6_init();
251
252 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
253
254 /* Skip global initialization stuff for non-default instances. */
255 #ifdef VIMAGE
256 if (!IS_DEFAULT_VNET(curvnet)) {
257 netisr_register_vnet(&ip6_nh);
258 #ifdef RSS
259 netisr_register_vnet(&ip6_direct_nh);
260 #endif
261 return;
262 }
263 #endif
264
265 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
266 if (pr == NULL)
267 panic("ip6_init");
268
269 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
270 for (i = 0; i < IPPROTO_MAX; i++)
271 ip6_protox[i] = pr - inet6sw;
272 /*
273 * Cycle through IP protocols and put them into the appropriate place
274 * in ip6_protox[].
275 */
276 for (pr = inet6domain.dom_protosw;
277 pr < inet6domain.dom_protoswNPROTOSW; pr++)
278 if (pr->pr_domain->dom_family == PF_INET6 &&
279 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
280 /* Be careful to only index valid IP protocols. */
281 if (pr->pr_protocol < IPPROTO_MAX)
282 ip6_protox[pr->pr_protocol] = pr - inet6sw;
283 }
284
285 netisr_register(&ip6_nh);
286 #ifdef RSS
287 netisr_register(&ip6_direct_nh);
288 #endif
289 }
290
291 /*
292 * The protocol to be inserted into ip6_protox[] must be already registered
293 * in inet6sw[], either statically or through pf_proto_register().
294 */
295 int
296 ip6proto_register(short ip6proto)
297 {
298 struct protosw *pr;
299
300 /* Sanity checks. */
301 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
302 return (EPROTONOSUPPORT);
303
304 /*
305 * The protocol slot must not be occupied by another protocol
306 * already. An index pointing to IPPROTO_RAW is unused.
307 */
308 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
309 if (pr == NULL)
310 return (EPFNOSUPPORT);
311 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */
312 return (EEXIST);
313
314 /*
315 * Find the protocol position in inet6sw[] and set the index.
316 */
317 for (pr = inet6domain.dom_protosw;
318 pr < inet6domain.dom_protoswNPROTOSW; pr++) {
319 if (pr->pr_domain->dom_family == PF_INET6 &&
320 pr->pr_protocol && pr->pr_protocol == ip6proto) {
321 ip6_protox[pr->pr_protocol] = pr - inet6sw;
322 return (0);
323 }
324 }
325 return (EPROTONOSUPPORT);
326 }
327
328 int
329 ip6proto_unregister(short ip6proto)
330 {
331 struct protosw *pr;
332
333 /* Sanity checks. */
334 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
335 return (EPROTONOSUPPORT);
336
337 /* Check if the protocol was indeed registered. */
338 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
339 if (pr == NULL)
340 return (EPFNOSUPPORT);
341 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */
342 return (ENOENT);
343
344 /* Reset the protocol slot to IPPROTO_RAW. */
345 ip6_protox[ip6proto] = pr - inet6sw;
346 return (0);
347 }
348
349 #ifdef VIMAGE
350 static void
351 ip6_destroy(void *unused __unused)
352 {
353 struct ifaddr *ifa, *nifa;
354 struct ifnet *ifp;
355 int error;
356
357 #ifdef RSS
358 netisr_unregister_vnet(&ip6_direct_nh);
359 #endif
360 netisr_unregister_vnet(&ip6_nh);
361
362 pfil_head_unregister(V_inet6_pfil_head);
363 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
364 if (error != 0) {
365 printf("%s: WARNING: unable to deregister input helper hook "
366 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
367 "error %d returned\n", __func__, error);
368 }
369 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
370 if (error != 0) {
371 printf("%s: WARNING: unable to deregister output helper hook "
372 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
373 "error %d returned\n", __func__, error);
374 }
375
376 /* Cleanup addresses. */
377 IFNET_RLOCK();
378 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
379 /* Cannot lock here - lock recursion. */
380 /* IF_ADDR_LOCK(ifp); */
381 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
382 if (ifa->ifa_addr->sa_family != AF_INET6)
383 continue;
384 in6_purgeaddr(ifa);
385 }
386 /* IF_ADDR_UNLOCK(ifp); */
387 in6_ifdetach_destroy(ifp);
388 mld_domifdetach(ifp);
389 }
390 IFNET_RUNLOCK();
391
392 /* Make sure any routes are gone as well. */
393 rib_flush_routes_family(AF_INET6);
394
395 frag6_destroy();
396 nd6_destroy();
397 in6_ifattach_destroy();
398
399 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
400 }
401
402 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL);
403 #endif
404
405 static int
406 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off,
407 int *nxt, int *ours)
408 {
409 struct mbuf *m;
410 struct ip6_hdr *ip6;
411 struct ip6_hbh *hbh;
412
413 if (ip6_hopopts_input(plen, rtalert, mp, off)) {
414 #if 0 /*touches NULL pointer*/
415 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard);
416 #endif
417 goto out; /* m have already been freed */
418 }
419
420 /* adjust pointer */
421 m = *mp;
422 ip6 = mtod(m, struct ip6_hdr *);
423
424 /*
425 * if the payload length field is 0 and the next header field
426 * indicates Hop-by-Hop Options header, then a Jumbo Payload
427 * option MUST be included.
