Initial import from FreeBSD RELENG_4:
[dragonfly.git] / sys / netinet6 / in6_pcb.c
1 /*      $FreeBSD: src/sys/netinet6/in6_pcb.c,v 1.10.2.9 2003/01/24 05:11:35 sam Exp $   */
2 /*      $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $        */
3   
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  */
33
34 /*
35  * Copyright (c) 1982, 1986, 1991, 1993
36  *      The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *      This product includes software developed by the University of
49  *      California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *      @(#)in_pcb.c    8.2 (Berkeley) 1/4/94
67  */
68
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/sockio.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/proc.h>
85 #include <sys/jail.h>
86
87 #include <vm/vm_zone.h>
88
89 #include <net/if.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92
93 #include <netinet/in.h>
94 #include <netinet/in_var.h>
95 #include <netinet/in_systm.h>
96 #include <netinet/ip6.h>
97 #include <netinet/ip_var.h>
98 #include <netinet6/ip6_var.h>
99 #include <netinet6/nd6.h>
100 #include <netinet/in_pcb.h>
101 #include <netinet6/in6_pcb.h>
102
103 #ifdef IPSEC
104 #include <netinet6/ipsec.h>
105 #ifdef INET6
106 #include <netinet6/ipsec6.h>
107 #endif
108 #include <netinet6/ah.h>
109 #ifdef INET6
110 #include <netinet6/ah6.h>
111 #endif
112 #include <netkey/key.h>
113 #endif /* IPSEC */
114
115 #ifdef FAST_IPSEC
116 #include <netipsec/ipsec.h>
117 #include <netipsec/ipsec6.h>
118 #include <netipsec/key.h>
119 #define IPSEC
120 #endif /* FAST_IPSEC */
121
122 struct  in6_addr zeroin6_addr;
123
124 int
125 in6_pcbbind(inp, nam, p)
126         register struct inpcb *inp;
127         struct sockaddr *nam;
128         struct proc *p;
129 {
130         struct socket *so = inp->inp_socket;
131         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
132         struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
133         u_short lport = 0;
134         int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
135
136         if (!in6_ifaddr) /* XXX broken! */
137                 return (EADDRNOTAVAIL);
138         if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
139                 return(EINVAL);
140         if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
141                 wild = 1;
142         if (nam) {
143                 sin6 = (struct sockaddr_in6 *)nam;
144                 if (nam->sa_len != sizeof(*sin6))
145                         return(EINVAL);
146                 /*
147                  * family check.
148                  */
149                 if (nam->sa_family != AF_INET6)
150                         return(EAFNOSUPPORT);
151
152                 /* KAME hack: embed scopeid */
153                 if (in6_embedscope(&sin6->sin6_addr, sin6, inp, NULL) != 0)
154                         return EINVAL;
155                 /* this must be cleared for ifa_ifwithaddr() */
156                 sin6->sin6_scope_id = 0;
157
158                 lport = sin6->sin6_port;
159                 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
160                         /*
161                          * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
162                          * allow compepte duplication of binding if
163                          * SO_REUSEPORT is set, or if SO_REUSEADDR is set
164                          * and a multicast address is bound on both
165                          * new and duplicated sockets.
166                          */
167                         if (so->so_options & SO_REUSEADDR)
168                                 reuseport = SO_REUSEADDR|SO_REUSEPORT;
169                 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
170                         struct ifaddr *ia = NULL;
171
172                         sin6->sin6_port = 0;            /* yech... */
173                         if ((ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0)
174                                 return(EADDRNOTAVAIL);
175
176                         /*
177                          * XXX: bind to an anycast address might accidentally
178                          * cause sending a packet with anycast source address.
179                          * We should allow to bind to a deprecated address, since
180                          * the application dare to use it.
