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
in6_pcbsetport(struct in6_addr * laddr,struct inpcb * inp,struct ucred * cred)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 /*
164 * Determine whether the inpcb can be bound to the specified address/port tuple.
165 */
166 static int
in6_pcbbind_avail(struct inpcb * inp,const struct sockaddr_in6 * sin6,int sooptions,int lookupflags,struct ucred * cred)167 in6_pcbbind_avail(struct inpcb *inp, const struct sockaddr_in6 *sin6,
168 int sooptions, int lookupflags, struct ucred *cred)
169 {
170 const struct in6_addr *laddr;
171 int reuseport, reuseport_lb;
172 u_short lport;
173
174 INP_LOCK_ASSERT(inp);
175 INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo);
176
177 laddr = &sin6->sin6_addr;
178 lport = sin6->sin6_port;
179
180 reuseport = (sooptions & SO_REUSEPORT);
181 reuseport_lb = (sooptions & SO_REUSEPORT_LB);
182
183 if (IN6_IS_ADDR_MULTICAST(laddr)) {
184 /*
185 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
186 * allow compepte duplication of binding if
187 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
188 * and a multicast address is bound on both
189 * new and duplicated sockets.
190 */
191 if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT)) != 0)
192 reuseport = SO_REUSEADDR | SO_REUSEPORT;
193 /*
194 * XXX: How to deal with SO_REUSEPORT_LB here?
195 * Treat same as SO_REUSEPORT for now.
196 */
197 if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT_LB)) != 0)
198 reuseport_lb = SO_REUSEADDR | SO_REUSEPORT_LB;
199 } else if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) {
200 struct sockaddr_in6 sin6;
201 struct epoch_tracker et;
202 struct ifaddr *ifa;
203
204 memset(&sin6, 0, sizeof(sin6));
205 sin6.sin6_family = AF_INET6;
206 sin6.sin6_len = sizeof(sin6);
207 sin6.sin6_addr = *laddr;
208
209 NET_EPOCH_ENTER(et);
210 if ((ifa = ifa_ifwithaddr((const struct sockaddr *)&sin6)) ==
211 NULL && (inp->inp_flags & INP_BINDANY) == 0) {
212 NET_EPOCH_EXIT(et);
213 return (EADDRNOTAVAIL);
214 }
215
216 /*
217 * XXX: bind to an anycast address might accidentally
218 * cause sending a packet with anycast source address.
219 * We should allow to bind to a deprecated address, since
220 * the application dares to use it.
221 */
222 if (ifa != NULL &&
223 ((struct in6_ifaddr *)ifa)->ia6_flags &
224 (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY | IN6_IFF_DETACHED)) {
225 NET_EPOCH_EXIT(et);
226 return (EADDRNOTAVAIL);
227 }
228 NET_EPOCH_EXIT(et);
229 }
230
231 if (lport != 0) {
232 struct inpcb *t;
233
234 if (ntohs(lport) <= V_ipport_reservedhigh &&
235 ntohs(lport) >= V_ipport_reservedlow &&
236 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT))
237 return (EACCES);
238
239 if (!IN6_IS_ADDR_MULTICAST(laddr) &&
240 priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT) !=
241 0) {
242 /*
243 * If a socket owned by a different user is already
244 * bound to this port, fail. In particular, SO_REUSE*
245 * can only be used to share a port among sockets owned
246 * by the same user.
247 *
248 * However, we can share a port with a connected socket
249 * which has a unique 4-tuple.
250 */
251 t = in6_pcblookup_local(inp->inp_pcbinfo, laddr, lport,
252 INPLOOKUP_WILDCARD, cred);
253 if (t != NULL &&
254 (inp->inp_socket->so_type != SOCK_STREAM ||
255 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
256 (inp->inp_cred->cr_uid != t->inp_cred->cr_uid))
257 return (EADDRINUSE);
258
259 #ifdef INET
260 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
261 IN6_IS_ADDR_UNSPECIFIED(laddr)) {
262 struct sockaddr_in sin;
263
264 in6_sin6_2_sin(&sin, sin6);
265 t = in_pcblookup_local(inp->inp_pcbinfo,
266 sin.sin_addr, lport, INPLOOKUP_WILDCARD,
267 cred);
268 if (t != NULL &&
269 (inp->inp_socket->so_type != SOCK_STREAM ||
270 in_nullhost(t->inp_faddr)) &&
271 (inp->inp_cred->cr_uid !=
272 t->inp_cred->cr_uid))
273 return (EADDRINUSE);
274 }
275 #endif
276 }
277 t = in6_pcblookup_local(inp->inp_pcbinfo, laddr, lport,
278 lookupflags, cred);
279 if (t != NULL && ((reuseport | reuseport_lb) &
280 t->inp_socket->so_options) == 0)
281 return (EADDRINUSE);
282 #ifdef INET
283 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
284 IN6_IS_ADDR_UNSPECIFIED(laddr)) {
285 struct sockaddr_in sin;
286
287 in6_sin6_2_sin(&sin, sin6);
288 t = in_pcblookup_local(inp->inp_pcbinfo, sin.sin_addr,
289 lport, lookupflags, cred);
290 if (t != NULL && ((reuseport | reuseport_lb) &
291 t->inp_socket->so_options) == 0 &&
292 (!in_nullhost(t->inp_laddr) ||
293 (t->inp_vflag & INP_IPV6PROTO) != 0)) {
294 return (EADDRINUSE);
295 }
296 }
297 #endif
298 }
299 return (0);
300 }
301
302 int
in6_pcbbind(struct inpcb * inp,struct sockaddr_in6 * sin6,struct ucred * cred)303 in6_pcbbind(struct inpcb *inp, struct sockaddr_in6 *sin6, struct ucred *cred)
