xref: /freebsd/sys/netinet6/in6_pcb.c (revision 9b04aee86ccd58a6f3c936e0cd95bb9305a69848)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * Copyright (c) 2010-2011 Juniper Networks, Inc.
6  * All rights reserved.
7  *
8  * Portions of this software were developed by Robert N. M. Watson under
9  * contract to Juniper Networks, Inc.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the project nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	$KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $
36  */
37 
38 /*-
39  * Copyright (c) 1982, 1986, 1991, 1993
40  *	The Regents of the University of California.  All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  */
66 
67 #include <sys/cdefs.h>
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #include "opt_route.h"
72 #include "opt_rss.h"
73 
74 #include <sys/hash.h>
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/domain.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/smr.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sockio.h>
86 #include <sys/sysctl.h>
87 #include <sys/errno.h>
88 #include <sys/time.h>
89 #include <sys/priv.h>
90 #include <sys/proc.h>
91 #include <sys/jail.h>
92 
93 #include <vm/uma.h>
94 
95 #include <net/if.h>
96 #include <net/if_var.h>
97 #include <net/if_llatbl.h>
98 #include <net/if_types.h>
99 #include <net/route.h>
100 #include <net/route/nhop.h>
101 #include <net/vnet.h>
102 
103 #include <netinet/in.h>
104 #include <netinet/in_var.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/ip6.h>
107 #include <netinet/ip_var.h>
108 
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet/in_pcb.h>
112 #include <netinet/in_pcb_var.h>
113 #include <netinet6/in6_pcb.h>
114 #include <netinet6/in6_fib.h>
115 #include <netinet6/scope6_var.h>
116 
117 SYSCTL_DECL(_net_inet6);
118 SYSCTL_DECL(_net_inet6_ip6);
119 VNET_DEFINE_STATIC(int, connect_in6addr_wild) = 1;
120 #define	V_connect_in6addr_wild	VNET(connect_in6addr_wild)
121 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, connect_in6addr_wild,
122     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(connect_in6addr_wild), 0,
123     "Allow connecting to the unspecified address for connect(2)");
124 
125 int
126 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
127 {
128 	struct socket *so = inp->inp_socket;
129 	u_int16_t lport = 0;
130 	int error, lookupflags = 0;
131 #ifdef INVARIANTS
132 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
133 #endif
134 
135 	INP_WLOCK_ASSERT(inp);
136 	INP_HASH_WLOCK_ASSERT(pcbinfo);
137 
138 	error = prison_local_ip6(cred, laddr,
139 	    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
140 	if (error)
141 		return(error);
142 
143 	/* XXX: this is redundant when called from in6_pcbbind */
144 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0)
145 		lookupflags = INPLOOKUP_WILDCARD;
146 
147 	inp->inp_flags |= INP_ANONPORT;
148 
149 	error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
150 	if (error != 0)
151 		return (error);
152 
153 	inp->inp_lport = lport;
154 	if (in_pcbinshash(inp) != 0) {
155 		inp->in6p_laddr = in6addr_any;
156 		inp->inp_lport = 0;
157 		return (EAGAIN);
158 	}
159 
160 	return (0);
161 }
162 
163 int
164 in6_pcbbind(struct inpcb *inp, struct sockaddr_in6 *sin6, struct ucred *cred)
165 {
166 	struct socket *so = inp->inp_socket;
167 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
168 	u_short	lport = 0;
169 	int error, lookupflags = 0;
170 	int reuseport = (so->so_options & SO_REUSEPORT);
171 
172 	/*
173 	 * XXX: Maybe we could let SO_REUSEPORT_LB set SO_REUSEPORT bit here
174 	 * so that we don't have to add to the (already messy) code below.
175 	 */
176 	int reuseport_lb = (so->so_options & SO_REUSEPORT_LB);
177 
178 	INP_WLOCK_ASSERT(inp);
179 	INP_HASH_WLOCK_ASSERT(pcbinfo);
180 
181 	if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
182 		return (EINVAL);
183 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0)
184 		lookupflags = INPLOOKUP_WILDCARD;
185 	if (sin6 == NULL) {
186 		if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
187 		    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
188 			return (error);
189 	} else {
190 		KASSERT(sin6->sin6_family == AF_INET6,
191 		    ("%s: invalid address family for %p", __func__, sin6));
192 		KASSERT(sin6->sin6_len == sizeof(*sin6),
193 		    ("%s: invalid address length for %p", __func__, sin6));
194 
195 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
196 			return(error);
197 
198 		if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
199 		    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
200 			return (error);
201 
202 		lport = sin6->sin6_port;
203 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
204 			/*
205 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
206 			 * allow compepte duplication of binding if
207 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
208 			 * and a multicast address is bound on both
209 			 * new and duplicated sockets.
210 			 */
211 			if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
212 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
213 			/*
214 			 * XXX: How to deal with SO_REUSEPORT_LB here?
215 			 * Treat same as SO_REUSEPORT for now.
216 			 */
217 			if ((so->so_options &
218 			    (SO_REUSEADDR|SO_REUSEPORT_LB)) != 0)
219 				reuseport_lb = SO_REUSEADDR|SO_REUSEPORT_LB;
220 		} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
221 			struct epoch_tracker et;
222 			struct ifaddr *ifa;
223 
224 			sin6->sin6_port = 0;		/* yech... */
225 			NET_EPOCH_ENTER(et);
226 			if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) ==
227 			    NULL &&
228 			    (inp->inp_flags & INP_BINDANY) == 0) {
229 				NET_EPOCH_EXIT(et);
230 				return (EADDRNOTAVAIL);
231 			}
232 
233 			/*
234 			 * XXX: bind to an anycast address might accidentally
235 			 * cause sending a packet with anycast source address.
