xref: /xnu-8020.121.3/bsd/netinet6/ip6_input.c (revision fdd8201d7b966f0c3ea610489d29bd841d358941)
1 /*
2  * Copyright (c) 2003-2021 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 
127 #include <netinet/in.h>
128 #include <netinet/in_systm.h>
129 #if INET
130 #include <netinet/ip.h>
131 #include <netinet/ip_icmp.h>
132 #endif /* INET */
133 #include <netinet/kpi_ipfilter_var.h>
134 #include <netinet/ip6.h>
135 #include <netinet/udp.h>
136 #include <netinet6/in6_var.h>
137 #include <netinet6/ip6_var.h>
138 #include <netinet/in_pcb.h>
139 #include <netinet/icmp6.h>
140 #include <netinet6/in6_ifattach.h>
141 #include <netinet6/nd6.h>
142 #include <netinet6/scope6_var.h>
143 #include <netinet6/ip6protosw.h>
144 
145 #if IPSEC
146 #include <netinet6/ipsec.h>
147 #include <netinet6/ipsec6.h>
148 extern int ipsec_bypass;
149 #endif /* IPSEC */
150 
151 #if DUMMYNET
152 #include <netinet/ip_dummynet.h>
153 #endif /* DUMMYNET */
154 
155 /* we need it for NLOOP. */
156 #include "loop.h"
157 
158 #if PF
159 #include <net/pfvar.h>
160 #endif /* PF */
161 
162 #include <os/log.h>
163 
164 struct ip6protosw *ip6_protox[IPPROTO_MAX];
165 
166 static LCK_GRP_DECLARE(in6_ifaddr_rwlock_grp, "in6_ifaddr_rwlock");
167 LCK_RW_DECLARE(in6_ifaddr_rwlock, &in6_ifaddr_rwlock_grp);
168 
169 /* Protected by in6_ifaddr_rwlock */
170 struct in6_ifaddrhead in6_ifaddrhead;
171 struct in6_ifaddrhashhead * in6_ifaddrhashtbl;
172 uint32_t in6_ifaddrhmask;
173 
174 #define IN6ADDR_NHASH    61
175 u_int32_t in6addr_nhash = 0;                  /* hash table size */
176 u_int32_t in6addr_hashp = 0;                  /* next largest prime */
177 
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 #if NSTF
205 extern void stfattach(void);
206 #endif /* NSTF */
207 
208 SYSCTL_DECL(_net_inet6_ip6);
209 
210 static uint32_t ip6_adj_clear_hwcksum = 0;
211 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, adj_clear_hwcksum,
212     CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_adj_clear_hwcksum, 0,
213     "Invalidate hwcksum info when adjusting length");
214 
215 static uint32_t ip6_adj_partial_sum = 1;
216 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, adj_partial_sum,
217     CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_adj_partial_sum, 0,
218     "Perform partial sum adjustment of trailing bytes at IP layer");
219 
220 static int ip6_input_measure = 0;
221 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, input_perf,
222     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
223     &ip6_input_measure, 0, sysctl_reset_ip6_input_stats, "I", "Do time measurement");
224 
225 static uint64_t ip6_input_measure_bins = 0;
226 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, input_perf_bins,
227     CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_input_measure_bins, 0,
228     sysctl_ip6_input_measure_bins, "I",
229     "bins for chaining performance data histogram");
230 
231 static net_perf_t net_perf;
232 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, input_perf_data,
233     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
234     0, 0, sysctl_ip6_input_getperf, "S,net_perf",
235     "IP6 input performance data (struct net_perf, net/net_perf.h)");
236 
237 /*
238  * ip6_checkinterface controls the receive side of the models for multihoming
239  * that are discussed in RFC 1122.
240  *
241  * sysctl_ip6_checkinterface values are:
242  *  IP6_CHECKINTERFACE_WEAK_ES:
243  *	This corresponds to the Weak End-System model where incoming packets from
244  *	any interface are accepted provided the destination address of the incoming packet
245  *	is assigned to some interface.
246  *
247  *  IP6_CHECKINTERFACE_HYBRID_ES:
248  *	The Hybrid End-System model use the Strong End-System for tunnel interfaces
249  *	(ipsec and utun) and the weak End-System model for other interfaces families.
250  *	This prevents a rogue middle box to probe for signs of TCP connections
251  *	that use the tunnel interface.
252  *
253  *  IP6_CHECKINTERFACE_STRONG_ES:
254  *	The Strong model model requires the packet arrived on an interface that
255  *	is assigned the destination address of the packet.
256  *
257  * Since the routing table and transmit implementation do not implement the Strong ES model,
258  * setting this to a value different from IP6_CHECKINTERFACE_WEAK_ES may lead to unexpected results.
259  *
260  * When forwarding is enabled, the system reverts to the Weak ES model as a router
261  * is expected by design to receive packets from several interfaces to the same address.
262  */
263 #define IP6_CHECKINTERFACE_WEAK_ES       0
264 #define IP6_CHECKINTERFACE_HYBRID_ES     1
265 #define IP6_CHECKINTERFACE_STRONG_ES     2
266 
267 static int ip6_checkinterface = IP6_CHECKINTERFACE_HYBRID_ES;
268 
269 static int sysctl_ip6_checkinterface SYSCTL_HANDLER_ARGS;
270 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, check_interface,
271     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
272     0, 0, sysctl_ip6_checkinterface, "I", "Verify packet arrives on correct interface");
273 
274 #if (DEBUG || DEVELOPMENT)
275 #define IP6_CHECK_IFDEBUG 1
276 #else
277 #define IP6_CHECK_IFDEBUG 0
278 #endif /* (DEBUG || DEVELOPMENT) */
279 static int ip6_checkinterface_debug = IP6_CHECK_IFDEBUG;
280 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, checkinterface_debug, CTLFLAG_RW | CTLFLAG_LOCKED,
281     &ip6_checkinterface_debug, IP6_CHECK_IFDEBUG, "");
282 
283 typedef enum ip6_check_if_result {
284 	IP6_CHECK_IF_NONE = 0,
285 	IP6_CHECK_IF_OURS = 1,
286 	IP6_CHECK_IF_DROP = 2,
287 	IP6_CHECK_IF_FORWARD = 3
288 } ip6_check_if_result_t;
289 
290 static ip6_check_if_result_t ip6_input_check_interface(struct mbuf *, struct ip6_hdr *, struct ifnet *, struct route_in6 *rin6, struct ifnet **);
291 
292 /*
293  * On platforms which require strict alignment (currently for anything but
294  * i386 or x86_64), check if the IP header pointer is 32-bit aligned; if not,
295  * copy the contents of the mbuf chain into a new chain, and free the original
296  * one.  Create some head room in the first mbuf of the new chain, in case
297  * it's needed later on.
298  *
299  * RFC 2460 says that IPv6 headers are 64-bit aligned, but network interfaces
300  * mostly align to 32-bit boundaries.  Care should be taken never to use 64-bit
301  * load/store operations on the fields in IPv6 headers.
302  */
303 #if defined(__i386__) || defined(__x86_64__)
304 #define IP6_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { } while (0)
305 #else /* !__i386__ && !__x86_64__ */
306 #define IP6_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do {                 \
307 	if (!IP6_HDR_ALIGNED_P(mtod(_m, caddr_t))) {                    \
308 	        struct mbuf *_n;                                        \
309 	        struct ifnet *__ifp = (_ifp);                           \
310 	        atomic_add_64(&(__ifp)->if_alignerrs, 1);               \
311 	        if (((_m)->m_flags & M_PKTHDR) &&                       \
312 	            (_m)->m_pkthdr.pkt_hdr != NULL)                     \
313 	                (_m)->m_pkthdr.pkt_hdr = NULL;                  \
314 	        _n = m_defrag_offset(_m, max_linkhdr, M_NOWAIT);        \
315 	        if (_n == NULL) {                                       \
316 	                ip6stat.ip6s_toosmall++;                        \
317 	                m_freem(_m);                                    \
318 	                (_m) = NULL;                                    \
319 	                _action;                                        \
320 	        } else {                                                \
321 	                VERIFY(_n != (_m));                             \
322 	                (_m) = _n;                                      \
323 	        }                                                       \
324 	}                                                               \
325 } while (0)
326 #endif /* !__i386__ && !__x86_64__ */
327 
328 static void
ip6_proto_input(protocol_family_t protocol,mbuf_t packet)329 ip6_proto_input(protocol_family_t protocol, mbuf_t packet)
330 {
331 #pragma unused(protocol)
332 #if INET
333 	struct timeval start_tv;
334 	if (ip6_input_measure) {
335 		net_perf_start_time(&net_perf, &start_tv);
336 	}
337 #endif /* INET */
338 	ip6_input(packet);
339 #if INET
340 	if (ip6_input_measure) {
341 		net_perf_measure_time(&net_perf, &start_tv, 1);
342 		net_perf_histogram(&net_perf, 1);
343 	}
344 #endif /* INET */
345 }
346 
347 /*
348  * IP6 initialization: fill in IP6 protocol switch table.
349  * All protocols not implemented in kernel go to raw IP6 protocol handler.
350  */
351 void
ip6_init(struct ip6protosw * pp,struct domain * dp)352 ip6_init(struct ip6protosw *pp, struct domain *dp)
353 {
354 	static int ip6_initialized = 0;
355 	struct protosw *pr;
356 	struct timeval tv;
357 	int i;
358 	domain_unguard_t unguard;
359 
360 	domain_proto_mtx_lock_assert_held();
361 	VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
362 
363 	_CASSERT((sizeof(struct ip6_hdr) +
364 	    sizeof(struct icmp6_hdr)) <= _MHLEN);
365 
366 	if (ip6_initialized) {
367 		return;
368 	}
369 	ip6_initialized = 1;
370 
371 	eventhandler_lists_ctxt_init(&in6_evhdlr_ctxt);
372 	(void)EVENTHANDLER_REGISTER(&in6_evhdlr_ctxt, in6_event,
373 	    in6_eventhdlr_callback, eventhandler_entry_dummy_arg,
374 	    EVENTHANDLER_PRI_ANY);
375 
376 	eventhandler_lists_ctxt_init(&in6_clat46_evhdlr_ctxt);
377 	(void)EVENTHANDLER_REGISTER(&in6_clat46_evhdlr_ctxt, in6_clat46_event,
378 	    in6_clat46_eventhdlr_callback, eventhandler_entry_dummy_arg,
379 	    EVENTHANDLER_PRI_ANY);
380 
381 	for (i = 0; i < IN6_EVENT_MAX; i++) {
382 		VERIFY(in6_event2kev_array[i].in6_event_code == i);
383 	}
384 
385 	pr = pffindproto_locked(PF_INET6, IPPROTO_RAW, SOCK_RAW);
386 	if (pr == NULL) {
387 		panic("%s: Unable to find [PF_INET6,IPPROTO_RAW,SOCK_RAW]",
388 		    __func__);
389 		/* NOTREACHED */
390 	}
391 
392 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
393 	for (i = 0; i < IPPROTO_MAX; i++) {
394 		ip6_protox[i] = (struct ip6protosw *)pr;
395 	}
396 	/*
397 	 * Cycle through IP protocols and put them into the appropriate place
398 	 * in ip6_protox[], skipping protocols IPPROTO_{IP,RAW}.
