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 "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70 #include "opt_route.h"
71 #include "opt_rss.h"
72
73 #include <sys/hash.h>
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/domain.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/smr.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/sockio.h>
85 #include <sys/sysctl.h>
86 #include <sys/errno.h>
87 #include <sys/time.h>
88 #include <sys/priv.h>
89 #include <sys/proc.h>
90 #include <sys/jail.h>
91
92 #include <vm/uma.h>
93
94 #include <net/if.h>
95 #include <net/if_var.h>
96 #include <net/if_llatbl.h>
97 #include <net/if_types.h>
98 #include <net/route.h>
99 #include <net/route/nhop.h>
100 #include <net/vnet.h>
101
102 #include <netinet/in.h>
103 #include <netinet/in_var.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/ip6.h>
106 #include <netinet/ip_var.h>
107
108 #include <netinet6/ip6_var.h>
109 #include <netinet6/nd6.h>
110 #include <netinet/in_pcb.h>
111 #include <netinet/in_pcb_var.h>
112 #include <netinet6/in6_pcb.h>
113 #include <netinet6/in6_fib.h>
114 #include <netinet6/scope6_var.h>
115
116 SYSCTL_DECL(_net_inet6);
117 SYSCTL_DECL(_net_inet6_ip6);
118 VNET_DEFINE_STATIC(int, connect_in6addr_wild) = 1;
119 #define V_connect_in6addr_wild VNET(connect_in6addr_wild)
120 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, connect_in6addr_wild,
121 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(connect_in6addr_wild), 0,
122 "Allow connecting to the unspecified address for connect(2)");
123
124 int
in6_pcbsetport(struct in6_addr * laddr,struct inpcb * inp,struct ucred * cred)125 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
126 {
127 struct socket *so = inp->inp_socket;
128 u_int16_t lport = 0;
129 int error, lookupflags = 0;
130 #ifdef INVARIANTS
131 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
132 #endif
133
134 INP_WLOCK_ASSERT(inp);
135 INP_HASH_WLOCK_ASSERT(pcbinfo);
136
137 error = prison_local_ip6(cred, laddr,
138 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
139 if (error)
140 return(error);
141
142 /* XXX: this is redundant when called from in6_pcbbind */
143 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0)
144 lookupflags = INPLOOKUP_WILDCARD;
145
146 inp->inp_flags |= INP_ANONPORT;
147
148 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
149 if (error != 0)
150 return (error);
151
152 inp->inp_lport = lport;
153 if (in_pcbinshash(inp) != 0) {
154 inp->in6p_laddr = in6addr_any;
155 inp->inp_lport = 0;
156 return (EAGAIN);
157 }
158
159 return (0);
160 }
161
162 /*
163 * Determine whether the inpcb can be bound to the specified address/port tuple.
164 */
165 static int
in6_pcbbind_avail(struct inpcb * inp,const struct sockaddr_in6 * sin6,int fib,int sooptions,int lookupflags,struct ucred * cred)166 in6_pcbbind_avail(struct inpcb *inp, const struct sockaddr_in6 *sin6, int fib,
167 int sooptions, int lookupflags, struct ucred *cred)
168 {
169 const struct in6_addr *laddr;
170 int reuseport, reuseport_lb;
171 u_short lport;
172
173 INP_LOCK_ASSERT(inp);
174 INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo);
175
176 laddr = &sin6->sin6_addr;
177 lport = sin6->sin6_port;
178
179 reuseport = (sooptions & SO_REUSEPORT);
180 reuseport_lb = (sooptions & SO_REUSEPORT_LB);
181
182 if (IN6_IS_ADDR_MULTICAST(laddr)) {
183 /*
184 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
185 * allow compepte duplication of binding if
186 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
187 * and a multicast address is bound on both
188 * new and duplicated sockets.
189 */
190 if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT)) != 0)
191 reuseport = SO_REUSEADDR | SO_REUSEPORT;
192 /*
193 * XXX: How to deal with SO_REUSEPORT_LB here?
194 * Treat same as SO_REUSEPORT for now.
195 */
196 if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT_LB)) != 0)
197 reuseport_lb = SO_REUSEADDR | SO_REUSEPORT_LB;
198 } else if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) {
199 struct sockaddr_in6 sin6;
200 struct epoch_tracker et;
201 struct ifaddr *ifa;
202
203 memset(&sin6, 0, sizeof(sin6));
204 sin6.sin6_family = AF_INET6;
205 sin6.sin6_len = sizeof(sin6);
206 sin6.sin6_addr = *laddr;
207
208 NET_EPOCH_ENTER(et);
209 if ((ifa = ifa_ifwithaddr((const struct sockaddr *)&sin6)) ==
210 NULL && (inp->inp_flags & INP_BINDANY) == 0) {
211 NET_EPOCH_EXIT(et);
212 return (EADDRNOTAVAIL);
213 }
214
215 /*
216 * We used to prohibit binding to an anycast address here,
217 * based on RFC3513, but that restriction was removed in
218 * RFC4291.
219 */
220 if (ifa != NULL &&
221 ((struct in6_ifaddr *)ifa)->ia6_flags &
222 (IN6_IFF_NOTREADY | IN6_IFF_DETACHED)) {
223 NET_EPOCH_EXIT(et);
224 return (EADDRNOTAVAIL);
225 }
226 NET_EPOCH_EXIT(et);
227 }
228
229 if (lport != 0) {
230 struct inpcb *t;
231
232 if (ntohs(lport) <= V_ipport_reservedhigh &&
233 ntohs(lport) >= V_ipport_reservedlow &&
234 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT))
235 return (EACCES);
236
237 if (!IN6_IS_ADDR_MULTICAST(laddr) &&
238 priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT) !=
239 0) {
240 /*
241 * If a socket owned by a different user is already
242 * bound to this port, fail. In particular, SO_REUSE*
243 * can only be used to share a port among sockets owned
244 * by the same user.