236 			 * We should allow to bind to a deprecated address, since
237 			 * the application dares to use it.
238 			 */
239 			if (ifa != NULL &&
240 			    ((struct in6_ifaddr *)ifa)->ia6_flags &
241 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
242 				NET_EPOCH_EXIT(et);
243 				return (EADDRNOTAVAIL);
244 			}
245 			NET_EPOCH_EXIT(et);
246 		}
247 		if (lport) {
248 			struct inpcb *t;
249 
250 			/* GROSS */
251 			if (ntohs(lport) <= V_ipport_reservedhigh &&
252 			    ntohs(lport) >= V_ipport_reservedlow &&
253 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT))
254 				return (EACCES);
255 			if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
256 			    priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT) != 0) {
257 				t = in6_pcblookup_local(pcbinfo,
258 				    &sin6->sin6_addr, lport,
259 				    INPLOOKUP_WILDCARD, cred);
260 				if (t != NULL &&
261 				    (so->so_type != SOCK_STREAM ||
262 				     IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
263 				    (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
264 				     !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
265 				     (t->inp_socket->so_options & SO_REUSEPORT) ||
266 				     (t->inp_socket->so_options & SO_REUSEPORT_LB) == 0) &&
267 				    (inp->inp_cred->cr_uid !=
268 				     t->inp_cred->cr_uid))
269 					return (EADDRINUSE);
270 
271 #ifdef INET
272 				if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
273 				    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
274 					struct sockaddr_in sin;
275 
276 					in6_sin6_2_sin(&sin, sin6);
277 					t = in_pcblookup_local(pcbinfo,
278 					    sin.sin_addr, lport,
279 					    INPLOOKUP_WILDCARD, cred);
280 					if (t != NULL &&
281 					    (so->so_type != SOCK_STREAM ||
282 					     ntohl(t->inp_faddr.s_addr) ==
283 					      INADDR_ANY) &&
284 					    (inp->inp_cred->cr_uid !=
285 					     t->inp_cred->cr_uid))
286 						return (EADDRINUSE);
287 				}
288 #endif
289 			}
290 			t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
291 			    lport, lookupflags, cred);
292 			if (t && (reuseport & t->inp_socket->so_options) == 0 &&
293 			    (reuseport_lb & t->inp_socket->so_options) == 0) {
294 				return (EADDRINUSE);
295 			}
296 #ifdef INET
297 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
298 			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
299 				struct sockaddr_in sin;
300 
301 				in6_sin6_2_sin(&sin, sin6);
302 				t = in_pcblookup_local(pcbinfo, sin.sin_addr,
303 				   lport, lookupflags, cred);
304 				if (t &&
305 				    (reuseport & t->inp_socket->so_options) == 0 &&
306 				    (reuseport_lb & t->inp_socket->so_options) == 0 &&
307 				    (ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
308 				        (t->inp_vflag & INP_IPV6PROTO) != 0)) {
309 					return (EADDRINUSE);
310 				}
311 			}
312 #endif
313 		}
314 		inp->in6p_laddr = sin6->sin6_addr;
315 	}
316 	if (lport == 0) {
317 		if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
318 			/* Undo an address bind that may have occurred. */
319 			inp->in6p_laddr = in6addr_any;
320 			return (error);
321 		}
322 	} else {
323 		inp->inp_lport = lport;
324 		if (in_pcbinshash(inp) != 0) {
325 			inp->in6p_laddr = in6addr_any;
326 			inp->inp_lport = 0;
327 			return (EAGAIN);
328 		}
329 	}
330 	return (0);
331 }
332 
333 /*
334  *   Transform old in6_pcbconnect() into an inner subroutine for new
335  *   in6_pcbconnect(): Do some validity-checking on the remote
336  *   address (in mbuf 'nam') and then determine local host address
337  *   (i.e., which interface) to use to access that remote host.
338  *
339  *   This preserves definition of in6_pcbconnect(), while supporting a
340  *   slightly different version for T/TCP.  (This is more than
341  *   a bit of a kludge, but cleaning up the internal interfaces would
342  *   have forced minor changes in every protocol).
343  */
344 static int
345 in6_pcbladdr(struct inpcb *inp, struct sockaddr_in6 *sin6,
346     struct in6_addr *plocal_addr6, bool sas_required)
347 {
348 	int error = 0;
349 	int scope_ambiguous = 0;
350 	struct in6_addr in6a;
351 
352 	NET_EPOCH_ASSERT();
353 	INP_WLOCK_ASSERT(inp);
354 	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);	/* XXXRW: why? */
355 
356 	if (sin6->sin6_port == 0)
357 		return (EADDRNOTAVAIL);
358 
359 	if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
360 		scope_ambiguous = 1;
361 	if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
362 		return(error);
363 
364 	if (V_connect_in6addr_wild && !CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) {
365 		/*
366 		 * If the destination address is UNSPECIFIED addr,
367 		 * use the loopback addr, e.g ::1.
