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 static_assert(!(offsetof(struct in6_ifstat, f) % sizeof (uint64_t)))
181
182 #define ICMP6_IFSTAT_REQUIRE_ALIGNED_64(f) \
183 static_assert(!(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 protosw * pp,struct domain * dp)355 ip6_init(struct protosw *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 static_assert((sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr)) <= _MHLEN);
367
368 static_assert(IP_RECV_LINK_ADDR_TYPE == IPV6_RECV_LINK_ADDR_TYPE);
369
370 if (!os_atomic_cmpxchg(&ip6_initialized, 0, 1, relaxed)) {
371 return;
372 }
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 static void
ip6_input_process_wake_packet(struct mbuf * m)780 ip6_input_process_wake_packet(struct mbuf *m)
781 {
782 struct ifnet *ifp = m->m_pkthdr.rcvif;
783
784 if (if_is_lpw_enabled(ifp)) {
785 if_exit_lpw(ifp, "IP6 packet");
786 }
787 }
788
789 void
ip6_input(struct mbuf * m)790 ip6_input(struct mbuf *m)
791 {
792 struct ip6_hdr *ip6;
793 int off = sizeof(struct ip6_hdr), nest;
794 u_int32_t plen;
795 u_int32_t rtalert = ~0;
796 int nxt = 0, ours = 0;
797 ifnet_ref_t inifp, deliverifp = NULL;
798 ipfilter_t __single inject_ipfref = NULL;
799 int seen = 1;
800 #if DUMMYNET
801 struct m_tag *__single tag;
802 struct ip_fw_args args = {};
803 #endif /* DUMMYNET */
804 struct route_in6 rin6 = {};
805
806 /*
807 * Check if the packet we received is valid after interface filter
808 * processing
809 */
810 MBUF_INPUT_CHECK(m, m->m_pkthdr.rcvif);
811 inifp = m->m_pkthdr.rcvif;
812 VERIFY(inifp != NULL);
813
814 /* Perform IP header alignment fixup, if needed */
815 IP6_HDR_ALIGNMENT_FIXUP(m, inifp, return );
816
817 m->m_pkthdr.pkt_flags &= ~PKTF_FORWARDED;
818 #if IPSEC
819 /*
820 * should the inner packet be considered authentic?
821 * see comment in ah4_input().
822 */
823 m->m_flags &= ~M_AUTHIPHDR;
824 m->m_flags &= ~M_AUTHIPDGM;
825 #endif /* IPSEC */
826
827 /*
828 * make sure we don't have onion peering information into m_aux.
829 */
830 ip6_delaux(m);
831
832 #if DUMMYNET
833 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
834 KERNEL_TAG_TYPE_DUMMYNET)) != NULL) {
835 struct dn_pkt_tag *__single dn_tag;
836
837 dn_tag = (struct dn_pkt_tag *)(tag->m_tag_data);
838
839 args.fwa_pf_rule = dn_tag->dn_pf_rule;
840
841 m_tag_delete(m, tag);
842 }
843
844 if (args.fwa_pf_rule) {
845 ip6 = mtod(m, struct ip6_hdr *); /* In case PF got disabled */
846
847 goto check_with_pf;
848 }
849 #endif /* DUMMYNET */
850
851 /*
852 * No need to process packet twice if we've already seen it.
853 */
854 inject_ipfref = ipf_get_inject_filter(m);
855 if (inject_ipfref != NULL) {
856 ip6 = mtod(m, struct ip6_hdr *);
857 nxt = ip6->ip6_nxt;
858 seen = 0;
859 goto injectit;
860 } else {
861 seen = 1;
862 }
863
864 if (__improbable(m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
865 if_ports_used_match_mbuf(inifp, PF_INET6, m);
866 }
867
868 /*
869 * mbuf statistics
870 */
871 if (m->m_flags & M_EXT) {
872 if (m->m_next != NULL) {
873 ip6stat.ip6s_mext2m++;
874 } else {
875 ip6stat.ip6s_mext1++;
876 }
877 } else {
878 #define M2MMAX (sizeof (ip6stat.ip6s_m2m) / sizeof (ip6stat.ip6s_m2m[0]))
879 if (m->m_next != NULL) {
880 if (m->m_pkthdr.pkt_flags & PKTF_LOOP) {
881 /* XXX */
882 ip6stat.ip6s_m2m[ifnet_index(lo_ifp)]++;
883 } else if (inifp->if_index < M2MMAX) {
884 ip6stat.ip6s_m2m[inifp->if_index]++;
885 } else {
886 ip6stat.ip6s_m2m[0]++;
887 }
888 } else {
889 ip6stat.ip6s_m1++;
890 }
891 #undef M2MMAX
892 }
893
894 /*
895 * Drop the packet if IPv6 operation is disabled on the interface.
896 */
897 if (inifp->if_eflags & IFEF_IPV6_DISABLED) {
898 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_IF_IPV6_DISABLED, NULL, 0);
899 goto bad;
900 }
901
902 in6_ifstat_inc_na(inifp, ifs6_in_receive);
903 ip6stat.ip6s_total++;
904
905 /*
906 * L2 bridge code and some other code can return mbuf chain
907 * that does not conform to KAME requirement. too bad.
908 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
909 */
910 if (m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
911 mbuf_ref_t n;
912
913 MGETHDR(n, M_DONTWAIT, MT_HEADER); /* MAC-OK */
914 if (n) {
915 M_COPY_PKTHDR(n, m);
916 }
917 if (n && m->m_pkthdr.len > MHLEN) {
918 MCLGET(n, M_DONTWAIT);
919 if ((n->m_flags & M_EXT) == 0) {
920 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0);
921 n = NULL;
922 }
923 }
924 if (n == NULL) {
925 goto bad;
926 }
927
928 m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
929 n->m_len = m->m_pkthdr.len;
930 m_freem(m);
931 m = n;
932 }
933 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), { goto done; });
934
935 if (m->m_len < sizeof(struct ip6_hdr)) {
936 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
937 ip6stat.ip6s_toosmall++;
938 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
939 goto done;
940 }
941 }
942
943 ip6 = mtod(m, struct ip6_hdr *);
944
945 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
946 ip6stat.ip6s_badvers++;
947 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
948 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_BAD_VERSION, NULL, 0);
949 goto bad;
950 }
951
952 ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
953
954 /*
955 * Check against address spoofing/corruption.
956 */
957 if (!(m->m_pkthdr.pkt_flags & PKTF_LOOP) &&
958 IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src)) {
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_MULTICAST(&ip6->ip6_src) ||
965 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
966 /*
967 * XXX: "badscope" is not very suitable for a multicast source.
968 */
969 ip6stat.ip6s_badscope++;
970 in6_ifstat_inc(inifp, ifs6_in_addrerr);
971 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
972 goto bad;
973 }
974 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
975 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
976 /*
977 * In this case, the packet should come from the loopback
978 * interface. However, we cannot just check the if_flags,
979 * because ip6_mloopback() passes the "actual" interface
980 * as the outgoing/incoming interface.
981 */
982 ip6stat.ip6s_badscope++;
983 in6_ifstat_inc(inifp, ifs6_in_addrerr);
984 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
985 goto bad;
986 }
987
988 /*
989 * The following check is not documented in specs. A malicious
990 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
991 * and bypass security checks (act as if it was from 127.0.0.1 by using
992 * IPv6 src ::ffff:127.0.0.1). Be cautious.
993 *
994 * This check chokes if we are in an SIIT cloud. As none of BSDs
995 * support IPv4-less kernel compilation, we cannot support SIIT
996 * environment at all. So, it makes more sense for us to reject any
997 * malicious packets for non-SIIT environment, than try to do a
998 * partial support for SIIT environment.
999 */
1000 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
1001 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
1002 ip6stat.ip6s_badscope++;
1003 in6_ifstat_inc(inifp, ifs6_in_addrerr);
1004 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1005 goto bad;
1006 }
1007
1008 if (((ntohl(ip6->ip6_flow & IPV6_FLOW_ECN_MASK) >> 20) & IPTOS_ECN_ECT1) == IPTOS_ECN_ECT1) {
1009 m->m_pkthdr.pkt_ext_flags |= PKTF_EXT_L4S;
1010 }
1011
1012 #if 0
1013 /*
1014 * Reject packets with IPv4 compatible addresses (auto tunnel).