181                          */
182                         if (ia &&
183                             ((struct in6_ifaddr *)ia)->ia6_flags &
184                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
185                                 return(EADDRNOTAVAIL);
186                         }
187                 }
188                 if (lport) {
189                         struct inpcb *t;
190
191                         /* GROSS */
192                         if (ntohs(lport) < IPV6PORT_RESERVED && p &&
193                             suser_xxx(0, p, PRISON_ROOT))
194                                 return(EACCES);
195                         if (so->so_cred->cr_uid != 0 &&
196                             !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
197                                 t = in6_pcblookup_local(pcbinfo,
198                                     &sin6->sin6_addr, lport,
199                                     INPLOOKUP_WILDCARD);
200                                 if (t &&
201                                     (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
202                                      !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
203                                      (t->inp_socket->so_options &
204                                       SO_REUSEPORT) == 0) &&
205                                     (so->so_cred->cr_uid !=
206                                      t->inp_socket->so_cred->cr_uid))
207                                         return (EADDRINUSE);
208                                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
209                                     IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
210                                         struct sockaddr_in sin;
211
212                                         in6_sin6_2_sin(&sin, sin6);
213                                         t = in_pcblookup_local(pcbinfo,
214                                                 sin.sin_addr, lport,
215                                                 INPLOOKUP_WILDCARD);
216                                         if (t &&
217                                             (so->so_cred->cr_uid !=
218                                              t->inp_socket->so_cred->cr_uid) &&
219                                             (ntohl(t->inp_laddr.s_addr) !=
220                                              INADDR_ANY ||
221                                              INP_SOCKAF(so) ==
222                                              INP_SOCKAF(t->inp_socket)))
223                                                 return (EADDRINUSE);
224                                 }
225                         }
226                         t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
227                                                 lport, wild);
228                         if (t && (reuseport & t->inp_socket->so_options) == 0)
229                                 return(EADDRINUSE);
230                         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
231                             IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
232                                 struct sockaddr_in sin;
233
234                                 in6_sin6_2_sin(&sin, sin6);
235                                 t = in_pcblookup_local(pcbinfo, sin.sin_addr,
236                                                        lport, wild);
237                                 if (t &&
238                                     (reuseport & t->inp_socket->so_options)
239                                     == 0 &&
240                                     (ntohl(t->inp_laddr.s_addr)
241                                      != INADDR_ANY ||
242                                      INP_SOCKAF(so) ==
243                                      INP_SOCKAF(t->inp_socket)))
244                                         return (EADDRINUSE);
245                         }
246                 }
247                 inp->in6p_laddr = sin6->sin6_addr;
248         }
249         if (lport == 0) {
250                 int e;
251                 if ((e = in6_pcbsetport(&inp->in6p_laddr, inp, p)) != 0)
252                         return(e);
253         }
254         else {
255                 inp->inp_lport = lport;
256                 if (in_pcbinshash(inp) != 0) {
257                         inp->in6p_laddr = in6addr_any;
258                         inp->inp_lport = 0;
259                         return (EAGAIN);
260                 }
261         }
262         return(0);
263 }
264
265 /*
266  *   Transform old in6_pcbconnect() into an inner subroutine for new
267  *   in6_pcbconnect(): Do some validity-checking on the remote
268  *   address (in mbuf 'nam') and then determine local host address
269  *   (i.e., which interface) to use to access that remote host.
270  *
271  *   This preserves definition of in6_pcbconnect(), while supporting a
272  *   slightly different version for T/TCP.  (This is more than
273  *   a bit of a kludge, but cleaning up the internal interfaces would
274  *   have forced minor changes in every protocol).
275  */
276
277 int
278 in6_pcbladdr(inp, nam, plocal_addr6)
279         register struct inpcb *inp;
280         struct sockaddr *nam;
281         struct in6_addr **plocal_addr6;
282 {
283         register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
284         struct ifnet *ifp = NULL;
285         int error = 0;
286
287         if (nam->sa_len != sizeof (*sin6))
288                 return (EINVAL);
289         if (sin6->sin6_family != AF_INET6)
290                 return (EAFNOSUPPORT);
291         if (sin6->sin6_port == 0)
292                 return (EADDRNOTAVAIL);
293
294         /* KAME hack: embed scopeid */
295         if (in6_embedscope(&sin6->sin6_addr, sin6, inp, &ifp) != 0)
296                 return EINVAL;
297
298         if (in6_ifaddr) {
299                 /*
300                  * If the destination address is UNSPECIFIED addr,
301                  * use the loopback addr, e.g ::1.
302                  */
303                 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
304                         sin6->sin6_addr = in6addr_loopback;
305         }
306         {
307                 /*
308                  * XXX: in6_selectsrc might replace the bound local address
309                  * with the address specified by setsockopt(IPV6_PKTINFO).
310                  * Is it the intended behavior?
311                  */
312                 *plocal_addr6 = in6_selectsrc(sin6, inp->in6p_outputopts,
313                                               inp->in6p_moptions,
314                                               &inp->in6p_route,
315                                               &inp->in6p_laddr, &error);
316                 if (*plocal_addr6 == 0) {
317                         if (error == 0)
318                                 error = EADDRNOTAVAIL;
319                         return(error);
320                 }
321                 /*
322                  * Don't do pcblookup call here; return interface in
323                  * plocal_addr6
324                  * and exit to caller, that will do the lookup.
