1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * Copyright (c) 2010-2011 Juniper Networks, Inc.
4 * All rights reserved.
5 *
6 * Portions of this software were developed by Robert N. M. Watson under
7 * contract to Juniper Networks, Inc.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $
34 */
35
36/*-
37 * Copyright (c) 1982, 1986, 1991, 1993
38 * The Regents of the University of California. All rights reserved.
39 *
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
42 * are met:
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 4. Neither the name of the University nor the names of its contributors
49 * may be used to endorse or promote products derived from this software
50 * without specific prior written permission.
51 *
52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * SUCH DAMAGE.
63 *
64 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
65 */
66
67#include <sys/cdefs.h>
| 1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * Copyright (c) 2010-2011 Juniper Networks, Inc.
4 * All rights reserved.
5 *
6 * Portions of this software were developed by Robert N. M. Watson under
7 * contract to Juniper Networks, Inc.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $
34 */
35
36/*-
37 * Copyright (c) 1982, 1986, 1991, 1993
38 * The Regents of the University of California. All rights reserved.
39 *
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
42 * are met:
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 4. Neither the name of the University nor the names of its contributors
49 * may be used to endorse or promote products derived from this software
50 * without specific prior written permission.
51 *
52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * SUCH DAMAGE.
63 *
64 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
65 */
66
67#include <sys/cdefs.h>
|
68__FBSDID("$FreeBSD: head/sys/netinet6/in6_pcb.c 268562 2014-07-12 05:46:33Z adrian $");
| 68__FBSDID("$FreeBSD: head/sys/netinet6/in6_pcb.c 271386 2014-09-10 12:35:42Z ae $");
|
69
70#include "opt_inet.h"
71#include "opt_inet6.h"
72#include "opt_ipsec.h"
73#include "opt_pcbgroup.h"
74#include "opt_rss.h"
75
76#include <sys/param.h>
77#include <sys/systm.h>
78#include <sys/malloc.h>
79#include <sys/mbuf.h>
80#include <sys/domain.h>
81#include <sys/protosw.h>
82#include <sys/socket.h>
83#include <sys/socketvar.h>
84#include <sys/sockio.h>
85#include <sys/errno.h>
86#include <sys/time.h>
87#include <sys/priv.h>
88#include <sys/proc.h>
89#include <sys/jail.h>
90
91#include <vm/uma.h>
92
93#include <net/if.h>
94#include <net/if_var.h>
95#include <net/if_types.h>
96#include <net/route.h>
97
98#include <netinet/in.h>
99#include <netinet/in_var.h>
100#include <netinet/in_systm.h>
101#include <netinet/tcp_var.h>
102#include <netinet/ip6.h>
103#include <netinet/ip_var.h>
104
105#include <netinet6/ip6_var.h>
106#include <netinet6/nd6.h>
107#include <netinet/in_pcb.h>
108#include <netinet6/in6_pcb.h>
109#include <netinet6/scope6_var.h>
110
111int
112in6_pcbbind(register struct inpcb *inp, struct sockaddr *nam,
113 struct ucred *cred)
114{
115 struct socket *so = inp->inp_socket;
116 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
117 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
118 u_short lport = 0;
119 int error, lookupflags = 0;
120 int reuseport = (so->so_options & SO_REUSEPORT);
121
122 INP_WLOCK_ASSERT(inp);
123 INP_HASH_WLOCK_ASSERT(pcbinfo);
124
125 if (TAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */
126 return (EADDRNOTAVAIL);
127 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
128 return (EINVAL);
129 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
130 lookupflags = INPLOOKUP_WILDCARD;
131 if (nam == NULL) {
132 if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
133 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
134 return (error);
135 } else {
136 sin6 = (struct sockaddr_in6 *)nam;
137 if (nam->sa_len != sizeof(*sin6))
138 return (EINVAL);
139 /*
140 * family check.
141 */
142 if (nam->sa_family != AF_INET6)
143 return (EAFNOSUPPORT);
144
145 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
146 return(error);
147
148 if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
149 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
150 return (error);
151
152 lport = sin6->sin6_port;
153 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
154 /*
155 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
156 * allow compepte duplication of binding if
157 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
158 * and a multicast address is bound on both
159 * new and duplicated sockets.
160 */
161 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
162 reuseport = SO_REUSEADDR|SO_REUSEPORT;
163 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
164 struct ifaddr *ifa;
165
166 sin6->sin6_port = 0; /* yech... */
167 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) ==
168 NULL &&
169 (inp->inp_flags & INP_BINDANY) == 0) {
170 return (EADDRNOTAVAIL);
171 }
172
173 /*
174 * XXX: bind to an anycast address might accidentally
175 * cause sending a packet with anycast source address.
176 * We should allow to bind to a deprecated address, since
177 * the application dares to use it.
178 */
179 if (ifa != NULL &&
180 ((struct in6_ifaddr *)ifa)->ia6_flags &
181 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
182 ifa_free(ifa);
183 return (EADDRNOTAVAIL);
184 }
185 if (ifa != NULL)
186 ifa_free(ifa);
187 }
188 if (lport) {
189 struct inpcb *t;
190 struct tcptw *tw;
191
192 /* GROSS */
193 if (ntohs(lport) <= V_ipport_reservedhigh &&
194 ntohs(lport) >= V_ipport_reservedlow &&
195 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
196 0))
197 return (EACCES);
198 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
199 priv_check_cred(inp->inp_cred,
200 PRIV_NETINET_REUSEPORT, 0) != 0) {
201 t = in6_pcblookup_local(pcbinfo,
202 &sin6->sin6_addr, lport,
203 INPLOOKUP_WILDCARD, cred);
204 if (t &&
205 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
206 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
207 (so->so_type != SOCK_STREAM ||
208 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
209 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
210 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
211 (t->inp_flags2 & INP_REUSEPORT) == 0) &&
212 (inp->inp_cred->cr_uid !=
213 t->inp_cred->cr_uid))
214 return (EADDRINUSE);
215
216 /*
217 * If the socket is a BINDMULTI socket, then
218 * the credentials need to match and the
219 * original socket also has to have been bound
220 * with BINDMULTI.
221 */
222 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
223 return (EADDRINUSE);
224
225#ifdef INET
226 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
227 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
228 struct sockaddr_in sin;
229
230 in6_sin6_2_sin(&sin, sin6);
231 t = in_pcblookup_local(pcbinfo,
232 sin.sin_addr, lport,
233 INPLOOKUP_WILDCARD, cred);
234 if (t &&
235 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
236 ((t->inp_flags &
237 INP_TIMEWAIT) == 0) &&
238 (so->so_type != SOCK_STREAM ||
239 ntohl(t->inp_faddr.s_addr) ==
240 INADDR_ANY) &&
241 (inp->inp_cred->cr_uid !=
242 t->inp_cred->cr_uid))
243 return (EADDRINUSE);
244
245 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
246 return (EADDRINUSE);
247 }
248#endif
249 }
250 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
251 lport, lookupflags, cred);
252 if (t && (t->inp_flags & INP_TIMEWAIT)) {
253 /*
254 * XXXRW: If an incpb has had its timewait
255 * state recycled, we treat the address as
256 * being in use (for now). This is better
257 * than a panic, but not desirable.
258 */
259 tw = intotw(t);
260 if (tw == NULL ||
261 (reuseport & tw->tw_so_options) == 0)
262 return (EADDRINUSE);
263 } else if (t && (reuseport & inp_so_options(t)) == 0) {
264 return (EADDRINUSE);
265 }
266#ifdef INET
267 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
268 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
269 struct sockaddr_in sin;
270
271 in6_sin6_2_sin(&sin, sin6);
272 t = in_pcblookup_local(pcbinfo, sin.sin_addr,
273 lport, lookupflags, cred);
274 if (t && t->inp_flags & INP_TIMEWAIT) {
275 tw = intotw(t);
276 if (tw == NULL)
277 return (EADDRINUSE);
278 if ((reuseport & tw->tw_so_options) == 0
279 && (ntohl(t->inp_laddr.s_addr) !=
280 INADDR_ANY || ((inp->inp_vflag &
281 INP_IPV6PROTO) ==
282 (t->inp_vflag & INP_IPV6PROTO))))
283 return (EADDRINUSE);
284 } else if (t &&
285 (reuseport & inp_so_options(t)) == 0 &&
286 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
287 (t->inp_vflag & INP_IPV6PROTO) != 0))
288 return (EADDRINUSE);
289 }
290#endif
291 }
292 inp->in6p_laddr = sin6->sin6_addr;
293 }
294 if (lport == 0) {
295 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
296 /* Undo an address bind that may have occurred. */
297 inp->in6p_laddr = in6addr_any;
298 return (error);
299 }
300 } else {
301 inp->inp_lport = lport;
302 if (in_pcbinshash(inp) != 0) {
303 inp->in6p_laddr = in6addr_any;
304 inp->inp_lport = 0;
305 return (EAGAIN);
306 }
307 }
308 return (0);
309}
310
311/*
312 * Transform old in6_pcbconnect() into an inner subroutine for new
313 * in6_pcbconnect(): Do some validity-checking on the remote
314 * address (in mbuf 'nam') and then determine local host address
315 * (i.e., which interface) to use to access that remote host.
