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