1015 *
1016 * The code forbids auto tunnel relay case in RFC1933 (the check is
1017 * stronger than RFC1933). We may want to re-enable it if mech-xx
1018 * is revised to forbid relaying case.
1019 */
1020 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
1021 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
1022 ip6stat.ip6s_badscope++;
1023 in6_ifstat_inc(inifp, ifs6_in_addrerr);
1024 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1025 goto bad;
1026 }
1027 #endif
1028
1029 /*
1030 * Naively assume we can attribute inbound data to the route we would
1031 * use to send to this destination. Asymetric routing breaks this
1032 * assumption, but it still allows us to account for traffic from
1033 * a remote node in the routing table.
1034 * this has a very significant performance impact so we bypass
1035 * if nstat_collect is disabled. We may also bypass if the
1036 * protocol is tcp in the future because tcp will have a route that
1037 * we can use to attribute the data to. That does mean we would not
1038 * account for forwarded tcp traffic.
1039 */
1040 if (nstat_collect) {
1041 rtentry_ref_t rte =
1042 ifnet_cached_rtlookup_inet6(inifp, &ip6->ip6_src);
1043 if (rte != NULL) {
1044 nstat_route_rx(rte, 1, m->m_pkthdr.len, 0);
1045 rtfree(rte);
1046 }
1047 }
1048
1049 #if DUMMYNET
1050 check_with_pf:
1051 #endif /* DUMMYNET */
1052 #if PF
1053 /* Invoke inbound packet filter */
1054 if (PF_IS_ENABLED) {
1055 int error;
1056 #if DUMMYNET
1057 error = pf_af_hook(inifp, NULL, &m, AF_INET6, TRUE, &args);
1058 #else /* !DUMMYNET */
1059 error = pf_af_hook(inifp, NULL, &m, AF_INET6, TRUE, NULL);
1060 #endif /* !DUMMYNET */
1061 if (error != 0 || m == NULL) {
1062 if (m != NULL) {
1063 panic("%s: unexpected packet %p",
1064 __func__, m);
1065 /* NOTREACHED */
1066 }
1067 /* Already freed by callee */
1068 goto done;
1069 }
1070 ip6 = mtod(m, struct ip6_hdr *);
1071 }
1072 #endif /* PF */
1073
1074 /* drop packets if interface ID portion is already filled */
1075 if (!(inifp->if_flags & IFF_LOOPBACK) &&
1076 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
1077 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
1078 ip6->ip6_src.s6_addr16[1]) {
1079 ip6stat.ip6s_badscope++;
1080 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1081 goto bad;
1082 }
1083 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) &&
1084 ip6->ip6_dst.s6_addr16[1]) {
1085 ip6stat.ip6s_badscope++;
1086 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1087 goto bad;
1088 }
1089 }
1090
1091 if ((m->m_pkthdr.pkt_flags & PKTF_IFAINFO) != 0 && in6_embedded_scope) {
1092 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1093 ip6->ip6_src.s6_addr16[1] =
1094 htons(m->m_pkthdr.src_ifindex);
1095 }
1096 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
1097 ip6->ip6_dst.s6_addr16[1] =
1098 htons(m->m_pkthdr.dst_ifindex);
1099 }
1100 } else if (in6_embedded_scope) {
1101 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1102 ip6->ip6_src.s6_addr16[1] = htons(inifp->if_index);
1103 }
1104 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
1105 ip6->ip6_dst.s6_addr16[1] = htons(inifp->if_index);
1106 }
1107 }
1108
1109 /*
1110 * Multicast check
1111 */
1112 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1113 struct in6_multi *__single in6m = NULL;
1114
1115 in6_ifstat_inc_na(inifp, ifs6_in_mcast);
1116 /*
1117 * See if we belong to the destination multicast group on the
1118 * arrival interface.
1119 */
1120 in6_multihead_lock_shared();
1121 IN6_LOOKUP_MULTI(&ip6->ip6_dst, inifp, in6m);
1122 in6_multihead_lock_done();
1123 if (in6m != NULL) {
1124 IN6M_REMREF(in6m);
1125 ours = 1;
1126 } else if (!nd6_prproxy) {
1127 ip6stat.ip6s_notmember++;
1128 ip6stat.ip6s_cantforward++;
1129 in6_ifstat_inc(inifp, ifs6_in_discard);
1130 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_UNKNOWN_MULTICAST_GROUP, NULL, 0);
1131 goto bad;
1132 }
1133 deliverifp = inifp;
1134 /*
1135 * record dst address information into mbuf, if we don't have one yet.
1136 * note that we are unable to record it, if the address is not listed
1137 * as our interface address (e.g. multicast addresses, etc.)
1138 */
1139 if (deliverifp != NULL) {
1140 struct in6_ifaddr *__single ia6 = NULL;
1141
1142 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
1143 if (ia6 != NULL) {
1144 (void) ip6_setdstifaddr_info(m, 0, ia6);
1145 (void) ip6_setsrcifaddr_info(m, ia6->ia_ifp->if_index, NULL);
1146 ifa_remref(&ia6->ia_ifa);
1147 } else {
1148 (void) ip6_setdstifaddr_info(m, inifp->if_index, NULL);
1149 (void) ip6_setsrcifaddr_info(m, inifp->if_index, NULL);
1150 }
1151 }
1152 goto hbhcheck;
1153 } else {
1154 /*
1155 * Unicast check
1156 */
1157 ip6_check_if_result_t check_if_result = IP6_CHECK_IF_NONE;
1158 check_if_result = ip6_input_check_interface(m, ip6, inifp, &rin6, &deliverifp);
1159 ASSERT(check_if_result != IP6_CHECK_IF_NONE);
1160 if (check_if_result == IP6_CHECK_IF_OURS) {
1161 ours = 1;
1162 goto hbhcheck;
1163 } else if (check_if_result == IP6_CHECK_IF_DROP) {
1164 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_RCV_IF_NO_MATCH, NULL, 0);
1165 goto bad;
1166 }
1167 }
1168
1169 /*
1170 * Now there is no reason to process the packet if it's not our own
1171 * and we're not a router.
1172 */
1173 if (!ip6_forwarding) {
1174 ip6stat.ip6s_cantforward++;
1175 in6_ifstat_inc(inifp, ifs6_in_discard);
1176 /*
1177 * Raise a kernel event if the packet received on cellular
1178 * interface is not intended for local host.
1179 * For now limit it to ICMPv6 packets.
1180 */
1181 if (inifp->if_type == IFT_CELLULAR &&
1182 ip6->ip6_nxt == IPPROTO_ICMPV6) {
1183 in6_ifstat_inc(inifp, ifs6_cantfoward_icmp6);
1184 }
1185 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_CANNOT_FORWARD, NULL, 0);
1186 goto bad;
1187 }
1188
1189 hbhcheck:
1190 /*
1191 * Process Hop-by-Hop options header if it's contained.
1192 * m may be modified in ip6_hopopts_input().
1193 * If a JumboPayload option is included, plen will also be modified.
1194 */
1195 plen = (u_int32_t)ntohs(ip6->ip6_plen);
1196 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1197 struct ip6_hbh *__single hbh;
1198
1199 /*
1200 * Mark the packet to imply that HBH option has been checked.
1201 * This can only be true is the packet came in unfragmented
1202 * or if the option is in the first fragment
1203 */
1204 m->m_pkthdr.pkt_flags |= PKTF_HBH_CHKED;
1205 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
1206 #if 0 /* touches NULL pointer */
1207 in6_ifstat_inc(inifp, ifs6_in_discard);
1208 #endif
1209 goto done; /* m have already been freed */
1210 }
1211
1212 /* adjust pointer */
1213 ip6 = mtod(m, struct ip6_hdr *);
1214
1215 /*
1216 * if the payload length field is 0 and the next header field
1217 * indicates Hop-by-Hop Options header, then a Jumbo Payload
1218 * option MUST be included.
1219 */
1220 if (ip6->ip6_plen == 0 && plen == 0) {
1221 /*
1222 * Note that if a valid jumbo payload option is
1223 * contained, ip6_hopopts_input() must set a valid
1224 * (non-zero) payload length to the variable plen.