316 *
317 * This preserves definition of in6_pcbconnect(), while supporting a
318 * slightly different version for T/TCP. (This is more than
319 * a bit of a kludge, but cleaning up the internal interfaces would
320 * have forced minor changes in every protocol).
321 */
322int
323in6_pcbladdr(register struct inpcb *inp, struct sockaddr *nam,
324 struct in6_addr *plocal_addr6)
325{
326 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
327 int error = 0;
328 struct ifnet *ifp = NULL;
329 int scope_ambiguous = 0;
330 struct in6_addr in6a;
331
332 INP_WLOCK_ASSERT(inp);
333 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */
334
335 if (nam->sa_len != sizeof (*sin6))
336 return (EINVAL);
337 if (sin6->sin6_family != AF_INET6)
338 return (EAFNOSUPPORT);
339 if (sin6->sin6_port == 0)
340 return (EADDRNOTAVAIL);
341
342 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
343 scope_ambiguous = 1;
344 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
345 return(error);
346
347 if (!TAILQ_EMPTY(&V_in6_ifaddrhead)) {
348 /*
349 * If the destination address is UNSPECIFIED addr,
350 * use the loopback addr, e.g ::1.
351 */
352 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
353 sin6->sin6_addr = in6addr_loopback;
354 }
355 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
356 return (error);
357
358 error = in6_selectsrc(sin6, inp->in6p_outputopts,
359 inp, NULL, inp->inp_cred, &ifp, &in6a);
360 if (error)
361 return (error);
362
363 if (ifp && scope_ambiguous &&
364 (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) {
365 return(error);
366 }
367
368 /*
369 * Do not update this earlier, in case we return with an error.
370 *
371 * XXX: this in6_selectsrc result might replace the bound local
372 * address with the address specified by setsockopt(IPV6_PKTINFO).
373 * Is it the intended behavior?
374 */
375 *plocal_addr6 = in6a;
376
377 /*
378 * Don't do pcblookup call here; return interface in
379 * plocal_addr6
380 * and exit to caller, that will do the lookup.
381 */
382
383 return (0);
384}
385
386/*
387 * Outer subroutine:
388 * Connect from a socket to a specified address.
389 * Both address and port must be specified in argument sin.
390 * If don't have a local address for this socket yet,
391 * then pick one.
392 */
393int
394in6_pcbconnect_mbuf(register struct inpcb *inp, struct sockaddr *nam,
395 struct ucred *cred, struct mbuf *m)
396{
397 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
398 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
399 struct in6_addr addr6;
400 int error;
401
402 INP_WLOCK_ASSERT(inp);
403 INP_HASH_WLOCK_ASSERT(pcbinfo);
404
405 /*
406 * Call inner routine, to assign local interface address.
407 * in6_pcbladdr() may automatically fill in sin6_scope_id.
408 */
409 if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0)
410 return (error);
411
412 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
413 sin6->sin6_port,
414 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
415 ? &addr6 : &inp->in6p_laddr,
416 inp->inp_lport, 0, NULL) != NULL) {
417 return (EADDRINUSE);
418 }
419 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
420 if (inp->inp_lport == 0) {
421 error = in6_pcbbind(inp, (struct sockaddr *)0, cred);
422 if (error)
423 return (error);
424 }
425 inp->in6p_laddr = addr6;
426 }
427 inp->in6p_faddr = sin6->sin6_addr;
428 inp->inp_fport = sin6->sin6_port;
429 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
430 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
431 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
432 inp->inp_flow |=
433 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
434
435 in_pcbrehash_mbuf(inp, m);
436
437 return (0);
438}
439
440int
441in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
442{
443
444 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL));
445}
446
447void
448in6_pcbdisconnect(struct inpcb *inp)
449{
450
451 INP_WLOCK_ASSERT(inp);
452 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
453
454 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
455 inp->inp_fport = 0;
456 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
457 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
458 in_pcbrehash(inp);
459}
460
461struct sockaddr *
462in6_sockaddr(in_port_t port, struct in6_addr *addr_p)
463{
464 struct sockaddr_in6 *sin6;
465
466 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK);
467 bzero(sin6, sizeof *sin6);
468 sin6->sin6_family = AF_INET6;
469 sin6->sin6_len = sizeof(*sin6);
470 sin6->sin6_port = port;
471 sin6->sin6_addr = *addr_p;
472 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */
473
474 return (struct sockaddr *)sin6;
475}
476
477struct sockaddr *
478in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p)
479{
480 struct sockaddr_in sin;
481 struct sockaddr_in6 *sin6_p;
482
483 bzero(&sin, sizeof sin);
484 sin.sin_family = AF_INET;
485 sin.sin_len = sizeof(sin);
486 sin.sin_port = port;
487 sin.sin_addr = *addr_p;
488
489 sin6_p = malloc(sizeof *sin6_p, M_SONAME,
490 M_WAITOK);
491 in6_sin_2_v4mapsin6(&sin, sin6_p);
492
493 return (struct sockaddr *)sin6_p;
494}
495
496int
497in6_getsockaddr(struct socket *so, struct sockaddr **nam)
498{
499 register struct inpcb *inp;
500 struct in6_addr addr;
501 in_port_t port;
502
503 inp = sotoinpcb(so);
504 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
505
506 INP_RLOCK(inp);
507 port = inp->inp_lport;
508 addr = inp->in6p_laddr;
509 INP_RUNLOCK(inp);
510
511 *nam = in6_sockaddr(port, &addr);
512 return 0;
513}
514
515int
516in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
517{
518 struct inpcb *inp;
519 struct in6_addr addr;
520 in_port_t port;
521
522 inp = sotoinpcb(so);
523 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
524
525 INP_RLOCK(inp);
526 port = inp->inp_fport;
527 addr = inp->in6p_faddr;
528 INP_RUNLOCK(inp);
529
530 *nam = in6_sockaddr(port, &addr);
531 return 0;
532}
533
534int
535in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
536{
537 struct inpcb *inp;
538 int error;
539
540 inp = sotoinpcb(so);
541 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
542
543#ifdef INET
544 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
545 error = in_getsockaddr(so, nam);
546 if (error == 0)
547 in6_sin_2_v4mapsin6_in_sock(nam);
548 } else
549#endif
550 {
551 /* scope issues will be handled in in6_getsockaddr(). */
552 error = in6_getsockaddr(so, nam);
553 }
554
555 return error;
556}
557
558int
559in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
560{
561 struct inpcb *inp;
562 int error;
563
564 inp = sotoinpcb(so);
565 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
566
567#ifdef INET
568 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
569 error = in_getpeeraddr(so, nam);
570 if (error == 0)
571 in6_sin_2_v4mapsin6_in_sock(nam);
572 } else
573#endif
574 /* scope issues will be handled in in6_getpeeraddr(). */
575 error = in6_getpeeraddr(so, nam);
576
577 return error;
578}
579
580/*
581 * Pass some notification to all connections of a protocol
582 * associated with address dst. The local address and/or port numbers
583 * may be specified to limit the search. The "usual action" will be
584 * taken, depending on the ctlinput cmd. The caller must filter any
585 * cmds that are uninteresting (e.g., no error in the map).
586 * Call the protocol specific routine (if any) to report
587 * any errors for each matching socket.
588 */
589void
590in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst,
591 u_int fport_arg, const struct sockaddr *src, u_int lport_arg,
592 int cmd, void *cmdarg,
593 struct inpcb *(*notify)(struct inpcb *, int))
594{
595 struct inpcb *inp, *inp_temp;
596 struct sockaddr_in6 sa6_src, *sa6_dst;
597 u_short fport = fport_arg, lport = lport_arg;
598 u_int32_t flowinfo;
599 int errno;
600
601 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
602 return;
603
604 sa6_dst = (struct sockaddr_in6 *)dst;
605 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
606 return;
607
608 /*
609 * note that src can be NULL when we get notify by local fragmentation.
610 */
611 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
612 flowinfo = sa6_src.sin6_flowinfo;
613
614 /*
615 * Redirects go to all references to the destination,
616 * and use in6_rtchange to invalidate the route cache.
617 * Dead host indications: also use in6_rtchange to invalidate
618 * the cache, and deliver the error to all the sockets.
619 * Otherwise, if we have knowledge of the local port and address,
620 * deliver only to that socket.
