xref: /xnu-11215.41.3/bsd/netinet6/ip6_output.c (revision 33de042d024d46de5ff4e89f2471de6608e37fa4)
1 /*
2  * Copyright (c) 2000-2024 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, 1990, 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_output.c	8.3 (Berkeley) 1/21/94
91  */
92 /*
93  * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
94  * support for mandatory and extensible security protections.  This notice
95  * is included in support of clause 2.2 (b) of the Apple Public License,
96  * Version 2.0.
97  */
98 
99 #include <sys/param.h>
100 #include <sys/malloc.h>
101 #include <sys/mbuf.h>
102 #include <sys/errno.h>
103 #include <sys/protosw.h>
104 #include <sys/socket.h>
105 #include <sys/socketvar.h>
106 #include <sys/systm.h>
107 #include <sys/kernel.h>
108 #include <sys/proc.h>
109 #include <sys/kauth.h>
110 #include <sys/mcache.h>
111 #include <sys/sysctl.h>
112 #include <kern/zalloc.h>
113 #include <libkern/OSByteOrder.h>
114 
115 #include <pexpert/pexpert.h>
116 #include <mach/sdt.h>
117 
118 #include <net/if.h>
119 #include <net/route.h>
120 #include <net/dlil.h>
121 #include <net/net_api_stats.h>
122 #include <net/net_osdep.h>
123 #include <net/net_perf.h>
124 #include <net/droptap.h>
125 
126 #include <netinet/ip.h>
127 #include <netinet/in.h>
128 #include <netinet/in_var.h>
129 #include <netinet/ip_var.h>
130 #include <netinet6/in6_var.h>
131 #include <netinet/ip6.h>
132 #include <netinet/kpi_ipfilter_var.h>
133 #include <netinet/in_tclass.h>
134 
135 #include <netinet6/ip6protosw.h>
136 #include <netinet/icmp6.h>
137 #include <netinet6/ip6_var.h>
138 #include <netinet/in_pcb.h>
139 #include <netinet6/nd6.h>
140 #include <netinet6/scope6_var.h>
141 #if IPSEC
142 #include <netinet6/ipsec.h>
143 #include <netinet6/ipsec6.h>
144 #include <netkey/key.h>
145 extern int ipsec_bypass;
146 #endif /* IPSEC */
147 
148 #if NECP
149 #include <net/necp.h>
150 #endif /* NECP */
151 
152 #if DUMMYNET
153 #include <netinet/ip_dummynet.h>
154 #endif /* DUMMYNET */
155 
156 #if PF
157 #include <net/pfvar.h>
158 #endif /* PF */
159 
160 #include <net/sockaddr_utils.h>
161 
162 static int sysctl_reset_ip6_output_stats SYSCTL_HANDLER_ARGS;
163 static int sysctl_ip6_output_measure_bins SYSCTL_HANDLER_ARGS;
164 static int sysctl_ip6_output_getperf SYSCTL_HANDLER_ARGS;
165 static int ip6_copyexthdr(struct mbuf **, caddr_t __indexable, int);
166 static void ip6_out_cksum_stats(int, u_int32_t);
167 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
168 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
169     struct ip6_frag **);
170 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
171     struct ifnet *, struct in6_addr *, uint32_t, u_int32_t *);
172 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *,
173     struct sockopt *sopt);
174 static int ip6_pcbopt(int, u_char *buf __sized_by(len), int len, struct ip6_pktopts **, int);
175 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
176 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, zalloc_flags_t);
177 static void im6o_trace(struct ip6_moptions *, int);
178 static int ip6_setpktopt(int, uint8_t * __sized_by(len), int len, struct ip6_pktopts *, int,
179     int, int);
180 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
181 static void ip6_output_checksum(struct ifnet *, uint32_t, struct mbuf *,
182     int, uint32_t, uint32_t);
183 extern int udp_ctloutput(struct socket *, struct sockopt *);
184 static int ip6_fragment_packet(struct mbuf **m,
185     struct ip6_pktopts *opt, struct ip6_out_args * ip6oa,
186     struct ip6_exthdrs *exthdrsp, struct ifnet *ifp,
187     uint32_t mtu, uint32_t unfragpartlen,
188     int nxt0, uint32_t optlen);
189 
190 SYSCTL_DECL(_net_inet6_ip6);
191 
192 static int ip6_output_measure = 0;
193 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf,
194     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
195     &ip6_output_measure, 0, sysctl_reset_ip6_output_stats, "I", "Do time measurement");
196 
197 static uint64_t ip6_output_measure_bins = 0;
198 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf_bins,
199     CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_output_measure_bins, 0,
200     sysctl_ip6_output_measure_bins, "I",
201     "bins for chaining performance data histogram");
202 
203 static net_perf_t net_perf;
204 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf_data,
205     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
206     0, 0, sysctl_ip6_output_getperf, "S,net_perf",
207     "IP6 output performance data (struct net_perf, net/net_perf.h)");
208 
209 #define IM6O_TRACE_HIST_SIZE    32      /* size of trace history */
210 
211 #define GET_IP6EXTHDR(s, t, f) \
212     __unsafe_forge_bidi_indexable(t, s->f, (((struct ip6_ext *)s->f)->ip6e_len + 1) << 3)
213 
214 /* For gdb */
215 __private_extern__ unsigned int im6o_trace_hist_size = IM6O_TRACE_HIST_SIZE;
216 
217 struct ip6_moptions_dbg {
218 	struct ip6_moptions     im6o;                   /* ip6_moptions */
219 	u_int16_t               im6o_refhold_cnt;       /* # of IM6O_ADDREF */
220 	u_int16_t               im6o_refrele_cnt;       /* # of IM6O_REMREF */
221 	/*
222 	 * Alloc and free callers.
223 	 */
224 	ctrace_t                im6o_alloc;
225 	ctrace_t                im6o_free;
226 	/*
227 	 * Circular lists of IM6O_ADDREF and IM6O_REMREF callers.
228 	 */
229 	ctrace_t                im6o_refhold[IM6O_TRACE_HIST_SIZE];
230 	ctrace_t                im6o_refrele[IM6O_TRACE_HIST_SIZE];
231 };
232 
233 #if DEBUG
234 static unsigned int im6o_debug = 1;     /* debugging (enabled) */
235 #else
236 static unsigned int im6o_debug;         /* debugging (disabled) */
237 #endif /* !DEBUG */
238 
239 ZONE_DECLARE(im6o_zone, struct ip6_moptions);
240 #define IM6O_ZONE_NAME          "ip6_moptions"  /* zone name */
241 zone_t im6o_zone = {0};                         /* zone for *ip6_moptions */
242 
243 /*
244  * ip6_output() calls ip6_output_list() to do the work
245  */
246 int
ip6_output(struct mbuf * m0,struct ip6_pktopts * opt,struct route_in6 * ro,int flags,struct ip6_moptions * im6o,struct ifnet ** ifpp,struct ip6_out_args * ip6oa)247 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
248     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
249     struct ifnet **ifpp, struct ip6_out_args *ip6oa)
250 {
251 	return ip6_output_list(m0, 0, opt, ro, flags, im6o, ifpp, ip6oa);
252 }
253 
254 /*
255  * IP6 output. Each packet in mbuf chain m contains a skeletal IP6
256  * header (with pri, len, nxt, hlim, src, dst).
257  * This function may modify ver and hlim only.
258  * The mbuf chain containing the packet will be freed.
259  * The mbuf opt, if present, will not be freed.
260  *
261  * If ro is non-NULL and has valid ro->ro_rt, route lookup would be
262  * skipped and ro->ro_rt would be used.  Otherwise the result of route
263  * lookup is stored in ro->ro_rt.
264  *
265  * type of "mtu": rt_rmx.rmx_mtu is u_int32_t, ifnet.ifr_mtu is int, and
266  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_int32_t to hold largest one,
267  * which is rt_rmx.rmx_mtu.
268  */
269 int
ip6_output_list(struct mbuf * m0,int packetchain,struct ip6_pktopts * opt,struct route_in6 * ro,int flags,struct ip6_moptions * im6o,struct ifnet ** ifpp,struct ip6_out_args * ip6oa)270 ip6_output_list(struct mbuf *m0, int packetchain, struct ip6_pktopts *opt,
271     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
272     struct ifnet **ifpp, struct ip6_out_args *ip6oa)
273 {
274 	struct ip6_hdr *__single ip6;
275 	u_char *__single nexthdrp;
276 	ifnet_ref_t ifp = NULL, origifp = NULL; /* refcnt'd */
277 	ifnet_ref_t *ifpp_save = ifpp;
278 	mbuf_ref_t m, mprev;
279 	mbuf_ref_t sendchain = NULL, sendchain_last = NULL;
280 	mbuf_ref_t inputchain = NULL;
281 	int nxt0 = 0;
282 	struct route_in6 *__single ro_pmtu = NULL;
283 	rtentry_ref_t rt = NULL;
284 	struct sockaddr_in6 *__single dst = NULL;
285 	struct sockaddr_in6 src_sa, dst_sa;
286 	int error = 0;
287 	struct in6_ifaddr *__single ia = NULL, *__single src_ia = NULL;
288 	u_int32_t mtu = 0;
289 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
290 	struct ip6_rthdr *__single rh;
291 	struct in6_addr finaldst;
292 	ipfilter_t __single inject_filter_ref;
293 	struct ipf_pktopts *__single ippo = NULL;
294 	struct flowadv *__single adv = NULL;
295 	uint32_t pktcnt = 0;
296 	uint32_t packets_processed = 0;
297 	struct timeval start_tv;
298 #if PF
299 	boolean_t skip_pf = (ip6oa != NULL) &&
300 	    (ip6oa->ip6oa_flags & IP6OAF_SKIP_PF);
301 #endif
302 
303 #if DUMMYNET
304 	struct m_tag *__single tag;
305 	struct ip6_out_args saved_ip6oa;
306 	struct sockaddr_in6 dst_buf;
307 #endif /* DUMMYNET */
308 #if IPSEC
309 	struct socket *__single so = NULL;
310 	struct secpolicy *__single sp = NULL;
311 	struct route_in6 *__single ipsec_saved_route = NULL;
312 	boolean_t needipsectun = FALSE;
313 #endif /* IPSEC */
314 #if NECP
315 	necp_kernel_policy_result necp_result = 0;
316 	necp_kernel_policy_result_parameter necp_result_parameter;
317 	necp_kernel_policy_id necp_matched_policy_id = 0;
318 #endif /* NECP */
319 	struct {
320 		struct ipf_pktopts ipf_pktopts;
321 		struct ip6_exthdrs exthdrs;
322 		struct route_in6 ip6route;
323 #if IPSEC
324 		struct ipsec_output_state ipsec_state;
325 #endif /* IPSEC */
326 #if NECP
327 		struct route_in6 necp_route;
328 #endif /* NECP */
329 #if DUMMYNET
330 		struct route_in6 saved_route;
331 		struct route_in6 saved_ro_pmtu;
332 		struct ip_fw_args args;
333 #endif /* DUMMYNET */
334 	} ip6obz;
335 #define ipf_pktopts     ip6obz.ipf_pktopts
336 #define exthdrs         ip6obz.exthdrs
337 #define ip6route        ip6obz.ip6route
338 #define ipsec_state     ip6obz.ipsec_state
339 #define necp_route      ip6obz.necp_route
340 #define saved_route     ip6obz.saved_route
341 #define saved_ro_pmtu   ip6obz.saved_ro_pmtu
342 #define args            ip6obz.args
343 	union {
344 		struct {
345 			boolean_t select_srcif : 1;
346 			boolean_t hdrsplit : 1;
347 			boolean_t route_selected : 1;
348 			boolean_t dontfrag : 1;
349 #if IPSEC
350 			boolean_t needipsec : 1;
351 			boolean_t noipsec : 1;
352 #endif /* IPSEC */
353 		};
354 		uint32_t raw;
355 	} ip6obf = { .raw = 0 };
356 	drop_reason_t drop_reason = DROP_REASON_UNSPECIFIED;
357 
358 	if (ip6_output_measure) {
359 		net_perf_start_time(&net_perf, &start_tv);
360 	}
361 
362 	VERIFY(m0->m_flags & M_PKTHDR);
363 
364 	/* zero out {saved_route, saved_ro_pmtu, ip6route, exthdrs, args} */
365 	bzero(&ip6obz, sizeof(ip6obz));
366 
367 #if DUMMYNET
368 	if (SLIST_EMPTY(&m0->m_pkthdr.tags)) {
369 		goto tags_done;
370 	}
371 
372 	/* Grab info from mtags prepended to the chain */
373 	if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID,
374 	    KERNEL_TAG_TYPE_DUMMYNET)) != NULL) {
375 		struct dn_pkt_tag *__single dn_tag;
376 
377 		/*
378 		 * ip6_output_list() cannot handle chains of packets reinjected
379 		 * by dummynet. The same restriction applies to
380 		 * ip_output_list().
381 		 */
382 		VERIFY(0 == packetchain);
383 
384 		dn_tag = (struct dn_pkt_tag *)(tag->m_tag_data);
385 		args.fwa_pf_rule = dn_tag->dn_pf_rule;
386 
387 		SOCKADDR_COPY(&dn_tag->dn_dst6, &dst_buf, sizeof(dst_buf));
388 		dst = &dst_buf;
389 		ifp = dn_tag->dn_ifp;
390 		if (ifp != NULL) {
391 			ifnet_reference(ifp);
392 		}
393 		flags = dn_tag->dn_flags;
394 		if (dn_tag->dn_flags & IPV6_OUTARGS) {
395 			saved_ip6oa = dn_tag->dn_ip6oa;
396 			ip6oa = &saved_ip6oa;
397 		}
398 
399 		saved_route = dn_tag->dn_ro6;
400 		ro = &saved_route;
401 		saved_ro_pmtu = dn_tag->dn_ro6_pmtu;
402 		ro_pmtu = &saved_ro_pmtu;
403 		origifp = dn_tag->dn_origifp;
404 		if (origifp != NULL) {
405 			ifnet_reference(origifp);
406 		}
407 		mtu = dn_tag->dn_mtu;
408 		unfragpartlen = dn_tag->dn_unfragpartlen;
409 
410 		bcopy(&dn_tag->dn_exthdrs, &exthdrs, sizeof(exthdrs));
411 
412 		m_tag_delete(m0, tag);
413 	}
414 
415 tags_done:
416 #endif /* DUMMYNET */
417 
418 	m = m0;
419 
420 #if IPSEC
421 	if (ipsec_bypass == 0) {
422 		so = ipsec_getsocket(m);
423 		if (so != NULL) {
424 			(void) ipsec_setsocket(m, NULL);
425 		}
426 		/* If packet is bound to an interface, check bound policies */
427 		if ((flags & IPV6_OUTARGS) &&
428 		    (ip6oa->ip6oa_flags & IP6OAF_BOUND_IF) &&
429 		    ip6oa->ip6oa_boundif != IFSCOPE_NONE) {
430 			/* ip6obf.noipsec is a bitfield, use temp integer */
431 			int noipsec = 0;
432 
433 			if (ipsec6_getpolicybyinterface(m, IPSEC_DIR_OUTBOUND,
434 			    flags, ip6oa, &noipsec, &sp) != 0) {
435 				drop_reason = DROP_REASON_IP_OUTBOUND_IPSEC_POLICY;
436 				goto bad;
437 			}
438 
439 			ip6obf.noipsec = (noipsec != 0);
440 		}
441 	}
442 #endif /* IPSEC */
443 
444 	ippo = &ipf_pktopts;
445 
446 	if (flags & IPV6_OUTARGS) {
447 		/*
448 		 * In the forwarding case, only the ifscope value is used,
449 		 * as source interface selection doesn't take place.
450 		 */
451 		if ((ip6obf.select_srcif = (!(flags & (IPV6_FORWARDING |
452 		    IPV6_UNSPECSRC | IPV6_FLAG_NOSRCIFSEL)) &&
453 		    (ip6oa->ip6oa_flags & IP6OAF_SELECT_SRCIF)))) {
454 			ipf_pktopts.ippo_flags |= IPPOF_SELECT_SRCIF;
455 		}
456 
457 		if ((ip6oa->ip6oa_flags & IP6OAF_BOUND_IF) &&
458 		    ip6oa->ip6oa_boundif != IFSCOPE_NONE) {
459 			ipf_pktopts.ippo_flags |= (IPPOF_BOUND_IF |
460 			    (ip6oa->ip6oa_boundif << IPPOF_SHIFT_IFSCOPE));
461 		}
462 
463 		if (ip6oa->ip6oa_flags & IP6OAF_BOUND_SRCADDR) {
464 			ipf_pktopts.ippo_flags |= IPPOF_BOUND_SRCADDR;
465 		}
466 	} else {
467 		ip6obf.select_srcif = FALSE;
468 		if (flags & IPV6_OUTARGS) {
469 			ip6oa->ip6oa_boundif = IFSCOPE_NONE;
470 			ip6oa->ip6oa_flags &= ~(IP6OAF_SELECT_SRCIF |
471 			    IP6OAF_BOUND_IF | IP6OAF_BOUND_SRCADDR);
472 		}
473 	}
474 
475 	if (flags & IPV6_OUTARGS) {
476 		if (ip6oa->ip6oa_flags & IP6OAF_NO_CELLULAR) {
477 			ipf_pktopts.ippo_flags |= IPPOF_NO_IFT_CELLULAR;
478 		}
479 		if (ip6oa->ip6oa_flags & IP6OAF_NO_EXPENSIVE) {
480 			ipf_pktopts.ippo_flags |= IPPOF_NO_IFF_EXPENSIVE;
481 		}
482 		if (ip6oa->ip6oa_flags & IP6OAF_NO_CONSTRAINED) {
483 			ipf_pktopts.ippo_flags |= IPPOF_NO_IFF_CONSTRAINED;
484 		}
485 		adv = &ip6oa->ip6oa_flowadv;
486 		adv->code = FADV_SUCCESS;
487 		ip6oa->ip6oa_flags &= ~IP6OAF_RET_MASK;
488 	}
489 
490 	/*
491 	 * Clear out ifpp to be filled in after determining route. ifpp_save is
492 	 * used to keep old value to release reference properly and dtrace
493 	 * ipsec tunnel traffic properly.
494 	 */
495 	if (ifpp != NULL && *ifpp != NULL) {
496 		*ifpp = NULL;
497 	}
498 
499 #if DUMMYNET
500 	if (args.fwa_pf_rule) {
501 		ip6 = mtod(m, struct ip6_hdr *);
502 		VERIFY(ro != NULL);     /* ro == saved_route */
503 		goto check_with_pf;
504 	}
505 #endif /* DUMMYNET */
506 
507 #if NECP
508 	/*
509 	 * Since all packets are assumed to come from same socket, necp lookup
510 	 * only needs to happen once per function entry.
511 	 */
512 	necp_matched_policy_id = necp_ip6_output_find_policy_match(m, flags,
513 	    (flags & IPV6_OUTARGS) ? ip6oa : NULL, ro ? ro->ro_rt : NULL, &necp_result,
514 	    &necp_result_parameter);
515 #endif /* NECP */
516 
517 	/*
518 	 * If a chain was passed in, prepare for ther first iteration. For all
519 	 * other iterations, this work will be done at evaluateloop: label.
520 	 */
521 	if (packetchain) {
522 		/*
523 		 * Remove m from the chain during processing to avoid
524 		 * accidental frees on entire list.
525 		 */
526 		inputchain = m->m_nextpkt;
527 		m->m_nextpkt = NULL;
528 	}
529 
530 loopit:
531 	packets_processed++;
532 	m->m_pkthdr.pkt_flags &= ~(PKTF_LOOP | PKTF_IFAINFO);
533 	ip6 = mtod(m, struct ip6_hdr *);
534 	nxt0 = ip6->ip6_nxt;
535 	finaldst = ip6->ip6_dst;
536 	ip6obf.hdrsplit = FALSE;
537 	ro_pmtu = NULL;
538 
539 	if (!SLIST_EMPTY(&m->m_pkthdr.tags)) {
540 		inject_filter_ref = ipf_get_inject_filter(m);
541 	} else {
542 		inject_filter_ref = NULL;
543 	}
544 
545 #define MAKE_EXTHDR(hp, mp) do {                                    \
546 	if (hp != NULL) {                                               \
547 	        struct ip6_ext *__single eh = (struct ip6_ext *)(hp);   \
548 	        error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
549 	            ((eh)->ip6e_len + 1) << 3);                         \
550 	        if (error)                                              \
551 	                goto freehdrs;                                  \
552 	}                                                               \
553 } while (0)
554 
555 	if (opt != NULL) {
556 		/* Hop-by-Hop options header */
557 		MAKE_EXTHDR(GET_IP6EXTHDR(opt, struct ip6_hbh*, ip6po_hbh), &exthdrs.ip6e_hbh);
558 		/* Destination options header(1st part) */
559 		if (opt->ip6po_rthdr) {
560 			/*
561 			 * Destination options header(1st part)
562 			 * This only makes sense with a routing header.
563 			 * See Section 9.2 of RFC 3542.
564 			 * Disabling this part just for MIP6 convenience is
565 			 * a bad idea.  We need to think carefully about a
566 			 * way to make the advanced API coexist with MIP6
567 			 * options, which might automatically be inserted in
568 			 * the kernel.