399 	 */
400 	VERIFY(dp == inet6domain && dp->dom_family == PF_INET6);
401 	TAILQ_FOREACH(pr, &dp->dom_protosw, pr_entry) {
402 		VERIFY(pr->pr_domain == dp);
403 		if (pr->pr_protocol != 0 && pr->pr_protocol != IPPROTO_RAW) {
404 			/* Be careful to only index valid IP protocols. */
405 			if (pr->pr_protocol < IPPROTO_MAX) {
406 				ip6_protox[pr->pr_protocol] =
407 				    (struct ip6protosw *)pr;
408 			}
409 		}
410 	}
411 
412 	TAILQ_INIT(&in6_ifaddrhead);
413 	in6_ifaddrhashtbl_init();
414 
415 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_receive);
416 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_hdrerr);
417 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_toobig);
418 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_noroute);
419 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_addrerr);
420 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_protounknown);
421 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_truncated);
422 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_discard);
423 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_deliver);
424 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_forward);
425 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_request);
426 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_discard);
427 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragok);
428 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragfail);
429 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragcreat);
430 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_reqd);
431 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_ok);
432 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_fail);
433 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mcast);
434 	IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mcast);
435 
436 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_msg);
437 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_error);
438 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_dstunreach);
439 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_adminprohib);
440 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_timeexceed);
441 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_paramprob);
442 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_pkttoobig);
443 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_echo);
444 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_echoreply);
445 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_routersolicit);
446 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_routeradvert);
447 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_neighborsolicit);
448 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_neighboradvert);
449 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_redirect);
450 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mldquery);
451 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mldreport);
452 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mlddone);
453 
454 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_msg);
455 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_error);
456 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_dstunreach);
457 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_adminprohib);
458 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_timeexceed);
459 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_paramprob);
460 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_pkttoobig);
461 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_echo);
462 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_echoreply);
463 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_routersolicit);
464 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_routeradvert);
465 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_neighborsolicit);
466 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_neighboradvert);
467 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_redirect);
468 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mldquery);
469 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mldreport);
470 	ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mlddone);
471 
472 	getmicrotime(&tv);
473 	ip6_desync_factor =
474 	    (RandomULong() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
475 
476 	PE_parse_boot_argn("in6_embedded_scope", &in6_embedded_scope, sizeof(in6_embedded_scope));
477 	PE_parse_boot_argn("ip6_checkinterface", &i, sizeof(i));
478 	switch (i) {
479 	case IP6_CHECKINTERFACE_WEAK_ES:
480 	case IP6_CHECKINTERFACE_HYBRID_ES:
481 	case IP6_CHECKINTERFACE_STRONG_ES:
482 		ip6_checkinterface = i;
483 		break;
484 	default:
485 		break;
486 	}
487 
488 	in6_ifaddr_init();
489 	ip6_moptions_init();
490 	nd6_init();
491 	frag6_init();
492 	icmp6_init(NULL, dp);
493 	addrsel_policy_init();
494 
495 	/*
496 	 * P2P interfaces often route the local address to the loopback
497 	 * interface. At this point, lo0 hasn't been initialized yet, which
498 	 * means that we need to delay the IPv6 configuration of lo0.
499 	 */
500 	net_init_add(ip6_init_delayed);
501 
502 	unguard = domain_unguard_deploy();
503 	i = proto_register_input(PF_INET6, ip6_proto_input, NULL, 0);
504 	if (i != 0) {
505 		panic("%s: failed to register PF_INET6 protocol: %d",
506 		    __func__, i);
507 		/* NOTREACHED */
508 	}
509 	domain_unguard_release(unguard);
510 }
511 
512 static void
ip6_init_delayed(void)513 ip6_init_delayed(void)
514 {
515 	(void) in6_ifattach_prelim(lo_ifp);
516 
517 	/* timer for regeneranation of temporary addresses randomize ID */
518 	timeout(in6_tmpaddrtimer, NULL,
519 	    (ip6_temp_preferred_lifetime - ip6_desync_factor -
520 	    ip6_temp_regen_advance) * hz);
521 
522 #if NSTF
523 	stfattach();
524 #endif /* NSTF */
525 }
526 
527 static void
ip6_input_adjust(struct mbuf * m,struct ip6_hdr * ip6,uint32_t plen,struct ifnet * inifp)528 ip6_input_adjust(struct mbuf *m, struct ip6_hdr *ip6, uint32_t plen,
529     struct ifnet *inifp)
530 {
531 	boolean_t adjust = TRUE;
532 	uint32_t tot_len = sizeof(*ip6) + plen;
533 
534 	ASSERT(m_pktlen(m) > tot_len);
535 
536 	/*
537 	 * Invalidate hardware checksum info if ip6_adj_clear_hwcksum
538 	 * is set; useful to handle buggy drivers.  Note that this
539 	 * should not be enabled by default, as we may get here due
540 	 * to link-layer padding.
541 	 */
542 	if (ip6_adj_clear_hwcksum &&
543 	    (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) &&
544 	    !(inifp->if_flags & IFF_LOOPBACK) &&
545 	    !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
546 		m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID;
547 		m->m_pkthdr.csum_data = 0;
548 		ip6stat.ip6s_adj_hwcsum_clr++;
549 	}
550 
551 	/*
552 	 * If partial checksum information is available, subtract
553 	 * out the partial sum of postpended extraneous bytes, and
554 	 * update the checksum metadata accordingly.  By doing it
555 	 * here, the upper layer transport only needs to adjust any
556 	 * prepended extraneous bytes (else it will do both.)
557 	 */
558 	if (ip6_adj_partial_sum &&
559 	    (m->m_pkthdr.csum_flags & (CSUM_DATA_VALID | CSUM_PARTIAL)) ==
560 	    (CSUM_DATA_VALID | CSUM_PARTIAL)) {
561 		m->m_pkthdr.csum_rx_val = m_adj_sum16(m,
562 		    m->m_pkthdr.csum_rx_start, m->m_pkthdr.csum_rx_start,
563 		    (tot_len - m->m_pkthdr.csum_rx_start),
564 		    m->m_pkthdr.csum_rx_val);
565 	} else if ((m->m_pkthdr.csum_flags &
566 	    (CSUM_DATA_VALID | CSUM_PARTIAL)) ==
567 	    (CSUM_DATA_VALID | CSUM_PARTIAL)) {
568 		/*
569 		 * If packet has partial checksum info and we decided not
570 		 * to subtract the partial sum of postpended extraneous
571 		 * bytes here (not the default case), leave that work to
572 		 * be handled by the other layers.  For now, only TCP, UDP
573 		 * layers are capable of dealing with this.  For all other
574 		 * protocols (including fragments), trim and ditch the
575 		 * partial sum as those layers might not implement partial
576 		 * checksumming (or adjustment) at all.
577 		 */
578 		if (ip6->ip6_nxt == IPPROTO_TCP ||
579 		    ip6->ip6_nxt == IPPROTO_UDP) {
580 			adjust = FALSE;
581 		} else {
582 			m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID;
583 			m->m_pkthdr.csum_data = 0;
584 			ip6stat.ip6s_adj_hwcsum_clr++;
585 		}
586 	}
587 
588 	if (adjust) {
589 		ip6stat.ip6s_adj++;
590 		if (m->m_len == m->m_pkthdr.len) {
591 			m->m_len = tot_len;
592 			m->m_pkthdr.len = tot_len;
593 		} else {
594 			m_adj(m, tot_len - m->m_pkthdr.len);
595 		}
596 	}
597 }
598 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)599 ip6_input_check_interface(struct mbuf *m, struct ip6_hdr *ip6, struct ifnet *inifp, struct route_in6 *rin6, struct ifnet **deliverifp)
600 {
601 	struct in6_ifaddr *ia6 = NULL;
602 	struct in6_addr tmp_dst = ip6->ip6_dst; /* copy to avoid unaligned access */
603 	struct in6_ifaddr *best_ia6 = NULL;
604 	uint32_t dst_ifscope = IFSCOPE_NONE;
605 	ip6_check_if_result_t result = IP6_CHECK_IF_NONE;
606 
607 	*deliverifp = NULL;
608 
609 	if (m->m_pkthdr.pkt_flags & PKTF_IFAINFO) {
610 		dst_ifscope = m->m_pkthdr.dst_ifindex;
611 	} else {
612 		dst_ifscope = inifp->if_index;
613 	}
614 	/*
615 	 * Check for exact addresses in the hash bucket.
616 	 */
617 	lck_rw_lock_shared(&in6_ifaddr_rwlock);
618 	TAILQ_FOREACH(ia6, IN6ADDR_HASH(&tmp_dst), ia6_hash) {
619 		/*
620 		 * TODO: should we accept loopback
621 		 */
622 		if (in6_are_addr_equal_scoped(&ia6->ia_addr.sin6_addr, &tmp_dst, ia6->ia_ifp->if_index, dst_ifscope)) {
623 			if ((ia6->ia6_flags & (IN6_IFF_NOTREADY | IN6_IFF_CLAT46))) {
624 				continue;
625 			}
626 			best_ia6 = ia6;
627 			if (ia6->ia_ifp == inifp) {
628 				/*
629 				 * TODO: should we also accept locally originated packets
630 				 * or from loopback ???
631 				 */
632 				break;
633 			}
634 			/*
635 			 * Continue the loop in case there's a exact match with another
636 			 * interface
637 			 */
638 		}
639 	}
640 	if (best_ia6 != NULL) {
641 		if (best_ia6->ia_ifp != inifp && ip6_forwarding == 0 &&
642 		    ((ip6_checkinterface == IP6_CHECKINTERFACE_HYBRID_ES &&
643 		    (best_ia6->ia_ifp->if_family == IFNET_FAMILY_IPSEC ||
644 		    best_ia6->ia_ifp->if_family == IFNET_FAMILY_UTUN)) ||
645 		    ip6_checkinterface == IP6_CHECKINTERFACE_STRONG_ES)) {
646 			/*
647 			 * Drop when interface address check is strict and forwarding
648 			 * is disabled
649 			 */
650 			result = IP6_CHECK_IF_DROP;
651 		} else {
652 			result = IP6_CHECK_IF_OURS;
653 			*deliverifp = best_ia6->ia_ifp;
654 			ip6_setdstifaddr_info(m, 0, best_ia6);
655 			ip6_setsrcifaddr_info(m, best_ia6->ia_ifp->if_index, NULL);
656 		}
657 	}
658 	lck_rw_done(&in6_ifaddr_rwlock);
659 
660 	if (result == IP6_CHECK_IF_NONE) {
661 		/*
662 		 * Slow path: route lookup.
663 		 */
664 		struct sockaddr_in6 *dst6;
665 
666 		dst6 = SIN6(&rin6->ro_dst);
667 		dst6->sin6_len = sizeof(struct sockaddr_in6);
668 		dst6->sin6_family = AF_INET6;
669 		dst6->sin6_addr = ip6->ip6_dst;
670 		if (!in6_embedded_scope && IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
671 			dst6->sin6_scope_id = dst_ifscope;
672 		}
673 		rtalloc_scoped_ign((struct route *)rin6,
674 		    RTF_PRCLONING, IFSCOPE_NONE);
675 		if (rin6->ro_rt != NULL) {
676 			RT_LOCK_SPIN(rin6->ro_rt);
677 		}
678 
679 #define rt6_key(r) (SIN6((r)->rt_nodes->rn_key))
680 
681 		/*
682 		 * Accept the packet if the forwarding interface to the destination
683 		 * according to the routing table is the loopback interface,
684 		 * unless the associated route has a gateway.