1225 */
1226 ip6stat.ip6s_badoptions++;
1227 in6_ifstat_inc(inifp, ifs6_in_discard);
1228 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
1229 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
1230 (int)((caddr_t)&ip6->ip6_plen - (caddr_t)ip6));
1231 goto done;
1232 }
1233 /* ip6_hopopts_input() ensures that mbuf is contiguous */
1234 hbh = (struct ip6_hbh *)(ip6 + 1);
1235 nxt = hbh->ip6h_nxt;
1236
1237 /*
1238 * If we are acting as a router and the packet contains a
1239 * router alert option, see if we know the option value.
1240 * Currently, we only support the option value for MLD, in which
1241 * case we should pass the packet to the multicast routing
1242 * daemon.
1243 */
1244 if (rtalert != ~0 && ip6_forwarding) {
1245 switch (rtalert) {
1246 case IP6OPT_RTALERT_MLD:
1247 ours = 1;
1248 break;
1249 default:
1250 /*
1251 * RFC2711 requires unrecognized values must be
1252 * silently ignored.
1253 */
1254 break;
1255 }
1256 }
1257 } else {
1258 nxt = ip6->ip6_nxt;
1259 }
1260
1261 /*
1262 * Check that the amount of data in the buffers
1263 * is as at least much as the IPv6 header would have us expect.
1264 * Trim mbufs if longer than we expect.
1265 * Drop packet if shorter than we expect.
1266 */
1267 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
1268 ip6stat.ip6s_tooshort++;
1269 in6_ifstat_inc(inifp, ifs6_in_truncated);
1270 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SHORT, NULL, 0);
1271 goto bad;
1272 }
1273
1274 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
1275 ip6_input_adjust(m, ip6, plen, inifp);
1276 }
1277
1278 /*
1279 * Forward if desirable.
1280 */
1281 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1282 if (!ours && nd6_prproxy) {
1283 /*
1284 * If this isn't for us, this might be a Neighbor
1285 * Solicitation (dst is solicited-node multicast)
1286 * against an address in one of the proxied prefixes;
1287 * if so, claim the packet and let icmp6_input()
1288 * handle the rest.
1289 */
1290 ours = nd6_prproxy_isours(m, ip6, NULL, IFSCOPE_NONE);
1291 VERIFY(!ours ||
1292 (m->m_pkthdr.pkt_flags & PKTF_PROXY_DST));
1293 }
1294 if (!ours) {
1295 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_UNPROXIED_NS, NULL, 0);
1296 goto bad;
1297 }
1298 } else if (!ours) {
1299 /*
1300 * The unicast forwarding function might return the packet
1301 * if we are proxying prefix(es), and if the packet is an
1302 * ICMPv6 packet that has failed the zone checks, but is
1303 * targetted towards a proxied address (this is optimized by
1304 * way of RTF_PROXY test.) If so, claim the packet as ours
1305 * and let icmp6_input() handle the rest. The packet's hop
1306 * limit value is kept intact (it's not decremented). This
1307 * is for supporting Neighbor Unreachability Detection between
1308 * proxied nodes on different links (src is link-local, dst
1309 * is target address.)
1310 */
1311 if ((m = ip6_forward(m, &rin6, 0)) == NULL) {
1312 goto done;
1313 }
1314 VERIFY(rin6.ro_rt != NULL);
1315 VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
1316 deliverifp = rin6.ro_rt->rt_ifp;
1317 ours = 1;
1318 }
1319
1320 ip6 = mtod(m, struct ip6_hdr *);
1321
1322 /*
1323 * Malicious party may be able to use IPv4 mapped addr to confuse
1324 * tcp/udp stack and bypass security checks (act as if it was from
1325 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious.
1326 *
1327 * For SIIT end node behavior, you may want to disable the check.
1328 * However, you will become vulnerable to attacks using IPv4 mapped
1329 * source.
1330 */
1331 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
1332 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
1333 ip6stat.ip6s_badscope++;
1334 in6_ifstat_inc(inifp, ifs6_in_addrerr);
1335 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1336 goto bad;
1337 }
1338
1339 /*
1340 * Tell launch routine the next header
1341 */
1342 ip6stat.ip6s_delivered++;
1343 in6_ifstat_inc_na(deliverifp, ifs6_in_deliver);
1344
1345 injectit:
1346 nest = 0;
1347
1348 /*
1349 * Perform IP header alignment fixup again, if needed. Note that
1350 * we do it once for the outermost protocol, and we assume each
1351 * protocol handler wouldn't mess with the alignment afterwards.
1352 */
1353 IP6_HDR_ALIGNMENT_FIXUP(m, inifp, return );
1354
1355 while (nxt != IPPROTO_DONE) {
1356 struct ipfilter *__single filter;
1357 int (*pr_input)(struct mbuf **, int *, int);
1358
1359 /*
1360 * This would imply either IPPROTO_HOPOPTS was not the first
1361 * option or it did not come in the first fragment.
1362 */
1363 if (nxt == IPPROTO_HOPOPTS &&
1364 (m->m_pkthdr.pkt_flags & PKTF_HBH_CHKED) == 0) {
1365 /*
1366 * This implies that HBH option was not contained
1367 * in the first fragment
1368 */
1369 ip6stat.ip6s_badoptions++;
1370 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_OPTION, NULL, 0);
1371 goto bad;
1372 }
1373
1374 if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
1375 ip6stat.ip6s_toomanyhdr++;
1376 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_TOO_MANY_OPTIONS, NULL, 0);
1377 goto bad;
1378 }
1379
1380 /*
1381 * protection against faulty packet - there should be
1382 * more sanity checks in header chain processing.
1383 */
1384 if (m->m_pkthdr.len < off) {
1385 ip6stat.ip6s_tooshort++;
1386 in6_ifstat_inc(inifp, ifs6_in_truncated);
1387 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SHORT, NULL, 0);
1388 goto bad;
1389 }
1390
1391 #if IPSEC
1392 /*
1393 * enforce IPsec policy checking if we are seeing last header.
1394 * note that we do not visit this with protocols with pcb layer
1395 * code - like udp/tcp/raw ip.
1396 */
1397 if ((ipsec_bypass == 0) &&
1398 (ip6_protox[nxt]->pr_flags & PR_LASTHDR) != 0) {
1399 if (ipsec6_in_reject(m, NULL)) {
1400 IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
1401 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IPSEC_REJECT, NULL, 0);
1402 goto bad;
1403 }
1404 }
1405 #endif /* IPSEC */
1406
1407 /*
1408 * Call IP filter
1409 */
1410 if (!TAILQ_EMPTY(&ipv6_filters) && !IFNET_IS_INTCOPROC(inifp)) {
1411 ipf_ref();
1412 TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) {
1413 if (seen == 0) {
1414 if ((struct ipfilter *)inject_ipfref ==
1415 filter) {
1416 seen = 1;
1417 }
1418 } else if (filter->ipf_filter.ipf_input) {
1419 errno_t result;
1420
1421 result = filter->ipf_filter.ipf_input(
1422 filter->ipf_filter.cookie,
1423 (mbuf_t *)&m, off, (uint8_t)nxt);
1424 if (result == EJUSTRETURN) {
1425 ipf_unref();
1426 goto done;
1427 }
1428 if (result != 0) {
1429 ipf_unref();
1430 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_FILTER_DROP, NULL, 0);
1431 goto bad;
1432 }
1433 }
1434 }
1435 ipf_unref();
1436 }
1437
1438 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
1439 struct ip6_hdr *, ip6, struct ifnet *, inifp,
1440 struct ip *, NULL, struct ip6_hdr *, ip6);
1441
1442 /*
1443 * Check if need to switch to full wake mode -- TCP knows about idle connections
1444 */
1445 if (__improbable(nxt != IPPROTO_TCP && (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT) != 0)) {
1446 ip6_input_process_wake_packet(m);
1447 }
1448
1449 if ((pr_input = ip6_protox[nxt]->pr_input) == NULL) {
1450 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_NO_PROTO, NULL, 0);
1451 m = NULL;
1452 nxt = IPPROTO_DONE;
1453 } else if (!(ip6_protox[nxt]->pr_flags & PR_PROTOLOCK)) {
1454 lck_mtx_lock(inet6_domain_mutex);
1455 nxt = pr_input(&m, &off, nxt);
1456 lck_mtx_unlock(inet6_domain_mutex);
1457 } else {
1458 nxt = pr_input(&m, &off, nxt);
1459 }
1460 }
1461 done:
1462 ROUTE_RELEASE(&rin6);
1463 return;
1464 bad:
1465 goto done;
1466 }
1467
1468 void
ip6_setsrcifaddr_info(struct mbuf * m,uint32_t src_idx,struct in6_ifaddr * ia6)1469 ip6_setsrcifaddr_info(struct mbuf *m, uint32_t src_idx, struct in6_ifaddr *ia6)
1470 {
1471 VERIFY(m->m_flags & M_PKTHDR);
1472 m->m_pkthdr.pkt_ext_flags &= ~PKTF_EXT_OUTPUT_SCOPE;
1473 /*
1474 * If the source ifaddr is specified, pick up the information
1475 * from there; otherwise just grab the passed-in ifindex as the
1476 * caller may not have the ifaddr available.