569 			 */
570 			MAKE_EXTHDR(GET_IP6EXTHDR(opt, struct ip6_dest*, ip6po_dest1), &exthdrs.ip6e_dest1);
571 		}
572 		/* Routing header */
573 		MAKE_EXTHDR(GET_IP6EXTHDR(opt, struct ip6_rthdr*, ip6po_rthdr), &exthdrs.ip6e_rthdr);
574 		/* Destination options header(2nd part) */
575 		MAKE_EXTHDR(GET_IP6EXTHDR(opt, struct ip6_dest*, ip6po_dest2), &exthdrs.ip6e_dest2);
576 	}
577 
578 #undef MAKE_EXTHDR
579 
580 #if NECP
581 	if (necp_matched_policy_id) {
582 		necp_mark_packet_from_ip(m, necp_matched_policy_id);
583 
584 		switch (necp_result) {
585 		case NECP_KERNEL_POLICY_RESULT_PASS:
586 			if (necp_result_parameter.pass_flags & NECP_KERNEL_POLICY_PASS_NO_SKIP_IPSEC) {
587 				break;
588 			}
589 			goto skip_ipsec;
590 		case NECP_KERNEL_POLICY_RESULT_DROP:
591 			error = EHOSTUNREACH;
592 			ip6stat.ip6s_necp_policy_drop++;
593 			drop_reason = DROP_REASON_IP_NECP_POLICY_DROP;
594 			goto freehdrs;
595 		case NECP_KERNEL_POLICY_RESULT_SOCKET_DIVERT:
596 			/*
597 			 * Flow divert packets should be blocked at the IP
598 			 * layer.
599 			 */
600 			error = EHOSTUNREACH;
601 			ip6stat.ip6s_necp_policy_drop++;
602 			drop_reason = DROP_REASON_IP_NECP_POLICY_SOCKET_DIVERT;
603 			goto freehdrs;
604 		case NECP_KERNEL_POLICY_RESULT_IP_TUNNEL: {
605 			/*
606 			 * Verify that the packet is being routed to the tunnel
607 			 */
608 			struct ifnet *__single policy_ifp =
609 			    necp_get_ifnet_from_result_parameter(
610 				&necp_result_parameter);
611 
612 			/*
613 			 * Update the QOS marking policy if
614 			 * 1. upper layer asks it to do so
615 			 * 2. net_qos_policy_restricted is not set
616 			 * 3. qos_marking_gencount doesn't match necp_kernel_socket_policies_gencount (checked in necp_lookup_current_qos_marking)
617 			 */
618 			if (ip6oa != NULL && (ip6oa->ip6oa_flags & IP6OAF_REDO_QOSMARKING_POLICY) &&
619 			    net_qos_policy_restricted != 0) {
620 				bool qos_marking = (ip6oa->ip6oa_flags & IP6OAF_QOSMARKING_ALLOWED) != 0;
621 				qos_marking = necp_lookup_current_qos_marking(&ip6oa->qos_marking_gencount, NULL, policy_ifp, necp_result_parameter.route_rule_id, qos_marking);
622 				if (qos_marking) {
623 					ip6oa->ip6oa_flags |= IP6OAF_QOSMARKING_ALLOWED;
624 				} else {
625 					ip6oa->ip6oa_flags &= ~IP6OAF_QOSMARKING_ALLOWED;
626 				}
627 			}
628 
629 			if (policy_ifp == ifp) {
630 				goto skip_ipsec;
631 			} else {
632 				if (necp_packet_can_rebind_to_ifnet(m,
633 				    policy_ifp, (struct route *)&necp_route,
634 				    AF_INET6)) {
635 					/*
636 					 * Set scoped index to the tunnel
637 					 * interface, since it is compatible
638 					 * with the packet. This will only work
639 					 * for callers who pass IPV6_OUTARGS,
640 					 * but that covers all of the clients
641 					 * we care about today.
642 					 */
643 					if (flags & IPV6_OUTARGS) {
644 						ip6oa->ip6oa_boundif =
645 						    policy_ifp->if_index;
646 						ip6oa->ip6oa_flags |=
647 						    IP6OAF_BOUND_IF;
648 					}
649 					if (opt != NULL
650 					    && opt->ip6po_pktinfo != NULL) {
651 						opt->ip6po_pktinfo->
652 						ipi6_ifindex =
653 						    policy_ifp->if_index;
654 					}
655 					ro = &necp_route;
656 					goto skip_ipsec;
657 				} else {
658 					error = ENETUNREACH;
659 					ip6stat.ip6s_necp_policy_drop++;
660 					drop_reason = DROP_REASON_IP_NECP_POLICY_TUN_NO_REBIND_IF;
661 					goto freehdrs;
662 				}
663 			}
664 		}
665 		default:
666 			break;
667 		}
668 	}
669 #endif /* NECP */
670 
671 #if IPSEC
672 	if (ipsec_bypass != 0 || ip6obf.noipsec) {
673 		goto skip_ipsec;
674 	}
675 
676 	if (sp == NULL) {
677 		/* get a security policy for this packet */
678 		if (so != NULL) {
679 			sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND,
680 			    so, &error);
681 		} else {
682 			sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
683 			    0, &error);
684 		}
685 		if (sp == NULL) {
686 			IPSEC_STAT_INCREMENT(ipsec6stat.out_inval);
687 			goto freehdrs;
688 		}
689 	}
690 
691 	error = 0;
692 
693 	/* check policy */
694 	switch (sp->policy) {
695 	case IPSEC_POLICY_DISCARD:
696 	case IPSEC_POLICY_GENERATE:
697 		/*
698 		 * This packet is just discarded.
699 		 */
700 		IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio);
701 		goto freehdrs;
702 
703 	case IPSEC_POLICY_BYPASS:
704 	case IPSEC_POLICY_NONE:
705 		/* no need to do IPsec. */
706 		ip6obf.needipsec = FALSE;
707 		break;
708 
709 	case IPSEC_POLICY_IPSEC:
710 		if (sp->req == NULL) {
711 			/* acquire a policy */
712 			error = key_spdacquire(sp);
713 			goto freehdrs;
714 		}
715 		if (sp->ipsec_if) {
716 			goto skip_ipsec;
717 		} else {
718 			ip6obf.needipsec = true;
719 		}
720 		break;
721 
722 	case IPSEC_POLICY_ENTRUST:
723 	default:
724 		printf("%s: Invalid policy found: %d\n", __func__, sp->policy);
725 		break;
726 	}
727 skip_ipsec:
728 #endif /* IPSEC */
729 
730 	/*
731 	 * Calculate the total length of the extension header chain.
732 	 * Keep the length of the unfragmentable part for fragmentation.
733 	 */
734 	optlen = 0;
735 	if (exthdrs.ip6e_hbh != NULL) {
736 		optlen += exthdrs.ip6e_hbh->m_len;
737 	}
738 	if (exthdrs.ip6e_dest1 != NULL) {
739 		optlen += exthdrs.ip6e_dest1->m_len;
740 	}
741 	if (exthdrs.ip6e_rthdr != NULL) {
742 		optlen += exthdrs.ip6e_rthdr->m_len;
743 	}
744 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
745 
746 	/* NOTE: we don't add AH/ESP length here. do that later. */
747 	if (exthdrs.ip6e_dest2 != NULL) {
748 		optlen += exthdrs.ip6e_dest2->m_len;
749 	}
750 
751 	/*
752 	 * If we need IPsec, or there is at least one extension header,
753 	 * separate IP6 header from the payload.
754 	 */
755 	if ((
756 #if IPSEC
757 		    ip6obf.needipsec ||
758 #endif /* IPSEC */
759 		    optlen) && !ip6obf.hdrsplit) {
760 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
761 			m = NULL;
762 			goto freehdrs;
763 		}
764 		m = exthdrs.ip6e_ip6;
765 		ip6obf.hdrsplit = true;
766 	}
767 
768 	/* adjust pointer */
769 	ip6 = mtod(m, struct ip6_hdr *);
770 
771 	/* adjust mbuf packet header length */
772 	m->m_pkthdr.len += optlen;
773 	plen = m->m_pkthdr.len - sizeof(*ip6);
774 
775 	/* If this is a jumbo payload, insert a jumbo payload option. */
776 	if (plen > IPV6_MAXPACKET) {
777 		if (!ip6obf.hdrsplit) {
778 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
779 				m = NULL;
780 				goto freehdrs;
781 			}
782 			m = exthdrs.ip6e_ip6;
783 			ip6obf.hdrsplit = true;
784 		}
785 		/* adjust pointer */
786 		ip6 = mtod(m, struct ip6_hdr *);
787 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) {
788 			goto freehdrs;
789 		}
790 		ip6->ip6_plen = 0;
791 	} else {
792 		ip6->ip6_plen = htons((uint16_t)plen);
793 	}
794 	/*
795 	 * Concatenate headers and fill in next header fields.
796 	 * Here we have, on "m"
797 	 *	IPv6 payload
798 	 * and we insert headers accordingly.  Finally, we should be getting:
799 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
800 	 *
801 	 * during the header composing process, "m" points to IPv6 header.
802 	 * "mprev" points to an extension header prior to esp.
803 	 */
804 	nexthdrp = &ip6->ip6_nxt;
805 	mprev = m;
806 
807 	/*
808 	 * we treat dest2 specially.  this makes IPsec processing
809 	 * much easier.  the goal here is to make mprev point the
810 	 * mbuf prior to dest2.
811 	 *
812 	 * result: IPv6 dest2 payload
813 	 * m and mprev will point to IPv6 header.
814 	 */
815 	if (exthdrs.ip6e_dest2 != NULL) {
816 		if (!ip6obf.hdrsplit) {
817 			panic("assumption failed: hdr not split");
818 			/* NOTREACHED */
819 		}
820 		exthdrs.ip6e_dest2->m_next = m->m_next;
821 		m->m_next = exthdrs.ip6e_dest2;
822 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
823 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
824 	}
825 
826 #define MAKE_CHAIN(m, mp, p, i) do {                                    \
827 	if (m != NULL) {                                                \
828 	        if (!ip6obf.hdrsplit) {                                 \
829 	                panic("assumption failed: hdr not split");      \
830 	/* NOTREACHED */                                \
831 	        }                                                       \
832 	        *mtod((m), u_char *) = *(p);                            \
833 	        *(p) = (i);                                             \
834 	        p = mtod((m), u_char *);                                \
835 	        (m)->m_next = (mp)->m_next;                             \
836 	        (mp)->m_next = (m);                                     \
837 	        (mp) = (m);                                             \
838 	}                                                               \
839 } while (0)
840 	/*
841 	 * result: IPv6 hbh dest1 rthdr dest2 payload
842 	 * m will point to IPv6 header.  mprev will point to the
843 	 * extension header prior to dest2 (rthdr in the above case).
844 	 */
845 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
846 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
847 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
848 
849 	/* It is no longer safe to free the pointers in exthdrs. */
850 	exthdrs.merged = TRUE;
851 
852 #undef MAKE_CHAIN
853 
854 #if IPSEC
855 	if (ip6obf.needipsec && (m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA)) {
856 		in6_delayed_cksum_offset(m, 0, optlen, nxt0);
857 	}
858 #endif /* IPSEC */
859 
860 	if (!TAILQ_EMPTY(&ipv6_filters) &&
861 	    !((flags & IPV6_OUTARGS) &&
862 	    (ip6oa->ip6oa_flags & IP6OAF_INTCOPROC_ALLOWED) &&
863 	    (ip6oa->ip6oa_flags & IP6OAF_MANAGEMENT_ALLOWED)
864 #if NECP
865 	    && !necp_packet_should_skip_filters(m)
866 #endif // NECP
867 	    )) {
868 		struct ipfilter *__single filter;
869 		int seen = (inject_filter_ref == NULL);
870 		int fixscope = 0;
871 
872 		if (im6o != NULL && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
873 			ippo->ippo_flags |= IPPOF_MCAST_OPTS;
874 			IM6O_LOCK(im6o);
875 			ippo->ippo_mcast_ifnet = im6o->im6o_multicast_ifp;
876 			ippo->ippo_mcast_ttl = im6o->im6o_multicast_hlim;
877 			ippo->ippo_mcast_loop = im6o->im6o_multicast_loop;
878 			IM6O_UNLOCK(im6o);
879 		}
880 
881 		/* Hack: embed the scope_id in the destination */
882 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
883 		    (ip6->ip6_dst.s6_addr16[1] == 0) && (ro != NULL)) {
884 			fixscope = 1;
885 			ip6->ip6_dst.s6_addr16[1] =
886 			    htons((uint16_t)ro->ro_dst.sin6_scope_id);
887 		}
888 
889 		ipf_ref();
890 		TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) {
891 			/*
892 			 * Don't process packet twice if we've already seen it.
893 			 */
894 			if (seen == 0) {
895 				if ((struct ipfilter *)inject_filter_ref ==
896 				    filter) {
897 					seen = 1;
898 				}
899 			} else if (filter->ipf_filter.ipf_output != NULL) {
900 				errno_t result;
901 
902 				result = filter->ipf_filter.ipf_output(
903 					filter->ipf_filter.cookie,
904 					(mbuf_t *)&m, ippo);
905 				if (result == EJUSTRETURN) {
906 					ipf_unref();
907 					m = NULL;
908 					goto evaluateloop;
909 				}
910 				if (result != 0) {
911 					ipf_unref();
912 					drop_reason = DROP_REASON_IP_FILTER_DROP;
913 					goto bad;
914 				}
915 			}
916 		}
917 		ipf_unref();
918 
919 		ip6 = mtod(m, struct ip6_hdr *);
920 		/* Hack: cleanup embedded scope_id if we put it there */
921 		if (fixscope) {
922 			ip6->ip6_dst.s6_addr16[1] = 0;
923 		}
924 	}
925 
926 #if IPSEC
927 	if (ip6obf.needipsec) {
928 		uint8_t segleft_org;
929 
930 		/*
931 		 * pointers after IPsec headers are not valid any more.
932 		 * other pointers need a great care too.
933 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
934 		 */
935 		exthdrs.ip6e_dest2 = NULL;
936 
937 		if (exthdrs.ip6e_rthdr != NULL) {
938 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
939 			segleft_org = rh->ip6r_segleft;
940 			rh->ip6r_segleft = 0;
941 		} else {
942 			rh = NULL;
943 			segleft_org = 0;
944 		}
945 
946 		ipsec_state.m = m;
947 		error = ipsec6_output_trans(&ipsec_state, nexthdrp, mprev,
948 		    sp, flags, &needipsectun);
949 		m = ipsec_state.m;
950 		if (error) {
951 			/* mbuf is already reclaimed in ipsec6_output_trans. */
952 			m = NULL;
953 			switch (error) {
954 			case EHOSTUNREACH:
955 			case ENETUNREACH:
956 			case EMSGSIZE:
957 			case ENOBUFS:
958 			case ENOMEM:
959 				break;
960 			default:
961 				printf("ip6_output (ipsec): error code %d\n",
962 				    error);
963 				OS_FALLTHROUGH;
964 			case ENOENT:
965 				/* don't show these error codes to the user */
966 				error = 0;
967 				break;
968 			}
969 			goto bad;
970 		}
971 		if (exthdrs.ip6e_rthdr != NULL) {
972 			/* ah6_output doesn't modify mbuf chain */
973 			rh->ip6r_segleft = segleft_org;
974 		}
975 	}
976 #endif /* IPSEC */
977 
978 	/* If there is a routing header, discard the packet. */
979 	if (exthdrs.ip6e_rthdr != NULL) {
980 		error = EINVAL;
981 		goto bad;
982 	}
983 
984 	/* Source address validation */
985 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
986 	    !(flags & IPV6_UNSPECSRC)) {
987 		error = EOPNOTSUPP;
988 		ip6stat.ip6s_badscope++;
989 		drop_reason = DROP_REASON_IP6_BAD_SCOPE;
990 		goto bad;
991 	}
992 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
993 		error = EOPNOTSUPP;
994 		ip6stat.ip6s_badscope++;
995 		drop_reason = DROP_REASON_IP6_BAD_SCOPE;
996 		goto bad;
997 	}
998 
999 	ip6stat.ip6s_localout++;
1000 
1001 	/*
1002 	 * Route packet.
1003 	 */
1004 	if (ro == NULL) {
1005 		ro = &ip6route;
1006 		bzero(ro, sizeof(*ro));
1007 	}
1008 	ro_pmtu = ro;
1009 	if (opt != NULL && opt->ip6po_rthdr) {
1010 		ro = &opt->ip6po_route;
1011 	}
1012 	dst = SIN6(&ro->ro_dst);
1013 
1014 	if (ro->ro_rt != NULL) {
1015 		RT_LOCK_ASSERT_NOTHELD(ro->ro_rt);
1016 	}
1017 	/*
1018 	 * if specified, try to fill in the traffic class field.
1019 	 * do not override if a non-zero value is already set.
1020 	 * we check the diffserv field and the ecn field separately.
1021 	 */
1022 	if (opt != NULL && opt->ip6po_tclass >= 0) {
1023 		int mask = 0;
1024 
1025 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) {
1026 			mask |= 0xfc;
1027 		}
1028 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) {
1029 			mask |= 0x03;
1030 		}
1031 		if (mask != 0) {
1032 			ip6->ip6_flow |=
1033 			    htonl((opt->ip6po_tclass & mask) << 20);
1034 		}
1035 	}
1036 
1037 	if (((ntohl(ip6->ip6_flow & IPV6_FLOW_ECN_MASK) >> 20) & IPTOS_ECN_ECT1) == IPTOS_ECN_ECT1) {
1038 		m->m_pkthdr.pkt_ext_flags |= PKTF_EXT_L4S;
1039 	}
1040 
1041 	/* fill in or override the hop limit field, if necessary. */
1042 	if (opt && opt->ip6po_hlim != -1) {
1043 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
1044 	} else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1045 		if (im6o != NULL) {
1046 			IM6O_LOCK(im6o);
1047 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
1048 			IM6O_UNLOCK(im6o);
1049 		} else {
1050 			ip6->ip6_hlim = (uint8_t)ip6_defmcasthlim;
1051 		}
1052 	}
1053 
1054 	/*
1055 	 * If there is a cached route, check that it is to the same
1056 	 * destination and is still up. If not, free it and try again.
1057 	 * Test rt_flags without holding rt_lock for performance reasons;
1058 	 * if the route is down it will hopefully be caught by the layer
1059 	 * below (since it uses this route as a hint) or during the
1060 	 * next transmit.
1061 	 */
1062 	if (ROUTE_UNUSABLE(ro) || dst->sin6_family != AF_INET6 ||
1063 	    !in6_are_addr_equal_scoped(&dst->sin6_addr, &ip6->ip6_dst, dst->sin6_scope_id, ip6_output_getdstifscope(m))) {
1064 		ROUTE_RELEASE(ro);
1065 	}
1066 
1067 	if (ro->ro_rt == NULL) {
1068 		SOCKADDR_ZERO(dst, sizeof(*dst));
1069 		dst->sin6_family = AF_INET6;
1070 		dst->sin6_len = sizeof(struct sockaddr_in6);
1071 		dst->sin6_addr = ip6->ip6_dst;
1072 	}
1073 #if IPSEC
1074 	if (ip6obf.needipsec && needipsectun) {
1075 #if CONFIG_DTRACE
1076 		struct ifnet *trace_ifp = (ifpp_save != NULL) ? (*ifpp_save) : NULL;
1077 #endif /* CONFIG_DTRACE */
1078 		/*
1079 		 * All the extension headers will become inaccessible
1080 		 * (since they can be encrypted).
1081 		 * Don't panic, we need no more updates to extension headers
1082 		 * on inner IPv6 packet (since they are now encapsulated).
1083 		 *
1084 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
1085 		 */
1086 		bzero(&exthdrs, sizeof(exthdrs));
1087 		exthdrs.ip6e_ip6 = m;
1088 
1089 		ipsec_state.m = m;
1090 		route_copyout((struct route *)&ipsec_state.ro, (struct route *)ro,
1091 		    sizeof(struct route_in6));
1092 		ipsec_state.dst = SA(dst);
1093 
1094 		/* So that we can see packets inside the tunnel */
1095 		DTRACE_IP6(send, struct mbuf *, m, struct inpcb *, NULL,
1096 		    struct ip6_hdr *, ip6, struct ifnet *, trace_ifp,
1097 		    struct ip *, NULL, struct ip6_hdr *, ip6);
1098 
1099 		error = ipsec6_output_tunnel(&ipsec_state, sp, flags);
1100 		/* tunneled in IPv4? packet is gone */
1101 		if (ipsec_state.tunneled == 4) {
1102 			m = NULL;
1103 			goto evaluateloop;
1104 		}
1105 		m = ipsec_state.m;
1106 		ipsec_saved_route = ro;
1107 		ro = (struct route_in6 *)&ipsec_state.ro;
1108 		dst = SIN6(ipsec_state.dst);
1109 		if (error) {
1110 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
1111 			m = NULL;
1112 			switch (error) {
1113 			case EHOSTUNREACH:
1114 			case ENETUNREACH:
1115 			case EMSGSIZE:
1116 			case ENOBUFS:
1117 			case ENOMEM:
1118 				break;
1119 			default:
1120 				printf("ip6_output (ipsec): error code %d\n",
1121 				    error);
1122 				OS_FALLTHROUGH;
1123 			case ENOENT:
1124 				/* don't show these error codes to the user */
1125 				error = 0;
1126 				break;
1127 			}
1128 			goto bad;
1129 		}
1130 		/*
1131 		 * The packet has been encapsulated so the ifscope
1132 		 * is no longer valid since it does not apply to the
1133 		 * outer address: ignore the ifscope.
1134 		 */
1135 		if (flags & IPV6_OUTARGS) {
1136 			ip6oa->ip6oa_boundif = IFSCOPE_NONE;
1137 			ip6oa->ip6oa_flags &= ~IP6OAF_BOUND_IF;
1138 		}
1139 		if (opt != NULL && opt->ip6po_pktinfo != NULL) {
1140 			if (opt->ip6po_pktinfo->ipi6_ifindex != IFSCOPE_NONE) {
1141 				opt->ip6po_pktinfo->ipi6_ifindex = IFSCOPE_NONE;
1142 			}
1143 		}
1144 		exthdrs.ip6e_ip6 = m;
1145 	}
1146 #endif /* IPSEC */
1147 
1148 	/*
1149 	 * ifp should only be filled in for dummy net packets which will jump
1150 	 * to check_with_pf label.