428 */
429 if (ip6->ip6_plen == 0 && *plen == 0) {
430 /*
431 * Note that if a valid jumbo payload option is
432 * contained, ip6_hopopts_input() must set a valid
433 * (non-zero) payload length to the variable plen.
434 */
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
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 = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
525 }
526 return;
527 bad:
528 m_freem(m);
529 }
530 #endif
531
532 void
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 ((ND_IFINFO(rcvif)->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 if (m->m_flags & M_LOOP) {
583 IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
584 } else if (rcvif->if_index < IP6S_M2MMAX)
585 IP6STAT_INC(ip6s_m2m[rcvif->if_index]);
586 else
587 IP6STAT_INC(ip6s_m2m[0]);
588 } else
589 IP6STAT_INC(ip6s_m1);
590 }
591
592 in6_ifstat_inc(rcvif, ifs6_in_receive);
593 IP6STAT_INC(ip6s_total);
594
595 /*
596 * L2 bridge code and some other code can return mbuf chain
597 * that does not conform to KAME requirement. too bad.
598 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
599 */
600 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
601 struct mbuf *n;
602
603 if (m->m_pkthdr.len > MHLEN)
604 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
605 else
606 n = m_gethdr(M_NOWAIT, MT_DATA);
607 if (n == NULL)
608 goto bad;
609
610 m_move_pkthdr(n, m);
611 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
612 n->m_len = n->m_pkthdr.len;
613 m_freem(m);
614 m = n;
615 }
616 if (m->m_len < sizeof(struct ip6_hdr)) {
617 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
618 IP6STAT_INC(ip6s_toosmall);
619 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
620 goto bad;
621 }
622 }
623
624 ip6 = mtod(m, struct ip6_hdr *);
625 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
626 IP6STAT_INC(ip6s_badvers);
627 in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
628 goto bad;
629 }
630
631 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
632 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6);
633
634 /*
635 * Check against address spoofing/corruption.
636 */
637 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
638 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
639 /*
640 * XXX: "badscope" is not very suitable for a multicast source.
641 */
642 IP6STAT_INC(ip6s_badscope);
643 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
644 goto bad;
645 }
646 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
647 !(m->m_flags & M_LOOP)) {
648 /*
649 * In this case, the packet should come from the loopback
650 * interface. However, we cannot just check the if_flags,
651 * because ip6_mloopback() passes the "actual" interface
652 * as the outgoing/incoming interface.
653 */
654 IP6STAT_INC(ip6s_badscope);
655 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
656 goto bad;
657 }
658 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
659 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
660 /*
661 * RFC4291 2.7:
662 * Nodes must not originate a packet to a multicast address
663 * whose scop field contains the reserved value 0; if such
664 * a packet is received, it must be silently dropped.
665 */
666 IP6STAT_INC(ip6s_badscope);
667 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
668 goto bad;
669 }
670 #ifdef ALTQ
671 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
672 /* packet is dropped by traffic conditioner */
673 return;
674 }
675 #endif
676 /*
677 * The following check is not documented in specs. A malicious
678 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
679 * and bypass security checks (act as if it was from 127.0.0.1 by using
680 * IPv6 src ::ffff:127.0.0.1). Be cautious.
681 *
682 * We have supported IPv6-only kernels for a few years and this issue
683 * has not come up. The world seems to move mostly towards not using
684 * v4mapped on the wire, so it makes sense for us to keep rejecting
685 * any such packets.
686 */
687 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
688 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
689 IP6STAT_INC(ip6s_badscope);
690 in6_ifstat_inc(rcvif, ifs6_in_addrerr);
691 goto bad;
692 }
693 #if 0
694 /*
695 * Reject packets with IPv4 compatible addresses (auto tunnel).
696 *
697 * The code forbids auto tunnel relay case in RFC1933 (the check is
698 * stronger than RFC1933). We may want to re-enable it if mech-xx
699 * is revised to forbid relaying case.
700 */
701 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
702 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
703 IP6STAT_INC(ip6s_badscope);
704 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
705 goto bad;
706 }
707 #endif
708 /*
709 * Try to forward the packet, but if we fail continue.
710 * ip6_tryforward() does not generate redirects, so fall
711 * through to normal processing if redirects are required.
712 * ip6_tryforward() does inbound and outbound packet firewall
713 * processing. If firewall has decided that destination becomes
714 * our local address, it sets M_FASTFWD_OURS flag. In this
715 * case skip another inbound firewall processing and update
716 * ip6 pointer.
