2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Copyright (c) 2010-2011 Juniper Networks, Inc.
8 * Portions of this software were developed by Robert N. M. Watson under
9 * contract to Juniper Networks, Inc.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
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.
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
35 * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $
39 * Copyright (c) 1982, 1986, 1991, 1993
40 * The Regents of the University of California. All rights reserved.
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
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.
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
66 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
73 #include "opt_inet6.h"
74 #include "opt_ipsec.h"
75 #include "opt_pcbgroup.h"
78 #include <sys/param.h>
79 #include <sys/systm.h>
80 #include <sys/malloc.h>
82 #include <sys/domain.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/sockio.h>
87 #include <sys/errno.h>
96 #include <net/if_var.h>
97 #include <net/if_llatbl.h>
98 #include <net/if_types.h>
99 #include <net/route.h>
101 #include <netinet/in.h>
102 #include <netinet/in_var.h>
103 #include <netinet/in_systm.h>
104 #include <netinet/tcp_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet/ip_var.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet6/nd6.h>
110 #include <netinet/in_pcb.h>
111 #include <netinet6/in6_pcb.h>
112 #include <netinet6/scope6_var.h>
114 static struct inpcb *in6_pcblookup_hash_locked(struct inpcbinfo *,
115 struct in6_addr *, u_int, struct in6_addr *, u_int, int, struct ifnet *);
118 in6_pcbbind(struct inpcb *inp, struct sockaddr *nam,
121 struct socket *so = inp->inp_socket;
122 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
123 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
125 int error, lookupflags = 0;
126 int reuseport = (so->so_options & SO_REUSEPORT);
129 * XXX: Maybe we could let SO_REUSEPORT_LB set SO_REUSEPORT bit here
130 * so that we don't have to add to the (already messy) code below.
132 int reuseport_lb = (so->so_options & SO_REUSEPORT_LB);
134 INP_WLOCK_ASSERT(inp);
135 INP_HASH_WLOCK_ASSERT(pcbinfo);
137 if (CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */
138 return (EADDRNOTAVAIL);
139 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
141 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0)
142 lookupflags = INPLOOKUP_WILDCARD;
144 if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
145 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
148 sin6 = (struct sockaddr_in6 *)nam;
149 if (nam->sa_len != sizeof(*sin6))
154 if (nam->sa_family != AF_INET6)
155 return (EAFNOSUPPORT);
157 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
160 if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
161 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
164 lport = sin6->sin6_port;
165 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
167 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
168 * allow compepte duplication of binding if
169 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
170 * and a multicast address is bound on both
171 * new and duplicated sockets.
173 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
174 reuseport = SO_REUSEADDR|SO_REUSEPORT;
176 * XXX: How to deal with SO_REUSEPORT_LB here?
177 * Treat same as SO_REUSEPORT for now.
179 if ((so->so_options &
180 (SO_REUSEADDR|SO_REUSEPORT_LB)) != 0)
181 reuseport_lb = SO_REUSEADDR|SO_REUSEPORT_LB;
182 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
185 sin6->sin6_port = 0; /* yech... */
187 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) ==
189 (inp->inp_flags & INP_BINDANY) == 0) {
191 return (EADDRNOTAVAIL);
195 * XXX: bind to an anycast address might accidentally
196 * cause sending a packet with anycast source address.
197 * We should allow to bind to a deprecated address, since
198 * the application dares to use it.
