1 /*
2 * Copyright (c) 2003-2025 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29 /*
30 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
31 * All rights reserved.
32 *
33 * Redistribution and use in source and binary forms, with or without
34 * modification, are permitted provided that the following conditions
35 * are met:
36 * 1. Redistributions of source code must retain the above copyright
37 * notice, this list of conditions and the following disclaimer.
38 * 2. Redistributions in binary form must reproduce the above copyright
39 * notice, this list of conditions and the following disclaimer in the
40 * documentation and/or other materials provided with the distribution.
41 * 3. Neither the name of the project nor the names of its contributors
42 * may be used to endorse or promote products derived from this software
43 * without specific prior written permission.
44 *
45 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
46 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
49 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
55 * SUCH DAMAGE.
56 */
57
58 /*
59 * Copyright (c) 1982, 1986, 1988, 1993
60 * The Regents of the University of California. All rights reserved.
61 *
62 * Redistribution and use in source and binary forms, with or without
63 * modification, are permitted provided that the following conditions
64 * are met:
65 * 1. Redistributions of source code must retain the above copyright
66 * notice, this list of conditions and the following disclaimer.
67 * 2. Redistributions in binary form must reproduce the above copyright
68 * notice, this list of conditions and the following disclaimer in the
69 * documentation and/or other materials provided with the distribution.
70 * 3. All advertising materials mentioning features or use of this software
71 * must display the following acknowledgement:
72 * This product includes software developed by the University of
73 * California, Berkeley and its contributors.
74 * 4. Neither the name of the University nor the names of its contributors
75 * may be used to endorse or promote products derived from this software
76 * without specific prior written permission.
77 *
78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88 * SUCH DAMAGE.
89 *
90 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
91 */
92
93 #include <sys/param.h>
94 #include <sys/systm.h>
95 #include <sys/malloc.h>
96 #include <sys/mbuf.h>
97 #include <sys/domain.h>
98 #include <sys/protosw.h>
99 #include <sys/socket.h>
100 #include <sys/socketvar.h>
101 #include <sys/errno.h>
102 #include <sys/time.h>
103 #include <sys/kernel.h>
104 #include <sys/syslog.h>
105 #include <sys/sysctl.h>
106 #include <sys/proc.h>
107 #include <sys/kauth.h>
108 #include <sys/mcache.h>
109
110 #include <mach/mach_time.h>
111 #include <mach/sdt.h>
112 #include <pexpert/pexpert.h>
113 #include <dev/random/randomdev.h>
114
115 #include <net/if.h>
116 #include <net/if_var.h>
117 #include <net/if_types.h>
118 #include <net/if_dl.h>
119 #include <net/route.h>
120 #include <net/kpi_protocol.h>
121 #include <net/ntstat.h>
122 #include <net/init.h>
123 #include <net/net_osdep.h>
124 #include <net/net_perf.h>
125 #include <net/if_ports_used.h>
126 #include <net/droptap.h>
127
128 #include <netinet/in.h>
129 #include <netinet/in_systm.h>
130 #if INET
131 #include <netinet/ip.h>
132 #include <netinet/ip_icmp.h>
133 #endif /* INET */
134 #include <netinet/kpi_ipfilter_var.h>
135 #include <netinet/ip6.h>
136 #include <netinet/udp.h>
137 #include <netinet6/in6_var.h>
138 #include <netinet6/ip6_var.h>
139 #include <netinet/in_pcb.h>
140 #include <netinet/icmp6.h>
141 #include <netinet6/in6_ifattach.h>
142 #include <netinet6/nd6.h>
143 #include <netinet6/scope6_var.h>
144 #include <netinet6/ip6protosw.h>
145
146 #if IPSEC
147 #include <netinet6/ipsec.h>
148 #include <netinet6/ipsec6.h>
149 extern int ipsec_bypass;
150 #endif /* IPSEC */
151
152 #if DUMMYNET
153 #include <netinet/ip_dummynet.h>
154 #endif /* DUMMYNET */
155
156 /* we need it for NLOOP. */
157 #include "loop.h"
158
159 #if PF
160 #include <net/pfvar.h>
161 #endif /* PF */
162
163 #include <os/log.h>
164
165 struct ip6protosw *ip6_protox[IPPROTO_MAX];
166
167 static LCK_GRP_DECLARE(in6_ifaddr_rwlock_grp, "in6_ifaddr_rwlock");
168 LCK_RW_DECLARE(in6_ifaddr_rwlock, &in6_ifaddr_rwlock_grp);
169
170 /* Protected by in6_ifaddr_rwlock */
171 struct in6_ifaddrhead in6_ifaddrhead;
172 uint32_t in6_ifaddrlist_genid = 0;
173
174 #define IN6ADDR_NHASH 61
175 u_int32_t in6addr_hashp = 0; /* next largest prime */
176 u_int32_t in6addr_nhash = 0; /* hash table size */
177 struct in6_ifaddrhashhead *__counted_by(in6addr_nhash) in6_ifaddrhashtbl = 0;
178
179 #define IN6_IFSTAT_REQUIRE_ALIGNED_64(f) \
180 _CASSERT(!(offsetof(struct in6_ifstat, f) % sizeof (uint64_t)))
181
182 #define ICMP6_IFSTAT_REQUIRE_ALIGNED_64(f) \
183 _CASSERT(!(offsetof(struct icmp6_ifstat, f) % sizeof (uint64_t)))
184
185 struct ip6stat ip6stat;
186
187 LCK_ATTR_DECLARE(ip6_mutex_attr, 0, 0);
188 LCK_GRP_DECLARE(ip6_mutex_grp, "ip6");
189
190 LCK_MTX_DECLARE_ATTR(proxy6_lock, &ip6_mutex_grp, &ip6_mutex_attr);
191 LCK_MTX_DECLARE_ATTR(nd6_mutex_data, &ip6_mutex_grp, &ip6_mutex_attr);
192
193 extern int loopattach_done;
194 extern void addrsel_policy_init(void);
195
196 static int sysctl_reset_ip6_input_stats SYSCTL_HANDLER_ARGS;
197 static int sysctl_ip6_input_measure_bins SYSCTL_HANDLER_ARGS;
198 static int sysctl_ip6_input_getperf SYSCTL_HANDLER_ARGS;
199 static void ip6_init_delayed(void);
200 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
201
202 static void in6_ifaddrhashtbl_init(void);
203
204 static struct m_tag *m_tag_kalloc_inet6(u_int32_t id, u_int16_t type, uint16_t len, int wait);
205 static void m_tag_kfree_inet6(struct m_tag *tag);
206
207 #if NSTF
208 extern void stfattach(void);
209 #endif /* NSTF */
210
211 SYSCTL_DECL(_net_inet6_ip6);
212
213 static uint32_t ip6_adj_clear_hwcksum = 0;
214 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, adj_clear_hwcksum,
215 CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_adj_clear_hwcksum, 0,
216 "Invalidate hwcksum info when adjusting length");
217
218 static uint32_t ip6_adj_partial_sum = 1;
219 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, adj_partial_sum,
220 CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_adj_partial_sum, 0,
221 "Perform partial sum adjustment of trailing bytes at IP layer");
222
223 static int ip6_input_measure = 0;
224 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, input_perf,
225 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
226 &ip6_input_measure, 0, sysctl_reset_ip6_input_stats, "I", "Do time measurement");
227
228 static uint64_t ip6_input_measure_bins = 0;
229 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, input_perf_bins,
230 CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_input_measure_bins, 0,
231 sysctl_ip6_input_measure_bins, "I",
232 "bins for chaining performance data histogram");
233
234 static net_perf_t net_perf;
235 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, input_perf_data,
236 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
237 0, 0, sysctl_ip6_input_getperf, "S,net_perf",
238 "IP6 input performance data (struct net_perf, net/net_perf.h)");
239
240 /*
241 * ip6_checkinterface controls the receive side of the models for multihoming
242 * that are discussed in RFC 1122.
243 *
244 * sysctl_ip6_checkinterface values are:
245 * IP6_CHECKINTERFACE_WEAK_ES:
246 * This corresponds to the Weak End-System model where incoming packets from
247 * any interface are accepted provided the destination address of the incoming packet
248 * is assigned to some interface.
249 *
250 * IP6_CHECKINTERFACE_HYBRID_ES:
251 * The Hybrid End-System model use the Strong End-System for tunnel interfaces
252 * (ipsec and utun) and the weak End-System model for other interfaces families.
253 * This prevents a rogue middle box to probe for signs of TCP connections
254 * that use the tunnel interface.
255 *
256 * IP6_CHECKINTERFACE_STRONG_ES:
257 * The Strong model model requires the packet arrived on an interface that
258 * is assigned the destination address of the packet.
259 *
260 * Since the routing table and transmit implementation do not implement the Strong ES model,
261 * setting this to a value different from IP6_CHECKINTERFACE_WEAK_ES may lead to unexpected results.
262 *
263 * When forwarding is enabled, the system reverts to the Weak ES model as a router
264 * is expected by design to receive packets from several interfaces to the same address.
265 */
266 #define IP6_CHECKINTERFACE_WEAK_ES 0
267 #define IP6_CHECKINTERFACE_HYBRID_ES 1
268 #define IP6_CHECKINTERFACE_STRONG_ES 2
269
270 static int ip6_checkinterface = IP6_CHECKINTERFACE_HYBRID_ES;
271
272 static int sysctl_ip6_checkinterface SYSCTL_HANDLER_ARGS;
273 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, check_interface,
274 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
275 0, 0, sysctl_ip6_checkinterface, "I", "Verify packet arrives on correct interface");
276
277 #if (DEBUG || DEVELOPMENT)
278 #define IP6_CHECK_IFDEBUG 1
279 #else
280 #define IP6_CHECK_IFDEBUG 0
281 #endif /* (DEBUG || DEVELOPMENT) */
282 static int ip6_checkinterface_debug = IP6_CHECK_IFDEBUG;
283 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, checkinterface_debug, CTLFLAG_RW | CTLFLAG_LOCKED,
284 &ip6_checkinterface_debug, IP6_CHECK_IFDEBUG, "");
285
286 typedef enum ip6_check_if_result {
287 IP6_CHECK_IF_NONE = 0,
288 IP6_CHECK_IF_OURS = 1,
289 IP6_CHECK_IF_DROP = 2,
290 IP6_CHECK_IF_FORWARD = 3
291 } ip6_check_if_result_t;
292
293 static ip6_check_if_result_t ip6_input_check_interface(struct mbuf *, struct ip6_hdr *, struct ifnet *, struct route_in6 *rin6, struct ifnet **);
294
295 /*
296 * On platforms which require strict alignment (currently for anything but
297 * i386 or x86_64 or arm64), check if the IP header pointer is 32-bit aligned; if not,
298 * copy the contents of the mbuf chain into a new chain, and free the original
299 * one. Create some head room in the first mbuf of the new chain, in case
300 * it's needed later on.
301 *
302 * RFC 2460 says that IPv6 headers are 64-bit aligned, but network interfaces
303 * mostly align to 32-bit boundaries. Care should be taken never to use 64-bit
304 * load/store operations on the fields in IPv6 headers.