1151 	 */
1152 	if (ifp != NULL) {
1153 		VERIFY(ip6obf.route_selected);
1154 	}
1155 
1156 	/* adjust pointer */
1157 	ip6 = mtod(m, struct ip6_hdr *);
1158 
1159 	if (ip6obf.select_srcif) {
1160 		SOCKADDR_ZERO(&src_sa, sizeof(src_sa));
1161 		src_sa.sin6_family = AF_INET6;
1162 		src_sa.sin6_len = sizeof(src_sa);
1163 		src_sa.sin6_addr = ip6->ip6_src;
1164 		src_sa.sin6_scope_id = (!in6_embedded_scope && IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) ? ip6_output_getsrcifscope(m) : IFSCOPE_NONE;
1165 	}
1166 	SOCKADDR_ZERO(&dst_sa, sizeof(dst_sa));
1167 	dst_sa.sin6_family = AF_INET6;
1168 	dst_sa.sin6_len = sizeof(dst_sa);
1169 	dst_sa.sin6_addr = ip6->ip6_dst;
1170 	dst_sa.sin6_scope_id = (!in6_embedded_scope && IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) ? ip6_output_getdstifscope(m) : IFSCOPE_NONE;
1171 
1172 	/*
1173 	 * Only call in6_selectroute() on first iteration to avoid taking
1174 	 * multiple references on ifp and rt.
1175 	 *
1176 	 * in6_selectroute() might return an ifp with its reference held
1177 	 * even in the error case, so make sure to release its reference.
1178 	 * ip6oa may be NULL if IPV6_OUTARGS isn't set.
1179 	 */
1180 	if (!ip6obf.route_selected) {
1181 		error = in6_selectroute( ip6obf.select_srcif ? &src_sa : NULL,
1182 		    &dst_sa, opt, im6o, &src_ia, ro, &ifp, &rt, 0, ip6oa);
1183 
1184 		if (error != 0) {
1185 			switch (error) {
1186 			case EHOSTUNREACH:
1187 				drop_reason = DROP_REASON_IP_NO_ROUTE;
1188 				ip6stat.ip6s_noroute++;
1189 				break;
1190 			case EADDRNOTAVAIL:
1191 				drop_reason = DROP_REASON_IP_SRC_ADDR_NO_AVAIL;
1192 			default:
1193 				break; /* XXX statistics? */
1194 			}
1195 			if (ifp != NULL) {
1196 				in6_ifstat_inc(ifp, ifs6_out_discard);
1197 			}
1198 			/* ifp (if non-NULL) will be released at the end */
1199 			goto bad;
1200 		}
1201 		ip6obf.route_selected = true;
1202 	}
1203 	if (rt == NULL) {
1204 		/*
1205 		 * If in6_selectroute() does not return a route entry,
1206 		 * dst may not have been updated.
1207 		 */
1208 		*dst = dst_sa;  /* XXX */
1209 	}
1210 
1211 #if NECP
1212 	/* Catch-all to check if the interface is allowed */
1213 	if (!necp_packet_is_allowed_over_interface(m, ifp)) {
1214 		error = EHOSTUNREACH;
1215 		ip6stat.ip6s_necp_policy_drop++;
1216 		drop_reason = DROP_REASON_IP_NECP_POLICY_NO_ALLOW_IF;
1217 		goto bad;
1218 	}
1219 #endif /* NECP */
1220 
1221 	/*
1222 	 * then rt (for unicast) and ifp must be non-NULL valid values.
1223 	 */
1224 	if (!(flags & IPV6_FORWARDING)) {
1225 		in6_ifstat_inc_na(ifp, ifs6_out_request);
1226 	}
1227 	if (rt != NULL) {
1228 		RT_LOCK(rt);
1229 		if (ia == NULL) {
1230 			ia = ifatoia6(rt->rt_ifa);
1231 			if (ia != NULL) {
1232 				ifa_addref(&ia->ia_ifa);
1233 			}
1234 		}
1235 		rt->rt_use++;
1236 		RT_UNLOCK(rt);
1237 	}
1238 
1239 	/*
1240 	 * The outgoing interface must be in the zone of source and
1241 	 * destination addresses (except local/loopback).  We should
1242 	 * use ia_ifp to support the case of sending packets to an
1243 	 * address of our own.
1244 	 */
1245 	if (ia != NULL && ia->ia_ifp) {
1246 		ifnet_reference(ia->ia_ifp);    /* for origifp */
1247 		if (origifp != NULL) {
1248 			ifnet_release(origifp);
1249 		}
1250 		origifp = ia->ia_ifp;
1251 	} else {
1252 		if (ifp != NULL) {
1253 			ifnet_reference(ifp);   /* for origifp */
1254 		}
1255 		if (origifp != NULL) {
1256 			ifnet_release(origifp);
1257 		}
1258 		origifp = ifp;
1259 	}
1260 
1261 	/* skip scope enforcements for local/loopback route */
1262 	if (rt == NULL || rt->rt_ifp == NULL || !(rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
1263 		struct in6_addr src0, dst0;
1264 		u_int32_t zone;
1265 
1266 		src0 = ip6->ip6_src;
1267 		if (in6_setscope(&src0, origifp, &zone)) {
1268 			goto badscope;
1269 		}
1270 		SOCKADDR_ZERO(&src_sa, sizeof(src_sa));
1271 		src_sa.sin6_family = AF_INET6;
1272 		src_sa.sin6_len = sizeof(src_sa);
1273 		src_sa.sin6_addr = ip6->ip6_src;
1274 		src_sa.sin6_scope_id = (!in6_embedded_scope && IN6_IS_SCOPE_EMBED(&src_sa.sin6_addr)) ? ip6_output_getsrcifscope(m) : IFSCOPE_NONE;
1275 		if ((sa6_recoverscope(&src_sa, TRUE) ||
1276 		    zone != src_sa.sin6_scope_id)) {
1277 			goto badscope;
1278 		}
1279 
1280 		dst0 = ip6->ip6_dst;
1281 		if ((in6_setscope(&dst0, origifp, &zone))) {
1282 			goto badscope;
1283 		}
1284 		/* re-initialize to be sure */
1285 		SOCKADDR_ZERO(&dst_sa, sizeof(dst_sa));
1286 		dst_sa.sin6_family = AF_INET6;
1287 		dst_sa.sin6_len = sizeof(dst_sa);
1288 		dst_sa.sin6_addr = ip6->ip6_dst;
1289 		dst_sa.sin6_scope_id = (!in6_embedded_scope && IN6_IS_SCOPE_EMBED(&dst_sa.sin6_addr)) ?  ip6_output_getdstifscope(m) : IFSCOPE_NONE;
1290 		if ((sa6_recoverscope(&dst_sa, TRUE) ||
1291 		    zone != dst_sa.sin6_scope_id)) {
1292 			goto badscope;
1293 		}
1294 
1295 		/* scope check is done. */
1296 		goto routefound;
1297 
1298 badscope:
1299 		ip6stat.ip6s_badscope++;
1300 		in6_ifstat_inc(origifp, ifs6_out_discard);
1301 		if (error == 0) {
1302 			error = EHOSTUNREACH; /* XXX */
1303 		}
1304 		drop_reason = DROP_REASON_IP6_BAD_SCOPE;
1305 		goto bad;
1306 	}
1307 
1308 routefound:
1309 	if (rt != NULL && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1310 		if (opt != NULL && opt->ip6po_nextroute.ro_rt) {
1311 			/*
1312 			 * The nexthop is explicitly specified by the
1313 			 * application.  We assume the next hop is an IPv6
1314 			 * address.
1315 			 */
1316 			dst = SIN6(opt->ip6po_nexthop);
1317 		} else if ((rt->rt_flags & RTF_GATEWAY)) {
1318 			dst = SIN6(rt->rt_gateway);
1319 		}
1320 		/*
1321 		 * For packets destined to local/loopback, record the
1322 		 * source the source interface (which owns the source
1323 		 * address), as well as the output interface.  This is
1324 		 * needed to reconstruct the embedded zone for the
1325 		 * link-local address case in ip6_input().
1326 		 */
1327 		if (ia != NULL && (ifp->if_flags & IFF_LOOPBACK)) {
1328 			uint32_t srcidx;
1329 
1330 			if (src_ia != NULL) {
1331 				srcidx = src_ia->ia_ifp->if_index;
1332 			} else if (ro->ro_srcia != NULL) {
1333 				srcidx = ro->ro_srcia->ifa_ifp->if_index;
1334 			} else {
1335 				srcidx = 0;
1336 			}
1337 
1338 			ip6_setsrcifaddr_info(m, srcidx, NULL);
1339 			ip6_setdstifaddr_info(m, 0, ia);
1340 		}
1341 	}
1342 
1343 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1344 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
1345 	} else {
1346 		struct in6_multi *__single in6m;
1347 
1348 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
1349 		in6_ifstat_inc_na(ifp, ifs6_out_mcast);
1350 
1351 		/*
1352 		 * Confirm that the outgoing interface supports multicast.
1353 		 */
1354 		if (!(ifp->if_flags & IFF_MULTICAST)) {
1355 			ip6stat.ip6s_noroute++;
1356 			in6_ifstat_inc(ifp, ifs6_out_discard);
1357 			error = ENETUNREACH;
1358 			drop_reason = DROP_REASON_IP_NO_ROUTE;
1359 			goto bad;
1360 		}
1361 		in6_multihead_lock_shared();
1362 		IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp, in6m);
1363 		in6_multihead_lock_done();
1364 		if (im6o != NULL) {
1365 			IM6O_LOCK(im6o);
1366 		}
1367 		if (in6m != NULL &&
1368 		    (im6o == NULL || im6o->im6o_multicast_loop)) {
1369 			if (im6o != NULL) {
1370 				IM6O_UNLOCK(im6o);
1371 			}
1372 			/*
1373 			 * If we belong to the destination multicast group
1374 			 * on the outgoing interface, and the caller did not
1375 			 * forbid loopback, loop back a copy.
1376 			 */
1377 			ip6_mloopback(NULL, ifp, m, dst, optlen, nxt0);
1378 		} else if (im6o != NULL) {
1379 			IM6O_UNLOCK(im6o);
1380 		}
1381 		if (in6m != NULL) {
1382 			IN6M_REMREF(in6m);
1383 		}
1384 		/*
1385 		 * Multicasts with a hoplimit of zero may be looped back,
1386 		 * above, but must not be transmitted on a network.
1387 		 * Also, multicasts addressed to the loopback interface
1388 		 * are not sent -- the above call to ip6_mloopback() will
1389 		 * loop back a copy if this host actually belongs to the
1390 		 * destination group on the loopback interface.
1391 		 */
1392 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
1393 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
1394 			/* remove m from the packetchain and continue looping */
1395 			if (m != NULL) {
1396 				m_freem(m);
1397 			}
1398 			m = NULL;
1399 			goto evaluateloop;
1400 		}
1401 	}
1402 
1403 	/*
1404 	 * Fill the outgoing inteface to tell the upper layer
1405 	 * to increment per-interface statistics.
1406 	 */
1407 	if (ifpp != NULL && *ifpp == NULL) {
1408 		ifnet_reference(ifp);   /* for caller */
1409 		*ifpp = ifp;
1410 	}
1411 
1412 	/* Determine path MTU. */
1413 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, ifp->if_index, &mtu)) != 0) {
1414 		drop_reason = DROP_REASON_IP6_BAD_PATH_MTU;
1415 		goto bad;
1416 	}
1417 
1418 	/*
1419 	 * The caller of this function may specify to use the minimum MTU
1420 	 * in some cases.
1421 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
1422 	 * setting.  The logic is a bit complicated; by default, unicast
1423 	 * packets will follow path MTU while multicast packets will be sent at
1424 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
1425 	 * including unicast ones will be sent at the minimum MTU.  Multicast
1426 	 * packets will always be sent at the minimum MTU unless
1427 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
1428 	 * See RFC 3542 for more details.
1429 	 */
1430 	if (mtu > IPV6_MMTU) {
1431 		if ((flags & IPV6_MINMTU)) {
1432 			mtu = IPV6_MMTU;
1433 		} else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) {
1434 			mtu = IPV6_MMTU;
1435 		} else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
1436 		    (opt == NULL ||
1437 		    opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
1438 			mtu = IPV6_MMTU;
1439 		}
1440 	}
1441 
1442 	/*
1443 	 * clear embedded scope identifiers if necessary.
1444 	 * in6_clearscope will touch the addresses only when necessary.
1445 	 */
1446 	in6_clearscope(&ip6->ip6_src);
1447 	in6_clearscope(&ip6->ip6_dst);
1448 	/*
1449 	 * If the outgoing packet contains a hop-by-hop options header,
1450 	 * it must be examined and processed even by the source node.
1451 	 * (RFC 2460, section 4.)
1452 	 */
1453 	if (exthdrs.ip6e_hbh != NULL) {
1454 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
1455 		u_int32_t dummy; /* XXX unused */
1456 		uint32_t oplen = 0; /* for ip6_process_hopopts() */
1457 #if DIAGNOSTIC
1458 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) {
1459 			panic("ip6e_hbh is not continuous");
1460 		}
1461 #endif
1462 		/*
1463 		 * XXX: If we have to send an ICMPv6 error to the sender,
1464 		 * we need the M_LOOP flag since icmp6_error() expects
1465 		 * the IPv6 and the hop-by-hop options header are
1466 		 * continuous unless the flag is set.
1467 		 */
1468 		m->m_flags |= M_LOOP;
1469 		m->m_pkthdr.rcvif = ifp;
1470 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1471 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1472 		    &dummy, &oplen) < 0) {
1473 			/*
1474 			 * m was already freed at this point. Set to NULL so it
1475 			 * is not re-freed at end of ip6_output_list.
1476 			 */
1477 			m = NULL;
1478 			error = EINVAL; /* better error? */
1479 			goto bad;
1480 		}
1481 		m->m_flags &= ~M_LOOP; /* XXX */
1482 		m->m_pkthdr.rcvif = NULL;
1483 	}
1484 
1485 #if DUMMYNET
1486 check_with_pf:
1487 #endif /* DUMMYNET */
1488 #if PF
1489 	if (PF_IS_ENABLED && !skip_pf) {
1490 #if DUMMYNET
1491 
1492 		/*
1493 		 * TODO: Need to save opt->ip6po_flags for reinjection
1494 		 * rdar://10434993
1495 		 */
1496 		args.fwa_oif = ifp;
1497 		args.fwa_oflags = flags;
1498 		if (flags & IPV6_OUTARGS) {
1499 			args.fwa_ip6oa = ip6oa;
1500 		}
1501 		args.fwa_ro6 = ro;
1502 		args.fwa_dst6 = dst;
1503 		args.fwa_ro6_pmtu = ro_pmtu;
1504 		args.fwa_origifp = origifp;
1505 		args.fwa_mtu = mtu;
1506 		args.fwa_unfragpartlen = unfragpartlen;
1507 		args.fwa_exthdrs = &exthdrs;
1508 		/* Invoke outbound packet filter */
1509 		error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, &args);
1510 #else /* !DUMMYNET */
1511 		error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL);
1512 #endif /* !DUMMYNET */
1513 
1514 		if (error != 0 || m == NULL) {
1515 			if (m != NULL) {
1516 				panic("%s: unexpected packet %p",
1517 				    __func__, m);
1518 				/* NOTREACHED */
1519 			}
1520 			/* m was already freed by callee and is now NULL.  */
1521 			goto evaluateloop;
1522 		}
1523 		ip6 = mtod(m, struct ip6_hdr *);
1524 	}
1525 #endif /* PF */
1526 
1527 #ifdef IPSEC
1528 	/* clean ipsec history before fragmentation */
1529 	ipsec_delaux(m);
1530 #endif /* IPSEC */
1531 
1532 	if (ip6oa != NULL) {
1533 		u_int8_t dscp;
1534 
1535 		dscp = (ntohl(ip6->ip6_flow) & IP6FLOW_DSCP_MASK) >> IP6FLOW_DSCP_SHIFT;
1536 
1537 		error = set_packet_qos(m, ifp,
1538 		    ip6oa->ip6oa_flags & IP6OAF_QOSMARKING_ALLOWED ? TRUE : FALSE,
1539 		    ip6oa->ip6oa_sotc, ip6oa->ip6oa_netsvctype, &dscp);
1540 		if (error == 0) {
1541 			ip6->ip6_flow &= ~htonl(IP6FLOW_DSCP_MASK);
1542 			ip6->ip6_flow |= htonl((u_int32_t)dscp << IP6FLOW_DSCP_SHIFT);
1543 		} else {
1544 			printf("%s if_dscp_for_mbuf() error %d\n", __func__, error);
1545 			error = 0;
1546 		}
1547 	}
1548 	/*
1549 	 * Determine whether fragmentation is necessary. If so, m is passed
1550 	 * back as a chain of packets and original mbuf is freed. Otherwise, m
1551 	 * is unchanged.
1552 	 */
1553 	error = ip6_fragment_packet(&m, opt, ip6oa,
1554 	    &exthdrs, ifp, mtu, unfragpartlen, nxt0,
1555 	    optlen);
1556 
1557 	if (error) {
1558 		drop_reason = DROP_REASON_IP_CANNOT_FRAGMENT;
1559 		goto bad;
1560 	}
1561 
1562 /*
1563  * The evaluateloop label is where we decide whether to continue looping over
1564  * packets or call into nd code to send.
1565  */
1566 evaluateloop:
1567 
1568 	/*
1569 	 * m may be NULL when we jump to the evaluateloop label from PF or
1570 	 * other code that can drop packets.
1571 	 */
1572 	if (m != NULL) {
1573 		/*
1574 		 * If we already have a chain to send, tack m onto the end.
1575 		 * Otherwise make m the start and end of the to-be-sent chain.
1576 		 */
1577 		if (sendchain != NULL) {
1578 			sendchain_last->m_nextpkt = m;
1579 		} else {
1580 			sendchain = m;
1581 		}
1582 
1583 		/* Fragmentation may mean m is a chain. Find the last packet. */
1584 		while (m->m_nextpkt) {
1585 			m = m->m_nextpkt;
1586 		}
1587 		sendchain_last = m;
1588 		pktcnt++;
1589 	}
1590 
1591 	/* Fill in next m from inputchain as appropriate. */
1592 	m = inputchain;
1593 	if (m != NULL) {
1594 		/* Isolate m from rest of input chain. */
1595 		inputchain = m->m_nextpkt;
1596 		m->m_nextpkt = NULL;
1597 
1598 		/*
1599 		 * Clear exthdrs and ipsec_state so stale contents are not
1600 		 * reused. Note this also clears the exthdrs.merged flag.
1601 		 */
1602 		bzero(&exthdrs, sizeof(exthdrs));
1603 		bzero(&ipsec_state, sizeof(ipsec_state));
1604 
1605 		/* Continue looping. */
1606 		goto loopit;
1607 	}
1608 
1609 	/*
1610 	 * If we get here, there's no more mbufs in inputchain, so send the
1611 	 * sendchain if there is one.
1612 	 */
1613 	if (pktcnt > 0) {
1614 		error = nd6_output_list(ifp, origifp, sendchain, dst,
1615 		    ro->ro_rt, adv);
1616 		/*
1617 		 * Fall through to done label even in error case because
1618 		 * nd6_output_list frees packetchain in both success and
1619 		 * failure cases.