717 */
718 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0
719 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
720 && (!IPSEC_ENABLED(ipv6) ||
721 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0)
722 #endif
723 ) {
724 if ((m = ip6_tryforward(m)) == NULL)
725 return;
726 if (m->m_flags & M_FASTFWD_OURS) {
727 ip6 = mtod(m, struct ip6_hdr *);
728 goto passin;
729 }
730 }
731 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
732 /*
733 * Bypass packet filtering for packets previously handled by IPsec.
734 */
735 if (IPSEC_ENABLED(ipv6) &&
736 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0)
737 goto passin;
738 #endif
739 /*
740 * Run through list of hooks for input packets.
741 *
742 * NB: Beware of the destination address changing
743 * (e.g. by NAT rewriting). When this happens,
744 * tell ip6_forward to do the right thing.
745 */
746
747 /* Jump over all PFIL processing if hooks are not active. */
748 if (!PFIL_HOOKED_IN(V_inet6_pfil_head))
749 goto passin;
750
751 odst = ip6->ip6_dst;
752 if (pfil_run_hooks(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, PFIL_IN,
753 NULL) != PFIL_PASS)
754 return;
755 ip6 = mtod(m, struct ip6_hdr *);
756 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
757 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP &&
758 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
759 /*
760 * Directly ship the packet on. This allows forwarding
761 * packets originally destined to us to some other directly
762 * connected host.
763 */
764 ip6_forward(m, 1);
765 return;
766 }
767
768 passin:
769 /*
770 * Disambiguate address scope zones (if there is ambiguity).
771 * We first make sure that the original source or destination address
772 * is not in our internal form for scoped addresses. Such addresses
773 * are not necessarily invalid spec-wise, but we cannot accept them due
774 * to the usage conflict.
775 * in6_setscope() then also checks and rejects the cases where src or
776 * dst are the loopback address and the receiving interface
777 * is not loopback.
778 */
779 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
780 IP6STAT_INC(ip6s_badscope); /* XXX */
781 goto bad;
782 }
783 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) ||
784 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) {
785 IP6STAT_INC(ip6s_badscope);
786 goto bad;
787 }
788 if (m->m_flags & M_FASTFWD_OURS) {
789 m->m_flags &= ~M_FASTFWD_OURS;
790 ours = 1;
791 goto hbhcheck;
792 }
793 /*
794 * Multicast check. Assume packet is for us to avoid
795 * prematurely taking locks.
796 */
797 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
798 ours = 1;
799 in6_ifstat_inc(rcvif, ifs6_in_mcast);
800 goto hbhcheck;
801 }
802 /*
803 * Unicast check
804 * XXX: For now we keep link-local IPv6 addresses with embedded
805 * scope zone id, therefore we use zero zoneid here.
806 */
807 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
808 if (ia != NULL) {
809 if (ia->ia6_flags & IN6_IFF_NOTREADY) {
810 char ip6bufs[INET6_ADDRSTRLEN];
811 char ip6bufd[INET6_ADDRSTRLEN];
812 /* address is not ready, so discard the packet. */
813 nd6log((LOG_INFO,
814 "ip6_input: packet to an unready address %s->%s\n",
815 ip6_sprintf(ip6bufs, &ip6->ip6_src),
816 ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
817 ifa_free(&ia->ia_ifa);
818 goto bad;
819 }
820 /* Count the packet in the ip address stats */
821 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
822 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
823 ifa_free(&ia->ia_ifa);
824 ours = 1;
825 goto hbhcheck;
826 }
827
828 /*
829 * Now there is no reason to process the packet if it's not our own
830 * and we're not a router.
831 */
832 if (!V_ip6_forwarding) {
833 IP6STAT_INC(ip6s_cantforward);
834 goto bad;
835 }
836
837 hbhcheck:
838 /*
839 * Process Hop-by-Hop options header if it's contained.
840 * m may be modified in ip6_hopopts_input().
841 * If a JumboPayload option is included, plen will also be modified.
842 */
843 plen = (u_int32_t)ntohs(ip6->ip6_plen);
844 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
845 if (ip6_input_hbh(&m, &plen, &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_mrouter &&
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 * Tell launch routine the next header
903 */
904 IP6STAT_INC(ip6s_delivered);
905 in6_ifstat_inc(rcvif, ifs6_in_deliver);
906 nest = 0;
907
908 while (nxt != IPPROTO_DONE) {
909 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
910 IP6STAT_INC(ip6s_toomanyhdr);
911 goto bad;
912 }
913
914 /*
915 * protection against faulty packet - there should be
916 * more sanity checks in header chain processing.
917 */
918 if (m->m_pkthdr.len < off) {
919 IP6STAT_INC(ip6s_tooshort);
920 in6_ifstat_inc(rcvif, ifs6_in_truncated);
921 goto bad;
922 }
923
924 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
925 if (IPSEC_ENABLED(ipv6)) {
926 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
927 return;
928 }
929 #endif /* IPSEC */
930
931 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
932 }
933 return;
934 bad:
935 in6_ifstat_inc(rcvif, ifs6_in_discard);
936 if (m != NULL)
937 m_freem(m);
938 }
939
940 /*
941 * Hop-by-Hop options header processing. If a valid jumbo payload option is
942 * included, the real payload length will be stored in plenp.