305 */
306 #if defined(__i386__) || defined(__x86_64__) || defined(__arm64__)
307 #define IP6_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { } while (0)
308 #else /* !__i386__ && !__x86_64__ && !__arm64__ */
309 #define IP6_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { \
310 if (!IP6_HDR_ALIGNED_P(mtod(_m, caddr_t))) { \
311 mbuf_ref_t _n; \
312 ifnet_ref_t __ifp = (_ifp); \
313 os_atomic_inc(&(__ifp)->if_alignerrs, relaxed); \
314 if (((_m)->m_flags & M_PKTHDR) && \
315 (_m)->m_pkthdr.pkt_hdr != NULL) \
316 (_m)->m_pkthdr.pkt_hdr = NULL; \
317 _n = m_defrag_offset(_m, max_linkhdr, M_NOWAIT); \
318 if (_n == NULL) { \
319 ip6stat.ip6s_toosmall++; \
320 m_drop(_m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0); \
321 (_m) = NULL; \
322 _action; \
323 } else { \
324 VERIFY(_n != (_m)); \
325 (_m) = _n; \
326 } \
327 } \
328 } while (0)
329 #endif /* !__i386__ && !__x86_64___ && !__arm64__ */
330
331 static void
ip6_proto_input(protocol_family_t protocol,mbuf_t packet)332 ip6_proto_input(protocol_family_t protocol, mbuf_t packet)
333 {
334 #pragma unused(protocol)
335 #if INET
336 struct timeval start_tv;
337 if (ip6_input_measure) {
338 net_perf_start_time(&net_perf, &start_tv);
339 }
340 #endif /* INET */
341 ip6_input(packet);
342 #if INET
343 if (ip6_input_measure) {
344 net_perf_measure_time(&net_perf, &start_tv, 1);
345 net_perf_histogram(&net_perf, 1);
346 }
347 #endif /* INET */
348 }
349
350 /*
351 * IP6 initialization: fill in IP6 protocol switch table.
352 * All protocols not implemented in kernel go to raw IP6 protocol handler.
353 */
354 void
ip6_init(struct ip6protosw * pp,struct domain * dp)355 ip6_init(struct ip6protosw *pp, struct domain *dp)
356 {
357 static int ip6_initialized = 0;
358 struct protosw *__single pr;
359 struct timeval tv;
360 int i;
361 domain_unguard_t __single unguard;
362
363 domain_proto_mtx_lock_assert_held();
364 VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
365
366 _CASSERT((sizeof(struct ip6_hdr) +
367 sizeof(struct icmp6_hdr)) <= _MHLEN);
368
369 if (ip6_initialized) {
370 return;
371 }
372 ip6_initialized = 1;
373
374 eventhandler_lists_ctxt_init(&in6_evhdlr_ctxt);
375 (void)EVENTHANDLER_REGISTER(&in6_evhdlr_ctxt, in6_event,
376 &in6_eventhdlr_callback, eventhandler_entry_dummy_arg,
377 EVENTHANDLER_PRI_ANY);
378
379 eventhandler_lists_ctxt_init(&in6_clat46_evhdlr_ctxt);
380 (void)EVENTHANDLER_REGISTER(&in6_clat46_evhdlr_ctxt, in6_clat46_event,
381 &in6_clat46_eventhdlr_callback, eventhandler_entry_dummy_arg,
382 EVENTHANDLER_PRI_ANY);
383
384 for (i = 0; i < IN6_EVENT_MAX; i++) {
385 VERIFY(in6_event2kev_array[i].in6_event_code == i);
386 }
387
388 pr = pffindproto_locked(PF_INET6, IPPROTO_RAW, SOCK_RAW);
389 if (pr == NULL) {
390 panic("%s: Unable to find [PF_INET6,IPPROTO_RAW,SOCK_RAW]",
391 __func__);
392 /* NOTREACHED */
393 }
394
395 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
396 for (i = 0; i < IPPROTO_MAX; i++) {
397 ip6_protox[i] = (struct ip6protosw *)pr;
398 }
399 /*
400 * Cycle through IP protocols and put them into the appropriate place
401 * in ip6_protox[], skipping protocols IPPROTO_{IP,RAW}.
402 */
403 VERIFY(dp == inet6domain && dp->dom_family == PF_INET6);
404 TAILQ_FOREACH(pr, &dp->dom_protosw, pr_entry) {
405 VERIFY(pr->pr_domain == dp);
406 if (pr->pr_protocol != 0 && pr->pr_protocol != IPPROTO_RAW) {
407 /* Be careful to only index valid IP protocols. */
408 if (pr->pr_protocol < IPPROTO_MAX) {
409 ip6_protox[pr->pr_protocol] =
410 (struct ip6protosw *)pr;
411 }
412 }
413 }
414
415 TAILQ_INIT(&in6_ifaddrhead);
416 in6_ifaddrhashtbl_init();
417
418 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_receive);
419 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_hdrerr);
420 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_toobig);
421 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_noroute);
422 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_addrerr);
423 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_protounknown);
424 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_truncated);
425 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_discard);
426 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_deliver);
427 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_forward);
428 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_request);
429 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_discard);
430 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragok);
431 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragfail);
432 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragcreat);
433 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_reqd);
434 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_ok);
435 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_fail);
436 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mcast);
437 IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mcast);
438
439 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_msg);
440 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_error);
441 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_dstunreach);
442 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_adminprohib);
443 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_timeexceed);
444 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_paramprob);
445 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_pkttoobig);
446 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_echo);
447 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_echoreply);
448 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_routersolicit);
449 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_routeradvert);
450 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_neighborsolicit);
451 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_neighboradvert);
452 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_redirect);
453 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mldquery);
454 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mldreport);
455 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mlddone);
456
457 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_msg);
458 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_error);
459 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_dstunreach);
460 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_adminprohib);
461 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_timeexceed);
462 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_paramprob);
463 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_pkttoobig);
464 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_echo);
465 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_echoreply);
466 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_routersolicit);
467 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_routeradvert);
468 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_neighborsolicit);
469 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_neighboradvert);
470 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_redirect);
471 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mldquery);
472 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mldreport);
473 ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mlddone);
474
475 getmicrotime(&tv);
476 ip6_desync_factor =
477 (RandomULong() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
478
479 PE_parse_boot_argn("in6_embedded_scope", &in6_embedded_scope, sizeof(in6_embedded_scope));
480 PE_parse_boot_argn("ip6_checkinterface", &i, sizeof(i));
481 switch (i) {
482 case IP6_CHECKINTERFACE_WEAK_ES:
483 case IP6_CHECKINTERFACE_HYBRID_ES:
484 case IP6_CHECKINTERFACE_STRONG_ES:
485 ip6_checkinterface = i;
486 break;
487 default:
488 break;
489 }
490
491 in6_ifaddr_init();
492 ip6_moptions_init();
493 nd6_init();
494 frag6_init();
495 icmp6_init(NULL, dp);
496 addrsel_policy_init();
497
498 /*
499 * P2P interfaces often route the local address to the loopback
500 * interface. At this point, lo0 hasn't been initialized yet, which
501 * means that we need to delay the IPv6 configuration of lo0.
502 */
503 net_init_add(ip6_init_delayed);
504
505 unguard = domain_unguard_deploy();
506 i = proto_register_input(PF_INET6, ip6_proto_input, NULL, 0);
507 if (i != 0) {
508 panic("%s: failed to register PF_INET6 protocol: %d",
509 __func__, i);
510 /* NOTREACHED */
511 }
512 domain_unguard_release(unguard);
513 }
514
515 static void
ip6_init_delayed(void)516 ip6_init_delayed(void)
517 {
518 (void) in6_ifattach_prelim(lo_ifp);
519
520 /* timer for regeneranation of temporary addresses randomize ID */
521 timeout(in6_tmpaddrtimer, NULL,
522 (ip6_temp_preferred_lifetime - ip6_desync_factor -
523 ip6_temp_regen_advance) * hz);
524
525 #if NSTF
526 stfattach();
527 #endif /* NSTF */
528 }
529
530 static void
ip6_input_adjust(struct mbuf * m,struct ip6_hdr * ip6,uint32_t plen,struct ifnet * inifp)531 ip6_input_adjust(struct mbuf *m, struct ip6_hdr *ip6, uint32_t plen,
532 struct ifnet *inifp)
533 {
534 boolean_t adjust = TRUE;
535 uint32_t tot_len = sizeof(*ip6) + plen;
536
537 ASSERT(m_pktlen(m) > tot_len);
538
539 /*
540 * Invalidate hardware checksum info if ip6_adj_clear_hwcksum
541 * is set; useful to handle buggy drivers. Note that this
542 * should not be enabled by default, as we may get here due
543 * to link-layer padding.
544 */
545 if (ip6_adj_clear_hwcksum &&
546 (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) &&
547 !(inifp->if_flags & IFF_LOOPBACK) &&
548 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
549 m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID;
550 m->m_pkthdr.csum_data = 0;
551 ip6stat.ip6s_adj_hwcsum_clr++;
552 }
553
554 /*
555 * If partial checksum information is available, subtract
556 * out the partial sum of postpended extraneous bytes, and
557 * update the checksum metadata accordingly. By doing it
558 * here, the upper layer transport only needs to adjust any
559 * prepended extraneous bytes (else it will do both.)
560 */
561 if (ip6_adj_partial_sum &&
562 (m->m_pkthdr.csum_flags & (CSUM_DATA_VALID | CSUM_PARTIAL)) ==
563 (CSUM_DATA_VALID | CSUM_PARTIAL)) {
564 m->m_pkthdr.csum_rx_val = m_adj_sum16(m,
565 m->m_pkthdr.csum_rx_start, m->m_pkthdr.csum_rx_start,
566 (tot_len - m->m_pkthdr.csum_rx_start),
567 m->m_pkthdr.csum_rx_val);
568 } else if ((m->m_pkthdr.csum_flags &
569 (CSUM_DATA_VALID | CSUM_PARTIAL)) ==
570 (CSUM_DATA_VALID | CSUM_PARTIAL)) {
571 /*
572 * If packet has partial checksum info and we decided not
573 * to subtract the partial sum of postpended extraneous
574 * bytes here (not the default case), leave that work to
575 * be handled by the other layers. For now, only TCP, UDP
576 * layers are capable of dealing with this. For all other
577 * protocols (including fragments), trim and ditch the
578 * partial sum as those layers might not implement partial
579 * checksumming (or adjustment) at all.
580 */
581 if (ip6->ip6_nxt == IPPROTO_TCP ||
582 ip6->ip6_nxt == IPPROTO_UDP) {
583 adjust = FALSE;
584 } else {
585 m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID;
586 m->m_pkthdr.csum_data = 0;
587 ip6stat.ip6s_adj_hwcsum_clr++;
588 }
589 }
590
591 if (adjust) {
592 ip6stat.ip6s_adj++;
593 if (m->m_len == m->m_pkthdr.len) {
594 m->m_len = tot_len;
595 m->m_pkthdr.len = tot_len;
596 } else {
597 m_adj(m, tot_len - m->m_pkthdr.len);
598 }
599 }
600 }
601
602 static ip6_check_if_result_t
ip6_input_check_interface(struct mbuf * m,struct ip6_hdr * ip6,struct ifnet * inifp,struct route_in6 * rin6,struct ifnet ** deliverifp)603 ip6_input_check_interface(struct mbuf *m, struct ip6_hdr *ip6, struct ifnet *inifp, struct route_in6 *rin6, struct ifnet **deliverifp)
604 {
605 struct in6_ifaddr *__single ia6 = NULL;
606 struct in6_addr tmp_dst = ip6->ip6_dst; /* copy to avoid unaligned access */
607 struct in6_ifaddr *__single best_ia6 = NULL;
608 uint32_t dst_ifscope = IFSCOPE_NONE;
609 ip6_check_if_result_t result = IP6_CHECK_IF_NONE;
610
611 *deliverifp = NULL;
612
613 if (m->m_pkthdr.pkt_flags & PKTF_IFAINFO) {
614 dst_ifscope = m->m_pkthdr.dst_ifindex;
615 } else {
616 dst_ifscope = inifp->if_index;
617 }
618 /*
619 * Check for exact addresses in the hash bucket.