685 		 * Note that this approach causes to accept a packet if there is a
686 		 * route to the loopback interface for the destination of the packet.
687 		 * But we think it's even useful in some situations, e.g. when using
688 		 * a special daemon which wants to intercept the packet.
689 		 *
690 		 * XXX: some OSes automatically make a cloned route for the destination
691 		 * of an outgoing packet.  If the outgoing interface of the packet
692 		 * is a loopback one, the kernel would consider the packet to be
693 		 * accepted, even if we have no such address assinged on the interface.
694 		 * We check the cloned flag of the route entry to reject such cases,
695 		 * assuming that route entries for our own addresses are not made by
696 		 * cloning (it should be true because in6_addloop explicitly installs
697 		 * the host route).  However, we might have to do an explicit check
698 		 * while it would be less efficient.  Or, should we rather install a
699 		 * reject route for such a case?
700 		 */
701 		if (rin6->ro_rt != NULL &&
702 		    (rin6->ro_rt->rt_flags & (RTF_HOST | RTF_GATEWAY)) == RTF_HOST &&
703 #if RTF_WASCLONED
704 		    !(rin6->ro_rt->rt_flags & RTF_WASCLONED) &&
705 #endif
706 		    rin6->ro_rt->rt_ifp->if_type == IFT_LOOP) {
707 			ia6 = (struct in6_ifaddr *)rin6->ro_rt->rt_ifa;
708 			/*
709 			 * Packets to a tentative, duplicated, or somehow invalid
710 			 * address must not be accepted.
711 			 *
712 			 * For performance, test without acquiring the address lock;
713 			 * a lot of things in the address are set once and never
714 			 * changed (e.g. ia_ifp.)
715 			 */
716 			if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
717 				/* this address is ready */
718 				result = IP6_CHECK_IF_OURS;
719 				*deliverifp = ia6->ia_ifp;       /* correct? */
720 				/*
721 				 * record dst address information into mbuf.
722 				 */
723 				(void) ip6_setdstifaddr_info(m, 0, ia6);
724 				(void) ip6_setsrcifaddr_info(m, ia6->ia_ifp->if_index, NULL);
725 			}
726 		}
727 
728 		if (rin6->ro_rt != NULL) {
729 			RT_UNLOCK(rin6->ro_rt);
730 		}
731 	}
732 
733 	if (result == IP6_CHECK_IF_NONE) {
734 		if (ip6_forwarding == 0) {
735 			result = IP6_CHECK_IF_DROP;
736 		} else {
737 			result = IP6_CHECK_IF_FORWARD;
738 			ip6_setdstifaddr_info(m, inifp->if_index, NULL);
739 			ip6_setsrcifaddr_info(m, inifp->if_index, NULL);
740 		}
741 	}
742 
743 	if (result == IP6_CHECK_IF_OURS && *deliverifp != inifp) {
744 		ASSERT(*deliverifp != NULL);
745 		ip6stat.ip6s_rcv_if_weak_match++;
746 
747 		/*  Logging is too noisy when forwarding is enabled */
748 		if (ip6_checkinterface_debug != IP6_CHECKINTERFACE_WEAK_ES && ip6_forwarding != 0) {
749 			char src_str[MAX_IPv6_STR_LEN];
750 			char dst_str[MAX_IPv6_STR_LEN];
751 
752 			inet_ntop(AF_INET6, &ip6->ip6_src, src_str, sizeof(src_str));
753 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst_str, sizeof(dst_str));
754 			os_log_info(OS_LOG_DEFAULT,
755 			    "%s: weak ES interface match to %s for packet from %s to %s proto %u received via %s",
756 			    __func__, (*deliverifp)->if_xname, src_str, dst_str, ip6->ip6_nxt, inifp->if_xname);
757 		}
758 	} else if (result == IP6_CHECK_IF_DROP) {
759 		ip6stat.ip6s_rcv_if_no_match++;
760 		if (ip6_checkinterface_debug > 0) {
761 			char src_str[MAX_IPv6_STR_LEN];
762 			char dst_str[MAX_IPv6_STR_LEN];
763 
764 			inet_ntop(AF_INET6, &ip6->ip6_src, src_str, sizeof(src_str));
765 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst_str, sizeof(dst_str));
766 			os_log(OS_LOG_DEFAULT,
767 			    "%s: no interface match for packet from %s to %s proto %u received via %s",
768 			    __func__, src_str, dst_str, ip6->ip6_nxt, inifp->if_xname);
769 		}
770 	}
771 
772 	return result;
773 }
774 
775 void
ip6_input(struct mbuf * m)776 ip6_input(struct mbuf *m)
777 {
778 	struct ip6_hdr *ip6;
779 	int off = sizeof(struct ip6_hdr), nest;
780 	u_int32_t plen;
781 	u_int32_t rtalert = ~0;
782 	int nxt = 0, ours = 0;
783 	struct ifnet *inifp, *deliverifp = NULL;
784 	ipfilter_t inject_ipfref = NULL;
785 	int seen = 1;
786 #if DUMMYNET
787 	struct m_tag *tag;
788 	struct ip_fw_args args = {};
789 #endif /* DUMMYNET */
790 	struct route_in6 rin6 = {};
791 
792 	/*
793 	 * Check if the packet we received is valid after interface filter
794 	 * processing
795 	 */
796 	MBUF_INPUT_CHECK(m, m->m_pkthdr.rcvif);
797 	inifp = m->m_pkthdr.rcvif;
798 	VERIFY(inifp != NULL);
799 
800 	/* Perform IP header alignment fixup, if needed */
801 	IP6_HDR_ALIGNMENT_FIXUP(m, inifp, return );
802 
803 	m->m_pkthdr.pkt_flags &= ~PKTF_FORWARDED;
804 #if IPSEC
805 	/*
806 	 * should the inner packet be considered authentic?
807 	 * see comment in ah4_input().
808 	 */
809 	m->m_flags &= ~M_AUTHIPHDR;
810 	m->m_flags &= ~M_AUTHIPDGM;
811 #endif /* IPSEC */
812 
813 	/*
814 	 * make sure we don't have onion peering information into m_aux.
815 	 */
816 	ip6_delaux(m);
817 
818 #if DUMMYNET
819 	if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
820 	    KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) {
821 		struct dn_pkt_tag       *dn_tag;
822 
823 		dn_tag = (struct dn_pkt_tag *)(tag + 1);
824 
825 		args.fwa_pf_rule = dn_tag->dn_pf_rule;
826 
827 		m_tag_delete(m, tag);
828 	}
829 
830 	if (args.fwa_pf_rule) {
831 		ip6 = mtod(m, struct ip6_hdr *); /* In case PF got disabled */
832 
833 		goto check_with_pf;
834 	}
835 #endif /* DUMMYNET */
836 
837 	/*
838 	 * No need to process packet twice if we've already seen it.
839 	 */
840 	inject_ipfref = ipf_get_inject_filter(m);
841 	if (inject_ipfref != NULL) {
842 		ip6 = mtod(m, struct ip6_hdr *);
843 		nxt = ip6->ip6_nxt;
844 		seen = 0;
845 		goto injectit;
846 	} else {
847 		seen = 1;
848 	}
849 
850 	if (__improbable(m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
851 		if_ports_used_match_mbuf(inifp, PF_INET6, m);
852 	}
853 
854 	/*
855 	 * mbuf statistics
856 	 */
857 	if (m->m_flags & M_EXT) {
858 		if (m->m_next != NULL) {
859 			ip6stat.ip6s_mext2m++;
860 		} else {
861 			ip6stat.ip6s_mext1++;
862 		}
863 	} else {
864 #define M2MMAX  (sizeof (ip6stat.ip6s_m2m) / sizeof (ip6stat.ip6s_m2m[0]))
865 		if (m->m_next != NULL) {
866 			if (m->m_pkthdr.pkt_flags & PKTF_LOOP) {
867 				/* XXX */
868 				ip6stat.ip6s_m2m[ifnet_index(lo_ifp)]++;
869 			} else if (inifp->if_index < M2MMAX) {
870 				ip6stat.ip6s_m2m[inifp->if_index]++;
871 			} else {
872 				ip6stat.ip6s_m2m[0]++;
873 			}
874 		} else {
875 			ip6stat.ip6s_m1++;
876 		}
877 #undef M2MMAX
878 	}
879 
880 	/*
881 	 * Drop the packet if IPv6 operation is disabled on the interface.
882 	 */
883 	if (inifp->if_eflags & IFEF_IPV6_DISABLED) {
884 		goto bad;
885 	}
886 
887 	in6_ifstat_inc_na(inifp, ifs6_in_receive);
888 	ip6stat.ip6s_total++;
889 
890 	/*
891 	 * L2 bridge code and some other code can return mbuf chain
892 	 * that does not conform to KAME requirement.  too bad.
893 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
894 	 */
895 	if (m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
896 		struct mbuf *n;
897 
898 		MGETHDR(n, M_DONTWAIT, MT_HEADER);      /* MAC-OK */
899 		if (n) {
900 			M_COPY_PKTHDR(n, m);
901 		}
902 		if (n && m->m_pkthdr.len > MHLEN) {
903 			MCLGET(n, M_DONTWAIT);
904 			if ((n->m_flags & M_EXT) == 0) {
905 				m_freem(n);
906 				n = NULL;
907 			}
908 		}
909 		if (n == NULL) {
910 			goto bad;
911 		}
912 
913 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
914 		n->m_len = m->m_pkthdr.len;
915 		m_freem(m);
916 		m = n;
917 	}
918 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), { goto done; });
919 
920 	if (m->m_len < sizeof(struct ip6_hdr)) {
921 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
922 			ip6stat.ip6s_toosmall++;
923 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
924 			goto done;
925 		}
926 	}
927 
928 	ip6 = mtod(m, struct ip6_hdr *);
929 
930 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
931 		ip6stat.ip6s_badvers++;
932 		in6_ifstat_inc(inifp, ifs6_in_hdrerr);
933 		goto bad;
934 	}
935 
936 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
937 
938 	/*
939 	 * Check against address spoofing/corruption.
940 	 */
941 	if (!(m->m_pkthdr.pkt_flags & PKTF_LOOP) &&
942 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src)) {
943 		ip6stat.ip6s_badscope++;
944 		in6_ifstat_inc(inifp, ifs6_in_addrerr);
945 		goto bad;
946 	}
947 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
948 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
949 		/*
950 		 * XXX: "badscope" is not very suitable for a multicast source.
951 		 */
952 		ip6stat.ip6s_badscope++;
953 		in6_ifstat_inc(inifp, ifs6_in_addrerr);
954 		goto bad;
955 	}
956 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
957 	    !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
958 		/*
959 		 * In this case, the packet should come from the loopback
960 		 * interface.  However, we cannot just check the if_flags,
961 		 * because ip6_mloopback() passes the "actual" interface
962 		 * as the outgoing/incoming interface.