201 ((struct in6_ifaddr *)ifa)->ia6_flags &
202 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
204 return (EADDRNOTAVAIL);
213 if (ntohs(lport) <= V_ipport_reservedhigh &&
214 ntohs(lport) >= V_ipport_reservedlow &&
215 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
218 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
219 priv_check_cred(inp->inp_cred,
220 PRIV_NETINET_REUSEPORT, 0) != 0) {
221 t = in6_pcblookup_local(pcbinfo,
222 &sin6->sin6_addr, lport,
223 INPLOOKUP_WILDCARD, cred);
225 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
226 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
227 (so->so_type != SOCK_STREAM ||
228 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
229 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
230 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
231 (t->inp_flags2 & INP_REUSEPORT) ||
232 (t->inp_flags2 & INP_REUSEPORT_LB) == 0) &&
233 (inp->inp_cred->cr_uid !=
234 t->inp_cred->cr_uid))
238 * If the socket is a BINDMULTI socket, then
239 * the credentials need to match and the
240 * original socket also has to have been bound
243 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
247 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
248 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
249 struct sockaddr_in sin;
251 in6_sin6_2_sin(&sin, sin6);
252 t = in_pcblookup_local(pcbinfo,
254 INPLOOKUP_WILDCARD, cred);
256 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
258 INP_TIMEWAIT) == 0) &&
259 (so->so_type != SOCK_STREAM ||
260 ntohl(t->inp_faddr.s_addr) ==
262 (inp->inp_cred->cr_uid !=
263 t->inp_cred->cr_uid))
266 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
271 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
272 lport, lookupflags, cred);
273 if (t && (t->inp_flags & INP_TIMEWAIT)) {
275 * XXXRW: If an incpb has had its timewait
276 * state recycled, we treat the address as
277 * being in use (for now). This is better
278 * than a panic, but not desirable.
282 ((reuseport & tw->tw_so_options) == 0 &&
283 (reuseport_lb & tw->tw_so_options) == 0))
285 } else if (t && (reuseport & inp_so_options(t)) == 0 &&
286 (reuseport_lb & inp_so_options(t)) == 0) {
290 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
291 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
292 struct sockaddr_in sin;
294 in6_sin6_2_sin(&sin, sin6);
295 t = in_pcblookup_local(pcbinfo, sin.sin_addr,
296 lport, lookupflags, cred);
297 if (t && t->inp_flags & INP_TIMEWAIT) {
301 if ((reuseport & tw->tw_so_options) == 0
302 && (reuseport_lb & tw->tw_so_options) == 0
303 && (ntohl(t->inp_laddr.s_addr) !=
304 INADDR_ANY || ((inp->inp_vflag &
306 (t->inp_vflag & INP_IPV6PROTO))))
309 (reuseport & inp_so_options(t)) == 0 &&
310 (reuseport_lb & inp_so_options(t)) == 0 &&
311 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
312 (t->inp_vflag & INP_IPV6PROTO) != 0)) {
318 inp->in6p_laddr = sin6->sin6_addr;
321 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
322 /* Undo an address bind that may have occurred. */
323 inp->in6p_laddr = in6addr_any;
327 inp->inp_lport = lport;
328 if (in_pcbinshash(inp) != 0) {
329 inp->in6p_laddr = in6addr_any;
338 * Transform old in6_pcbconnect() into an inner subroutine for new
339 * in6_pcbconnect(): Do some validity-checking on the remote
340 * address (in mbuf 'nam') and then determine local host address
341 * (i.e., which interface) to use to access that remote host.
343 * This preserves definition of in6_pcbconnect(), while supporting a
344 * slightly different version for T/TCP. (This is more than
345 * a bit of a kludge, but cleaning up the internal interfaces would
346 * have forced minor changes in every protocol).
349 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
350 struct in6_addr *plocal_addr6)
352 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
354 int scope_ambiguous = 0;
355 struct in6_addr in6a;
357 INP_WLOCK_ASSERT(inp);
358 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */
360 if (nam->sa_len != sizeof (*sin6))
362 if (sin6->sin6_family != AF_INET6)
363 return (EAFNOSUPPORT);
364 if (sin6->sin6_port == 0)
365 return (EADDRNOTAVAIL);
367 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
369 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
372 if (!CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) {
374 * If the destination address is UNSPECIFIED addr,
375 * use the loopback addr, e.g ::1.
377 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
378 sin6->sin6_addr = in6addr_loopback;
380 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
383 error = in6_selectsrc_socket(sin6, inp->in6p_outputopts,
384 inp, inp->inp_cred, scope_ambiguous, &in6a, NULL);
389 * Do not update this earlier, in case we return with an error.
391 * XXX: this in6_selectsrc_socket result might replace the bound local
392 * address with the address specified by setsockopt(IPV6_PKTINFO).