1620 		 */
1621 	}
1622 
1623 done:
1624 	if (ifpp_save != NULL && *ifpp_save != NULL) {
1625 		ifnet_release(*ifpp_save);
1626 		*ifpp_save = NULL;
1627 	}
1628 	ROUTE_RELEASE(&ip6route);
1629 #if IPSEC
1630 	ROUTE_RELEASE(&ipsec_state.ro);
1631 	if (sp != NULL) {
1632 		key_freesp(sp, KEY_SADB_UNLOCKED);
1633 	}
1634 #endif /* IPSEC */
1635 #if NECP
1636 	ROUTE_RELEASE(&necp_route);
1637 #endif /* NECP */
1638 #if DUMMYNET
1639 	ROUTE_RELEASE(&saved_route);
1640 	ROUTE_RELEASE(&saved_ro_pmtu);
1641 #endif /* DUMMYNET */
1642 
1643 	if (ia != NULL) {
1644 		ifa_remref(&ia->ia_ifa);
1645 	}
1646 	if (src_ia != NULL) {
1647 		ifa_remref(&src_ia->ia_ifa);
1648 	}
1649 	if (ifp != NULL) {
1650 		ifnet_release(ifp);
1651 	}
1652 	if (origifp != NULL) {
1653 		ifnet_release(origifp);
1654 	}
1655 	if (ip6_output_measure) {
1656 		net_perf_measure_time(&net_perf, &start_tv, packets_processed);
1657 		net_perf_histogram(&net_perf, packets_processed);
1658 	}
1659 	return error;
1660 
1661 freehdrs:
1662 	if (exthdrs.ip6e_hbh != NULL) {
1663 		if (exthdrs.merged) {
1664 			panic("Double free of ip6e_hbh");
1665 		}
1666 		m_freem(exthdrs.ip6e_hbh);
1667 	}
1668 	if (exthdrs.ip6e_dest1 != NULL) {
1669 		if (exthdrs.merged) {
1670 			panic("Double free of ip6e_dest1");
1671 		}
1672 		m_freem(exthdrs.ip6e_dest1);
1673 	}
1674 	if (exthdrs.ip6e_rthdr != NULL) {
1675 		if (exthdrs.merged) {
1676 			panic("Double free of ip6e_rthdr");
1677 		}
1678 		m_freem(exthdrs.ip6e_rthdr);
1679 	}
1680 	if (exthdrs.ip6e_dest2 != NULL) {
1681 		if (exthdrs.merged) {
1682 			panic("Double free of ip6e_dest2");
1683 		}
1684 		m_freem(exthdrs.ip6e_dest2);
1685 	}
1686 	/* FALLTHRU */
1687 bad:
1688 	if (inputchain != NULL) {
1689 		m_freem_list(inputchain);
1690 	}
1691 	if (sendchain != NULL) {
1692 		m_freem_list(sendchain);
1693 	}
1694 	if (m != NULL) {
1695 		m_drop(m, DROPTAP_FLAG_DIR_OUT | DROPTAP_FLAG_L2_MISSING, drop_reason, NULL, 0);
1696 	}
1697 
1698 	goto done;
1699 
1700 #undef ipf_pktopts
1701 #undef exthdrs
1702 #undef ip6route
1703 #undef ipsec_state
1704 #undef saved_route
1705 #undef saved_ro_pmtu
1706 #undef args
1707 }
1708 
1709 /* ip6_fragment_packet
1710  *
1711  * The fragmentation logic is rather complex:
1712  * 1: normal case (dontfrag == 0)
1713  * 1-a:	send as is if tlen <= path mtu
1714  * 1-b:	fragment if tlen > path mtu
1715  *
1716  * 2: if user asks us not to fragment (dontfrag == 1)
1717  * 2-a:	send as is if tlen <= interface mtu
1718  * 2-b:	error if tlen > interface mtu
1719  */
1720 
1721 static int
ip6_fragment_packet(struct mbuf ** mptr,struct ip6_pktopts * opt,struct ip6_out_args * ip6oa,struct ip6_exthdrs * exthdrsp,struct ifnet * ifp,uint32_t mtu,uint32_t unfragpartlen,int nxt0,uint32_t optlen)1722 ip6_fragment_packet(struct mbuf **mptr, struct ip6_pktopts *opt,
1723     struct ip6_out_args *ip6oa, struct ip6_exthdrs *exthdrsp,
1724     struct ifnet *ifp, uint32_t mtu, uint32_t unfragpartlen,
1725     int nxt0, uint32_t optlen)
1726 {
1727 	VERIFY(NULL != mptr);
1728 	mbuf_ref_t m = *mptr;
1729 	int error = 0;
1730 	uint32_t tlen = m->m_pkthdr.len;
1731 	boolean_t dontfrag = (opt != NULL && (opt->ip6po_flags & IP6PO_DONTFRAG)) ||
1732 	    (ip6oa != NULL && (ip6oa->ip6oa_flags & IP6OAF_DONT_FRAG));
1733 
1734 	if (m->m_pkthdr.pkt_flags & PKTF_FORWARDED) {
1735 		dontfrag = TRUE;
1736 		/*
1737 		 * Discard partial sum information if this packet originated
1738 		 * from another interface; the packet would already have the
1739 		 * final checksum and we shouldn't recompute it.
1740 		 */
1741 		if ((m->m_pkthdr.csum_flags & (CSUM_DATA_VALID | CSUM_PARTIAL)) ==
1742 		    (CSUM_DATA_VALID | CSUM_PARTIAL)) {
1743 			m->m_pkthdr.csum_flags &= ~CSUM_TX_FLAGS;
1744 			m->m_pkthdr.csum_data = 0;
1745 		}
1746 	}
1747 
1748 	/* Access without acquiring nd_ifinfo lock for performance */
1749 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
1750 		/*
1751 		 * We do not notify the connection in the same outbound path
1752 		 * to avoid lock ordering issues.
1753 		 * The returned error should imply that the packet is too big
1754 		 * and the application should query the PMTU for a given destination.
1755 		 */
1756 		return EMSGSIZE;
1757 	}
1758 
1759 	/*
1760 	 * transmit packet without fragmentation
1761 	 */
1762 	if (dontfrag ||
1763 	    (tlen <= mtu || TSO_IPV6_OK(ifp, m) ||
1764 	    (ifp->if_hwassist & CSUM_FRAGMENT_IPV6))) {
1765 		/*
1766 		 * mppn not updated in this case because no new chain is formed
1767 		 * and inserted
1768 		 */
1769 		ip6_output_checksum(ifp, mtu, m, nxt0, tlen, optlen);
1770 	} else {
1771 		/*
1772 		 * time to fragment - cases 1-b is handled inside
1773 		 * ip6_do_fragmentation().
1774 		 * mppn is passed down to be updated to point at fragment chain.
1775 		 */
1776 		u_int8_t *lexthdrsp;
1777 
1778 		if (exthdrsp->ip6e_rthdr != NULL) {
1779 			lexthdrsp = mtod(exthdrsp->ip6e_rthdr, uint8_t *);
1780 		} else if (exthdrsp->ip6e_dest1 != NULL) {
1781 			lexthdrsp = mtod(exthdrsp->ip6e_dest1, uint8_t *);
1782 		} else if (exthdrsp->ip6e_hbh != NULL) {
1783 			lexthdrsp = mtod(exthdrsp->ip6e_hbh, uint8_t *);
1784 		} else {
1785 			lexthdrsp = NULL;
1786 		}
1787 		error = ip6_do_fragmentation(mptr, optlen, ifp,
1788 		    unfragpartlen, mtod(m, struct ip6_hdr *), lexthdrsp, mtu,
1789 		    nxt0, htonl(ip6_randomid((uint64_t)m)));
1790 	}
1791 
1792 	return error;
1793 }
1794 
1795 /*
1796  * ip6_do_fragmentation() is called by ip6_fragment_packet() after determining
1797  * the packet needs to be fragmented. on success, morig is freed and a chain
1798  * of fragments is linked into the packet chain where morig existed. Otherwise,
1799  * an errno is returned.
1800  * optlen:        total length of all extension headers (excludes the IPv6 header).
1801  * unfragpartlen: length of the per-fragment headers which consist of the IPv6
1802  *                header plus any extension headers that must be processed by nodes
1803  *                en route to the destination.
1804  * lexthdrsp:     pointer to the last extension header in the unfragmentable part
1805  *                or NULL.
1806  * nxt0:          upper-layer protocol number.
1807  * id:            Identification value to be used in the fragment header.
1808  */
1809 int
ip6_do_fragmentation(struct mbuf ** mptr,uint32_t optlen,struct ifnet * ifp,uint32_t unfragpartlen,struct ip6_hdr * ip6,uint8_t * lexthdrsp,uint32_t mtu,int nxt0,uint32_t id)1810 ip6_do_fragmentation(struct mbuf **mptr, uint32_t optlen, struct ifnet *ifp,
1811     uint32_t unfragpartlen, struct ip6_hdr *ip6, uint8_t *lexthdrsp,
1812     uint32_t mtu, int nxt0, uint32_t id)
1813 {
1814 	VERIFY(NULL != mptr);
1815 	int error = 0;
1816 
1817 	mbuf_ref_t morig = *mptr;
1818 	mbuf_ref_t first_mbufp = NULL;
1819 	mbuf_ref_t last_mbufp = NULL;
1820 
1821 	uint32_t tlen = morig->m_pkthdr.len;
1822 
1823 	/* try to fragment the packet. case 1-b */
1824 	if ((morig->m_pkthdr.csum_flags & CSUM_TSO_IPV6)) {
1825 		/* TSO and fragment aren't compatible */
1826 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1827 		return EMSGSIZE;
1828 	} else if (mtu < IPV6_MMTU) {
1829 		/* path MTU cannot be less than IPV6_MMTU */
1830 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1831 		return EMSGSIZE;
1832 	} else if (ip6->ip6_plen == 0) {
1833 		/* jumbo payload cannot be fragmented */
1834 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1835 		return EMSGSIZE;
1836 	} else {
1837 		uint32_t hlen, off, len;
1838 		mbuf_ref_t *mnext = NULL;
1839 		struct ip6_frag *__single ip6f;
1840 		u_char nextproto;
1841 
1842 		/*
1843 		 * Too large for the destination or interface;
1844 		 * fragment if possible.
1845 		 * Must be able to put at least 8 bytes per fragment.
1846 		 */
1847 		hlen = unfragpartlen;
1848 		if (mtu > IPV6_MAXPACKET) {
1849 			mtu = IPV6_MAXPACKET;
1850 		}
1851 
1852 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1853 		if (len < 8) {
1854 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1855 			return EMSGSIZE;
1856 		}
1857 
1858 		/*
1859 		 * Change the next header field of the last header in the
1860 		 * unfragmentable part.
1861 		 */
1862 		if (lexthdrsp != NULL) {
1863 			nextproto = *lexthdrsp;
1864 			*lexthdrsp = IPPROTO_FRAGMENT;
1865 		} else {
1866 			nextproto = ip6->ip6_nxt;
1867 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1868 		}
1869 
1870 		if (morig->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) {
1871 			in6_delayed_cksum_offset(morig, 0, optlen, nxt0);
1872 		}
1873 
1874 		/*
1875 		 * Loop through length of segment after first fragment,
1876 		 * make new header and copy data of each part and link onto
1877 		 * chain.
1878 		 */
1879 		for (off = hlen; off < tlen; off += len) {
1880 			struct ip6_hdr *__single new_mhip6;
1881 			mbuf_ref_t new_m;
1882 			mbuf_ref_t m_frgpart;
1883 
1884 			MGETHDR(new_m, M_DONTWAIT, MT_HEADER);  /* MAC-OK */
1885 			if (new_m == NULL) {
1886 				error = ENOBUFS;
1887 				ip6stat.ip6s_odropped++;
1888 				break;
1889 			}
1890 			new_m->m_pkthdr.rcvif = NULL;
1891 			new_m->m_flags = morig->m_flags & M_COPYFLAGS;
1892 
1893 			if (first_mbufp != NULL) {
1894 				/* Every pass through loop but first */
1895 				*mnext = new_m;
1896 				last_mbufp = new_m;
1897 			} else {
1898 				/* This is the first element of the fragment chain */
1899 				first_mbufp = new_m;
1900 				last_mbufp = new_m;
1901 			}
1902 			mnext = &new_m->m_nextpkt;
1903 
1904 			new_m->m_data += max_linkhdr;
1905 			new_mhip6 = mtod(new_m, struct ip6_hdr *);
1906 			*new_mhip6 = *ip6;
1907 			new_m->m_len = sizeof(*new_mhip6);
1908 
1909 			error = ip6_insertfraghdr(morig, new_m, hlen, &ip6f);
1910 			if (error) {
1911 				ip6stat.ip6s_odropped++;
1912 				break;
1913 			}
1914 
1915 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1916 			if (off + len >= tlen) {
1917 				len = tlen - off;
1918 			} else {
1919 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1920 			}
1921 			new_mhip6->ip6_plen = htons((u_short)(len + hlen +
1922 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1923 
1924 			if ((m_frgpart = m_copy(morig, off, len)) == NULL) {
1925 				error = ENOBUFS;
1926 				ip6stat.ip6s_odropped++;
1927 				break;
1928 			}
1929 			m_cat(new_m, m_frgpart);
1930 			new_m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1931 			new_m->m_pkthdr.rcvif = NULL;
1932 
1933 			M_COPY_CLASSIFIER(new_m, morig);
1934 			M_COPY_PFTAG(new_m, morig);
1935 			M_COPY_NECPTAG(new_m, morig);
1936 
1937 			ip6f->ip6f_reserved = 0;
1938 			ip6f->ip6f_ident = id;
1939 			ip6f->ip6f_nxt = nextproto;
1940 			ip6stat.ip6s_ofragments++;
1941 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1942 		}
1943 
1944 		if (error) {
1945 			/* free all the fragments created */
1946 			if (first_mbufp != NULL) {
1947 				m_freem_list(first_mbufp);
1948 				first_mbufp = NULL;
1949 			}
1950 			last_mbufp = NULL;
1951 		} else {
1952 			/* successful fragmenting */
1953 			m_freem(morig);
1954 			*mptr = first_mbufp;
1955 			last_mbufp->m_nextpkt = NULL;
1956 			ip6stat.ip6s_fragmented++;
1957 			in6_ifstat_inc(ifp, ifs6_out_fragok);
1958 		}
1959 	}
1960 	return error;
1961 }
1962 
1963 static int
ip6_copyexthdr(struct mbuf ** mp,caddr_t __indexable hdr,int hlen)1964 ip6_copyexthdr(struct mbuf **mp, caddr_t __indexable hdr, int hlen)
1965 {
1966 	mbuf_ref_t m;
1967 
1968 	if (hlen > MCLBYTES) {
1969 		return ENOBUFS; /* XXX */
1970 	}
1971 	MGET(m, M_DONTWAIT, MT_DATA);
1972 	if (m == NULL) {
1973 		return ENOBUFS;
1974 	}
1975 
1976 	if (hlen > MLEN) {
1977 		MCLGET(m, M_DONTWAIT);
1978 		if (!(m->m_flags & M_EXT)) {
1979 			m_free(m);
1980 			return ENOBUFS;
1981 		}
1982 	}
1983 	m->m_len = hlen;
1984 	if (hdr != NULL) {
1985 		bcopy(hdr, mtod(m, caddr_t), hlen);
1986 	}
1987 
1988 	*mp = m;
1989 	return 0;
1990 }
1991 
1992 static void
ip6_out_cksum_stats(int proto,u_int32_t len)1993 ip6_out_cksum_stats(int proto, u_int32_t len)
1994 {
1995 	switch (proto) {
1996 	case IPPROTO_TCP:
1997 		tcp_out6_cksum_stats(len);
1998 		break;
1999 	case IPPROTO_UDP:
2000 		udp_out6_cksum_stats(len);
2001 		break;
2002 	default:
2003 		/* keep only TCP or UDP stats for now */
2004 		break;
2005 	}
2006 }
2007 
2008 /*
2009  * Process a delayed payload checksum calculation (outbound path.)
2010  *
2011  * hoff is the number of bytes beyond the mbuf data pointer which
2012  * points to the IPv6 header.  optlen is the number of bytes, if any,
2013  * between the end of IPv6 header and the beginning of the ULP payload
2014  * header, which represents the extension headers.  If optlen is less
2015  * than zero, this routine will bail when it detects extension headers.
2016  *
2017  * Returns a bitmask representing all the work done in software.
2018  */
2019 uint32_t
in6_finalize_cksum(struct mbuf * m,uint32_t hoff,int32_t optlen,int32_t nxt0,uint32_t csum_flags)2020 in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen,
2021     int32_t nxt0, uint32_t csum_flags)
2022 {
2023 	unsigned char buf[sizeof(struct ip6_hdr)] __attribute__((aligned(8)));
2024 	struct ip6_hdr *__single ip6;
2025 	uint32_t offset, mlen, hlen, olen, sw_csum;
2026 	uint16_t csum, ulpoff, plen;
2027 	uint8_t nxt;
2028 
2029 	_CASSERT(sizeof(csum) == sizeof(uint16_t));
2030 	VERIFY(m->m_flags & M_PKTHDR);
2031 
2032 	sw_csum = (csum_flags & m->m_pkthdr.csum_flags);
2033 
2034 	if ((sw_csum &= CSUM_DELAY_IPV6_DATA) == 0) {
2035 		goto done;
2036 	}
2037 
2038 	mlen = m->m_pkthdr.len;                         /* total mbuf len */
2039 	hlen = sizeof(*ip6);                            /* IPv6 header len */
2040 
2041 	/* sanity check (need at least IPv6 header) */
2042 	if (mlen < (hoff + hlen)) {
2043 		panic("%s: mbuf %p pkt len (%u) < hoff+ip6_hdr "
2044 		    "(%u+%u)\n", __func__, m, mlen, hoff, hlen);
2045 		/* NOTREACHED */
2046 	}
2047 
2048 	/*
2049 	 * In case the IPv6 header is not contiguous, or not 32-bit
2050 	 * aligned, copy it to a local buffer.
2051 	 */
2052 	if ((hoff + hlen) > m->m_len ||
2053 	    !IP6_HDR_ALIGNED_P(mtod(m, caddr_t) + hoff)) {
2054 		m_copydata(m, hoff, hlen, (caddr_t)buf);
2055 		ip6 = (struct ip6_hdr *)(void *)buf;
2056 	} else {
2057 		ip6 = (struct ip6_hdr *)(m_mtod_current(m) + hoff);
2058 	}
2059 
2060 	nxt = ip6->ip6_nxt;
2061 	plen = ntohs(ip6->ip6_plen);
2062 	if (plen != (mlen - (hoff + hlen))) {
2063 		plen = OSSwapInt16(plen);
2064 		if (plen != (mlen - (hoff + hlen))) {
2065 			/* Don't complain for jumbograms */
2066 			if (plen != 0 || nxt != IPPROTO_HOPOPTS) {
2067 				printf("%s: mbuf 0x%llx proto %d IPv6 "
2068 				    "plen %d (%x) [swapped %d (%x)] doesn't "
2069 				    "match actual packet length; %d is used "
2070 				    "instead\n", __func__,
2071 				    (uint64_t)VM_KERNEL_ADDRPERM(m), nxt,
2072 				    ip6->ip6_plen, ip6->ip6_plen, plen, plen,
2073 				    (mlen - (hoff + hlen)));
2074 			}
2075 			plen = (uint16_t)(mlen - (hoff + hlen));
2076 		}
2077 	}
2078 
2079 	if (optlen < 0) {
2080 		/* next header isn't TCP/UDP and we don't know optlen, bail */
2081 		if (nxt != IPPROTO_TCP && nxt != IPPROTO_UDP) {
2082 			sw_csum = 0;
2083 			goto done;
2084 		}
2085 		olen = 0;
2086 	} else {
2087 		/* caller supplied the original transport number; use it */
2088 		if (nxt0 >= 0) {
2089 			nxt = (uint8_t)nxt0;
2090 		}
2091 		olen = optlen;
2092 	}
2093 
2094 	offset = hoff + hlen + olen;                    /* ULP header */
2095 
2096 	/* sanity check */
2097 	if (mlen < offset) {
2098 		panic("%s: mbuf %p pkt len (%u) < hoff+ip6_hdr+ext_hdr "
2099 		    "(%u+%u+%u)\n", __func__, m, mlen, hoff, hlen, olen);
2100 		/* NOTREACHED */
2101 	}
2102 
2103 	/*
2104 	 * offset is added to the lower 16-bit value of csum_data,
2105 	 * which is expected to contain the ULP offset; therefore
2106 	 * CSUM_PARTIAL offset adjustment must be undone.
2107 	 */
2108 	if ((m->m_pkthdr.csum_flags & (CSUM_PARTIAL | CSUM_DATA_VALID)) ==
2109 	    (CSUM_PARTIAL | CSUM_DATA_VALID)) {
2110 		/*
2111 		 * Get back the original ULP offset (this will
2112 		 * undo the CSUM_PARTIAL logic in ip6_output.)
2113 		 */
2114 		m->m_pkthdr.csum_data = (m->m_pkthdr.csum_tx_stuff -
2115 		    m->m_pkthdr.csum_tx_start);
2116 	}
2117 
2118 	ulpoff = (m->m_pkthdr.csum_data & 0xffff);      /* ULP csum offset */
2119 
2120 	if (mlen < (ulpoff + sizeof(csum))) {
2121 		panic("%s: mbuf %p pkt len (%u) proto %d invalid ULP "
2122 		    "cksum offset (%u) cksum flags 0x%x\n", __func__,
2123 		    m, mlen, nxt, ulpoff, m->m_pkthdr.csum_flags);
2124 		/* NOTREACHED */
2125 	}
2126 
2127 	csum = inet6_cksum(m, 0, offset, plen - olen);
2128 
2129 	/* Update stats */
2130 	ip6_out_cksum_stats(nxt, plen - olen);
2131 
2132 	/* RFC1122 4.1.3.4 */
2133 	if (csum == 0 &&
2134 	    (m->m_pkthdr.csum_flags & (CSUM_UDPIPV6 | CSUM_ZERO_INVERT))) {
2135 		csum = 0xffff;
2136 	}
2137 
2138 	/* Insert the checksum in the ULP csum field */
2139 	offset += ulpoff;
2140 	if ((offset + sizeof(csum)) > m->m_len) {
2141 		m_copyback(m, offset, sizeof(csum), &csum);
2142 	} else if (IP6_HDR_ALIGNED_P(mtod(m, char *) + hoff)) {
2143 		*(uint16_t *)(void *)(mtod(m, char *) + offset) = csum;
2144 	} else {
2145 		bcopy(&csum, (mtod(m, char *) + offset), sizeof(csum));
2146 	}
2147 	m->m_pkthdr.csum_flags &= ~(CSUM_DELAY_IPV6_DATA | CSUM_DATA_VALID |
2148 	    CSUM_PARTIAL | CSUM_ZERO_INVERT);
2149 
2150 done:
2151 	return sw_csum;
2152 }
2153 
2154 /*
2155  * Insert jumbo payload option.
2156  */
2157 static int
ip6_insert_jumboopt(struct ip6_exthdrs * exthdrs,u_int32_t plen)2158 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
2159 {
2160 	mbuf_ref_t mopt;
2161 	u_char *optbuf;
2162 	u_int32_t v;
2163 
2164 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
2165 
2166 	/*
2167 	 * If there is no hop-by-hop options header, allocate new one.
2168 	 * If there is one but it doesn't have enough space to store the
2169 	 * jumbo payload option, allocate a cluster to store the whole options.
2170 	 * Otherwise, use it to store the options.
2171 	 */
2172 	if (exthdrs->ip6e_hbh == NULL) {
2173 		MGET(mopt, M_DONTWAIT, MT_DATA);
2174 		if (mopt == NULL) {
2175 			return ENOBUFS;
2176 		}
2177 		mopt->m_len = JUMBOOPTLEN;
2178 		optbuf = mtod(mopt, u_char *);
2179 		optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
2180 		exthdrs->ip6e_hbh = mopt;
2181 	} else {
2182 		struct ip6_hbh *hbh;
2183 
2184 		mopt = exthdrs->ip6e_hbh;
2185 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
2186 			/*
2187 			 * XXX assumption:
2188 			 * - exthdrs->ip6e_hbh is not referenced from places
2189 			 *   other than exthdrs.