621 */
622 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
623 fport = 0;
624 lport = 0;
625 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
626
627 if (cmd != PRC_HOSTDEAD)
628 notify = in6_rtchange;
629 }
630 errno = inet6ctlerrmap[cmd];
631 INP_INFO_WLOCK(pcbinfo);
632 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
633 INP_WLOCK(inp);
634 if ((inp->inp_vflag & INP_IPV6) == 0) {
635 INP_WUNLOCK(inp);
636 continue;
637 }
638
639 /*
640 * If the error designates a new path MTU for a destination
641 * and the application (associated with this socket) wanted to
642 * know the value, notify. Note that we notify for all
643 * disconnected sockets if the corresponding application
644 * wanted. This is because some UDP applications keep sending
645 * sockets disconnected.
646 * XXX: should we avoid to notify the value to TCP sockets?
647 */
648 if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 &&
649 (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
650 IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) {
651 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst,
652 (u_int32_t *)cmdarg);
653 }
654
655 /*
656 * Detect if we should notify the error. If no source and
657 * destination ports are specifed, 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 INP_WUNLOCK(inp);
677 continue;
678 }
679
680 do_notify:
681 if (notify) {
682 if ((*notify)(inp, errno))
683 INP_WUNLOCK(inp);
684 } else
685 INP_WUNLOCK(inp);
686 }
687 INP_INFO_WUNLOCK(pcbinfo);
688}
689
690/*
691 * Lookup a PCB based on the local address and port. Caller must hold the
692 * hash lock. No inpcb locks or references are acquired.
693 */
694struct inpcb *
695in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
696 u_short lport, int lookupflags, struct ucred *cred)
697{
698 register struct inpcb *inp;
699 int matchwild = 3, wildcard;
700
701 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
702 ("%s: invalid lookup flags %d", __func__, lookupflags));
703
704 INP_HASH_WLOCK_ASSERT(pcbinfo);
705
706 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
707 struct inpcbhead *head;
708 /*
709 * Look for an unconnected (wildcard foreign addr) PCB that
710 * matches the local address and port we're looking for.
711 */
| 69
70#include "opt_inet.h"
71#include "opt_inet6.h"
72#include "opt_ipsec.h"
73#include "opt_pcbgroup.h"
74#include "opt_rss.h"
75
76#include <sys/param.h>
77#include <sys/systm.h>
78#include <sys/malloc.h>
79#include <sys/mbuf.h>
80#include <sys/domain.h>
81#include <sys/protosw.h>
82#include <sys/socket.h>
83#include <sys/socketvar.h>
84#include <sys/sockio.h>
85#include <sys/errno.h>
86#include <sys/time.h>
87#include <sys/priv.h>
88#include <sys/proc.h>
89#include <sys/jail.h>
90
91#include <vm/uma.h>
92
93#include <net/if.h>
94#include <net/if_var.h>
95#include <net/if_types.h>
96#include <net/route.h>
97
98#include <netinet/in.h>
99#include <netinet/in_var.h>
100#include <netinet/in_systm.h>
101#include <netinet/tcp_var.h>
102#include <netinet/ip6.h>
103#include <netinet/ip_var.h>
104
105#include <netinet6/ip6_var.h>
106#include <netinet6/nd6.h>
107#include <netinet/in_pcb.h>
108#include <netinet6/in6_pcb.h>
109#include <netinet6/scope6_var.h>
110
111int
112in6_pcbbind(register struct inpcb *inp, struct sockaddr *nam,
113 struct ucred *cred)
114{
115 struct socket *so = inp->inp_socket;
116 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
117 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
118 u_short lport = 0;
119 int error, lookupflags = 0;
120 int reuseport = (so->so_options & SO_REUSEPORT);
121
122 INP_WLOCK_ASSERT(inp);
123 INP_HASH_WLOCK_ASSERT(pcbinfo);
124
125 if (TAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */
126 return (EADDRNOTAVAIL);
127 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
128 return (EINVAL);
129 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
130 lookupflags = INPLOOKUP_WILDCARD;
131 if (nam == NULL) {
132 if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
133 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
134 return (error);
135 } else {
136 sin6 = (struct sockaddr_in6 *)nam;
137 if (nam->sa_len != sizeof(*sin6))
138 return (EINVAL);
139 /*
140 * family check.
141 */
142 if (nam->sa_family != AF_INET6)
143 return (EAFNOSUPPORT);
144
145 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
146 return(error);
147
148 if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
149 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
150 return (error);
151
152 lport = sin6->sin6_port;
153 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
154 /*
155 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
156 * allow compepte duplication of binding if
157 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
158 * and a multicast address is bound on both
159 * new and duplicated sockets.
160 */
161 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
162 reuseport = SO_REUSEADDR|SO_REUSEPORT;
163 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
164 struct ifaddr *ifa;
165
166 sin6->sin6_port = 0; /* yech... */
167 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) ==
168 NULL &&
169 (inp->inp_flags & INP_BINDANY) == 0) {
170 return (EADDRNOTAVAIL);
171 }
172
173 /*
174 * XXX: bind to an anycast address might accidentally
175 * cause sending a packet with anycast source address.
176 * We should allow to bind to a deprecated address, since
177 * the application dares to use it.
178 */
179 if (ifa != NULL &&
180 ((struct in6_ifaddr *)ifa)->ia6_flags &
181 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
182 ifa_free(ifa);
183 return (EADDRNOTAVAIL);
184 }
185 if (ifa != NULL)
186 ifa_free(ifa);
187 }
188 if (lport) {
189 struct inpcb *t;
190 struct tcptw *tw;
191
192 /* GROSS */
193 if (ntohs(lport) <= V_ipport_reservedhigh &&
194 ntohs(lport) >= V_ipport_reservedlow &&
195 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
196 0))
197 return (EACCES);
198 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
199 priv_check_cred(inp->inp_cred,
200 PRIV_NETINET_REUSEPORT, 0) != 0) {
201 t = in6_pcblookup_local(pcbinfo,
202 &sin6->sin6_addr, lport,
203 INPLOOKUP_WILDCARD, cred);
204 if (t &&
205 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
206 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
207 (so->so_type != SOCK_STREAM ||
208 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
209 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
210 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
211 (t->inp_flags2 & INP_REUSEPORT) == 0) &&
212 (inp->inp_cred->cr_uid !=
213 t->inp_cred->cr_uid))
214 return (EADDRINUSE);
215
216 /*
217 * If the socket is a BINDMULTI socket, then
218 * the credentials need to match and the
219 * original socket also has to have been bound
220 * with BINDMULTI.
221 */
222 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
223 return (EADDRINUSE);
224
225#ifdef INET
226 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
227 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
228 struct sockaddr_in sin;
229
230 in6_sin6_2_sin(&sin, sin6);
231 t = in_pcblookup_local(pcbinfo,
232 sin.sin_addr, lport,
233 INPLOOKUP_WILDCARD, cred);
234 if (t &&
235 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
236 ((t->inp_flags &
237 INP_TIMEWAIT) == 0) &&
238 (so->so_type != SOCK_STREAM ||
239 ntohl(t->inp_faddr.s_addr) ==
240 INADDR_ANY) &&
241 (inp->inp_cred->cr_uid !=
242 t->inp_cred->cr_uid))
243 return (EADDRINUSE);
244
245 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
246 return (EADDRINUSE);
247 }
248#endif
249 }
250 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
251 lport, lookupflags, cred);
252 if (t && (t->inp_flags & INP_TIMEWAIT)) {
253 /*
254 * XXXRW: If an incpb has had its timewait
255 * state recycled, we treat the address as
256 * being in use (for now). This is better
257 * than a panic, but not desirable.
258 */
259 tw = intotw(t);
260 if (tw == NULL ||
261 (reuseport & tw->tw_so_options) == 0)
262 return (EADDRINUSE);
263 } else if (t && (reuseport & inp_so_options(t)) == 0) {
264 return (EADDRINUSE);
265 }
266#ifdef INET
267 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
268 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
269 struct sockaddr_in sin;
270
271 in6_sin6_2_sin(&sin, sin6);
272 t = in_pcblookup_local(pcbinfo, sin.sin_addr,
273 lport, lookupflags, cred);
274 if (t && t->inp_flags & INP_TIMEWAIT) {
275 tw = intotw(t);
276 if (tw == NULL)
277 return (EADDRINUSE);
278 if ((reuseport & tw->tw_so_options) == 0
279 && (ntohl(t->inp_laddr.s_addr) !=
280 INADDR_ANY || ((inp->inp_vflag &
281 INP_IPV6PROTO) ==
282 (t->inp_vflag & INP_IPV6PROTO))))
283 return (EADDRINUSE);
284 } else if (t &&
285 (reuseport & inp_so_options(t)) == 0 &&
286 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
287 (t->inp_vflag & INP_IPV6PROTO) != 0))
288 return (EADDRINUSE);
289 }
290#endif
291 }
292 inp->in6p_laddr = sin6->sin6_addr;
293 }
294 if (lport == 0) {
295 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
296 /* Undo an address bind that may have occurred. */
297 inp->in6p_laddr = in6addr_any;
298 return (error);
299 }
300 } else {
301 inp->inp_lport = lport;
302 if (in_pcbinshash(inp) != 0) {
303 inp->in6p_laddr = in6addr_any;
304 inp->inp_lport = 0;
305 return (EAGAIN);
306 }
307 }
308 return (0);
309}
310
311/*
312 * Transform old in6_pcbconnect() into an inner subroutine for new
313 * in6_pcbconnect(): Do some validity-checking on the remote
314 * address (in mbuf 'nam') and then determine local host address
315 * (i.e., which interface) to use to access that remote host.