393 * Is it the intended behavior?
395 *plocal_addr6 = in6a;
398 * Don't do pcblookup call here; return interface in
400 * and exit to caller, that will do the lookup.
408 * Connect from a socket to a specified address.
409 * Both address and port must be specified in argument sin.
410 * If don't have a local address for this socket yet,
414 in6_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
415 struct ucred *cred, struct mbuf *m)
417 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
418 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
419 struct in6_addr addr6;
422 INP_WLOCK_ASSERT(inp);
423 INP_HASH_WLOCK_ASSERT(pcbinfo);
426 * Call inner routine, to assign local interface address.
427 * in6_pcbladdr() may automatically fill in sin6_scope_id.
429 if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0)
432 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
434 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
435 ? &addr6 : &inp->in6p_laddr,
436 inp->inp_lport, 0, NULL) != NULL) {
439 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
440 if (inp->inp_lport == 0) {
441 error = in6_pcbbind(inp, (struct sockaddr *)0, cred);
445 inp->in6p_laddr = addr6;
447 inp->in6p_faddr = sin6->sin6_addr;
448 inp->inp_fport = sin6->sin6_port;
449 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
450 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
451 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
453 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
455 in_pcbrehash_mbuf(inp, m);
461 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
464 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL));
468 in6_pcbdisconnect(struct inpcb *inp)
471 INP_WLOCK_ASSERT(inp);
472 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
474 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
476 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
477 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
482 in6_sockaddr(in_port_t port, struct in6_addr *addr_p)
484 struct sockaddr_in6 *sin6;
486 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK);
487 bzero(sin6, sizeof *sin6);
488 sin6->sin6_family = AF_INET6;
489 sin6->sin6_len = sizeof(*sin6);
490 sin6->sin6_port = port;
491 sin6->sin6_addr = *addr_p;
492 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */
494 return (struct sockaddr *)sin6;
498 in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p)
500 struct sockaddr_in sin;
501 struct sockaddr_in6 *sin6_p;
503 bzero(&sin, sizeof sin);
504 sin.sin_family = AF_INET;
505 sin.sin_len = sizeof(sin);
507 sin.sin_addr = *addr_p;
509 sin6_p = malloc(sizeof *sin6_p, M_SONAME,
511 in6_sin_2_v4mapsin6(&sin, sin6_p);
513 return (struct sockaddr *)sin6_p;
517 in6_getsockaddr(struct socket *so, struct sockaddr **nam)
520 struct in6_addr addr;
524 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
527 port = inp->inp_lport;
528 addr = inp->in6p_laddr;
531 *nam = in6_sockaddr(port, &addr);
536 in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
539 struct in6_addr addr;
543 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
546 port = inp->inp_fport;
547 addr = inp->in6p_faddr;
550 *nam = in6_sockaddr(port, &addr);
555 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
561 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
564 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
565 error = in_getsockaddr(so, nam);
567 in6_sin_2_v4mapsin6_in_sock(nam);
571 /* scope issues will be handled in in6_getsockaddr(). */
572 error = in6_getsockaddr(so, nam);
579 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
585 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
588 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
589 error = in_getpeeraddr(so, nam);
591 in6_sin_2_v4mapsin6_in_sock(nam);
594 /* scope issues will be handled in in6_getpeeraddr(). */
595 error = in6_getpeeraddr(so, nam);
601 * Pass some notification to all connections of a protocol
602 * associated with address dst. The local address and/or port numbers
603 * may be specified to limit the search. The "usual action" will be
604 * taken, depending on the ctlinput cmd. The caller must filter any
605 * cmds that are uninteresting (e.g., no error in the map).
606 * Call the protocol specific routine (if any) to report
607 * any errors for each matching socket.
610 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst,
611 u_int fport_arg, const struct sockaddr *src, u_int lport_arg,
612 int cmd, void *cmdarg,
613 struct inpcb *(*notify)(struct inpcb *, int))
615 struct inpcb *inp, *inp_temp;
616 struct sockaddr_in6 sa6_src, *sa6_dst;
617 u_short fport = fport_arg, lport = lport_arg;
621 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
624 sa6_dst = (struct sockaddr_in6 *)dst;
625 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
629 * note that src can be NULL when we get notify by local fragmentation.