943 *
944 * rtalertp - XXX: should be stored more smart way
945 */
946 static int
947 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
948 struct mbuf **mp, int *offp)
949 {
950 struct mbuf *m = *mp;
951 int off = *offp, hbhlen;
952 struct ip6_hbh *hbh;
953
954 /* validation of the length of the header */
955 if (m->m_len < off + sizeof(*hbh)) {
956 m = m_pullup(m, off + sizeof(*hbh));
957 if (m == NULL) {
958 IP6STAT_INC(ip6s_exthdrtoolong);
959 *mp = NULL;
960 return (-1);
961 }
962 }
963 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
964 hbhlen = (hbh->ip6h_len + 1) << 3;
965
966 if (m->m_len < off + hbhlen) {
967 m = m_pullup(m, off + hbhlen);
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 off += hbhlen;
976 hbhlen -= sizeof(struct ip6_hbh);
977 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
978 hbhlen, rtalertp, plenp) < 0) {
979 *mp = NULL;
980 return (-1);
981 }
982
983 *offp = off;
984 *mp = m;
985 return (0);
986 }
987
988 /*
989 * Search header for all Hop-by-hop options and process each option.
990 * This function is separate from ip6_hopopts_input() in order to
991 * handle a case where the sending node itself process its hop-by-hop
992 * options header. In such a case, the function is called from ip6_output().
993 *
994 * The function assumes that hbh header is located right after the IPv6 header
995 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
996 * opthead + hbhlen is located in contiguous memory region.
997 */
998 int
999 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1000 u_int32_t *rtalertp, u_int32_t *plenp)
1001 {
1002 struct ip6_hdr *ip6;
1003 int optlen = 0;
1004 u_int8_t *opt = opthead;
1005 u_int16_t rtalert_val;
1006 u_int32_t jumboplen;
1007 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1008
1009 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1010 switch (*opt) {
1011 case IP6OPT_PAD1:
1012 optlen = 1;
1013 break;
1014 case IP6OPT_PADN:
1015 if (hbhlen < IP6OPT_MINLEN) {
1016 IP6STAT_INC(ip6s_toosmall);
1017 goto bad;
1018 }
1019 optlen = *(opt + 1) + 2;
1020 break;
1021 case IP6OPT_ROUTER_ALERT:
1022 /* XXX may need check for alignment */
1023 if (hbhlen < IP6OPT_RTALERT_LEN) {
1024 IP6STAT_INC(ip6s_toosmall);
1025 goto bad;
1026 }
1027 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1028 /* XXX stat */
1029 icmp6_error(m, ICMP6_PARAM_PROB,
1030 ICMP6_PARAMPROB_HEADER,
1031 erroff + opt + 1 - opthead);
1032 return (-1);
1033 }
1034 optlen = IP6OPT_RTALERT_LEN;
1035 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1036 *rtalertp = ntohs(rtalert_val);
1037 break;
1038 case IP6OPT_JUMBO:
1039 /* XXX may need check for alignment */
1040 if (hbhlen < IP6OPT_JUMBO_LEN) {
1041 IP6STAT_INC(ip6s_toosmall);
1042 goto bad;
1043 }
1044 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1045 /* XXX stat */
1046 icmp6_error(m, ICMP6_PARAM_PROB,
1047 ICMP6_PARAMPROB_HEADER,
1048 erroff + opt + 1 - opthead);
1049 return (-1);
1050 }
1051 optlen = IP6OPT_JUMBO_LEN;
1052
1053 /*
1054 * IPv6 packets that have non 0 payload length
1055 * must not contain a jumbo payload option.
1056 */
1057 ip6 = mtod(m, struct ip6_hdr *);
1058 if (ip6->ip6_plen) {
1059 IP6STAT_INC(ip6s_badoptions);
1060 icmp6_error(m, ICMP6_PARAM_PROB,
1061 ICMP6_PARAMPROB_HEADER,
1062 erroff + opt - opthead);
1063 return (-1);
1064 }
1065
1066 /*
1067 * We may see jumbolen in unaligned location, so
1068 * we'd need to perform bcopy().
1069 */
1070 bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1071 jumboplen = (u_int32_t)htonl(jumboplen);
1072
1073 #if 1
1074 /*
1075 * if there are multiple jumbo payload options,
1076 * *plenp will be non-zero and the packet will be
1077 * rejected.
1078 * the behavior may need some debate in ipngwg -
1079 * multiple options does not make sense, however,
1080 * there's no explicit mention in specification.
1081 */
1082 if (*plenp != 0) {
1083 IP6STAT_INC(ip6s_badoptions);
1084 icmp6_error(m, ICMP6_PARAM_PROB,
1085 ICMP6_PARAMPROB_HEADER,
1086 erroff + opt + 2 - opthead);
1087 return (-1);
1088 }
1089 #endif
1090
1091 /*
1092 * jumbo payload length must be larger than 65535.