620 */
621 lck_rw_lock_shared(&in6_ifaddr_rwlock);
622 TAILQ_FOREACH(ia6, IN6ADDR_HASH(&tmp_dst), ia6_hash) {
623 /*
624 * TODO: should we accept loopback
625 */
626 if (in6_are_addr_equal_scoped(&ia6->ia_addr.sin6_addr, &tmp_dst, ia6->ia_ifp->if_index, dst_ifscope)) {
627 if ((ia6->ia6_flags & IN6_IFF_NOTREADY) != 0) {
628 continue;
629 }
630 best_ia6 = ia6;
631 if (ia6->ia_ifp == inifp) {
632 /*
633 * TODO: should we also accept locally originated packets
634 * or from loopback ???
635 */
636 break;
637 }
638 /*
639 * Continue the loop in case there's a exact match with another
640 * interface
641 */
642 }
643 }
644 if (best_ia6 != NULL) {
645 if (best_ia6->ia_ifp != inifp && ip6_forwarding == 0 &&
646 ((ip6_checkinterface == IP6_CHECKINTERFACE_HYBRID_ES &&
647 (best_ia6->ia_ifp->if_family == IFNET_FAMILY_IPSEC ||
648 best_ia6->ia_ifp->if_family == IFNET_FAMILY_UTUN)) ||
649 ip6_checkinterface == IP6_CHECKINTERFACE_STRONG_ES)) {
650 /*
651 * Drop when interface address check is strict and forwarding
652 * is disabled
653 */
654 result = IP6_CHECK_IF_DROP;
655 } else {
656 result = IP6_CHECK_IF_OURS;
657 *deliverifp = best_ia6->ia_ifp;
658 ip6_setdstifaddr_info(m, 0, best_ia6);
659 ip6_setsrcifaddr_info(m, best_ia6->ia_ifp->if_index, NULL);
660 }
661 }
662 lck_rw_done(&in6_ifaddr_rwlock);
663
664 if (result == IP6_CHECK_IF_NONE) {
665 /*
666 * Slow path: route lookup.
667 */
668 struct sockaddr_in6 *__single dst6;
669
670 dst6 = SIN6(&rin6->ro_dst);
671 dst6->sin6_len = sizeof(struct sockaddr_in6);
672 dst6->sin6_family = AF_INET6;
673 dst6->sin6_addr = ip6->ip6_dst;
674 if (!in6_embedded_scope && IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
675 dst6->sin6_scope_id = dst_ifscope;
676 }
677 rtalloc_scoped_ign((struct route *)rin6,
678 RTF_PRCLONING, IFSCOPE_NONE);
679 if (rin6->ro_rt != NULL) {
680 RT_LOCK_SPIN(rin6->ro_rt);
681 }
682
683 #define rt6_key(r) (SIN6(rn_get_key((r)->rt_nodes)))
684
685 /*
686 * Accept the packet if the forwarding interface to the destination
687 * according to the routing table is the loopback interface,
688 * unless the associated route has a gateway.
689 * Note that this approach causes to accept a packet if there is a
690 * route to the loopback interface for the destination of the packet.
691 * But we think it's even useful in some situations, e.g. when using
692 * a special daemon which wants to intercept the packet.
693 *
694 * XXX: some OSes automatically make a cloned route for the destination
695 * of an outgoing packet. If the outgoing interface of the packet
696 * is a loopback one, the kernel would consider the packet to be
697 * accepted, even if we have no such address assinged on the interface.
698 * We check the cloned flag of the route entry to reject such cases,
699 * assuming that route entries for our own addresses are not made by
700 * cloning (it should be true because in6_addloop explicitly installs
701 * the host route). However, we might have to do an explicit check
702 * while it would be less efficient. Or, should we rather install a
703 * reject route for such a case?
704 */
705 if (rin6->ro_rt != NULL &&
706 (rin6->ro_rt->rt_flags & (RTF_HOST | RTF_GATEWAY)) == RTF_HOST &&
707 #if RTF_WASCLONED
708 !(rin6->ro_rt->rt_flags & RTF_WASCLONED) &&
709 #endif
710 rin6->ro_rt->rt_ifp->if_type == IFT_LOOP) {
711 ia6 = ifatoia6(rin6->ro_rt->rt_ifa);
712 /*
713 * Packets to a tentative, duplicated, or somehow invalid
714 * address must not be accepted.
715 *
716 * For performance, test without acquiring the address lock;
717 * a lot of things in the address are set once and never
718 * changed (e.g. ia_ifp.)
719 */
720 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
721 /* this address is ready */
722 result = IP6_CHECK_IF_OURS;
723 *deliverifp = ia6->ia_ifp; /* correct? */
724 /*
725 * record dst address information into mbuf.
726 */
727 (void) ip6_setdstifaddr_info(m, 0, ia6);
728 (void) ip6_setsrcifaddr_info(m, ia6->ia_ifp->if_index, NULL);
729 }
730 }
731
732 if (rin6->ro_rt != NULL) {
733 RT_UNLOCK(rin6->ro_rt);
734 }
735 }
736
737 if (result == IP6_CHECK_IF_NONE) {
738 if (ip6_forwarding == 0) {
739 result = IP6_CHECK_IF_DROP;
740 } else {
741 result = IP6_CHECK_IF_FORWARD;
742 ip6_setdstifaddr_info(m, inifp->if_index, NULL);
743 ip6_setsrcifaddr_info(m, inifp->if_index, NULL);
744 }
745 }
746
747 if (result == IP6_CHECK_IF_OURS && *deliverifp != inifp) {
748 ASSERT(*deliverifp != NULL);
749 ip6stat.ip6s_rcv_if_weak_match++;
750
751 /* Logging is too noisy when forwarding is enabled */
752 if (ip6_checkinterface_debug != IP6_CHECKINTERFACE_WEAK_ES && ip6_forwarding != 0) {
753 char src_str[MAX_IPv6_STR_LEN];
754 char dst_str[MAX_IPv6_STR_LEN];
755
756 inet_ntop(AF_INET6, &ip6->ip6_src, src_str, sizeof(src_str));
757 inet_ntop(AF_INET6, &ip6->ip6_dst, dst_str, sizeof(dst_str));
758 os_log_info(OS_LOG_DEFAULT,
759 "%s: weak ES interface match to %s for packet from %s to %s proto %u received via %s",
760 __func__, (*deliverifp)->if_xname, src_str, dst_str, ip6->ip6_nxt, inifp->if_xname);
761 }
762 } else if (result == IP6_CHECK_IF_DROP) {
763 ip6stat.ip6s_rcv_if_no_match++;
764 if (ip6_checkinterface_debug > 0) {
765 char src_str[MAX_IPv6_STR_LEN];
766 char dst_str[MAX_IPv6_STR_LEN];
767
768 inet_ntop(AF_INET6, &ip6->ip6_src, src_str, sizeof(src_str));
769 inet_ntop(AF_INET6, &ip6->ip6_dst, dst_str, sizeof(dst_str));
770 os_log(OS_LOG_DEFAULT,
771 "%s: no interface match for packet from %s to %s proto %u received via %s",
772 __func__, src_str, dst_str, ip6->ip6_nxt, inifp->if_xname);
773 }
774 }
775
776 return result;
777 }
778
779 void
ip6_input(struct mbuf * m)780 ip6_input(struct mbuf *m)
781 {
782 struct ip6_hdr *ip6;
783 int off = sizeof(struct ip6_hdr), nest;
784 u_int32_t plen;
785 u_int32_t rtalert = ~0;
786 int nxt = 0, ours = 0;
787 ifnet_ref_t inifp, deliverifp = NULL;
788 ipfilter_t __single inject_ipfref = NULL;
789 int seen = 1;
790 #if DUMMYNET
791 struct m_tag *__single tag;
792 struct ip_fw_args args = {};
793 #endif /* DUMMYNET */
794 struct route_in6 rin6 = {};
795
796 /*
797 * Check if the packet we received is valid after interface filter
798 * processing
799 */
800 MBUF_INPUT_CHECK(m, m->m_pkthdr.rcvif);
801 inifp = m->m_pkthdr.rcvif;
802 VERIFY(inifp != NULL);
803
804 /* Perform IP header alignment fixup, if needed */
805 IP6_HDR_ALIGNMENT_FIXUP(m, inifp, return );
806
807 m->m_pkthdr.pkt_flags &= ~PKTF_FORWARDED;
808 #if IPSEC
809 /*
810 * should the inner packet be considered authentic?
811 * see comment in ah4_input().
812 */
813 m->m_flags &= ~M_AUTHIPHDR;
814 m->m_flags &= ~M_AUTHIPDGM;
815 #endif /* IPSEC */
816
817 /*
818 * make sure we don't have onion peering information into m_aux.
819 */
820 ip6_delaux(m);
821
822 #if DUMMYNET
823 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
824 KERNEL_TAG_TYPE_DUMMYNET)) != NULL) {
825 struct dn_pkt_tag *__single dn_tag;
826
827 dn_tag = (struct dn_pkt_tag *)(tag->m_tag_data);
828
829 args.fwa_pf_rule = dn_tag->dn_pf_rule;
830
831 m_tag_delete(m, tag);
832 }
833
834 if (args.fwa_pf_rule) {
835 ip6 = mtod(m, struct ip6_hdr *); /* In case PF got disabled */
836
837 goto check_with_pf;
838 }
839 #endif /* DUMMYNET */
840
841 /*
842 * No need to process packet twice if we've already seen it.
843 */
844 inject_ipfref = ipf_get_inject_filter(m);
845 if (inject_ipfref != NULL) {
846 ip6 = mtod(m, struct ip6_hdr *);
847 nxt = ip6->ip6_nxt;
848 seen = 0;
849 goto injectit;
850 } else {
851 seen = 1;
852 }
853
854 if (__improbable(m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
855 if_ports_used_match_mbuf(inifp, PF_INET6, m);
856 }
857
858 /*
859 * mbuf statistics
860 */
861 if (m->m_flags & M_EXT) {
862 if (m->m_next != NULL) {
863 ip6stat.ip6s_mext2m++;
864 } else {
865 ip6stat.ip6s_mext1++;
866 }
867 } else {
868 #define M2MMAX (sizeof (ip6stat.ip6s_m2m) / sizeof (ip6stat.ip6s_m2m[0]))
869 if (m->m_next != NULL) {
870 if (m->m_pkthdr.pkt_flags & PKTF_LOOP) {
871 /* XXX */
872 ip6stat.ip6s_m2m[ifnet_index(lo_ifp)]++;
873 } else if (inifp->if_index < M2MMAX) {
874 ip6stat.ip6s_m2m[inifp->if_index]++;
875 } else {
876 ip6stat.ip6s_m2m[0]++;
877 }
878 } else {
879 ip6stat.ip6s_m1++;
880 }
881 #undef M2MMAX
882 }
883
884 /*
885 * Drop the packet if IPv6 operation is disabled on the interface.
886 */
887 if (inifp->if_eflags & IFEF_IPV6_DISABLED) {
888 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_IF_IPV6_DISABLED, NULL, 0);
889 goto bad;
890 }
891
892 in6_ifstat_inc_na(inifp, ifs6_in_receive);
893 ip6stat.ip6s_total++;
894
895 /*
896 * L2 bridge code and some other code can return mbuf chain
897 * that does not conform to KAME requirement. too bad.