631 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
632 flowinfo = sa6_src.sin6_flowinfo;
635 * Redirects go to all references to the destination,
636 * and use in6_rtchange to invalidate the route cache.
637 * Dead host indications: also use in6_rtchange to invalidate
638 * the cache, and deliver the error to all the sockets.
639 * Otherwise, if we have knowledge of the local port and address,
640 * deliver only to that socket.
642 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
645 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
647 if (cmd != PRC_HOSTDEAD)
648 notify = in6_rtchange;
650 errno = inet6ctlerrmap[cmd];
651 INP_INFO_WLOCK(pcbinfo);
652 CK_LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
654 if ((inp->inp_vflag & INP_IPV6) == 0) {
660 * If the error designates a new path MTU for a destination
661 * and the application (associated with this socket) wanted to
662 * know the value, notify.
663 * XXX: should we avoid to notify the value to TCP sockets?
665 if (cmd == PRC_MSGSIZE && cmdarg != NULL)
666 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst,
667 *(u_int32_t *)cmdarg);
670 * Detect if we should notify the error. If no source and
671 * destination ports are specifed, but non-zero flowinfo and
672 * local address match, notify the error. This is the case
673 * when the error is delivered with an encrypted buffer
674 * by ESP. Otherwise, just compare addresses and ports
677 if (lport == 0 && fport == 0 && flowinfo &&
678 inp->inp_socket != NULL &&
679 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
680 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
682 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
683 &sa6_dst->sin6_addr) ||
684 inp->inp_socket == 0 ||
685 (lport && inp->inp_lport != lport) ||
686 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
687 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
688 &sa6_src.sin6_addr)) ||
689 (fport && inp->inp_fport != fport)) {
696 if ((*notify)(inp, errno))
701 INP_INFO_WUNLOCK(pcbinfo);
705 * Lookup a PCB based on the local address and port. Caller must hold the
706 * hash lock. No inpcb locks or references are acquired.
709 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
710 u_short lport, int lookupflags, struct ucred *cred)
713 int matchwild = 3, wildcard;
715 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
716 ("%s: invalid lookup flags %d", __func__, lookupflags));
718 INP_HASH_WLOCK_ASSERT(pcbinfo);
720 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
721 struct inpcbhead *head;
723 * Look for an unconnected (wildcard foreign addr) PCB that
724 * matches the local address and port we're looking for.
726 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
727 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
728 pcbinfo->ipi_hashmask)];
729 CK_LIST_FOREACH(inp, head, inp_hash) {
730 /* XXX inp locking */
731 if ((inp->inp_vflag & INP_IPV6) == 0)
733 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
734 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
735 inp->inp_lport == lport) {
738 prison_equal_ip6(cred->cr_prison,
739 inp->inp_cred->cr_prison))
748 struct inpcbporthead *porthash;
749 struct inpcbport *phd;
750 struct inpcb *match = NULL;
752 * Best fit PCB lookup.
754 * First see if this local port is in use by looking on the
757 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
758 pcbinfo->ipi_porthashmask)];
759 CK_LIST_FOREACH(phd, porthash, phd_hash) {
760 if (phd->phd_port == lport)
765 * Port is in use by one or more PCBs. Look for best
768 CK_LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
771 !prison_equal_ip6(cred->cr_prison,
772 inp->inp_cred->cr_prison))
774 /* XXX inp locking */
775 if ((inp->inp_vflag & INP_IPV6) == 0)
777 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
779 if (!IN6_IS_ADDR_UNSPECIFIED(
781 if (IN6_IS_ADDR_UNSPECIFIED(laddr))
783 else if (!IN6_ARE_ADDR_EQUAL(
784 &inp->in6p_laddr, laddr))
787 if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
790 if (wildcard < matchwild) {
792 matchwild = wildcard;
803 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
806 struct ip6_moptions *im6o;
809 INP_INFO_WLOCK(pcbinfo);
810 CK_LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) {
812 if (__predict_false(in6p->inp_flags2 & INP_FREED)) {
816 im6o = in6p->in6p_moptions;
817 if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) {
819 * Unselect the outgoing ifp for multicast if it
822 if (im6o->im6o_multicast_ifp == ifp)
823 im6o->im6o_multicast_ifp = NULL;
825 * Drop multicast group membership if we joined
826 * through the interface being detached.