1093 */
1094 if (jumboplen <= IPV6_MAXPACKET) {
1095 IP6STAT_INC(ip6s_badoptions);
1096 icmp6_error(m, ICMP6_PARAM_PROB,
1097 ICMP6_PARAMPROB_HEADER,
1098 erroff + opt + 2 - opthead);
1099 return (-1);
1100 }
1101 *plenp = jumboplen;
1102
1103 break;
1104 default: /* unknown option */
1105 if (hbhlen < IP6OPT_MINLEN) {
1106 IP6STAT_INC(ip6s_toosmall);
1107 goto bad;
1108 }
1109 optlen = ip6_unknown_opt(opt, m,
1110 erroff + opt - opthead);
1111 if (optlen == -1)
1112 return (-1);
1113 optlen += 2;
1114 break;
1115 }
1116 }
1117
1118 return (0);
1119
1120 bad:
1121 m_freem(m);
1122 return (-1);
1123 }
1124
1125 /*
1126 * Unknown option processing.
1127 * The third argument `off' is the offset from the IPv6 header to the option,
1128 * which is necessary if the IPv6 header the and option header and IPv6 header
1129 * is not contiguous in order to return an ICMPv6 error.
1130 */
1131 int
1132 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1133 {
1134 struct ip6_hdr *ip6;
1135
1136 switch (IP6OPT_TYPE(*optp)) {
1137 case IP6OPT_TYPE_SKIP: /* ignore the option */
1138 return ((int)*(optp + 1));
1139 case IP6OPT_TYPE_DISCARD: /* silently discard */
1140 m_freem(m);
1141 return (-1);
1142 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1143 IP6STAT_INC(ip6s_badoptions);
1144 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1145 return (-1);
1146 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1147 IP6STAT_INC(ip6s_badoptions);
1148 ip6 = mtod(m, struct ip6_hdr *);
1149 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1150 (m->m_flags & (M_BCAST|M_MCAST)))
1151 m_freem(m);
1152 else
1153 icmp6_error(m, ICMP6_PARAM_PROB,
1154 ICMP6_PARAMPROB_OPTION, off);
1155 return (-1);
1156 }
1157
1158 m_freem(m); /* XXX: NOTREACHED */
1159 return (-1);
1160 }
1161
1162 /*
1163 * Create the "control" list for this pcb.
1164 * These functions will not modify mbuf chain at all.
1165 *
1166 * The routine will be called from upper layer handlers like tcp6_input().
1167 * Thus the routine assumes that the caller (tcp6_input) have already
1168 * called m_pullup() and all the extension headers are located in the
1169 * very first mbuf on the mbuf chain.
1170 *
1171 * ip6_savecontrol_v4 will handle those options that are possible to be
1172 * set on a v4-mapped socket.
1173 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1174 * options and handle the v6-only ones itself.
1175 */
1176 struct mbuf **
1177 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1178 int *v4only)
1179 {
1180 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1181
1182 #ifdef SO_TIMESTAMP
1183 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1184 union {
1185 struct timeval tv;
1186 struct bintime bt;
1187 struct timespec ts;
1188 } t;
1189 struct bintime boottimebin, bt1;
1190 struct timespec ts1;
1191 bool stamped;
1192
1193 stamped = false;
1194 switch (inp->inp_socket->so_ts_clock) {
1195 case SO_TS_REALTIME_MICRO:
1196 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1197 M_TSTMP)) {
1198 mbuf_tstmp2timespec(m, &ts1);
1199 timespec2bintime(&ts1, &bt1);
1200 getboottimebin(&boottimebin);
1201 bintime_add(&bt1, &boottimebin);
1202 bintime2timeval(&bt1, &t.tv);
1203 } else {
1204 microtime(&t.tv);
1205 }
1206 *mp = sbcreatecontrol((caddr_t) &t.tv, sizeof(t.tv),
1207 SCM_TIMESTAMP, SOL_SOCKET);
1208 if (*mp != NULL) {
1209 mp = &(*mp)->m_next;
1210 stamped = true;
1211 }
1212 break;
1213
1214 case SO_TS_BINTIME:
1215 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1216 M_TSTMP)) {
1217 mbuf_tstmp2timespec(m, &ts1);
1218 timespec2bintime(&ts1, &t.bt);
1219 getboottimebin(&boottimebin);
1220 bintime_add(&t.bt, &boottimebin);
1221 } else {
1222 bintime(&t.bt);
1223 }
1224 *mp = sbcreatecontrol((caddr_t)&t.bt, sizeof(t.bt),
1225 SCM_BINTIME, SOL_SOCKET);
1226 if (*mp != NULL) {
1227 mp = &(*mp)->m_next;
1228 stamped = true;
1229 }
1230 break;
1231
1232 case SO_TS_REALTIME:
1233 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1234 M_TSTMP)) {
1235 mbuf_tstmp2timespec(m, &t.ts);
1236 getboottimebin(&boottimebin);
1237 bintime2timespec(&boottimebin, &ts1);
1238 timespecadd(&t.ts, &ts1, &t.ts);
1239 } else {
1240 nanotime(&t.ts);
1241 }
1242 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts),
1243 SCM_REALTIME, SOL_SOCKET);
1244 if (*mp != NULL) {
1245 mp = &(*mp)->m_next;
1246 stamped = true;
1247 }
1248 break;
1249
1250 case SO_TS_MONOTONIC:
1251 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1252 M_TSTMP))