368 		 */
369 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
370 			sin6->sin6_addr = in6addr_loopback;
371 	} else if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
372 		return (ENETUNREACH);
373 	}
374 
375 	if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
376 		return (error);
377 
378 	if (sas_required) {
379 		error = in6_selectsrc_socket(sin6, inp->in6p_outputopts,
380 		    inp, inp->inp_cred, scope_ambiguous, &in6a, NULL);
381 		if (error)
382 			return (error);
383 	} else {
384 		/*
385 		 * Source address selection isn't required when syncache
386 		 * has already established connection and both source and
387 		 * destination addresses was chosen.
388 		 *
389 		 * This also includes the case when fwd_tag was used to
390 		 * select source address in tcp_input().
391 		 */
392 		in6a = inp->in6p_laddr;
393 	}
394 
395 	if (IN6_IS_ADDR_UNSPECIFIED(&in6a))
396 		return (EHOSTUNREACH);
397 	/*
398 	 * Do not update this earlier, in case we return with an error.
399 	 *
400 	 * XXX: this in6_selectsrc_socket result might replace the bound local
401 	 * address with the address specified by setsockopt(IPV6_PKTINFO).
402 	 * Is it the intended behavior?
403 	 */
404 	*plocal_addr6 = in6a;
405 
406 	/*
407 	 * Don't do pcblookup call here; return interface in
408 	 * plocal_addr6
409 	 * and exit to caller, that will do the lookup.
410 	 */
411 
412 	return (0);
413 }
414 
415 /*
416  * Outer subroutine:
417  * Connect from a socket to a specified address.
418  * Both address and port must be specified in argument sin.
419  * If don't have a local address for this socket yet,
420  * then pick one.
421  */
422 int
423 in6_pcbconnect(struct inpcb *inp, struct sockaddr_in6 *sin6, struct ucred *cred,
424     bool sas_required)
425 {
426 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
427 	struct sockaddr_in6 laddr6;
428 	int error;
429 
430 	NET_EPOCH_ASSERT();
431 	INP_WLOCK_ASSERT(inp);
432 	INP_HASH_WLOCK_ASSERT(pcbinfo);
433 	KASSERT(sin6->sin6_family == AF_INET6,
434 	    ("%s: invalid address family for %p", __func__, sin6));
435 	KASSERT(sin6->sin6_len == sizeof(*sin6),
436 	    ("%s: invalid address length for %p", __func__, sin6));
437 	KASSERT(IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr),
438 	    ("%s: inp is already connected", __func__));
439 
440 	bzero(&laddr6, sizeof(laddr6));
441 	laddr6.sin6_family = AF_INET6;
442 
443 #ifdef ROUTE_MPATH
444 	if (CALC_FLOWID_OUTBOUND) {
445 		uint32_t hash_type, hash_val;
446 
447 		hash_val = fib6_calc_software_hash(&inp->in6p_laddr,
448 		    &sin6->sin6_addr, 0, sin6->sin6_port,
449 		    inp->inp_socket->so_proto->pr_protocol, &hash_type);
450 		inp->inp_flowid = hash_val;
451 		inp->inp_flowtype = hash_type;
452 	}
453 #endif
454 	/*
455 	 * Call inner routine, to assign local interface address.
456 	 * in6_pcbladdr() may automatically fill in sin6_scope_id.
457 	 */
458 	if ((error = in6_pcbladdr(inp, sin6, &laddr6.sin6_addr,
459 	    sas_required)) != 0)
460 		return (error);
461 
462 	if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
463 	    sin6->sin6_port, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
464 	    &laddr6.sin6_addr : &inp->in6p_laddr, inp->inp_lport, 0,
465 	    M_NODOM) != NULL)
466 		return (EADDRINUSE);
467 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
468 		if (inp->inp_lport == 0) {
469 			error = in_pcb_lport_dest(inp,
470 			    (struct sockaddr *) &laddr6, &inp->inp_lport,
471 			    (struct sockaddr *) sin6, sin6->sin6_port, cred,
472 			    INPLOOKUP_WILDCARD);
473 			if (error)
474 				return (error);
475 		}
476 		inp->in6p_laddr = laddr6.sin6_addr;
477 	}
478 	inp->in6p_faddr = sin6->sin6_addr;
479 	inp->inp_fport = sin6->sin6_port;
480 	/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
481 	inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
482 	if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
483 		inp->inp_flow |=
484 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
485 
486 	if ((inp->inp_flags & INP_INHASHLIST) != 0) {
487 		in_pcbrehash(inp);
488 	} else {
489 		in_pcbinshash(inp);
490 	}
491 
492 	return (0);
493 }
494 
495 void
496 in6_pcbdisconnect(struct inpcb *inp)
497 {
498 
499 	INP_WLOCK_ASSERT(inp);
500 	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
501 	KASSERT(inp->inp_smr == SMR_SEQ_INVALID,
502 	    ("%s: inp %p was already disconnected", __func__, inp));
503 
504 	in_pcbremhash_locked(inp);
505 
506 	/* See the comment in in_pcbinshash(). */
507 	inp->inp_smr = smr_advance(inp->inp_pcbinfo->ipi_smr);
508 
509 	/* XXX-MJ torn writes are visible to SMR lookup */
510 	memset(&inp->in6p_laddr, 0, sizeof(inp->in6p_laddr));