245 *
246 * However, we can share a port with a connected socket
247 * which has a unique 4-tuple.
248 */
249 t = in6_pcblookup_local(inp->inp_pcbinfo, laddr, lport,
250 RT_ALL_FIBS, INPLOOKUP_WILDCARD, cred);
251 if (t != NULL &&
252 (inp->inp_socket->so_type != SOCK_STREAM ||
253 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
254 (inp->inp_cred->cr_uid != t->inp_cred->cr_uid))
255 return (EADDRINUSE);
256
257 #ifdef INET
258 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
259 IN6_IS_ADDR_UNSPECIFIED(laddr)) {
260 struct sockaddr_in sin;
261
262 in6_sin6_2_sin(&sin, sin6);
263 t = in_pcblookup_local(inp->inp_pcbinfo,
264 sin.sin_addr, lport, RT_ALL_FIBS,
265 INPLOOKUP_WILDCARD, cred);
266 if (t != NULL &&
267 (inp->inp_socket->so_type != SOCK_STREAM ||
268 in_nullhost(t->inp_faddr)) &&
269 (inp->inp_cred->cr_uid !=
270 t->inp_cred->cr_uid))
271 return (EADDRINUSE);
272 }
273 #endif
274 }
275 t = in6_pcblookup_local(inp->inp_pcbinfo, laddr, lport,
276 fib, lookupflags, cred);
277 if (t != NULL && ((reuseport | reuseport_lb) &
278 t->inp_socket->so_options) == 0)
279 return (EADDRINUSE);
280 #ifdef INET
281 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
282 IN6_IS_ADDR_UNSPECIFIED(laddr)) {
283 struct sockaddr_in sin;
284
285 in6_sin6_2_sin(&sin, sin6);
286 t = in_pcblookup_local(inp->inp_pcbinfo, sin.sin_addr,
287 lport, RT_ALL_FIBS, lookupflags, cred);
288 if (t != NULL && ((reuseport | reuseport_lb) &
289 t->inp_socket->so_options) == 0 &&
290 (!in_nullhost(t->inp_laddr) ||
291 (t->inp_vflag & INP_IPV6PROTO) != 0)) {
292 return (EADDRINUSE);
293 }
294 }
295 #endif
296 }
297 return (0);
298 }
299
300 int
in6_pcbbind(struct inpcb * inp,struct sockaddr_in6 * sin6,int flags,struct ucred * cred)301 in6_pcbbind(struct inpcb *inp, struct sockaddr_in6 *sin6, int flags,
302 struct ucred *cred)
303 {
304 struct socket *so = inp->inp_socket;
305 u_short lport = 0;
306 int error, fib, lookupflags, sooptions;
307
308 INP_WLOCK_ASSERT(inp);
309 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
310
311 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
312 return (EINVAL);
313
314 lookupflags = 0;
315 sooptions = atomic_load_int(&so->so_options);
316 if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT | SO_REUSEPORT_LB)) == 0)
317 lookupflags = INPLOOKUP_WILDCARD;
318 if (sin6 == NULL) {
319 if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
320 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
321 return (error);
322 } else {
323 KASSERT(sin6->sin6_family == AF_INET6,
324 ("%s: invalid address family for %p", __func__, sin6));
325 KASSERT(sin6->sin6_len == sizeof(*sin6),
326 ("%s: invalid address length for %p", __func__, sin6));
327
328 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
329 return(error);
330
331 if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
332 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
333 return (error);
334
335 fib = (flags & INPBIND_FIB) != 0 ? inp->inp_inc.inc_fibnum :
336 RT_ALL_FIBS;
337
338 /* See if this address/port combo is available. */
339 error = in6_pcbbind_avail(inp, sin6, fib, sooptions, lookupflags,
340 cred);
341 if (error != 0)
342 return (error);
343
344 lport = sin6->sin6_port;
345 inp->in6p_laddr = sin6->sin6_addr;
346 }
347 if ((flags & INPBIND_FIB) != 0)
348 inp->inp_flags |= INP_BOUNDFIB;
349 if (lport == 0) {
350 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
351 /* Undo an address bind that may have occurred. */
352 inp->inp_flags &= ~INP_BOUNDFIB;
353 inp->in6p_laddr = in6addr_any;
354 return (error);
355 }
356 } else {
357 inp->inp_lport = lport;
358 if (in_pcbinshash(inp) != 0) {
359 inp->inp_flags &= ~INP_BOUNDFIB;
360 inp->in6p_laddr = in6addr_any;
361 inp->inp_lport = 0;
362 return (EAGAIN);
363 }
364 }
365 return (0);
366 }
367
368 /*
369 * Transform old in6_pcbconnect() into an inner subroutine for new
370 * in6_pcbconnect(): Do some validity-checking on the remote
371 * address (in mbuf 'nam') and then determine local host address
372 * (i.e., which interface) to use to access that remote host.
373 *
374 * This preserves definition of in6_pcbconnect(), while supporting a
375 * slightly different version for T/TCP. (This is more than
376 * a bit of a kludge, but cleaning up the internal interfaces would
377 * have forced minor changes in every protocol).