316 *
317 * This preserves definition of in6_pcbconnect(), while supporting a
318 * slightly different version for T/TCP. (This is more than
319 * a bit of a kludge, but cleaning up the internal interfaces would
320 * have forced minor changes in every protocol).
321 */
322int
323in6_pcbladdr(register struct inpcb *inp, struct sockaddr *nam,
324 struct in6_addr *plocal_addr6)
325{
326 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
327 int error = 0;
328 struct ifnet *ifp = NULL;
329 int scope_ambiguous = 0;
330 struct in6_addr in6a;
331
332 INP_WLOCK_ASSERT(inp);
333 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */
334
335 if (nam->sa_len != sizeof (*sin6))
336 return (EINVAL);
337 if (sin6->sin6_family != AF_INET6)
338 return (EAFNOSUPPORT);
339 if (sin6->sin6_port == 0)
340 return (EADDRNOTAVAIL);
341
342 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
343 scope_ambiguous = 1;
344 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
345 return(error);
346
347 if (!TAILQ_EMPTY(&V_in6_ifaddrhead)) {
348 /*
349 * If the destination address is UNSPECIFIED addr,
350 * use the loopback addr, e.g ::1.
351 */
352 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
353 sin6->sin6_addr = in6addr_loopback;
354 }
355 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
356 return (error);
357
358 error = in6_selectsrc(sin6, inp->in6p_outputopts,
359 inp, NULL, inp->inp_cred, &ifp, &in6a);
360 if (error)
361 return (error);
362
363 if (ifp && scope_ambiguous &&
364 (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) {
365 return(error);
366 }
367
368 /*
369 * Do not update this earlier, in case we return with an error.
370 *
371 * XXX: this in6_selectsrc result might replace the bound local
372 * address with the address specified by setsockopt(IPV6_PKTINFO).
373 * Is it the intended behavior?
374 */
375 *plocal_addr6 = in6a;
376
377 /*
378 * Don't do pcblookup call here; return interface in
379 * plocal_addr6
380 * and exit to caller, that will do the lookup.
381 */
382
383 return (0);
384}
385
386/*
387 * Outer subroutine:
388 * Connect from a socket to a specified address.
389 * Both address and port must be specified in argument sin.
390 * If don't have a local address for this socket yet,
391 * then pick one.
392 */
393int
394in6_pcbconnect_mbuf(register struct inpcb *inp, struct sockaddr *nam,
395 struct ucred *cred, struct mbuf *m)
396{
397 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
398 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
399 struct in6_addr addr6;
400 int error;
401
402 INP_WLOCK_ASSERT(inp);
403 INP_HASH_WLOCK_ASSERT(pcbinfo);
404
405 /*
406 * Call inner routine, to assign local interface address.
407 * in6_pcbladdr() may automatically fill in sin6_scope_id.
408 */
409 if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0)
410 return (error);
411
412 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
413 sin6->sin6_port,
414 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
415 ? &addr6 : &inp->in6p_laddr,
416 inp->inp_lport, 0, NULL) != NULL) {
417 return (EADDRINUSE);
418 }
419 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
420 if (inp->inp_lport == 0) {
421 error = in6_pcbbind(inp, (struct sockaddr *)0, cred);
422 if (error)
423 return (error);
424 }
425 inp->in6p_laddr = addr6;
426 }
427 inp->in6p_faddr = sin6->sin6_addr;
428 inp->inp_fport = sin6->sin6_port;
429 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
430 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
431 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
432 inp->inp_flow |=
433 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
434
435 in_pcbrehash_mbuf(inp, m);
436
437 return (0);
438}
439
440int
441in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
442{
443
444 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL));
445}
446
447void
448in6_pcbdisconnect(struct inpcb *inp)
449{
450
451 INP_WLOCK_ASSERT(inp);
452 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
453
454 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
455 inp->inp_fport = 0;
456 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
457 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
458 in_pcbrehash(inp);
459}
460
461struct sockaddr *
462in6_sockaddr(in_port_t port, struct in6_addr *addr_p)
463{
464 struct sockaddr_in6 *sin6;
465
466 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK);
467 bzero(sin6, sizeof *sin6);
468 sin6->sin6_family = AF_INET6;
469 sin6->sin6_len = sizeof(*sin6);
470 sin6->sin6_port = port;
471 sin6->sin6_addr = *addr_p;
472 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */
473
474 return (struct sockaddr *)sin6;
475}
476
477struct sockaddr *
478in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p)
479{
480 struct sockaddr_in sin;
481 struct sockaddr_in6 *sin6_p;
482
483 bzero(&sin, sizeof sin);
484 sin.sin_family = AF_INET;
485 sin.sin_len = sizeof(sin);
486 sin.sin_port = port;
487 sin.sin_addr = *addr_p;
488
489 sin6_p = malloc(sizeof *sin6_p, M_SONAME,
490 M_WAITOK);
491 in6_sin_2_v4mapsin6(&sin, sin6_p);
492
493 return (struct sockaddr *)sin6_p;
494}
495
496int
497in6_getsockaddr(struct socket *so, struct sockaddr **nam)
498{
499 register struct inpcb *inp;
500 struct in6_addr addr;
501 in_port_t port;
502
503 inp = sotoinpcb(so);
504 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
505
506 INP_RLOCK(inp);
507 port = inp->inp_lport;
508 addr = inp->in6p_laddr;
509 INP_RUNLOCK(inp);
510
511 *nam = in6_sockaddr(port, &addr);
512 return 0;
513}
514
515int
516in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
517{
518 struct inpcb *inp;
519 struct in6_addr addr;
520 in_port_t port;
521
522 inp = sotoinpcb(so);
523 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
524
525 INP_RLOCK(inp);
526 port = inp->inp_fport;
527 addr = inp->in6p_faddr;
528 INP_RUNLOCK(inp);
529
530 *nam = in6_sockaddr(port, &addr);
531 return 0;
532}
533
534int
535in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
536{
537 struct inpcb *inp;
538 int error;
539
540 inp = sotoinpcb(so);
541 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
542
543#ifdef INET
544 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
545 error = in_getsockaddr(so, nam);
546 if (error == 0)
547 in6_sin_2_v4mapsin6_in_sock(nam);
548 } else
549#endif
550 {
551 /* scope issues will be handled in in6_getsockaddr(). */
552 error = in6_getsockaddr(so, nam);
553 }
554
555 return error;
556}
557
558int
559in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
560{
561 struct inpcb *inp;
562 int error;
563
564 inp = sotoinpcb(so);
565 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
566
567#ifdef INET
568 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
569 error = in_getpeeraddr(so, nam);
570 if (error == 0)
571 in6_sin_2_v4mapsin6_in_sock(nam);
572 } else
573#endif
574 /* scope issues will be handled in in6_getpeeraddr(). */
575 error = in6_getpeeraddr(so, nam);
576
577 return error;
578}
579
580/*
581 * Pass some notification to all connections of a protocol
582 * associated with address dst. The local address and/or port numbers
583 * may be specified to limit the search. The "usual action" will be
584 * taken, depending on the ctlinput cmd. The caller must filter any
585 * cmds that are uninteresting (e.g., no error in the map).
586 * Call the protocol specific routine (if any) to report
587 * any errors for each matching socket.
588 */
589void
590in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst,
591 u_int fport_arg, const struct sockaddr *src, u_int lport_arg,
592 int cmd, void *cmdarg,
593 struct inpcb *(*notify)(struct inpcb *, int))
594{
595 struct inpcb *inp, *inp_temp;
596 struct sockaddr_in6 sa6_src, *sa6_dst;
597 u_short fport = fport_arg, lport = lport_arg;
598 u_int32_t flowinfo;
599 int errno;
600
601 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
602 return;
603
604 sa6_dst = (struct sockaddr_in6 *)dst;
605 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
606 return;
607
608 /*
609 * note that src can be NULL when we get notify by local fragmentation.
610 */
611 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
612 flowinfo = sa6_src.sin6_flowinfo;
613
614 /*
615 * Redirects go to all references to the destination,
616 * and use in6_rtchange to invalidate the route cache.
617 * Dead host indications: also use in6_rtchange to invalidate
618 * the cache, and deliver the error to all the sockets.
619 * Otherwise, if we have knowledge of the local port and address,
620 * deliver only to that socket.