963 		 */
964 		ip6stat.ip6s_badscope++;
965 		in6_ifstat_inc(inifp, ifs6_in_addrerr);
966 		goto bad;
967 	}
968 
969 	/*
970 	 * The following check is not documented in specs.  A malicious
971 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
972 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
973 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
974 	 *
975 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
976 	 * support IPv4-less kernel compilation, we cannot support SIIT
977 	 * environment at all.  So, it makes more sense for us to reject any
978 	 * malicious packets for non-SIIT environment, than try to do a
979 	 * partial support for SIIT environment.
980 	 */
981 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
982 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
983 		ip6stat.ip6s_badscope++;
984 		in6_ifstat_inc(inifp, ifs6_in_addrerr);
985 		goto bad;
986 	}
987 #if 0
988 	/*
989 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
990 	 *
991 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
992 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
993 	 * is revised to forbid relaying case.
994 	 */
995 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
996 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
997 		ip6stat.ip6s_badscope++;
998 		in6_ifstat_inc(inifp, ifs6_in_addrerr);
999 		goto bad;
1000 	}
1001 #endif
1002 
1003 	/*
1004 	 * Naively assume we can attribute inbound data to the route we would
1005 	 * use to send to this destination. Asymetric routing breaks this
1006 	 * assumption, but it still allows us to account for traffic from
1007 	 * a remote node in the routing table.
1008 	 * this has a very significant performance impact so we bypass
1009 	 * if nstat_collect is disabled. We may also bypass if the
1010 	 * protocol is tcp in the future because tcp will have a route that
1011 	 * we can use to attribute the data to. That does mean we would not
1012 	 * account for forwarded tcp traffic.
1013 	 */
1014 	if (nstat_collect) {
1015 		struct rtentry *rte =
1016 		    ifnet_cached_rtlookup_inet6(inifp, &ip6->ip6_src);
1017 		if (rte != NULL) {
1018 			nstat_route_rx(rte, 1, m->m_pkthdr.len, 0);
1019 			rtfree(rte);
1020 		}
1021 	}
1022 
1023 #if DUMMYNET
1024 check_with_pf:
1025 #endif /* DUMMYNET */
1026 #if PF
1027 	/* Invoke inbound packet filter */
1028 	if (PF_IS_ENABLED) {
1029 		int error;
1030 #if DUMMYNET
1031 		error = pf_af_hook(inifp, NULL, &m, AF_INET6, TRUE, &args);
1032 #else /* !DUMMYNET */
1033 		error = pf_af_hook(inifp, NULL, &m, AF_INET6, TRUE, NULL);
1034 #endif /* !DUMMYNET */
1035 		if (error != 0 || m == NULL) {
1036 			if (m != NULL) {
1037 				panic("%s: unexpected packet %p",
1038 				    __func__, m);
1039 				/* NOTREACHED */
1040 			}
1041 			/* Already freed by callee */
1042 			goto done;
1043 		}
1044 		ip6 = mtod(m, struct ip6_hdr *);
1045 	}
1046 #endif /* PF */
1047 
1048 	/* drop packets if interface ID portion is already filled */
1049 	if (!(inifp->if_flags & IFF_LOOPBACK) &&
1050 	    !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
1051 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
1052 		    ip6->ip6_src.s6_addr16[1]) {
1053 			ip6stat.ip6s_badscope++;
1054 			goto bad;
1055 		}
1056 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) &&
1057 		    ip6->ip6_dst.s6_addr16[1]) {
1058 			ip6stat.ip6s_badscope++;
1059 			goto bad;
1060 		}
1061 	}
1062 
1063 	if ((m->m_pkthdr.pkt_flags & PKTF_IFAINFO) != 0 && in6_embedded_scope) {
1064 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1065 			ip6->ip6_src.s6_addr16[1] =
1066 			    htons(m->m_pkthdr.src_ifindex);
1067 		}
1068 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
1069 			ip6->ip6_dst.s6_addr16[1] =
1070 			    htons(m->m_pkthdr.dst_ifindex);
1071 		}
1072 	} else if (in6_embedded_scope) {
1073 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1074 			ip6->ip6_src.s6_addr16[1] = htons(inifp->if_index);
1075 		}
1076 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
1077 			ip6->ip6_dst.s6_addr16[1] = htons(inifp->if_index);
1078 		}
1079 	}
1080 
1081 	/*
1082 	 * Multicast check
1083 	 */
1084 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1085 		struct  in6_multi *in6m = NULL;
1086 
1087 		in6_ifstat_inc_na(inifp, ifs6_in_mcast);
1088 		/*
1089 		 * See if we belong to the destination multicast group on the
1090 		 * arrival interface.
1091 		 */
1092 		in6_multihead_lock_shared();
1093 		IN6_LOOKUP_MULTI(&ip6->ip6_dst, inifp, in6m);
1094 		in6_multihead_lock_done();
1095 		if (in6m != NULL) {
1096 			IN6M_REMREF(in6m);
1097 			ours = 1;
1098 		} else if (!nd6_prproxy) {
1099 			ip6stat.ip6s_notmember++;
1100 			ip6stat.ip6s_cantforward++;
1101 			in6_ifstat_inc(inifp, ifs6_in_discard);
1102 			goto bad;
1103 		}
1104 		deliverifp = inifp;
1105 		/*
1106 		 * record dst address information into mbuf, if we don't have one yet.
1107 		 * note that we are unable to record it, if the address is not listed
1108 		 * as our interface address (e.g. multicast addresses, etc.)
1109 		 */
1110 		if (deliverifp != NULL) {
1111 			struct in6_ifaddr *ia6 = NULL;
1112 
1113 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
1114 			if (ia6 != NULL) {
1115 				(void) ip6_setdstifaddr_info(m, 0, ia6);
1116 				(void) ip6_setsrcifaddr_info(m, ia6->ia_ifp->if_index, NULL);
1117 				IFA_REMREF(&ia6->ia_ifa);
1118 			} else {
1119 				(void) ip6_setdstifaddr_info(m, inifp->if_index, NULL);
1120 				(void) ip6_setsrcifaddr_info(m, inifp->if_index, NULL);
1121 			}
1122 		}
1123 		goto hbhcheck;
1124 	} else {
1125 		/*
1126 		 * Unicast check
1127 		 */
1128 		ip6_check_if_result_t check_if_result = IP6_CHECK_IF_NONE;
1129 		check_if_result = ip6_input_check_interface(m, ip6, inifp, &rin6, &deliverifp);
1130 		ASSERT(check_if_result != IP6_CHECK_IF_NONE);
1131 		if (check_if_result == IP6_CHECK_IF_OURS) {
1132 			ours = 1;
1133 			goto hbhcheck;
1134 		} else if (check_if_result == IP6_CHECK_IF_DROP) {
1135 			goto bad;
1136 		}
1137 	}
1138 
1139 	/*
1140 	 * Now there is no reason to process the packet if it's not our own
1141 	 * and we're not a router.
1142 	 */
1143 	if (!ip6_forwarding) {
1144 		ip6stat.ip6s_cantforward++;
1145 		in6_ifstat_inc(inifp, ifs6_in_discard);
1146 		/*
1147 		 * Raise a kernel event if the packet received on cellular
1148 		 * interface is not intended for local host.
1149 		 * For now limit it to ICMPv6 packets.
1150 		 */
1151 		if (inifp->if_type == IFT_CELLULAR &&
1152 		    ip6->ip6_nxt == IPPROTO_ICMPV6) {
1153 			in6_ifstat_inc(inifp, ifs6_cantfoward_icmp6);
1154 		}
1155 		goto bad;
1156 	}
1157 
1158 hbhcheck:
1159 	/*
1160 	 * Process Hop-by-Hop options header if it's contained.
1161 	 * m may be modified in ip6_hopopts_input().
1162 	 * If a JumboPayload option is included, plen will also be modified.
1163 	 */
1164 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
1165 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1166 		struct ip6_hbh *hbh;
1167 
1168 		/*
1169 		 * Mark the packet to imply that HBH option has been checked.
1170 		 * This can only be true is the packet came in unfragmented
1171 		 * or if the option is in the first fragment
1172 		 */
1173 		m->m_pkthdr.pkt_flags |= PKTF_HBH_CHKED;
1174 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
1175 #if 0   /* touches NULL pointer */
1176 			in6_ifstat_inc(inifp, ifs6_in_discard);
1177 #endif
1178 			goto done;      /* m have already been freed */
1179 		}
1180 
1181 		/* adjust pointer */
1182 		ip6 = mtod(m, struct ip6_hdr *);
1183 
1184 		/*
1185 		 * if the payload length field is 0 and the next header field
1186 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
1187 		 * option MUST be included.
1188 		 */
1189 		if (ip6->ip6_plen == 0 && plen == 0) {
1190 			/*
1191 			 * Note that if a valid jumbo payload option is
1192 			 * contained, ip6_hopopts_input() must set a valid
1193 			 * (non-zero) payload length to the variable plen.
1194 			 */
1195 			ip6stat.ip6s_badoptions++;
1196 			in6_ifstat_inc(inifp, ifs6_in_discard);
1197 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
1198 			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
1199 			    (int)((caddr_t)&ip6->ip6_plen - (caddr_t)ip6));
1200 			goto done;
1201 		}
1202 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
1203 		hbh = (struct ip6_hbh *)(ip6 + 1);
1204 		nxt = hbh->ip6h_nxt;
1205 
1206 		/*
1207 		 * If we are acting as a router and the packet contains a
1208 		 * router alert option, see if we know the option value.
1209 		 * Currently, we only support the option value for MLD, in which
1210 		 * case we should pass the packet to the multicast routing
1211 		 * daemon.
1212 		 */
1213 		if (rtalert != ~0 && ip6_forwarding) {
1214 			switch (rtalert) {
1215 			case IP6OPT_RTALERT_MLD:
1216 				ours = 1;
1217 				break;
1218 			default:
1219 				/*
1220 				 * RFC2711 requires unrecognized values must be
1221 				 * silently ignored.
1222 				 */
1223 				break;
1224 			}
1225 		}
1226 	} else {
1227 		nxt = ip6->ip6_nxt;
1228 	}
1229 
1230 	/*
1231 	 * Check that the amount of data in the buffers
1232 	 * is as at least much as the IPv6 header would have us expect.
1233 	 * Trim mbufs if longer than we expect.
1234 	 * Drop packet if shorter than we expect.
1235 	 */
1236 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
1237 		ip6stat.ip6s_tooshort++;
1238 		in6_ifstat_inc(inifp, ifs6_in_truncated);
1239 		goto bad;
1240 	}
1241 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
1242 		ip6_input_adjust(m, ip6, plen, inifp);
1243 	}
1244 
1245 	/*
1246 	 * Forward if desirable.
1247 	 */
1248 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1249 		if (!ours && nd6_prproxy) {
1250 			/*
1251 			 * If this isn't for us, this might be a Neighbor
1252 			 * Solicitation (dst is solicited-node multicast)
1253 			 * against an address in one of the proxied prefixes;
1254 			 * if so, claim the packet and let icmp6_input()
1255 			 * handle the rest.