378 */
379 static int
in6_pcbladdr(struct inpcb * inp,struct sockaddr_in6 * sin6,struct in6_addr * plocal_addr6,bool sas_required)380 in6_pcbladdr(struct inpcb *inp, struct sockaddr_in6 *sin6,
381 struct in6_addr *plocal_addr6, bool sas_required)
382 {
383 int error = 0;
384 int scope_ambiguous = 0;
385 struct in6_addr in6a;
386
387 NET_EPOCH_ASSERT();
388 INP_WLOCK_ASSERT(inp);
389 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */
390
391 if (sin6->sin6_port == 0)
392 return (EADDRNOTAVAIL);
393
394 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
395 scope_ambiguous = 1;
396 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
397 return(error);
398
399 if (V_connect_in6addr_wild && !CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) {
400 /*
401 * If the destination address is UNSPECIFIED addr,
402 * use the loopback addr, e.g ::1.
403 */
404 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
405 sin6->sin6_addr = in6addr_loopback;
406 } else if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
407 return (ENETUNREACH);
408 }
409
410 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
411 return (error);
412
413 if (sas_required) {
414 error = in6_selectsrc_socket(sin6, inp->in6p_outputopts,
415 inp, inp->inp_cred, scope_ambiguous, &in6a, NULL);
416 if (error)
417 return (error);
418 } else {
419 /*
420 * Source address selection isn't required when syncache
421 * has already established connection and both source and
422 * destination addresses was chosen.
423 *
424 * This also includes the case when fwd_tag was used to
425 * select source address in tcp_input().
426 */
427 in6a = inp->in6p_laddr;
428 }
429
430 if (IN6_IS_ADDR_UNSPECIFIED(&in6a))
431 return (EHOSTUNREACH);
432 /*
433 * Do not update this earlier, in case we return with an error.
434 *
435 * XXX: this in6_selectsrc_socket result might replace the bound local
436 * address with the address specified by setsockopt(IPV6_PKTINFO).
437 * Is it the intended behavior?
438 */
439 *plocal_addr6 = in6a;
440
441 /*
442 * Don't do pcblookup call here; return interface in
443 * plocal_addr6
444 * and exit to caller, that will do the lookup.
445 */
446
447 return (0);
448 }
449
450 /*
451 * Outer subroutine:
452 * Connect from a socket to a specified address.
453 * Both address and port must be specified in argument sin.
454 * If don't have a local address for this socket yet,
455 * then pick one.
456 */
457 int
in6_pcbconnect(struct inpcb * inp,struct sockaddr_in6 * sin6,struct ucred * cred,bool sas_required)458 in6_pcbconnect(struct inpcb *inp, struct sockaddr_in6 *sin6, struct ucred *cred,
459 bool sas_required)
460 {
461 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
462 struct sockaddr_in6 laddr6;
463 int error;
464
465 NET_EPOCH_ASSERT();
466 INP_WLOCK_ASSERT(inp);
467 INP_HASH_WLOCK_ASSERT(pcbinfo);
468 KASSERT(sin6->sin6_family == AF_INET6,
469 ("%s: invalid address family for %p", __func__, sin6));
470 KASSERT(sin6->sin6_len == sizeof(*sin6),
471 ("%s: invalid address length for %p", __func__, sin6));
472 KASSERT(IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr),
473 ("%s: inp is already connected", __func__));
474
475 bzero(&laddr6, sizeof(laddr6));
476 laddr6.sin6_family = AF_INET6;
477
478 #ifdef ROUTE_MPATH
479 if (CALC_FLOWID_OUTBOUND) {
480 uint32_t hash_type, hash_val;
481
482 hash_val = fib6_calc_software_hash(&inp->in6p_laddr,
483 &sin6->sin6_addr, 0, sin6->sin6_port,
484 inp->inp_socket->so_proto->pr_protocol, &hash_type);
485 inp->inp_flowid = hash_val;
486 inp->inp_flowtype = hash_type;
487 }
488 #endif
489 /*
490 * Call inner routine, to assign local interface address.
491 * in6_pcbladdr() may automatically fill in sin6_scope_id.
492 */
493 if ((error = in6_pcbladdr(inp, sin6, &laddr6.sin6_addr,
494 sas_required)) != 0)
495 return (error);
496
497 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
498 sin6->sin6_port, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
499 &laddr6.sin6_addr : &inp->in6p_laddr, inp->inp_lport, 0,
500 M_NODOM, RT_ALL_FIBS) != NULL)
501 return (EADDRINUSE);
502 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
503 if (inp->inp_lport == 0) {
504 error = in_pcb_lport_dest(inp,
505 (struct sockaddr *) &laddr6, &inp->inp_lport,
506 (struct sockaddr *) sin6, sin6->sin6_port, cred,
507 INPLOOKUP_WILDCARD);
508 if (error)
509 return (error);
510 }
511 inp->in6p_laddr = laddr6.sin6_addr;
512 }
513 inp->in6p_faddr = sin6->sin6_addr;
514 inp->inp_fport = sin6->sin6_port;
515 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
516 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
517 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
518 inp->inp_flow |=
519 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
520
521 if ((inp->inp_flags & INP_INHASHLIST) != 0) {
522 in_pcbrehash(inp);
523 } else {
524 error = in_pcbinshash(inp);
525 MPASS(error == 0);
526 }
527
528 return (0);
529 }
530
531 void
in6_pcbdisconnect(struct inpcb * inp)532 in6_pcbdisconnect(struct inpcb *inp)
533 {
534
535 INP_WLOCK_ASSERT(inp);
536 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
537 KASSERT(inp->inp_smr == SMR_SEQ_INVALID,