621 */
622 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
623 fport = 0;
624 lport = 0;
625 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
626
627 if (cmd != PRC_HOSTDEAD)
628 notify = in6_rtchange;
629 }
630 errno = inet6ctlerrmap[cmd];
631 INP_INFO_WLOCK(pcbinfo);
632 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
633 INP_WLOCK(inp);
634 if ((inp->inp_vflag & INP_IPV6) == 0) {
635 INP_WUNLOCK(inp);
636 continue;
637 }
638
639 /*
640 * If the error designates a new path MTU for a destination
641 * and the application (associated with this socket) wanted to
642 * know the value, notify. Note that we notify for all
643 * disconnected sockets if the corresponding application
644 * wanted. This is because some UDP applications keep sending
645 * sockets disconnected.
646 * XXX: should we avoid to notify the value to TCP sockets?
647 */
648 if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 &&
649 (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
650 IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) {
651 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst,
652 (u_int32_t *)cmdarg);
653 }
654
655 /*
656 * Detect if we should notify the error. If no source and
657 * destination ports are specifed, 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 INP_WUNLOCK(inp);
677 continue;
678 }
679
680 do_notify:
681 if (notify) {
682 if ((*notify)(inp, errno))
683 INP_WUNLOCK(inp);
684 } else
685 INP_WUNLOCK(inp);
686 }
687 INP_INFO_WUNLOCK(pcbinfo);
688}
689
690/*
691 * Lookup a PCB based on the local address and port. Caller must hold the
692 * hash lock. No inpcb locks or references are acquired.
693 */
694struct inpcb *
695in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
696 u_short lport, int lookupflags, struct ucred *cred)
697{
698 register struct inpcb *inp;
699 int matchwild = 3, wildcard;
700
701 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
702 ("%s: invalid lookup flags %d", __func__, lookupflags));
703
704 INP_HASH_WLOCK_ASSERT(pcbinfo);
705
706 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
707 struct inpcbhead *head;
708 /*
709 * Look for an unconnected (wildcard foreign addr) PCB that
710 * matches the local address and port we're looking for.
711 */
|
712 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
713 0, pcbinfo->ipi_hashmask)];
| 712 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
713 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
714 pcbinfo->ipi_hashmask)];
|
714 LIST_FOREACH(inp, head, inp_hash) {
715 /* XXX inp locking */
716 if ((inp->inp_vflag & INP_IPV6) == 0)
717 continue;
718 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
719 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
720 inp->inp_lport == lport) {
721 /* Found. */
722 if (cred == NULL ||
723 prison_equal_ip6(cred->cr_prison,
724 inp->inp_cred->cr_prison))
725 return (inp);
726 }
727 }
728 /*
729 * Not found.
730 */
731 return (NULL);
732 } else {
733 struct inpcbporthead *porthash;
734 struct inpcbport *phd;
735 struct inpcb *match = NULL;
736 /*
737 * Best fit PCB lookup.
738 *
739 * First see if this local port is in use by looking on the
740 * port hash list.
741 */
742 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
743 pcbinfo->ipi_porthashmask)];
744 LIST_FOREACH(phd, porthash, phd_hash) {
745 if (phd->phd_port == lport)
746 break;
747 }
748 if (phd != NULL) {
749 /*
750 * Port is in use by one or more PCBs. Look for best
751 * fit.
752 */
753 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
754 wildcard = 0;
755 if (cred != NULL &&
756 !prison_equal_ip6(cred->cr_prison,
757 inp->inp_cred->cr_prison))
758 continue;
759 /* XXX inp locking */
760 if ((inp->inp_vflag & INP_IPV6) == 0)
761 continue;
762 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
763 wildcard++;
764 if (!IN6_IS_ADDR_UNSPECIFIED(
765 &inp->in6p_laddr)) {
766 if (IN6_IS_ADDR_UNSPECIFIED(laddr))
767 wildcard++;
768 else if (!IN6_ARE_ADDR_EQUAL(
769 &inp->in6p_laddr, laddr))
770 continue;
771 } else {
772 if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
773 wildcard++;
774 }
775 if (wildcard < matchwild) {
776 match = inp;
777 matchwild = wildcard;
778 if (matchwild == 0)
779 break;
780 }
781 }
782 }
783 return (match);
784 }
785}
786
787void
788in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
789{
790 struct inpcb *in6p;
791 struct ip6_moptions *im6o;
792 int i, gap;
793
794 INP_INFO_RLOCK(pcbinfo);
795 LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) {
796 INP_WLOCK(in6p);
797 im6o = in6p->in6p_moptions;
798 if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) {
799 /*
800 * Unselect the outgoing ifp for multicast if it
801 * is being detached.
802 */
803 if (im6o->im6o_multicast_ifp == ifp)
804 im6o->im6o_multicast_ifp = NULL;
805 /*
806 * Drop multicast group membership if we joined
807 * through the interface being detached.
808 */
809 gap = 0;
810 for (i = 0; i < im6o->im6o_num_memberships; i++) {
811 if (im6o->im6o_membership[i]->in6m_ifp ==
812 ifp) {
813 in6_mc_leave(im6o->im6o_membership[i],
814 NULL);
815 gap++;
816 } else if (gap != 0) {
817 im6o->im6o_membership[i - gap] =
818 im6o->im6o_membership[i];
819 }
820 }
821 im6o->im6o_num_memberships -= gap;
822 }
823 INP_WUNLOCK(in6p);
824 }
825 INP_INFO_RUNLOCK(pcbinfo);
826}
827
828/*
829 * Check for alternatives when higher level complains
830 * about service problems. For now, invalidate cached
831 * routing information. If the route was created dynamically
832 * (by a redirect), time to try a default gateway again.
833 */
834void
835in6_losing(struct inpcb *in6p)
836{
837
838 /*
839 * We don't store route pointers in the routing table anymore
840 */
841 return;
842}
843
844/*
845 * After a routing change, flush old routing
846 * and allocate a (hopefully) better one.
847 */
848struct inpcb *
849in6_rtchange(struct inpcb *inp, int errno)
850{
851 /*
852 * We don't store route pointers in the routing table anymore
853 */
854 return inp;
855}
856
857#ifdef PCBGROUP
858/*
859 * Lookup PCB in hash list, using pcbgroup tables.
860 */
861static struct inpcb *
862in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
863 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr,
864 u_int lport_arg, int lookupflags, struct ifnet *ifp)
865{
866 struct inpcbhead *head;
867 struct inpcb *inp, *tmpinp;
868 u_short fport = fport_arg, lport = lport_arg;
869 int faith;
870
871 if (faithprefix_p != NULL)
872 faith = (*faithprefix_p)(laddr);
873 else
874 faith = 0;
875
876 /*
877 * First look for an exact match.
878 */
879 tmpinp = NULL;
880 INP_GROUP_LOCK(pcbgroup);
| 715 LIST_FOREACH(inp, head, inp_hash) {
716 /* XXX inp locking */
717 if ((inp->inp_vflag & INP_IPV6) == 0)
718 continue;
719 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
720 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
721 inp->inp_lport == lport) {
722 /* Found. */
723 if (cred == NULL ||
724 prison_equal_ip6(cred->cr_prison,
725 inp->inp_cred->cr_prison))
726 return (inp);
727 }
728 }
729 /*
730 * Not found.
731 */
732 return (NULL);
733 } else {
734 struct inpcbporthead *porthash;
735 struct inpcbport *phd;
736 struct inpcb *match = NULL;
737 /*
738 * Best fit PCB lookup.
739 *
740 * First see if this local port is in use by looking on the
741 * port hash list.
742 */
743 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
744 pcbinfo->ipi_porthashmask)];
745 LIST_FOREACH(phd, porthash, phd_hash) {
746 if (phd->phd_port == lport)
747 break;
748 }
749 if (phd != NULL) {
750 /*
751 * Port is in use by one or more PCBs. Look for best
752 * fit.
753 */
754 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
755 wildcard = 0;
756 if (cred != NULL &&
757 !prison_equal_ip6(cred->cr_prison,
758 inp->inp_cred->cr_prison))
759 continue;
760 /* XXX inp locking */
761 if ((inp->inp_vflag & INP_IPV6) == 0)
762 continue;
763 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
764 wildcard++;
765 if (!IN6_IS_ADDR_UNSPECIFIED(
766 &inp->in6p_laddr)) {
767 if (IN6_IS_ADDR_UNSPECIFIED(laddr))
768 wildcard++;
769 else if (!IN6_ARE_ADDR_EQUAL(
770 &inp->in6p_laddr, laddr))
771 continue;
772 } else {
773 if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
774 wildcard++;
775 }
776 if (wildcard < matchwild) {
777 match = inp;
778 matchwild = wildcard;
779 if (matchwild == 0)
780 break;
781 }
782 }
783 }
784 return (match);
785 }
786}
787
788void
789in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
790{
791 struct inpcb *in6p;
792 struct ip6_moptions *im6o;
793 int i, gap;
794
795 INP_INFO_RLOCK(pcbinfo);
796 LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) {
797 INP_WLOCK(in6p);
798 im6o = in6p->in6p_moptions;
799 if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) {
800 /*
801 * Unselect the outgoing ifp for multicast if it
802 * is being detached.