898 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
899 */
900 if (m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
901 mbuf_ref_t n;
902
903 MGETHDR(n, M_DONTWAIT, MT_HEADER); /* MAC-OK */
904 if (n) {
905 M_COPY_PKTHDR(n, m);
906 }
907 if (n && m->m_pkthdr.len > MHLEN) {
908 MCLGET(n, M_DONTWAIT);
909 if ((n->m_flags & M_EXT) == 0) {
910 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0);
911 n = NULL;
912 }
913 }
914 if (n == NULL) {
915 goto bad;
916 }
917
918 m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
919 n->m_len = m->m_pkthdr.len;
920 m_freem(m);
921 m = n;
922 }
923 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), { goto done; });
924
925 if (m->m_len < sizeof(struct ip6_hdr)) {
926 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
927 ip6stat.ip6s_toosmall++;
928 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
929 goto done;
930 }
931 }
932
933 ip6 = mtod(m, struct ip6_hdr *);
934
935 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
936 ip6stat.ip6s_badvers++;
937 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
938 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_BAD_VERSION, NULL, 0);
939 goto bad;
940 }
941
942 ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
943
944 /*
945 * Check against address spoofing/corruption.
946 */
947 if (!(m->m_pkthdr.pkt_flags & PKTF_LOOP) &&
948 IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src)) {
949 ip6stat.ip6s_badscope++;
950 in6_ifstat_inc(inifp, ifs6_in_addrerr);
951 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
952 goto bad;
953 }
954 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
955 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
956 /*
957 * XXX: "badscope" is not very suitable for a multicast source.
958 */
959 ip6stat.ip6s_badscope++;
960 in6_ifstat_inc(inifp, ifs6_in_addrerr);
961 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
962 goto bad;
963 }
964 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
965 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
966 /*
967 * In this case, the packet should come from the loopback
968 * interface. However, we cannot just check the if_flags,
969 * because ip6_mloopback() passes the "actual" interface
970 * as the outgoing/incoming interface.
971 */
972 ip6stat.ip6s_badscope++;
973 in6_ifstat_inc(inifp, ifs6_in_addrerr);
974 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
975 goto bad;
976 }
977
978 /*
979 * The following check is not documented in specs. A malicious
980 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
981 * and bypass security checks (act as if it was from 127.0.0.1 by using
982 * IPv6 src ::ffff:127.0.0.1). Be cautious.
983 *
984 * This check chokes if we are in an SIIT cloud. As none of BSDs
985 * support IPv4-less kernel compilation, we cannot support SIIT
986 * environment at all. So, it makes more sense for us to reject any
987 * malicious packets for non-SIIT environment, than try to do a
988 * partial support for SIIT environment.
989 */
990 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
991 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
992 ip6stat.ip6s_badscope++;
993 in6_ifstat_inc(inifp, ifs6_in_addrerr);
994 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
995 goto bad;
996 }
997
998 if (((ntohl(ip6->ip6_flow & IPV6_FLOW_ECN_MASK) >> 20) & IPTOS_ECN_ECT1) == IPTOS_ECN_ECT1) {
999 m->m_pkthdr.pkt_ext_flags |= PKTF_EXT_L4S;
1000 }
1001
1002 #if 0
1003 /*
1004 * Reject packets with IPv4 compatible addresses (auto tunnel).
1005 *
1006 * The code forbids auto tunnel relay case in RFC1933 (the check is
1007 * stronger than RFC1933). We may want to re-enable it if mech-xx
1008 * is revised to forbid relaying case.
1009 */
1010 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
1011 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
1012 ip6stat.ip6s_badscope++;
1013 in6_ifstat_inc(inifp, ifs6_in_addrerr);
1014 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1015 goto bad;
1016 }
1017 #endif
1018
1019 /*
1020 * Naively assume we can attribute inbound data to the route we would
1021 * use to send to this destination. Asymetric routing breaks this
1022 * assumption, but it still allows us to account for traffic from
1023 * a remote node in the routing table.
1024 * this has a very significant performance impact so we bypass
1025 * if nstat_collect is disabled. We may also bypass if the
1026 * protocol is tcp in the future because tcp will have a route that
1027 * we can use to attribute the data to. That does mean we would not
1028 * account for forwarded tcp traffic.
1029 */
1030 if (nstat_collect) {
1031 rtentry_ref_t rte =
1032 ifnet_cached_rtlookup_inet6(inifp, &ip6->ip6_src);
1033 if (rte != NULL) {
1034 nstat_route_rx(rte, 1, m->m_pkthdr.len, 0);
1035 rtfree(rte);
1036 }
1037 }
1038
1039 #if DUMMYNET
1040 check_with_pf:
1041 #endif /* DUMMYNET */
1042 #if PF
1043 /* Invoke inbound packet filter */
1044 if (PF_IS_ENABLED) {
1045 int error;
1046 #if DUMMYNET
1047 error = pf_af_hook(inifp, NULL, &m, AF_INET6, TRUE, &args);
1048 #else /* !DUMMYNET */
1049 error = pf_af_hook(inifp, NULL, &m, AF_INET6, TRUE, NULL);
1050 #endif /* !DUMMYNET */
1051 if (error != 0 || m == NULL) {
1052 if (m != NULL) {
1053 panic("%s: unexpected packet %p",
1054 __func__, m);
1055 /* NOTREACHED */
1056 }
1057 /* Already freed by callee */
1058 goto done;
1059 }
1060 ip6 = mtod(m, struct ip6_hdr *);
1061 }
1062 #endif /* PF */
1063
1064 /* drop packets if interface ID portion is already filled */
1065 if (!(inifp->if_flags & IFF_LOOPBACK) &&
1066 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
1067 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
1068 ip6->ip6_src.s6_addr16[1]) {
1069 ip6stat.ip6s_badscope++;
1070 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1071 goto bad;
1072 }
1073 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) &&
1074 ip6->ip6_dst.s6_addr16[1]) {
1075 ip6stat.ip6s_badscope++;
1076 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1077 goto bad;
1078 }
1079 }
1080
1081 if ((m->m_pkthdr.pkt_flags & PKTF_IFAINFO) != 0 && in6_embedded_scope) {
1082 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1083 ip6->ip6_src.s6_addr16[1] =
1084 htons(m->m_pkthdr.src_ifindex);
1085 }
1086 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
1087 ip6->ip6_dst.s6_addr16[1] =
1088 htons(m->m_pkthdr.dst_ifindex);
1089 }
1090 } else if (in6_embedded_scope) {
1091 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1092 ip6->ip6_src.s6_addr16[1] = htons(inifp->if_index);
1093 }
1094 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
1095 ip6->ip6_dst.s6_addr16[1] = htons(inifp->if_index);
1096 }
1097 }
1098
1099 /*
1100 * Multicast check
1101 */
1102 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1103 struct in6_multi *__single in6m = NULL;
1104
1105 in6_ifstat_inc_na(inifp, ifs6_in_mcast);
1106 /*
1107 * See if we belong to the destination multicast group on the
1108 * arrival interface.
1109 */
1110 in6_multihead_lock_shared();
1111 IN6_LOOKUP_MULTI(&ip6->ip6_dst, inifp, in6m);
1112 in6_multihead_lock_done();
1113 if (in6m != NULL) {
1114 IN6M_REMREF(in6m);
1115 ours = 1;
1116 } else if (!nd6_prproxy) {
1117 ip6stat.ip6s_notmember++;
1118 ip6stat.ip6s_cantforward++;
1119 in6_ifstat_inc(inifp, ifs6_in_discard);
1120 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_UNKNOWN_MULTICAST_GROUP, NULL, 0);
1121 goto bad;
1122 }
1123 deliverifp = inifp;
1124 /*
1125 * record dst address information into mbuf, if we don't have one yet.
1126 * note that we are unable to record it, if the address is not listed
1127 * as our interface address (e.g. multicast addresses, etc.)
1128 */
1129 if (deliverifp != NULL) {
1130 struct in6_ifaddr *__single ia6 = NULL;
1131
1132 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
1133 if (ia6 != NULL) {
1134 (void) ip6_setdstifaddr_info(m, 0, ia6);
1135 (void) ip6_setsrcifaddr_info(m, ia6->ia_ifp->if_index, NULL);
1136 ifa_remref(&ia6->ia_ifa);
1137 } else {
1138 (void) ip6_setdstifaddr_info(m, inifp->if_index, NULL);
1139 (void) ip6_setsrcifaddr_info(m, inifp->if_index, NULL);
1140 }
1141 }
1142 goto hbhcheck;
1143 } else {
1144 /*
1145 * Unicast check
1146 */
1147 ip6_check_if_result_t check_if_result = IP6_CHECK_IF_NONE;
1148 check_if_result = ip6_input_check_interface(m, ip6, inifp, &rin6, &deliverifp);
1149 ASSERT(check_if_result != IP6_CHECK_IF_NONE);
1150 if (check_if_result == IP6_CHECK_IF_OURS) {
1151 ours = 1;
1152 goto hbhcheck;
1153 } else if (check_if_result == IP6_CHECK_IF_DROP) {
1154 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_RCV_IF_NO_MATCH, NULL, 0);
1155 goto bad;
1156 }
1157 }
1158
1159 /*
1160 * Now there is no reason to process the packet if it's not our own
1161 * and we're not a router.
1162 */
1163 if (!ip6_forwarding) {
1164 ip6stat.ip6s_cantforward++;
1165 in6_ifstat_inc(inifp, ifs6_in_discard);
1166 /*
1167 * Raise a kernel event if the packet received on cellular
1168 * interface is not intended for local host.
1169 * For now limit it to ICMPv6 packets.
1170 */
1171 if (inifp->if_type == IFT_CELLULAR &&
1172 ip6->ip6_nxt == IPPROTO_ICMPV6) {
1173 in6_ifstat_inc(inifp, ifs6_cantfoward_icmp6);
1174 }
1175 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_CANNOT_FORWARD, NULL, 0);
1176 goto bad;
1177 }
1178
1179 hbhcheck:
1180 /*
1181 * Process Hop-by-Hop options header if it's contained.
1182 * m may be modified in ip6_hopopts_input().
1183 * If a JumboPayload option is included, plen will also be modified.
1184 */
1185 plen = (u_int32_t)ntohs(ip6->ip6_plen);
1186 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1187 struct ip6_hbh *__single hbh;
1188
1189 /*
1190 * Mark the packet to imply that HBH option has been checked.
1191 * This can only be true is the packet came in unfragmented
1192 * or if the option is in the first fragment
1193 */
1194 m->m_pkthdr.pkt_flags |= PKTF_HBH_CHKED;
1195 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
1196 #if 0 /* touches NULL pointer */
1197 in6_ifstat_inc(inifp, ifs6_in_discard);
1198 #endif
1199 goto done; /* m have already been freed */
1200 }
1201
1202 /* adjust pointer */
1203 ip6 = mtod(m, struct ip6_hdr *);
1204
1205 /*
1206 * if the payload length field is 0 and the next header field
1207 * indicates Hop-by-Hop Options header, then a Jumbo Payload
1208 * option MUST be included.
1209 */
1210 if (ip6->ip6_plen == 0 && plen == 0) {
1211 /*
1212 * Note that if a valid jumbo payload option is
1213 * contained, ip6_hopopts_input() must set a valid
1214 * (non-zero) payload length to the variable plen.
1215 */
1216 ip6stat.ip6s_badoptions++;
1217 in6_ifstat_inc(inifp, ifs6_in_discard);
1218 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
1219 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
1220 (int)((caddr_t)&ip6->ip6_plen - (caddr_t)ip6));
1221 goto done;
1222 }
1223 /* ip6_hopopts_input() ensures that mbuf is contiguous */
1224 hbh = (struct ip6_hbh *)(ip6 + 1);
1225 nxt = hbh->ip6h_nxt;
1226
1227 /*
1228 * If we are acting as a router and the packet contains a
1229 * router alert option, see if we know the option value.
1230 * Currently, we only support the option value for MLD, in which
1231 * case we should pass the packet to the multicast routing
1232 * daemon.