538 ("%s: inp %p was already disconnected", __func__, inp));
539
540 in_pcbremhash_locked(inp);
541
542 /* See the comment in in_pcbinshash(). */
543 inp->inp_smr = smr_advance(inp->inp_pcbinfo->ipi_smr);
544
545 /* XXX-MJ torn writes are visible to SMR lookup */
546 memset(&inp->in6p_laddr, 0, sizeof(inp->in6p_laddr));
547 memset(&inp->in6p_faddr, 0, sizeof(inp->in6p_faddr));
548 inp->inp_fport = 0;
549 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
550 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
551 }
552
553 int
in6_getsockaddr(struct socket * so,struct sockaddr * sa)554 in6_getsockaddr(struct socket *so, struct sockaddr *sa)
555 {
556 struct inpcb *inp;
557
558 inp = sotoinpcb(so);
559 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
560
561 *(struct sockaddr_in6 *)sa = (struct sockaddr_in6 ){
562 .sin6_len = sizeof(struct sockaddr_in6),
563 .sin6_family = AF_INET6,
564 .sin6_port = inp->inp_lport,
565 .sin6_addr = inp->in6p_laddr,
566 };
567 /* XXX: should catch errors */
568 (void)sa6_recoverscope((struct sockaddr_in6 *)sa);
569
570 return (0);
571 }
572
573 int
in6_getpeeraddr(struct socket * so,struct sockaddr * sa)574 in6_getpeeraddr(struct socket *so, struct sockaddr *sa)
575 {
576 struct inpcb *inp;
577
578 inp = sotoinpcb(so);
579 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
580
581 *(struct sockaddr_in6 *)sa = (struct sockaddr_in6 ){
582 .sin6_len = sizeof(struct sockaddr_in6),
583 .sin6_family = AF_INET6,
584 .sin6_port = inp->inp_fport,
585 .sin6_addr = inp->in6p_faddr,
586 };
587 /* XXX: should catch errors */
588 (void)sa6_recoverscope((struct sockaddr_in6 *)sa);
589
590 return (0);
591 }
592
593 int
in6_mapped_sockaddr(struct socket * so,struct sockaddr * sa)594 in6_mapped_sockaddr(struct socket *so, struct sockaddr *sa)
595 {
596 int error;
597 #ifdef INET
598 struct inpcb *inp;
599
600 inp = sotoinpcb(so);
601 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
602
603 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
604 struct sockaddr_in sin;
605
606 error = in_getsockaddr(so, (struct sockaddr *)&sin);
607 if (error == 0)
608 in6_sin_2_v4mapsin6(&sin, (struct sockaddr_in6 *)sa);
609 } else
610 #endif
611 {
612 /* scope issues will be handled in in6_getsockaddr(). */
613 error = in6_getsockaddr(so, sa);
614 }
615
616 return error;
617 }
618
619 int
in6_mapped_peeraddr(struct socket * so,struct sockaddr * sa)620 in6_mapped_peeraddr(struct socket *so, struct sockaddr *sa)
621 {
622 int error;
623 #ifdef INET
624 struct inpcb *inp;
625
626 inp = sotoinpcb(so);
627 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
628
629 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
630 struct sockaddr_in sin;
631
632 error = in_getpeeraddr(so, (struct sockaddr *)&sin);
633 if (error == 0)
634 in6_sin_2_v4mapsin6(&sin, (struct sockaddr_in6 *)sa);
635 } else
636 #endif
637 {
638 /* scope issues will be handled in in6_getpeeraddr(). */
639 error = in6_getpeeraddr(so, sa);
640 }
641
642 return error;
643 }
644
645 /*
646 * Pass some notification to all connections of a protocol
647 * associated with address dst. The local address and/or port numbers
648 * may be specified to limit the search. The "usual action" will be
649 * taken, depending on the ctlinput cmd. The caller must filter any
650 * cmds that are uninteresting (e.g., no error in the map).
651 * Call the protocol specific routine (if any) to report
652 * any errors for each matching socket.
653 */
654 static bool
inp_match6(const struct inpcb * inp,void * v __unused)655 inp_match6(const struct inpcb *inp, void *v __unused)
656 {
657
658 return ((inp->inp_vflag & INP_IPV6) != 0);
659 }
660
661 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))662 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr_in6 *sa6_dst,
663 u_int fport_arg, const struct sockaddr_in6 *src, u_int lport_arg,
664 int errno, void *cmdarg,
665 struct inpcb *(*notify)(struct inpcb *, int))
666 {
667 struct inpcb_iterator inpi = INP_ITERATOR(pcbinfo, INPLOOKUP_WLOCKPCB,
668 inp_match6, NULL);
669 struct inpcb *inp;
670 struct sockaddr_in6 sa6_src;
671 u_short fport = fport_arg, lport = lport_arg;
672 u_int32_t flowinfo;
673
674 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
675 return;
676
677 /*
678 * note that src can be NULL when we get notify by local fragmentation.
679 */
680 sa6_src = (src == NULL) ? sa6_any : *src;
681 flowinfo = sa6_src.sin6_flowinfo;
682
683 while ((inp = inp_next(&inpi)) != NULL) {
684 INP_WLOCK_ASSERT(inp);
685 /*
686 * If the error designates a new path MTU for a destination
687 * and the application (associated with this socket) wanted to
688 * know the value, notify.
689 * XXX: should we avoid to notify the value to TCP sockets?
690 */
691 if (errno == EMSGSIZE && cmdarg != NULL)
692 ip6_notify_pmtu(inp, sa6_dst, *(uint32_t *)cmdarg);
693
694 /*
695 * Detect if we should notify the error. If no source and
696 * destination ports are specified, but non-zero flowinfo and
697 * local address match, notify the error. This is the case
698 * when the error is delivered with an encrypted buffer
699 * by ESP. Otherwise, just compare addresses and ports
700 * as usual.