803 */
804 if (im6o->im6o_multicast_ifp == ifp)
805 im6o->im6o_multicast_ifp = NULL;
806 /*
807 * Drop multicast group membership if we joined
808 * through the interface being detached.
809 */
810 gap = 0;
811 for (i = 0; i < im6o->im6o_num_memberships; i++) {
812 if (im6o->im6o_membership[i]->in6m_ifp ==
813 ifp) {
814 in6_mc_leave(im6o->im6o_membership[i],
815 NULL);
816 gap++;
817 } else if (gap != 0) {
818 im6o->im6o_membership[i - gap] =
819 im6o->im6o_membership[i];
820 }
821 }
822 im6o->im6o_num_memberships -= gap;
823 }
824 INP_WUNLOCK(in6p);
825 }
826 INP_INFO_RUNLOCK(pcbinfo);
827}
828
829/*
830 * Check for alternatives when higher level complains
831 * about service problems. For now, invalidate cached
832 * routing information. If the route was created dynamically
833 * (by a redirect), time to try a default gateway again.
834 */
835void
836in6_losing(struct inpcb *in6p)
837{
838
839 /*
840 * We don't store route pointers in the routing table anymore
841 */
842 return;
843}
844
845/*
846 * After a routing change, flush old routing
847 * and allocate a (hopefully) better one.
848 */
849struct inpcb *
850in6_rtchange(struct inpcb *inp, int errno)
851{
852 /*
853 * We don't store route pointers in the routing table anymore
854 */
855 return inp;
856}
857
858#ifdef PCBGROUP
859/*
860 * Lookup PCB in hash list, using pcbgroup tables.
861 */
862static struct inpcb *
863in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
864 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr,
865 u_int lport_arg, int lookupflags, struct ifnet *ifp)
866{
867 struct inpcbhead *head;
868 struct inpcb *inp, *tmpinp;
869 u_short fport = fport_arg, lport = lport_arg;
870 int faith;
871
872 if (faithprefix_p != NULL)
873 faith = (*faithprefix_p)(laddr);
874 else
875 faith = 0;
876
877 /*
878 * First look for an exact match.
879 */
880 tmpinp = NULL;
881 INP_GROUP_LOCK(pcbgroup);
|
881 head = &pcbgroup->ipg_hashbase[
882 INP_PCBHASH(faddr->s6_addr32[3] /* XXX */, lport, fport,
883 pcbgroup->ipg_hashmask)];
| 882 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(
883 INP6_PCBHASHKEY(faddr), lport, fport, pcbgroup->ipg_hashmask)];
|
884 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
885 /* XXX inp locking */
886 if ((inp->inp_vflag & INP_IPV6) == 0)
887 continue;
888 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
889 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
890 inp->inp_fport == fport &&
891 inp->inp_lport == lport) {
892 /*
893 * XXX We should be able to directly return
894 * the inp here, without any checks.
895 * Well unless both bound with SO_REUSEPORT?
896 */
897 if (prison_flag(inp->inp_cred, PR_IP6))
898 goto found;
899 if (tmpinp == NULL)
900 tmpinp = inp;
901 }
902 }
903 if (tmpinp != NULL) {
904 inp = tmpinp;
905 goto found;
906 }
907
908 /*
909 * Then look for a wildcard match in the pcbgroup.
910 */
911 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
912 struct inpcb *local_wild = NULL, *local_exact = NULL;
913 struct inpcb *jail_wild = NULL;
914 int injail;
915
916 /*
917 * Order of socket selection - we always prefer jails.
918 * 1. jailed, non-wild.
919 * 2. jailed, wild.
920 * 3. non-jailed, non-wild.
921 * 4. non-jailed, wild.
922 */
923 head = &pcbgroup->ipg_hashbase[
924 INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)];
925 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
926 /* XXX inp locking */
927 if ((inp->inp_vflag & INP_IPV6) == 0)
928 continue;
929
930 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
931 inp->inp_lport != lport) {
932 continue;
933 }
934
935 /* XXX inp locking */
936 if (faith && (inp->inp_flags & INP_FAITH) == 0)
937 continue;
938
939 injail = prison_flag(inp->inp_cred, PR_IP6);
940 if (injail) {
941 if (prison_check_ip6(inp->inp_cred,
942 laddr) != 0)
943 continue;
944 } else {
945 if (local_exact != NULL)
946 continue;
947 }
948
949 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
950 if (injail)
951 goto found;
952 else
953 local_exact = inp;
954 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
955 if (injail)
956 jail_wild = inp;
957 else
958 local_wild = inp;
959 }
960 } /* LIST_FOREACH */
961
962 inp = jail_wild;
963 if (inp == NULL)
964 inp = jail_wild;
965 if (inp == NULL)
966 inp = local_exact;
967 if (inp == NULL)
968 inp = local_wild;
969 if (inp != NULL)
970 goto found;
971 }
972
973 /*
974 * Then look for a wildcard match, if requested.
975 */
976 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
977 struct inpcb *local_wild = NULL, *local_exact = NULL;
978 struct inpcb *jail_wild = NULL;
979 int injail;
980
981 /*
982 * Order of socket selection - we always prefer jails.
983 * 1. jailed, non-wild.
984 * 2. jailed, wild.
985 * 3. non-jailed, non-wild.
986 * 4. non-jailed, wild.
987 */
| 884 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
885 /* XXX inp locking */
886 if ((inp->inp_vflag & INP_IPV6) == 0)
887 continue;
888 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
889 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
890 inp->inp_fport == fport &&
891 inp->inp_lport == lport) {
892 /*
893 * XXX We should be able to directly return
894 * the inp here, without any checks.
895 * Well unless both bound with SO_REUSEPORT?
896 */
897 if (prison_flag(inp->inp_cred, PR_IP6))
898 goto found;
899 if (tmpinp == NULL)
900 tmpinp = inp;
901 }
902 }
903 if (tmpinp != NULL) {
904 inp = tmpinp;
905 goto found;
906 }
907
908 /*
909 * Then look for a wildcard match in the pcbgroup.
910 */
911 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
912 struct inpcb *local_wild = NULL, *local_exact = NULL;
913 struct inpcb *jail_wild = NULL;
914 int injail;
915
916 /*
917 * Order of socket selection - we always prefer jails.
918 * 1. jailed, non-wild.
919 * 2. jailed, wild.
920 * 3. non-jailed, non-wild.
921 * 4. non-jailed, wild.
922 */
923 head = &pcbgroup->ipg_hashbase[
924 INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)];
925 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
926 /* XXX inp locking */
927 if ((inp->inp_vflag & INP_IPV6) == 0)
928 continue;
929
930 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
931 inp->inp_lport != lport) {
932 continue;
933 }
934
935 /* XXX inp locking */
936 if (faith && (inp->inp_flags & INP_FAITH) == 0)
937 continue;
938
939 injail = prison_flag(inp->inp_cred, PR_IP6);
940 if (injail) {
941 if (prison_check_ip6(inp->inp_cred,
942 laddr) != 0)
943 continue;
944 } else {
945 if (local_exact != NULL)
946 continue;
947 }
948
949 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
950 if (injail)
951 goto found;
952 else
953 local_exact = inp;
954 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
955 if (injail)
956 jail_wild = inp;
957 else
958 local_wild = inp;
959 }
960 } /* LIST_FOREACH */
961
962 inp = jail_wild;
963 if (inp == NULL)
964 inp = jail_wild;
965 if (inp == NULL)
966 inp = local_exact;
967 if (inp == NULL)
968 inp = local_wild;
969 if (inp != NULL)
970 goto found;
971 }
972
973 /*
974 * Then look for a wildcard match, if requested.
975 */
976 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
977 struct inpcb *local_wild = NULL, *local_exact = NULL;
978 struct inpcb *jail_wild = NULL;
979 int injail;
980
981 /*
982 * Order of socket selection - we always prefer jails.
983 * 1. jailed, non-wild.
984 * 2. jailed, wild.
985 * 3. non-jailed, non-wild.
986 * 4. non-jailed, wild.