1233 */
1234 if (rtalert != ~0 && ip6_forwarding) {
1235 switch (rtalert) {
1236 case IP6OPT_RTALERT_MLD:
1237 ours = 1;
1238 break;
1239 default:
1240 /*
1241 * RFC2711 requires unrecognized values must be
1242 * silently ignored.
1243 */
1244 break;
1245 }
1246 }
1247 } else {
1248 nxt = ip6->ip6_nxt;
1249 }
1250
1251 /*
1252 * Check that the amount of data in the buffers
1253 * is as at least much as the IPv6 header would have us expect.
1254 * Trim mbufs if longer than we expect.
1255 * Drop packet if shorter than we expect.
1256 */
1257 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
1258 ip6stat.ip6s_tooshort++;
1259 in6_ifstat_inc(inifp, ifs6_in_truncated);
1260 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SHORT, NULL, 0);
1261 goto bad;
1262 }
1263
1264 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
1265 ip6_input_adjust(m, ip6, plen, inifp);
1266 }
1267
1268 /*
1269 * Forward if desirable.
1270 */
1271 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1272 if (!ours && nd6_prproxy) {
1273 /*
1274 * If this isn't for us, this might be a Neighbor
1275 * Solicitation (dst is solicited-node multicast)
1276 * against an address in one of the proxied prefixes;
1277 * if so, claim the packet and let icmp6_input()
1278 * handle the rest.
1279 */
1280 ours = nd6_prproxy_isours(m, ip6, NULL, IFSCOPE_NONE);
1281 VERIFY(!ours ||
1282 (m->m_pkthdr.pkt_flags & PKTF_PROXY_DST));
1283 }
1284 if (!ours) {
1285 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_UNPROXIED_NS, NULL, 0);
1286 goto bad;
1287 }
1288 } else if (!ours) {
1289 /*
1290 * The unicast forwarding function might return the packet
1291 * if we are proxying prefix(es), and if the packet is an
1292 * ICMPv6 packet that has failed the zone checks, but is
1293 * targetted towards a proxied address (this is optimized by
1294 * way of RTF_PROXY test.) If so, claim the packet as ours
1295 * and let icmp6_input() handle the rest. The packet's hop
1296 * limit value is kept intact (it's not decremented). This
1297 * is for supporting Neighbor Unreachability Detection between
1298 * proxied nodes on different links (src is link-local, dst
1299 * is target address.)
1300 */
1301 if ((m = ip6_forward(m, &rin6, 0)) == NULL) {
1302 goto done;
1303 }
1304 VERIFY(rin6.ro_rt != NULL);
1305 VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
1306 deliverifp = rin6.ro_rt->rt_ifp;
1307 ours = 1;
1308 }
1309
1310 ip6 = mtod(m, struct ip6_hdr *);
1311
1312 /*
1313 * Malicious party may be able to use IPv4 mapped addr to confuse
1314 * tcp/udp stack and bypass security checks (act as if it was from
1315 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious.
1316 *
1317 * For SIIT end node behavior, you may want to disable the check.
1318 * However, you will become vulnerable to attacks using IPv4 mapped
1319 * source.
1320 */
1321 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
1322 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
1323 ip6stat.ip6s_badscope++;
1324 in6_ifstat_inc(inifp, ifs6_in_addrerr);
1325 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1326 goto bad;
1327 }
1328
1329 /*
1330 * Tell launch routine the next header
1331 */
1332 ip6stat.ip6s_delivered++;
1333 in6_ifstat_inc_na(deliverifp, ifs6_in_deliver);
1334
1335 injectit:
1336 nest = 0;
1337
1338 /*
1339 * Perform IP header alignment fixup again, if needed. Note that
1340 * we do it once for the outermost protocol, and we assume each
1341 * protocol handler wouldn't mess with the alignment afterwards.
1342 */
1343 IP6_HDR_ALIGNMENT_FIXUP(m, inifp, return );
1344
1345 while (nxt != IPPROTO_DONE) {
1346 struct ipfilter *__single filter;
1347 int (*pr_input)(struct mbuf **, int *, int);
1348
1349 /*
1350 * This would imply either IPPROTO_HOPOPTS was not the first
1351 * option or it did not come in the first fragment.
1352 */
1353 if (nxt == IPPROTO_HOPOPTS &&
1354 (m->m_pkthdr.pkt_flags & PKTF_HBH_CHKED) == 0) {
1355 /*
1356 * This implies that HBH option was not contained
1357 * in the first fragment
1358 */
1359 ip6stat.ip6s_badoptions++;
1360 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_OPTION, NULL, 0);
1361 goto bad;
1362 }
1363
1364 if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
1365 ip6stat.ip6s_toomanyhdr++;
1366 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_TOO_MANY_OPTIONS, NULL, 0);
1367 goto bad;
1368 }
1369
1370 /*
1371 * protection against faulty packet - there should be
1372 * more sanity checks in header chain processing.
1373 */
1374 if (m->m_pkthdr.len < off) {
1375 ip6stat.ip6s_tooshort++;
1376 in6_ifstat_inc(inifp, ifs6_in_truncated);
1377 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SHORT, NULL, 0);
1378 goto bad;
1379 }
1380
1381 #if IPSEC
1382 /*
1383 * enforce IPsec policy checking if we are seeing last header.
1384 * note that we do not visit this with protocols with pcb layer
1385 * code - like udp/tcp/raw ip.
1386 */
1387 if ((ipsec_bypass == 0) &&
1388 (ip6_protox[nxt]->pr_flags & PR_LASTHDR) != 0) {
1389 if (ipsec6_in_reject(m, NULL)) {
1390 IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
1391 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IPSEC_REJECT, NULL, 0);
1392 goto bad;
1393 }
1394 }
1395 #endif /* IPSEC */
1396
1397 /*
1398 * Call IP filter
1399 */
1400 if (!TAILQ_EMPTY(&ipv6_filters) && !IFNET_IS_INTCOPROC(inifp)) {
1401 ipf_ref();
1402 TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) {
1403 if (seen == 0) {
1404 if ((struct ipfilter *)inject_ipfref ==
1405 filter) {
1406 seen = 1;
1407 }
1408 } else if (filter->ipf_filter.ipf_input) {
1409 errno_t result;
1410
1411 result = filter->ipf_filter.ipf_input(
1412 filter->ipf_filter.cookie,
1413 (mbuf_t *)&m, off, (uint8_t)nxt);
1414 if (result == EJUSTRETURN) {
1415 ipf_unref();
1416 goto done;
1417 }
1418 if (result != 0) {
1419 ipf_unref();
1420 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_FILTER_DROP, NULL, 0);
1421 goto bad;
1422 }
1423 }
1424 }
1425 ipf_unref();
1426 }
1427
1428 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
1429 struct ip6_hdr *, ip6, struct ifnet *, inifp,
1430 struct ip *, NULL, struct ip6_hdr *, ip6);
1431
1432 if ((pr_input = ip6_protox[nxt]->pr_input) == NULL) {
1433 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_NO_PROTO, NULL, 0);
1434 m = NULL;
1435 nxt = IPPROTO_DONE;
1436 } else if (!(ip6_protox[nxt]->pr_flags & PR_PROTOLOCK)) {
1437 lck_mtx_lock(inet6_domain_mutex);
1438 nxt = pr_input(&m, &off, nxt);
1439 lck_mtx_unlock(inet6_domain_mutex);
1440 } else {
1441 nxt = pr_input(&m, &off, nxt);
1442 }
1443 }
1444 done:
1445 ROUTE_RELEASE(&rin6);
1446 return;
1447 bad:
1448 goto done;
1449 }
1450
1451 void
ip6_setsrcifaddr_info(struct mbuf * m,uint32_t src_idx,struct in6_ifaddr * ia6)1452 ip6_setsrcifaddr_info(struct mbuf *m, uint32_t src_idx, struct in6_ifaddr *ia6)
1453 {
1454 VERIFY(m->m_flags & M_PKTHDR);
1455 m->m_pkthdr.pkt_ext_flags &= ~PKTF_EXT_OUTPUT_SCOPE;
1456 /*
1457 * If the source ifaddr is specified, pick up the information
1458 * from there; otherwise just grab the passed-in ifindex as the
1459 * caller may not have the ifaddr available.
1460 */
1461 if (ia6 != NULL) {
1462 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1463 m->m_pkthdr.src_ifindex = ia6->ia_ifp->if_index;
1464
1465 /* See IN6_IFF comments in in6_var.h */
1466 m->m_pkthdr.src_iff = (ia6->ia6_flags & 0xffff);
1467 } else {
1468 m->m_pkthdr.src_iff = 0;
1469 m->m_pkthdr.src_ifindex = (uint16_t)src_idx;
1470 if (src_idx != 0) {
1471 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1472 }
1473 }
1474 }
1475
1476 void
ip6_setdstifaddr_info(struct mbuf * m,uint32_t dst_idx,struct in6_ifaddr * ia6)1477 ip6_setdstifaddr_info(struct mbuf *m, uint32_t dst_idx, struct in6_ifaddr *ia6)
1478 {
1479 VERIFY(m->m_flags & M_PKTHDR);
1480 m->m_pkthdr.pkt_ext_flags &= ~PKTF_EXT_OUTPUT_SCOPE;
1481
1482 /*
1483 * If the destination ifaddr is specified, pick up the information
1484 * from there; otherwise just grab the passed-in ifindex as the
1485 * caller may not have the ifaddr available.
1486 */
1487 if (ia6 != NULL) {
1488 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1489 m->m_pkthdr.dst_ifindex = ia6->ia_ifp->if_index;
1490
1491 /* See IN6_IFF comments in in6_var.h */
1492 m->m_pkthdr.dst_iff = (ia6->ia6_flags & 0xffff);
1493 } else {
1494 m->m_pkthdr.dst_iff = 0;
1495 m->m_pkthdr.dst_ifindex = (uint16_t)dst_idx;
1496 if (dst_idx != 0) {
1497 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1498 }
1499 }
1500 }
1501
1502 int
ip6_getsrcifaddr_info(struct mbuf * m,uint32_t * src_idx,uint32_t * ia6f)1503 ip6_getsrcifaddr_info(struct mbuf *m, uint32_t *src_idx, uint32_t *ia6f)
1504 {
1505 VERIFY(m->m_flags & M_PKTHDR);
1506
1507 if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO)) {
1508 return -1;
1509 }
1510
1511 if (src_idx != NULL) {
1512 *src_idx = m->m_pkthdr.src_ifindex;
1513 }
1514
1515 if (ia6f != NULL) {
1516 *ia6f = m->m_pkthdr.src_iff;
1517 }
1518
1519 return 0;
1520 }
1521
1522 int
ip6_getdstifaddr_info(struct mbuf * m,uint32_t * dst_idx,uint32_t * ia6f)1523 ip6_getdstifaddr_info(struct mbuf *m, uint32_t *dst_idx, uint32_t *ia6f)
1524 {
1525 VERIFY(m->m_flags & M_PKTHDR);
1526
1527 if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO)) {
1528 return -1;
1529 }
1530
1531 if (dst_idx != NULL) {
1532 *dst_idx = m->m_pkthdr.dst_ifindex;
1533 }
1534
1535 if (ia6f != NULL) {
1536 *ia6f = m->m_pkthdr.dst_iff;
1537 }
1538
1539 return 0;
1540 }
1541
1542 uint32_t
ip6_input_getsrcifscope(struct mbuf * m)1543 ip6_input_getsrcifscope(struct mbuf *m)
1544 {
1545 VERIFY(m->m_flags & M_PKTHDR);
1546
1547 if (m->m_pkthdr.rcvif != NULL) {
1548 return m->m_pkthdr.rcvif->if_index;
1549 }
1550
1551 uint32_t src_ifscope = IFSCOPE_NONE;
1552 ip6_getsrcifaddr_info(m, &src_ifscope, NULL);
1553 return src_ifscope;
1554 }
1555
1556 uint32_t
ip6_input_getdstifscope(struct mbuf * m)1557 ip6_input_getdstifscope(struct mbuf *m)
1558 {
1559 VERIFY(m->m_flags & M_PKTHDR);
1560
1561 if (m->m_pkthdr.rcvif != NULL) {
1562 return m->m_pkthdr.rcvif->if_index;
1563 }
1564
1565 uint32_t dst_ifscope = IFSCOPE_NONE;
1566 ip6_getdstifaddr_info(m, &dst_ifscope, NULL);
1567 return dst_ifscope;
1568 }
1569
1570 /*
1571 * Hop-by-Hop options header processing. If a valid jumbo payload option is
1572 * included, the real payload length will be stored in plenp.