2190 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
2191 			 */
2192 			u_int32_t oldoptlen = mopt->m_len;
2193 			mbuf_ref_t n;
2194 
2195 			/*
2196 			 * XXX: give up if the whole (new) hbh header does
2197 			 * not fit even in an mbuf cluster.
2198 			 */
2199 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES) {
2200 				return ENOBUFS;
2201 			}
2202 
2203 			/*
2204 			 * As a consequence, we must always prepare a cluster
2205 			 * at this point.
2206 			 */
2207 			MGET(n, M_DONTWAIT, MT_DATA);
2208 			if (n != NULL) {
2209 				MCLGET(n, M_DONTWAIT);
2210 				if (!(n->m_flags & M_EXT)) {
2211 					m_freem(n);
2212 					n = NULL;
2213 				}
2214 			}
2215 			if (n == NULL) {
2216 				return ENOBUFS;
2217 			}
2218 			n->m_len = oldoptlen + JUMBOOPTLEN;
2219 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
2220 			    oldoptlen);
2221 			optbuf = mtod(n, u_char *) + oldoptlen;
2222 			m_freem(mopt);
2223 			mopt = exthdrs->ip6e_hbh = n;
2224 		} else {
2225 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
2226 			mopt->m_len += JUMBOOPTLEN;
2227 		}
2228 		optbuf[0] = IP6OPT_PADN;
2229 		optbuf[1] = 1;
2230 
2231 		/*
2232 		 * Adjust the header length according to the pad and
2233 		 * the jumbo payload option.
2234 		 */
2235 		hbh = mtod(mopt, struct ip6_hbh *);
2236 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
2237 	}
2238 
2239 	/* fill in the option. */
2240 	optbuf[2] = IP6OPT_JUMBO;
2241 	optbuf[3] = 4;
2242 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
2243 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
2244 
2245 	/* finally, adjust the packet header length */
2246 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
2247 
2248 	return 0;
2249 #undef JUMBOOPTLEN
2250 }
2251 
2252 /*
2253  * Insert fragment header and copy unfragmentable header portions.
2254  */
2255 static int
ip6_insertfraghdr(struct mbuf * m0,struct mbuf * m,int hlen,struct ip6_frag ** frghdrp)2256 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
2257     struct ip6_frag **frghdrp)
2258 {
2259 	mbuf_ref_t n, mlast;
2260 
2261 	if (hlen > sizeof(struct ip6_hdr)) {
2262 		n = m_copym(m0, sizeof(struct ip6_hdr),
2263 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
2264 		if (n == NULL) {
2265 			return ENOBUFS;
2266 		}
2267 		m->m_next = n;
2268 	} else {
2269 		n = m;
2270 	}
2271 
2272 	/* Search for the last mbuf of unfragmentable part. */
2273 	for (mlast = n; mlast->m_next; mlast = mlast->m_next) {
2274 		;
2275 	}
2276 
2277 	if (!(mlast->m_flags & M_EXT) &&
2278 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
2279 		/* use the trailing space of the last mbuf for the frag hdr */
2280 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
2281 		    mlast->m_len);
2282 		mlast->m_len += sizeof(struct ip6_frag);
2283 		m->m_pkthdr.len += sizeof(struct ip6_frag);
2284 	} else {
2285 		/* allocate a new mbuf for the fragment header */
2286 		mbuf_ref_t mfrg;
2287 
2288 		MGET(mfrg, M_DONTWAIT, MT_DATA);
2289 		if (mfrg == NULL) {
2290 			return ENOBUFS;
2291 		}
2292 		mfrg->m_len = sizeof(struct ip6_frag);
2293 		*frghdrp = mtod(mfrg, struct ip6_frag *);
2294 		mlast->m_next = mfrg;
2295 	}
2296 
2297 	return 0;
2298 }
2299 
2300 static int
ip6_getpmtu(struct route_in6 * ro_pmtu,struct route_in6 * ro,struct ifnet * ifp,struct in6_addr * dst,uint32_t dst_ifscope,u_int32_t * mtup)2301 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
2302     struct ifnet *ifp, struct in6_addr *dst, uint32_t dst_ifscope, u_int32_t *mtup)
2303 {
2304 	u_int32_t mtu = 0;
2305 	int error = 0;
2306 
2307 	if (ro_pmtu != ro) {
2308 		/* The first hop and the final destination may differ. */
2309 		struct sockaddr_in6 *__single sa6_dst = SIN6(&ro_pmtu->ro_dst);
2310 		if (ROUTE_UNUSABLE(ro_pmtu) ||
2311 		    !in6_are_addr_equal_scoped(&sa6_dst->sin6_addr, dst, sa6_dst->sin6_scope_id, dst_ifscope)) {
2312 			ROUTE_RELEASE(ro_pmtu);
2313 		}
2314 
2315 		if (ro_pmtu->ro_rt == NULL) {
2316 			SOCKADDR_ZERO(sa6_dst, sizeof(*sa6_dst));
2317 			sa6_dst->sin6_family = AF_INET6;
2318 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
2319 			sa6_dst->sin6_addr = *dst;
2320 
2321 			rtalloc_scoped((struct route *)ro_pmtu,
2322 			    ifp != NULL ? ifp->if_index : IFSCOPE_NONE);
2323 		}
2324 	}
2325 
2326 	if (ro_pmtu->ro_rt != NULL) {
2327 		u_int32_t ifmtu;
2328 
2329 		if (ifp == NULL) {
2330 			ifp = ro_pmtu->ro_rt->rt_ifp;
2331 		}
2332 		/* Access without acquiring nd_ifinfo lock for performance */
2333 		ifmtu = IN6_LINKMTU(ifp);
2334 
2335 		/*
2336 		 * Access rmx_mtu without holding the route entry lock,
2337 		 * for performance; this isn't something that changes
2338 		 * often, so optimize.
2339 		 */
2340 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
2341 		if (mtu > ifmtu || mtu == 0) {
2342 			/*
2343 			 * The MTU on the route is larger than the MTU on
2344 			 * the interface!  This shouldn't happen, unless the
2345 			 * MTU of the interface has been changed after the
2346 			 * interface was brought up.  Change the MTU in the
2347 			 * route to match the interface MTU (as long as the
2348 			 * field isn't locked).
2349 			 *
2350 			 * if MTU on the route is 0, we need to fix the MTU.
2351 			 * this case happens with path MTU discovery timeouts.
2352 			 */
2353 			mtu = ifmtu;
2354 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) {
2355 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
2356 			}
2357 		}
2358 	} else {
2359 		if (ifp) {
2360 			/* Don't hold nd_ifinfo lock for performance */
2361 			mtu = IN6_LINKMTU(ifp);
2362 		} else {
2363 			error = EHOSTUNREACH; /* XXX */
2364 		}
2365 	}
2366 
2367 	*mtup = mtu;
2368 	return error;
2369 }
2370 
2371 /*
2372  * IP6 socket option processing.
2373  */
2374 int
ip6_ctloutput(struct socket * so,struct sockopt * sopt)2375 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
2376 {
2377 	int optdatalen, uproto;
2378 	void *optdata;
2379 	int privileged;
2380 	struct inpcb *__single in6p = sotoinpcb(so);
2381 	int error = 0, optval = 0;
2382 	int level, op = -1, optname = 0;
2383 	size_t optlen = 0;
2384 	struct proc *__single p;
2385 
2386 	VERIFY(sopt != NULL);
2387 
2388 	level = sopt->sopt_level;
2389 	op = sopt->sopt_dir;
2390 	optname = sopt->sopt_name;
2391 	optlen = sopt->sopt_valsize;
2392 	p = sopt->sopt_p;
2393 	uproto = (int)SOCK_PROTO(so);
2394 
2395 	privileged = (proc_suser(p) == 0);
2396 
2397 	if (level == IPPROTO_IPV6) {
2398 		boolean_t capture_exthdrstat_in = FALSE;
2399 		switch (op) {
2400 		case SOPT_SET:
2401 			/*
2402 			 * Wait if we are in the middle of ip6_output
2403 			 * as we unlocked the socket there and don't
2404 			 * want to overwrite the IP options
2405 			 */
2406 			if (in6p->inp_sndinprog_cnt > 0) {
2407 				lck_mtx_t *__single mutex_held = NULL;
2408 
2409 				mutex_held = socket_getlock(so, PR_F_WILLUNLOCK);
2410 				in6p->inp_sndingprog_waiters++;
2411 
2412 				while (in6p->inp_sndinprog_cnt > 0) {
2413 					msleep(&in6p->inp_sndinprog_cnt, mutex_held,
2414 					    PSOCK | PCATCH, "inp_sndinprog_cnt",
2415 					    NULL);
2416 				}
2417 				in6p->inp_sndingprog_waiters--;
2418 			}
2419 			switch (optname) {
2420 			case IPV6_2292PKTOPTIONS: {
2421 				mbuf_ref_t m;
2422 
2423 				error = soopt_getm(sopt, &m);
2424 				if (error != 0) {
2425 					break;
2426 				}
2427 				error = soopt_mcopyin(sopt, m);
2428 				if (error != 0) {
2429 					break;
2430 				}
2431 				error = ip6_pcbopts(&in6p->in6p_outputopts,
2432 				    m, so, sopt);
2433 				m_freem(m);
2434 				break;
2435 			}
2436 
2437 			/*
2438 			 * Use of some Hop-by-Hop options or some
2439 			 * Destination options, might require special
2440 			 * privilege.  That is, normal applications
2441 			 * (without special privilege) might be forbidden
2442 			 * from setting certain options in outgoing packets,
2443 			 * and might never see certain options in received
2444 			 * packets. [RFC 2292 Section 6]
2445 			 * KAME specific note:
2446 			 *  KAME prevents non-privileged users from sending or
2447 			 *  receiving ANY hbh/dst options in order to avoid
2448 			 *  overhead of parsing options in the kernel.
2449 			 */
2450 			case IPV6_RECVHOPOPTS:
2451 			case IPV6_RECVDSTOPTS:
2452 			case IPV6_RECVRTHDRDSTOPTS:
2453 				if (!privileged) {
2454 					break;
2455 				}
2456 				OS_FALLTHROUGH;
2457 			case IPV6_UNICAST_HOPS:
2458 			case IPV6_HOPLIMIT:
2459 			case IPV6_RECVPKTINFO:
2460 			case IPV6_RECVHOPLIMIT:
2461 			case IPV6_RECVRTHDR:
2462 			case IPV6_RECVPATHMTU:
2463 			case IPV6_RECVTCLASS:
2464 			case IPV6_V6ONLY:
2465 			case IPV6_AUTOFLOWLABEL:
2466 				if (optlen != sizeof(int)) {
2467 					error = EINVAL;
2468 					break;
2469 				}
2470 				error = sooptcopyin(sopt, &optval,
2471 				    sizeof(optval), sizeof(optval));
2472 				if (error) {
2473 					break;
2474 				}
2475 
2476 				switch (optname) {
2477 				case IPV6_UNICAST_HOPS:
2478 					if (optval < -1 || optval >= 256) {
2479 						error = EINVAL;
2480 					} else {
2481 						/* -1 = kernel default */
2482 						in6p->in6p_hops = (short)optval;
2483 						if (in6p->inp_vflag &
2484 						    INP_IPV4) {
2485 							in6p->inp_ip_ttl =
2486 							    (uint8_t)optval;
2487 						}
2488 					}
2489 					break;
2490 #define OPTSET(bit) do {                                                \
2491 	if (optval)                                                     \
2492 	        in6p->inp_flags |= (bit);                               \
2493 	else                                                            \
2494 	        in6p->inp_flags &= ~(bit);                              \
2495 } while (0)
2496 
2497 #define OPTSET2(bit) do {                                           \
2498 	if (optval) {                                                   \
2499 	        in6p->inp_flags2 |= (bit);                              \
2500 	} else {                                                        \
2501 	        in6p->inp_flags2 &= ~(bit);                             \
2502 	}                                                               \
2503 } while (0)
2504 
2505 #define OPTSET2292(bit) do {                                            \
2506 	in6p->inp_flags |= IN6P_RFC2292;                                \
2507 	if (optval)                                                     \
2508 	        in6p->inp_flags |= (bit);                               \
2509 	else                                                            \
2510 	        in6p->inp_flags &= ~(bit);                              \
2511 } while (0)
2512 
2513 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
2514 
2515 				case IPV6_RECVPKTINFO:
2516 					/* cannot mix with RFC2292 */
2517 					if (OPTBIT(IN6P_RFC2292)) {
2518 						error = EINVAL;
2519 						break;
2520 					}
2521 					OPTSET(IN6P_PKTINFO);
2522 					break;
2523 
2524 				case IPV6_HOPLIMIT: {
2525 					struct ip6_pktopts **optp;
2526 
2527 					/* cannot mix with RFC2292 */
2528 					if (OPTBIT(IN6P_RFC2292)) {
2529 						error = EINVAL;
2530 						break;
2531 					}
2532 					optp = &in6p->in6p_outputopts;
2533 					error = ip6_pcbopt(IPV6_HOPLIMIT,
2534 					    (u_char *)&optval, sizeof(optval),
2535 					    optp, uproto);
2536 					break;
2537 				}
2538 
2539 				case IPV6_RECVHOPLIMIT:
2540 					/* cannot mix with RFC2292 */
2541 					if (OPTBIT(IN6P_RFC2292)) {
2542 						error = EINVAL;
2543 						break;
2544 					}
2545 					OPTSET(IN6P_HOPLIMIT);
2546 					break;
2547 
2548 				case IPV6_RECVHOPOPTS:
2549 					/* cannot mix with RFC2292 */
2550 					if (OPTBIT(IN6P_RFC2292)) {
2551 						error = EINVAL;
2552 						break;
2553 					}
2554 					OPTSET(IN6P_HOPOPTS);
2555 					capture_exthdrstat_in = TRUE;
2556 					break;
2557 
2558 				case IPV6_RECVDSTOPTS:
2559 					/* cannot mix with RFC2292 */
2560 					if (OPTBIT(IN6P_RFC2292)) {
2561 						error = EINVAL;
2562 						break;
2563 					}
2564 					OPTSET(IN6P_DSTOPTS);
2565 					capture_exthdrstat_in = TRUE;
2566 					break;
2567 
2568 				case IPV6_RECVRTHDRDSTOPTS:
2569 					/* cannot mix with RFC2292 */
2570 					if (OPTBIT(IN6P_RFC2292)) {
2571 						error = EINVAL;
2572 						break;
2573 					}
2574 					OPTSET(IN6P_RTHDRDSTOPTS);
2575 					capture_exthdrstat_in = TRUE;
2576 					break;
2577 
2578 				case IPV6_RECVRTHDR:
2579 					/* cannot mix with RFC2292 */
2580 					if (OPTBIT(IN6P_RFC2292)) {
2581 						error = EINVAL;
2582 						break;
2583 					}
2584 					OPTSET(IN6P_RTHDR);
2585 					capture_exthdrstat_in = TRUE;
2586 					break;
2587 
2588 				case IPV6_RECVPATHMTU:
2589 					/*
2590 					 * We ignore this option for TCP
2591 					 * sockets.
2592 					 * (RFC3542 leaves this case
2593 					 * unspecified.)
2594 					 */
2595 					if (uproto != IPPROTO_TCP) {
2596 						OPTSET(IN6P_MTU);
2597 					}
2598 					break;
2599 
2600 				case IPV6_V6ONLY:
2601 					/*
2602 					 * make setsockopt(IPV6_V6ONLY)
2603 					 * available only prior to bind(2).
2604 					 * see ipng mailing list, Jun 22 2001.
2605 					 */
2606 					if (in6p->inp_lport ||
2607 					    !IN6_IS_ADDR_UNSPECIFIED(
2608 						    &in6p->in6p_laddr)) {
2609 						error = EINVAL;
2610 						break;
2611 					}
2612 					OPTSET(IN6P_IPV6_V6ONLY);
2613 					if (optval) {
2614 						in6p->inp_vflag &= ~INP_IPV4;
2615 					} else {
2616 						in6p->inp_vflag |= INP_IPV4;
2617 					}
2618 					break;
2619 
2620 				case IPV6_RECVTCLASS:
2621 					/* we can mix with RFC2292 */
2622 					OPTSET(IN6P_TCLASS);
2623 					break;
2624 
2625 				case IPV6_AUTOFLOWLABEL:
2626 					OPTSET(IN6P_AUTOFLOWLABEL);
2627 					break;
2628 				}
2629 				break;
2630 
2631 			case IPV6_TCLASS:
2632 			case IPV6_DONTFRAG:
2633 			case IPV6_USE_MIN_MTU:
2634 			case IPV6_PREFER_TEMPADDR: {
2635 				struct ip6_pktopts **optp;
2636 
2637 				if (optlen != sizeof(optval)) {
2638 					error = EINVAL;
2639 					break;
2640 				}
2641 				error = sooptcopyin(sopt, &optval,
2642 				    sizeof(optval), sizeof(optval));
2643 				if (error) {
2644 					break;
2645 				}
2646 
2647 				optp = &in6p->in6p_outputopts;
2648 				error = ip6_pcbopt(optname, (u_char *)&optval,
2649 				    sizeof(optval), optp, uproto);
2650 				switch (optname) {
2651 				case  IPV6_TCLASS: {
2652 					// Add in the ECN flags
2653 					u_int8_t tos = (in6p->inp_ip_tos & ~IPTOS_ECN_MASK);
2654 					u_int8_t ecn = optval & IPTOS_ECN_MASK;
2655 					in6p->inp_ip_tos = tos | ecn;
2656 				}
2657 				break;
2658 				case IPV6_DONTFRAG:
2659 					if (in6p->inp_vflag & INP_IPV4) {
2660 						OPTSET2(INP2_DONTFRAG);
2661 					}
2662 					break;
2663 				}
2664 				break;
2665 			}
2666 
2667 			case IPV6_2292PKTINFO:
2668 			case IPV6_2292HOPLIMIT:
2669 			case IPV6_2292HOPOPTS:
2670 			case IPV6_2292DSTOPTS:
2671 			case IPV6_2292RTHDR:
2672 				/* RFC 2292 */
2673 				if (optlen != sizeof(int)) {
2674 					error = EINVAL;
2675 					break;
2676 				}
2677 				error = sooptcopyin(sopt, &optval,
2678 				    sizeof(optval), sizeof(optval));
2679 				if (error) {
2680 					break;
2681 				}
2682 				switch (optname) {
2683 				case IPV6_2292PKTINFO:
2684 					OPTSET2292(IN6P_PKTINFO);
2685 					break;
2686 				case IPV6_2292HOPLIMIT:
2687 					OPTSET2292(IN6P_HOPLIMIT);
2688 					break;
2689 				case IPV6_2292HOPOPTS:
2690 					/*
2691 					 * Check super-user privilege.
2692 					 * See comments for IPV6_RECVHOPOPTS.
2693 					 */
2694 					if (!privileged) {
2695 						return EPERM;
2696 					}
2697 					OPTSET2292(IN6P_HOPOPTS);
2698 					capture_exthdrstat_in = TRUE;
2699 					break;
2700 				case IPV6_2292DSTOPTS:
2701 					if (!privileged) {
2702 						return EPERM;
2703 					}
2704 					OPTSET2292(IN6P_DSTOPTS |
2705 					    IN6P_RTHDRDSTOPTS); /* XXX */
2706 					capture_exthdrstat_in = TRUE;
2707 					break;
2708 				case IPV6_2292RTHDR:
2709 					OPTSET2292(IN6P_RTHDR);
2710 					capture_exthdrstat_in = TRUE;
2711 					break;
2712 				}
2713 				break;
2714 
2715 			case IPV6_3542PKTINFO:
2716 			case IPV6_3542HOPOPTS:
2717 			case IPV6_3542RTHDR:
2718 			case IPV6_3542DSTOPTS:
2719 			case IPV6_RTHDRDSTOPTS:
2720 			case IPV6_3542NEXTHOP: {
2721 				struct ip6_pktopts **optp;
2722 				/* new advanced API (RFC3542) */
2723 				mbuf_ref_t m;
2724 
2725 				/* cannot mix with RFC2292 */
2726 				if (OPTBIT(IN6P_RFC2292)) {
2727 					error = EINVAL;
2728 					break;
2729 				}
2730 				error = soopt_getm(sopt, &m);
2731 				if (error != 0) {
2732 					break;
2733 				}
2734 				error = soopt_mcopyin(sopt, m);
2735 				if (error != 0) {
2736 					break;
2737 				}
2738 
2739 				optp = &in6p->in6p_outputopts;
2740 				error = ip6_pcbopt(optname, mtod(m, u_char *),
2741 				    m->m_len, optp, uproto);
2742 				m_freem(m);
2743 				break;
2744 			}
2745 #undef OPTSET
2746 #undef OPTSET2
2747 			case IPV6_MULTICAST_IF:
2748 			case IPV6_MULTICAST_HOPS:
2749 			case IPV6_MULTICAST_LOOP:
2750 			case IPV6_JOIN_GROUP:
2751 			case IPV6_LEAVE_GROUP:
2752 			case IPV6_MSFILTER:
2753 			case MCAST_BLOCK_SOURCE:
2754 			case MCAST_UNBLOCK_SOURCE:
2755 			case MCAST_JOIN_GROUP:
2756 			case MCAST_LEAVE_GROUP:
2757 			case MCAST_JOIN_SOURCE_GROUP:
2758 			case MCAST_LEAVE_SOURCE_GROUP:
2759 				error = ip6_setmoptions(in6p, sopt);
2760 				break;
2761 
2762 			case IPV6_PORTRANGE:
2763 				error = sooptcopyin(sopt, &optval,
2764 				    sizeof(optval), sizeof(optval));
2765 				if (error) {
2766 					break;
2767 				}
2768 
2769 				switch (optval) {
2770 				case IPV6_PORTRANGE_DEFAULT:
2771 					in6p->inp_flags &= ~(INP_LOWPORT);
2772 					in6p->inp_flags &= ~(INP_HIGHPORT);
2773 					break;
2774 
2775 				case IPV6_PORTRANGE_HIGH:
2776 					in6p->inp_flags &= ~(INP_LOWPORT);
2777 					in6p->inp_flags |= INP_HIGHPORT;
2778 					break;
2779 
2780 				case IPV6_PORTRANGE_LOW:
2781 					in6p->inp_flags &= ~(INP_HIGHPORT);
2782 					in6p->inp_flags |= INP_LOWPORT;
2783 					break;
2784 
2785 				default:
2786 					error = EINVAL;
2787 					break;
2788 				}
2789 				break;
2790 #if IPSEC
2791 			case IPV6_IPSEC_POLICY: {
2792 				caddr_t req = NULL;
2793 				size_t len = 0;
2794 				mbuf_ref_t m;
2795 
2796 				if ((error = soopt_getm(sopt, &m)) != 0) {
2797 					break;
2798 				}
2799 				if ((error = soopt_mcopyin(sopt, m)) != 0) {
2800 					break;
2801 				}
2802 
2803 				req = mtod(m, caddr_t);
2804 				len = m->m_len;
2805 				error = ipsec6_set_policy(in6p, optname, req,
2806 				    len, privileged);
2807 				m_freem(m);
2808 				break;
2809 			}
2810 #endif /* IPSEC */
2811 			/*
2812 			 * IPv6 variant of IP_BOUND_IF; for details see
2813 			 * comments on IP_BOUND_IF in ip_ctloutput().