1253 mbuf_tstmp2timespec(m, &t.ts);
1254 else
1255 nanouptime(&t.ts);
1256 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts),
1257 SCM_MONOTONIC, SOL_SOCKET);
1258 if (*mp != NULL) {
1259 mp = &(*mp)->m_next;
1260 stamped = true;
1261 }
1262 break;
1263
1264 default:
1265 panic("unknown (corrupted) so_ts_clock");
1266 }
1267 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) ==
1268 (M_PKTHDR | M_TSTMP)) {
1269 struct sock_timestamp_info sti;
1270
1271 bzero(&sti, sizeof(sti));
1272 sti.st_info_flags = ST_INFO_HW;
1273 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1274 sti.st_info_flags |= ST_INFO_HW_HPREC;
1275 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti),
1276 SCM_TIME_INFO, SOL_SOCKET);
1277 if (*mp != NULL)
1278 mp = &(*mp)->m_next;
1279 }
1280 }
1281 #endif
1282
1283 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1284 /* RFC 2292 sec. 5 */
1285 if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1286 struct in6_pktinfo pi6;
1287
1288 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1289 #ifdef INET
1290 struct ip *ip;
1291
1292 ip = mtod(m, struct ip *);
1293 pi6.ipi6_addr.s6_addr32[0] = 0;
1294 pi6.ipi6_addr.s6_addr32[1] = 0;
1295 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1296 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1297 #else
1298 /* We won't hit this code */
1299 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1300 #endif
1301 } else {
1302 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1303 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1304 }
1305 pi6.ipi6_ifindex =
1306 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1307
1308 *mp = sbcreatecontrol((caddr_t) &pi6,
1309 sizeof(struct in6_pktinfo),
1310 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1311 if (*mp)
1312 mp = &(*mp)->m_next;
1313 }
1314
1315 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1316 int hlim;
1317
1318 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1319 #ifdef INET
1320 struct ip *ip;
1321
1322 ip = mtod(m, struct ip *);
1323 hlim = ip->ip_ttl;
1324 #else
1325 /* We won't hit this code */
1326 hlim = 0;
1327 #endif
1328 } else {
1329 hlim = ip6->ip6_hlim & 0xff;
1330 }
1331 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1332 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1333 IPPROTO_IPV6);
1334 if (*mp)
1335 mp = &(*mp)->m_next;
1336 }
1337
1338 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1339 int tclass;
1340
1341 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1342 #ifdef INET
1343 struct ip *ip;
1344
1345 ip = mtod(m, struct ip *);
1346 tclass = ip->ip_tos;
1347 #else
1348 /* We won't hit this code */
1349 tclass = 0;
1350 #endif
1351 } else {
1352 u_int32_t flowinfo;
1353
1354 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1355 flowinfo >>= 20;
1356 tclass = flowinfo & 0xff;
1357 }
1358 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1359 IPV6_TCLASS, IPPROTO_IPV6);
1360 if (*mp)
1361 mp = &(*mp)->m_next;
1362 }
1363
1364 if (v4only != NULL) {
1365 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1366 *v4only = 1;
1367 } else {
1368 *v4only = 0;
1369 }
1370 }
1371
1372 return (mp);
1373 }
1374
1375 void
1376 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp)
1377 {
1378 struct ip6_hdr *ip6;
1379 int v4only = 0;
1380
1381 mp = ip6_savecontrol_v4(inp, m, mp, &v4only);
1382 if (v4only)
1383 return;
1384
1385 ip6 = mtod(m, struct ip6_hdr *);
1386 /*
1387 * IPV6_HOPOPTS socket option. Recall that we required super-user
1388 * privilege for the option (see ip6_ctloutput), but it might be too
1389 * strict, since there might be some hop-by-hop options which can be
1390 * returned to normal user.
1391 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1392 */
1393 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) {
1394 /*
1395 * Check if a hop-by-hop options header is contatined in the
1396 * received packet, and if so, store the options as ancillary
1397 * data. Note that a hop-by-hop options header must be
1398 * just after the IPv6 header, which is assured through the
1399 * IPv6 input processing.
1400 */
1401 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1402 struct ip6_hbh *hbh;
1403 int hbhlen;
1404
1405 hbh = (struct ip6_hbh *)(ip6 + 1);
1406 hbhlen = (hbh->ip6h_len + 1) << 3;
1407
1408 /*
1409 * XXX: We copy the whole header even if a
1410 * jumbo payload option is included, the option which
1411 * is to be removed before returning according to
1412 * RFC2292.