511 	memset(&inp->in6p_faddr, 0, sizeof(inp->in6p_faddr));
512 	inp->inp_fport = 0;
513 	/* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
514 	inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
515 }
516 
517 int
518 in6_getsockaddr(struct socket *so, struct sockaddr *sa)
519 {
520 	struct inpcb *inp;
521 
522 	inp = sotoinpcb(so);
523 	KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
524 
525 	*(struct sockaddr_in6 *)sa = (struct sockaddr_in6 ){
526 		.sin6_len = sizeof(struct sockaddr_in6),
527 		.sin6_family = AF_INET6,
528 		.sin6_port = inp->inp_lport,
529 		.sin6_addr = inp->in6p_laddr,
530 	};
531 	/* XXX: should catch errors */
532 	(void)sa6_recoverscope((struct sockaddr_in6 *)sa);
533 
534 	return (0);
535 }
536 
537 int
538 in6_getpeeraddr(struct socket *so, struct sockaddr *sa)
539 {
540 	struct inpcb *inp;
541 
542 	inp = sotoinpcb(so);
543 	KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
544 
545 	*(struct sockaddr_in6 *)sa = (struct sockaddr_in6 ){
546 		.sin6_len = sizeof(struct sockaddr_in6),
547 		.sin6_family = AF_INET6,
548 		.sin6_port = inp->inp_fport,
549 		.sin6_addr = inp->in6p_faddr,
550 	};
551 	/* XXX: should catch errors */
552 	(void)sa6_recoverscope((struct sockaddr_in6 *)sa);
553 
554 	return (0);
555 }
556 
557 int
558 in6_mapped_sockaddr(struct socket *so, struct sockaddr *sa)
559 {
560 	int	error;
561 #ifdef INET
562 	struct	inpcb *inp;
563 
564 	inp = sotoinpcb(so);
565 	KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
566 
567 	if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
568 		struct sockaddr_in sin;
569 
570 		error = in_getsockaddr(so, (struct sockaddr *)&sin);
571 		if (error == 0)
572 			in6_sin_2_v4mapsin6(&sin, (struct sockaddr_in6 *)sa);
573 	} else
574 #endif
575 	{
576 		/* scope issues will be handled in in6_getsockaddr(). */
577 		error = in6_getsockaddr(so, sa);
578 	}
579 
580 	return error;
581 }
582 
583 int
584 in6_mapped_peeraddr(struct socket *so, struct sockaddr *sa)
585 {
586 	int	error;
587 #ifdef INET
588 	struct	inpcb *inp;
589 
590 	inp = sotoinpcb(so);
591 	KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
592 
593 	if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
594 		struct sockaddr_in sin;
595 
596 		error = in_getpeeraddr(so, (struct sockaddr *)&sin);
597 		if (error == 0)
598 			in6_sin_2_v4mapsin6(&sin, (struct sockaddr_in6 *)sa);
599 	} else
600 #endif
601 	{
602 		/* scope issues will be handled in in6_getpeeraddr(). */
603 		error = in6_getpeeraddr(so, sa);
604 	}
605 
606 	return error;
607 }
608 
609 /*
610  * Pass some notification to all connections of a protocol
611  * associated with address dst.  The local address and/or port numbers
612  * may be specified to limit the search.  The "usual action" will be
613  * taken, depending on the ctlinput cmd.  The caller must filter any
614  * cmds that are uninteresting (e.g., no error in the map).
615  * Call the protocol specific routine (if any) to report
616  * any errors for each matching socket.
617  */
618 static bool
619 inp_match6(const struct inpcb *inp, void *v __unused)
620 {
621 
622 	return ((inp->inp_vflag & INP_IPV6) != 0);
623 }
624 
625 void
626 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr_in6 *sa6_dst,
627     u_int fport_arg, const struct sockaddr_in6 *src, u_int lport_arg,
628     int errno, void *cmdarg,
629     struct inpcb *(*notify)(struct inpcb *, int))
630 {
631 	struct inpcb_iterator inpi = INP_ITERATOR(pcbinfo, INPLOOKUP_WLOCKPCB,
632 	    inp_match6, NULL);
633 	struct inpcb *inp;
634 	struct sockaddr_in6 sa6_src;
635 	u_short	fport = fport_arg, lport = lport_arg;
636 	u_int32_t flowinfo;
637 
638 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
639 		return;
640 
641 	/*
642 	 * note that src can be NULL when we get notify by local fragmentation.
643 	 */
644 	sa6_src = (src == NULL) ? sa6_any : *src;
645 	flowinfo = sa6_src.sin6_flowinfo;
646 
647 	while ((inp = inp_next(&inpi)) != NULL) {
648 		INP_WLOCK_ASSERT(inp);
649 		/*
650 		 * If the error designates a new path MTU for a destination
651 		 * and the application (associated with this socket) wanted to
652 		 * know the value, notify.
653 		 * XXX: should we avoid to notify the value to TCP sockets?
654 		 */
655 		if (errno == EMSGSIZE && cmdarg != NULL)
656 			ip6_notify_pmtu(inp, sa6_dst, *(uint32_t *)cmdarg);
657 
658 		/*
659 		 * Detect if we should notify the error. If no source and
660 		 * destination ports are specified, but non-zero flowinfo and
661 		 * local address match, notify the error. This is the case
662 		 * when the error is delivered with an encrypted buffer
663 		 * by ESP. Otherwise, just compare addresses and ports
664 		 * as usual.