2814 			 */
2815 			case IPV6_BOUND_IF:
2816 				/* This option is settable only on IPv6 */
2817 				if (!(in6p->inp_vflag & INP_IPV6)) {
2818 					error = EINVAL;
2819 					break;
2820 				}
2821 
2822 				error = sooptcopyin(sopt, &optval,
2823 				    sizeof(optval), sizeof(optval));
2824 
2825 				if (error) {
2826 					break;
2827 				}
2828 
2829 				error = inp_bindif(in6p, optval, NULL);
2830 				break;
2831 
2832 			case IPV6_NO_IFT_CELLULAR:
2833 				/* This option is settable only for IPv6 */
2834 				if (!(in6p->inp_vflag & INP_IPV6)) {
2835 					error = EINVAL;
2836 					break;
2837 				}
2838 
2839 				error = sooptcopyin(sopt, &optval,
2840 				    sizeof(optval), sizeof(optval));
2841 
2842 				if (error) {
2843 					break;
2844 				}
2845 
2846 				/* once set, it cannot be unset */
2847 				if (!optval && INP_NO_CELLULAR(in6p)) {
2848 					error = EINVAL;
2849 					break;
2850 				}
2851 
2852 				error = so_set_restrictions(so,
2853 				    SO_RESTRICT_DENY_CELLULAR);
2854 				break;
2855 
2856 			case IPV6_OUT_IF:
2857 				/* This option is not settable */
2858 				error = EINVAL;
2859 				break;
2860 
2861 			default:
2862 				error = ENOPROTOOPT;
2863 				break;
2864 			}
2865 			if (capture_exthdrstat_in) {
2866 				if (uproto == IPPROTO_TCP) {
2867 					INC_ATOMIC_INT64_LIM(net_api_stats.nas_sock_inet6_stream_exthdr_in);
2868 				} else if (uproto == IPPROTO_UDP) {
2869 					INC_ATOMIC_INT64_LIM(net_api_stats.nas_sock_inet6_dgram_exthdr_in);
2870 				}
2871 			}
2872 			break;
2873 
2874 		case SOPT_GET:
2875 			switch (optname) {
2876 			case IPV6_2292PKTOPTIONS:
2877 				/*
2878 				 * RFC3542 (effectively) deprecated the
2879 				 * semantics of the 2292-style pktoptions.
2880 				 * Since it was not reliable in nature (i.e.,
2881 				 * applications had to expect the lack of some
2882 				 * information after all), it would make sense
2883 				 * to simplify this part by always returning
2884 				 * empty data.
2885 				 */
2886 				sopt->sopt_valsize = 0;
2887 				break;
2888 
2889 			case IPV6_RECVHOPOPTS:
2890 			case IPV6_RECVDSTOPTS:
2891 			case IPV6_RECVRTHDRDSTOPTS:
2892 			case IPV6_UNICAST_HOPS:
2893 			case IPV6_RECVPKTINFO:
2894 			case IPV6_RECVHOPLIMIT:
2895 			case IPV6_RECVRTHDR:
2896 			case IPV6_RECVPATHMTU:
2897 			case IPV6_V6ONLY:
2898 			case IPV6_PORTRANGE:
2899 			case IPV6_RECVTCLASS:
2900 			case IPV6_AUTOFLOWLABEL:
2901 				switch (optname) {
2902 				case IPV6_RECVHOPOPTS:
2903 					optval = OPTBIT(IN6P_HOPOPTS);
2904 					break;
2905 
2906 				case IPV6_RECVDSTOPTS:
2907 					optval = OPTBIT(IN6P_DSTOPTS);
2908 					break;
2909 
2910 				case IPV6_RECVRTHDRDSTOPTS:
2911 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2912 					break;
2913 
2914 				case IPV6_UNICAST_HOPS:
2915 					optval = in6p->in6p_hops;
2916 					break;
2917 
2918 				case IPV6_RECVPKTINFO:
2919 					optval = OPTBIT(IN6P_PKTINFO);
2920 					break;
2921 
2922 				case IPV6_RECVHOPLIMIT:
2923 					optval = OPTBIT(IN6P_HOPLIMIT);
2924 					break;
2925 
2926 				case IPV6_RECVRTHDR:
2927 					optval = OPTBIT(IN6P_RTHDR);
2928 					break;
2929 
2930 				case IPV6_RECVPATHMTU:
2931 					optval = OPTBIT(IN6P_MTU);
2932 					break;
2933 
2934 				case IPV6_V6ONLY:
2935 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2936 					break;
2937 
2938 				case IPV6_PORTRANGE: {
2939 					int flags;
2940 					flags = in6p->inp_flags;
2941 					if (flags & INP_HIGHPORT) {
2942 						optval = IPV6_PORTRANGE_HIGH;
2943 					} else if (flags & INP_LOWPORT) {
2944 						optval = IPV6_PORTRANGE_LOW;
2945 					} else {
2946 						optval = 0;
2947 					}
2948 					break;
2949 				}
2950 				case IPV6_RECVTCLASS:
2951 					optval = OPTBIT(IN6P_TCLASS);
2952 					break;
2953 
2954 				case IPV6_AUTOFLOWLABEL:
2955 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2956 					break;
2957 				}
2958 				if (error) {
2959 					break;
2960 				}
2961 				error = sooptcopyout(sopt, &optval,
2962 				    sizeof(optval));
2963 				break;
2964 
2965 			case IPV6_PATHMTU: {
2966 				u_int32_t pmtu = 0;
2967 				struct ip6_mtuinfo mtuinfo;
2968 				struct route_in6 sro;
2969 
2970 				bzero(&sro, sizeof(sro));
2971 
2972 				if (!(so->so_state & SS_ISCONNECTED)) {
2973 					return ENOTCONN;
2974 				}
2975 				/*
2976 				 * XXX: we dot not consider the case of source
2977 				 * routing, or optional information to specify
2978 				 * the outgoing interface.
2979 				 */
2980 				error = ip6_getpmtu(&sro, NULL, NULL,
2981 				    &in6p->in6p_faddr, in6p->inp_fifscope, &pmtu);
2982 				ROUTE_RELEASE(&sro);
2983 				if (error) {
2984 					break;
2985 				}
2986 				if (pmtu > IPV6_MAXPACKET) {
2987 					pmtu = IPV6_MAXPACKET;
2988 				}
2989 
2990 				bzero(&mtuinfo, sizeof(mtuinfo));
2991 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2992 				optdata = (void *)&mtuinfo;
2993 				optdatalen = sizeof(mtuinfo);
2994 				error = sooptcopyout(sopt, optdata,
2995 				    optdatalen);
2996 				break;
2997 			}
2998 
2999 			case IPV6_2292PKTINFO:
3000 			case IPV6_2292HOPLIMIT:
3001 			case IPV6_2292HOPOPTS:
3002 			case IPV6_2292RTHDR:
3003 			case IPV6_2292DSTOPTS:
3004 				switch (optname) {
3005 				case IPV6_2292PKTINFO:
3006 					optval = OPTBIT(IN6P_PKTINFO);
3007 					break;
3008 				case IPV6_2292HOPLIMIT:
3009 					optval = OPTBIT(IN6P_HOPLIMIT);
3010 					break;
3011 				case IPV6_2292HOPOPTS:
3012 					optval = OPTBIT(IN6P_HOPOPTS);
3013 					break;
3014 				case IPV6_2292RTHDR:
3015 					optval = OPTBIT(IN6P_RTHDR);
3016 					break;
3017 				case IPV6_2292DSTOPTS:
3018 					optval = OPTBIT(IN6P_DSTOPTS |
3019 					    IN6P_RTHDRDSTOPTS);
3020 					break;
3021 				}
3022 				error = sooptcopyout(sopt, &optval,
3023 				    sizeof(optval));
3024 				break;
3025 
3026 			case IPV6_PKTINFO:
3027 			case IPV6_HOPOPTS:
3028 			case IPV6_RTHDR:
3029 			case IPV6_DSTOPTS:
3030 			case IPV6_RTHDRDSTOPTS:
3031 			case IPV6_NEXTHOP:
3032 			case IPV6_TCLASS:
3033 			case IPV6_DONTFRAG:
3034 			case IPV6_USE_MIN_MTU:
3035 			case IPV6_PREFER_TEMPADDR:
3036 				error = ip6_getpcbopt(in6p->in6p_outputopts,
3037 				    optname, sopt);
3038 				break;
3039 
3040 			case IPV6_MULTICAST_IF:
3041 			case IPV6_MULTICAST_HOPS:
3042 			case IPV6_MULTICAST_LOOP:
3043 			case IPV6_MSFILTER:
3044 				error = ip6_getmoptions(in6p, sopt);
3045 				break;
3046 #if IPSEC
3047 			case IPV6_IPSEC_POLICY: {
3048 				error = 0; /* This option is no longer supported */
3049 				break;
3050 			}
3051 #endif /* IPSEC */
3052 			case IPV6_BOUND_IF:
3053 				if (in6p->inp_flags & INP_BOUND_IF) {
3054 					optval = in6p->inp_boundifp->if_index;
3055 				}
3056 				error = sooptcopyout(sopt, &optval,
3057 				    sizeof(optval));
3058 				break;
3059 
3060 			case IPV6_NO_IFT_CELLULAR:
3061 				optval = INP_NO_CELLULAR(in6p) ? 1 : 0;
3062 				error = sooptcopyout(sopt, &optval,
3063 				    sizeof(optval));
3064 				break;
3065 
3066 			case IPV6_OUT_IF:
3067 				optval = (in6p->in6p_last_outifp != NULL) ?
3068 				    in6p->in6p_last_outifp->if_index : 0;
3069 				error = sooptcopyout(sopt, &optval,
3070 				    sizeof(optval));
3071 				break;
3072 
3073 			default:
3074 				error = ENOPROTOOPT;
3075 				break;
3076 			}
3077 			break;
3078 		}
3079 	} else if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_BINDTODEVICE) {
3080 		char namebuf[IFNAMSIZ + 1] = {};
3081 
3082 		switch (op) {
3083 		case SOPT_SET:
3084 			error = sooptcopyin_bindtodevice(sopt, namebuf, sizeof(namebuf));
3085 			if (error != 0) {
3086 				break;
3087 			}
3088 			namebuf[IFNAMSIZ] = 0;
3089 
3090 			error = inp_bindtodevice(in6p, __unsafe_null_terminated_from_indexable(namebuf));
3091 
3092 			break;
3093 		case SOPT_GET:
3094 			if (in6p->inp_flags & INP_BOUND_IF) {
3095 				strlcpy(namebuf, in6p->inp_boundifp->if_xname, IFNAMSIZ);
3096 			}
3097 			error = sooptcopyout(sopt, &namebuf, IFNAMSIZ);
3098 			break;
3099 		}
3100 	} else {
3101 		error = EINVAL;
3102 	}
3103 	return error;
3104 }
3105 
3106 int
ip6_raw_ctloutput(struct socket * so,struct sockopt * sopt)3107 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
3108 {
3109 	int error = 0, optval;
3110 	size_t optlen;
3111 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
3112 	struct inpcb *__single in6p = sotoinpcb(so);
3113 	int level, op, optname;
3114 
3115 	level = sopt->sopt_level;
3116 	op = sopt->sopt_dir;
3117 	optname = sopt->sopt_name;
3118 	optlen = sopt->sopt_valsize;
3119 
3120 	if (level != IPPROTO_IPV6) {
3121 		return EINVAL;
3122 	}
3123 
3124 	switch (optname) {
3125 	case IPV6_CHECKSUM:
3126 		/*
3127 		 * For ICMPv6 sockets, no modification allowed for checksum
3128 		 * offset, permit "no change" values to help existing apps.
3129 		 *
3130 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
3131 		 * for an ICMPv6 socket will fail."
3132 		 * The current behavior does not meet RFC3542.
3133 		 */
3134 		switch (op) {
3135 		case SOPT_SET:
3136 			if (optlen != sizeof(int)) {
3137 				error = EINVAL;
3138 				break;
3139 			}
3140 			error = sooptcopyin(sopt, &optval, sizeof(optval),
3141 			    sizeof(optval));
3142 			if (error) {
3143 				break;
3144 			}
3145 			if ((optval % 2) != 0) {
3146 				/* the API assumes even offset values */
3147 				error = EINVAL;
3148 			} else if (SOCK_PROTO(so) == IPPROTO_ICMPV6) {
3149 				if (optval != icmp6off) {
3150 					error = EINVAL;
3151 				}
3152 			} else {
3153 				in6p->in6p_cksum = optval;
3154 			}
3155 			break;
3156 
3157 		case SOPT_GET:
3158 			if (SOCK_PROTO(so) == IPPROTO_ICMPV6) {
3159 				optval = icmp6off;
3160 			} else {
3161 				optval = in6p->in6p_cksum;
3162 			}
3163 
3164 			error = sooptcopyout(sopt, &optval, sizeof(optval));
3165 			break;
3166 
3167 		default:
3168 			error = EINVAL;
3169 			break;
3170 		}
3171 		break;
3172 	default:
3173 		error = ENOPROTOOPT;
3174 		break;
3175 	}
3176 
3177 	return error;
3178 }
3179 
3180 /*
3181  * Set up IP6 options in pcb for insertion in output packets or
3182  * specifying behavior of outgoing packets.
3183  */
3184 static int
ip6_pcbopts(struct ip6_pktopts ** pktopt,struct mbuf * m,struct socket * so,struct sockopt * sopt)3185 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so,
3186     struct sockopt *sopt)
3187 {
3188 #pragma unused(sopt)
3189 	struct ip6_pktopts *__single opt = *pktopt;
3190 	int error = 0;
3191 
3192 	/* turn off any old options. */
3193 	if (opt != NULL) {
3194 #if DIAGNOSTIC
3195 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
3196 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
3197 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr) {
3198 			printf("%s: all specified options are cleared.\n",
3199 			    __func__);
3200 		}
3201 #endif
3202 		ip6_clearpktopts(opt, -1);
3203 	} else {
3204 		opt = kalloc_type(struct ip6_pktopts, Z_WAITOK | Z_NOFAIL);
3205 	}
3206 	*pktopt = NULL;
3207 
3208 	if (m == NULL || m->m_len == 0) {
3209 		/*
3210 		 * Only turning off any previous options, regardless of
3211 		 * whether the opt is just created or given.
3212 		 */
3213 		if (opt != NULL) {
3214 			kfree_type(struct ip6_pktopts, opt);
3215 		}
3216 		return 0;
3217 	}
3218 
3219 	/*  set options specified by user. */
3220 	if ((error = ip6_setpktopts(m, opt, NULL, SOCK_PROTO(so))) != 0) {
3221 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
3222 		kfree_type(struct ip6_pktopts, opt);
3223 		return error;
3224 	}
3225 	*pktopt = opt;
3226 	return 0;
3227 }
3228 
3229 /*
3230  * initialize ip6_pktopts.  beware that there are non-zero default values in
3231  * the struct.
3232  */
3233 void
ip6_initpktopts(struct ip6_pktopts * opt)3234 ip6_initpktopts(struct ip6_pktopts *opt)
3235 {
3236 	bzero(opt, sizeof(*opt));
3237 	opt->ip6po_hlim = -1;   /* -1 means default hop limit */
3238 	opt->ip6po_tclass = -1; /* -1 means default traffic class */
3239 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
3240 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
3241 }
3242 
3243 static int
ip6_pcbopt(int optname,u_char * buf __sized_by (len),int len,struct ip6_pktopts ** pktopt,int uproto)3244 ip6_pcbopt(int optname, u_char *buf __sized_by(len), int len, struct ip6_pktopts **pktopt,
3245     int uproto)
3246 {
3247 	struct ip6_pktopts *__single opt;
3248 
3249 	opt = *pktopt;
3250 	if (opt == NULL) {
3251 		opt = kalloc_type(struct ip6_pktopts, Z_WAITOK | Z_NOFAIL);
3252 		ip6_initpktopts(opt);
3253 		*pktopt = opt;
3254 	}
3255 
3256 	return ip6_setpktopt(optname, buf, len, opt, 1, 0, uproto);
3257 }
3258 
3259 static int
ip6_getpcbopt(struct ip6_pktopts * pktopt,int optname,struct sockopt * sopt)3260 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
3261 {
3262 	struct ip6_ext *ip6e;
3263 
3264 	/*
3265 	 * Buffer for serialization of empty IPV6_PKTINO option.
3266 	 * Initialized only once.