1477 */
1478 if (ia6 != NULL) {
1479 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1480 m->m_pkthdr.src_ifindex = ia6->ia_ifp->if_index;
1481
1482 /* See IN6_IFF comments in in6_var.h */
1483 m->m_pkthdr.src_iff = (ia6->ia6_flags & 0xffff);
1484 } else {
1485 m->m_pkthdr.src_iff = 0;
1486 m->m_pkthdr.src_ifindex = (uint16_t)src_idx;
1487 if (src_idx != 0) {
1488 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1489 }
1490 }
1491 }
1492
1493 void
ip6_setdstifaddr_info(struct mbuf * m,uint32_t dst_idx,struct in6_ifaddr * ia6)1494 ip6_setdstifaddr_info(struct mbuf *m, uint32_t dst_idx, struct in6_ifaddr *ia6)
1495 {
1496 VERIFY(m->m_flags & M_PKTHDR);
1497 m->m_pkthdr.pkt_ext_flags &= ~PKTF_EXT_OUTPUT_SCOPE;
1498
1499 /*
1500 * If the destination ifaddr is specified, pick up the information
1501 * from there; otherwise just grab the passed-in ifindex as the
1502 * caller may not have the ifaddr available.
1503 */
1504 if (ia6 != NULL) {
1505 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1506 m->m_pkthdr.dst_ifindex = ia6->ia_ifp->if_index;
1507
1508 /* See IN6_IFF comments in in6_var.h */
1509 m->m_pkthdr.dst_iff = (ia6->ia6_flags & 0xffff);
1510 } else {
1511 m->m_pkthdr.dst_iff = 0;
1512 m->m_pkthdr.dst_ifindex = (uint16_t)dst_idx;
1513 if (dst_idx != 0) {
1514 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1515 }
1516 }
1517 }
1518
1519 int
ip6_getsrcifaddr_info(struct mbuf * m,uint32_t * src_idx,uint32_t * ia6f)1520 ip6_getsrcifaddr_info(struct mbuf *m, uint32_t *src_idx, uint32_t *ia6f)
1521 {
1522 VERIFY(m->m_flags & M_PKTHDR);
1523
1524 if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO)) {
1525 return -1;
1526 }
1527
1528 if (src_idx != NULL) {
1529 *src_idx = m->m_pkthdr.src_ifindex;
1530 }
1531
1532 if (ia6f != NULL) {
1533 *ia6f = m->m_pkthdr.src_iff;
1534 }
1535
1536 return 0;
1537 }
1538
1539 int
ip6_getdstifaddr_info(struct mbuf * m,uint32_t * dst_idx,uint32_t * ia6f)1540 ip6_getdstifaddr_info(struct mbuf *m, uint32_t *dst_idx, uint32_t *ia6f)
1541 {
1542 VERIFY(m->m_flags & M_PKTHDR);
1543
1544 if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO)) {
1545 return -1;
1546 }
1547
1548 if (dst_idx != NULL) {
1549 *dst_idx = m->m_pkthdr.dst_ifindex;
1550 }
1551
1552 if (ia6f != NULL) {
1553 *ia6f = m->m_pkthdr.dst_iff;
1554 }
1555
1556 return 0;
1557 }
1558
1559 uint32_t
ip6_input_getsrcifscope(struct mbuf * m)1560 ip6_input_getsrcifscope(struct mbuf *m)
1561 {
1562 VERIFY(m->m_flags & M_PKTHDR);
1563
1564 if (m->m_pkthdr.rcvif != NULL) {
1565 return m->m_pkthdr.rcvif->if_index;
1566 }
1567
1568 uint32_t src_ifscope = IFSCOPE_NONE;
1569 ip6_getsrcifaddr_info(m, &src_ifscope, NULL);
1570 return src_ifscope;
1571 }
1572
1573 uint32_t
ip6_input_getdstifscope(struct mbuf * m)1574 ip6_input_getdstifscope(struct mbuf *m)
1575 {
1576 VERIFY(m->m_flags & M_PKTHDR);
1577
1578 if (m->m_pkthdr.rcvif != NULL) {
1579 return m->m_pkthdr.rcvif->if_index;
1580 }
1581
1582 uint32_t dst_ifscope = IFSCOPE_NONE;
1583 ip6_getdstifaddr_info(m, &dst_ifscope, NULL);
1584 return dst_ifscope;
1585 }
1586
1587 /*
1588 * Hop-by-Hop options header processing. If a valid jumbo payload option is
1589 * included, the real payload length will be stored in plenp.
1590 */
1591 static int
ip6_hopopts_input(uint32_t * plenp,uint32_t * rtalertp,struct mbuf ** mp,int * offp)1592 ip6_hopopts_input(uint32_t *plenp, uint32_t *rtalertp, struct mbuf **mp,
1593 int *offp)
1594 {
1595 mbuf_ref_t m = *mp;
1596 int off = *offp, hbhlen;
1597 struct ip6_hbh *hbh;
1598 u_int8_t *opt;
1599
1600 /* validation of the length of the header */
1601 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), return (-1));
1602 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1603 hbhlen = (hbh->ip6h_len + 1) << 3;
1604
1605 IP6_EXTHDR_CHECK(m, off, hbhlen, return (-1));
1606 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1607 off += hbhlen;
1608 hbhlen -= sizeof(struct ip6_hbh);
1609 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1610
1611 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1612 hbhlen, rtalertp, plenp) < 0) {
1613 return -1;
1614 }
1615
1616 *offp = off;
1617 *mp = m;
1618 return 0;
1619 }
1620
1621 /*
1622 * Search header for all Hop-by-hop options and process each option.
1623 * This function is separate from ip6_hopopts_input() in order to
1624 * handle a case where the sending node itself process its hop-by-hop
1625 * options header. In such a case, the function is called from ip6_output().
1626 *
1627 * The function assumes that hbh header is located right after the IPv6 header
1628 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1629 * opthead + hbhlen is located in continuous memory region.