701 */
702 if (lport == 0 && fport == 0 && flowinfo &&
703 inp->inp_socket != NULL &&
704 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
705 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
706 goto do_notify;
707 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
708 &sa6_dst->sin6_addr) ||
709 inp->inp_socket == 0 ||
710 (lport && inp->inp_lport != lport) ||
711 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
712 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
713 &sa6_src.sin6_addr)) ||
714 (fport && inp->inp_fport != fport)) {
715 continue;
716 }
717
718 do_notify:
719 if (notify)
720 (*notify)(inp, errno);
721 }
722 }
723
724 /*
725 * Lookup a PCB based on the local address and port. Caller must hold the
726 * hash lock. No inpcb locks or references are acquired.
727 */
728 struct inpcb *
in6_pcblookup_local(struct inpcbinfo * pcbinfo,const struct in6_addr * laddr,u_short lport,int fib,int lookupflags,struct ucred * cred)729 in6_pcblookup_local(struct inpcbinfo *pcbinfo, const struct in6_addr *laddr,
730 u_short lport, int fib, int lookupflags, struct ucred *cred)
731 {
732 struct inpcb *inp;
733 int matchwild = 3, wildcard;
734
735 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
736 ("%s: invalid lookup flags %d", __func__, lookupflags));
737 KASSERT(fib == RT_ALL_FIBS || (fib >= 0 && fib < V_rt_numfibs),
738 ("%s: invalid fib %d", __func__, fib));
739
740 INP_HASH_LOCK_ASSERT(pcbinfo);
741
742 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
743 struct inpcbhead *head;
744 /*
745 * Look for an unconnected (wildcard foreign addr) PCB that
746 * matches the local address and port we're looking for.
747 */
748 head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
749 pcbinfo->ipi_hashmask)];
750 CK_LIST_FOREACH(inp, head, inp_hash_wild) {
751 /* XXX inp locking */
752 if ((inp->inp_vflag & INP_IPV6) == 0)
753 continue;
754 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
755 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
756 inp->inp_lport == lport && (fib == RT_ALL_FIBS ||
757 inp->inp_inc.inc_fibnum == fib)) {
758 /* Found. */
759 if (prison_equal_ip6(cred->cr_prison,
760 inp->inp_cred->cr_prison))
761 return (inp);
762 }
763 }
764 /*
765 * Not found.
766 */
767 return (NULL);
768 } else {
769 struct inpcbhead *porthash;
770 struct inpcb *match = NULL;
771
772 /*
773 * Port is in use by one or more PCBs. Look for best
774 * fit.
775 */
776 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
777 pcbinfo->ipi_porthashmask)];
778 CK_LIST_FOREACH(inp, porthash, inp_portlist) {
779 if (inp->inp_lport != lport)
780 continue;
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 (fib != RT_ALL_FIBS &&
788 inp->inp_inc.inc_fibnum != fib)
789 continue;
790 wildcard = 0;
791 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
792 wildcard++;
793 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
794 if (IN6_IS_ADDR_UNSPECIFIED(laddr))
795 wildcard++;
796 else if (!IN6_ARE_ADDR_EQUAL(
797 &inp->in6p_laddr, laddr))
798 continue;
799 } else {
800 if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
801 wildcard++;
802 }
803 if (wildcard < matchwild) {
804 match = inp;
805 matchwild = wildcard;
806 if (matchwild == 0)
807 break;
808 }
809 }
810 return (match);
811 }
812 }
813
814 static bool
in6_multi_match(const struct inpcb * inp,void * v __unused)815 in6_multi_match(const struct inpcb *inp, void *v __unused)
816 {
817
818 if ((inp->inp_vflag & INP_IPV6) && inp->in6p_moptions != NULL)
819 return (true);
820 else
821 return (false);
822 }
823
824 void
in6_pcbpurgeif0(struct inpcbinfo * pcbinfo,struct ifnet * ifp)825 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
826 {
827 struct inpcb_iterator inpi = INP_ITERATOR(pcbinfo, INPLOOKUP_RLOCKPCB,
828 in6_multi_match, NULL);
829 struct inpcb *inp;
830 struct in6_multi *inm;
831 struct in6_mfilter *imf;
832 struct ip6_moptions *im6o;
833
834 IN6_MULTI_LOCK_ASSERT();
835
836 while ((inp = inp_next(&inpi)) != NULL) {
837 INP_RLOCK_ASSERT(inp);
838
839 im6o = inp->in6p_moptions;
840 /*
841 * Unselect the outgoing ifp for multicast if it
842 * is being detached.
843 */
844 if (im6o->im6o_multicast_ifp == ifp)
845 im6o->im6o_multicast_ifp = NULL;
846 /*
847 * Drop multicast group membership if we joined
848 * through the interface being detached.
849 */
850 restart:
851 IP6_MFILTER_FOREACH(imf, &im6o->im6o_head) {
852 if ((inm = imf->im6f_in6m) == NULL)
853 continue;
854 if (inm->in6m_ifp != ifp)
855 continue;
856 ip6_mfilter_remove(&im6o->im6o_head, imf);
857 in6_leavegroup_locked(inm, NULL);
858 ip6_mfilter_free(imf);
859 goto restart;
860 }
861 }
862 }
863
864 /*
865 * Check for alternatives when higher level complains
866 * about service problems. For now, invalidate cached
867 * routing information. If the route was created dynamically
868 * (by a redirect), time to try a default gateway again.
869 */
870 void
in6_losing(struct inpcb * inp)871 in6_losing(struct inpcb *inp)
872 {
873
874 RO_INVALIDATE_CACHE(&inp->inp_route6);
875 }
876
877 /*
878 * After a routing change, flush old routing
879 * and allocate a (hopefully) better one.