987 */
|
988 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport,
989 0, pcbinfo->ipi_wildmask)];
| 988 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(
989 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
990 pcbinfo->ipi_wildmask)];
|
990 LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
991 /* XXX inp locking */
992 if ((inp->inp_vflag & INP_IPV6) == 0)
993 continue;
994
995 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
996 inp->inp_lport != lport) {
997 continue;
998 }
999
1000 /* XXX inp locking */
1001 if (faith && (inp->inp_flags & INP_FAITH) == 0)
1002 continue;
1003
1004 injail = prison_flag(inp->inp_cred, PR_IP6);
1005 if (injail) {
1006 if (prison_check_ip6(inp->inp_cred,
1007 laddr) != 0)
1008 continue;
1009 } else {
1010 if (local_exact != NULL)
1011 continue;
1012 }
1013
1014 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1015 if (injail)
1016 goto found;
1017 else
1018 local_exact = inp;
1019 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1020 if (injail)
1021 jail_wild = inp;
1022 else
1023 local_wild = inp;
1024 }
1025 } /* LIST_FOREACH */
1026
1027 inp = jail_wild;
1028 if (inp == NULL)
1029 inp = jail_wild;
1030 if (inp == NULL)
1031 inp = local_exact;
1032 if (inp == NULL)
1033 inp = local_wild;
1034 if (inp != NULL)
1035 goto found;
1036 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1037 INP_GROUP_UNLOCK(pcbgroup);
1038 return (NULL);
1039
1040found:
1041 in_pcbref(inp);
1042 INP_GROUP_UNLOCK(pcbgroup);
1043 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1044 INP_WLOCK(inp);
1045 if (in_pcbrele_wlocked(inp))
1046 return (NULL);
1047 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1048 INP_RLOCK(inp);
1049 if (in_pcbrele_rlocked(inp))
1050 return (NULL);
1051 } else
1052 panic("%s: locking buf", __func__);
1053 return (inp);
1054}
1055#endif /* PCBGROUP */
1056
1057/*
1058 * Lookup PCB in hash list.
1059 */
1060struct inpcb *
1061in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1062 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
1063 int lookupflags, struct ifnet *ifp)
1064{
1065 struct inpcbhead *head;
1066 struct inpcb *inp, *tmpinp;
1067 u_short fport = fport_arg, lport = lport_arg;
1068 int faith;
1069
1070 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1071 ("%s: invalid lookup flags %d", __func__, lookupflags));
1072
1073 INP_HASH_LOCK_ASSERT(pcbinfo);
1074
1075 if (faithprefix_p != NULL)
1076 faith = (*faithprefix_p)(laddr);
1077 else
1078 faith = 0;
1079
1080 /*
1081 * First look for an exact match.
1082 */
1083 tmpinp = NULL;
| 991 LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
992 /* XXX inp locking */
993 if ((inp->inp_vflag & INP_IPV6) == 0)
994 continue;
995
996 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
997 inp->inp_lport != lport) {
998 continue;
999 }
1000
1001 /* XXX inp locking */
1002 if (faith && (inp->inp_flags & INP_FAITH) == 0)
1003 continue;
1004
1005 injail = prison_flag(inp->inp_cred, PR_IP6);
1006 if (injail) {
1007 if (prison_check_ip6(inp->inp_cred,
1008 laddr) != 0)
1009 continue;
1010 } else {
1011 if (local_exact != NULL)
1012 continue;
1013 }
1014
1015 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1016 if (injail)
1017 goto found;
1018 else
1019 local_exact = inp;
1020 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1021 if (injail)
1022 jail_wild = inp;
1023 else
1024 local_wild = inp;
1025 }
1026 } /* LIST_FOREACH */
1027
1028 inp = jail_wild;
1029 if (inp == NULL)
1030 inp = jail_wild;
1031 if (inp == NULL)
1032 inp = local_exact;
1033 if (inp == NULL)
1034 inp = local_wild;
1035 if (inp != NULL)
1036 goto found;
1037 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1038 INP_GROUP_UNLOCK(pcbgroup);
1039 return (NULL);
1040
1041found:
1042 in_pcbref(inp);
1043 INP_GROUP_UNLOCK(pcbgroup);
1044 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1045 INP_WLOCK(inp);
1046 if (in_pcbrele_wlocked(inp))
1047 return (NULL);
1048 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1049 INP_RLOCK(inp);
1050 if (in_pcbrele_rlocked(inp))
1051 return (NULL);
1052 } else
1053 panic("%s: locking buf", __func__);
1054 return (inp);
1055}
1056#endif /* PCBGROUP */
1057
1058/*
1059 * Lookup PCB in hash list.
1060 */
1061struct inpcb *
1062in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1063 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
1064 int lookupflags, struct ifnet *ifp)
1065{
1066 struct inpcbhead *head;
1067 struct inpcb *inp, *tmpinp;
1068 u_short fport = fport_arg, lport = lport_arg;
1069 int faith;
1070
1071 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1072 ("%s: invalid lookup flags %d", __func__, lookupflags));
1073
1074 INP_HASH_LOCK_ASSERT(pcbinfo);
1075
1076 if (faithprefix_p != NULL)
1077 faith = (*faithprefix_p)(laddr);
1078 else
1079 faith = 0;
1080
1081 /*
1082 * First look for an exact match.
1083 */
1084 tmpinp = NULL;
|
1084 head = &pcbinfo->ipi_hashbase[
1085 INP_PCBHASH(faddr->s6_addr32[3] /* XXX */, lport, fport,
1086 pcbinfo->ipi_hashmask)];
| 1085 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1086 INP6_PCBHASHKEY(faddr), lport, fport, pcbinfo->ipi_hashmask)];
|
1087 LIST_FOREACH(inp, head, inp_hash) {
1088 /* XXX inp locking */
1089 if ((inp->inp_vflag & INP_IPV6) == 0)
1090 continue;
1091 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1092 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1093 inp->inp_fport == fport &&
1094 inp->inp_lport == lport) {
1095 /*
1096 * XXX We should be able to directly return
1097 * the inp here, without any checks.
1098 * Well unless both bound with SO_REUSEPORT?
1099 */
1100 if (prison_flag(inp->inp_cred, PR_IP6))
1101 return (inp);
1102 if (tmpinp == NULL)
1103 tmpinp = inp;
1104 }
1105 }
1106 if (tmpinp != NULL)
1107 return (tmpinp);
1108
1109 /*
1110 * Then look for a wildcard match, if requested.
1111 */
1112 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1113 struct inpcb *local_wild = NULL, *local_exact = NULL;
1114 struct inpcb *jail_wild = NULL;
1115 int injail;
1116
1117 /*
1118 * Order of socket selection - we always prefer jails.
1119 * 1. jailed, non-wild.
1120 * 2. jailed, wild.
1121 * 3. non-jailed, non-wild.
1122 * 4. non-jailed, wild.
1123 */
| 1087 LIST_FOREACH(inp, head, inp_hash) {
1088 /* XXX inp locking */
1089 if ((inp->inp_vflag & INP_IPV6) == 0)
1090 continue;
1091 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1092 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1093 inp->inp_fport == fport &&
1094 inp->inp_lport == lport) {
1095 /*
1096 * XXX We should be able to directly return
1097 * the inp here, without any checks.
1098 * Well unless both bound with SO_REUSEPORT?
1099 */
1100 if (prison_flag(inp->inp_cred, PR_IP6))
1101 return (inp);
1102 if (tmpinp == NULL)
1103 tmpinp = inp;
1104 }
1105 }
1106 if (tmpinp != NULL)
1107 return (tmpinp);
1108
1109 /*
1110 * Then look for a wildcard match, if requested.
1111 */
1112 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1113 struct inpcb *local_wild = NULL, *local_exact = NULL;
1114 struct inpcb *jail_wild = NULL;
1115 int injail;
1116
1117 /*
1118 * Order of socket selection - we always prefer jails.
1119 * 1. jailed, non-wild.
1120 * 2. jailed, wild.
1121 * 3. non-jailed, non-wild.
1122 * 4. non-jailed, wild.
1123 */
|
1124 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1125 0, pcbinfo->ipi_hashmask)];
| 1124 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1125 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
1126 pcbinfo->ipi_hashmask)];
|
1126 LIST_FOREACH(inp, head, inp_hash) {
1127 /* XXX inp locking */
1128 if ((inp->inp_vflag & INP_IPV6) == 0)
1129 continue;
1130
1131 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1132 inp->inp_lport != lport) {
1133 continue;
1134 }
1135
1136 /* XXX inp locking */
1137 if (faith && (inp->inp_flags & INP_FAITH) == 0)
1138 continue;
1139
1140 injail = prison_flag(inp->inp_cred, PR_IP6);
1141 if (injail) {
1142 if (prison_check_ip6(inp->inp_cred,
1143 laddr) != 0)
1144 continue;
1145 } else {
1146 if (local_exact != NULL)
1147 continue;
1148 }
1149
1150 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1151 if (injail)
1152 return (inp);
1153 else
1154 local_exact = inp;
1155 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1156 if (injail)
1157 jail_wild = inp;
1158 else
1159 local_wild = inp;
1160 }
1161 } /* LIST_FOREACH */
1162
1163 if (jail_wild != NULL)
1164 return (jail_wild);
1165 if (local_exact != NULL)
1166 return (local_exact);
1167 if (local_wild != NULL)
1168 return (local_wild);
1169 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1170
1171 /*
1172 * Not found.