1630 */
1631 int
ip6_process_hopopts(struct mbuf * m,u_int8_t * __sized_by (inhbhlen)opthead,int inhbhlen,u_int32_t * rtalertp,u_int32_t * plenp)1632 ip6_process_hopopts(struct mbuf *m, u_int8_t *__sized_by(inhbhlen)opthead, int inhbhlen,
1633 u_int32_t *rtalertp, u_int32_t *plenp)
1634 {
1635 struct ip6_hdr *__single ip6;
1636 int optlen = 0;
1637 int hbhlen = inhbhlen;
1638 u_int8_t *opt = opthead;
1639 u_int16_t rtalert_val;
1640 u_int32_t jumboplen;
1641 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1642
1643 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1644 switch (*opt) {
1645 case IP6OPT_PAD1:
1646 optlen = 1;
1647 break;
1648 case IP6OPT_PADN:
1649 if (hbhlen < IP6OPT_MINLEN) {
1650 ip6stat.ip6s_toosmall++;
1651 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0);
1652 goto bad;
1653 }
1654 optlen = *(opt + 1) + 2;
1655 break;
1656 case IP6OPT_ROUTER_ALERT:
1657 /* XXX may need check for alignment */
1658 if (hbhlen < IP6OPT_RTALERT_LEN) {
1659 ip6stat.ip6s_toosmall++;
1660 goto bad;
1661 }
1662 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1663 /* XXX stat */
1664 icmp6_error(m, ICMP6_PARAM_PROB,
1665 ICMP6_PARAMPROB_HEADER,
1666 (int)(erroff + opt + 1 - opthead));
1667 return -1;
1668 }
1669 optlen = IP6OPT_RTALERT_LEN;
1670 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1671 *rtalertp = ntohs(rtalert_val);
1672 break;
1673 case IP6OPT_JUMBO:
1674 /* XXX may need check for alignment */
1675 if (hbhlen < IP6OPT_JUMBO_LEN) {
1676 ip6stat.ip6s_toosmall++;
1677 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0);
1678 goto bad;
1679 }
1680 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1681 /* XXX stat */
1682 icmp6_error(m, ICMP6_PARAM_PROB,
1683 ICMP6_PARAMPROB_HEADER,
1684 (int)(erroff + opt + 1 - opthead));
1685 return -1;
1686 }
1687 optlen = IP6OPT_JUMBO_LEN;
1688
1689 /*
1690 * IPv6 packets that have non 0 payload length
1691 * must not contain a jumbo payload option.
1692 */
1693 ip6 = mtod(m, struct ip6_hdr *);
1694 if (ip6->ip6_plen) {
1695 ip6stat.ip6s_badoptions++;
1696 icmp6_error(m, ICMP6_PARAM_PROB,
1697 ICMP6_PARAMPROB_HEADER,
1698 (int)(erroff + opt - opthead));
1699 return -1;
1700 }
1701
1702 /*
1703 * We may see jumbolen in unaligned location, so
1704 * we'd need to perform bcopy().
1705 */
1706 bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1707 jumboplen = (u_int32_t)htonl(jumboplen);
1708
1709 #if 1
1710 /*
1711 * if there are multiple jumbo payload options,
1712 * *plenp will be non-zero and the packet will be
1713 * rejected.
1714 * the behavior may need some debate in ipngwg -
1715 * multiple options does not make sense, however,
1716 * there's no explicit mention in specification.
1717 */
1718 if (*plenp != 0) {
1719 ip6stat.ip6s_badoptions++;
1720 icmp6_error(m, ICMP6_PARAM_PROB,
1721 ICMP6_PARAMPROB_HEADER,
1722 (int)(erroff + opt + 2 - opthead));
1723 return -1;
1724 }
1725 #endif
1726
1727 /*
1728 * jumbo payload length must be larger than 65535.
1729 */
1730 if (jumboplen <= IPV6_MAXPACKET) {
1731 ip6stat.ip6s_badoptions++;
1732 icmp6_error(m, ICMP6_PARAM_PROB,
1733 ICMP6_PARAMPROB_HEADER,
1734 (int)(erroff + opt + 2 - opthead));
1735 return -1;
1736 }
1737 *plenp = jumboplen;
1738
1739 break;
1740 default: /* unknown option */
1741 if (hbhlen < IP6OPT_MINLEN) {
1742 ip6stat.ip6s_toosmall++;
1743 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_TOO_SMALL, NULL, 0);
1744 goto bad;
1745 }
1746 optlen = ip6_unknown_opt(opt, hbhlen, m,
1747 erroff + opt - opthead);
1748 if (optlen == -1) {
1749 return -1;
1750 }
1751 optlen += 2;
1752 break;
1753 }
1754 }
1755
1756 return 0;
1757
1758 bad:
1759 return -1;
1760 }
1761
1762 /*
1763 * Unknown option processing.
1764 * The fourth argument `off' is the offset from the IPv6 header to the option,
1765 * which is necessary if the IPv6 header the and option header and IPv6 header
1766 * is not continuous in order to return an ICMPv6 error.
1767 */
1768 int
ip6_unknown_opt(uint8_t * __counted_by (optplen)optp,size_t optplen,struct mbuf * m,size_t off)1769 ip6_unknown_opt(uint8_t *__counted_by(optplen) optp, size_t optplen, struct mbuf *m, size_t off)
1770 {
1771 #pragma unused(optplen)
1772
1773 struct ip6_hdr *__single ip6;
1774
1775 switch (IP6OPT_TYPE(*optp)) {
1776 case IP6OPT_TYPE_SKIP: /* ignore the option */
1777 return (int)*(optp + 1);
1778
1779 case IP6OPT_TYPE_DISCARD: /* silently discard */
1780 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_OPT_DISCARD, NULL, 0);
1781 return -1;
1782
1783 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1784 ip6stat.ip6s_badoptions++;
1785 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, (int)off);
1786 return -1;
1787
1788 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1789 ip6stat.ip6s_badoptions++;
1790 ip6 = mtod(m, struct ip6_hdr *);
1791 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1792 (m->m_flags & (M_BCAST | M_MCAST))) {
1793 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_OPTION, NULL, 0);
1794 } else {
1795 icmp6_error(m, ICMP6_PARAM_PROB,
1796 ICMP6_PARAMPROB_OPTION, (int)off);
1797 }
1798 return -1;
1799 }
1800
1801 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_UNSPECIFIED, NULL, 0); /* XXX: NOTREACHED */
1802 return -1;
1803 }
1804
1805 /*
1806 * Create the "control" list for this pcb.
1807 * These functions will not modify mbuf chain at all.
1808 *
1809 * With KAME mbuf chain restriction:
1810 * The routine will be called from upper layer handlers like tcp6_input().
1811 * Thus the routine assumes that the caller (tcp6_input) have already
1812 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1813 * very first mbuf on the mbuf chain.
1814 *
1815 * ip6_savecontrol_v4 will handle those options that are possible to be
1816 * set on a v4-mapped socket.
1817 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1818 * options and handle the v6-only ones itself.