880 */
881 struct inpcb *
in6_rtchange(struct inpcb * inp,int errno __unused)882 in6_rtchange(struct inpcb *inp, int errno __unused)
883 {
884
885 RO_INVALIDATE_CACHE(&inp->inp_route6);
886 return inp;
887 }
888
889 static bool
in6_pcblookup_lb_match(const struct inpcblbgroup * grp,int domain,int fib)890 in6_pcblookup_lb_match(const struct inpcblbgroup *grp, int domain, int fib)
891 {
892 return ((domain == M_NODOM || domain == grp->il_numa_domain) &&
893 (fib == RT_ALL_FIBS || fib == grp->il_fibnum));
894 }
895
896 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,int fib)897 in6_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo,
898 const struct in6_addr *faddr, uint16_t fport, const struct in6_addr *laddr,
899 uint16_t lport, uint8_t domain, int fib)
900 {
901 const struct inpcblbgrouphead *hdr;
902 struct inpcblbgroup *grp;
903 struct inpcblbgroup *jail_exact, *jail_wild, *local_exact, *local_wild;
904 struct inpcb *inp;
905 u_int count;
906
907 INP_HASH_LOCK_ASSERT(pcbinfo);
908 NET_EPOCH_ASSERT();
909
910 hdr = &pcbinfo->ipi_lbgrouphashbase[
911 INP_PCBPORTHASH(lport, pcbinfo->ipi_lbgrouphashmask)];
912
913 /*
914 * Search for an LB group match based on the following criteria:
915 * - prefer jailed groups to non-jailed groups
916 * - prefer exact source address matches to wildcard matches
917 * - prefer groups bound to the specified NUMA domain
918 */
919 jail_exact = jail_wild = local_exact = local_wild = NULL;
920 CK_LIST_FOREACH(grp, hdr, il_list) {
921 bool injail;
922
923 #ifdef INET
924 if (!(grp->il_vflag & INP_IPV6))
925 continue;
926 #endif
927 if (grp->il_lport != lport)
928 continue;
929
930 injail = prison_flag(grp->il_cred, PR_IP6) != 0;
931 if (injail && prison_check_ip6_locked(grp->il_cred->cr_prison,
932 laddr) != 0)
933 continue;
934
935 if (IN6_ARE_ADDR_EQUAL(&grp->il6_laddr, laddr)) {
936 if (injail) {
937 jail_exact = grp;
938 if (in6_pcblookup_lb_match(grp, domain, fib))
939 /* This is a perfect match. */
940 goto out;
941 } else if (local_exact == NULL ||
942 in6_pcblookup_lb_match(grp, domain, fib)) {
943 local_exact = grp;
944 }
945 } else if (IN6_IS_ADDR_UNSPECIFIED(&grp->il6_laddr)) {
946 if (injail) {
947 if (jail_wild == NULL ||
948 in6_pcblookup_lb_match(grp, domain, fib))
949 jail_wild = grp;
950 } else if (local_wild == NULL ||
951 in6_pcblookup_lb_match(grp, domain, fib)) {
952 local_wild = grp;
953 }
954 }
955 }
956
957 if (jail_exact != NULL)
958 grp = jail_exact;
959 else if (jail_wild != NULL)
960 grp = jail_wild;
961 else if (local_exact != NULL)
962 grp = local_exact;
963 else
964 grp = local_wild;
965 if (grp == NULL)
966 return (NULL);
967 out:
968 /*
969 * Synchronize with in_pcblbgroup_insert().
970 */
971 count = atomic_load_acq_int(&grp->il_inpcnt);
972 if (count == 0)
973 return (NULL);
974 inp = grp->il_inp[INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport) % count];
975 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
976 return (inp);
977 }
978
979 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)980 in6_pcblookup_exact_match(const struct inpcb *inp, const struct in6_addr *faddr,
981 u_short fport, const struct in6_addr *laddr, u_short lport)
982 {
983 /* XXX inp locking */
984 if ((inp->inp_vflag & INP_IPV6) == 0)
985 return (false);
986 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
987 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
988 inp->inp_fport == fport && inp->inp_lport == lport)
989 return (true);
990 return (false);
991 }
992
993 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)994 in6_pcblookup_hash_exact(struct inpcbinfo *pcbinfo,
995 const struct in6_addr *faddr, u_short fport,
996 const struct in6_addr *laddr, u_short lport)
997 {
998 struct inpcbhead *head;
999 struct inpcb *inp;
1000
1001 INP_HASH_LOCK_ASSERT(pcbinfo);
1002
1003 /*
1004 * First look for an exact match.