325                  */
326         }
327
328         if (inp->in6p_route.ro_rt)
329                 ifp = inp->in6p_route.ro_rt->rt_ifp;
330
331         return(0);
332 }
333
334 /*
335  * Outer subroutine:
336  * Connect from a socket to a specified address.
337  * Both address and port must be specified in argument sin.
338  * If don't have a local address for this socket yet,
339  * then pick one.
340  */
341 int
342 in6_pcbconnect(inp, nam, p)
343         register struct inpcb *inp;
344         struct sockaddr *nam;
345         struct proc *p;
346 {
347         struct in6_addr *addr6;
348         register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
349         int error;
350
351         /*
352          * Call inner routine, to assign local interface address.
353          * in6_pcbladdr() may automatically fill in sin6_scope_id.
354          */
355         if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0)
356                 return(error);
357
358         if (in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
359                                sin6->sin6_port,
360                               IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
361                               ? addr6 : &inp->in6p_laddr,
362                               inp->inp_lport, 0, NULL) != NULL) {
363                 return (EADDRINUSE);
364         }
365         if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
366                 if (inp->inp_lport == 0) {
367                         error = in6_pcbbind(inp, (struct sockaddr *)0, p);
368                         if (error)
369                                 return (error);
370                 }
371                 inp->in6p_laddr = *addr6;
372         }
373         inp->in6p_faddr = sin6->sin6_addr;
374         inp->inp_fport = sin6->sin6_port;
375         /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
376         inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
377         if (inp->in6p_flags & IN6P_AUTOFLOWLABEL)
378                 inp->in6p_flowinfo |=
379                     (htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK);
380
381         in_pcbrehash(inp);
382         return (0);
383 }
384
385 #if 0
386 /*
387  * Return an IPv6 address, which is the most appropriate for given
388  * destination and user specified options.
389  * If necessary, this function lookups the routing table and return
390  * an entry to the caller for later use.
391  */
392 struct in6_addr *
393 in6_selectsrc(dstsock, opts, mopts, ro, laddr, errorp)
394         struct sockaddr_in6 *dstsock;
395         struct ip6_pktopts *opts;
396         struct ip6_moptions *mopts;
397         struct route_in6 *ro;
398         struct in6_addr *laddr;
399         int *errorp;
400 {
401         struct in6_addr *dst;
402         struct in6_ifaddr *ia6 = 0;
403         struct in6_pktinfo *pi = NULL;
404
405         dst = &dstsock->sin6_addr;
406         *errorp = 0;
407
408         /*
409          * If the source address is explicitly specified by the caller,
410          * use it.
411          */
412         if (opts && (pi = opts->ip6po_pktinfo) &&
413             !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr))
414                 return(&pi->ipi6_addr);
415
416         /*
417          * If the source address is not specified but the socket(if any)
418          * is already bound, use the bound address.
419          */
420         if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr))
421                 return(laddr);
422
423         /*
424          * If the caller doesn't specify the source address but
425          * the outgoing interface, use an address associated with
426          * the interface.
427          */
428         if (pi && pi->ipi6_ifindex) {
429                 /* XXX boundary check is assumed to be already done. */
430                 ia6 = in6_ifawithscope(ifindex2ifnet[pi->ipi6_ifindex],
431                                        dst);
432                 if (ia6 == 0) {
433                         *errorp = EADDRNOTAVAIL;
434                         return(0);
435                 }
436                 return(&satosin6(&ia6->ia_addr)->sin6_addr);
437         }
438
439         /*
440          * If the destination address is a link-local unicast address or
441          * a multicast address, and if the outgoing interface is specified
442          * by the sin6_scope_id filed, use an address associated with the
443          * interface.
444          * XXX: We're now trying to define more specific semantics of
445          *      sin6_scope_id field, so this part will be rewritten in
446          *      the near future.
447          */
448         if ((IN6_IS_ADDR_LINKLOCAL(dst) || IN6_IS_ADDR_MULTICAST(dst)) &&
449             dstsock->sin6_scope_id) {
450                 /*
451                  * I'm not sure if boundary check for scope_id is done
452                  * somewhere...