1573 */
1574 static int
ip6_hopopts_input(uint32_t * plenp,uint32_t * rtalertp,struct mbuf ** mp,int * offp)1575 ip6_hopopts_input(uint32_t *plenp, uint32_t *rtalertp, struct mbuf **mp,
1576 int *offp)
1577 {
1578 mbuf_ref_t m = *mp;
1579 int off = *offp, hbhlen;
1580 struct ip6_hbh *hbh;
1581 u_int8_t *opt;
1582
1583 /* validation of the length of the header */
1584 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), return (-1));
1585 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1586 hbhlen = (hbh->ip6h_len + 1) << 3;
1587
1588 IP6_EXTHDR_CHECK(m, off, hbhlen, return (-1));
1589 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1590 off += hbhlen;
1591 hbhlen -= sizeof(struct ip6_hbh);
1592 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1593
1594 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1595 hbhlen, rtalertp, plenp) < 0) {
1596 return -1;
1597 }
1598
1599 *offp = off;
1600 *mp = m;
1601 return 0;
1602 }
1603
1604 /*
1605 * Search header for all Hop-by-hop options and process each option.
1606 * This function is separate from ip6_hopopts_input() in order to
1607 * handle a case where the sending node itself process its hop-by-hop
1608 * options header. In such a case, the function is called from ip6_output().
1609 *
1610 * The function assumes that hbh header is located right after the IPv6 header
1611 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1612 * opthead + hbhlen is located in continuous memory region.
1613 */
1614 int
ip6_process_hopopts(struct mbuf * m,u_int8_t * __sized_by (inhbhlen)opthead,int inhbhlen,u_int32_t * rtalertp,u_int32_t * plenp)1615 ip6_process_hopopts(struct mbuf *m, u_int8_t *__sized_by(inhbhlen)opthead, int inhbhlen,
1616 u_int32_t *rtalertp, u_int32_t *plenp)
1617 {
1618 struct ip6_hdr *__single ip6;
1619 int optlen = 0;
1620 int hbhlen = inhbhlen;
1621 u_int8_t *opt = opthead;
1622 u_int16_t rtalert_val;
1623 u_int32_t jumboplen;
1624 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1625
1626 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1627 switch (*opt) {
1628 case IP6OPT_PAD1:
1629 optlen = 1;
1630 break;
1631 case IP6OPT_PADN:
1632 if (hbhlen < IP6OPT_MINLEN) {
1633 ip6stat.ip6s_toosmall++;
1634 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0);
1635 goto bad;
1636 }
1637 optlen = *(opt + 1) + 2;
1638 break;
1639 case IP6OPT_ROUTER_ALERT:
1640 /* XXX may need check for alignment */
1641 if (hbhlen < IP6OPT_RTALERT_LEN) {
1642 ip6stat.ip6s_toosmall++;
1643 goto bad;
1644 }
1645 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1646 /* XXX stat */
1647 icmp6_error(m, ICMP6_PARAM_PROB,
1648 ICMP6_PARAMPROB_HEADER,
1649 (int)(erroff + opt + 1 - opthead));
1650 return -1;
1651 }
1652 optlen = IP6OPT_RTALERT_LEN;
1653 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1654 *rtalertp = ntohs(rtalert_val);
1655 break;
1656 case IP6OPT_JUMBO:
1657 /* XXX may need check for alignment */
1658 if (hbhlen < IP6OPT_JUMBO_LEN) {
1659 ip6stat.ip6s_toosmall++;
1660 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0);
1661 goto bad;
1662 }
1663 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1664 /* XXX stat */
1665 icmp6_error(m, ICMP6_PARAM_PROB,
1666 ICMP6_PARAMPROB_HEADER,
1667 (int)(erroff + opt + 1 - opthead));
1668 return -1;
1669 }
1670 optlen = IP6OPT_JUMBO_LEN;
1671
1672 /*
1673 * IPv6 packets that have non 0 payload length
1674 * must not contain a jumbo payload option.
1675 */
1676 ip6 = mtod(m, struct ip6_hdr *);
1677 if (ip6->ip6_plen) {
1678 ip6stat.ip6s_badoptions++;
1679 icmp6_error(m, ICMP6_PARAM_PROB,
1680 ICMP6_PARAMPROB_HEADER,
1681 (int)(erroff + opt - opthead));
1682 return -1;
1683 }
1684
1685 /*
1686 * We may see jumbolen in unaligned location, so
1687 * we'd need to perform bcopy().
1688 */
1689 bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1690 jumboplen = (u_int32_t)htonl(jumboplen);
1691
1692 #if 1
1693 /*
1694 * if there are multiple jumbo payload options,
1695 * *plenp will be non-zero and the packet will be
1696 * rejected.
1697 * the behavior may need some debate in ipngwg -
1698 * multiple options does not make sense, however,
1699 * there's no explicit mention in specification.
1700 */
1701 if (*plenp != 0) {
1702 ip6stat.ip6s_badoptions++;
1703 icmp6_error(m, ICMP6_PARAM_PROB,
1704 ICMP6_PARAMPROB_HEADER,
1705 (int)(erroff + opt + 2 - opthead));
1706 return -1;
1707 }
1708 #endif
1709
1710 /*
1711 * jumbo payload length must be larger than 65535.
1712 */
1713 if (jumboplen <= IPV6_MAXPACKET) {
1714 ip6stat.ip6s_badoptions++;
1715 icmp6_error(m, ICMP6_PARAM_PROB,
1716 ICMP6_PARAMPROB_HEADER,
1717 (int)(erroff + opt + 2 - opthead));
1718 return -1;
1719 }
1720 *plenp = jumboplen;
1721
1722 break;
1723 default: /* unknown option */
1724 if (hbhlen < IP6OPT_MINLEN) {
1725 ip6stat.ip6s_toosmall++;
1726 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0);
1727 goto bad;
1728 }
1729 optlen = ip6_unknown_opt(opt, hbhlen, m,
1730 erroff + opt - opthead);
1731 if (optlen == -1) {
1732 return -1;
1733 }
1734 optlen += 2;
1735 break;
1736 }
1737 }
1738
1739 return 0;
1740
1741 bad:
1742 return -1;
1743 }
1744
1745 /*
1746 * Unknown option processing.
1747 * The fourth argument `off' is the offset from the IPv6 header to the option,
1748 * which is necessary if the IPv6 header the and option header and IPv6 header
1749 * is not continuous in order to return an ICMPv6 error.
1750 */
1751 int
ip6_unknown_opt(uint8_t * __counted_by (optplen)optp,size_t optplen,struct mbuf * m,size_t off)1752 ip6_unknown_opt(uint8_t *__counted_by(optplen) optp, size_t optplen, struct mbuf *m, size_t off)
1753 {
1754 #pragma unused(optplen)
1755
1756 struct ip6_hdr *__single ip6;
1757
1758 switch (IP6OPT_TYPE(*optp)) {
1759 case IP6OPT_TYPE_SKIP: /* ignore the option */
1760 return (int)*(optp + 1);
1761
1762 case IP6OPT_TYPE_DISCARD: /* silently discard */
1763 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_OPT_DISCARD, NULL, 0);
1764 return -1;
1765
1766 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1767 ip6stat.ip6s_badoptions++;
1768 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, (int)off);
1769 return -1;
1770
1771 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1772 ip6stat.ip6s_badoptions++;
1773 ip6 = mtod(m, struct ip6_hdr *);
1774 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1775 (m->m_flags & (M_BCAST | M_MCAST))) {
1776 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_OPTION, NULL, 0);
1777 } else {
1778 icmp6_error(m, ICMP6_PARAM_PROB,
1779 ICMP6_PARAMPROB_OPTION, (int)off);
1780 }
1781 return -1;
1782 }
1783
1784 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_UNSPECIFIED, NULL, 0); /* XXX: NOTREACHED */
1785 return -1;
1786 }
1787
1788 /*
1789 * Create the "control" list for this pcb.
1790 * These functions will not modify mbuf chain at all.
1791 *
1792 * With KAME mbuf chain restriction:
1793 * The routine will be called from upper layer handlers like tcp6_input().
1794 * Thus the routine assumes that the caller (tcp6_input) have already
1795 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1796 * very first mbuf on the mbuf chain.
1797 *
1798 * ip6_savecontrol_v4 will handle those options that are possible to be
1799 * set on a v4-mapped socket.
1800 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1801 * options and handle the v6-only ones itself.