1256 			 */
1257 			ours = nd6_prproxy_isours(m, ip6, NULL, IFSCOPE_NONE);
1258 			VERIFY(!ours ||
1259 			    (m->m_pkthdr.pkt_flags & PKTF_PROXY_DST));
1260 		}
1261 		if (!ours) {
1262 			goto bad;
1263 		}
1264 	} else if (!ours) {
1265 		/*
1266 		 * The unicast forwarding function might return the packet
1267 		 * if we are proxying prefix(es), and if the packet is an
1268 		 * ICMPv6 packet that has failed the zone checks, but is
1269 		 * targetted towards a proxied address (this is optimized by
1270 		 * way of RTF_PROXY test.)  If so, claim the packet as ours
1271 		 * and let icmp6_input() handle the rest.  The packet's hop
1272 		 * limit value is kept intact (it's not decremented).  This
1273 		 * is for supporting Neighbor Unreachability Detection between
1274 		 * proxied nodes on different links (src is link-local, dst
1275 		 * is target address.)
1276 		 */
1277 		if ((m = ip6_forward(m, &rin6, 0)) == NULL) {
1278 			goto done;
1279 		}
1280 		VERIFY(rin6.ro_rt != NULL);
1281 		VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
1282 		deliverifp = rin6.ro_rt->rt_ifp;
1283 		ours = 1;
1284 	}
1285 
1286 	ip6 = mtod(m, struct ip6_hdr *);
1287 
1288 	/*
1289 	 * Malicious party may be able to use IPv4 mapped addr to confuse
1290 	 * tcp/udp stack and bypass security checks (act as if it was from
1291 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
1292 	 *
1293 	 * For SIIT end node behavior, you may want to disable the check.
1294 	 * However, you will  become vulnerable to attacks using IPv4 mapped
1295 	 * source.
1296 	 */
1297 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
1298 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
1299 		ip6stat.ip6s_badscope++;
1300 		in6_ifstat_inc(inifp, ifs6_in_addrerr);
1301 		goto bad;
1302 	}
1303 
1304 	/*
1305 	 * Tell launch routine the next header
1306 	 */
1307 	ip6stat.ip6s_delivered++;
1308 	in6_ifstat_inc_na(deliverifp, ifs6_in_deliver);
1309 
1310 injectit:
1311 	nest = 0;
1312 
1313 	/*
1314 	 * Perform IP header alignment fixup again, if needed.  Note that
1315 	 * we do it once for the outermost protocol, and we assume each
1316 	 * protocol handler wouldn't mess with the alignment afterwards.
1317 	 */
1318 	IP6_HDR_ALIGNMENT_FIXUP(m, inifp, return );
1319 
1320 	while (nxt != IPPROTO_DONE) {
1321 		struct ipfilter *filter;
1322 		int (*pr_input)(struct mbuf **, int *, int);
1323 
1324 		/*
1325 		 * This would imply either IPPROTO_HOPOPTS was not the first
1326 		 * option or it did not come in the first fragment.
1327 		 */
1328 		if (nxt == IPPROTO_HOPOPTS &&
1329 		    (m->m_pkthdr.pkt_flags & PKTF_HBH_CHKED) == 0) {
1330 			/*
1331 			 * This implies that HBH option was not contained
1332 			 * in the first fragment
1333 			 */
1334 			ip6stat.ip6s_badoptions++;
1335 			goto bad;
1336 		}
1337 
1338 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
1339 			ip6stat.ip6s_toomanyhdr++;
1340 			goto bad;
1341 		}
1342 
1343 		/*
1344 		 * protection against faulty packet - there should be
1345 		 * more sanity checks in header chain processing.
1346 		 */
1347 		if (m->m_pkthdr.len < off) {
1348 			ip6stat.ip6s_tooshort++;
1349 			in6_ifstat_inc(inifp, ifs6_in_truncated);
1350 			goto bad;
1351 		}
1352 
1353 #if IPSEC
1354 		/*
1355 		 * enforce IPsec policy checking if we are seeing last header.
1356 		 * note that we do not visit this with protocols with pcb layer
1357 		 * code - like udp/tcp/raw ip.
1358 		 */
1359 		if ((ipsec_bypass == 0) &&
1360 		    (ip6_protox[nxt]->pr_flags & PR_LASTHDR) != 0) {
1361 			if (ipsec6_in_reject(m, NULL)) {
1362 				IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
1363 				goto bad;
1364 			}
1365 		}
1366 #endif /* IPSEC */
1367 
1368 		/*
1369 		 * Call IP filter
1370 		 */
1371 		if (!TAILQ_EMPTY(&ipv6_filters) && !IFNET_IS_INTCOPROC(inifp)) {
1372 			ipf_ref();
1373 			TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) {
1374 				if (seen == 0) {
1375 					if ((struct ipfilter *)inject_ipfref ==
1376 					    filter) {
1377 						seen = 1;
1378 					}
1379 				} else if (filter->ipf_filter.ipf_input) {
1380 					errno_t result;
1381 
1382 					result = filter->ipf_filter.ipf_input(
1383 						filter->ipf_filter.cookie,
1384 						(mbuf_t *)&m, off, (uint8_t)nxt);
1385 					if (result == EJUSTRETURN) {
1386 						ipf_unref();
1387 						goto done;
1388 					}
1389 					if (result != 0) {
1390 						ipf_unref();
1391 						goto bad;
1392 					}
1393 				}
1394 			}
1395 			ipf_unref();
1396 		}
1397 
1398 		DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
1399 		    struct ip6_hdr *, ip6, struct ifnet *, inifp,
1400 		    struct ip *, NULL, struct ip6_hdr *, ip6);
1401 
1402 		if ((pr_input = ip6_protox[nxt]->pr_input) == NULL) {
1403 			m_freem(m);
1404 			m = NULL;
1405 			nxt = IPPROTO_DONE;
1406 		} else if (!(ip6_protox[nxt]->pr_flags & PR_PROTOLOCK)) {
1407 			lck_mtx_lock(inet6_domain_mutex);
1408 			nxt = pr_input(&m, &off, nxt);
1409 			lck_mtx_unlock(inet6_domain_mutex);
1410 		} else {
1411 			nxt = pr_input(&m, &off, nxt);
1412 		}
1413 	}
1414 done:
1415 	ROUTE_RELEASE(&rin6);
1416 	return;
1417 bad:
1418 	m_freem(m);
1419 	goto done;
1420 }
1421 
1422 void
ip6_setsrcifaddr_info(struct mbuf * m,uint32_t src_idx,struct in6_ifaddr * ia6)1423 ip6_setsrcifaddr_info(struct mbuf *m, uint32_t src_idx, struct in6_ifaddr *ia6)
1424 {
1425 	VERIFY(m->m_flags & M_PKTHDR);
1426 	m->m_pkthdr.pkt_ext_flags &= ~PKTF_EXT_OUTPUT_SCOPE;
1427 	/*
1428 	 * If the source ifaddr is specified, pick up the information
1429 	 * from there; otherwise just grab the passed-in ifindex as the
1430 	 * caller may not have the ifaddr available.
1431 	 */
1432 	if (ia6 != NULL) {
1433 		m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1434 		m->m_pkthdr.src_ifindex = ia6->ia_ifp->if_index;
1435 
1436 		/* See IN6_IFF comments in in6_var.h */
1437 		m->m_pkthdr.src_iff = (ia6->ia6_flags & 0xffff);
1438 	} else {
1439 		m->m_pkthdr.src_iff = 0;
1440 		m->m_pkthdr.src_ifindex = (uint16_t)src_idx;
1441 		if (src_idx != 0) {
1442 			m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1443 		}
1444 	}
1445 }
1446 
1447 void
ip6_setdstifaddr_info(struct mbuf * m,uint32_t dst_idx,struct in6_ifaddr * ia6)1448 ip6_setdstifaddr_info(struct mbuf *m, uint32_t dst_idx, struct in6_ifaddr *ia6)
1449 {
1450 	VERIFY(m->m_flags & M_PKTHDR);
1451 	m->m_pkthdr.pkt_ext_flags &= ~PKTF_EXT_OUTPUT_SCOPE;
1452 
1453 	/*
1454 	 * If the destination ifaddr is specified, pick up the information
1455 	 * from there; otherwise just grab the passed-in ifindex as the
1456 	 * caller may not have the ifaddr available.
1457 	 */
1458 	if (ia6 != NULL) {
1459 		m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1460 		m->m_pkthdr.dst_ifindex = ia6->ia_ifp->if_index;
1461 
1462 		/* See IN6_IFF comments in in6_var.h */
1463 		m->m_pkthdr.dst_iff = (ia6->ia6_flags & 0xffff);
1464 	} else {
1465 		m->m_pkthdr.dst_iff = 0;
1466 		m->m_pkthdr.dst_ifindex = (uint16_t)dst_idx;
1467 		if (dst_idx != 0) {
1468 			m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
1469 		}
1470 	}
1471 }
1472 
1473 int
ip6_getsrcifaddr_info(struct mbuf * m,uint32_t * src_idx,uint32_t * ia6f)1474 ip6_getsrcifaddr_info(struct mbuf *m, uint32_t *src_idx, uint32_t *ia6f)
1475 {
1476 	VERIFY(m->m_flags & M_PKTHDR);
1477 
1478 	if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO)) {
1479 		return -1;
1480 	}
1481 
1482 	if (src_idx != NULL) {
1483 		*src_idx = m->m_pkthdr.src_ifindex;
1484 	}
1485 
1486 	if (ia6f != NULL) {
1487 		*ia6f = m->m_pkthdr.src_iff;
1488 	}
1489 
1490 	return 0;
1491 }
1492 
1493 int
ip6_getdstifaddr_info(struct mbuf * m,uint32_t * dst_idx,uint32_t * ia6f)1494 ip6_getdstifaddr_info(struct mbuf *m, uint32_t *dst_idx, uint32_t *ia6f)
1495 {
1496 	VERIFY(m->m_flags & M_PKTHDR);
1497 
1498 	if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO)) {
1499 		return -1;
1500 	}
1501 
1502 	if (dst_idx != NULL) {
1503 		*dst_idx = m->m_pkthdr.dst_ifindex;
1504 	}
1505 
1506 	if (ia6f != NULL) {
1507 		*ia6f = m->m_pkthdr.dst_iff;
1508 	}
1509 
1510 	return 0;
1511 }
1512 
1513 uint32_t
ip6_input_getsrcifscope(struct mbuf * m)1514 ip6_input_getsrcifscope(struct mbuf *m)
1515 {
1516 	VERIFY(m->m_flags & M_PKTHDR);
1517 
1518 	if (m->m_pkthdr.rcvif != NULL) {
1519 		return m->m_pkthdr.rcvif->if_index;
1520 	}
1521 
1522 	uint32_t src_ifscope = IFSCOPE_NONE;
1523 	ip6_getsrcifaddr_info(m, &src_ifscope, NULL);
1524 	return src_ifscope;
1525 }
1526 
1527 uint32_t
ip6_input_getdstifscope(struct mbuf * m)1528 ip6_input_getdstifscope(struct mbuf *m)
1529 {
1530 	VERIFY(m->m_flags & M_PKTHDR);
1531 
1532 	if (m->m_pkthdr.rcvif != NULL) {
1533 		return m->m_pkthdr.rcvif->if_index;
1534 	}
1535 
1536 	uint32_t dst_ifscope = IFSCOPE_NONE;
1537 	ip6_getdstifaddr_info(m, &dst_ifscope, NULL);
1538 	return dst_ifscope;
1539 }
1540 
1541 /*
1542  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1543  * included, the real payload length will be stored in plenp.