453                  */
454                 if (dstsock->sin6_scope_id < 0 ||
455                     if_index < dstsock->sin6_scope_id) {
456                         *errorp = ENXIO; /* XXX: better error? */
457                         return(0);
458                 }
459                 ia6 = in6_ifawithscope(ifindex2ifnet[dstsock->sin6_scope_id],
460                                        dst);
461                 if (ia6 == 0) {
462                         *errorp = EADDRNOTAVAIL;
463                         return(0);
464                 }
465                 return(&satosin6(&ia6->ia_addr)->sin6_addr);
466         }
467
468         /*
469          * If the destination address is a multicast address and
470          * the outgoing interface for the address is specified
471          * by the caller, use an address associated with the interface.
472          * There is a sanity check here; if the destination has node-local
473          * scope, the outgoing interfacde should be a loopback address.
474          * Even if the outgoing interface is not specified, we also
475          * choose a loopback interface as the outgoing interface.
476          */
477         if (IN6_IS_ADDR_MULTICAST(dst)) {
478                 struct ifnet *ifp = mopts ? mopts->im6o_multicast_ifp : NULL;
479
480                 if (ifp == NULL && IN6_IS_ADDR_MC_NODELOCAL(dst)) {
481                         ifp = &loif[0];
482                 }
483
484                 if (ifp) {
485                         ia6 = in6_ifawithscope(ifp, dst);
486                         if (ia6 == 0) {
487                                 *errorp = EADDRNOTAVAIL;
488                                 return(0);
489                         }
490                         return(&ia6->ia_addr.sin6_addr);
491                 }
492         }
493
494         /*
495          * If the next hop address for the packet is specified
496          * by caller, use an address associated with the route
497          * to the next hop.
498          */
499         {
500                 struct sockaddr_in6 *sin6_next;
501                 struct rtentry *rt;
502
503                 if (opts && opts->ip6po_nexthop) {
504                         sin6_next = satosin6(opts->ip6po_nexthop);
505                         rt = nd6_lookup(&sin6_next->sin6_addr, 1, NULL);
506                         if (rt) {
507                                 ia6 = in6_ifawithscope(rt->rt_ifp, dst);
508                                 if (ia6 == 0)
509                                         ia6 = ifatoia6(rt->rt_ifa);
510                         }
511                         if (ia6 == 0) {
512                                 *errorp = EADDRNOTAVAIL;
513                                 return(0);
514                         }
515                         return(&satosin6(&ia6->ia_addr)->sin6_addr);
516                 }
517         }
518
519         /*
520          * If route is known or can be allocated now,
521          * our src addr is taken from the i/f, else punt.
522          */
523         if (ro) {
524                 if (ro->ro_rt &&
525                     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst)) {
526                         RTFREE(ro->ro_rt);
527                         ro->ro_rt = (struct rtentry *)0;
528                 }
529                 if (ro->ro_rt == (struct rtentry *)0 ||
530                     ro->ro_rt->rt_ifp == (struct ifnet *)0) {
531                         struct sockaddr_in6 *dst6;
532
533                         /* No route yet, so try to acquire one */
534                         bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
535                         dst6 = (struct sockaddr_in6 *)&ro->ro_dst;
536                         dst6->sin6_family = AF_INET6;
537                         dst6->sin6_len = sizeof(struct sockaddr_in6);
538                         dst6->sin6_addr = *dst;
539                         if (IN6_IS_ADDR_MULTICAST(dst)) {
540                                 ro->ro_rt = rtalloc1(&((struct route *)ro)
541                                                      ->ro_dst, 0, 0UL);
542                         } else {
543                                 rtalloc((struct route *)ro);
544                         }
545                 }
546
547                 /*
548                  * in_pcbconnect() checks out IFF_LOOPBACK to skip using
549                  * the address. But we don't know why it does so.
550                  * It is necessary to ensure the scope even for lo0
551                  * so doesn't check out IFF_LOOPBACK.
552                  */
553
554                 if (ro->ro_rt) {
555                         ia6 = in6_ifawithscope(ro->ro_rt->rt_ifa->ifa_ifp, dst);
556                         if (ia6 == 0) /* xxx scope error ?*/
557                                 ia6 = ifatoia6(ro->ro_rt->rt_ifa);
558                 }
559                 if (ia6 == 0) {
560                         *errorp = EHOSTUNREACH; /* no route */
561                         return(0);
562                 }
563                 return(&satosin6(&ia6->ia_addr)->sin6_addr);
564         }
565
566         *errorp = EADDRNOTAVAIL;
567         return(0);
568 }
569
570 /*
571  * Default hop limit selection. The precedence is as follows:
572  * 1. Hoplimit valued specified via ioctl.