665 		 */
666 		if (lport == 0 && fport == 0 && flowinfo &&
667 		    inp->inp_socket != NULL &&
668 		    flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
669 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
670 			goto do_notify;
671 		else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
672 					     &sa6_dst->sin6_addr) ||
673 			 inp->inp_socket == 0 ||
674 			 (lport && inp->inp_lport != lport) ||
675 			 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
676 			  !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
677 					      &sa6_src.sin6_addr)) ||
678 			 (fport && inp->inp_fport != fport)) {
679 			continue;
680 		}
681 
682 	  do_notify:
683 		if (notify)
684 			(*notify)(inp, errno);
685 	}
686 }
687 
688 /*
689  * Lookup a PCB based on the local address and port.  Caller must hold the
690  * hash lock.  No inpcb locks or references are acquired.
691  */
692 struct inpcb *
693 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
694     u_short lport, int lookupflags, struct ucred *cred)
695 {
696 	struct inpcb *inp;
697 	int matchwild = 3, wildcard;
698 
699 	KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
700 	    ("%s: invalid lookup flags %d", __func__, lookupflags));
701 
702 	INP_HASH_LOCK_ASSERT(pcbinfo);
703 
704 	if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
705 		struct inpcbhead *head;
706 		/*
707 		 * Look for an unconnected (wildcard foreign addr) PCB that
708 		 * matches the local address and port we're looking for.
709 		 */
710 		head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
711 		    pcbinfo->ipi_hashmask)];
712 		CK_LIST_FOREACH(inp, head, inp_hash_wild) {
713 			/* XXX inp locking */
714 			if ((inp->inp_vflag & INP_IPV6) == 0)
715 				continue;
716 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
717 			    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
718 			    inp->inp_lport == lport) {
719 				/* Found. */
720 				if (prison_equal_ip6(cred->cr_prison,
721 				    inp->inp_cred->cr_prison))
722 					return (inp);
723 			}
724 		}
725 		/*
726 		 * Not found.
727 		 */
728 		return (NULL);
729 	} else {
730 		struct inpcbporthead *porthash;
731 		struct inpcbport *phd;
732 		struct inpcb *match = NULL;
733 		/*
734 		 * Best fit PCB lookup.
735 		 *
736 		 * First see if this local port is in use by looking on the
737 		 * port hash list.
738 		 */
739 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
740 		    pcbinfo->ipi_porthashmask)];
741 		CK_LIST_FOREACH(phd, porthash, phd_hash) {
742 			if (phd->phd_port == lport)
743 				break;
744 		}
745 		if (phd != NULL) {
746 			/*
747 			 * Port is in use by one or more PCBs. Look for best
748 			 * fit.
749 			 */
750 			CK_LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
751 				wildcard = 0;
752 				if (!prison_equal_ip6(cred->cr_prison,
753 				    inp->inp_cred->cr_prison))
754 					continue;
755 				/* XXX inp locking */
756 				if ((inp->inp_vflag & INP_IPV6) == 0)
757 					continue;
758 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
759 					wildcard++;
760 				if (!IN6_IS_ADDR_UNSPECIFIED(
761 					&inp->in6p_laddr)) {
762 					if (IN6_IS_ADDR_UNSPECIFIED(laddr))
763 						wildcard++;
764 					else if (!IN6_ARE_ADDR_EQUAL(
765 					    &inp->in6p_laddr, laddr))
766 						continue;
767 				} else {
768 					if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
769 						wildcard++;
770 				}
771 				if (wildcard < matchwild) {
772 					match = inp;
773 					matchwild = wildcard;
774 					if (matchwild == 0)
775 						break;
776 				}
777 			}
778 		}
779 		return (match);
780 	}
781 }
782 
783 static bool
784 in6_multi_match(const struct inpcb *inp, void *v __unused)
785 {
786 
787 	if ((inp->inp_vflag & INP_IPV6) && inp->in6p_moptions != NULL)
788 		return (true);
789 	else
790 		return (false);
791 }
792 
793 void
794 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
795 {
796 	struct inpcb_iterator inpi = INP_ITERATOR(pcbinfo, INPLOOKUP_RLOCKPCB,
797 	    in6_multi_match, NULL);
798 	struct inpcb *inp;
799 	struct in6_multi *inm;
800 	struct in6_mfilter *imf;
801 	struct ip6_moptions *im6o;
802 
803 	IN6_MULTI_LOCK_ASSERT();
804 
805 	while ((inp = inp_next(&inpi)) != NULL) {
806 		INP_RLOCK_ASSERT(inp);
807 
808 		im6o = inp->in6p_moptions;
809 		/*
810 		 * Unselect the outgoing ifp for multicast if it
811 		 * is being detached.
812 		 */
813 		if (im6o->im6o_multicast_ifp == ifp)
814 			im6o->im6o_multicast_ifp = NULL;
815 		/*
816 		 * Drop multicast group membership if we joined
817 		 * through the interface being detached.
818 		 */
819 restart:
820 		IP6_MFILTER_FOREACH(imf, &im6o->im6o_head) {
821 			if ((inm = imf->im6f_in6m) == NULL)
822 				continue;
823 			if (inm->in6m_ifp != ifp)
824 				continue;
825 			ip6_mfilter_remove(&im6o->im6o_head, imf);
826 			in6_leavegroup_locked(inm, NULL);
827 			ip6_mfilter_free(imf);
828 			goto restart;
829 		}
830 	}
831 }
832 
833 /*
834  * Check for alternatives when higher level complains
835  * about service problems.  For now, invalidate cached
836  * routing information.  If the route was created dynamically
837  * (by a redirect), time to try a default gateway again.