1802 */
1803 struct mbuf **
ip6_savecontrol_v4(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp,int * v4only)1804 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1805 int *v4only)
1806 {
1807 struct ip6_hdr *__single ip6 = mtod(m, struct ip6_hdr *);
1808
1809 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1810 struct timeval tv;
1811
1812 getmicrotime(&tv);
1813 mp = sbcreatecontrol_mbuf((caddr_t)&tv, sizeof(tv),
1814 SCM_TIMESTAMP, SOL_SOCKET, mp);
1815 if (*mp == NULL) {
1816 return NULL;
1817 }
1818 }
1819 if ((inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) {
1820 uint64_t time;
1821
1822 time = mach_absolute_time();
1823 mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof(time),
1824 SCM_TIMESTAMP_MONOTONIC, SOL_SOCKET, mp);
1825 if (*mp == NULL) {
1826 return NULL;
1827 }
1828 }
1829 if ((inp->inp_socket->so_options & SO_TIMESTAMP_CONTINUOUS) != 0) {
1830 uint64_t time;
1831
1832 time = mach_continuous_time();
1833 mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof(time),
1834 SCM_TIMESTAMP_CONTINUOUS, SOL_SOCKET, mp);
1835 if (*mp == NULL) {
1836 return NULL;
1837 }
1838 }
1839 if ((inp->inp_socket->so_flags & SOF_RECV_TRAFFIC_CLASS) != 0) {
1840 int tc = m_get_traffic_class(m);
1841
1842 mp = sbcreatecontrol_mbuf((caddr_t)&tc, sizeof(tc),
1843 SO_TRAFFIC_CLASS, SOL_SOCKET, mp);
1844 if (*mp == NULL) {
1845 return NULL;
1846 }
1847 }
1848
1849 if ((inp->inp_socket->so_flags & SOF_RECV_WAKE_PKT) &&
1850 (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
1851 int flag = 1;
1852
1853 mp = sbcreatecontrol_mbuf((caddr_t)&flag, sizeof(flag),
1854 SO_RECV_WAKE_PKT, SOL_SOCKET, mp);
1855 if (*mp == NULL) {
1856 return NULL;
1857 }
1858 }
1859
1860 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1861 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1862 if (v4only != NULL) {
1863 *v4only = 1;
1864 }
1865
1866 // Send ECN flags for v4-mapped addresses
1867 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1868 struct ip *__single ip_header = mtod(m, struct ip *);
1869
1870 int tclass = (int)(ip_header->ip_tos);
1871 mp = sbcreatecontrol_mbuf((caddr_t)&tclass, sizeof(tclass),
1872 IPV6_TCLASS, IPPROTO_IPV6, mp);
1873 if (*mp == NULL) {
1874 return NULL;
1875 }
1876 }
1877
1878 // Send IN6P_PKTINFO for v4-mapped address
1879 if ((inp->inp_flags & IN6P_PKTINFO) != 0 || SOFLOW_ENABLED(inp->inp_socket)) {
1880 struct in6_pktinfo pi6 = {
1881 .ipi6_addr = IN6ADDR_V4MAPPED_INIT,
1882 .ipi6_ifindex = (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0,
1883 };
1884
1885 struct ip *__single ip_header = mtod(m, struct ip *);
1886 bcopy(&ip_header->ip_dst, &pi6.ipi6_addr.s6_addr32[3], sizeof(struct in_addr));
1887
1888 mp = sbcreatecontrol_mbuf((caddr_t)&pi6,
1889 sizeof(struct in6_pktinfo),
1890 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO),
1891 IPPROTO_IPV6, mp);
1892 if (*mp == NULL) {
1893 return NULL;
1894 }
1895 }
1896 return mp;
1897 }
1898
1899 /* RFC 2292 sec. 5 */
1900 if ((inp->inp_flags & IN6P_PKTINFO) != 0 || SOFLOW_ENABLED(inp->inp_socket)) {
1901 struct in6_pktinfo pi6;
1902
1903 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1904 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1905 pi6.ipi6_ifindex =
1906 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1907
1908 mp = sbcreatecontrol_mbuf((caddr_t)&pi6,
1909 sizeof(struct in6_pktinfo),
1910 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO),
1911 IPPROTO_IPV6, mp);
1912 if (*mp == NULL) {
1913 return NULL;
1914 }
1915 }
1916
1917 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1918 int hlim = ip6->ip6_hlim & 0xff;
1919
1920 mp = sbcreatecontrol_mbuf((caddr_t)&hlim, sizeof(int),
1921 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1922 IPPROTO_IPV6, mp);
1923 if (*mp == NULL) {
1924 return NULL;
1925 }
1926 }
1927
1928 if (v4only != NULL) {
1929 *v4only = 0;
1930 }
1931 return mp;
1932 }
1933
1934 int
ip6_savecontrol(struct inpcb * in6p,struct mbuf * m,struct mbuf ** mp)1935 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1936 {
1937 struct mbuf **np;
1938 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1939 int v4only = 0;
1940
1941 *mp = NULL;
1942 np = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1943 if (np == NULL) {
1944 goto no_mbufs;
1945 }
1946
1947 mp = np;
1948 if (v4only) {
1949 return 0;
1950 }
1951
1952 if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1953 u_int32_t flowinfo;
1954 int tclass;
1955
1956 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1957 flowinfo >>= 20;
1958
1959 tclass = flowinfo & 0xff;
1960 mp = sbcreatecontrol_mbuf((caddr_t)&tclass, sizeof(tclass),
1961 IPV6_TCLASS, IPPROTO_IPV6, mp);
1962 if (*mp == NULL) {
1963 goto no_mbufs;
1964 }
1965 }
1966
1967 /*
1968 * IPV6_HOPOPTS socket option. Recall that we required super-user
1969 * privilege for the option (see ip6_ctloutput), but it might be too
1970 * strict, since there might be some hop-by-hop options which can be
1971 * returned to normal user.
1972 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1973 */
1974 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1975 /*
1976 * Check if a hop-by-hop options header is contatined in the
1977 * received packet, and if so, store the options as ancillary
1978 * data. Note that a hop-by-hop options header must be
1979 * just after the IPv6 header, which is assured through the
1980 * IPv6 input processing.
1981 */
1982 ip6 = mtod(m, struct ip6_hdr *);
1983 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1984 struct ip6_hbh *hbh;
1985 int hbhlen = 0;
1986 hbh = (struct ip6_hbh *)(ip6 + 1);
1987 hbhlen = (hbh->ip6h_len + 1) << 3;
1988
1989 /*
1990 * XXX: We copy the whole header even if a
1991 * jumbo payload option is included, the option which
1992 * is to be removed before returning according to
1993 * RFC2292.
1994 * Note: this constraint is removed in RFC3542
1995 */
1996 mp = sbcreatecontrol_mbuf((caddr_t)hbh, hbhlen,
1997 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1998 IPPROTO_IPV6, mp);
1999
2000 if (*mp == NULL) {
2001 goto no_mbufs;
2002 }
2003 }
2004 }
2005
2006 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
2007 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
2008
2009 /*
2010 * Search for destination options headers or routing
2011 * header(s) through the header chain, and stores each
2012 * header as ancillary data.
2013 * Note that the order of the headers remains in
2014 * the chain of ancillary data.
2015 */
2016 while (1) { /* is explicit loop prevention necessary? */
2017 struct ip6_ext *ip6e = NULL;
2018 int elen;
2019
2020 /*
2021 * if it is not an extension header, don't try to
2022 * pull it from the chain.
2023 */
2024 switch (nxt) {
2025 case IPPROTO_DSTOPTS:
2026 case IPPROTO_ROUTING:
2027 case IPPROTO_HOPOPTS:
2028 case IPPROTO_AH: /* is it possible? */
2029 break;
2030 default:
2031 goto loopend;
2032 }
2033
2034 if (off + sizeof(*ip6e) > m->m_len) {
2035 goto loopend;
2036 }
2037 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
2038 if (nxt == IPPROTO_AH) {
2039 elen = (ip6e->ip6e_len + 2) << 2;
2040 } else {
2041 elen = (ip6e->ip6e_len + 1) << 3;
2042 }
2043 if (off + elen > m->m_len) {
2044 goto loopend;
2045 }
2046
2047 switch (nxt) {
2048 case IPPROTO_DSTOPTS:
2049 if (!(in6p->inp_flags & IN6P_DSTOPTS)) {
2050 break;
2051 }
2052
2053 mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
2054 IS2292(in6p, IPV6_2292DSTOPTS,
2055 IPV6_DSTOPTS), IPPROTO_IPV6, mp);
2056 if (*mp == NULL) {
2057 goto no_mbufs;
2058 }
2059 break;
2060 case IPPROTO_ROUTING:
2061 if (!(in6p->inp_flags & IN6P_RTHDR)) {
2062 break;
2063 }
2064
2065 mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
2066 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
2067 IPPROTO_IPV6, mp);
2068 if (*mp == NULL) {
2069 goto no_mbufs;
2070 }
2071 break;
2072 case IPPROTO_HOPOPTS:
2073 case IPPROTO_AH: /* is it possible? */
2074 break;
2075
2076 default:
2077 /*
2078 * other cases have been filtered in the above.
2079 * none will visit this case. here we supply
2080 * the code just in case (nxt overwritten or
2081 * other cases).
2082 */
2083 goto loopend;
2084 }
2085
2086 /* proceed with the next header. */
2087 off += elen;
2088 nxt = ip6e->ip6e_nxt;
2089 ip6e = NULL;
2090 }
2091 loopend:
2092 ;
2093 }
2094 return 0;
2095 no_mbufs:
2096 ip6stat.ip6s_pktdropcntrl++;
2097 /* XXX increment a stat to show the failure */
2098 return ENOBUFS;
2099 }
2100 #undef IS2292
2101
2102 void
ip6_notify_pmtu(struct inpcb * in6p,struct sockaddr_in6 * dst,u_int32_t * mtu)2103 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
2104 {
2105 struct socket *__single so;
2106 mbuf_ref_t m_mtu;
2107 struct ip6_mtuinfo mtuctl;
2108
2109 so = in6p->inp_socket;
2110
2111 if ((in6p->inp_flags & IN6P_MTU) == 0) {
2112 return;
2113 }
2114
2115 if (mtu == NULL) {
2116 return;
2117 }
2118
2119 if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && SOCK_CHECK_PROTO(so, IPPROTO_TCP)) {
2120 return;
2121 }
2122
2123 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2124 !in6_are_addr_equal_scoped(&in6p->in6p_faddr, &dst->sin6_addr, in6p->inp_fifscope, dst->sin6_scope_id)) {
2125 return;
2126 }
2127
2128 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */
2129 mtuctl.ip6m_mtu = *mtu;
2130 mtuctl.ip6m_addr = *dst;
2131 if (!in6_embedded_scope) {
2132 mtuctl.ip6m_addr.sin6_scope_id = dst->sin6_scope_id;
2133 }
2134 if (sa6_recoverscope(&mtuctl.ip6m_addr, TRUE)) {
2135 return;
2136 }
2137
2138 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
2139 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) {
2140 return;
2141 }
2142
2143 if (sbappendaddr(&so->so_rcv, SA(dst), NULL, m_mtu, NULL) == 0) {
2144 return;
2145 }
2146 sorwakeup(so);
2147 }
2148
2149 /*
2150 * Get pointer to the previous header followed by the header
2151 * currently processed.
2152 * XXX: This function supposes that
2153 * M includes all headers,
2154 * the next header field and the header length field of each header
2155 * are valid, and
2156 * the sum of each header length equals to OFF.
2157 * Because of these assumptions, this function must be called very
2158 * carefully. Moreover, it will not be used in the near future when
2159 * we develop `neater' mechanism to process extension headers.
2160 */
2161 char *
ip6_get_prevhdr(struct mbuf * m,int off)2162 ip6_get_prevhdr(struct mbuf *m, int off)
2163 {
2164 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2165
2166 if (off == sizeof(struct ip6_hdr)) {
2167 return (char *)&ip6->ip6_nxt;
2168 } else {
2169 int len, nxt;
2170 struct ip6_ext *ip6e = NULL;
2171
2172 nxt = ip6->ip6_nxt;
2173 len = sizeof(struct ip6_hdr);
2174 while (len < off) {
2175 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
2176
2177 switch (nxt) {
2178 case IPPROTO_FRAGMENT:
2179 len += sizeof(struct ip6_frag);
2180 break;
2181 case IPPROTO_AH:
2182 len += (ip6e->ip6e_len + 2) << 2;
2183 break;
2184 default:
2185 len += (ip6e->ip6e_len + 1) << 3;
2186 break;
2187 }
2188 nxt = ip6e->ip6e_nxt;
2189 }
2190 if (ip6e) {
2191 return (char *)&ip6e->ip6e_nxt;
2192 } else {
2193 return NULL;
2194 }
2195 }
2196 }
2197
2198 /*
2199 * get next header offset. m will be retained.
2200 */
2201 int
ip6_nexthdr(struct mbuf * m,int off,int proto,int * nxtp)2202 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
2203 {
2204 struct ip6_hdr ip6;
2205 struct ip6_ext ip6e;
2206 struct ip6_frag fh;
2207
2208 /* just in case */
2209 VERIFY(m != NULL);
2210 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) {
2211 return -1;
2212 }
2213
2214 switch (proto) {
2215 case IPPROTO_IPV6:
2216 if (m->m_pkthdr.len < off + sizeof(ip6)) {
2217 return -1;
2218 }
2219 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
2220 if (nxtp) {
2221 *nxtp = ip6.ip6_nxt;
2222 }
2223 off += sizeof(ip6);
2224 return off;
2225
2226 case IPPROTO_FRAGMENT:
2227 /*
2228 * terminate parsing if it is not the first fragment,
2229 * it does not make sense to parse through it.