573  * 2. (If the outgoing interface is detected) the current
574  *     hop limit of the interface specified by router advertisement.
575  * 3. The system default hoplimit.
576 */
577 int
578 in6_selecthlim(in6p, ifp)
579         struct in6pcb *in6p;
580         struct ifnet *ifp;
581 {
582         if (in6p && in6p->in6p_hops >= 0)
583                 return(in6p->in6p_hops);
584         else if (ifp)
585                 return(nd_ifinfo[ifp->if_index].chlim);
586         else
587                 return(ip6_defhlim);
588 }
589 #endif
590
591 void
592 in6_pcbdisconnect(inp)
593         struct inpcb *inp;
594 {
595         bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
596         inp->inp_fport = 0;
597         /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
598         inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
599         in_pcbrehash(inp);
600         if (inp->inp_socket->so_state & SS_NOFDREF)
601                 in6_pcbdetach(inp);
602 }
603
604 void
605 in6_pcbdetach(inp)
606         struct inpcb *inp;
607 {
608         struct socket *so = inp->inp_socket;
609         struct inpcbinfo *ipi = inp->inp_pcbinfo;
610
611 #ifdef IPSEC
612         if (inp->in6p_sp != NULL)
613                 ipsec6_delete_pcbpolicy(inp);
614 #endif /* IPSEC */
615         inp->inp_gencnt = ++ipi->ipi_gencnt;
616         in_pcbremlists(inp);
617         sotoinpcb(so) = 0;
618         sofree(so);
619
620         if (inp->in6p_options)
621                 m_freem(inp->in6p_options);
622         ip6_freepcbopts(inp->in6p_outputopts);
623         ip6_freemoptions(inp->in6p_moptions);
624         if (inp->in6p_route.ro_rt)
625                 rtfree(inp->in6p_route.ro_rt);
626         /* Check and free IPv4 related resources in case of mapped addr */
627         if (inp->inp_options)
628                 (void)m_free(inp->inp_options);
629         ip_freemoptions(inp->inp_moptions);
630
631         inp->inp_vflag = 0;
632         zfreei(ipi->ipi_zone, inp);
633 }
634
635 /*
636  * The calling convention of in6_setsockaddr() and in6_setpeeraddr() was
637  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
638  * in struct pr_usrreqs, so that protocols can just reference then directly
639  * without the need for a wrapper function.  The socket must have a valid
640  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
641  * except through a kernel programming error, so it is acceptable to panic
642  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
643  * because there actually /is/ a programming error somewhere... XXX)
644  */
645 int
646 in6_setsockaddr(so, nam)
647         struct socket *so;
648         struct sockaddr **nam;
649 {
650         int s;
651         register struct inpcb *inp;
652         register struct sockaddr_in6 *sin6;
653
654         /*
655          * Do the malloc first in case it blocks.
656          */
657         MALLOC(sin6, struct sockaddr_in6 *, sizeof *sin6, M_SONAME, M_WAITOK);
658         bzero(sin6, sizeof *sin6);
659         sin6->sin6_family = AF_INET6;
660         sin6->sin6_len = sizeof(*sin6);
661
662         s = splnet();
663         inp = sotoinpcb(so);
664         if (!inp) {
665                 splx(s);
666                 free(sin6, M_SONAME);
667                 return EINVAL;
668         }
669         sin6->sin6_port = inp->inp_lport;
670         sin6->sin6_addr = inp->in6p_laddr;
671         splx(s);
672         if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
673                 sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
674         else
675                 sin6->sin6_scope_id = 0;        /*XXX*/
676         if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
677                 sin6->sin6_addr.s6_addr16[1] = 0;
678
679         *nam = (struct sockaddr *)sin6;
680         return 0;
681 }
682
683 int
684 in6_setpeeraddr(so, nam)
685         struct socket *so;
686         struct sockaddr **nam;
687 {
688         int s;
689         struct inpcb *inp;
690         register struct sockaddr_in6 *sin6;
691
692         /*
693          * Do the malloc first in case it blocks.