1173 */
1174 return (NULL);
1175}
1176
1177/*
1178 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1179 * hash list lock, and will return the inpcb locked (i.e., requires
1180 * INPLOOKUP_LOCKPCB).
1181 */
1182static struct inpcb *
1183in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1184 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1185 struct ifnet *ifp)
1186{
1187 struct inpcb *inp;
1188
1189 INP_HASH_RLOCK(pcbinfo);
1190 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1191 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1192 if (inp != NULL) {
1193 in_pcbref(inp);
1194 INP_HASH_RUNLOCK(pcbinfo);
1195 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1196 INP_WLOCK(inp);
1197 if (in_pcbrele_wlocked(inp))
1198 return (NULL);
1199 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1200 INP_RLOCK(inp);
1201 if (in_pcbrele_rlocked(inp))
1202 return (NULL);
1203 } else
1204 panic("%s: locking bug", __func__);
1205 } else
1206 INP_HASH_RUNLOCK(pcbinfo);
1207 return (inp);
1208}
1209
1210/*
1211 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1212 * from which a pre-calculated hash value may be extracted.
1213 *
1214 * Possibly more of this logic should be in in6_pcbgroup.c.
1215 */
1216struct inpcb *
1217in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport,
1218 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1219{
1220#if defined(PCBGROUP) && !defined(RSS)
1221 struct inpcbgroup *pcbgroup;
1222#endif
1223
1224 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1225 ("%s: invalid lookup flags %d", __func__, lookupflags));
1226 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1227 ("%s: LOCKPCB not set", __func__));
1228
1229 /*
1230 * When not using RSS, use connection groups in preference to the
1231 * reservation table when looking up 4-tuples. When using RSS, just
1232 * use the reservation table, due to the cost of the Toeplitz hash
1233 * in software.
1234 *
1235 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1236 * we could be doing RSS with a non-Toeplitz hash that is affordable
1237 * in software.
1238 */
1239#if defined(PCBGROUP) && !defined(RSS)
1240 if (in_pcbgroup_enabled(pcbinfo)) {
1241 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1242 fport);
1243 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1244 laddr, lport, lookupflags, ifp));
1245 }
1246#endif
1247 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1248 lookupflags, ifp));
1249}
1250
1251struct inpcb *
1252in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1253 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1254 struct ifnet *ifp, struct mbuf *m)
1255{
1256#ifdef PCBGROUP
1257 struct inpcbgroup *pcbgroup;
1258#endif
1259
1260 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1261 ("%s: invalid lookup flags %d", __func__, lookupflags));
1262 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1263 ("%s: LOCKPCB not set", __func__));
1264
1265#ifdef PCBGROUP
1266 /*
1267 * If we can use a hardware-generated hash to look up the connection
1268 * group, use that connection group to find the inpcb. Otherwise
1269 * fall back on a software hash -- or the reservation table if we're
1270 * using RSS.
1271 *
1272 * XXXRW: As above, that policy belongs in the pcbgroup code.
1273 */
1274 if (in_pcbgroup_enabled(pcbinfo) &&
1275 !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) {
1276 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1277 m->m_pkthdr.flowid);
1278 if (pcbgroup != NULL)
1279 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr,
1280 fport, laddr, lport, lookupflags, ifp));
1281#ifndef RSS
1282 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1283 fport);
1284 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1285 laddr, lport, lookupflags, ifp));
1286#endif
1287 }
1288#endif
1289 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1290 lookupflags, ifp));
1291}
1292
1293void
1294init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
1295{
1296 struct ip6_hdr *ip;
1297
1298 ip = mtod(m, struct ip6_hdr *);
1299 bzero(sin6, sizeof(*sin6));
1300 sin6->sin6_len = sizeof(*sin6);
1301 sin6->sin6_family = AF_INET6;
1302 sin6->sin6_addr = ip->ip6_src;
1303
1304 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */
1305
1306 return;
1307}
| 1127 LIST_FOREACH(inp, head, inp_hash) {
1128 /* XXX inp locking */
1129 if ((inp->inp_vflag & INP_IPV6) == 0)
1130 continue;
1131
1132 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1133 inp->inp_lport != lport) {
1134 continue;
1135 }
1136
1137 /* XXX inp locking */
1138 if (faith && (inp->inp_flags & INP_FAITH) == 0)
1139 continue;
1140
1141 injail = prison_flag(inp->inp_cred, PR_IP6);
1142 if (injail) {
1143 if (prison_check_ip6(inp->inp_cred,
1144 laddr) != 0)
1145 continue;
1146 } else {
1147 if (local_exact != NULL)
1148 continue;
1149 }
1150
1151 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1152 if (injail)
1153 return (inp);
1154 else
1155 local_exact = inp;
1156 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1157 if (injail)
1158 jail_wild = inp;
1159 else
1160 local_wild = inp;
1161 }
1162 } /* LIST_FOREACH */
1163
1164 if (jail_wild != NULL)
1165 return (jail_wild);
1166 if (local_exact != NULL)
1167 return (local_exact);
1168 if (local_wild != NULL)
1169 return (local_wild);
1170 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1171
1172 /*
1173 * Not found.
1174 */
1175 return (NULL);
1176}
1177
1178/*
1179 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1180 * hash list lock, and will return the inpcb locked (i.e., requires
1181 * INPLOOKUP_LOCKPCB).
1182 */
1183static struct inpcb *
1184in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1185 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1186 struct ifnet *ifp)
1187{
1188 struct inpcb *inp;
1189
1190 INP_HASH_RLOCK(pcbinfo);
1191 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1192 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1193 if (inp != NULL) {
1194 in_pcbref(inp);
1195 INP_HASH_RUNLOCK(pcbinfo);
1196 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1197 INP_WLOCK(inp);
1198 if (in_pcbrele_wlocked(inp))
1199 return (NULL);
1200 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1201 INP_RLOCK(inp);
1202 if (in_pcbrele_rlocked(inp))
1203 return (NULL);
1204 } else
1205 panic("%s: locking bug", __func__);
1206 } else
1207 INP_HASH_RUNLOCK(pcbinfo);
1208 return (inp);
1209}
1210
1211/*
1212 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1213 * from which a pre-calculated hash value may be extracted.
1214 *
1215 * Possibly more of this logic should be in in6_pcbgroup.c.
1216 */
1217struct inpcb *
1218in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport,
1219 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1220{
1221#if defined(PCBGROUP) && !defined(RSS)
1222 struct inpcbgroup *pcbgroup;
1223#endif
1224
1225 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1226 ("%s: invalid lookup flags %d", __func__, lookupflags));
1227 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1228 ("%s: LOCKPCB not set", __func__));
1229
1230 /*
1231 * When not using RSS, use connection groups in preference to the
1232 * reservation table when looking up 4-tuples. When using RSS, just
1233 * use the reservation table, due to the cost of the Toeplitz hash
1234 * in software.
1235 *
1236 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1237 * we could be doing RSS with a non-Toeplitz hash that is affordable
1238 * in software.
1239 */
1240#if defined(PCBGROUP) && !defined(RSS)
1241 if (in_pcbgroup_enabled(pcbinfo)) {
1242 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1243 fport);
1244 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1245 laddr, lport, lookupflags, ifp));
1246 }
1247#endif
1248 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1249 lookupflags, ifp));
1250}
1251
1252struct inpcb *
1253in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1254 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1255 struct ifnet *ifp, struct mbuf *m)
1256{
1257#ifdef PCBGROUP
1258 struct inpcbgroup *pcbgroup;
1259#endif
1260
1261 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1262 ("%s: invalid lookup flags %d", __func__, lookupflags));
1263 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1264 ("%s: LOCKPCB not set", __func__));
1265
1266#ifdef PCBGROUP
1267 /*
1268 * If we can use a hardware-generated hash to look up the connection
1269 * group, use that connection group to find the inpcb. Otherwise
1270 * fall back on a software hash -- or the reservation table if we're
1271 * using RSS.
1272 *
1273 * XXXRW: As above, that policy belongs in the pcbgroup code.
1274 */
1275 if (in_pcbgroup_enabled(pcbinfo) &&
1276 !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) {
1277 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1278 m->m_pkthdr.flowid);
1279 if (pcbgroup != NULL)
1280 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr,
1281 fport, laddr, lport, lookupflags, ifp));
1282#ifndef RSS
1283 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1284 fport);
1285 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1286 laddr, lport, lookupflags, ifp));
1287#endif
1288 }
1289#endif
1290 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1291 lookupflags, ifp));
1292}
1293
1294void
1295init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
1296{
1297 struct ip6_hdr *ip;
1298
1299 ip = mtod(m, struct ip6_hdr *);
1300 bzero(sin6, sizeof(*sin6));
1301 sin6->sin6_len = sizeof(*sin6);
1302 sin6->sin6_family = AF_INET6;
1303 sin6->sin6_addr = ip->ip6_src;
1304
1305 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */
1306
1307 return;
1308}
|