838  */
839 void
840 in6_losing(struct inpcb *inp)
841 {
842 
843 	RO_INVALIDATE_CACHE(&inp->inp_route6);
844 }
845 
846 /*
847  * After a routing change, flush old routing
848  * and allocate a (hopefully) better one.
849  */
850 struct inpcb *
851 in6_rtchange(struct inpcb *inp, int errno __unused)
852 {
853 
854 	RO_INVALIDATE_CACHE(&inp->inp_route6);
855 	return inp;
856 }
857 
858 static bool
859 in6_pcblookup_lb_numa_match(const struct inpcblbgroup *grp, int domain)
860 {
861 	return (domain == M_NODOM || domain == grp->il_numa_domain);
862 }
863 
864 static struct inpcb *
865 in6_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo,
866     const struct in6_addr *faddr, uint16_t fport, const struct in6_addr *laddr,
867     uint16_t lport, uint8_t domain)
868 {
869 	const struct inpcblbgrouphead *hdr;
870 	struct inpcblbgroup *grp;
871 	struct inpcblbgroup *jail_exact, *jail_wild, *local_exact, *local_wild;
872 
873 	INP_HASH_LOCK_ASSERT(pcbinfo);
874 
875 	hdr = &pcbinfo->ipi_lbgrouphashbase[
876 	    INP_PCBPORTHASH(lport, pcbinfo->ipi_lbgrouphashmask)];
877 
878 	/*
879 	 * Search for an LB group match based on the following criteria:
880 	 * - prefer jailed groups to non-jailed groups
881 	 * - prefer exact source address matches to wildcard matches
882 	 * - prefer groups bound to the specified NUMA domain
883 	 */
884 	jail_exact = jail_wild = local_exact = local_wild = NULL;
885 	CK_LIST_FOREACH(grp, hdr, il_list) {
886 		bool injail;
887 
888 #ifdef INET
889 		if (!(grp->il_vflag & INP_IPV6))
890 			continue;
891 #endif
892 		if (grp->il_lport != lport)
893 			continue;
894 
895 		injail = prison_flag(grp->il_cred, PR_IP6) != 0;
896 		if (injail && prison_check_ip6_locked(grp->il_cred->cr_prison,
897 		    laddr) != 0)
898 			continue;
899 
900 		if (IN6_ARE_ADDR_EQUAL(&grp->il6_laddr, laddr)) {
901 			if (injail) {
902 				jail_exact = grp;
903 				if (in6_pcblookup_lb_numa_match(grp, domain))
904 					/* This is a perfect match. */
905 					goto out;
906 			} else if (local_exact == NULL ||
907 			    in6_pcblookup_lb_numa_match(grp, domain)) {
908 				local_exact = grp;
909 			}
910 		} else if (IN6_IS_ADDR_UNSPECIFIED(&grp->il6_laddr)) {
911 			if (injail) {
912 				if (jail_wild == NULL ||
913 				    in6_pcblookup_lb_numa_match(grp, domain))
914 					jail_wild = grp;
915 			} else if (local_wild == NULL ||
916 			    in6_pcblookup_lb_numa_match(grp, domain)) {
917 				local_wild = grp;
918 			}
919 		}
920 	}
921 
922 	if (jail_exact != NULL)
923 		grp = jail_exact;
924 	else if (jail_wild != NULL)
925 		grp = jail_wild;
926 	else if (local_exact != NULL)
927 		grp = local_exact;
928 	else
929 		grp = local_wild;
930 	if (grp == NULL)
931 		return (NULL);
932 out:
933 	return (grp->il_inp[INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport) %
934 	    grp->il_inpcnt]);
935 }
936 
937 static bool
938 in6_pcblookup_exact_match(const struct inpcb *inp, const struct in6_addr *faddr,
939     u_short fport, const struct in6_addr *laddr, u_short lport)
940 {
941 	/* XXX inp locking */
942 	if ((inp->inp_vflag & INP_IPV6) == 0)
943 		return (false);
944 	if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
945 	    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
946 	    inp->inp_fport == fport && inp->inp_lport == lport)
947 		return (true);
948 	return (false);
949 }
950 
951 static struct inpcb *
952 in6_pcblookup_hash_exact(struct inpcbinfo *pcbinfo,
953     const struct in6_addr *faddr, u_short fport,
954     const struct in6_addr *laddr, u_short lport)
955 {
956 	struct inpcbhead *head;
957 	struct inpcb *inp;
958 
959 	INP_HASH_LOCK_ASSERT(pcbinfo);
960 
961 	/*
962 	 * First look for an exact match.