1413 * Note: this constraint is removed in RFC3542
1414 */
1415 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1416 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1417 IPPROTO_IPV6);
1418 if (*mp)
1419 mp = &(*mp)->m_next;
1420 }
1421 }
1422
1423 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1424 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1425
1426 /*
1427 * Search for destination options headers or routing
1428 * header(s) through the header chain, and stores each
1429 * header as ancillary data.
1430 * Note that the order of the headers remains in
1431 * the chain of ancillary data.
1432 */
1433 while (1) { /* is explicit loop prevention necessary? */
1434 struct ip6_ext *ip6e = NULL;
1435 int elen;
1436
1437 /*
1438 * if it is not an extension header, don't try to
1439 * pull it from the chain.
1440 */
1441 switch (nxt) {
1442 case IPPROTO_DSTOPTS:
1443 case IPPROTO_ROUTING:
1444 case IPPROTO_HOPOPTS:
1445 case IPPROTO_AH: /* is it possible? */
1446 break;
1447 default:
1448 goto loopend;
1449 }
1450
1451 if (off + sizeof(*ip6e) > m->m_len)
1452 goto loopend;
1453 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1454 if (nxt == IPPROTO_AH)
1455 elen = (ip6e->ip6e_len + 2) << 2;
1456 else
1457 elen = (ip6e->ip6e_len + 1) << 3;
1458 if (off + elen > m->m_len)
1459 goto loopend;
1460
1461 switch (nxt) {
1462 case IPPROTO_DSTOPTS:
1463 if (!(inp->inp_flags & IN6P_DSTOPTS))
1464 break;
1465
1466 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1467 IS2292(inp,
1468 IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1469 IPPROTO_IPV6);
1470 if (*mp)
1471 mp = &(*mp)->m_next;
1472 break;
1473 case IPPROTO_ROUTING:
1474 if (!(inp->inp_flags & IN6P_RTHDR))
1475 break;
1476
1477 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1478 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR),
1479 IPPROTO_IPV6);
1480 if (*mp)
1481 mp = &(*mp)->m_next;
1482 break;
1483 case IPPROTO_HOPOPTS:
1484 case IPPROTO_AH: /* is it possible? */
1485 break;
1486
1487 default:
1488 /*
1489 * other cases have been filtered in the above.
1490 * none will visit this case. here we supply
1491 * the code just in case (nxt overwritten or
1492 * other cases).
1493 */
1494 goto loopend;
1495 }
1496
1497 /* proceed with the next header. */
1498 off += elen;
1499 nxt = ip6e->ip6e_nxt;
1500 ip6e = NULL;
1501 }
1502 loopend:
1503 ;
1504 }
1505
1506 if (inp->inp_flags2 & INP_RECVFLOWID) {
1507 uint32_t flowid, flow_type;
1508
1509 flowid = m->m_pkthdr.flowid;
1510 flow_type = M_HASHTYPE_GET(m);
1511
1512 /*
1513 * XXX should handle the failure of one or the
1514 * other - don't populate both?
1515 */
1516 *mp = sbcreatecontrol((caddr_t) &flowid,
1517 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6);
1518 if (*mp)
1519 mp = &(*mp)->m_next;
1520 *mp = sbcreatecontrol((caddr_t) &flow_type,
1521 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6);
1522 if (*mp)
1523 mp = &(*mp)->m_next;
1524 }
1525
1526 #ifdef RSS
1527 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1528 uint32_t flowid, flow_type;
1529 uint32_t rss_bucketid;
1530
1531 flowid = m->m_pkthdr.flowid;
1532 flow_type = M_HASHTYPE_GET(m);
1533
1534 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1535 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1536 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6);
1537 if (*mp)
1538 mp = &(*mp)->m_next;
1539 }
1540 }
1541 #endif
1542
1543 }
1544 #undef IS2292
1545
1546 void
1547 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1548 {
1549 struct socket *so;
1550 struct mbuf *m_mtu;
1551 struct ip6_mtuinfo mtuctl;
1552
1553 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1554 /*
1555 * Notify the error by sending IPV6_PATHMTU ancillary data if
1556 * application wanted to know the MTU value.
1557 * NOTE: we notify disconnected sockets, because some udp
1558 * applications keep sending sockets disconnected.
1559 * NOTE: our implementation doesn't notify connected sockets that has
1560 * foreign address that is different than given destination addresses
1561 * (this is permitted by RFC 3542).
1562 */
1563 if ((inp->inp_flags & IN6P_MTU) == 0 || (
1564 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1565 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1566 return;
1567
1568 mtuctl.ip6m_mtu = mtu;
1569 mtuctl.ip6m_addr = *dst;
1570 if (sa6_recoverscope(&mtuctl.ip6m_addr))
1571 return;
1572
1573 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1574 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1575 return;
1576
1577 so = inp->inp_socket;
1578 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1579 == 0) {
1580 m_freem(m_mtu);
1581 /* XXX: should count statistics */
1582 } else
1583 sorwakeup(so);
1584 }
1585
1586 /*
1587 * Get pointer to the previous header followed by the header
1588 * currently processed.