1819 */
1820 struct mbuf **
ip6_savecontrol_v4(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp,int * v4only)1821 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1822 int *v4only)
1823 {
1824 struct ip6_hdr *__single ip6 = mtod(m, struct ip6_hdr *);
1825
1826 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1827 struct timeval tv;
1828
1829 getmicrotime(&tv);
1830 mp = sbcreatecontrol_mbuf((caddr_t)&tv, sizeof(tv),
1831 SCM_TIMESTAMP, SOL_SOCKET, mp);
1832 if (*mp == NULL) {
1833 return NULL;
1834 }
1835 }
1836 if ((inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) {
1837 uint64_t time;
1838
1839 time = mach_absolute_time();
1840 mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof(time),
1841 SCM_TIMESTAMP_MONOTONIC, SOL_SOCKET, mp);
1842 if (*mp == NULL) {
1843 return NULL;
1844 }
1845 }
1846 if ((inp->inp_socket->so_options & SO_TIMESTAMP_CONTINUOUS) != 0) {
1847 uint64_t time;
1848
1849 time = mach_continuous_time();
1850 mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof(time),
1851 SCM_TIMESTAMP_CONTINUOUS, SOL_SOCKET, mp);
1852 if (*mp == NULL) {
1853 return NULL;
1854 }
1855 }
1856 if ((inp->inp_socket->so_flags & SOF_RECV_TRAFFIC_CLASS) != 0) {
1857 int tc = m_get_traffic_class(m);
1858
1859 mp = sbcreatecontrol_mbuf((caddr_t)&tc, sizeof(tc),
1860 SO_TRAFFIC_CLASS, SOL_SOCKET, mp);
1861 if (*mp == NULL) {
1862 return NULL;
1863 }
1864 }
1865
1866 if ((inp->inp_socket->so_flags & SOF_RECV_WAKE_PKT) &&
1867 (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
1868 int flag = 1;
1869
1870 mp = sbcreatecontrol_mbuf((caddr_t)&flag, sizeof(flag),
1871 SO_RECV_WAKE_PKT, SOL_SOCKET, mp);
1872 if (*mp == NULL) {
1873 return NULL;
1874 }
1875 }
1876
1877 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1878 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1879 if (v4only != NULL) {
1880 *v4only = 1;
1881 }
1882
1883 // Send ECN flags for v4-mapped addresses
1884 if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1885 struct ip *__single ip_header = mtod(m, struct ip *);
1886
1887 int tclass = (int)(ip_header->ip_tos);
1888 mp = sbcreatecontrol_mbuf((caddr_t)&tclass, sizeof(tclass),
1889 IPV6_TCLASS, IPPROTO_IPV6, mp);
1890 if (*mp == NULL) {
1891 return NULL;
1892 }
1893 }
1894
1895 // Send IN6P_PKTINFO for v4-mapped address
1896 if ((inp->inp_flags & IN6P_PKTINFO) != 0 || SOFLOW_ENABLED(inp->inp_socket)) {
1897 struct in6_pktinfo pi6 = {
1898 .ipi6_addr = IN6ADDR_V4MAPPED_INIT,
1899 .ipi6_ifindex = (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0,
1900 };
1901
1902 struct ip *__single ip_header = mtod(m, struct ip *);
1903 bcopy(&ip_header->ip_dst, &pi6.ipi6_addr.s6_addr32[3], sizeof(struct in_addr));
1904
1905 mp = sbcreatecontrol_mbuf((caddr_t)&pi6,
1906 sizeof(struct in6_pktinfo),
1907 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO),
1908 IPPROTO_IPV6, mp);
1909 if (*mp == NULL) {
1910 return NULL;
1911 }
1912 }
1913 return mp;
1914 }
1915
1916 /* RFC 2292 sec. 5 */
1917 if ((inp->inp_flags & IN6P_PKTINFO) != 0 || SOFLOW_ENABLED(inp->inp_socket)) {
1918 struct in6_pktinfo pi6;
1919
1920 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1921 in6_clearscope(&pi6.ipi6_addr); /* XXX */
1922 pi6.ipi6_ifindex =
1923 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1924
1925 mp = sbcreatecontrol_mbuf((caddr_t)&pi6,
1926 sizeof(struct in6_pktinfo),
1927 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO),
1928 IPPROTO_IPV6, mp);
1929 if (*mp == NULL) {
1930 return NULL;
1931 }
1932 }
1933
1934 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1935 int hlim = ip6->ip6_hlim & 0xff;
1936
1937 mp = sbcreatecontrol_mbuf((caddr_t)&hlim, sizeof(int),
1938 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1939 IPPROTO_IPV6, mp);
1940 if (*mp == NULL) {
1941 return NULL;
1942 }
1943 }
1944
1945 if (inp->inp_flags2 & INP2_RECV_LINK_ADDR_TYPE) {
1946 int mode = IP_RECV_LINK_ADDR_UNICAST;
1947
1948 /* There is no broadcast for IPv6 */
1949 if (m->m_flags & M_MCAST) {
1950 mode = IP_RECV_LINK_ADDR_MULTICAST;
1951 }
1952
1953 mp = sbcreatecontrol_mbuf((caddr_t)&mode,
1954 sizeof(int), IPV6_RECV_LINK_ADDR_TYPE, IPPROTO_IPV6, mp);
1955 if (*mp == NULL) {
1956 return NULL;
1957 }
1958 }
1959
1960 if (v4only != NULL) {
1961 *v4only = 0;
1962 }
1963 return mp;
1964 }
1965
1966 int
ip6_savecontrol(struct inpcb * in6p,struct mbuf * m,struct mbuf ** mp)1967 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1968 {
1969 struct mbuf **np;
1970 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1971 int v4only = 0;
1972
1973 *mp = NULL;
1974 np = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1975 if (np == NULL) {
1976 goto no_mbufs;
1977 }
1978
1979 mp = np;
1980 if (v4only) {
1981 return 0;
1982 }
1983
1984 if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1985 u_int32_t flowinfo;
1986 int tclass;
1987
1988 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1989 flowinfo >>= 20;
1990
1991 tclass = flowinfo & 0xff;
1992 mp = sbcreatecontrol_mbuf((caddr_t)&tclass, sizeof(tclass),
1993 IPV6_TCLASS, IPPROTO_IPV6, mp);
1994 if (*mp == NULL) {
1995 goto no_mbufs;
1996 }
1997 }
1998
1999 /*
2000 * IPV6_HOPOPTS socket option. Recall that we required super-user
2001 * privilege for the option (see ip6_ctloutput), but it might be too
2002 * strict, since there might be some hop-by-hop options which can be
2003 * returned to normal user.
2004 * See also RFC 2292 section 6 (or RFC 3542 section 8).
2005 */
2006 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
2007 /*
2008 * Check if a hop-by-hop options header is contatined in the
2009 * received packet, and if so, store the options as ancillary
2010 * data. Note that a hop-by-hop options header must be
2011 * just after the IPv6 header, which is assured through the
2012 * IPv6 input processing.
2013 */
2014 ip6 = mtod(m, struct ip6_hdr *);
2015 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
2016 struct ip6_hbh *hbh;
2017 int hbhlen = 0;
2018 hbh = (struct ip6_hbh *)(ip6 + 1);
2019 hbhlen = (hbh->ip6h_len + 1) << 3;
2020
2021 /*
2022 * XXX: We copy the whole header even if a
2023 * jumbo payload option is included, the option which
2024 * is to be removed before returning according to
2025 * RFC2292.
2026 * Note: this constraint is removed in RFC3542
2027 */
2028 mp = sbcreatecontrol_mbuf((caddr_t)hbh, hbhlen,
2029 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
2030 IPPROTO_IPV6, mp);
2031
2032 if (*mp == NULL) {
2033 goto no_mbufs;
2034 }
2035 }
2036 }
2037
2038 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
2039 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
2040
2041 /*
2042 * Search for destination options headers or routing
2043 * header(s) through the header chain, and stores each
2044 * header as ancillary data.
2045 * Note that the order of the headers remains in
2046 * the chain of ancillary data.
2047 */
2048 while (1) { /* is explicit loop prevention necessary? */
2049 struct ip6_ext *ip6e = NULL;
2050 int elen;
2051
2052 /*
2053 * if it is not an extension header, don't try to
2054 * pull it from the chain.
2055 */
2056 switch (nxt) {
2057 case IPPROTO_DSTOPTS:
2058 case IPPROTO_ROUTING:
2059 case IPPROTO_HOPOPTS:
2060 case IPPROTO_AH: /* is it possible? */
2061 break;
2062 default:
2063 goto loopend;
2064 }
2065
2066 if (off + sizeof(*ip6e) > m->m_len) {
2067 goto loopend;
2068 }
2069 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
2070 if (nxt == IPPROTO_AH) {
2071 elen = (ip6e->ip6e_len + 2) << 2;
2072 } else {
2073 elen = (ip6e->ip6e_len + 1) << 3;
2074 }
2075 if (off + elen > m->m_len) {
2076 goto loopend;
2077 }
2078
2079 switch (nxt) {
2080 case IPPROTO_DSTOPTS:
2081 if (!(in6p->inp_flags & IN6P_DSTOPTS)) {
2082 break;
2083 }
2084
2085 mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
2086 IS2292(in6p, IPV6_2292DSTOPTS,
2087 IPV6_DSTOPTS), IPPROTO_IPV6, mp);
2088 if (*mp == NULL) {
2089 goto no_mbufs;
2090 }
2091 break;
2092 case IPPROTO_ROUTING:
2093 if (!(in6p->inp_flags & IN6P_RTHDR)) {
2094 break;
2095 }
2096
2097 mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
2098 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
2099 IPPROTO_IPV6, mp);
2100 if (*mp == NULL) {
2101 goto no_mbufs;
2102 }
2103 break;
2104 case IPPROTO_HOPOPTS:
2105 case IPPROTO_AH: /* is it possible? */
2106 break;
2107
2108 default:
2109 /*
2110 * other cases have been filtered in the above.
2111 * none will visit this case. here we supply
2112 * the code just in case (nxt overwritten or
2113 * other cases).