963 	 */
964 	head = &pcbinfo->ipi_hash_exact[INP6_PCBHASH(faddr, lport, fport,
965 	    pcbinfo->ipi_hashmask)];
966 	CK_LIST_FOREACH(inp, head, inp_hash_exact) {
967 		if (in6_pcblookup_exact_match(inp, faddr, fport, laddr, lport))
968 			return (inp);
969 	}
970 	return (NULL);
971 }
972 
973 typedef enum {
974 	INPLOOKUP_MATCH_NONE = 0,
975 	INPLOOKUP_MATCH_WILD = 1,
976 	INPLOOKUP_MATCH_LADDR = 2,
977 } inp_lookup_match_t;
978 
979 static inp_lookup_match_t
980 in6_pcblookup_wild_match(const struct inpcb *inp, const struct in6_addr *laddr,
981     u_short lport)
982 {
983 	/* XXX inp locking */
984 	if ((inp->inp_vflag & INP_IPV6) == 0)
985 		return (INPLOOKUP_MATCH_NONE);
986 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
987 	    inp->inp_lport != lport)
988 		return (INPLOOKUP_MATCH_NONE);
989 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
990 		return (INPLOOKUP_MATCH_WILD);
991 	if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr))
992 		return (INPLOOKUP_MATCH_LADDR);
993 	return (INPLOOKUP_MATCH_NONE);
994 }
995 
996 #define	INP_LOOKUP_AGAIN	((struct inpcb *)(uintptr_t)-1)
997 
998 static struct inpcb *
999 in6_pcblookup_hash_wild_smr(struct inpcbinfo *pcbinfo,
1000     const struct in6_addr *faddr, u_short fport, const struct in6_addr *laddr,
1001     u_short lport, const inp_lookup_t lockflags)
1002 {
1003 	struct inpcbhead *head;
1004 	struct inpcb *inp;
1005 
1006 	KASSERT(SMR_ENTERED(pcbinfo->ipi_smr),
1007 	    ("%s: not in SMR read section", __func__));
1008 
1009 	head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
1010 	    pcbinfo->ipi_hashmask)];
1011 	CK_LIST_FOREACH(inp, head, inp_hash_wild) {
1012 		inp_lookup_match_t match;
1013 
1014 		match = in6_pcblookup_wild_match(inp, laddr, lport);
1015 		if (match == INPLOOKUP_MATCH_NONE)
1016 			continue;
1017 
1018 		if (__predict_true(inp_smr_lock(inp, lockflags))) {
1019 			match = in6_pcblookup_wild_match(inp, laddr, lport);
1020 			if (match != INPLOOKUP_MATCH_NONE &&
1021 			    prison_check_ip6_locked(inp->inp_cred->cr_prison,
1022 			    laddr) == 0)
1023 				return (inp);
1024 			inp_unlock(inp, lockflags);
1025 		}
1026 
1027 		/*
1028 		 * The matching socket disappeared out from under us.  Fall back
1029 		 * to a serialized lookup.
1030 		 */
1031 		return (INP_LOOKUP_AGAIN);
1032 	}
1033 	return (NULL);
1034 }
1035 
1036 static struct inpcb *
1037 in6_pcblookup_hash_wild_locked(struct inpcbinfo *pcbinfo,
1038     const struct in6_addr *faddr, u_short fport, const struct in6_addr *laddr,
1039     u_short lport)
1040 {
1041 	struct inpcbhead *head;
1042 	struct inpcb *inp, *jail_wild, *local_exact, *local_wild;
1043 
1044 	INP_HASH_LOCK_ASSERT(pcbinfo);
1045 
1046 	/*
1047 	 * Order of socket selection - we always prefer jails.
1048 	 *      1. jailed, non-wild.
1049 	 *      2. jailed, wild.
1050 	 *      3. non-jailed, non-wild.
1051 	 *      4. non-jailed, wild.
1052 	 */
1053 	head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
1054 	    pcbinfo->ipi_hashmask)];
1055 	local_wild = local_exact = jail_wild = NULL;
1056 	CK_LIST_FOREACH(inp, head, inp_hash_wild) {
1057 		inp_lookup_match_t match;
1058 		bool injail;
1059 
1060 		match = in6_pcblookup_wild_match(inp, laddr, lport);
1061 		if (match == INPLOOKUP_MATCH_NONE)
1062 			continue;
1063 
1064 		injail = prison_flag(inp->inp_cred, PR_IP6) != 0;
1065 		if (injail) {
1066 			if (prison_check_ip6_locked(
1067 			    inp->inp_cred->cr_prison, laddr) != 0)
1068 				continue;
1069 		} else {
1070 			if (local_exact != NULL)
1071 				continue;
1072 		}
1073 
1074 		if (match == INPLOOKUP_MATCH_LADDR) {
1075 			if (injail)
1076 				return (inp);
1077 			else
1078 				local_exact = inp;
1079 		} else {
1080 			if (injail)
1081 				jail_wild = inp;
1082 			else
1083 				local_wild = inp;
1084 		}
1085 	}
1086 
1087 	if (jail_wild != NULL)
1088 		return (jail_wild);
1089 	if (local_exact != NULL)
1090 		return (local_exact);
1091 	if (local_wild != NULL)
1092 		return (local_wild);
1093 	return (NULL);
1094 }
1095 
1096 struct inpcb *
1097 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo,
1098     const struct in6_addr *faddr, u_int fport_arg,
1099     const struct in6_addr *laddr, u_int lport_arg,
1100     int lookupflags, uint8_t numa_domain)
1101 {
1102 	struct inpcb *inp;
1103 	u_short fport = fport_arg, lport = lport_arg;
1104 
1105 	KASSERT((lookupflags & ~INPLOOKUP_WILDCARD) == 0,
1106 	    ("%s: invalid lookup flags %d", __func__, lookupflags));
1107 	KASSERT(!IN6_IS_ADDR_UNSPECIFIED(faddr),
1108 	    ("%s: invalid foreign address", __func__));
1109 	KASSERT(!IN6_IS_ADDR_UNSPECIFIED(laddr),
1110 	    ("%s: invalid local address", __func__));
1111 	INP_HASH_LOCK_ASSERT(pcbinfo);