694          */
695         MALLOC(sin6, struct sockaddr_in6 *, sizeof(*sin6), M_SONAME, M_WAITOK);
696         bzero((caddr_t)sin6, sizeof (*sin6));
697         sin6->sin6_family = AF_INET6;
698         sin6->sin6_len = sizeof(struct sockaddr_in6);
699
700         s = splnet();
701         inp = sotoinpcb(so);
702         if (!inp) {
703                 splx(s);
704                 free(sin6, M_SONAME);
705                 return EINVAL;
706         }
707         sin6->sin6_port = inp->inp_fport;
708         sin6->sin6_addr = inp->in6p_faddr;
709         splx(s);
710         if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
711                 sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
712         else
713                 sin6->sin6_scope_id = 0;        /*XXX*/
714         if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
715                 sin6->sin6_addr.s6_addr16[1] = 0;
716
717         *nam = (struct sockaddr *)sin6;
718         return 0;
719 }
720
721 int
722 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
723 {
724         struct  inpcb *inp = sotoinpcb(so);
725         int     error;
726
727         if (inp == NULL)
728                 return EINVAL;
729         if (inp->inp_vflag & INP_IPV4) {
730                 error = in_setsockaddr(so, nam);
731                 if (error == 0)
732                         in6_sin_2_v4mapsin6_in_sock(nam);
733         } else
734         /* scope issues will be handled in in6_setsockaddr(). */
735         error = in6_setsockaddr(so, nam);
736
737         return error;
738 }
739
740 int
741 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
742 {
743         struct  inpcb *inp = sotoinpcb(so);
744         int     error;
745
746         if (inp == NULL)
747                 return EINVAL;
748         if (inp->inp_vflag & INP_IPV4) {
749                 error = in_setpeeraddr(so, nam);
750                 if (error == 0)
751                         in6_sin_2_v4mapsin6_in_sock(nam);
752         } else
753         /* scope issues will be handled in in6_setpeeraddr(). */
754         error = in6_setpeeraddr(so, nam);
755
756         return error;
757 }
758
759 /*
760  * Pass some notification to all connections of a protocol
761  * associated with address dst.  The local address and/or port numbers
762  * may be specified to limit the search.  The "usual action" will be
763  * taken, depending on the ctlinput cmd.  The caller must filter any
764  * cmds that are uninteresting (e.g., no error in the map).
765  * Call the protocol specific routine (if any) to report
766  * any errors for each matching socket.
767  *
768  * Must be called at splnet.
769  */
770 void
771 in6_pcbnotify(head, dst, fport_arg, src, lport_arg, cmd, notify)
772         struct inpcbhead *head;
773         struct sockaddr *dst, *src;
774         u_int fport_arg, lport_arg;
775         int cmd;
776         void (*notify) __P((struct inpcb *, int));
777 {
778         struct inpcb *inp, *ninp;
779         struct sockaddr_in6 sa6_src, *sa6_dst;
780         u_short fport = fport_arg, lport = lport_arg;
781         u_int32_t flowinfo;
782         int errno, s;
783
784         if ((unsigned)cmd > PRC_NCMDS || dst->sa_family != AF_INET6)
785                 return;
786
787         sa6_dst = (struct sockaddr_in6 *)dst;
788         if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
789                 return;
790
791         /*
792          * note that src can be NULL when we get notify by local fragmentation.
793          */
794         sa6_src = (src == NULL) ? sa6_any : *(struct sockaddr_in6 *)src;
795         flowinfo = sa6_src.sin6_flowinfo;
796
797         /*
798          * Redirects go to all references to the destination,
799          * and use in6_rtchange to invalidate the route cache.
800          * Dead host indications: also use in6_rtchange to invalidate
801          * the cache, and deliver the error to all the sockets.
802          * Otherwise, if we have knowledge of the local port and address,
803          * deliver only to that socket.
804          */
805         if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
806                 fport = 0;
807                 lport = 0;
808                 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
809
810                 if (cmd != PRC_HOSTDEAD)
811                         notify = in6_rtchange;
812         }
813         errno = inet6ctlerrmap[cmd];
814         s = splnet();
815         for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
816                 ninp = LIST_NEXT(inp, inp_list);
817
818                 if ((inp->inp_vflag & INP_IPV6) == 0)
819                         continue;
820
821                 /*
822                  * Detect if we should notify the error. If no source and
823                  * destination ports are specifed, but non-zero flowinfo and
824                  * local address match, notify the error. This is the case
825                  * when the error is delivered with an encrypted buffer
826                  * by ESP. Otherwise, just compare addresses and ports
827                  * as usual.