2114 */
2115 goto loopend;
2116 }
2117
2118 /* proceed with the next header. */
2119 off += elen;
2120 nxt = ip6e->ip6e_nxt;
2121 ip6e = NULL;
2122 }
2123 loopend:
2124 ;
2125 }
2126 return 0;
2127 no_mbufs:
2128 ip6stat.ip6s_pktdropcntrl++;
2129 /* XXX increment a stat to show the failure */
2130 return ENOBUFS;
2131 }
2132 #undef IS2292
2133
2134 void
ip6_notify_pmtu(struct inpcb * in6p,struct sockaddr_in6 * dst,u_int32_t * mtu)2135 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
2136 {
2137 struct socket *__single so;
2138 mbuf_ref_t m_mtu;
2139 struct ip6_mtuinfo mtuctl;
2140
2141 so = in6p->inp_socket;
2142
2143 if ((in6p->inp_flags & IN6P_MTU) == 0) {
2144 return;
2145 }
2146
2147 if (mtu == NULL) {
2148 return;
2149 }
2150
2151 if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && SOCK_CHECK_PROTO(so, IPPROTO_TCP)) {
2152 return;
2153 }
2154
2155 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2156 !in6_are_addr_equal_scoped(&in6p->in6p_faddr, &dst->sin6_addr, in6p->inp_fifscope, dst->sin6_scope_id)) {
2157 return;
2158 }
2159
2160 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */
2161 mtuctl.ip6m_mtu = *mtu;
2162 mtuctl.ip6m_addr = *dst;
2163 if (!in6_embedded_scope) {
2164 mtuctl.ip6m_addr.sin6_scope_id = dst->sin6_scope_id;
2165 }
2166 if (sa6_recoverscope(&mtuctl.ip6m_addr, TRUE)) {
2167 return;
2168 }
2169
2170 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
2171 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) {
2172 return;
2173 }
2174
2175 if (sbappendaddr(&so->so_rcv, SA(dst), NULL, m_mtu, NULL) == 0) {
2176 return;
2177 }
2178 sorwakeup(so);
2179 }
2180
2181 /*
2182 * Get pointer to the previous header followed by the header
2183 * currently processed.
2184 * XXX: This function supposes that
2185 * M includes all headers,
2186 * the next header field and the header length field of each header
2187 * are valid, and
2188 * the sum of each header length equals to OFF.
2189 * Because of these assumptions, this function must be called very
2190 * carefully. Moreover, it will not be used in the near future when
2191 * we develop `neater' mechanism to process extension headers.
2192 */
2193 char *
ip6_get_prevhdr(struct mbuf * m,int off)2194 ip6_get_prevhdr(struct mbuf *m, int off)
2195 {
2196 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2197
2198 if (off == sizeof(struct ip6_hdr)) {
2199 return (char *)&ip6->ip6_nxt;
2200 } else {
2201 int len, nxt;
2202 struct ip6_ext *ip6e = NULL;
2203
2204 nxt = ip6->ip6_nxt;
2205 len = sizeof(struct ip6_hdr);
2206 while (len < off) {
2207 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
2208
2209 switch (nxt) {
2210 case IPPROTO_FRAGMENT:
2211 len += sizeof(struct ip6_frag);
2212 break;
2213 case IPPROTO_AH:
2214 len += (ip6e->ip6e_len + 2) << 2;
2215 break;
2216 default:
2217 len += (ip6e->ip6e_len + 1) << 3;
2218 break;
2219 }
2220 nxt = ip6e->ip6e_nxt;
2221 }
2222 if (ip6e) {
2223 return (char *)&ip6e->ip6e_nxt;
2224 } else {
2225 return NULL;
2226 }
2227 }
2228 }
2229
2230 /*
2231 * get next header offset. m will be retained.
2232 */
2233 int
ip6_nexthdr(struct mbuf * m,int off,int proto,int * nxtp)2234 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
2235 {
2236 struct ip6_hdr ip6;
2237 struct ip6_ext ip6e;
2238 struct ip6_frag fh;
2239
2240 /* just in case */
2241 VERIFY(m != NULL);
2242 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) {
2243 return -1;
2244 }
2245
2246 switch (proto) {
2247 case IPPROTO_IPV6:
2248 if (m->m_pkthdr.len < off + sizeof(ip6)) {
2249 return -1;
2250 }
2251 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
2252 if (nxtp) {
2253 *nxtp = ip6.ip6_nxt;
2254 }
2255 off += sizeof(ip6);
2256 return off;
2257
2258 case IPPROTO_FRAGMENT:
2259 /*
2260 * terminate parsing if it is not the first fragment,
2261 * it does not make sense to parse through it.
2262 */
2263 if (m->m_pkthdr.len < off + sizeof(fh)) {
2264 return -1;
2265 }
2266 m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
2267 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
2268 if (fh.ip6f_offlg & IP6F_OFF_MASK) {
2269 return -1;
2270 }
2271 if (nxtp) {
2272 *nxtp = fh.ip6f_nxt;
2273 }
2274 off += sizeof(struct ip6_frag);
2275 return off;
2276
2277 case IPPROTO_AH:
2278 if (m->m_pkthdr.len < off + sizeof(ip6e)) {
2279 return -1;
2280 }
2281 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
2282 if (nxtp) {
2283 *nxtp = ip6e.ip6e_nxt;
2284 }
2285 off += (ip6e.ip6e_len + 2) << 2;
2286 return off;
2287
2288 case IPPROTO_HOPOPTS:
2289 case IPPROTO_ROUTING:
2290 case IPPROTO_DSTOPTS:
2291 if (m->m_pkthdr.len < off + sizeof(ip6e)) {
2292 return -1;
2293 }
2294 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
2295 if (nxtp) {
2296 *nxtp = ip6e.ip6e_nxt;
2297 }
2298 off += (ip6e.ip6e_len + 1) << 3;
2299 return off;
2300
2301 case IPPROTO_NONE:
2302 case IPPROTO_ESP:
2303 case IPPROTO_IPCOMP:
2304 /* give up */
2305 return -1;
2306
2307 default:
2308 return -1;
2309 }
2310 }
2311
2312 /*
2313 * get offset for the last header in the chain. m will be kept untainted.