829 for (i = 0; i < im6o->im6o_num_memberships; i++) {
830 if (im6o->im6o_membership[i]->in6m_ifp ==
832 in6_leavegroup(im6o->im6o_membership[i], NULL);
834 } else if (gap != 0) {
835 im6o->im6o_membership[i - gap] =
836 im6o->im6o_membership[i];
839 im6o->im6o_num_memberships -= gap;
843 INP_INFO_WUNLOCK(pcbinfo);
847 * Check for alternatives when higher level complains
848 * about service problems. For now, invalidate cached
849 * routing information. If the route was created dynamically
850 * (by a redirect), time to try a default gateway again.
853 in6_losing(struct inpcb *inp)
856 RO_INVALIDATE_CACHE(&inp->inp_route6);
860 * After a routing change, flush old routing
861 * and allocate a (hopefully) better one.
864 in6_rtchange(struct inpcb *inp, int errno __unused)
867 RO_INVALIDATE_CACHE(&inp->inp_route6);
871 static struct inpcb *
872 in6_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo,
873 const struct in6_addr *laddr, uint16_t lport, const struct in6_addr *faddr,
874 uint16_t fport, int lookupflags)
876 struct inpcb *local_wild = NULL;
877 const struct inpcblbgrouphead *hdr;
878 struct inpcblbgroup *grp;
879 struct inpcblbgroup *grp_local_wild;
882 INP_HASH_LOCK_ASSERT(pcbinfo);
884 hdr = &pcbinfo->ipi_lbgrouphashbase[INP_PCBLBGROUP_PORTHASH(
885 lport, pcbinfo->ipi_lbgrouphashmask)];
888 * Order of socket selection:
890 * 2. wild (if lookupflags contains INPLOOKUP_WILDCARD).
893 * - Load balanced group does not contain jailed sockets.
894 * - Load balanced does not contain IPv4 mapped INET6 wild sockets.
896 CK_LIST_FOREACH(grp, hdr, il_list) {
898 if (!(grp->il_vflag & INP_IPV6))
901 if (grp->il_lport == lport) {
903 int pkt_hash = INP_PCBLBGROUP_PKTHASH(
904 INP6_PCBHASHKEY(faddr), lport, fport);
906 idx = pkt_hash % grp->il_inpcnt;
908 if (IN6_ARE_ADDR_EQUAL(&grp->il6_laddr, laddr)) {
909 return (grp->il_inp[idx]);
911 if (IN6_IS_ADDR_UNSPECIFIED(&grp->il6_laddr) &&
912 (lookupflags & INPLOOKUP_WILDCARD)) {
913 local_wild = grp->il_inp[idx];
914 grp_local_wild = grp;
919 if (local_wild != NULL) {
927 * Lookup PCB in hash list, using pcbgroup tables.
929 static struct inpcb *
930 in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
931 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr,
932 u_int lport_arg, int lookupflags, struct ifnet *ifp)
934 struct inpcbhead *head;
935 struct inpcb *inp, *tmpinp;
936 u_short fport = fport_arg, lport = lport_arg;
940 * First look for an exact match.
943 INP_GROUP_LOCK(pcbgroup);
944 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(
945 INP6_PCBHASHKEY(faddr), lport, fport, pcbgroup->ipg_hashmask)];
946 CK_LIST_FOREACH(inp, head, inp_pcbgrouphash) {
947 /* XXX inp locking */
948 if ((inp->inp_vflag & INP_IPV6) == 0)
950 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
951 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
952 inp->inp_fport == fport &&
953 inp->inp_lport == lport) {
955 * XXX We should be able to directly return
956 * the inp here, without any checks.
957 * Well unless both bound with SO_REUSEPORT?