3267 	 */
3268 	static uint8_t null_pktinfo_bytes[sizeof(struct in6_pktinfo)] = {0, };
3269 
3270 	switch (optname) {
3271 	case IPV6_PKTINFO:
3272 		if (pktopt && pktopt->ip6po_pktinfo) {
3273 			return sooptcopyout(sopt, &pktopt->ip6po_pktinfo, sizeof(struct in6_pktinfo));
3274 		} else {
3275 			return sooptcopyout(sopt, &null_pktinfo_bytes, sizeof(struct in6_pktinfo));
3276 		}
3277 	case IPV6_TCLASS: {
3278 		int optval;
3279 		if (pktopt && pktopt->ip6po_tclass >= 0) {
3280 			optval = pktopt->ip6po_tclass;
3281 		} else {
3282 			optval = 0;
3283 		}
3284 		return sooptcopyout(sopt, &optval, sizeof(optval));
3285 	}
3286 	case IPV6_HOPOPTS:
3287 		if (pktopt && pktopt->ip6po_hbh) {
3288 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
3289 			return sooptcopyout(sopt, &pktopt->ip6po_hbh, IP6_EXT_LEN(ip6e));
3290 		}
3291 		break;
3292 	case IPV6_RTHDR:
3293 		if (pktopt && pktopt->ip6po_rthdr) {
3294 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
3295 			return sooptcopyout(sopt, &pktopt->ip6po_rthdr, IP6_EXT_LEN(ip6e));
3296 		}
3297 		break;
3298 	case IPV6_RTHDRDSTOPTS:
3299 		if (pktopt && pktopt->ip6po_dest1) {
3300 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
3301 			return sooptcopyout(sopt, &pktopt->ip6po_dest1, IP6_EXT_LEN(ip6e));
3302 		}
3303 		break;
3304 	case IPV6_DSTOPTS:
3305 		if (pktopt && pktopt->ip6po_dest2) {
3306 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
3307 			return sooptcopyout(sopt, &pktopt->ip6po_dest2, IP6_EXT_LEN(ip6e));
3308 		}
3309 		break;
3310 	case IPV6_NEXTHOP:
3311 		if (pktopt && pktopt->ip6po_nexthop) {
3312 			return sooptcopyout(sopt, &pktopt->ip6po_nexthop, pktopt->ip6po_nexthop->sa_len);
3313 		}
3314 		break;
3315 	case IPV6_USE_MIN_MTU: {
3316 		int optval;
3317 		if (pktopt) {
3318 			optval = pktopt->ip6po_minmtu;
3319 		} else {
3320 			optval = IP6PO_MINMTU_MCASTONLY;
3321 		}
3322 		return sooptcopyout(sopt, &optval, sizeof(optval));
3323 	}
3324 	case IPV6_DONTFRAG: {
3325 		int on;
3326 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) {
3327 			on = 1;
3328 		} else {
3329 			on = 0;
3330 		}
3331 		return sooptcopyout(sopt, &on, sizeof(on));
3332 	}
3333 	case IPV6_PREFER_TEMPADDR: {
3334 		int optval;
3335 		if (pktopt) {
3336 			optval = pktopt->ip6po_prefer_tempaddr;
3337 		} else {
3338 			optval = IP6PO_TEMPADDR_SYSTEM;
3339 		}
3340 		return sooptcopyout(sopt, &optval, sizeof(optval));
3341 	}
3342 	default:                /* should not happen */
3343 #ifdef DIAGNOSTIC
3344 		panic("ip6_getpcbopt: unexpected option");
3345 #endif
3346 		return ENOPROTOOPT;
3347 	}
3348 
3349 	// for bincompat, update sopt->sopt_valsize to be 0
3350 	return sooptcopyout(sopt, NULL, 0);
3351 }
3352 
3353 void
ip6_clearpktopts(struct ip6_pktopts * pktopt,int optname)3354 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
3355 {
3356 	if (pktopt == NULL) {
3357 		return;
3358 	}
3359 
3360 	if (optname == -1 || optname == IPV6_PKTINFO) {
3361 		if (pktopt->ip6po_pktinfo) {
3362 			kfree_type(struct in6_pktinfo, pktopt->ip6po_pktinfo);
3363 		}
3364 		pktopt->ip6po_pktinfo = NULL;
3365 	}
3366 	if (optname == -1 || optname == IPV6_HOPLIMIT) {
3367 		pktopt->ip6po_hlim = -1;
3368 	}
3369 	if (optname == -1 || optname == IPV6_TCLASS) {
3370 		pktopt->ip6po_tclass = -1;
3371 	}
3372 	if (optname == -1 || optname == IPV6_NEXTHOP) {
3373 		ROUTE_RELEASE(&pktopt->ip6po_nextroute);
3374 		if (pktopt->ip6po_nexthop) {
3375 			kfree_data_addr(pktopt->ip6po_nexthop);
3376 		}
3377 		pktopt->ip6po_nexthop = NULL;
3378 	}
3379 	if (optname == -1 || optname == IPV6_HOPOPTS) {
3380 		if (pktopt->ip6po_hbh) {
3381 			kfree_data_addr(pktopt->ip6po_hbh);
3382 		}
3383 		pktopt->ip6po_hbh = NULL;
3384 	}
3385 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
3386 		if (pktopt->ip6po_dest1) {
3387 			kfree_data_addr(pktopt->ip6po_dest1);
3388 		}
3389 		pktopt->ip6po_dest1 = NULL;
3390 	}
3391 	if (optname == -1 || optname == IPV6_RTHDR) {
3392 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) {
3393 			kfree_data_addr(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr);
3394 		}
3395 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
3396 		ROUTE_RELEASE(&pktopt->ip6po_route);
3397 	}
3398 	if (optname == -1 || optname == IPV6_DSTOPTS) {
3399 		if (pktopt->ip6po_dest2) {
3400 			kfree_data_addr(pktopt->ip6po_dest2);
3401 		}
3402 		pktopt->ip6po_dest2 = NULL;
3403 	}
3404 }
3405 
3406 #define PKTOPT_EXTHDRCPY(t, f) do {                              \
3407 	if (src->f != NULL) {                                        \
3408 	        int hlen =                                           \
3409 	            (((struct ip6_ext *)src->f)->ip6e_len + 1) << 3; \
3410 	        dst->f = kalloc_data(hlen, canwait);                 \
3411 	        if (dst->f == NULL && canwait == Z_NOWAIT)           \
3412 	            goto bad;                                        \
3413 	        bcopy(GET_IP6EXTHDR(src, t, f),                      \
3414 	              GET_IP6EXTHDR(dst, t, f),                      \
3415 	              hlen);                                         \
3416 	}                                                            \
3417 } while (0)
3418 
3419 #define GET_IP6_DEST_OPT(b, d, dlen) {                           \
3420     do {                                                         \
3421 	        struct ip6_dest *t = (struct ip6_dest *)(void *)buf; \
3422 	        (dlen) = (t->ip6d_len + 1) << 3;                     \
3423 	        (d) = __unsafe_forge_bidi_indexable(struct ip6_dest *, (b), len); \
3424     } while(0);                                                  \
3425 }
3426 
3427 static int
copypktopts(struct ip6_pktopts * dst,struct ip6_pktopts * src,zalloc_flags_t canwait)3428 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, zalloc_flags_t canwait)
3429 {
3430 	if (dst == NULL || src == NULL) {
3431 		printf("copypktopts: invalid argument\n");
3432 		return EINVAL;
3433 	}
3434 
3435 	dst->ip6po_hlim = src->ip6po_hlim;
3436 	dst->ip6po_tclass = src->ip6po_tclass;
3437 	dst->ip6po_flags = src->ip6po_flags;
3438 	if (src->ip6po_pktinfo) {
3439 		dst->ip6po_pktinfo = kalloc_type(struct in6_pktinfo, canwait);
3440 		if (dst->ip6po_pktinfo == NULL && canwait == Z_NOWAIT) {
3441 			goto bad;
3442 		}
3443 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
3444 	}
3445 	if (src->ip6po_nexthop) {
3446 		dst->ip6po_nexthop = kalloc_data(src->ip6po_nexthop->sa_len, canwait);
3447 		if (dst->ip6po_nexthop == NULL && canwait == Z_NOWAIT) {
3448 			goto bad;
3449 		}
3450 		SOCKADDR_COPY(src->ip6po_nexthop, dst->ip6po_nexthop,
3451 		    src->ip6po_nexthop->sa_len);
3452 	}
3453 	PKTOPT_EXTHDRCPY(struct ip6_hbh*, ip6po_hbh);
3454 	PKTOPT_EXTHDRCPY(struct ip6_dest*, ip6po_dest1);
3455 	PKTOPT_EXTHDRCPY(struct ip6_dest*, ip6po_dest2);
3456 	PKTOPT_EXTHDRCPY(struct ip6_rthdr*, ip6po_rthdr);         /* not copy the cached route */
3457 	return 0;
3458 
3459 bad:
3460 	ip6_clearpktopts(dst, -1);
3461 	return ENOBUFS;
3462 }
3463 #undef PKTOPT_EXTHDRCPY
3464 
3465 struct ip6_pktopts *
ip6_copypktopts(struct ip6_pktopts * src,zalloc_flags_t canwait)3466 ip6_copypktopts(struct ip6_pktopts *src, zalloc_flags_t canwait)
3467 {
3468 	int error;
3469 	struct ip6_pktopts *__single dst;
3470 
3471 	dst = kalloc_type(struct ip6_pktopts, canwait);
3472 	if (dst == NULL) {
3473 		return NULL;
3474 	}
3475 	ip6_initpktopts(dst);
3476 
3477 	if ((error = copypktopts(dst, src, canwait)) != 0) {
3478 		kfree_type(struct ip6_pktopts, dst);
3479 		return NULL;
3480 	}
3481 
3482 	return dst;
3483 }
3484 
3485 void
ip6_freepcbopts(struct ip6_pktopts * pktopt)3486 ip6_freepcbopts(struct ip6_pktopts *pktopt)
3487 {
3488 	if (pktopt == NULL) {
3489 		return;
3490 	}
3491 
3492 	ip6_clearpktopts(pktopt, -1);
3493 
3494 	kfree_type(struct ip6_pktopts, pktopt);
3495 }
3496 
3497 void
ip6_moptions_init(void)3498 ip6_moptions_init(void)
3499 {
3500 	PE_parse_boot_argn("ifa_debug", &im6o_debug, sizeof(im6o_debug));
3501 
3502 	vm_size_t im6o_size = (im6o_debug == 0) ? sizeof(struct ip6_moptions) :
3503 	    sizeof(struct ip6_moptions_dbg);
3504 
3505 	im6o_zone = zone_create(IM6O_ZONE_NAME, im6o_size, ZC_ZFREE_CLEARMEM);
3506 }
3507 
3508 void
im6o_addref(struct ip6_moptions * im6o,int locked)3509 im6o_addref(struct ip6_moptions *im6o, int locked)
3510 {
3511 	if (!locked) {
3512 		IM6O_LOCK(im6o);
3513 	} else {
3514 		IM6O_LOCK_ASSERT_HELD(im6o);
3515 	}
3516 
3517 	if (++im6o->im6o_refcnt == 0) {
3518 		panic("%s: im6o %p wraparound refcnt", __func__, im6o);
3519 		/* NOTREACHED */
3520 	} else if (im6o->im6o_trace != NULL) {
3521 		(*im6o->im6o_trace)(im6o, TRUE);
3522 	}
3523 
3524 	if (!locked) {
3525 		IM6O_UNLOCK(im6o);
3526 	}
3527 }
3528 
3529 void
im6o_remref(struct ip6_moptions * im6o)3530 im6o_remref(struct ip6_moptions *im6o)
3531 {
3532 	int i;
3533 
3534 	IM6O_LOCK(im6o);
3535 	if (im6o->im6o_refcnt == 0) {
3536 		panic("%s: im6o %p negative refcnt", __func__, im6o);
3537 		/* NOTREACHED */
3538 	} else if (im6o->im6o_trace != NULL) {
3539 		(*im6o->im6o_trace)(im6o, FALSE);
3540 	}
3541 
3542 	--im6o->im6o_refcnt;
3543 	if (im6o->im6o_refcnt > 0) {
3544 		IM6O_UNLOCK(im6o);
3545 		return;
3546 	}
3547 
3548 	for (i = 0; i < im6o->im6o_num_memberships; ++i) {
3549 		struct in6_mfilter *__single imf;
3550 
3551 		imf = im6o->im6o_mfilters ? &im6o->im6o_mfilters[i] : NULL;
3552 		if (imf != NULL) {
3553 			im6f_leave(imf);
3554 		}
3555 
3556 		(void) in6_mc_leave(im6o->im6o_membership[i], imf);
3557 
3558 		if (imf != NULL) {
3559 			im6f_purge(imf);
3560 		}
3561 
3562 		IN6M_REMREF(im6o->im6o_membership[i]);
3563 		im6o->im6o_membership[i] = NULL;
3564 	}
3565 	im6o->im6o_num_memberships = 0;
3566 	IM6O_UNLOCK(im6o);
3567 
3568 	kfree_type_counted_by(struct in6_multi *, im6o->im6o_max_memberships, im6o->im6o_membership);
3569 	kfree_type_counted_by(struct in6_mfilter, im6o->im6o_max_filters, im6o->im6o_mfilters);
3570 
3571 	lck_mtx_destroy(&im6o->im6o_lock, &ifa_mtx_grp);
3572 
3573 	if (!(im6o->im6o_debug & IFD_ALLOC)) {
3574 		panic("%s: im6o %p cannot be freed", __func__, im6o);
3575 		/* NOTREACHED */
3576 	}
3577 	zfree(im6o_zone, im6o);
3578 }
3579 
3580 static void
im6o_trace(struct ip6_moptions * im6o,int refhold)3581 im6o_trace(struct ip6_moptions *im6o, int refhold)
3582 {
3583 	struct ip6_moptions_dbg *im6o_dbg = (struct ip6_moptions_dbg *)im6o;
3584 	ctrace_t *tr;
3585 	u_int32_t idx;
3586 	u_int16_t *cnt;
3587 
3588 	if (!(im6o->im6o_debug & IFD_DEBUG)) {
3589 		panic("%s: im6o %p has no debug structure", __func__, im6o);
3590 		/* NOTREACHED */
3591 	}
3592 	if (refhold) {
3593 		cnt = &im6o_dbg->im6o_refhold_cnt;
3594 		tr = im6o_dbg->im6o_refhold;
3595 	} else {
3596 		cnt = &im6o_dbg->im6o_refrele_cnt;
3597 		tr = im6o_dbg->im6o_refrele;
3598 	}
3599 
3600 	idx = os_atomic_inc_orig(cnt, relaxed) % IM6O_TRACE_HIST_SIZE;
3601 	ctrace_record(&tr[idx]);
3602 }
3603 
3604 struct ip6_moptions *
ip6_allocmoptions(zalloc_flags_t how)3605 ip6_allocmoptions(zalloc_flags_t how)
3606 {
3607 	struct ip6_moptions *im6o;
3608 
3609 	im6o = zalloc_flags(im6o_zone, how | Z_ZERO);
3610 	if (im6o != NULL) {
3611 		lck_mtx_init(&im6o->im6o_lock, &ifa_mtx_grp, &ifa_mtx_attr);
3612 		im6o->im6o_debug |= IFD_ALLOC;
3613 		if (im6o_debug != 0) {
3614 			im6o->im6o_debug |= IFD_DEBUG;
3615 			im6o->im6o_trace = im6o_trace;
3616 		}
3617 		IM6O_ADDREF(im6o);
3618 	}
3619 
3620 	return im6o;
3621 }
3622 
3623 /*
3624  * Set IPv6 outgoing packet options based on advanced API.
3625  */
3626 int
ip6_setpktopts(struct mbuf * control,struct ip6_pktopts * opt,struct ip6_pktopts * stickyopt,int uproto)3627 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
3628     struct ip6_pktopts *stickyopt, int uproto)
3629 {
3630 	struct cmsghdr *cm = NULL;
3631 
3632 	if (control == NULL || opt == NULL) {
3633 		return EINVAL;
3634 	}
3635 
3636 	ip6_initpktopts(opt);
3637 	if (stickyopt != NULL) {
3638 		int error;
3639 
3640 		/*
3641 		 * If stickyopt is provided, make a local copy of the options
3642 		 * for this particular packet, then override them by ancillary
3643 		 * objects.
3644 		 * XXX: copypktopts() does not copy the cached route to a next
3645 		 * hop (if any).  This is not very good in terms of efficiency,
3646 		 * but we can allow this since this option should be rarely
3647 		 * used.
3648 		 */
3649 
3650 		if ((error = copypktopts(opt, stickyopt, Z_NOWAIT)) != 0) {
3651 			return error;
3652 		}
3653 	}
3654 
3655 	/*
3656 	 * XXX: Currently, we assume all the optional information is stored
3657 	 * in a single mbuf.
3658 	 */
3659 	if (control->m_next) {
3660 		return EINVAL;
3661 	}
3662 
3663 	if (control->m_len < CMSG_LEN(0)) {
3664 		return EINVAL;
3665 	}
3666 
3667 	for (cm = M_FIRST_CMSGHDR(control);
3668 	    is_cmsg_valid(control, cm);
3669 	    cm = M_NXT_CMSGHDR(control, cm)) {
3670 		int error;
3671 
3672 		if (cm->cmsg_level != IPPROTO_IPV6) {
3673 			continue;
3674 		}
3675 
3676 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
3677 		    cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto);
3678 		if (error) {
3679 			return error;
3680 		}
3681 	}
3682 
3683 	return 0;
3684 }
3685 /*
3686  * Set a particular packet option, as a sticky option or an ancillary data
3687  * item.  "len" can be 0 only when it's a sticky option.
3688  * We have 4 cases of combination of "sticky" and "cmsg":
3689  * "sticky=0, cmsg=0": impossible
3690  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
3691  * "sticky=1, cmsg=0": RFC3542 socket option
3692  * "sticky=1, cmsg=1": RFC2292 socket option
3693  */
3694 static int
ip6_setpktopt(int optname,uint8_t * buf __sized_by (len),int len,struct ip6_pktopts * opt,int sticky,int cmsg,int uproto)3695 ip6_setpktopt(int optname, uint8_t *buf __sized_by(len), int len, struct ip6_pktopts *opt,
3696     int sticky, int cmsg, int uproto)
3697 {
3698 	int minmtupolicy, preftemp;
3699 	int error;
3700 	boolean_t capture_exthdrstat_out = FALSE;
3701 
3702 	if (!sticky && !cmsg) {
3703 #ifdef DIAGNOSTIC
3704 		printf("ip6_setpktopt: impossible case\n");
3705 #endif
3706 		return EINVAL;
3707 	}
3708 
3709 	/*
3710 	 * Caller must have ensured that the buffer is at least
3711 	 * aligned on 32-bit boundary.
3712 	 */
3713 	VERIFY(IS_P2ALIGNED(buf, sizeof(u_int32_t)));
3714 
3715 	/*
3716 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
3717 	 * not be specified in the context of RFC3542.  Conversely,
3718 	 * RFC3542 types should not be specified in the context of RFC2292.
3719 	 */
3720 	if (!cmsg) {
3721 		switch (optname) {
3722 		case IPV6_2292PKTINFO:
3723 		case IPV6_2292HOPLIMIT:
3724 		case IPV6_2292NEXTHOP:
3725 		case IPV6_2292HOPOPTS:
3726 		case IPV6_2292DSTOPTS:
3727 		case IPV6_2292RTHDR:
3728 		case IPV6_2292PKTOPTIONS:
3729 			return ENOPROTOOPT;
3730 		}
3731 	}
3732 	if (sticky && cmsg) {
3733 		switch (optname) {
3734 		case IPV6_PKTINFO:
3735 		case IPV6_HOPLIMIT:
3736 		case IPV6_NEXTHOP:
3737 		case IPV6_HOPOPTS:
3738 		case IPV6_DSTOPTS:
3739 		case IPV6_RTHDRDSTOPTS:
3740 		case IPV6_RTHDR:
3741 		case IPV6_USE_MIN_MTU:
3742 		case IPV6_DONTFRAG:
3743 		case IPV6_TCLASS:
3744 		case IPV6_PREFER_TEMPADDR:         /* XXX: not an RFC3542 option */
3745 			return ENOPROTOOPT;
3746 		}
3747 	}
3748 
3749 	switch (optname) {
3750 	case IPV6_2292PKTINFO:
3751 	case IPV6_PKTINFO: {
3752 		struct ifnet *__single ifp = NULL;
3753 		struct in6_pktinfo *pktinfo;
3754 
3755 		if (len != sizeof(struct in6_pktinfo)) {
3756 			return EINVAL;
3757 		}
3758 
3759 		pktinfo = (struct in6_pktinfo *)(void *)buf;
3760 
3761 		/*
3762 		 * An application can clear any sticky IPV6_PKTINFO option by
3763 		 * doing a "regular" setsockopt with ipi6_addr being
3764 		 * in6addr_any and ipi6_ifindex being zero.
3765 		 * [RFC 3542, Section 6]
3766 		 */
3767 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
3768 		    pktinfo->ipi6_ifindex == 0 &&
3769 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3770 			ip6_clearpktopts(opt, optname);
3771 			break;
3772 		}
3773 
3774 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
3775 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3776 			return EINVAL;
3777 		}
3778 
3779 		/* validate the interface index if specified. */
3780 		ifnet_head_lock_shared();
3781 
3782 		if (pktinfo->ipi6_ifindex > if_index) {
3783 			ifnet_head_done();
3784 			return ENXIO;
3785 		}
3786 
3787 		if (pktinfo->ipi6_ifindex) {
3788 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
3789 			if (ifp == NULL) {
3790 				ifnet_head_done();
3791 				return ENXIO;
3792 			}
3793 		}
3794 
3795 		ifnet_head_done();
3796 
3797 		/*
3798 		 * We store the address anyway, and let in6_selectsrc()
3799 		 * validate the specified address.  This is because ipi6_addr
3800 		 * may not have enough information about its scope zone, and
3801 		 * we may need additional information (such as outgoing
3802 		 * interface or the scope zone of a destination address) to
3803 		 * disambiguate the scope.
3804 		 * XXX: the delay of the validation may confuse the
3805 		 * application when it is used as a sticky option.
3806 		 */
3807 		if (opt->ip6po_pktinfo == NULL) {
3808 			opt->ip6po_pktinfo = kalloc_type(struct in6_pktinfo, Z_NOWAIT);
3809 			if (opt->ip6po_pktinfo == NULL) {
3810 				return ENOBUFS;
3811 			}
3812 		}
3813 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3814 		break;
3815 	}
3816 
3817 	case IPV6_2292HOPLIMIT:
3818 	case IPV6_HOPLIMIT: {
3819 		int *hlimp;
3820 
3821 		/*
3822 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3823 		 * to simplify the ordering among hoplimit options.
3824 		 */
3825 		if (optname == IPV6_HOPLIMIT && sticky) {
3826 			return ENOPROTOOPT;
3827 		}
3828 
3829 		if (len != sizeof(int)) {
3830 			return EINVAL;
3831 		}
3832 		hlimp = (int *)(void *)buf;
3833 		if (*hlimp < -1 || *hlimp > IPV6_MAXHLIM) {
3834 			return EINVAL;
3835 		}
3836 
3837 		opt->ip6po_hlim = *hlimp;
3838 		break;
3839 	}
3840 
3841 	case IPV6_TCLASS: {
3842 		int tclass;
3843 
3844 		if (len != sizeof(int)) {
3845 			return EINVAL;
3846 		}
3847 		tclass = *(int *)(void *)buf;
3848 		if (tclass < -1 || tclass > 255) {
3849 			return EINVAL;
3850 		}
3851 
3852 		opt->ip6po_tclass = tclass;
3853 		break;
3854 	}
3855 
3856 	case IPV6_2292NEXTHOP:
3857 	case IPV6_NEXTHOP:
3858 		error = suser(kauth_cred_get(), 0);
3859 		if (error) {
3860 			return EACCES;
3861 		}
3862 
3863 		if (len == 0) {         /* just remove the option */
3864 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3865 			break;
3866 		}
3867 
3868 		/* check if cmsg_len is large enough for sa_len */
3869 		if (len < sizeof(struct sockaddr) || len < *buf) {
3870 			return EINVAL;
3871 		}
3872 
3873 		switch (SA(buf)->sa_family) {
3874 		case AF_INET6: {
3875 			struct sockaddr_in6 *sa6 = SIN6(buf);
3876 
3877 			if (sa6->sin6_len != sizeof(struct sockaddr_in6)) {
3878 				return EINVAL;
3879 			}
3880 
3881 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3882 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3883 				return EINVAL;
3884 			}
3885 			if ((error = sa6_embedscope(sa6, ip6_use_defzone, IN6_NULL_IF_EMBEDDED_SCOPE(&sa6->sin6_scope_id)))
3886 			    != 0) {
3887 				return error;
3888 			}
3889 			break;
3890 		}
3891 		case AF_LINK:         /* should eventually be supported */
3892 		default:
3893 			return EAFNOSUPPORT;
3894 		}
3895 
3896 		/* turn off the previous option, then set the new option. */
3897 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3898 		opt->ip6po_nexthop = kalloc_data(*buf, Z_NOWAIT);
3899 		if (opt->ip6po_nexthop == NULL) {
3900 			return ENOBUFS;
3901 		}
3902 		SOCKADDR_COPY(buf, opt->ip6po_nexthop, *buf);
3903 		break;
3904 
3905 	case IPV6_2292HOPOPTS:
3906 	case IPV6_HOPOPTS: {
3907 		struct ip6_hbh *hbh, *thbh = NULL;
3908 		int hbhlen;
3909 
3910 		/*
3911 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3912 		 * options, since per-option restriction has too much
3913 		 * overhead.