1589 */
1590 int
1591 ip6_get_prevhdr(const struct mbuf *m, int off)
1592 {
1593 struct ip6_ext ip6e;
1594 struct ip6_hdr *ip6;
1595 int len, nlen, nxt;
1596
1597 if (off == sizeof(struct ip6_hdr))
1598 return (offsetof(struct ip6_hdr, ip6_nxt));
1599 if (off < sizeof(struct ip6_hdr))
1600 panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1601
1602 ip6 = mtod(m, struct ip6_hdr *);
1603 nxt = ip6->ip6_nxt;
1604 len = sizeof(struct ip6_hdr);
1605 nlen = 0;
1606 while (len < off) {
1607 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1608 switch (nxt) {
1609 case IPPROTO_FRAGMENT:
1610 nlen = sizeof(struct ip6_frag);
1611 break;
1612 case IPPROTO_AH:
1613 nlen = (ip6e.ip6e_len + 2) << 2;
1614 break;
1615 default:
1616 nlen = (ip6e.ip6e_len + 1) << 3;
1617 }
1618 len += nlen;
1619 nxt = ip6e.ip6e_nxt;
1620 }
1621 return (len - nlen);
1622 }
1623
1624 /*
1625 * get next header offset. m will be retained.
1626 */
1627 int
1628 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1629 {
1630 struct ip6_hdr ip6;
1631 struct ip6_ext ip6e;
1632 struct ip6_frag fh;
1633
1634 /* just in case */
1635 if (m == NULL)
1636 panic("ip6_nexthdr: m == NULL");
1637 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1638 return -1;
1639
1640 switch (proto) {
1641 case IPPROTO_IPV6:
1642 if (m->m_pkthdr.len < off + sizeof(ip6))
1643 return -1;
1644 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1645 if (nxtp)
1646 *nxtp = ip6.ip6_nxt;
1647 off += sizeof(ip6);
1648 return off;
1649
1650 case IPPROTO_FRAGMENT:
1651 /*
1652 * terminate parsing if it is not the first fragment,
1653 * it does not make sense to parse through it.
1654 */
1655 if (m->m_pkthdr.len < off + sizeof(fh))
1656 return -1;
1657 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1658 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1659 if (fh.ip6f_offlg & IP6F_OFF_MASK)
1660 return -1;
1661 if (nxtp)
1662 *nxtp = fh.ip6f_nxt;
1663 off += sizeof(struct ip6_frag);
1664 return off;
1665
1666 case IPPROTO_AH:
1667 if (m->m_pkthdr.len < off + sizeof(ip6e))
1668 return -1;
1669 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1670 if (nxtp)
1671 *nxtp = ip6e.ip6e_nxt;
1672 off += (ip6e.ip6e_len + 2) << 2;
1673 return off;
1674
1675 case IPPROTO_HOPOPTS:
1676 case IPPROTO_ROUTING:
1677 case IPPROTO_DSTOPTS:
1678 if (m->m_pkthdr.len < off + sizeof(ip6e))
1679 return -1;
1680 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1681 if (nxtp)
1682 *nxtp = ip6e.ip6e_nxt;
1683 off += (ip6e.ip6e_len + 1) << 3;
1684 return off;
1685
1686 case IPPROTO_NONE:
1687 case IPPROTO_ESP:
1688 case IPPROTO_IPCOMP:
1689 /* give up */
1690 return -1;
1691
1692 default:
1693 return -1;
1694 }
1695
1696 /* NOTREACHED */
1697 }
1698
1699 /*
1700 * get offset for the last header in the chain. m will be kept untainted.
1701 */
1702 int
1703 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1704 {
1705 int newoff;
1706 int nxt;
1707
1708 if (!nxtp) {
1709 nxt = -1;
1710 nxtp = &nxt;
1711 }
1712 while (1) {
1713 newoff = ip6_nexthdr(m, off, proto, nxtp);
1714 if (newoff < 0)
1715 return off;
1716 else if (newoff < off)
1717 return -1; /* invalid */
1718 else if (newoff == off)
1719 return newoff;
1720
1721 off = newoff;
1722 proto = *nxtp;
1723 }
1724 }
1725
1726 /*
1727 * System control for IP6
1728 */
1729
1730 u_char inet6ctlerrmap[PRC_NCMDS] = {
1731 0, 0, 0, 0,
1732 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1733 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1734 EMSGSIZE, EHOSTUNREACH, 0, 0,
1735 0, 0, EHOSTUNREACH, 0,
1736 ENOPROTOOPT, ECONNREFUSED
1737 };
Cache object: 6ffc45935ea615ab30ed7af2d0dac427
|