2230 */
2231 if (m->m_pkthdr.len < off + sizeof(fh)) {
2232 return -1;
2233 }
2234 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
2235 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
2236 if (fh.ip6f_offlg & IP6F_OFF_MASK) {
2237 return -1;
2238 }
2239 if (nxtp) {
2240 *nxtp = fh.ip6f_nxt;
2241 }
2242 off += sizeof(struct ip6_frag);
2243 return off;
2244
2245 case IPPROTO_AH:
2246 if (m->m_pkthdr.len < off + sizeof(ip6e)) {
2247 return -1;
2248 }
2249 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
2250 if (nxtp) {
2251 *nxtp = ip6e.ip6e_nxt;
2252 }
2253 off += (ip6e.ip6e_len + 2) << 2;
2254 return off;
2255
2256 case IPPROTO_HOPOPTS:
2257 case IPPROTO_ROUTING:
2258 case IPPROTO_DSTOPTS:
2259 if (m->m_pkthdr.len < off + sizeof(ip6e)) {
2260 return -1;
2261 }
2262 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
2263 if (nxtp) {
2264 *nxtp = ip6e.ip6e_nxt;
2265 }
2266 off += (ip6e.ip6e_len + 1) << 3;
2267 return off;
2268
2269 case IPPROTO_NONE:
2270 case IPPROTO_ESP:
2271 case IPPROTO_IPCOMP:
2272 /* give up */
2273 return -1;
2274
2275 default:
2276 return -1;
2277 }
2278 }
2279
2280 /*
2281 * get offset for the last header in the chain. m will be kept untainted.
2282 */
2283 int
ip6_lasthdr(struct mbuf * m,int off,int proto,int * nxtp)2284 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
2285 {
2286 int newoff;
2287 int nxt;
2288
2289 if (!nxtp) {
2290 nxt = -1;
2291 nxtp = &nxt;
2292 }
2293 while (1) {
2294 newoff = ip6_nexthdr(m, off, proto, nxtp);
2295 if (newoff < 0) {
2296 return off;
2297 } else if (newoff < off) {
2298 return -1; /* invalid */
2299 } else if (newoff == off) {
2300 return newoff;
2301 }
2302
2303 off = newoff;
2304 proto = *nxtp;
2305 }
2306 }
2307
2308 boolean_t
ip6_pkt_has_ulp(struct mbuf * m)2309 ip6_pkt_has_ulp(struct mbuf *m)
2310 {
2311 int off = 0, nxt = IPPROTO_NONE;
2312
2313 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
2314 if (off < 0 || m->m_pkthdr.len < off) {
2315 return FALSE;
2316 }
2317
2318 switch (nxt) {
2319 case IPPROTO_TCP:
2320 if (off + sizeof(struct tcphdr) > m->m_pkthdr.len) {
2321 return FALSE;
2322 }
2323 break;
2324 case IPPROTO_UDP:
2325 if (off + sizeof(struct udphdr) > m->m_pkthdr.len) {
2326 return FALSE;
2327 }
2328 break;
2329 case IPPROTO_ICMPV6:
2330 if (off + sizeof(uint32_t) > m->m_pkthdr.len) {
2331 return FALSE;
2332 }
2333 break;
2334 case IPPROTO_NONE:
2335 return TRUE;
2336 case IPPROTO_ESP:
2337 return TRUE;
2338 case IPPROTO_IPCOMP:
2339 return TRUE;
2340 default:
2341 return FALSE;
2342 }
2343 return TRUE;
2344 }
2345
2346 struct ip6aux *
ip6_addaux(struct mbuf * m)2347 ip6_addaux(struct mbuf *m)
2348 {
2349 struct m_tag *__single tag;
2350
2351 /* Check if one is already allocated */
2352 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2353 KERNEL_TAG_TYPE_INET6);
2354 if (tag == NULL) {
2355 /* Allocate a tag */
2356 tag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6,
2357 sizeof(struct ip6aux), M_DONTWAIT, m);
2358
2359 /* Attach it to the mbuf */
2360 if (tag) {
2361 m_tag_prepend(m, tag);
2362 }
2363 }
2364
2365 return tag ? (struct ip6aux *)(tag->m_tag_data) : NULL;
2366 }
2367
2368 struct ip6aux *
ip6_findaux(struct mbuf * m)2369 ip6_findaux(struct mbuf *m)
2370 {
2371 struct m_tag *__single tag;
2372
2373 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2374 KERNEL_TAG_TYPE_INET6);
2375
2376 return tag != NULL ? (struct ip6aux *)(tag->m_tag_data) : NULL;
2377 }
2378
2379 void
ip6_delaux(struct mbuf * m)2380 ip6_delaux(struct mbuf *m)
2381 {
2382 struct m_tag *__single tag;
2383
2384 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2385 KERNEL_TAG_TYPE_INET6);
2386 if (tag != NULL) {
2387 m_tag_delete(m, tag);
2388 }
2389 }
2390
2391 struct inet6_tag_container {
2392 struct m_tag inet6_m_tag;
2393 struct ip6aux inet6_ip6a;
2394 };
2395
2396 struct m_tag *
m_tag_kalloc_inet6(u_int32_t id,u_int16_t type,uint16_t len,int wait)2397 m_tag_kalloc_inet6(u_int32_t id, u_int16_t type, uint16_t len, int wait)
2398 {
2399 struct inet6_tag_container *tag_container;
2400 struct m_tag *tag = NULL;
2401
2402 assert3u(id, ==, KERNEL_MODULE_TAG_ID);
2403 assert3u(type, ==, KERNEL_TAG_TYPE_INET6);
2404 assert3u(len, ==, sizeof(struct ip6aux));
2405
2406 if (len != sizeof(struct ip6aux)) {
2407 return NULL;
2408 }
2409
2410 tag_container = kalloc_type(struct inet6_tag_container, wait | M_ZERO);
2411 if (tag_container != NULL) {
2412 tag = &tag_container->inet6_m_tag;
2413
2414 assert3p(tag, ==, tag_container);
2415
2416 M_TAG_INIT(tag, id, type, len, &tag_container->inet6_ip6a, NULL);
2417 }
2418
2419 return tag;
2420 }
2421
2422 void
m_tag_kfree_inet6(struct m_tag * tag)2423 m_tag_kfree_inet6(struct m_tag *tag)
2424 {
2425 struct inet6_tag_container *__single tag_container = (struct inet6_tag_container *)tag;
2426
2427 assert3u(tag->m_tag_len, ==, sizeof(struct ip6aux));
2428
2429 kfree_type(struct inet6_tag_container, tag_container);
2430 }
2431
2432 void
ip6_register_m_tag(void)2433 ip6_register_m_tag(void)
2434 {
2435 int error;
2436
2437 error = m_register_internal_tag_type(KERNEL_TAG_TYPE_INET6, sizeof(struct ip6aux),
2438 m_tag_kalloc_inet6, m_tag_kfree_inet6);
2439
2440 assert3u(error, ==, 0);
2441 }
2442
2443 /*
2444 * Drain callback
2445 */
2446 void
ip6_drain(void)2447 ip6_drain(void)
2448 {
2449 frag6_drain(); /* fragments */
2450 in6_rtqdrain(); /* protocol cloned routes */
2451 nd6_drain(NULL); /* cloned routes: ND6 */
2452 }
2453
2454 /*
2455 * System control for IP6
2456 */
2457
2458 u_char inet6ctlerrmap[PRC_NCMDS] = {
2459 0, 0, 0, 0,
2460 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
2461 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
2462 EMSGSIZE, EHOSTUNREACH, 0, 0,
2463 0, 0, 0, 0,
2464 ENOPROTOOPT, ECONNREFUSED
2465 };
2466
2467 static int
2468 sysctl_reset_ip6_input_stats SYSCTL_HANDLER_ARGS
2469 {
2470 #pragma unused(arg1, arg2)
2471 int error, i;
2472
2473 i = ip6_input_measure;
2474 error = sysctl_handle_int(oidp, &i, 0, req);
2475 if (error || req->newptr == USER_ADDR_NULL) {
2476 goto done;
2477 }
2478 /* impose bounds */
2479 if (i < 0 || i > 1) {
2480 error = EINVAL;
2481 goto done;
2482 }
2483 if (ip6_input_measure != i && i == 1) {
2484 net_perf_initialize(&net_perf, ip6_input_measure_bins);
2485 }
2486 ip6_input_measure = i;
2487 done:
2488 return error;
2489 }
2490
2491 static int
2492 sysctl_ip6_input_measure_bins SYSCTL_HANDLER_ARGS
2493 {
2494 #pragma unused(arg1, arg2)
2495 int error;
2496 uint64_t i;
2497
2498 i = ip6_input_measure_bins;
2499 error = sysctl_handle_quad(oidp, &i, 0, req);
2500 if (error || req->newptr == USER_ADDR_NULL) {
2501 goto done;
2502 }
2503 /* validate data */
2504 if (!net_perf_validate_bins(i)) {
2505 error = EINVAL;
2506 goto done;
2507 }
2508 ip6_input_measure_bins = i;
2509 done:
2510 return error;
2511 }
2512
2513 static int
2514 sysctl_ip6_input_getperf SYSCTL_HANDLER_ARGS
2515 {
2516 #pragma unused(oidp, arg1, arg2)
2517 if (req->oldptr == USER_ADDR_NULL) {
2518 req->oldlen = (size_t)sizeof(struct net_perf);
2519 }
2520
2521 return SYSCTL_OUT(req, &net_perf, MIN(sizeof(net_perf), req->oldlen));
2522 }
2523
2524
2525 /*
2526 * Initialize IPv6 source address hash table.
2527 */
2528 static void
in6_ifaddrhashtbl_init(void)2529 in6_ifaddrhashtbl_init(void)
2530 {
2531 int i, k, p;
2532 uint32_t nhash = 0;
2533 uint32_t hash_size;
2534
2535 if (in6_ifaddrhashtbl != NULL) {
2536 return;
2537 }
2538
2539 PE_parse_boot_argn("ina6ddr_nhash", &nhash,
2540 sizeof(in6addr_nhash));
2541 if (nhash == 0) {
2542 nhash = IN6ADDR_NHASH;
2543 }
2544
2545 hash_size = nhash * sizeof(*in6_ifaddrhashtbl);
2546
2547 in6_ifaddrhashtbl = zalloc_permanent(
2548 hash_size,
2549 ZALIGN_PTR);
2550 in6addr_nhash = nhash;
2551
2552 /*
2553 * Generate the next largest prime greater than in6addr_nhash.
2554 */
2555 k = (in6addr_nhash % 2 == 0) ? in6addr_nhash + 1 : in6addr_nhash + 2;
2556 for (;;) {
2557 p = 1;
2558 for (i = 3; i * i <= k; i += 2) {
2559 if (k % i == 0) {
2560 p = 0;
2561 }
2562 }
2563 if (p == 1) {
2564 break;
2565 }
2566 k += 2;
2567 }
2568 in6addr_hashp = k;
2569 }
2570
2571 static int
2572 sysctl_ip6_checkinterface SYSCTL_HANDLER_ARGS
2573 {
2574 #pragma unused(arg1, arg2)
2575 int error, i;
2576
2577 i = ip6_checkinterface;
2578 error = sysctl_handle_int(oidp, &i, 0, req);
2579 if (error || req->newptr == USER_ADDR_NULL) {
2580 return error;
2581 }
2582
2583 switch (i) {
2584 case IP6_CHECKINTERFACE_WEAK_ES:
2585 case IP6_CHECKINTERFACE_HYBRID_ES:
2586 case IP6_CHECKINTERFACE_STRONG_ES:
2587 if (ip6_checkinterface != i) {
2588 ip6_checkinterface = i;
2589 os_log(OS_LOG_DEFAULT, "%s: ip6_checkinterface is now %d\n",
2590 __func__, ip6_checkinterface);
2591 }
2592 break;
2593 default:
2594 error = EINVAL;
2595 break;
2596 }
2597 return error;
2598 }
2599