1544  */
1545 static int
ip6_hopopts_input(uint32_t * plenp,uint32_t * rtalertp,struct mbuf ** mp,int * offp)1546 ip6_hopopts_input(uint32_t *plenp, uint32_t *rtalertp, struct mbuf **mp,
1547     int *offp)
1548 {
1549 	struct mbuf *m = *mp;
1550 	int off = *offp, hbhlen;
1551 	struct ip6_hbh *hbh;
1552 	u_int8_t *opt;
1553 
1554 	/* validation of the length of the header */
1555 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), return (-1));
1556 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1557 	hbhlen = (hbh->ip6h_len + 1) << 3;
1558 
1559 	IP6_EXTHDR_CHECK(m, off, hbhlen, return (-1));
1560 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1561 	off += hbhlen;
1562 	hbhlen -= sizeof(struct ip6_hbh);
1563 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1564 
1565 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1566 	    hbhlen, rtalertp, plenp) < 0) {
1567 		return -1;
1568 	}
1569 
1570 	*offp = off;
1571 	*mp = m;
1572 	return 0;
1573 }
1574 
1575 /*
1576  * Search header for all Hop-by-hop options and process each option.
1577  * This function is separate from ip6_hopopts_input() in order to
1578  * handle a case where the sending node itself process its hop-by-hop
1579  * options header. In such a case, the function is called from ip6_output().
1580  *
1581  * The function assumes that hbh header is located right after the IPv6 header
1582  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1583  * opthead + hbhlen is located in continuous memory region.
1584  */
1585 int
ip6_process_hopopts(struct mbuf * m,u_int8_t * opthead,int hbhlen,u_int32_t * rtalertp,u_int32_t * plenp)1586 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1587     u_int32_t *rtalertp, u_int32_t *plenp)
1588 {
1589 	struct ip6_hdr *ip6;
1590 	int optlen = 0;
1591 	u_int8_t *opt = opthead;
1592 	u_int16_t rtalert_val;
1593 	u_int32_t jumboplen;
1594 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1595 
1596 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1597 		switch (*opt) {
1598 		case IP6OPT_PAD1:
1599 			optlen = 1;
1600 			break;
1601 		case IP6OPT_PADN:
1602 			if (hbhlen < IP6OPT_MINLEN) {
1603 				ip6stat.ip6s_toosmall++;
1604 				goto bad;
1605 			}
1606 			optlen = *(opt + 1) + 2;
1607 			break;
1608 		case IP6OPT_ROUTER_ALERT:
1609 			/* XXX may need check for alignment */
1610 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1611 				ip6stat.ip6s_toosmall++;
1612 				goto bad;
1613 			}
1614 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1615 				/* XXX stat */
1616 				icmp6_error(m, ICMP6_PARAM_PROB,
1617 				    ICMP6_PARAMPROB_HEADER,
1618 				    (int)(erroff + opt + 1 - opthead));
1619 				return -1;
1620 			}
1621 			optlen = IP6OPT_RTALERT_LEN;
1622 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1623 			*rtalertp = ntohs(rtalert_val);
1624 			break;
1625 		case IP6OPT_JUMBO:
1626 			/* XXX may need check for alignment */
1627 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1628 				ip6stat.ip6s_toosmall++;
1629 				goto bad;
1630 			}
1631 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1632 				/* XXX stat */
1633 				icmp6_error(m, ICMP6_PARAM_PROB,
1634 				    ICMP6_PARAMPROB_HEADER,
1635 				    (int)(erroff + opt + 1 - opthead));
1636 				return -1;
1637 			}
1638 			optlen = IP6OPT_JUMBO_LEN;
1639 
1640 			/*
1641 			 * IPv6 packets that have non 0 payload length
1642 			 * must not contain a jumbo payload option.
1643 			 */
1644 			ip6 = mtod(m, struct ip6_hdr *);
1645 			if (ip6->ip6_plen) {
1646 				ip6stat.ip6s_badoptions++;
1647 				icmp6_error(m, ICMP6_PARAM_PROB,
1648 				    ICMP6_PARAMPROB_HEADER,
1649 				    (int)(erroff + opt - opthead));
1650 				return -1;
1651 			}
1652 
1653 			/*
1654 			 * We may see jumbolen in unaligned location, so
1655 			 * we'd need to perform bcopy().
1656 			 */
1657 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1658 			jumboplen = (u_int32_t)htonl(jumboplen);
1659 
1660 #if 1
1661 			/*
1662 			 * if there are multiple jumbo payload options,
1663 			 * *plenp will be non-zero and the packet will be
1664 			 * rejected.
1665 			 * the behavior may need some debate in ipngwg -
1666 			 * multiple options does not make sense, however,
1667 			 * there's no explicit mention in specification.
1668 			 */
1669 			if (*plenp != 0) {
1670 				ip6stat.ip6s_badoptions++;
1671 				icmp6_error(m, ICMP6_PARAM_PROB,
1672 				    ICMP6_PARAMPROB_HEADER,
1673 				    (int)(erroff + opt + 2 - opthead));
1674 				return -1;
1675 			}
1676 #endif
1677 
1678 			/*
1679 			 * jumbo payload length must be larger than 65535.
1680 			 */
1681 			if (jumboplen <= IPV6_MAXPACKET) {
1682 				ip6stat.ip6s_badoptions++;
1683 				icmp6_error(m, ICMP6_PARAM_PROB,
1684 				    ICMP6_PARAMPROB_HEADER,
1685 				    (int)(erroff + opt + 2 - opthead));
1686 				return -1;
1687 			}
1688 			*plenp = jumboplen;
1689 
1690 			break;
1691 		default:                /* unknown option */
1692 			if (hbhlen < IP6OPT_MINLEN) {
1693 				ip6stat.ip6s_toosmall++;
1694 				goto bad;
1695 			}
1696 			optlen = ip6_unknown_opt(opt, m,
1697 			    erroff + opt - opthead);
1698 			if (optlen == -1) {
1699 				return -1;
1700 			}
1701 			optlen += 2;
1702 			break;
1703 		}
1704 	}
1705 
1706 	return 0;
1707 
1708 bad:
1709 	m_freem(m);
1710 	return -1;
1711 }
1712 
1713 /*
1714  * Unknown option processing.
1715  * The third argument `off' is the offset from the IPv6 header to the option,
1716  * which is necessary if the IPv6 header the and option header and IPv6 header
1717  * is not continuous in order to return an ICMPv6 error.
1718  */
1719 int
ip6_unknown_opt(uint8_t * optp,struct mbuf * m,size_t off)1720 ip6_unknown_opt(uint8_t *optp, struct mbuf *m, size_t off)
1721 {
1722 	struct ip6_hdr *ip6;
1723 
1724 	switch (IP6OPT_TYPE(*optp)) {
1725 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1726 		return (int)*(optp + 1);
1727 
1728 	case IP6OPT_TYPE_DISCARD:       /* silently discard */
1729 		m_freem(m);
1730 		return -1;
1731 
1732 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1733 		ip6stat.ip6s_badoptions++;
1734 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, (int)off);
1735 		return -1;
1736 
1737 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1738 		ip6stat.ip6s_badoptions++;
1739 		ip6 = mtod(m, struct ip6_hdr *);
1740 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1741 		    (m->m_flags & (M_BCAST | M_MCAST))) {
1742 			m_freem(m);
1743 		} else {
1744 			icmp6_error(m, ICMP6_PARAM_PROB,
1745 			    ICMP6_PARAMPROB_OPTION, (int)off);
1746 		}
1747 		return -1;
1748 	}
1749 
1750 	m_freem(m);             /* XXX: NOTREACHED */
1751 	return -1;
1752 }
1753 
1754 /*
1755  * Create the "control" list for this pcb.
1756  * These functions will not modify mbuf chain at all.
1757  *
1758  * With KAME mbuf chain restriction:
1759  * The routine will be called from upper layer handlers like tcp6_input().
1760  * Thus the routine assumes that the caller (tcp6_input) have already
1761  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1762  * very first mbuf on the mbuf chain.
1763  *
1764  * ip6_savecontrol_v4 will handle those options that are possible to be
1765  * set on a v4-mapped socket.
1766  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1767  * options and handle the v6-only ones itself.