3914 		 */
3915 		error = suser(kauth_cred_get(), 0);
3916 		if (error) {
3917 			return EACCES;
3918 		}
3919 
3920 		if (len == 0) {
3921 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3922 			break;         /* just remove the option */
3923 		}
3924 
3925 		/* message length validation */
3926 		if (len < sizeof(struct ip6_hbh)) {
3927 			return EINVAL;
3928 		}
3929 		hbh = (struct ip6_hbh *)(void *)buf;
3930 		hbhlen = (hbh->ip6h_len + 1) << 3;
3931 		if (len != hbhlen) {
3932 			return EINVAL;
3933 		}
3934 
3935 		/* turn off the previous option, then set the new option. */
3936 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3937 		thbh = kalloc_data(hbhlen, Z_NOWAIT);
3938 		if (thbh == NULL) {
3939 			return ENOBUFS;
3940 		}
3941 
3942 		bcopy(hbh, thbh, hbhlen);
3943 		opt->ip6po_hbh = thbh;
3944 		capture_exthdrstat_out = TRUE;
3945 		break;
3946 	}
3947 
3948 	case IPV6_2292DSTOPTS:
3949 	case IPV6_DSTOPTS:
3950 	case IPV6_RTHDRDSTOPTS: {
3951 		struct ip6_dest *dest, *__single * newdest = NULL;
3952 		int destlen;
3953 
3954 		error = suser(kauth_cred_get(), 0);
3955 		if (error) {
3956 			return EACCES;
3957 		}
3958 
3959 		if (len == 0) {
3960 			ip6_clearpktopts(opt, optname);
3961 			break;         /* just remove the option */
3962 		}
3963 
3964 		/* message length validation */
3965 		if (len < sizeof(struct ip6_dest)) {
3966 			return EINVAL;
3967 		}
3968 
3969 		GET_IP6_DEST_OPT(buf, dest, destlen);
3970 		if (len != destlen) {
3971 			return EINVAL;
3972 		}
3973 
3974 		/*
3975 		 * Determine the position that the destination options header
3976 		 * should be inserted; before or after the routing header.
3977 		 */
3978 		switch (optname) {
3979 		case IPV6_2292DSTOPTS:
3980 			/*
3981 			 * The old advanced API is ambiguous on this point.
3982 			 * Our approach is to determine the position based
3983 			 * according to the existence of a routing header.
3984 			 * Note, however, that this depends on the order of the
3985 			 * extension headers in the ancillary data; the 1st
3986 			 * part of the destination options header must appear
3987 			 * before the routing header in the ancillary data,
3988 			 * too.
3989 			 * RFC3542 solved the ambiguity by introducing
3990 			 * separate ancillary data or option types.
3991 			 */
3992 			if (opt->ip6po_rthdr == NULL) {
3993 				newdest = &opt->ip6po_dest1;
3994 			} else {
3995 				newdest = &opt->ip6po_dest2;
3996 			}
3997 			break;
3998 		case IPV6_RTHDRDSTOPTS:
3999 			newdest = &opt->ip6po_dest1;
4000 			break;
4001 		case IPV6_DSTOPTS:
4002 			newdest = &opt->ip6po_dest2;
4003 			break;
4004 		}
4005 
4006 		/* turn off the previous option, then set the new option. */
4007 		ip6_clearpktopts(opt, optname);
4008 
4009 		void *destbytes = kalloc_data(destlen, Z_NOWAIT);
4010 		if (destbytes == NULL) {
4011 			return ENOBUFS;
4012 		}
4013 		bcopy(dest, destbytes, destlen);
4014 		*newdest = destbytes;
4015 
4016 		capture_exthdrstat_out = TRUE;
4017 		break;
4018 	}
4019 
4020 	case IPV6_2292RTHDR:
4021 	case IPV6_RTHDR: {
4022 		struct ip6_rthdr *rth, *trth = NULL;
4023 		int rthlen;
4024 
4025 		if (len == 0) {
4026 			ip6_clearpktopts(opt, IPV6_RTHDR);
4027 			break;         /* just remove the option */
4028 		}
4029 
4030 		/* message length validation */
4031 		if (len < sizeof(struct ip6_rthdr)) {
4032 			return EINVAL;
4033 		}
4034 		rth = (struct ip6_rthdr *)(void *)buf;
4035 		rthlen = (rth->ip6r_len + 1) << 3;
4036 		if (len != rthlen) {
4037 			return EINVAL;
4038 		}
4039 
4040 		switch (rth->ip6r_type) {
4041 		case IPV6_RTHDR_TYPE_0:
4042 			if (rth->ip6r_len == 0) {         /* must contain one addr */
4043 				return EINVAL;
4044 			}
4045 			if (rth->ip6r_len % 2) {         /* length must be even */
4046 				return EINVAL;
4047 			}
4048 			if (rth->ip6r_len / 2 != rth->ip6r_segleft) {
4049 				return EINVAL;
4050 			}
4051 			break;
4052 		default:
4053 			return EINVAL;         /* not supported */
4054 		}
4055 
4056 		/* turn off the previous option */
4057 		ip6_clearpktopts(opt, IPV6_RTHDR);
4058 
4059 		trth = kalloc_data(rthlen, Z_NOWAIT);
4060 		if (trth == NULL) {
4061 			return ENOBUFS;
4062 		}
4063 		bcopy(rth, trth, rthlen);
4064 		opt->ip6po_rthdr = trth;
4065 		capture_exthdrstat_out = TRUE;
4066 		break;
4067 	}
4068 
4069 	case IPV6_USE_MIN_MTU:
4070 		if (len != sizeof(int)) {
4071 			return EINVAL;
4072 		}
4073 		minmtupolicy = *(int *)(void *)buf;
4074 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
4075 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
4076 		    minmtupolicy != IP6PO_MINMTU_ALL) {
4077 			return EINVAL;
4078 		}
4079 		opt->ip6po_minmtu = minmtupolicy;
4080 		break;
4081 
4082 	case IPV6_DONTFRAG:
4083 		if (len != sizeof(int)) {
4084 			return EINVAL;
4085 		}
4086 
4087 		if (uproto == IPPROTO_TCP || *(int *)(void *)buf == 0) {
4088 			/*
4089 			 * we ignore this option for TCP sockets.
4090 			 * (RFC3542 leaves this case unspecified.)
4091 			 */
4092 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
4093 		} else {
4094 			opt->ip6po_flags |= IP6PO_DONTFRAG;
4095 		}
4096 		break;
4097 
4098 	case IPV6_PREFER_TEMPADDR:
4099 		if (len != sizeof(int)) {
4100 			return EINVAL;
4101 		}
4102 		preftemp = *(int *)(void *)buf;
4103 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
4104 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
4105 		    preftemp != IP6PO_TEMPADDR_PREFER) {
4106 			return EINVAL;
4107 		}
4108 		opt->ip6po_prefer_tempaddr = preftemp;
4109 		break;
4110 
4111 	default:
4112 		return ENOPROTOOPT;
4113 	}         /* end of switch */
4114 
4115 	if (capture_exthdrstat_out) {
4116 		if (uproto == IPPROTO_TCP) {
4117 			INC_ATOMIC_INT64_LIM(net_api_stats.nas_sock_inet6_stream_exthdr_out);
4118 		} else if (uproto == IPPROTO_UDP) {
4119 			INC_ATOMIC_INT64_LIM(net_api_stats.nas_sock_inet6_dgram_exthdr_out);
4120 		}
4121 	}
4122 
4123 	return 0;
4124 }
4125 
4126 /*
4127  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
4128  * packet to the input queue of a specified interface.  Note that this
4129  * calls the output routine of the loopback "driver", but with an interface
4130  * pointer that might NOT be &loif -- easier than replicating that code here.
4131  */
4132 void
ip6_mloopback(struct ifnet * srcifp,struct ifnet * origifp,struct mbuf * m,struct sockaddr_in6 * dst,uint32_t optlen,int32_t nxt0)4133 ip6_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m,
4134     struct sockaddr_in6 *dst, uint32_t optlen, int32_t nxt0)
4135 {
4136 	mbuf_ref_t copym;
4137 	struct ip6_hdr *__single ip6;
4138 	struct in6_addr src;
4139 
4140 	if (lo_ifp == NULL) {
4141 		return;
4142 	}
4143 
4144 	/*
4145 	 * Copy the packet header as it's needed for the checksum.
4146 	 * Make sure to deep-copy IPv6 header portion in case the data
4147 	 * is in an mbuf cluster, so that we can safely override the IPv6
4148 	 * header portion later.
4149 	 */
4150 	copym = m_copym_mode(m, 0, M_COPYALL, M_DONTWAIT, NULL, NULL, M_COPYM_COPY_HDR);
4151 	if (copym != NULL && ((copym->m_flags & M_EXT) ||
4152 	    copym->m_len < sizeof(struct ip6_hdr))) {
4153 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
4154 	}
4155 
4156 	if (copym == NULL) {
4157 		return;
4158 	}
4159 
4160 	ip6 = mtod(copym, struct ip6_hdr *);
4161 	src = ip6->ip6_src;
4162 	/*
4163 	 * clear embedded scope identifiers if necessary.
4164 	 * in6_clearscope will touch the addresses only when necessary.
4165 	 */
4166 	in6_clearscope(&ip6->ip6_src);
4167 	in6_clearscope(&ip6->ip6_dst);
4168 
4169 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) {
4170 		in6_delayed_cksum_offset(copym, 0, optlen, nxt0);
4171 	}
4172 
4173 	/*
4174 	 * Stuff the 'real' ifp into the pkthdr, to be used in matching
4175 	 * in ip6_input(); we need the loopback ifp/dl_tag passed as args
4176 	 * to make the loopback driver compliant with the data link
4177 	 * requirements.
4178 	 */
4179 	copym->m_pkthdr.rcvif = origifp;
4180 
4181 	/*
4182 	 * Also record the source interface (which owns the source address).
4183 	 * This is basically a stripped down version of ifa_foraddr6().
4184 	 */
4185 	if (srcifp == NULL) {
4186 		struct in6_ifaddr *__single ia;
4187 
4188 		lck_rw_lock_shared(&in6_ifaddr_rwlock);
4189 		TAILQ_FOREACH(ia, IN6ADDR_HASH(&src), ia6_hash) {
4190 			IFA_LOCK_SPIN(&ia->ia_ifa);
4191 			/* compare against src addr with embedded scope */
4192 			if (in6_are_addr_equal_scoped(&ia->ia_addr.sin6_addr, &src, ia->ia_addr.sin6_scope_id, ip6_output_getsrcifscope(m))) {
4193 				srcifp = ia->ia_ifp;
4194 				IFA_UNLOCK(&ia->ia_ifa);
4195 				break;
4196 			}
4197 			IFA_UNLOCK(&ia->ia_ifa);
4198 		}
4199 		lck_rw_done(&in6_ifaddr_rwlock);
4200 	}
4201 	if (srcifp != NULL) {
4202 		ip6_setsrcifaddr_info(copym, srcifp->if_index, NULL);
4203 	}
4204 	ip6_setdstifaddr_info(copym, origifp->if_index, NULL);
4205 
4206 	dlil_output(lo_ifp, PF_INET6, copym, NULL, SA(dst), 0, NULL);
4207 }
4208 
4209 /*
4210  * Chop IPv6 header off from the payload.
4211  */
4212 static int
ip6_splithdr(struct mbuf * m,struct ip6_exthdrs * exthdrs)4213 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
4214 {
4215 	mbuf_ref_t mh;
4216 	struct ip6_hdr *ip6;
4217 
4218 	ip6 = mtod(m, struct ip6_hdr *);
4219 	if (m->m_len > sizeof(*ip6)) {
4220 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);         /* MAC-OK */
4221 		if (mh == NULL) {
4222 			m_freem(m);
4223 			return ENOBUFS;
4224 		}
4225 		M_COPY_PKTHDR(mh, m);
4226 		MH_ALIGN(mh, sizeof(*ip6));
4227 		m->m_flags &= ~M_PKTHDR;
4228 		m->m_len -= sizeof(*ip6);
4229 		m->m_data += sizeof(*ip6);
4230 		mh->m_next = m;
4231 		m = mh;
4232 		m->m_len = sizeof(*ip6);
4233 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
4234 	}
4235 	exthdrs->ip6e_ip6 = m;
4236 	return 0;
4237 }
4238 
4239 static void
ip6_output_checksum(struct ifnet * ifp,uint32_t mtu,struct mbuf * m,int nxt0,uint32_t tlen,uint32_t optlen)4240 ip6_output_checksum(struct ifnet *ifp, uint32_t mtu, struct mbuf *m,
4241     int nxt0, uint32_t tlen, uint32_t optlen)
4242 {
4243 	uint32_t sw_csum, hwcap = ifp->if_hwassist;
4244 
4245 	if (!hwcksum_tx) {
4246 		/* do all in software; checksum offload is disabled */
4247 		sw_csum = CSUM_DELAY_IPV6_DATA & m->m_pkthdr.csum_flags;
4248 	} else {
4249 		/* do in software what the hardware cannot */
4250 		sw_csum = m->m_pkthdr.csum_flags & ~IF_HWASSIST_CSUM_FLAGS(hwcap);
4251 	}
4252 
4253 	if (optlen != 0) {
4254 		sw_csum |= (CSUM_DELAY_IPV6_DATA &
4255 		    m->m_pkthdr.csum_flags);
4256 	} else if ((sw_csum & CSUM_DELAY_IPV6_DATA) && (hwcap & CSUM_PARTIAL)) {
4257 		/*
4258 		 * Partial checksum offload, ere), if no extension headers,
4259 		 * and TCP only (no UDP support, as the hardware may not be
4260 		 * able to convert +0 to -0 (0xffff) per RFC1122 4.1.3.4.
4261 		 * unless the interface supports "invert zero" capability.)
4262 		 */
4263 		if (hwcksum_tx &&
4264 		    ((m->m_pkthdr.csum_flags & CSUM_TCPIPV6) ||
4265 		    ((hwcap & CSUM_ZERO_INVERT) &&
4266 		    (m->m_pkthdr.csum_flags & CSUM_ZERO_INVERT))) &&
4267 		    tlen <= mtu) {
4268 			uint16_t start = sizeof(struct ip6_hdr);
4269 			uint16_t ulpoff =
4270 			    m->m_pkthdr.csum_data & 0xffff;
4271 			m->m_pkthdr.csum_flags |=
4272 			    (CSUM_DATA_VALID | CSUM_PARTIAL);
4273 			m->m_pkthdr.csum_tx_stuff = (ulpoff + start);
4274 			m->m_pkthdr.csum_tx_start = start;
4275 			sw_csum = 0;
4276 		} else {
4277 			sw_csum |= (CSUM_DELAY_IPV6_DATA &
4278 			    m->m_pkthdr.csum_flags);
4279 		}
4280 	}
4281 
4282 	if (sw_csum & CSUM_DELAY_IPV6_DATA) {
4283 		in6_delayed_cksum_offset(m, 0, optlen, nxt0);
4284 		sw_csum &= ~CSUM_DELAY_IPV6_DATA;
4285 	}
4286 
4287 	if (hwcksum_tx) {
4288 		uint32_t delay_data = m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA;
4289 		uint32_t hw_csum = IF_HWASSIST_CSUM_FLAGS(hwcap);
4290 
4291 		/*
4292 		 * Drop off bits that aren't supported by hardware;
4293 		 * also make sure to preserve non-checksum related bits.
4294 		 */
4295 		m->m_pkthdr.csum_flags =
4296 		    ((m->m_pkthdr.csum_flags & (hw_csum | CSUM_DATA_VALID)) |
4297 		    (m->m_pkthdr.csum_flags & ~IF_HWASSIST_CSUM_MASK));
4298 
4299 		/*
4300 		 * If hardware supports partial checksum but not delay_data,
4301 		 * add back delay_data.
4302 		 */
4303 		if ((hw_csum & CSUM_PARTIAL) != 0 &&
4304 		    (hw_csum & delay_data) == 0) {
4305 			m->m_pkthdr.csum_flags |= delay_data;
4306 		}
4307 	} else {
4308 		/* drop all bits; checksum offload is disabled */
4309 		m->m_pkthdr.csum_flags = 0;
4310 	}
4311 }
4312 
4313 /*
4314  * Compute IPv6 extension header length.
4315  */
4316 int
ip6_optlen(struct in6pcb * in6p)4317 ip6_optlen(struct in6pcb *in6p)
4318 {
4319 	int len;
4320 
4321 	if (!in6p->in6p_outputopts) {
4322 		return 0;
4323 	}
4324 
4325 	len = 0;
4326 #define elen(x)                                                         \
4327 	(((struct ip6_ext *)(x)) ?                                      \
4328 	(((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
4329 
4330 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
4331 	if (in6p->in6p_outputopts->ip6po_rthdr) {
4332 		/* dest1 is valid with rthdr only */
4333 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
4334 	}
4335 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
4336 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
4337 	return len;
4338 #undef elen
4339 }
4340 
4341 static int
4342 sysctl_reset_ip6_output_stats SYSCTL_HANDLER_ARGS
4343 {
4344 #pragma unused(arg1, arg2)
4345 	int error, i;
4346 
4347 	i = ip6_output_measure;
4348 	error = sysctl_handle_int(oidp, &i, 0, req);
4349 	if (error || req->newptr == USER_ADDR_NULL) {
4350 		goto done;
4351 	}
4352 	/* impose bounds */
4353 	if (i < 0 || i > 1) {
4354 		error = EINVAL;
4355 		goto done;
4356 	}
4357 	if (ip6_output_measure != i && i == 1) {
4358 		net_perf_initialize(&net_perf, ip6_output_measure_bins);
4359 	}
4360 	ip6_output_measure = i;
4361 done:
4362 	return error;
4363 }
4364 
4365 static int
4366 sysctl_ip6_output_measure_bins SYSCTL_HANDLER_ARGS
4367 {
4368 #pragma unused(arg1, arg2)
4369 	int error;
4370 	uint64_t i;
4371 
4372 	i = ip6_output_measure_bins;
4373 	error = sysctl_handle_quad(oidp, &i, 0, req);
4374 	if (error || req->newptr == USER_ADDR_NULL) {
4375 		goto done;
4376 	}
4377 	/* validate data */
4378 	if (!net_perf_validate_bins(i)) {
4379 		error = EINVAL;
4380 		goto done;
4381 	}
4382 	ip6_output_measure_bins = i;
4383 done:
4384 	return error;
4385 }
4386 
4387 static int
4388 sysctl_ip6_output_getperf SYSCTL_HANDLER_ARGS
4389 {
4390 #pragma unused(oidp, arg1, arg2)
4391 	if (req->oldptr == USER_ADDR_NULL) {
4392 		req->oldlen = (size_t)sizeof(struct ipstat);
4393 	}
4394 
4395 	return SYSCTL_OUT(req, &net_perf, MIN(sizeof(net_perf), req->oldlen));
4396 }
4397 
4398 void
ip6_output_setsrcifscope(struct mbuf * m,uint32_t src_idx,struct in6_ifaddr * ia6)4399 ip6_output_setsrcifscope(struct mbuf *m, uint32_t src_idx, struct in6_ifaddr *ia6)
4400 {
4401 	VERIFY(m->m_flags & M_PKTHDR);
4402 
4403 	m->m_pkthdr.pkt_ext_flags |= PKTF_EXT_OUTPUT_SCOPE;
4404 	if (ia6 != NULL) {
4405 		m->m_pkthdr.src_ifindex = ia6->ia_ifp->if_index;
4406 	} else {
4407 		m->m_pkthdr.src_ifindex = (uint16_t)src_idx;
4408 	}
4409 }
4410 
4411 void
ip6_output_setdstifscope(struct mbuf * m,uint32_t dst_idx,struct in6_ifaddr * ia6)4412 ip6_output_setdstifscope(struct mbuf *m, uint32_t dst_idx, struct in6_ifaddr *ia6)
4413 {
4414 	VERIFY(m->m_flags & M_PKTHDR);
4415 
4416 	m->m_pkthdr.pkt_ext_flags |= PKTF_EXT_OUTPUT_SCOPE;
4417 	if (ia6 != NULL) {
4418 		m->m_pkthdr.dst_ifindex = ia6->ia_ifp->if_index;
4419 	} else {
4420 		m->m_pkthdr.dst_ifindex = (uint16_t)dst_idx;
4421 	}
4422 }
4423 
4424 uint32_t
ip6_output_getsrcifscope(struct mbuf * m)4425 ip6_output_getsrcifscope(struct mbuf *m)
4426 {
4427 	VERIFY(m->m_flags & M_PKTHDR);
4428 	if (in6_embedded_scope_debug) {
4429 		VERIFY(m->m_pkthdr.pkt_ext_flags & PKTF_EXT_OUTPUT_SCOPE);
4430 		VERIFY((m->m_pkthdr.pkt_flags & PKTF_IFAINFO) == 0);
4431 	}
4432 
4433 	return m->m_pkthdr.src_ifindex;
4434 }
4435 
4436 uint32_t
ip6_output_getdstifscope(struct mbuf * m)4437 ip6_output_getdstifscope(struct mbuf *m)
4438 {
4439 	VERIFY(m->m_flags & M_PKTHDR);
4440 	if (in6_embedded_scope_debug) {
4441 		VERIFY(m->m_pkthdr.pkt_ext_flags & PKTF_EXT_OUTPUT_SCOPE);
4442 		VERIFY((m->m_pkthdr.pkt_flags & PKTF_IFAINFO) == 0);
4443 	}
4444 
4445 	return m->m_pkthdr.dst_ifindex;
4446 }
4447