959 if (prison_flag(inp->inp_cred, PR_IP6))
965 if (tmpinp != NULL) {
971 * Then look for a wildcard match in the pcbgroup.
973 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
974 struct inpcb *local_wild = NULL, *local_exact = NULL;
975 struct inpcb *jail_wild = NULL;
979 * Order of socket selection - we always prefer jails.
980 * 1. jailed, non-wild.
982 * 3. non-jailed, non-wild.
983 * 4. non-jailed, wild.
985 head = &pcbgroup->ipg_hashbase[
986 INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)];
987 CK_LIST_FOREACH(inp, head, inp_pcbgrouphash) {
988 /* XXX inp locking */
989 if ((inp->inp_vflag & INP_IPV6) == 0)
992 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
993 inp->inp_lport != lport) {
997 injail = prison_flag(inp->inp_cred, PR_IP6);
999 if (prison_check_ip6(inp->inp_cred,
1003 if (local_exact != NULL)
1007 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1012 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1018 } /* LIST_FOREACH */
1032 * Then look for a wildcard match, if requested.
1034 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1035 struct inpcb *local_wild = NULL, *local_exact = NULL;
1036 struct inpcb *jail_wild = NULL;
1040 * Order of socket selection - we always prefer jails.
1041 * 1. jailed, non-wild.
1043 * 3. non-jailed, non-wild.
1044 * 4. non-jailed, wild.
1046 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(
1047 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
1048 pcbinfo->ipi_wildmask)];
1049 CK_LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1050 /* XXX inp locking */
1051 if ((inp->inp_vflag & INP_IPV6) == 0)
1054 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1055 inp->inp_lport != lport) {
1059 injail = prison_flag(inp->inp_cred, PR_IP6);
1061 if (prison_check_ip6(inp->inp_cred,
1065 if (local_exact != NULL)
1069 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1074 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1080 } /* LIST_FOREACH */
1091 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1092 INP_GROUP_UNLOCK(pcbgroup);
1096 if (lookupflags & INPLOOKUP_WLOCKPCB)
1097 locked = INP_TRY_WLOCK(inp);
1098 else if (lookupflags & INPLOOKUP_RLOCKPCB)
1099 locked = INP_TRY_RLOCK(inp);
1101 panic("%s: locking buf", __func__);
1104 INP_GROUP_UNLOCK(pcbgroup);
1106 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1108 if (in_pcbrele_wlocked(inp))
1112 if (in_pcbrele_rlocked(inp))
1117 if (lookupflags & INPLOOKUP_WLOCKPCB)
1118 INP_WLOCK_ASSERT(inp);
1120 INP_RLOCK_ASSERT(inp);
1124 #endif /* PCBGROUP */
1127 * Lookup PCB in hash list.
1129 static struct inpcb *
1130 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1131 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
1132 int lookupflags, struct ifnet *ifp)
1134 struct inpcbhead *head;
1135 struct inpcb *inp, *tmpinp;
1136 u_short fport = fport_arg, lport = lport_arg;
1138 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1139 ("%s: invalid lookup flags %d", __func__, lookupflags));
1141 INP_HASH_LOCK_ASSERT(pcbinfo);
1144 * First look for an exact match.
1147 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1148 INP6_PCBHASHKEY(faddr), lport, fport, pcbinfo->ipi_hashmask)];
1149 CK_LIST_FOREACH(inp, head, inp_hash) {
1150 /* XXX inp locking */
1151 if ((inp->inp_vflag & INP_IPV6) == 0)
1153 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1154 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1155 inp->inp_fport == fport &&
1156 inp->inp_lport == lport) {
1158 * XXX We should be able to directly return
1159 * the inp here, without any checks.
1160 * Well unless both bound with SO_REUSEPORT?
1162 if (prison_flag(inp->inp_cred, PR_IP6))
1172 * Then look in lb group (for wildcard match).
1174 if (pcbinfo->ipi_lbgrouphashbase != NULL &&
1175 (lookupflags & INPLOOKUP_WILDCARD)) {
1176 inp = in6_pcblookup_lbgroup(pcbinfo, laddr, lport, faddr,
1177 fport, lookupflags);
1184 * Then look for a wildcard match, if requested.