1768  */
1769 struct mbuf **
ip6_savecontrol_v4(struct inpcb * inp,struct mbuf * m,struct mbuf ** mp,int * v4only)1770 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1771     int *v4only)
1772 {
1773 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1774 
1775 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1776 		struct timeval tv;
1777 
1778 		getmicrotime(&tv);
1779 		mp = sbcreatecontrol_mbuf((caddr_t)&tv, sizeof(tv),
1780 		    SCM_TIMESTAMP, SOL_SOCKET, mp);
1781 		if (*mp == NULL) {
1782 			return NULL;
1783 		}
1784 	}
1785 	if ((inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) {
1786 		uint64_t time;
1787 
1788 		time = mach_absolute_time();
1789 		mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof(time),
1790 		    SCM_TIMESTAMP_MONOTONIC, SOL_SOCKET, mp);
1791 		if (*mp == NULL) {
1792 			return NULL;
1793 		}
1794 	}
1795 	if ((inp->inp_socket->so_options & SO_TIMESTAMP_CONTINUOUS) != 0) {
1796 		uint64_t time;
1797 
1798 		time = mach_continuous_time();
1799 		mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof(time),
1800 		    SCM_TIMESTAMP_CONTINUOUS, SOL_SOCKET, mp);
1801 		if (*mp == NULL) {
1802 			return NULL;
1803 		}
1804 	}
1805 	if ((inp->inp_socket->so_flags & SOF_RECV_TRAFFIC_CLASS) != 0) {
1806 		int tc = m_get_traffic_class(m);
1807 
1808 		mp = sbcreatecontrol_mbuf((caddr_t)&tc, sizeof(tc),
1809 		    SO_TRAFFIC_CLASS, SOL_SOCKET, mp);
1810 		if (*mp == NULL) {
1811 			return NULL;
1812 		}
1813 	}
1814 
1815 	if ((inp->inp_socket->so_flags & SOF_RECV_WAKE_PKT) &&
1816 	    (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
1817 		int flag = 1;
1818 
1819 		mp = sbcreatecontrol_mbuf((caddr_t)&flag, sizeof(flag),
1820 		    SO_RECV_WAKE_PKT, SOL_SOCKET, mp);
1821 		if (*mp == NULL) {
1822 			return NULL;
1823 		}
1824 	}
1825 
1826 #define IS2292(inp, x, y)       (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1827 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1828 		if (v4only != NULL) {
1829 			*v4only = 1;
1830 		}
1831 
1832 		// Send ECN flags for v4-mapped addresses
1833 		if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1834 			struct ip *ip_header = mtod(m, struct ip *);
1835 
1836 			int tclass = (int)(ip_header->ip_tos);
1837 			mp = sbcreatecontrol_mbuf((caddr_t)&tclass, sizeof(tclass),
1838 			    IPV6_TCLASS, IPPROTO_IPV6, mp);
1839 			if (*mp == NULL) {
1840 				return NULL;
1841 			}
1842 		}
1843 
1844 		// Send IN6P_PKTINFO for v4-mapped address
1845 		if ((inp->inp_flags & IN6P_PKTINFO) != 0 || SOFLOW_ENABLED(inp->inp_socket)) {
1846 			struct in6_pktinfo pi6 = {
1847 				.ipi6_addr = IN6ADDR_V4MAPPED_INIT,
1848 				.ipi6_ifindex = (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0,
1849 			};
1850 
1851 			struct ip *ip_header = mtod(m, struct ip *);
1852 			bcopy(&ip_header->ip_dst, &pi6.ipi6_addr.s6_addr32[3], sizeof(struct in_addr));
1853 
1854 			mp = sbcreatecontrol_mbuf((caddr_t)&pi6,
1855 			    sizeof(struct in6_pktinfo),
1856 			    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO),
1857 			    IPPROTO_IPV6, mp);
1858 			if (*mp == NULL) {
1859 				return NULL;
1860 			}
1861 		}
1862 		return mp;
1863 	}
1864 
1865 	/* RFC 2292 sec. 5 */
1866 	if ((inp->inp_flags & IN6P_PKTINFO) != 0 || SOFLOW_ENABLED(inp->inp_socket)) {
1867 		struct in6_pktinfo pi6;
1868 
1869 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1870 		in6_clearscope(&pi6.ipi6_addr); /* XXX */
1871 		pi6.ipi6_ifindex =
1872 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1873 
1874 		mp = sbcreatecontrol_mbuf((caddr_t)&pi6,
1875 		    sizeof(struct in6_pktinfo),
1876 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO),
1877 		    IPPROTO_IPV6, mp);
1878 		if (*mp == NULL) {
1879 			return NULL;
1880 		}
1881 	}
1882 
1883 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1884 		int hlim = ip6->ip6_hlim & 0xff;
1885 
1886 		mp = sbcreatecontrol_mbuf((caddr_t)&hlim, sizeof(int),
1887 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1888 		    IPPROTO_IPV6, mp);
1889 		if (*mp == NULL) {
1890 			return NULL;
1891 		}
1892 	}
1893 
1894 	if (v4only != NULL) {
1895 		*v4only = 0;
1896 	}
1897 	return mp;
1898 }
1899 
1900 int
ip6_savecontrol(struct inpcb * in6p,struct mbuf * m,struct mbuf ** mp)1901 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1902 {
1903 	struct mbuf **np;
1904 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1905 	int v4only = 0;
1906 
1907 	*mp = NULL;
1908 	np = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1909 	if (np == NULL) {
1910 		goto no_mbufs;
1911 	}
1912 
1913 	mp = np;
1914 	if (v4only) {
1915 		return 0;
1916 	}
1917 
1918 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1919 		u_int32_t flowinfo;
1920 		int tclass;
1921 
1922 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1923 		flowinfo >>= 20;
1924 
1925 		tclass = flowinfo & 0xff;
1926 		mp = sbcreatecontrol_mbuf((caddr_t)&tclass, sizeof(tclass),
1927 		    IPV6_TCLASS, IPPROTO_IPV6, mp);
1928 		if (*mp == NULL) {
1929 			goto no_mbufs;
1930 		}
1931 	}
1932 
1933 	/*
1934 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1935 	 * privilege for the option (see ip6_ctloutput), but it might be too
1936 	 * strict, since there might be some hop-by-hop options which can be
1937 	 * returned to normal user.
1938 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1939 	 */
1940 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1941 		/*
1942 		 * Check if a hop-by-hop options header is contatined in the
1943 		 * received packet, and if so, store the options as ancillary
1944 		 * data. Note that a hop-by-hop options header must be
1945 		 * just after the IPv6 header, which is assured through the
1946 		 * IPv6 input processing.
1947 		 */
1948 		ip6 = mtod(m, struct ip6_hdr *);
1949 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1950 			struct ip6_hbh *hbh;
1951 			int hbhlen = 0;
1952 			hbh = (struct ip6_hbh *)(ip6 + 1);
1953 			hbhlen = (hbh->ip6h_len + 1) << 3;
1954 
1955 			/*
1956 			 * XXX: We copy the whole header even if a
1957 			 * jumbo payload option is included, the option which
1958 			 * is to be removed before returning according to
1959 			 * RFC2292.
1960 			 * Note: this constraint is removed in RFC3542
1961 			 */
1962 			mp = sbcreatecontrol_mbuf((caddr_t)hbh, hbhlen,
1963 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1964 			    IPPROTO_IPV6, mp);
1965 
1966 			if (*mp == NULL) {
1967 				goto no_mbufs;
1968 			}
1969 		}
1970 	}
1971 
1972 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1973 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1974 
1975 		/*
1976 		 * Search for destination options headers or routing
1977 		 * header(s) through the header chain, and stores each
1978 		 * header as ancillary data.
1979 		 * Note that the order of the headers remains in
1980 		 * the chain of ancillary data.
1981 		 */
1982 		while (1) {     /* is explicit loop prevention necessary? */
1983 			struct ip6_ext *ip6e = NULL;
1984 			int elen;
1985 
1986 			/*
1987 			 * if it is not an extension header, don't try to
1988 			 * pull it from the chain.
1989 			 */
1990 			switch (nxt) {
1991 			case IPPROTO_DSTOPTS:
1992 			case IPPROTO_ROUTING:
1993 			case IPPROTO_HOPOPTS:
1994 			case IPPROTO_AH: /* is it possible? */
1995 				break;
1996 			default:
1997 				goto loopend;
1998 			}
1999 
2000 			if (off + sizeof(*ip6e) > m->m_len) {
2001 				goto loopend;
2002 			}
2003 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
2004 			if (nxt == IPPROTO_AH) {
2005 				elen = (ip6e->ip6e_len + 2) << 2;
2006 			} else {
2007 				elen = (ip6e->ip6e_len + 1) << 3;
2008 			}
2009 			if (off + elen > m->m_len) {
2010 				goto loopend;
2011 			}
2012 
2013 			switch (nxt) {
2014 			case IPPROTO_DSTOPTS:
2015 				if (!(in6p->inp_flags & IN6P_DSTOPTS)) {
2016 					break;
2017 				}
2018 
2019 				mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
2020 				    IS2292(in6p, IPV6_2292DSTOPTS,
2021 				    IPV6_DSTOPTS), IPPROTO_IPV6, mp);
2022 				if (*mp == NULL) {
2023 					goto no_mbufs;
2024 				}
2025 				break;
2026 			case IPPROTO_ROUTING:
2027 				if (!(in6p->inp_flags & IN6P_RTHDR)) {
2028 					break;
2029 				}
2030 
2031 				mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
2032 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
2033 				    IPPROTO_IPV6, mp);
2034 				if (*mp == NULL) {
2035 					goto no_mbufs;
2036 				}
2037 				break;
2038 			case IPPROTO_HOPOPTS:
2039 			case IPPROTO_AH: /* is it possible? */
2040 				break;
2041 
2042 			default:
2043 				/*
2044 				 * other cases have been filtered in the above.
2045 				 * none will visit this case.  here we supply
2046 				 * the code just in case (nxt overwritten or
2047 				 * other cases).
2048 				 */
2049 				goto loopend;
2050 			}
2051 
2052 			/* proceed with the next header. */
2053 			off += elen;
2054 			nxt = ip6e->ip6e_nxt;
2055 			ip6e = NULL;
2056 		}
2057 loopend:
2058 		;
2059 	}
2060 	return 0;
2061 no_mbufs:
2062 	ip6stat.ip6s_pktdropcntrl++;
2063 	/* XXX increment a stat to show the failure */
2064 	return ENOBUFS;
2065 }
2066 #undef IS2292
2067 
2068 void
ip6_notify_pmtu(struct inpcb * in6p,struct sockaddr_in6 * dst,u_int32_t * mtu)2069 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
2070 {
2071 	struct socket *so;
2072 	struct mbuf *m_mtu;
2073 	struct ip6_mtuinfo mtuctl;
2074 
2075 	so =  in6p->inp_socket;
2076 
2077 	if ((in6p->inp_flags & IN6P_MTU) == 0) {
2078 		return;
2079 	}
2080 
2081 	if (mtu == NULL) {
2082 		return;
2083 	}
2084 
2085 #ifdef DIAGNOSTIC
2086 	if (so == NULL) {               /* I believe this is impossible */
2087 		panic("ip6_notify_pmtu: socket is NULL");
2088 		/* NOTREACHED */
2089 	}
2090 #endif
2091 
2092 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2093 	    (so->so_proto == NULL || so->so_proto->pr_protocol == IPPROTO_TCP)) {
2094 		return;
2095 	}
2096 
2097 	if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2098 	    !in6_are_addr_equal_scoped(&in6p->in6p_faddr, &dst->sin6_addr, in6p->inp_fifscope, dst->sin6_scope_id)) {
2099 		return;
2100 	}
2101 
2102 	bzero(&mtuctl, sizeof(mtuctl));         /* zero-clear for safety */
2103 	mtuctl.ip6m_mtu = *mtu;
2104 	mtuctl.ip6m_addr = *dst;
2105 	if (!in6_embedded_scope) {
2106 		mtuctl.ip6m_addr.sin6_scope_id = dst->sin6_scope_id;
2107 	}
2108 	if (sa6_recoverscope(&mtuctl.ip6m_addr, TRUE)) {
2109 		return;
2110 	}
2111 
2112 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
2113 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) {
2114 		return;
2115 	}
2116 
2117 	if (sbappendaddr(&so->so_rcv, SA(dst), NULL, m_mtu, NULL) == 0) {
2118 		return;
2119 	}
2120 	sorwakeup(so);
2121 }
2122 
2123 /*
2124  * Get pointer to the previous header followed by the header
2125  * currently processed.
2126  * XXX: This function supposes that
2127  *	M includes all headers,
2128  *	the next header field and the header length field of each header
2129  *	are valid, and
2130  *	the sum of each header length equals to OFF.
2131  * Because of these assumptions, this function must be called very
2132  * carefully. Moreover, it will not be used in the near future when
2133  * we develop `neater' mechanism to process extension headers.
2134  */
2135 char *
ip6_get_prevhdr(struct mbuf * m,int off)2136 ip6_get_prevhdr(struct mbuf *m, int off)
2137 {
2138 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2139 
2140 	if (off == sizeof(struct ip6_hdr)) {
2141 		return (char *)&ip6->ip6_nxt;
2142 	} else {
2143 		int len, nxt;
2144 		struct ip6_ext *ip6e = NULL;
2145 
2146 		nxt = ip6->ip6_nxt;
2147 		len = sizeof(struct ip6_hdr);
2148 		while (len < off) {
2149 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
2150 
2151 			switch (nxt) {
2152 			case IPPROTO_FRAGMENT:
2153 				len += sizeof(struct ip6_frag);
2154 				break;
2155 			case IPPROTO_AH:
2156 				len += (ip6e->ip6e_len + 2) << 2;
2157 				break;
2158 			default:
2159 				len += (ip6e->ip6e_len + 1) << 3;
2160 				break;
2161 			}
2162 			nxt = ip6e->ip6e_nxt;
2163 		}
2164 		if (ip6e) {
2165 			return (char *)&ip6e->ip6e_nxt;
2166 		} else {
2167 			return NULL;
2168 		}
2169 	}
2170 }
2171 
2172 /*
2173  * get next header offset.  m will be retained.