2314 */
2315 int
ip6_lasthdr(struct mbuf * m,int off,int proto,int * nxtp)2316 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
2317 {
2318 int newoff;
2319 int nxt;
2320
2321 if (!nxtp) {
2322 nxt = -1;
2323 nxtp = &nxt;
2324 }
2325 while (1) {
2326 newoff = ip6_nexthdr(m, off, proto, nxtp);
2327 if (newoff < 0) {
2328 return off;
2329 } else if (newoff < off) {
2330 return -1; /* invalid */
2331 } else if (newoff == off) {
2332 return newoff;
2333 }
2334
2335 off = newoff;
2336 proto = *nxtp;
2337 }
2338 }
2339
2340 boolean_t
ip6_pkt_has_ulp(struct mbuf * m)2341 ip6_pkt_has_ulp(struct mbuf *m)
2342 {
2343 int off = 0, nxt = IPPROTO_NONE;
2344
2345 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
2346 if (off < 0 || m->m_pkthdr.len < off) {
2347 return FALSE;
2348 }
2349
2350 switch (nxt) {
2351 case IPPROTO_TCP:
2352 if (off + sizeof(struct tcphdr) > m->m_pkthdr.len) {
2353 return FALSE;
2354 }
2355 break;
2356 case IPPROTO_UDP:
2357 if (off + sizeof(struct udphdr) > m->m_pkthdr.len) {
2358 return FALSE;
2359 }
2360 break;
2361 case IPPROTO_ICMPV6:
2362 if (off + sizeof(uint32_t) > m->m_pkthdr.len) {
2363 return FALSE;
2364 }
2365 break;
2366 case IPPROTO_NONE:
2367 return TRUE;
2368 case IPPROTO_ESP:
2369 return TRUE;
2370 case IPPROTO_IPCOMP:
2371 return TRUE;
2372 default:
2373 return FALSE;
2374 }
2375 return TRUE;
2376 }
2377
2378 struct ip6aux *
ip6_addaux(struct mbuf * m)2379 ip6_addaux(struct mbuf *m)
2380 {
2381 struct m_tag *__single tag;
2382
2383 /* Check if one is already allocated */
2384 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2385 KERNEL_TAG_TYPE_INET6);
2386 if (tag == NULL) {
2387 /* Allocate a tag */
2388 tag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6,
2389 sizeof(struct ip6aux), M_DONTWAIT, m);
2390
2391 /* Attach it to the mbuf */
2392 if (tag) {
2393 m_tag_prepend(m, tag);
2394 }
2395 }
2396
2397 return tag ? (struct ip6aux *)(tag->m_tag_data) : NULL;
2398 }
2399
2400 struct ip6aux *
ip6_findaux(struct mbuf * m)2401 ip6_findaux(struct mbuf *m)
2402 {
2403 struct m_tag *__single tag;
2404
2405 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2406 KERNEL_TAG_TYPE_INET6);
2407
2408 return tag != NULL ? (struct ip6aux *)(tag->m_tag_data) : NULL;
2409 }
2410
2411 void
ip6_delaux(struct mbuf * m)2412 ip6_delaux(struct mbuf *m)
2413 {
2414 struct m_tag *__single tag;
2415
2416 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2417 KERNEL_TAG_TYPE_INET6);
2418 if (tag != NULL) {
2419 m_tag_delete(m, tag);
2420 }
2421 }
2422
2423 struct inet6_tag_container {
2424 struct m_tag inet6_m_tag;
2425 struct ip6aux inet6_ip6a;
2426 };
2427
2428 struct m_tag *
m_tag_kalloc_inet6(u_int32_t id,u_int16_t type,uint16_t len,int wait)2429 m_tag_kalloc_inet6(u_int32_t id, u_int16_t type, uint16_t len, int wait)
2430 {
2431 struct inet6_tag_container *tag_container;
2432 struct m_tag *tag = NULL;
2433
2434 assert3u(id, ==, KERNEL_MODULE_TAG_ID);
2435 assert3u(type, ==, KERNEL_TAG_TYPE_INET6);
2436 assert3u(len, ==, sizeof(struct ip6aux));
2437
2438 if (len != sizeof(struct ip6aux)) {
2439 return NULL;
2440 }
2441
2442 tag_container = kalloc_type(struct inet6_tag_container, wait | M_ZERO);
2443 if (tag_container != NULL) {
2444 tag = &tag_container->inet6_m_tag;
2445
2446 assert3p(tag, ==, tag_container);
2447
2448 M_TAG_INIT(tag, id, type, len, &tag_container->inet6_ip6a, NULL);
2449 }
2450
2451 return tag;
2452 }
2453
2454 void
m_tag_kfree_inet6(struct m_tag * tag)2455 m_tag_kfree_inet6(struct m_tag *tag)
2456 {
2457 struct inet6_tag_container *__single tag_container = (struct inet6_tag_container *)tag;
2458
2459 assert3u(tag->m_tag_len, ==, sizeof(struct ip6aux));
2460
2461 kfree_type(struct inet6_tag_container, tag_container);
2462 }
2463
2464 void
ip6_register_m_tag(void)2465 ip6_register_m_tag(void)
2466 {
2467 int error;
2468
2469 error = m_register_internal_tag_type(KERNEL_TAG_TYPE_INET6, sizeof(struct ip6aux),
2470 m_tag_kalloc_inet6, m_tag_kfree_inet6);
2471
2472 assert3u(error, ==, 0);
2473 }
2474
2475 /*
2476 * Drain callback
2477 */
2478 void
ip6_drain(void)2479 ip6_drain(void)
2480 {
2481 frag6_drain(); /* fragments */
2482 in6_rtqdrain(); /* protocol cloned routes */
2483 nd6_drain(NULL); /* cloned routes: ND6 */
2484 }
2485
2486 /*
2487 * System control for IP6
2488 */
2489
2490 u_char inet6ctlerrmap[PRC_NCMDS] = {
2491 0, 0, 0, 0,
2492 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
2493 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
2494 EMSGSIZE, EHOSTUNREACH, 0, 0,
2495 0, 0, 0, 0,
2496 ENOPROTOOPT, ECONNREFUSED
2497 };
2498
2499 static int
2500 sysctl_reset_ip6_input_stats SYSCTL_HANDLER_ARGS
2501 {
2502 #pragma unused(arg1, arg2)
2503 int error, i;
2504
2505 i = ip6_input_measure;
2506 error = sysctl_handle_int(oidp, &i, 0, req);
2507 if (error || req->newptr == USER_ADDR_NULL) {
2508 goto done;
2509 }
2510 /* impose bounds */
2511 if (i < 0 || i > 1) {
2512 error = EINVAL;
2513 goto done;
2514 }
2515 if (ip6_input_measure != i && i == 1) {
2516 net_perf_initialize(&net_perf, ip6_input_measure_bins);
2517 }
2518 ip6_input_measure = i;
2519 done:
2520 return error;
2521 }
2522
2523 static int
2524 sysctl_ip6_input_measure_bins SYSCTL_HANDLER_ARGS
2525 {
2526 #pragma unused(arg1, arg2)
2527 int error;
2528 uint64_t i;
2529
2530 i = ip6_input_measure_bins;
2531 error = sysctl_handle_quad(oidp, &i, 0, req);
2532 if (error || req->newptr == USER_ADDR_NULL) {
2533 goto done;
2534 }
2535 /* validate data */
2536 if (!net_perf_validate_bins(i)) {
2537 error = EINVAL;
2538 goto done;
2539 }
2540 ip6_input_measure_bins = i;
2541 done:
2542 return error;
2543 }
2544
2545 static int
2546 sysctl_ip6_input_getperf SYSCTL_HANDLER_ARGS
2547 {
2548 #pragma unused(oidp, arg1, arg2)
2549 if (req->oldptr == USER_ADDR_NULL) {
2550 req->oldlen = (size_t)sizeof(struct net_perf);
2551 }
2552
2553 return SYSCTL_OUT(req, &net_perf, MIN(sizeof(net_perf), req->oldlen));
2554 }
2555
2556
2557 /*
2558 * Initialize IPv6 source address hash table.
2559 */
2560 static void
in6_ifaddrhashtbl_init(void)2561 in6_ifaddrhashtbl_init(void)
2562 {
2563 int i, k, p;
2564 uint32_t nhash = 0;
2565 uint32_t hash_size;
2566
2567 if (in6_ifaddrhashtbl != NULL) {
2568 return;
2569 }
2570
2571 PE_parse_boot_argn("ina6ddr_nhash", &nhash,
2572 sizeof(in6addr_nhash));
2573 if (nhash == 0) {
2574 nhash = IN6ADDR_NHASH;
2575 }
2576
2577 hash_size = nhash * sizeof(*in6_ifaddrhashtbl);
2578
2579 in6_ifaddrhashtbl = zalloc_permanent(
2580 hash_size,
2581 ZALIGN_PTR);
2582 in6addr_nhash = nhash;
2583
2584 /*
2585 * Generate the next largest prime greater than in6addr_nhash.
2586 */
2587 k = (in6addr_nhash % 2 == 0) ? in6addr_nhash + 1 : in6addr_nhash + 2;
2588 for (;;) {
2589 p = 1;
2590 for (i = 3; i * i <= k; i += 2) {
2591 if (k % i == 0) {
2592 p = 0;
2593 }
2594 }
2595 if (p == 1) {
2596 break;
2597 }
2598 k += 2;
2599 }
2600 in6addr_hashp = k;
2601 }
2602
2603 static int
2604 sysctl_ip6_checkinterface SYSCTL_HANDLER_ARGS
2605 {
2606 #pragma unused(arg1, arg2)
2607 int error, i;
2608
2609 i = ip6_checkinterface;
2610 error = sysctl_handle_int(oidp, &i, 0, req);
2611 if (error || req->newptr == USER_ADDR_NULL) {
2612 return error;
2613 }
2614
2615 switch (i) {
2616 case IP6_CHECKINTERFACE_WEAK_ES:
2617 case IP6_CHECKINTERFACE_HYBRID_ES:
2618 case IP6_CHECKINTERFACE_STRONG_ES:
2619 if (ip6_checkinterface != i) {
2620 ip6_checkinterface = i;
2621 os_log(OS_LOG_DEFAULT, "%s: ip6_checkinterface is now %d\n",
2622 __func__, ip6_checkinterface);
2623 }
2624 break;
2625 default:
2626 error = EINVAL;
2627 break;
2628 }
2629 return error;
2630 }
2631