1112 
1113 	inp = in6_pcblookup_hash_exact(pcbinfo, faddr, fport, laddr, lport);
1114 	if (inp != NULL)
1115 		return (inp);
1116 
1117 	if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1118 		inp = in6_pcblookup_lbgroup(pcbinfo, faddr, fport, laddr,
1119 		    lport, numa_domain);
1120 		if (inp == NULL) {
1121 			inp = in6_pcblookup_hash_wild_locked(pcbinfo, faddr,
1122 			    fport, laddr, lport);
1123 		}
1124 	}
1125 	return (inp);
1126 }
1127 
1128 static struct inpcb *
1129 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1130     u_int fport, const struct in6_addr *laddr, u_int lport, int lookupflags,
1131     uint8_t numa_domain)
1132 {
1133 	struct inpcb *inp;
1134 	const inp_lookup_t lockflags = lookupflags & INPLOOKUP_LOCKMASK;
1135 
1136 	KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1137 	    ("%s: LOCKPCB not set", __func__));
1138 
1139 	INP_HASH_WLOCK(pcbinfo);
1140 	inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1141 	    lookupflags & ~INPLOOKUP_LOCKMASK, numa_domain);
1142 	if (inp != NULL && !inp_trylock(inp, lockflags)) {
1143 		in_pcbref(inp);
1144 		INP_HASH_WUNLOCK(pcbinfo);
1145 		inp_lock(inp, lockflags);
1146 		if (in_pcbrele(inp, lockflags))
1147 			/* XXX-MJ or retry until we get a negative match? */
1148 			inp = NULL;
1149 	} else {
1150 		INP_HASH_WUNLOCK(pcbinfo);
1151 	}
1152 	return (inp);
1153 }
1154 
1155 static struct inpcb *
1156 in6_pcblookup_hash_smr(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1157     u_int fport_arg, struct in6_addr *laddr, u_int lport_arg, int lookupflags,
1158     uint8_t numa_domain)
1159 {
1160 	struct inpcb *inp;
1161 	const inp_lookup_t lockflags = lookupflags & INPLOOKUP_LOCKMASK;
1162 	const u_short fport = fport_arg, lport = lport_arg;
1163 
1164 	KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1165 	    ("%s: invalid lookup flags %d", __func__, lookupflags));
1166 	KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1167 	    ("%s: LOCKPCB not set", __func__));
1168 
1169 	smr_enter(pcbinfo->ipi_smr);
1170 	inp = in6_pcblookup_hash_exact(pcbinfo, faddr, fport, laddr, lport);
1171 	if (inp != NULL) {
1172 		if (__predict_true(inp_smr_lock(inp, lockflags))) {
1173 			if (__predict_true(in6_pcblookup_exact_match(inp,
1174 			    faddr, fport, laddr, lport)))
1175 				return (inp);
1176 			inp_unlock(inp, lockflags);
1177 		}
1178 		/*
1179 		 * We failed to lock the inpcb, or its connection state changed
1180 		 * out from under us.  Fall back to a precise search.
1181 		 */
1182 		return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1183 		    lookupflags, numa_domain));
1184 	}
1185 
1186 	if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1187 		inp = in6_pcblookup_lbgroup(pcbinfo, faddr, fport,
1188 		    laddr, lport, numa_domain);
1189 		if (inp != NULL) {
1190 			if (__predict_true(inp_smr_lock(inp, lockflags))) {
1191 				if (__predict_true(in6_pcblookup_wild_match(inp,
1192 				    laddr, lport) != INPLOOKUP_MATCH_NONE))
1193 					return (inp);
1194 				inp_unlock(inp, lockflags);
1195 			}
1196 			inp = INP_LOOKUP_AGAIN;
1197 		} else {
1198 			inp = in6_pcblookup_hash_wild_smr(pcbinfo, faddr, fport,
1199 			    laddr, lport, lockflags);
1200 		}
1201 		if (inp == INP_LOOKUP_AGAIN) {
1202 			return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr,
1203 			    lport, lookupflags, numa_domain));
1204 		}
1205 	}
1206 
1207 	if (inp == NULL)
1208 		smr_exit(pcbinfo->ipi_smr);
1209 
1210 	return (inp);
1211 }
1212 
1213 /*
1214  * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1215  * from which a pre-calculated hash value may be extracted.
1216  */
1217 struct inpcb *
1218 in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport,
1219     struct in6_addr *laddr, u_int lport, int lookupflags,
1220     struct ifnet *ifp __unused)
1221 {
1222 	return (in6_pcblookup_hash_smr(pcbinfo, faddr, fport, laddr, lport,
1223 	    lookupflags, M_NODOM));
1224 }
1225 
1226 struct inpcb *
1227 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1228     u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1229     struct ifnet *ifp __unused, struct mbuf *m)
1230 {
1231 	return (in6_pcblookup_hash_smr(pcbinfo, faddr, fport, laddr, lport,
1232 	    lookupflags, m->m_pkthdr.numa_domain));
1233 }
1234 
1235 void
1236 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m, int srcordst)
1237 {
1238 	struct ip6_hdr *ip;
1239 
1240 	ip = mtod(m, struct ip6_hdr *);
1241 	bzero(sin6, sizeof(*sin6));
1242 	sin6->sin6_len = sizeof(*sin6);
1243 	sin6->sin6_family = AF_INET6;
1244 	sin6->sin6_addr = srcordst ? ip->ip6_dst : ip->ip6_src;
1245 
1246 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors... */
1247 
1248 	return;
1249 }
1250