2174  */
2175 int
ip6_nexthdr(struct mbuf * m,int off,int proto,int * nxtp)2176 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
2177 {
2178 	struct ip6_hdr ip6;
2179 	struct ip6_ext ip6e;
2180 	struct ip6_frag fh;
2181 
2182 	/* just in case */
2183 	VERIFY(m != NULL);
2184 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) {
2185 		return -1;
2186 	}
2187 
2188 	switch (proto) {
2189 	case IPPROTO_IPV6:
2190 		if (m->m_pkthdr.len < off + sizeof(ip6)) {
2191 			return -1;
2192 		}
2193 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
2194 		if (nxtp) {
2195 			*nxtp = ip6.ip6_nxt;
2196 		}
2197 		off += sizeof(ip6);
2198 		return off;
2199 
2200 	case IPPROTO_FRAGMENT:
2201 		/*
2202 		 * terminate parsing if it is not the first fragment,
2203 		 * it does not make sense to parse through it.
2204 		 */
2205 		if (m->m_pkthdr.len < off + sizeof(fh)) {
2206 			return -1;
2207 		}
2208 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
2209 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
2210 		if (fh.ip6f_offlg & IP6F_OFF_MASK) {
2211 			return -1;
2212 		}
2213 		if (nxtp) {
2214 			*nxtp = fh.ip6f_nxt;
2215 		}
2216 		off += sizeof(struct ip6_frag);
2217 		return off;
2218 
2219 	case IPPROTO_AH:
2220 		if (m->m_pkthdr.len < off + sizeof(ip6e)) {
2221 			return -1;
2222 		}
2223 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
2224 		if (nxtp) {
2225 			*nxtp = ip6e.ip6e_nxt;
2226 		}
2227 		off += (ip6e.ip6e_len + 2) << 2;
2228 		return off;
2229 
2230 	case IPPROTO_HOPOPTS:
2231 	case IPPROTO_ROUTING:
2232 	case IPPROTO_DSTOPTS:
2233 		if (m->m_pkthdr.len < off + sizeof(ip6e)) {
2234 			return -1;
2235 		}
2236 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
2237 		if (nxtp) {
2238 			*nxtp = ip6e.ip6e_nxt;
2239 		}
2240 		off += (ip6e.ip6e_len + 1) << 3;
2241 		return off;
2242 
2243 	case IPPROTO_NONE:
2244 	case IPPROTO_ESP:
2245 	case IPPROTO_IPCOMP:
2246 		/* give up */
2247 		return -1;
2248 
2249 	default:
2250 		return -1;
2251 	}
2252 }
2253 
2254 /*
2255  * get offset for the last header in the chain.  m will be kept untainted.
2256  */
2257 int
ip6_lasthdr(struct mbuf * m,int off,int proto,int * nxtp)2258 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
2259 {
2260 	int newoff;
2261 	int nxt;
2262 
2263 	if (!nxtp) {
2264 		nxt = -1;
2265 		nxtp = &nxt;
2266 	}
2267 	while (1) {
2268 		newoff = ip6_nexthdr(m, off, proto, nxtp);
2269 		if (newoff < 0) {
2270 			return off;
2271 		} else if (newoff < off) {
2272 			return -1;    /* invalid */
2273 		} else if (newoff == off) {
2274 			return newoff;
2275 		}
2276 
2277 		off = newoff;
2278 		proto = *nxtp;
2279 	}
2280 }
2281 
2282 boolean_t
ip6_pkt_has_ulp(struct mbuf * m)2283 ip6_pkt_has_ulp(struct mbuf *m)
2284 {
2285 	int off = 0, nxt = IPPROTO_NONE;
2286 
2287 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
2288 	if (off < 0 || m->m_pkthdr.len < off) {
2289 		return FALSE;
2290 	}
2291 
2292 	switch (nxt) {
2293 	case IPPROTO_TCP:
2294 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len) {
2295 			return FALSE;
2296 		}
2297 		break;
2298 	case IPPROTO_UDP:
2299 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len) {
2300 			return FALSE;
2301 		}
2302 		break;
2303 	case IPPROTO_ICMPV6:
2304 		if (off + sizeof(uint32_t) > m->m_pkthdr.len) {
2305 			return FALSE;
2306 		}
2307 		break;
2308 	case IPPROTO_NONE:
2309 		return TRUE;
2310 	case IPPROTO_ESP:
2311 		return TRUE;
2312 	case IPPROTO_IPCOMP:
2313 		return TRUE;
2314 	default:
2315 		return FALSE;
2316 	}
2317 	return TRUE;
2318 }
2319 
2320 struct ip6aux *
ip6_addaux(struct mbuf * m)2321 ip6_addaux(struct mbuf *m)
2322 {
2323 	struct m_tag            *tag;
2324 
2325 	/* Check if one is already allocated */
2326 	tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2327 	    KERNEL_TAG_TYPE_INET6, NULL);
2328 	if (tag == NULL) {
2329 		/* Allocate a tag */
2330 		tag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6,
2331 		    sizeof(struct ip6aux), M_DONTWAIT, m);
2332 
2333 		/* Attach it to the mbuf */
2334 		if (tag) {
2335 			m_tag_prepend(m, tag);
2336 		}
2337 	}
2338 
2339 	return tag ? (struct ip6aux *)(tag + 1) : NULL;
2340 }
2341 
2342 struct ip6aux *
ip6_findaux(struct mbuf * m)2343 ip6_findaux(struct mbuf *m)
2344 {
2345 	struct m_tag    *tag;
2346 
2347 	tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2348 	    KERNEL_TAG_TYPE_INET6, NULL);
2349 
2350 	return tag ? (struct ip6aux *)(tag + 1) : NULL;
2351 }
2352 
2353 void
ip6_delaux(struct mbuf * m)2354 ip6_delaux(struct mbuf *m)
2355 {
2356 	struct m_tag    *tag;
2357 
2358 	tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
2359 	    KERNEL_TAG_TYPE_INET6, NULL);
2360 	if (tag) {
2361 		m_tag_delete(m, tag);
2362 	}
2363 }
2364 
2365 /*
2366  * Drain callback
2367  */
2368 void
ip6_drain(void)2369 ip6_drain(void)
2370 {
2371 	frag6_drain();          /* fragments */
2372 	in6_rtqdrain();         /* protocol cloned routes */
2373 	nd6_drain(NULL);        /* cloned routes: ND6 */
2374 }
2375 
2376 /*
2377  * System control for IP6
2378  */
2379 
2380 u_char  inet6ctlerrmap[PRC_NCMDS] = {
2381 	0, 0, 0, 0,
2382 	0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
2383 	EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
2384 	EMSGSIZE, EHOSTUNREACH, 0, 0,
2385 	0, 0, 0, 0,
2386 	ENOPROTOOPT, ECONNREFUSED
2387 };
2388 
2389 static int
2390 sysctl_reset_ip6_input_stats SYSCTL_HANDLER_ARGS
2391 {
2392 #pragma unused(arg1, arg2)
2393 	int error, i;
2394 
2395 	i = ip6_input_measure;
2396 	error = sysctl_handle_int(oidp, &i, 0, req);
2397 	if (error || req->newptr == USER_ADDR_NULL) {
2398 		goto done;
2399 	}
2400 	/* impose bounds */
2401 	if (i < 0 || i > 1) {
2402 		error = EINVAL;
2403 		goto done;
2404 	}
2405 	if (ip6_input_measure != i && i == 1) {
2406 		net_perf_initialize(&net_perf, ip6_input_measure_bins);
2407 	}
2408 	ip6_input_measure = i;
2409 done:
2410 	return error;
2411 }
2412 
2413 static int
2414 sysctl_ip6_input_measure_bins SYSCTL_HANDLER_ARGS
2415 {
2416 #pragma unused(arg1, arg2)
2417 	int error;
2418 	uint64_t i;
2419 
2420 	i = ip6_input_measure_bins;
2421 	error = sysctl_handle_quad(oidp, &i, 0, req);
2422 	if (error || req->newptr == USER_ADDR_NULL) {
2423 		goto done;
2424 	}
2425 	/* validate data */
2426 	if (!net_perf_validate_bins(i)) {
2427 		error = EINVAL;
2428 		goto done;
2429 	}
2430 	ip6_input_measure_bins = i;
2431 done:
2432 	return error;
2433 }
2434 
2435 static int
2436 sysctl_ip6_input_getperf SYSCTL_HANDLER_ARGS
2437 {
2438 #pragma unused(oidp, arg1, arg2)
2439 	if (req->oldptr == USER_ADDR_NULL) {
2440 		req->oldlen = (size_t)sizeof(struct net_perf);
2441 	}
2442 
2443 	return SYSCTL_OUT(req, &net_perf, MIN(sizeof(net_perf), req->oldlen));
2444 }
2445 
2446 
2447 /*
2448  * Initialize IPv6 source address hash table.
2449  */
2450 static void
in6_ifaddrhashtbl_init(void)2451 in6_ifaddrhashtbl_init(void)
2452 {
2453 	int i, k, p;
2454 
2455 	if (in6_ifaddrhashtbl != NULL) {
2456 		return;
2457 	}
2458 
2459 	PE_parse_boot_argn("ina6ddr_nhash", &in6addr_nhash,
2460 	    sizeof(in6addr_nhash));
2461 	if (in6addr_nhash == 0) {
2462 		in6addr_nhash = IN6ADDR_NHASH;
2463 	}
2464 
2465 	in6_ifaddrhashtbl = zalloc_permanent(
2466 		in6addr_nhash * sizeof(*in6_ifaddrhashtbl),
2467 		ZALIGN_PTR);
2468 
2469 	/*
2470 	 * Generate the next largest prime greater than in6addr_nhash.
2471 	 */
2472 	k = (in6addr_nhash % 2 == 0) ? in6addr_nhash + 1 : in6addr_nhash + 2;
2473 	for (;;) {
2474 		p = 1;
2475 		for (i = 3; i * i <= k; i += 2) {
2476 			if (k % i == 0) {
2477 				p = 0;
2478 			}
2479 		}
2480 		if (p == 1) {
2481 			break;
2482 		}
2483 		k += 2;
2484 	}
2485 	in6addr_hashp = k;
2486 }
2487 
2488 static int
2489 sysctl_ip6_checkinterface SYSCTL_HANDLER_ARGS
2490 {
2491 #pragma unused(arg1, arg2)
2492 	int error, i;
2493 
2494 	i = ip6_checkinterface;
2495 	error = sysctl_handle_int(oidp, &i, 0, req);
2496 	if (error || req->newptr == USER_ADDR_NULL) {
2497 		return error;
2498 	}
2499 
2500 	switch (i) {
2501 	case IP6_CHECKINTERFACE_WEAK_ES:
2502 	case IP6_CHECKINTERFACE_HYBRID_ES:
2503 	case IP6_CHECKINTERFACE_STRONG_ES:
2504 		if (ip6_checkinterface != i) {
2505 			ip6_checkinterface = i;
2506 			os_log(OS_LOG_DEFAULT, "%s: ip6_checkinterface is now %d\n",
2507 			    __func__, ip6_checkinterface);
2508 		}
2509 		break;
2510 	default:
2511 		error = EINVAL;
2512 		break;
2513 	}
2514 	return error;
2515 }
2516