1005 */
1006 head = &pcbinfo->ipi_hash_exact[INP6_PCBHASH(faddr, lport, fport,
1007 pcbinfo->ipi_hashmask)];
1008 CK_LIST_FOREACH(inp, head, inp_hash_exact) {
1009 if (in6_pcblookup_exact_match(inp, faddr, fport, laddr, lport))
1010 return (inp);
1011 }
1012 return (NULL);
1013 }
1014
1015 typedef enum {
1016 INPLOOKUP_MATCH_NONE = 0,
1017 INPLOOKUP_MATCH_WILD = 1,
1018 INPLOOKUP_MATCH_LADDR = 2,
1019 } inp_lookup_match_t;
1020
1021 static inp_lookup_match_t
in6_pcblookup_wild_match(const struct inpcb * inp,const struct in6_addr * laddr,u_short lport,int fib)1022 in6_pcblookup_wild_match(const struct inpcb *inp, const struct in6_addr *laddr,
1023 u_short lport, int fib)
1024 {
1025 /* XXX inp locking */
1026 if ((inp->inp_vflag & INP_IPV6) == 0)
1027 return (INPLOOKUP_MATCH_NONE);
1028 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1029 inp->inp_lport != lport)
1030 return (INPLOOKUP_MATCH_NONE);
1031 if (fib != RT_ALL_FIBS && inp->inp_inc.inc_fibnum != fib)
1032 return (INPLOOKUP_MATCH_NONE);
1033 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1034 return (INPLOOKUP_MATCH_WILD);
1035 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr))
1036 return (INPLOOKUP_MATCH_LADDR);
1037 return (INPLOOKUP_MATCH_NONE);
1038 }
1039
1040 #define INP_LOOKUP_AGAIN ((struct inpcb *)(uintptr_t)-1)
1041
1042 static struct inpcb *
in6_pcblookup_hash_wild_smr(struct inpcbinfo * pcbinfo,const struct in6_addr * laddr,u_short lport,int fib,const inp_lookup_t lockflags)1043 in6_pcblookup_hash_wild_smr(struct inpcbinfo *pcbinfo,
1044 const struct in6_addr *laddr, u_short lport, int fib,
1045 const inp_lookup_t lockflags)
1046 {
1047 struct inpcbhead *head;
1048 struct inpcb *inp;
1049
1050 KASSERT(SMR_ENTERED(pcbinfo->ipi_smr),
1051 ("%s: not in SMR read section", __func__));
1052
1053 head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
1054 pcbinfo->ipi_hashmask)];
1055 CK_LIST_FOREACH(inp, head, inp_hash_wild) {
1056 inp_lookup_match_t match;
1057
1058 match = in6_pcblookup_wild_match(inp, laddr, lport, fib);
1059 if (match == INPLOOKUP_MATCH_NONE)
1060 continue;
1061
1062 if (__predict_true(inp_smr_lock(inp, lockflags))) {
1063 match = in6_pcblookup_wild_match(inp, laddr, lport,
1064 fib);
1065 if (match != INPLOOKUP_MATCH_NONE &&
1066 prison_check_ip6_locked(inp->inp_cred->cr_prison,
1067 laddr) == 0)
1068 return (inp);
1069 inp_unlock(inp, lockflags);
1070 }
1071
1072 /*
1073 * The matching socket disappeared out from under us. Fall back
1074 * to a serialized lookup.
1075 */
1076 return (INP_LOOKUP_AGAIN);
1077 }
1078 return (NULL);
1079 }
1080
1081 static struct inpcb *
in6_pcblookup_hash_wild_locked(struct inpcbinfo * pcbinfo,const struct in6_addr * laddr,u_short lport,int fib)1082 in6_pcblookup_hash_wild_locked(struct inpcbinfo *pcbinfo,
1083 const struct in6_addr *laddr, u_short lport, int fib)
1084 {
1085 struct inpcbhead *head;
1086 struct inpcb *inp, *jail_wild, *local_exact, *local_wild;
1087
1088 INP_HASH_LOCK_ASSERT(pcbinfo);
1089
1090 /*
1091 * Order of socket selection - we always prefer jails.
1092 * 1. jailed, non-wild.
1093 * 2. jailed, wild.
1094 * 3. non-jailed, non-wild.
1095 * 4. non-jailed, wild.
1096 */
1097 head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport,
1098 pcbinfo->ipi_hashmask)];
1099 local_wild = local_exact = jail_wild = NULL;
1100 CK_LIST_FOREACH(inp, head, inp_hash_wild) {
1101 inp_lookup_match_t match;
1102 bool injail;
1103
1104 match = in6_pcblookup_wild_match(inp, laddr, lport, fib);
1105 if (match == INPLOOKUP_MATCH_NONE)
1106 continue;
1107
1108 injail = prison_flag(inp->inp_cred, PR_IP6) != 0;
1109 if (injail) {
1110 if (prison_check_ip6_locked(
1111 inp->inp_cred->cr_prison, laddr) != 0)
1112 continue;
1113 } else {
1114 if (local_exact != NULL)
1115 continue;
1116 }
1117
1118 if (match == INPLOOKUP_MATCH_LADDR) {
1119 if (injail)
1120 return (inp);
1121 else
1122 local_exact = inp;
1123 } else {
1124 if (injail)
1125 jail_wild = inp;
1126 else
1127 local_wild = inp;
1128 }
1129 }
1130
1131 if (jail_wild != NULL)
1132 return (jail_wild);
1133 if (local_exact != NULL)
1134 return (local_exact);
1135 if (local_wild != NULL)
1136 return (local_wild);
1137 return (NULL);
1138 }
1139
1140 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,int fib)1141 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo,
1142 const struct in6_addr *faddr, u_int fport_arg,
1143 const struct in6_addr *laddr, u_int lport_arg,
1144 int lookupflags, uint8_t numa_domain, int fib)
1145 {
1146 struct inpcb *inp;
1147 u_short fport = fport_arg, lport = lport_arg;
1148
1149 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD | INPLOOKUP_FIB)) == 0,
1150 ("%s: invalid lookup flags %d", __func__, lookupflags));
1151 KASSERT(!IN6_IS_ADDR_UNSPECIFIED(faddr),
1152 ("%s: invalid foreign address", __func__));
1153 KASSERT(!IN6_IS_ADDR_UNSPECIFIED(laddr),
1154 ("%s: invalid local address", __func__));
1155 INP_HASH_LOCK_ASSERT(pcbinfo);
1156