304 {
305 struct socket *so = inp->inp_socket;
306 u_short lport = 0;
307 int error, lookupflags, sooptions;
308
309 INP_WLOCK_ASSERT(inp);
310 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
311
312 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
313 return (EINVAL);
314
315 lookupflags = 0;
316 sooptions = atomic_load_int(&so->so_options);
317 if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT | SO_REUSEPORT_LB)) == 0)
318 lookupflags = INPLOOKUP_WILDCARD;
319 if (sin6 == NULL) {
320 if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
321 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
322 return (error);
323 } else {
324 KASSERT(sin6->sin6_family == AF_INET6,
325 ("%s: invalid address family for %p", __func__, sin6));
326 KASSERT(sin6->sin6_len == sizeof(*sin6),
327 ("%s: invalid address length for %p", __func__, sin6));
328
329 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
330 return(error);
331
332 if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
333 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
334 return (error);
335
336 /* See if this address/port combo is available. */
337 error = in6_pcbbind_avail(inp, sin6, sooptions, lookupflags,
338 cred);
339 if (error != 0)
340 return (error);
341
342 lport = sin6->sin6_port;
343 inp->in6p_laddr = sin6->sin6_addr;
344 }
345 if (lport == 0) {
346 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
347 /* Undo an address bind that may have occurred. */
348 inp->in6p_laddr = in6addr_any;
349 return (error);
350 }
351 } else {
352 inp->inp_lport = lport;
353 if (in_pcbinshash(inp) != 0) {
354 inp->in6p_laddr = in6addr_any;
355 inp->inp_lport = 0;
356 return (EAGAIN);
357 }
358 }
359 return (0);
360 }
361
362 /*
363 * Transform old in6_pcbconnect() into an inner subroutine for new
364 * in6_pcbconnect(): Do some validity-checking on the remote
365 * address (in mbuf 'nam') and then determine local host address
366 * (i.e., which interface) to use to access that remote host.
367 *
368 * This preserves definition of in6_pcbconnect(), while supporting a
369 * slightly different version for T/TCP. (This is more than
370 * a bit of a kludge, but cleaning up the internal interfaces would
371 * have forced minor changes in every protocol).
372 */
373 static int
in6_pcbladdr(struct inpcb * inp,struct sockaddr_in6 * sin6,struct in6_addr * plocal_addr6,bool sas_required)374 in6_pcbladdr(struct inpcb *inp, struct sockaddr_in6 *sin6,
375 struct in6_addr *plocal_addr6, bool sas_required)
376 {
377 int error = 0;
378 int scope_ambiguous = 0;
379 struct in6_addr in6a;
380
381 NET_EPOCH_ASSERT();
382 INP_WLOCK_ASSERT(inp);
383 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */
384
385 if (sin6->sin6_port == 0)
386 return (EADDRNOTAVAIL);
387
388 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
389 scope_ambiguous = 1;
390 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
391 return(error);
392
393 if (V_connect_in6addr_wild && !CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) {
394 /*
395 * If the destination address is UNSPECIFIED addr,
396 * use the loopback addr, e.g ::1.
397 */
398 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
399 sin6->sin6_addr = in6addr_loopback;
400 } else if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
401 return (ENETUNREACH);
402 }
403
404 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
405 return (error);
406
407 if (sas_required) {
408 error = in6_selectsrc_socket(sin6, inp->in6p_outputopts,
409 inp, inp->inp_cred, scope_ambiguous, &in6a, NULL);
410 if (error)
411 return (error);
412 } else {
413 /*
414 * Source address selection isn't required when syncache
415 * has already established connection and both source and
416 * destination addresses was chosen.
417 *
418 * This also includes the case when fwd_tag was used to
419 * select source address in tcp_input().
420 */
421 in6a = inp->in6p_laddr;
422 }
423
424 if (IN6_IS_ADDR_UNSPECIFIED(&in6a))
425 return (EHOSTUNREACH);
426 /*
427 * Do not update this earlier, in case we return with an error.
428 *
429 * XXX: this in6_selectsrc_socket result might replace the bound local
430 * address with the address specified by setsockopt(IPV6_PKTINFO).
431 * Is it the intended behavior?
432 */
433 *plocal_addr6 = in6a;
434
435 /*
436 * Don't do pcblookup call here; return interface in
437 * plocal_addr6
438 * and exit to caller, that will do the lookup.
439 */
440
441 return (0);
442 }
443
444 /*
445 * Outer subroutine:
446 * Connect from a socket to a specified address.
447 * Both address and port must be specified in argument sin.
448 * If don't have a local address for this socket yet,
449 * then pick one.