828                  */
829                 if (lport == 0 && fport == 0 && flowinfo &&
830                     inp->inp_socket != NULL &&
831                     flowinfo == (inp->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
832                     IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
833                         goto do_notify;
834                 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
835                                              &sa6_dst->sin6_addr) ||
836                          inp->inp_socket == 0 ||
837                          (lport && inp->inp_lport != lport) ||
838                          (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
839                           !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
840                                               &sa6_src.sin6_addr)) ||
841                          (fport && inp->inp_fport != fport))
842                         continue;
843
844           do_notify:
845                 if (notify)
846                         (*notify)(inp, errno);
847         }
848         splx(s);
849 }
850
851 /*
852  * Lookup a PCB based on the local address and port.
853  */
854 struct inpcb *
855 in6_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay)
856         struct inpcbinfo *pcbinfo;
857         struct in6_addr *laddr;
858         u_int lport_arg;
859         int wild_okay;
860 {
861         register struct inpcb *inp;
862         int matchwild = 3, wildcard;
863         u_short lport = lport_arg;
864
865         if (!wild_okay) {
866                 struct inpcbhead *head;
867                 /*
868                  * Look for an unconnected (wildcard foreign addr) PCB that
869                  * matches the local address and port we're looking for.
870                  */
871                 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
872                                                       pcbinfo->hashmask)];
873                 LIST_FOREACH(inp, head, inp_hash) {
874                         if ((inp->inp_vflag & INP_IPV6) == 0)
875                                 continue;
876                         if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
877                             IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
878                             inp->inp_lport == lport) {
879                                 /*
880                                  * Found.
881                                  */
882                                 return (inp);
883                         }
884                 }
885                 /*
886                  * Not found.
887                  */
888                 return (NULL);
889         } else {
890                 struct inpcbporthead *porthash;
891                 struct inpcbport *phd;
892                 struct inpcb *match = NULL;
893                 /*
894                  * Best fit PCB lookup.
895                  *
896                  * First see if this local port is in use by looking on the
897                  * port hash list.
898                  */
899                 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
900                     pcbinfo->porthashmask)];
901                 LIST_FOREACH(phd, porthash, phd_hash) {
902                         if (phd->phd_port == lport)
903                                 break;
904                 }
905                 if (phd != NULL) {
906                         /*
907                          * Port is in use by one or more PCBs. Look for best
908                          * fit.
909                          */
910                         LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
911                                 wildcard = 0;
912                                 if ((inp->inp_vflag & INP_IPV6) == 0)
913                                         continue;
914                                 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
915                                         wildcard++;
916                                 if (!IN6_IS_ADDR_UNSPECIFIED(
917                                         &inp->in6p_laddr)) {
918                                         if (IN6_IS_ADDR_UNSPECIFIED(laddr))
919                                                 wildcard++;
920                                         else if (!IN6_ARE_ADDR_EQUAL(
921                                                 &inp->in6p_laddr, laddr))
922                                                 continue;
923                                 } else {
924                                         if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
925                                                 wildcard++;
926                                 }
927                                 if (wildcard < matchwild) {
928                                         match = inp;
929                                         matchwild = wildcard;
930                                         if (matchwild == 0) {
931                                                 break;
932                                         }
933                                 }
934                         }
935                 }
936                 return (match);
937         }
938 }
939
940 void
941 in6_pcbpurgeif0(head, ifp)
942         struct in6pcb *head;
943         struct ifnet *ifp;
944 {
945         struct in6pcb *in6p;
946         struct ip6_moptions *im6o;
947         struct in6_multi_mship *imm, *nimm;
948
949         for (in6p = head; in6p != NULL; in6p = LIST_NEXT(in6p, inp_list)) {
950                 im6o = in6p->in6p_moptions;
951                 if ((in6p->inp_vflag & INP_IPV6) &&
952                     im6o) {
953                         /*
954                          * Unselect the outgoing interface if it is being
955                          * detached.
956                          */
957                         if (im6o->im6o_multicast_ifp == ifp)
958                                 im6o->im6o_multicast_ifp = NULL;
959
960                         /*
961                          * Drop multicast group membership if we joined
962                          * through the interface being detached.
963                          * XXX controversial - is it really legal for kernel
964                          * to force this?
965                          */
966                         for (imm = im6o->im6o_memberships.lh_first;
967                              imm != NULL; imm = nimm) {
968                                 nimm = imm->i6mm_chain.le_next;
969                                 if (imm->i6mm_maddr->in6m_ifp == ifp) {
970                                         LIST_REMOVE(imm, i6mm_chain);
971                                         in6_delmulti(imm->i6mm_maddr);
972                                         free(imm, M_IPMADDR);
973                                 }
974                         }
975                 }
976         }
977 }
978
979 /*
980  * Check for alternatives when higher level complains
981  * about service problems.  For now, invalidate cached
982  * routing information.  If the route was created dynamically
983  * (by a redirect), time to try a default gateway again.