1157 inp = in6_pcblookup_hash_exact(pcbinfo, faddr, fport, laddr, lport);
1158 if (inp != NULL)
1159 return (inp);
1160
1161 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1162 inp = in6_pcblookup_lbgroup(pcbinfo, faddr, fport, laddr,
1163 lport, numa_domain, fib);
1164 if (inp == NULL) {
1165 inp = in6_pcblookup_hash_wild_locked(pcbinfo,
1166 laddr, lport, fib);
1167 }
1168 }
1169 return (inp);
1170 }
1171
1172 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,int fib)1173 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1174 u_int fport, const struct in6_addr *laddr, u_int lport, int lookupflags,
1175 uint8_t numa_domain, int fib)
1176 {
1177 struct inpcb *inp;
1178 const inp_lookup_t lockflags = lookupflags & INPLOOKUP_LOCKMASK;
1179
1180 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1181 ("%s: LOCKPCB not set", __func__));
1182
1183 INP_HASH_WLOCK(pcbinfo);
1184 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1185 lookupflags & ~INPLOOKUP_LOCKMASK, numa_domain, fib);
1186 if (inp != NULL && !inp_trylock(inp, lockflags)) {
1187 in_pcbref(inp);
1188 INP_HASH_WUNLOCK(pcbinfo);
1189 inp_lock(inp, lockflags);
1190 if (in_pcbrele(inp, lockflags))
1191 /* XXX-MJ or retry until we get a negative match? */
1192 inp = NULL;
1193 } else {
1194 INP_HASH_WUNLOCK(pcbinfo);
1195 }
1196 return (inp);
1197 }
1198
1199 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,int fib)1200 in6_pcblookup_hash_smr(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1201 u_int fport_arg, const struct in6_addr *laddr, u_int lport_arg,
1202 int lookupflags, uint8_t numa_domain, int fib)
1203 {
1204 struct inpcb *inp;
1205 const inp_lookup_t lockflags = lookupflags & INPLOOKUP_LOCKMASK;
1206 const u_short fport = fport_arg, lport = lport_arg;
1207
1208 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1209 ("%s: invalid lookup flags %d", __func__, lookupflags));
1210 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1211 ("%s: LOCKPCB not set", __func__));
1212
1213 smr_enter(pcbinfo->ipi_smr);
1214 inp = in6_pcblookup_hash_exact(pcbinfo, faddr, fport, laddr, lport);
1215 if (inp != NULL) {
1216 if (__predict_true(inp_smr_lock(inp, lockflags))) {
1217 if (__predict_true(in6_pcblookup_exact_match(inp,
1218 faddr, fport, laddr, lport)))
1219 return (inp);
1220 inp_unlock(inp, lockflags);
1221 }
1222 /*
1223 * We failed to lock the inpcb, or its connection state changed
1224 * out from under us. Fall back to a precise search.
1225 */
1226 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1227 lookupflags, numa_domain, fib));
1228 }
1229
1230 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1231 inp = in6_pcblookup_lbgroup(pcbinfo, faddr, fport,
1232 laddr, lport, numa_domain, fib);
1233 if (inp != NULL) {
1234 if (__predict_true(inp_smr_lock(inp, lockflags))) {
1235 if (__predict_true(in6_pcblookup_wild_match(inp,
1236 laddr, lport, fib) != INPLOOKUP_MATCH_NONE))
1237 return (inp);
1238 inp_unlock(inp, lockflags);
1239 }
1240 inp = INP_LOOKUP_AGAIN;
1241 } else {
1242 inp = in6_pcblookup_hash_wild_smr(pcbinfo, laddr, lport,
1243 fib, lockflags);
1244 }
1245 if (inp == INP_LOOKUP_AGAIN) {
1246 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr,
1247 lport, lookupflags, numa_domain, fib));
1248 }
1249 }
1250
1251 if (inp == NULL)
1252 smr_exit(pcbinfo->ipi_smr);
1253
1254 return (inp);
1255 }
1256
1257 /*
1258 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1259 * from which a pre-calculated hash value may be extracted.
1260 */
1261 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)1262 in6_pcblookup(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1263 u_int fport, const struct in6_addr *laddr, u_int lport, int lookupflags,
1264 struct ifnet *ifp)
1265 {
1266 int fib;
1267
1268 fib = (lookupflags & INPLOOKUP_FIB) ? if_getfib(ifp) : RT_ALL_FIBS;
1269 return (in6_pcblookup_hash_smr(pcbinfo, faddr, fport, laddr, lport,
1270 lookupflags, M_NODOM, fib));
1271 }
1272
1273 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)1274 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, const struct in6_addr *faddr,
1275 u_int fport, const struct in6_addr *laddr, u_int lport, int lookupflags,
1276 struct ifnet *ifp __unused, struct mbuf *m)
1277 {
1278 int fib;
1279
1280 M_ASSERTPKTHDR(m);
1281 fib = (lookupflags & INPLOOKUP_FIB) ? M_GETFIB(m) : RT_ALL_FIBS;
1282 return (in6_pcblookup_hash_smr(pcbinfo, faddr, fport, laddr, lport,
1283 lookupflags, m->m_pkthdr.numa_domain, fib));
1284 }
1285
1286 void
init_sin6(struct sockaddr_in6 * sin6,struct mbuf * m,int srcordst)1287 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m, int srcordst)
1288 {
1289 struct ip6_hdr *ip;
1290
1291 ip = mtod(m, struct ip6_hdr *);
1292 bzero(sin6, sizeof(*sin6));
1293 sin6->sin6_len = sizeof(*sin6);
1294 sin6->sin6_family = AF_INET6;
1295 sin6->sin6_addr = srcordst ? ip->ip6_dst : ip->ip6_src;
1296
1297 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */
1298
1299 return;
1300 }
1301