450 */
451 int
in6_pcbconnect(struct inpcb * inp,struct sockaddr_in6 * sin6,struct ucred * cred,bool sas_required)452 in6_pcbconnect(struct inpcb *inp, struct sockaddr_in6 *sin6, struct ucred *cred,
453 bool sas_required)
454 {
455 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
456 struct sockaddr_in6 laddr6;
457 int error;
458
459 NET_EPOCH_ASSERT();
460 INP_WLOCK_ASSERT(inp);
461 INP_HASH_WLOCK_ASSERT(pcbinfo);
462 KASSERT(sin6->sin6_family == AF_INET6,
463 ("%s: invalid address family for %p", __func__, sin6));
464 KASSERT(sin6->sin6_len == sizeof(*sin6),
465 ("%s: invalid address length for %p", __func__, sin6));
466 KASSERT(IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr),
467 ("%s: inp is already connected", __func__));
468
469 bzero(&laddr6, sizeof(laddr6));
470 laddr6.sin6_family = AF_INET6;
471
472 #ifdef ROUTE_MPATH
473 if (CALC_FLOWID_OUTBOUND) {
474 uint32_t hash_type, hash_val;
475
476 hash_val = fib6_calc_software_hash(&inp->in6p_laddr,
477 &sin6->sin6_addr, 0, sin6->sin6_port,
478 inp->inp_socket->so_proto->pr_protocol, &hash_type);
479 inp->inp_flowid = hash_val;
480 inp->inp_flowtype = hash_type;
481 }
482 #endif
483 /*
484 * Call inner routine, to assign local interface address.
485 * in6_pcbladdr() may automatically fill in sin6_scope_id.
486 */
487 if ((error = in6_pcbladdr(inp, sin6, &laddr6.sin6_addr,
488 sas_required)) != 0)
489 return (error);
490
491 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
492 sin6->sin6_port, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
493 &laddr6.sin6_addr : &inp->in6p_laddr, inp->inp_lport, 0,
494 M_NODOM) != NULL)
495 return (EADDRINUSE);
496 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
497 if (inp->inp_lport == 0) {
498 error = in_pcb_lport_dest(inp,
499 (struct sockaddr *) &laddr6, &inp->inp_lport,
500 (struct sockaddr *) sin6, sin6->sin6_port, cred,
501 INPLOOKUP_WILDCARD);
502 if (error)
503 return (error);
504 }
505 inp->in6p_laddr = laddr6.sin6_addr;
506 }
507 inp->in6p_faddr = sin6->sin6_addr;
508 inp->inp_fport = sin6->sin6_port;
509 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
510 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
511 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
512 inp->inp_flow |=
513 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
514
515 if ((inp->inp_flags & INP_INHASHLIST) != 0) {
516 in_pcbrehash(inp);
517 } else {
518 in_pcbinshash(inp);
519 }
520
521 return (0);
522 }
523
524 void
in6_pcbdisconnect(struct inpcb * inp)525 in6_pcbdisconnect(struct inpcb *inp)
526 {
527
528 INP_WLOCK_ASSERT(inp);
529 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
530 KASSERT(inp->inp_smr == SMR_SEQ_INVALID,
531 ("%s: inp %p was already disconnected", __func__, inp));
532
533 in_pcbremhash_locked(inp);
534
535 /* See the comment in in_pcbinshash(). */
536 inp->inp_smr = smr_advance(inp->inp_pcbinfo->ipi_smr);
537
538 /* XXX-MJ torn writes are visible to SMR lookup */
539 memset(&inp->in6p_laddr, 0, sizeof(inp->in6p_laddr));
540 memset(&inp->in6p_faddr, 0, sizeof(inp->in6p_faddr));
541 inp->inp_fport = 0;
542 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
543 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
544 }
545
546 int
in6_getsockaddr(struct socket * so,struct sockaddr * sa)547 in6_getsockaddr(struct socket *so, struct sockaddr *sa)
548 {
549 struct inpcb *inp;
550
551 inp = sotoinpcb(so);
552 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
553
554 *(struct sockaddr_in6 *)sa = (struct sockaddr_in6 ){
555 .sin6_len = sizeof(struct sockaddr_in6),
556 .sin6_family = AF_INET6,
557 .sin6_port = inp->inp_lport,
558 .sin6_addr = inp->in6p_laddr,
559 };
560 /* XXX: should catch errors */
561 (void)sa6_recoverscope((struct sockaddr_in6 *)sa);
562
563 return (0);
564 }
565
566 int
in6_getpeeraddr(struct socket * so,struct sockaddr * sa)567 in6_getpeeraddr(struct socket *so, struct sockaddr *sa)
568 {
569 struct inpcb *inp;
570
571 inp = sotoinpcb(so);
572 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
573
574 *(struct sockaddr_in6 *)sa = (struct sockaddr_in6 ){
575 .sin6_len = sizeof(struct sockaddr_in6),
576 .sin6_family = AF_INET6,
577 .sin6_port = inp->inp_fport,
578 .sin6_addr = inp->in6p_faddr,
579 };
580 /* XXX: should catch errors */
581 (void)sa6_recoverscope((struct sockaddr_in6 *)sa);
582
583 return (0);
584 }
585
586 int
in6_mapped_sockaddr(struct socket * so,struct sockaddr * sa)587 in6_mapped_sockaddr(struct socket *so, struct sockaddr *sa)
588 {
589 int error;
590 #ifdef INET
591 struct inpcb *inp;
592
593 inp = sotoinpcb(so);
594 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
595
596 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
597 struct sockaddr_in sin;
598
599 error = in_getsockaddr(so, (struct sockaddr *)&sin);
600 if (error == 0)
601 in6_sin_2_v4mapsin6(&sin, (struct sockaddr_in6 *)sa);
602 } else
603 #endif
604 {
605 /* scope issues will be handled in in6_getsockaddr(). */
606 error = in6_getsockaddr(so, sa);
607 }
608
609 return error;
610 }
611
612 int
in6_mapped_peeraddr(struct socket * so,struct sockaddr * sa)613 in6_mapped_peeraddr(struct socket *so, struct sockaddr *sa)
614 {
615 int error;
616 #ifdef INET
617 struct inpcb *inp;
618
619 inp = sotoinpcb(so);
620 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
621
622 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
623 struct sockaddr_in sin;
624
625 error = in_getpeeraddr(so, (struct sockaddr *)&sin);
626 if (error == 0)
627 in6_sin_2_v4mapsin6(&sin, (struct sockaddr_in6 *)sa);
628 } else
629 #endif
630 {
631 /* scope issues will be handled in in6_getpeeraddr(). */
632 error = in6_getpeeraddr(so, sa);
633 }
634
635 return error;
636 }
637
638 /*
639 * Pass some notification to all connections of a protocol
640 * associated with address dst. The local address and/or port numbers
641 * may be specified to limit the search. The "usual action" will be
642 * taken, depending on the ctlinput cmd. The caller must filter any
643 * cmds that are uninteresting (e.g., no error in the map).
644 * Call the protocol specific routine (if any) to report
645 * any errors for each matching socket.
646 */
647 static bool
inp_match6(const struct inpcb * inp,void * v __unused)648 inp_match6(const struct inpcb *inp, void *v __unused)
649 {
650
651 return ((inp->inp_vflag & INP_IPV6) != 0);
652 }
653
654 void
in6_pcbnotify(struct inpcbinfo * pcbinfo,struct sockaddr_in6 * sa6_dst,u_int fport_arg,const struct sockaddr_in6 * src,u_int lport_arg,int errno,void * cmdarg,struct inpcb * (* notify)(struct inpcb *,int))655 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr_in6 *sa6_dst,
656 u_int fport_arg, const struct sockaddr_in6 *src, u_int lport_arg,
657 int errno, void *cmdarg,
658 struct inpcb *(*notify)(struct inpcb *, int))
659 {
660 struct inpcb_iterator inpi = INP_ITERATOR(pcbinfo, INPLOOKUP_WLOCKPCB,
661 inp_match6, NULL);
662 struct inpcb *inp;
663 struct sockaddr_in6 sa6_src;
664 u_short fport = fport_arg, lport = lport_arg;
665 u_int32_t flowinfo;
666
667 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
668 return;
669
670 /*
671 * note that src can be NULL when we get notify by local fragmentation.
672 */
673 sa6_src = (src == NULL) ? sa6_any : *src;
674 flowinfo = sa6_src.sin6_flowinfo;
675
676 while ((inp = inp_next(&inpi)) != NULL) {
677 INP_WLOCK_ASSERT(inp);
678 /*
679 * If the error designates a new path MTU for a destination
680 * and the application (associated with this socket) wanted to
681 * know the value, notify.
682 * XXX: should we avoid to notify the value to TCP sockets?
683 */
684 if (errno == EMSGSIZE && cmdarg != NULL)
685 ip6_notify_pmtu(inp, sa6_dst, *(uint32_t *)cmdarg);
686
687 /*
688 * Detect if we should notify the error. If no source and
689 * destination ports are specified, but non-zero flowinfo and
690 * local address match, notify the error. This is the case
691 * when the error is delivered with an encrypted buffer
692 * by ESP. Otherwise, just compare addresses and ports
693 * as usual.
694 */
695 if (lport == 0 && fport == 0 && flowinfo &&
696 inp->inp_socket != NULL &&
697 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
698 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
699 goto do_notify;
700 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
701 &sa6_dst->sin6_addr) ||
702 inp->inp_socket == 0 ||
703 (lport && inp->inp_lport != lport) ||
704 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
705 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
706 &sa6_src.sin6_addr)) ||
707 (fport && inp->inp_fport != fport)) {
708 continue;
709 }
710
711 do_notify:
712 if (notify)
713 (*notify)(inp, errno);
714 }
715 }
716
717 /*
718 * Lookup a PCB based on the local address and port. Caller must hold the
719 * hash lock. No inpcb locks or references are acquired.
720 */
721 struct inpcb *
in6_pcblookup_local(struct inpcbinfo * pcbinfo,const struct in6_addr * laddr,u_short lport,int lookupflags,struct ucred * cred)722 in6_pcblookup_local(struct inpcbinfo *pcbinfo, const struct in6_addr *laddr,
723 u_short lport, int lookupflags, struct ucred *cred)
724 {
725 struct inpcb *inp;
726 int matchwild = 3, wildcard;
727
728 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
729 ("%s: invalid lookup flags %d", __func__, lookupflags));
730
731 INP_HASH_LOCK_ASSERT(pcbinfo);
732
733 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
734 struct inpcbhead *head;
735 /*
736 * Look for an unconnected (wildcard foreign addr) PCB that
737 * matches the local address and port we're looking for.
738 */
739 head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
740 pcbinfo->ipi_hashmask)];
741 CK_LIST_FOREACH(inp, head, inp_hash_wild) {
742 /* XXX inp locking */
743 if ((inp->inp_vflag & INP_IPV6) == 0)
744 continue;
745 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
746 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
747 inp->inp_lport == lport) {
748 /* Found. */
749 if (prison_equal_ip6(cred->cr_prison,
750 inp->inp_cred->cr_prison))
751 return (inp);
752 }
753 }
754 /*
755 * Not found.
756 */
757 return (NULL);
758 } else {
759 struct inpcbporthead *porthash;
760 struct inpcbport *phd;
761 struct inpcb *match = NULL;
762 /*
763 * Best fit PCB lookup.
764 *
765 * First see if this local port is in use by looking on the
766 * port hash list.
767 */
768 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
769 pcbinfo->ipi_porthashmask)];
770 CK_LIST_FOREACH(phd, porthash, phd_hash) {
771 if (phd->phd_port == lport)
772 break;
773 }
774 if (phd != NULL) {
775 /*
776 * Port is in use by one or more PCBs. Look for best
777 * fit.
778 */
779 CK_LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
780 wildcard = 0;
781 if (!prison_equal_ip6(cred->cr_prison,
782 inp->inp_cred->cr_prison))
783 continue;
784 /* XXX inp locking */
785 if ((inp->inp_vflag & INP_IPV6) == 0)
786 continue;
787 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
788 wildcard++;
789 if (!IN6_IS_ADDR_UNSPECIFIED(
790 &inp->in6p_laddr)) {
791 if (IN6_IS_ADDR_UNSPECIFIED(laddr))
792 wildcard++;
793 else if (!IN6_ARE_ADDR_EQUAL(
794 &inp->in6p_laddr, laddr))
795 continue;
796 } else {
797 if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
798 wildcard++;
799 }
800 if (wildcard < matchwild) {
801 match = inp;
802 matchwild = wildcard;
803 if (matchwild == 0)
804 break;
805 }
806 }
807 }
808 return (match);
809 }
810 }
811
812 static bool
in6_multi_match(const struct inpcb * inp,void * v __unused)813 in6_multi_match(const struct inpcb *inp, void *v __unused)
814 {
815
816 if ((inp->inp_vflag & INP_IPV6) && inp->in6p_moptions != NULL)
817 return (true);
818 else
819 return (false);
820 }
821
822 void
in6_pcbpurgeif0(struct inpcbinfo * pcbinfo,struct ifnet * ifp)823 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
824 {
825 struct inpcb_iterator inpi = INP_ITERATOR(pcbinfo, INPLOOKUP_RLOCKPCB,
826 in6_multi_match, NULL);
827 struct inpcb *inp;
828 struct in6_multi *inm;
829 struct in6_mfilter *imf;
830 struct ip6_moptions *im6o;
831
832 IN6_MULTI_LOCK_ASSERT();
833
834 while ((inp = inp_next(&inpi)) != NULL) {
835 INP_RLOCK_ASSERT(inp);
836
837 im6o = inp->in6p_moptions;
838 /*
839 * Unselect the outgoing ifp for multicast if it
840 * is being detached.
841 */
842 if (im6o->im6o_multicast_ifp == ifp)
843 im6o->im6o_multicast_ifp = NULL;
844 /*
845 * Drop multicast group membership if we joined
846 * through the interface being detached.
847 */
848 restart:
849 IP6_MFILTER_FOREACH(imf, &im6o->im6o_head) {
850 if ((inm = imf->im6f_in6m) == NULL)
851 continue;
852 if (inm->in6m_ifp != ifp)
853 continue;
854 ip6_mfilter_remove(&im6o->im6o_head, imf);
855 in6_leavegroup_locked(inm, NULL);
856 ip6_mfilter_free(imf);
857 goto restart;
858 }
859 }
860 }
861
862 /*
863 * Check for alternatives when higher level complains
864 * about service problems. For now, invalidate cached
865 * routing information. If the route was created dynamically
866 * (by a redirect), time to try a default gateway again.
867 */
868 void
in6_losing(struct inpcb * inp)869 in6_losing(struct inpcb *inp)
870 {
871
872 RO_INVALIDATE_CACHE(&inp->inp_route6);
873 }
874
875 /*
876 * After a routing change, flush old routing
877 * and allocate a (hopefully) better one.
878 */
879 struct inpcb *
in6_rtchange(struct inpcb * inp,int errno __unused)880 in6_rtchange(struct inpcb *inp, int errno __unused)
881 {
882
883 RO_INVALIDATE_CACHE(&inp->inp_route6);
884 return inp;
885 }
886
887 static bool
in6_pcblookup_lb_numa_match(const struct inpcblbgroup * grp,int domain)888 in6_pcblookup_lb_numa_match(const struct inpcblbgroup *grp, int domain)
889 {
890 return (domain == M_NODOM || domain == grp->il_numa_domain);
891 }
892
893 static struct inpcb *
in6_pcblookup_lbgroup(const struct inpcbinfo * pcbinfo,const struct in6_addr * faddr,uint16_t fport,const struct in6_addr * laddr,uint16_t lport,uint8_t domain)894 in6_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo,
895 const struct in6_addr *faddr, uint16_t fport, const struct in6_addr *laddr,
896 uint16_t lport, uint8_t domain)
897 {
898 const struct inpcblbgrouphead *hdr;
899 struct inpcblbgroup *grp;
900 struct inpcblbgroup *jail_exact, *jail_wild, *local_exact, *local_wild;
901 struct inpcb *inp;
902 u_int count;
903
904 INP_HASH_LOCK_ASSERT(pcbinfo);
905
906 hdr = &pcbinfo->ipi_lbgrouphashbase[
907 INP_PCBPORTHASH(lport, pcbinfo->ipi_lbgrouphashmask)];
908
909 /*
910 * Search for an LB group match based on the following criteria:
911 * - prefer jailed groups to non-jailed groups
912 * - prefer exact source address matches to wildcard matches
913 * - prefer groups bound to the specified NUMA domain
914 */
915 jail_exact = jail_wild = local_exact = local_wild = NULL;
916 CK_LIST_FOREACH(grp, hdr, il_list) {
917 bool injail;
918
919 #ifdef INET
920 if (!(grp->il_vflag & INP_IPV6))
921 continue;
922 #endif
923 if (grp->il_lport != lport)
924 continue;
925
926 injail = prison_flag(grp->il_cred, PR_IP6) != 0;
927 if (injail && prison_check_ip6_locked(grp->il_cred->cr_prison,
928 laddr) != 0)
929 continue;
930
931 if (IN6_ARE_ADDR_EQUAL(&grp->il6_laddr, laddr)) {
932 if (injail) {
933 jail_exact = grp;
934 if (in6_pcblookup_lb_numa_match(grp, domain))
935 /* This is a perfect match. */
936 goto out;
937 } else if (local_exact == NULL ||
938 in6_pcblookup_lb_numa_match(grp, domain)) {
939 local_exact = grp;
940 }
941 } else if (IN6_IS_ADDR_UNSPECIFIED(&grp->il6_laddr)) {
942 if (injail) {
943 if (jail_wild == NULL ||
944 in6_pcblookup_lb_numa_match(grp, domain))
945 jail_wild = grp;
946 } else if (local_wild == NULL ||
947 in6_pcblookup_lb_numa_match(grp, domain)) {
948 local_wild = grp;
949 }
950 }
951 }
952
953 if (jail_exact != NULL)
954 grp = jail_exact;
955 else if (jail_wild != NULL)
956 grp = jail_wild;
957 else if (local_exact != NULL)
958 grp = local_exact;
959 else
960 grp = local_wild;
961 if (grp == NULL)
962 return (NULL);
963 out:
964 /*
965 * Synchronize with in_pcblbgroup_insert().
966 */
967 count = atomic_load_acq_int(&grp->il_inpcnt);
968 if (count == 0)
969 return (NULL);
970 inp = grp->il_inp[INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport) % count];
971 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
972 return (inp);
973 }
974
975 static bool
in6_pcblookup_exact_match(const struct inpcb * inp,const struct in6_addr * faddr,u_short fport,const struct in6_addr * laddr,u_short lport)976 in6_pcblookup_exact_match(const struct inpcb *inp, const struct in6_addr *faddr,
977 u_short fport, const struct in6_addr *laddr, u_short lport)
978 {
979 /* XXX inp locking */
980 if ((inp->inp_vflag & INP_IPV6) == 0)
981 return (false);
982 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
983 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
984 inp->inp_fport == fport && inp->inp_lport == lport)
985 return (true);
986 return (false);
987 }
988
989 static struct inpcb *
in6_pcblookup_hash_exact(struct inpcbinfo * pcbinfo,const struct in6_addr * faddr,u_short fport,const struct in6_addr * laddr,u_short lport)990 in6_pcblookup_hash_exact(struct inpcbinfo *pcbinfo,
991 const struct in6_addr *faddr, u_short fport,
992 const struct in6_addr *laddr, u_short lport)
993 {
994 struct inpcbhead *head;
995 struct inpcb *inp;
996
997 INP_HASH_LOCK_ASSERT(pcbinfo);
998
999 /*
1000 * First look for an exact match.
1001 */
1002 head = &pcbinfo->ipi_hash_exact[INP6_PCBHASH(faddr, lport, fport,
1003 pcbinfo->ipi_hashmask)];
1004 CK_LIST_FOREACH(inp, head, inp_hash_exact) {
1005 if (in6_pcblookup_exact_match(inp, faddr, fport, laddr, lport))
1006 return (inp);
1007 }
1008 return (NULL);
1009 }
1010
1011 typedef enum {
1012 INPLOOKUP_MATCH_NONE = 0,
1013 INPLOOKUP_MATCH_WILD = 1,
1014 INPLOOKUP_MATCH_LADDR = 2,
1015 } inp_lookup_match_t;
1016
1017 static inp_lookup_match_t
in6_pcblookup_wild_match(const struct inpcb * inp,const struct in6_addr * laddr,u_short lport)1018 in6_pcblookup_wild_match(const struct inpcb *inp, const struct in6_addr *laddr,
1019 u_short lport)
1020 {
1021 /* XXX inp locking */
1022 if ((inp->inp_vflag & INP_IPV6) == 0)
1023 return (INPLOOKUP_MATCH_NONE);
1024 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1025 inp->inp_lport != lport)
1026 return (INPLOOKUP_MATCH_NONE);
1027 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1028 return (INPLOOKUP_MATCH_WILD);
1029 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr))
1030 return (INPLOOKUP_MATCH_LADDR);
1031 return (INPLOOKUP_MATCH_NONE);
1032 }
1033
1034 #define INP_LOOKUP_AGAIN ((struct inpcb *)(uintptr_t)-1)
1035
1036 static struct inpcb *
in6_pcblookup_hash_wild_smr(struct inpcbinfo * pcbinfo,const struct in6_addr * laddr,u_short lport,const inp_lookup_t lockflags)1037 in6_pcblookup_hash_wild_smr(struct inpcbinfo *pcbinfo,
1038 const struct in6_addr *laddr, u_short lport, const inp_lookup_t lockflags)
1039 {
1040 struct inpcbhead *head;
1041 struct inpcb *inp;
1042
1043 KASSERT(SMR_ENTERED(pcbinfo->ipi_smr),
1044 ("%s: not in SMR read section", __func__));
1045
1046 head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
1047 pcbinfo->ipi_hashmask)];
1048 CK_LIST_FOREACH(inp, head, inp_hash_wild) {
1049 inp_lookup_match_t match;
1050
1051 match = in6_pcblookup_wild_match(inp, laddr, lport);
1052 if (match == INPLOOKUP_MATCH_NONE)
1053 continue;
1054
1055 if (__predict_true(inp_smr_lock(inp, lockflags))) {
1056 match = in6_pcblookup_wild_match(inp, laddr, lport);
1057 if (match != INPLOOKUP_MATCH_NONE &&
1058 prison_check_ip6_locked(inp->inp_cred->cr_prison,
1059 laddr) == 0)
1060 return (inp);
1061 inp_unlock(inp, lockflags);
1062 }
1063
1064 /*
1065 * The matching socket disappeared out from under us. Fall back
1066 * to a serialized lookup.
1067 */
1068 return (INP_LOOKUP_AGAIN);
1069 }
1070 return (NULL);
1071 }
1072
1073 static struct inpcb *
in6_pcblookup_hash_wild_locked(struct inpcbinfo * pcbinfo,const struct in6_addr * laddr,u_short lport)1074 in6_pcblookup_hash_wild_locked(struct inpcbinfo *pcbinfo,
1075 const struct in6_addr *laddr, u_short lport)
1076 {
1077 struct inpcbhead *head;
1078 struct inpcb *inp, *jail_wild, *local_exact, *local_wild;
1079
1080 INP_HASH_LOCK_ASSERT(pcbinfo);
1081
1082 /*
1083 * Order of socket selection - we always prefer jails.
1084 * 1. jailed, non-wild.
1085 * 2. jailed, wild.
1086 * 3. non-jailed, non-wild.
1087 * 4. non-jailed, wild.
1088 */
1089 head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
1090 pcbinfo->ipi_hashmask)];
1091 local_wild = local_exact = jail_wild = NULL;
1092 CK_LIST_FOREACH(inp, head, inp_hash_wild) {
1093 inp_lookup_match_t match;
1094 bool injail;
1095
1096 match = in6_pcblookup_wild_match(inp, laddr, lport);
1097 if (match == INPLOOKUP_MATCH_NONE)
1098 continue;
1099
1100 injail = prison_flag(inp->inp_cred, PR_IP6) != 0;
1101 if (injail) {
1102 if (prison_check_ip6_locked(
1103 inp->inp_cred->cr_prison, laddr) != 0)
1104 continue;
1105 } else {
1106 if (local_exact != NULL)
1107 continue;
1108 }
1109
1110 if (match == INPLOOKUP_MATCH_LADDR) {
1111 if (injail)
1112 return (inp);
1113 else
1114 local_exact = inp;
1115 } else {
1116 if (injail)
1117 jail_wild = inp;
1118 else
1119 local_wild = inp;
1120 }
1121 }
1122
1123 if (jail_wild != NULL)
1124 return (jail_wild);
1125 if (local_exact != NULL)
1126 return (local_exact);
1127 if (local_wild != NULL)
1128 return (local_wild);
1129 return (NULL);
1130 }
1131
1132 struct inpcb *
in6_pcblookup_hash_locked(struct inpcbinfo * pcbinfo,const struct in6_addr * faddr,u_int fport_arg,const struct in6_addr * laddr,u_int lport_arg,int lookupflags,uint8_t numa_domain)1133 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo,
1134 const struct in6_addr *faddr, u_int fport_arg,
1135 const struct in6_addr *laddr, u_int lport_arg,
1136 int lookupflags, uint8_t numa_domain)
1137 {
1138 struct inpcb *inp;
1139 u_short fport = fport_arg, lport = lport_arg;
1140
1141 KASSERT((lookupflags & ~INPLOOKUP_WILDCARD) == 0,
1142 ("%s: invalid lookup flags %d", __func__, lookupflags));
1143 KASSERT(!IN6_IS_ADDR_UNSPECIFIED(faddr),
1144 ("%s: invalid foreign address", __func__));
1145 KASSERT(!IN6_IS_ADDR_UNSPECIFIED(laddr),
1146 ("%s: invalid local address", __func__));
1147 INP_HASH_LOCK_ASSERT(pcbinfo);
1148
1149 inp = in6_pcblookup_hash_exact(pcbinfo, faddr, fport, laddr, lport);
1150 if (inp != NULL)
1151 return (inp);
1152
1153 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1154 inp = in6_pcblookup_lbgroup(pcbinfo, faddr, fport, laddr,
1155 lport, numa_domain);
1156 if (inp == NULL) {
1157 inp = in6_pcblookup_hash_wild_locked(pcbinfo,
1158 laddr, lport);
1159 }
1160 }
1161 return (inp);
1162 }
1163
1164 static struct inpcb *
in6_pcblookup_hash(struct inpcbinfo * pcbinfo,const struct in6_addr * faddr,u_int fport,const struct in6_addr * laddr,u_int lport,int lookupflags,uint8_t numa_domain)1165 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1166 u_int fport, const struct in6_addr *laddr, u_int lport, int lookupflags,
1167 uint8_t numa_domain)
1168 {
1169 struct inpcb *inp;
1170 const inp_lookup_t lockflags = lookupflags & INPLOOKUP_LOCKMASK;
1171
1172 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1173 ("%s: LOCKPCB not set", __func__));
1174
1175 INP_HASH_WLOCK(pcbinfo);
1176 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1177 lookupflags & ~INPLOOKUP_LOCKMASK, numa_domain);
1178 if (inp != NULL && !inp_trylock(inp, lockflags)) {
1179 in_pcbref(inp);
1180 INP_HASH_WUNLOCK(pcbinfo);
1181 inp_lock(inp, lockflags);
1182 if (in_pcbrele(inp, lockflags))
1183 /* XXX-MJ or retry until we get a negative match? */
1184 inp = NULL;
1185 } else {
1186 INP_HASH_WUNLOCK(pcbinfo);
1187 }
1188 return (inp);
1189 }
1190
1191 static struct inpcb *
in6_pcblookup_hash_smr(struct inpcbinfo * pcbinfo,const struct in6_addr * faddr,u_int fport_arg,const struct in6_addr * laddr,u_int lport_arg,int lookupflags,uint8_t numa_domain)1192 in6_pcblookup_hash_smr(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1193 u_int fport_arg, const struct in6_addr *laddr, u_int lport_arg,
1194 int lookupflags, uint8_t numa_domain)
1195 {
1196 struct inpcb *inp;
1197 const inp_lookup_t lockflags = lookupflags & INPLOOKUP_LOCKMASK;
1198 const u_short fport = fport_arg, lport = lport_arg;
1199
1200 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1201 ("%s: invalid lookup flags %d", __func__, lookupflags));
1202 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1203 ("%s: LOCKPCB not set", __func__));
1204
1205 smr_enter(pcbinfo->ipi_smr);
1206 inp = in6_pcblookup_hash_exact(pcbinfo, faddr, fport, laddr, lport);
1207 if (inp != NULL) {
1208 if (__predict_true(inp_smr_lock(inp, lockflags))) {
1209 if (__predict_true(in6_pcblookup_exact_match(inp,
1210 faddr, fport, laddr, lport)))
1211 return (inp);
1212 inp_unlock(inp, lockflags);
1213 }
1214 /*
1215 * We failed to lock the inpcb, or its connection state changed
1216 * out from under us. Fall back to a precise search.
1217 */
1218 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1219 lookupflags, numa_domain));
1220 }
1221
1222 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1223 inp = in6_pcblookup_lbgroup(pcbinfo, faddr, fport,
1224 laddr, lport, numa_domain);
1225 if (inp != NULL) {
1226 if (__predict_true(inp_smr_lock(inp, lockflags))) {
1227 if (__predict_true(in6_pcblookup_wild_match(inp,
1228 laddr, lport) != INPLOOKUP_MATCH_NONE))
1229 return (inp);
1230 inp_unlock(inp, lockflags);
1231 }
1232 inp = INP_LOOKUP_AGAIN;
1233 } else {
1234 inp = in6_pcblookup_hash_wild_smr(pcbinfo, laddr, lport,
1235 lockflags);
1236 }
1237 if (inp == INP_LOOKUP_AGAIN) {
1238 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr,
1239 lport, lookupflags, numa_domain));
1240 }
1241 }
1242
1243 if (inp == NULL)
1244 smr_exit(pcbinfo->ipi_smr);
1245
1246 return (inp);
1247 }
1248
1249 /*
1250 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1251 * from which a pre-calculated hash value may be extracted.
1252 */
1253 struct inpcb *
in6_pcblookup(struct inpcbinfo * pcbinfo,const struct in6_addr * faddr,u_int fport,const struct in6_addr * laddr,u_int lport,int lookupflags,struct ifnet * ifp __unused)1254 in6_pcblookup(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1255 u_int fport, const struct in6_addr *laddr, u_int lport, int lookupflags,
1256 struct ifnet *ifp __unused)
1257 {
1258 return (in6_pcblookup_hash_smr(pcbinfo, faddr, fport, laddr, lport,
1259 lookupflags, M_NODOM));
1260 }
1261
1262 struct inpcb *
in6_pcblookup_mbuf(struct inpcbinfo * pcbinfo,const struct in6_addr * faddr,u_int fport,const struct in6_addr * laddr,u_int lport,int lookupflags,struct ifnet * ifp __unused,struct mbuf * m)1263 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1264 u_int fport, const struct in6_addr *laddr, u_int lport, int lookupflags,
1265 struct ifnet *ifp __unused, struct mbuf *m)
1266 {
1267 return (in6_pcblookup_hash_smr(pcbinfo, faddr, fport, laddr, lport,
1268 lookupflags, m->m_pkthdr.numa_domain));
1269 }
1270
1271 void
init_sin6(struct sockaddr_in6 * sin6,struct mbuf * m,int srcordst)1272 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m, int srcordst)
1273 {
1274 struct ip6_hdr *ip;
1275
1276 ip = mtod(m, struct ip6_hdr *);
1277 bzero(sin6, sizeof(*sin6));
1278 sin6->sin6_len = sizeof(*sin6);
1279 sin6->sin6_family = AF_INET6;
1280 sin6->sin6_addr = srcordst ? ip->ip6_dst : ip->ip6_src;
1281
1282 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */
1283
1284 return;
1285 }
1286