984  */
985 void
986 in6_losing(in6p)
987         struct inpcb *in6p;
988 {
989         struct rtentry *rt;
990         struct rt_addrinfo info;
991
992         if ((rt = in6p->in6p_route.ro_rt) != NULL) {
993                 bzero((caddr_t)&info, sizeof(info));
994                 info.rti_flags = rt->rt_flags;
995                 info.rti_info[RTAX_DST] = rt_key(rt);
996                 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
997                 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
998                 rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
999                 if (rt->rt_flags & RTF_DYNAMIC)
1000                         (void)rtrequest1(RTM_DELETE, &info, NULL);
1001                 in6p->in6p_route.ro_rt = NULL;
1002                 rtfree(rt);
1003                 /*
1004                  * A new route can be allocated
1005                  * the next time output is attempted.
1006                  */
1007         }
1008 }
1009
1010 /*
1011  * After a routing change, flush old routing
1012  * and allocate a (hopefully) better one.
1013  */
1014 void
1015 in6_rtchange(inp, errno)
1016         struct inpcb *inp;
1017         int errno;
1018 {
1019         if (inp->in6p_route.ro_rt) {
1020                 rtfree(inp->in6p_route.ro_rt);
1021                 inp->in6p_route.ro_rt = 0;
1022                 /*
1023                  * A new route can be allocated the next time
1024                  * output is attempted.
1025                  */
1026         }
1027 }
1028
1029 /*
1030  * Lookup PCB in hash list.
1031  */
1032 struct inpcb *
1033 in6_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, ifp)
1034         struct inpcbinfo *pcbinfo;
1035         struct in6_addr *faddr, *laddr;
1036         u_int fport_arg, lport_arg;
1037         int wildcard;
1038         struct ifnet *ifp;
1039 {
1040         struct inpcbhead *head;
1041         register struct inpcb *inp;
1042         u_short fport = fport_arg, lport = lport_arg;
1043         int faith;
1044
1045         if (faithprefix_p != NULL)
1046                 faith = (*faithprefix_p)(laddr);
1047         else
1048                 faith = 0;
1049
1050         /*
1051          * First look for an exact match.
1052          */
1053         head = &pcbinfo->hashbase[INP_PCBHASH(faddr->s6_addr32[3] /* XXX */,
1054                                               lport, fport,
1055                                               pcbinfo->hashmask)];
1056         LIST_FOREACH(inp, head, inp_hash) {
1057                 if ((inp->inp_vflag & INP_IPV6) == 0)
1058                         continue;
1059                 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1060                     IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1061                     inp->inp_fport == fport &&
1062                     inp->inp_lport == lport) {
1063                         /*
1064                          * Found.
1065                          */
1066                         return (inp);
1067                 }
1068         }
1069         if (wildcard) {
1070                 struct inpcb *local_wild = NULL;
1071
1072                 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1073                                                       pcbinfo->hashmask)];
1074                 LIST_FOREACH(inp, head, inp_hash) {
1075                         if ((inp->inp_vflag & INP_IPV6) == 0)
1076                                 continue;
1077                         if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1078                             inp->inp_lport == lport) {
1079                                 if (faith && (inp->inp_flags & INP_FAITH) == 0)
1080                                         continue;
1081                                 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
1082                                                        laddr))
1083                                         return (inp);
1084                                 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1085                                         local_wild = inp;
1086                         }
1087                 }
1088                 return (local_wild);
1089         }
1090
1091         /*
1092          * Not found.
1093          */
1094         return (NULL);
1095 }
1096
1097 void
1098 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
1099 {
1100         struct ip6_hdr *ip;
1101
1102         ip = mtod(m, struct ip6_hdr *);
1103         bzero(sin6, sizeof(*sin6));
1104         sin6->sin6_len = sizeof(*sin6);
1105         sin6->sin6_family = AF_INET6;
1106         sin6->sin6_addr = ip->ip6_src;
1107         if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
1108                 sin6->sin6_addr.s6_addr16[1] = 0;
1109         sin6->sin6_scope_id =
1110                 (m->m_pkthdr.rcvif && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
1111                 ? m->m_pkthdr.rcvif->if_index : 0;
1112
1113         return;
1114 }