1186 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1187 struct inpcb *local_wild = NULL, *local_exact = NULL;
1188 struct inpcb *jail_wild = NULL;
1192 * Order of socket selection - we always prefer jails.
1193 * 1. jailed, non-wild.
1195 * 3. non-jailed, non-wild.
1196 * 4. non-jailed, wild.
1198 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1199 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
1200 pcbinfo->ipi_hashmask)];
1201 CK_LIST_FOREACH(inp, head, inp_hash) {
1202 /* XXX inp locking */
1203 if ((inp->inp_vflag & INP_IPV6) == 0)
1206 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1207 inp->inp_lport != lport) {
1211 injail = prison_flag(inp->inp_cred, PR_IP6);
1213 if (prison_check_ip6(inp->inp_cred,
1217 if (local_exact != NULL)
1221 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1226 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1232 } /* LIST_FOREACH */
1234 if (jail_wild != NULL)
1236 if (local_exact != NULL)
1237 return (local_exact);
1238 if (local_wild != NULL)
1239 return (local_wild);
1240 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1249 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1250 * hash list lock, and will return the inpcb locked (i.e., requires
1251 * INPLOOKUP_LOCKPCB).
1253 static struct inpcb *
1254 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1255 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1260 INP_HASH_RLOCK(pcbinfo);
1261 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1262 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1264 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1266 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1270 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1272 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1277 panic("%s: locking bug", __func__);
1280 if (lookupflags & INPLOOKUP_WLOCKPCB)
1281 INP_WLOCK_ASSERT(inp);
1283 INP_RLOCK_ASSERT(inp);
1287 INP_HASH_RUNLOCK(pcbinfo);
1292 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1293 * from which a pre-calculated hash value may be extracted.
1295 * Possibly more of this logic should be in in6_pcbgroup.c.
1298 in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport,
1299 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1301 #if defined(PCBGROUP) && !defined(RSS)
1302 struct inpcbgroup *pcbgroup;
1305 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1306 ("%s: invalid lookup flags %d", __func__, lookupflags));
1307 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1308 ("%s: LOCKPCB not set", __func__));
1311 * When not using RSS, use connection groups in preference to the
1312 * reservation table when looking up 4-tuples. When using RSS, just
1313 * use the reservation table, due to the cost of the Toeplitz hash
1316 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1317 * we could be doing RSS with a non-Toeplitz hash that is affordable
1320 #if defined(PCBGROUP) && !defined(RSS)
1321 if (in_pcbgroup_enabled(pcbinfo)) {
1322 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1324 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1325 laddr, lport, lookupflags, ifp));
1328 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1333 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1334 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1335 struct ifnet *ifp, struct mbuf *m)
1338 struct inpcbgroup *pcbgroup;
1341 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1342 ("%s: invalid lookup flags %d", __func__, lookupflags));
1343 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1344 ("%s: LOCKPCB not set", __func__));
1348 * If we can use a hardware-generated hash to look up the connection
1349 * group, use that connection group to find the inpcb. Otherwise
1350 * fall back on a software hash -- or the reservation table if we're
1353 * XXXRW: As above, that policy belongs in the pcbgroup code.
1355 if (in_pcbgroup_enabled(pcbinfo) &&
1356 M_HASHTYPE_TEST(m, M_HASHTYPE_NONE) == 0) {
1357 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1358 m->m_pkthdr.flowid);
1359 if (pcbgroup != NULL)
1360 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr,
1361 fport, laddr, lport, lookupflags, ifp));
1363 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1365 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1366 laddr, lport, lookupflags, ifp));
1370 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1375 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m, int srcordst)
1379 ip = mtod(m, struct ip6_hdr *);
1380 bzero(sin6, sizeof(*sin6));
1381 sin6->sin6_len = sizeof(*sin6);
1382 sin6->sin6_family = AF_INET6;
1383 sin6->sin6_addr = srcordst ? ip->ip6_dst : ip->ip6_src;
1385 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */