]> CyberLeo.Net >> Repos - FreeBSD/releng/10.0.git/blob - sys/netinet6/udp6_usrreq.c
- Copy stable/10 (r259064) to releng/10.0 as part of the
[FreeBSD/releng/10.0.git] / sys / netinet6 / udp6_usrreq.c
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: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
34  *      $KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
35  */
36
37 /*-
38  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
39  *      The Regents of the University of California.
40  * All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 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  *      @(#)udp_usrreq.c        8.6 (Berkeley) 5/23/95
67  */
68
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71
72 #include "opt_inet.h"
73 #include "opt_inet6.h"
74 #include "opt_ipfw.h"
75 #include "opt_ipsec.h"
76 #include "opt_kdtrace.h"
77
78 #include <sys/param.h>
79 #include <sys/jail.h>
80 #include <sys/kernel.h>
81 #include <sys/lock.h>
82 #include <sys/mbuf.h>
83 #include <sys/priv.h>
84 #include <sys/proc.h>
85 #include <sys/protosw.h>
86 #include <sys/sdt.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sx.h>
91 #include <sys/sysctl.h>
92 #include <sys/syslog.h>
93 #include <sys/systm.h>
94
95 #include <net/if.h>
96 #include <net/if_types.h>
97 #include <net/route.h>
98
99 #include <netinet/in.h>
100 #include <netinet/in_kdtrace.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/in_systm.h>
103 #include <netinet/in_var.h>
104 #include <netinet/ip.h>
105 #include <netinet/ip_icmp.h>
106 #include <netinet/ip6.h>
107 #include <netinet/icmp_var.h>
108 #include <netinet/icmp6.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/udp.h>
111 #include <netinet/udp_var.h>
112
113 #include <netinet6/ip6protosw.h>
114 #include <netinet6/ip6_var.h>
115 #include <netinet6/in6_pcb.h>
116 #include <netinet6/udp6_var.h>
117 #include <netinet6/scope6_var.h>
118
119 #ifdef IPSEC
120 #include <netipsec/ipsec.h>
121 #include <netipsec/ipsec6.h>
122 #endif /* IPSEC */
123
124 #include <security/mac/mac_framework.h>
125
126 /*
127  * UDP protocol implementation.
128  * Per RFC 768, August, 1980.
129  */
130
131 extern struct protosw   inetsw[];
132 static void             udp6_detach(struct socket *so);
133
134 static void
135 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
136     struct sockaddr_in6 *fromsa)
137 {
138         struct socket *so;
139         struct mbuf *opts;
140
141         INP_LOCK_ASSERT(inp);
142
143 #ifdef IPSEC
144         /* Check AH/ESP integrity. */
145         if (ipsec6_in_reject(n, inp)) {
146                 m_freem(n);
147                 IPSEC6STAT_INC(ips_in_polvio);
148                 return;
149         }
150 #endif /* IPSEC */
151 #ifdef MAC
152         if (mac_inpcb_check_deliver(inp, n) != 0) {
153                 m_freem(n);
154                 return;
155         }
156 #endif
157         opts = NULL;
158         if (inp->inp_flags & INP_CONTROLOPTS ||
159             inp->inp_socket->so_options & SO_TIMESTAMP)
160                 ip6_savecontrol(inp, n, &opts);
161         m_adj(n, off + sizeof(struct udphdr));
162
163         so = inp->inp_socket;
164         SOCKBUF_LOCK(&so->so_rcv);
165         if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
166             opts) == 0) {
167                 SOCKBUF_UNLOCK(&so->so_rcv);
168                 m_freem(n);
169                 if (opts)
170                         m_freem(opts);
171                 UDPSTAT_INC(udps_fullsock);
172         } else
173                 sorwakeup_locked(so);
174 }
175
176 int
177 udp6_input(struct mbuf **mp, int *offp, int proto)
178 {
179         struct mbuf *m = *mp;
180         struct ifnet *ifp;
181         struct ip6_hdr *ip6;
182         struct udphdr *uh;
183         struct inpcb *inp;
184         struct udpcb *up;
185         int off = *offp;
186         int plen, ulen;
187         struct sockaddr_in6 fromsa;
188         struct m_tag *fwd_tag;
189         uint16_t uh_sum;
190
191         ifp = m->m_pkthdr.rcvif;
192         ip6 = mtod(m, struct ip6_hdr *);
193
194         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
195                 /* XXX send icmp6 host/port unreach? */
196                 m_freem(m);
197                 return (IPPROTO_DONE);
198         }
199
200 #ifndef PULLDOWN_TEST
201         IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
202         ip6 = mtod(m, struct ip6_hdr *);
203         uh = (struct udphdr *)((caddr_t)ip6 + off);
204 #else
205         IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
206         if (!uh)
207                 return (IPPROTO_DONE);
208 #endif
209
210         UDPSTAT_INC(udps_ipackets);
211
212         /*
213          * Destination port of 0 is illegal, based on RFC768.
214          */
215         if (uh->uh_dport == 0)
216                 goto badunlocked;
217
218         plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
219         ulen = ntohs((u_short)uh->uh_ulen);
220
221         if (plen != ulen) {
222                 UDPSTAT_INC(udps_badlen);
223                 goto badunlocked;
224         }
225
226         /*
227          * Checksum extended UDP header and data.
228          */
229         if (uh->uh_sum == 0) {
230                 UDPSTAT_INC(udps_nosum);
231                 goto badunlocked;
232         }
233
234         if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) {
235                 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
236                         uh_sum = m->m_pkthdr.csum_data;
237                 else
238                         uh_sum = in6_cksum_pseudo(ip6, ulen,
239                             IPPROTO_UDP, m->m_pkthdr.csum_data);
240                 uh_sum ^= 0xffff;
241         } else
242                 uh_sum = in6_cksum(m, IPPROTO_UDP, off, ulen);
243
244         if (uh_sum != 0) {
245                 UDPSTAT_INC(udps_badsum);
246                 goto badunlocked;
247         }
248
249         /*
250          * Construct sockaddr format source address.
251          */
252         init_sin6(&fromsa, m);
253         fromsa.sin6_port = uh->uh_sport;
254
255         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
256                 struct inpcb *last;
257                 struct ip6_moptions *imo;
258
259                 INP_INFO_RLOCK(&V_udbinfo);
260                 /*
261                  * In the event that laddr should be set to the link-local
262                  * address (this happens in RIPng), the multicast address
263                  * specified in the received packet will not match laddr.  To
264                  * handle this situation, matching is relaxed if the
265                  * receiving interface is the same as one specified in the
266                  * socket and if the destination multicast address matches
267                  * one of the multicast groups specified in the socket.
268                  */
269
270                 /*
271                  * KAME note: traditionally we dropped udpiphdr from mbuf
272                  * here.  We need udphdr for IPsec processing so we do that
273                  * later.
274                  */
275                 last = NULL;
276                 LIST_FOREACH(inp, &V_udb, inp_list) {
277                         if ((inp->inp_vflag & INP_IPV6) == 0)
278                                 continue;
279                         if (inp->inp_lport != uh->uh_dport)
280                                 continue;
281                         if (inp->inp_fport != 0 &&
282                             inp->inp_fport != uh->uh_sport)
283                                 continue;
284                         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
285                                 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
286                                                         &ip6->ip6_dst))
287                                         continue;
288                         }
289                         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
290                                 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
291                                                         &ip6->ip6_src) ||
292                                     inp->inp_fport != uh->uh_sport)
293                                         continue;
294                         }
295
296                         /*
297                          * XXXRW: Because we weren't holding either the inpcb
298                          * or the hash lock when we checked for a match 
299                          * before, we should probably recheck now that the 
300                          * inpcb lock is (supposed to be) held.
301                          */
302
303                         /*
304                          * Handle socket delivery policy for any-source
305                          * and source-specific multicast. [RFC3678]
306                          */
307                         imo = inp->in6p_moptions;
308                         if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
309                                 struct sockaddr_in6      mcaddr;
310                                 int                      blocked;
311
312                                 INP_RLOCK(inp);
313
314                                 bzero(&mcaddr, sizeof(struct sockaddr_in6));
315                                 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
316                                 mcaddr.sin6_family = AF_INET6;
317                                 mcaddr.sin6_addr = ip6->ip6_dst;
318
319                                 blocked = im6o_mc_filter(imo, ifp,
320                                         (struct sockaddr *)&mcaddr,
321                                         (struct sockaddr *)&fromsa);
322                                 if (blocked != MCAST_PASS) {
323                                         if (blocked == MCAST_NOTGMEMBER)
324                                                 IP6STAT_INC(ip6s_notmember);
325                                         if (blocked == MCAST_NOTSMEMBER ||
326                                             blocked == MCAST_MUTED)
327                                                 UDPSTAT_INC(udps_filtermcast);
328                                         INP_RUNLOCK(inp); /* XXX */
329                                         continue;
330                                 }
331
332                                 INP_RUNLOCK(inp);
333                         }
334                         if (last != NULL) {
335                                 struct mbuf *n;
336
337                                 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
338                                         INP_RLOCK(last);
339                                         up = intoudpcb(last);
340                                         if (up->u_tun_func == NULL) {
341                                                 udp6_append(last, n, off, &fromsa);
342                                         } else {
343                                                 /*
344                                                  * Engage the tunneling
345                                                  * protocol we will have to
346                                                  * leave the info_lock up,
347                                                  * since we are hunting
348                                                  * through multiple UDP's.
349                                                  * 
350                                                  */
351                                                 (*up->u_tun_func)(n, off, last);
352                                         }
353                                         INP_RUNLOCK(last);
354                                 }
355                         }
356                         last = inp;
357                         /*
358                          * Don't look for additional matches if this one does
359                          * not have either the SO_REUSEPORT or SO_REUSEADDR
360                          * socket options set.  This heuristic avoids
361                          * searching through all pcbs in the common case of a
362                          * non-shared port.  It assumes that an application
363                          * will never clear these options after setting them.
364                          */
365                         if ((last->inp_socket->so_options &
366                              (SO_REUSEPORT|SO_REUSEADDR)) == 0)
367                                 break;
368                 }
369
370                 if (last == NULL) {
371                         /*
372                          * No matching pcb found; discard datagram.  (No need
373                          * to send an ICMP Port Unreachable for a broadcast
374                          * or multicast datgram.)
375                          */
376                         UDPSTAT_INC(udps_noport);
377                         UDPSTAT_INC(udps_noportmcast);
378                         goto badheadlocked;
379                 }
380                 INP_RLOCK(last);
381                 INP_INFO_RUNLOCK(&V_udbinfo);
382                 up = intoudpcb(last);
383                 UDP_PROBE(receive, NULL, last, ip6, last, uh);
384                 if (up->u_tun_func == NULL) {
385                         udp6_append(last, m, off, &fromsa);
386                 } else {
387                         /*
388                          * Engage the tunneling protocol.
389                          */
390                         (*up->u_tun_func)(m, off, last);
391                 }
392                 INP_RUNLOCK(last);
393                 return (IPPROTO_DONE);
394         }
395         /*
396          * Locate pcb for datagram.
397          */
398
399         /*
400          * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
401          */
402         if ((m->m_flags & M_IP6_NEXTHOP) &&
403             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
404                 struct sockaddr_in6 *next_hop6;
405
406                 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
407
408                 /*
409                  * Transparently forwarded. Pretend to be the destination.
410                  * Already got one like this?
411                  */
412                 inp = in6_pcblookup_mbuf(&V_udbinfo,
413                     &ip6->ip6_src, uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
414                     INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
415                 if (!inp) {
416                         /*
417                          * It's new.  Try to find the ambushing socket.
418                          * Because we've rewritten the destination address,
419                          * any hardware-generated hash is ignored.
420                          */
421                         inp = in6_pcblookup(&V_udbinfo, &ip6->ip6_src,
422                             uh->uh_sport, &next_hop6->sin6_addr,
423                             next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
424                             uh->uh_dport, INPLOOKUP_WILDCARD |
425                             INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
426                 }
427                 /* Remove the tag from the packet. We don't need it anymore. */
428                 m_tag_delete(m, fwd_tag);
429                 m->m_flags &= ~M_IP6_NEXTHOP;
430         } else
431                 inp = in6_pcblookup_mbuf(&V_udbinfo, &ip6->ip6_src,
432                     uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
433                     INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
434                     m->m_pkthdr.rcvif, m);
435         if (inp == NULL) {
436                 if (udp_log_in_vain) {
437                         char ip6bufs[INET6_ADDRSTRLEN];
438                         char ip6bufd[INET6_ADDRSTRLEN];
439
440                         log(LOG_INFO,
441                             "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
442                             ip6_sprintf(ip6bufd, &ip6->ip6_dst),
443                             ntohs(uh->uh_dport),
444                             ip6_sprintf(ip6bufs, &ip6->ip6_src),
445                             ntohs(uh->uh_sport));
446                 }
447                 UDPSTAT_INC(udps_noport);
448                 if (m->m_flags & M_MCAST) {
449                         printf("UDP6: M_MCAST is set in a unicast packet.\n");
450                         UDPSTAT_INC(udps_noportmcast);
451                         goto badunlocked;
452                 }
453                 if (V_udp_blackhole)
454                         goto badunlocked;
455                 if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
456                         goto badunlocked;
457                 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
458                 return (IPPROTO_DONE);
459         }
460         INP_RLOCK_ASSERT(inp);
461         up = intoudpcb(inp);
462         UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
463         if (up->u_tun_func == NULL) {
464                 udp6_append(inp, m, off, &fromsa);
465         } else {
466                 /*
467                  * Engage the tunneling protocol.
468                  */
469
470                 (*up->u_tun_func)(m, off, inp);
471         }
472         INP_RUNLOCK(inp);
473         return (IPPROTO_DONE);
474
475 badheadlocked:
476         INP_INFO_RUNLOCK(&V_udbinfo);
477 badunlocked:
478         if (m)
479                 m_freem(m);
480         return (IPPROTO_DONE);
481 }
482
483 void
484 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
485 {
486         struct udphdr uh;
487         struct ip6_hdr *ip6;
488         struct mbuf *m;
489         int off = 0;
490         struct ip6ctlparam *ip6cp = NULL;
491         const struct sockaddr_in6 *sa6_src = NULL;
492         void *cmdarg;
493         struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
494         struct udp_portonly {
495                 u_int16_t uh_sport;
496                 u_int16_t uh_dport;
497         } *uhp;
498
499         if (sa->sa_family != AF_INET6 ||
500             sa->sa_len != sizeof(struct sockaddr_in6))
501                 return;
502
503         if ((unsigned)cmd >= PRC_NCMDS)
504                 return;
505         if (PRC_IS_REDIRECT(cmd))
506                 notify = in6_rtchange, d = NULL;
507         else if (cmd == PRC_HOSTDEAD)
508                 d = NULL;
509         else if (inet6ctlerrmap[cmd] == 0)
510                 return;
511
512         /* if the parameter is from icmp6, decode it. */
513         if (d != NULL) {
514                 ip6cp = (struct ip6ctlparam *)d;
515                 m = ip6cp->ip6c_m;
516                 ip6 = ip6cp->ip6c_ip6;
517                 off = ip6cp->ip6c_off;
518                 cmdarg = ip6cp->ip6c_cmdarg;
519                 sa6_src = ip6cp->ip6c_src;
520         } else {
521                 m = NULL;
522                 ip6 = NULL;
523                 cmdarg = NULL;
524                 sa6_src = &sa6_any;
525         }
526
527         if (ip6) {
528                 /*
529                  * XXX: We assume that when IPV6 is non NULL,
530                  * M and OFF are valid.
531                  */
532
533                 /* Check if we can safely examine src and dst ports. */
534                 if (m->m_pkthdr.len < off + sizeof(*uhp))
535                         return;
536
537                 bzero(&uh, sizeof(uh));
538                 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
539
540                 (void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
541                     (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
542                     cmdarg, notify);
543         } else
544                 (void) in6_pcbnotify(&V_udbinfo, sa, 0,
545                     (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
546 }
547
548 static int
549 udp6_getcred(SYSCTL_HANDLER_ARGS)
550 {
551         struct xucred xuc;
552         struct sockaddr_in6 addrs[2];
553         struct inpcb *inp;
554         int error;
555
556         error = priv_check(req->td, PRIV_NETINET_GETCRED);
557         if (error)
558                 return (error);
559
560         if (req->newlen != sizeof(addrs))
561                 return (EINVAL);
562         if (req->oldlen != sizeof(struct xucred))
563                 return (EINVAL);
564         error = SYSCTL_IN(req, addrs, sizeof(addrs));
565         if (error)
566                 return (error);
567         if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
568             (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
569                 return (error);
570         }
571         inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
572             addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
573             INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
574         if (inp != NULL) {
575                 INP_RLOCK_ASSERT(inp);
576                 if (inp->inp_socket == NULL)
577                         error = ENOENT;
578                 if (error == 0)
579                         error = cr_canseesocket(req->td->td_ucred,
580                             inp->inp_socket);
581                 if (error == 0)
582                         cru2x(inp->inp_cred, &xuc);
583                 INP_RUNLOCK(inp);
584         } else
585                 error = ENOENT;
586         if (error == 0)
587                 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
588         return (error);
589 }
590
591 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
592     0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
593
594 static int
595 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
596     struct mbuf *control, struct thread *td)
597 {
598         u_int32_t ulen = m->m_pkthdr.len;
599         u_int32_t plen = sizeof(struct udphdr) + ulen;
600         struct ip6_hdr *ip6;
601         struct udphdr *udp6;
602         struct in6_addr *laddr, *faddr, in6a;
603         struct sockaddr_in6 *sin6 = NULL;
604         struct ifnet *oifp = NULL;
605         int scope_ambiguous = 0;
606         u_short fport;
607         int error = 0;
608         struct ip6_pktopts *optp, opt;
609         int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
610         int flags;
611         struct sockaddr_in6 tmp;
612
613         INP_WLOCK_ASSERT(inp);
614         INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
615
616         if (addr6) {
617                 /* addr6 has been validated in udp6_send(). */
618                 sin6 = (struct sockaddr_in6 *)addr6;
619
620                 /* protect *sin6 from overwrites */
621                 tmp = *sin6;
622                 sin6 = &tmp;
623
624                 /*
625                  * Application should provide a proper zone ID or the use of
626                  * default zone IDs should be enabled.  Unfortunately, some
627                  * applications do not behave as it should, so we need a
628                  * workaround.  Even if an appropriate ID is not determined,
629                  * we'll see if we can determine the outgoing interface.  If we
630                  * can, determine the zone ID based on the interface below.
631                  */
632                 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
633                         scope_ambiguous = 1;
634                 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
635                         return (error);
636         }
637
638         if (control) {
639                 if ((error = ip6_setpktopts(control, &opt,
640                     inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
641                         goto release;
642                 optp = &opt;
643         } else
644                 optp = inp->in6p_outputopts;
645
646         if (sin6) {
647                 faddr = &sin6->sin6_addr;
648
649                 /*
650                  * Since we saw no essential reason for calling in_pcbconnect,
651                  * we get rid of such kind of logic, and call in6_selectsrc
652                  * and in6_pcbsetport in order to fill in the local address
653                  * and the local port.
654                  */
655                 if (sin6->sin6_port == 0) {
656                         error = EADDRNOTAVAIL;
657                         goto release;
658                 }
659
660                 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
661                         /* how about ::ffff:0.0.0.0 case? */
662                         error = EISCONN;
663                         goto release;
664                 }
665
666                 fport = sin6->sin6_port; /* allow 0 port */
667
668                 if (IN6_IS_ADDR_V4MAPPED(faddr)) {
669                         if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
670                                 /*
671                                  * I believe we should explicitly discard the
672                                  * packet when mapped addresses are disabled,
673                                  * rather than send the packet as an IPv6 one.
674                                  * If we chose the latter approach, the packet
675                                  * might be sent out on the wire based on the
676                                  * default route, the situation which we'd
677                                  * probably want to avoid.
678                                  * (20010421 jinmei@kame.net)
679                                  */
680                                 error = EINVAL;
681                                 goto release;
682                         }
683                         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
684                             !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
685                                 /*
686                                  * when remote addr is an IPv4-mapped address,
687                                  * local addr should not be an IPv6 address,
688                                  * since you cannot determine how to map IPv6
689                                  * source address to IPv4.
690                                  */
691                                 error = EINVAL;
692                                 goto release;
693                         }
694
695                         af = AF_INET;
696                 }
697
698                 if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
699                         error = in6_selectsrc(sin6, optp, inp, NULL,
700                             td->td_ucred, &oifp, &in6a);
701                         if (error)
702                                 goto release;
703                         if (oifp && scope_ambiguous &&
704                             (error = in6_setscope(&sin6->sin6_addr,
705                             oifp, NULL))) {
706                                 goto release;
707                         }
708                         laddr = &in6a;
709                 } else
710                         laddr = &inp->in6p_laddr;       /* XXX */
711                 if (laddr == NULL) {
712                         if (error == 0)
713                                 error = EADDRNOTAVAIL;
714                         goto release;
715                 }
716                 if (inp->inp_lport == 0 &&
717                     (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
718                         /* Undo an address bind that may have occurred. */
719                         inp->in6p_laddr = in6addr_any;
720                         goto release;
721                 }
722         } else {
723                 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
724                         error = ENOTCONN;
725                         goto release;
726                 }
727                 if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
728                         if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
729                                 /*
730                                  * XXX: this case would happen when the
731                                  * application sets the V6ONLY flag after
732                                  * connecting the foreign address.
733                                  * Such applications should be fixed,
734                                  * so we bark here.
735                                  */
736                                 log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
737                                     "option was set for a connected socket\n");
738                                 error = EINVAL;
739                                 goto release;
740                         } else
741                                 af = AF_INET;
742                 }
743                 laddr = &inp->in6p_laddr;
744                 faddr = &inp->in6p_faddr;
745                 fport = inp->inp_fport;
746         }
747
748         if (af == AF_INET)
749                 hlen = sizeof(struct ip);
750
751         /*
752          * Calculate data length and get a mbuf
753          * for UDP and IP6 headers.
754          */
755         M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
756         if (m == 0) {
757                 error = ENOBUFS;
758                 goto release;
759         }
760
761         /*
762          * Stuff checksum and output datagram.
763          */
764         udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
765         udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
766         udp6->uh_dport = fport;
767         if (plen <= 0xffff)
768                 udp6->uh_ulen = htons((u_short)plen);
769         else
770                 udp6->uh_ulen = 0;
771         udp6->uh_sum = 0;
772
773         switch (af) {
774         case AF_INET6:
775                 ip6 = mtod(m, struct ip6_hdr *);
776                 ip6->ip6_flow   = inp->inp_flow & IPV6_FLOWINFO_MASK;
777                 ip6->ip6_vfc    &= ~IPV6_VERSION_MASK;
778                 ip6->ip6_vfc    |= IPV6_VERSION;
779                 ip6->ip6_plen   = htons((u_short)plen);
780                 ip6->ip6_nxt    = IPPROTO_UDP;
781                 ip6->ip6_hlim   = in6_selecthlim(inp, NULL);
782                 ip6->ip6_src    = *laddr;
783                 ip6->ip6_dst    = *faddr;
784
785                 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, IPPROTO_UDP, 0);
786                 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
787                 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
788
789                 flags = 0;
790
791                 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
792                 UDPSTAT_INC(udps_opackets);
793                 error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
794                     NULL, inp);
795                 break;
796         case AF_INET:
797                 error = EAFNOSUPPORT;
798                 goto release;
799         }
800         goto releaseopt;
801
802 release:
803         m_freem(m);
804
805 releaseopt:
806         if (control) {
807                 ip6_clearpktopts(&opt, -1);
808                 m_freem(control);
809         }
810         return (error);
811 }
812
813 static void
814 udp6_abort(struct socket *so)
815 {
816         struct inpcb *inp;
817
818         inp = sotoinpcb(so);
819         KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
820
821 #ifdef INET
822         if (inp->inp_vflag & INP_IPV4) {
823                 struct pr_usrreqs *pru;
824
825                 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
826                 (*pru->pru_abort)(so);
827                 return;
828         }
829 #endif
830
831         INP_WLOCK(inp);
832         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
833                 INP_HASH_WLOCK(&V_udbinfo);
834                 in6_pcbdisconnect(inp);
835                 inp->in6p_laddr = in6addr_any;
836                 INP_HASH_WUNLOCK(&V_udbinfo);
837                 soisdisconnected(so);
838         }
839         INP_WUNLOCK(inp);
840 }
841
842 static int
843 udp6_attach(struct socket *so, int proto, struct thread *td)
844 {
845         struct inpcb *inp;
846         int error;
847
848         inp = sotoinpcb(so);
849         KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
850
851         if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
852                 error = soreserve(so, udp_sendspace, udp_recvspace);
853                 if (error)
854                         return (error);
855         }
856         INP_INFO_WLOCK(&V_udbinfo);
857         error = in_pcballoc(so, &V_udbinfo);
858         if (error) {
859                 INP_INFO_WUNLOCK(&V_udbinfo);
860                 return (error);
861         }
862         inp = (struct inpcb *)so->so_pcb;
863         inp->inp_vflag |= INP_IPV6;
864         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
865                 inp->inp_vflag |= INP_IPV4;
866         inp->in6p_hops = -1;    /* use kernel default */
867         inp->in6p_cksum = -1;   /* just to be sure */
868         /*
869          * XXX: ugly!!
870          * IPv4 TTL initialization is necessary for an IPv6 socket as well,
871          * because the socket may be bound to an IPv6 wildcard address,
872          * which may match an IPv4-mapped IPv6 address.
873          */
874         inp->inp_ip_ttl = V_ip_defttl;
875
876         error = udp_newudpcb(inp);
877         if (error) {
878                 in_pcbdetach(inp);
879                 in_pcbfree(inp);
880                 INP_INFO_WUNLOCK(&V_udbinfo);
881                 return (error);
882         }
883         INP_WUNLOCK(inp);
884         INP_INFO_WUNLOCK(&V_udbinfo);
885         return (0);
886 }
887
888 static int
889 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
890 {
891         struct inpcb *inp;
892         int error;
893
894         inp = sotoinpcb(so);
895         KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
896
897         INP_WLOCK(inp);
898         INP_HASH_WLOCK(&V_udbinfo);
899         inp->inp_vflag &= ~INP_IPV4;
900         inp->inp_vflag |= INP_IPV6;
901         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
902                 struct sockaddr_in6 *sin6_p;
903
904                 sin6_p = (struct sockaddr_in6 *)nam;
905
906                 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
907                         inp->inp_vflag |= INP_IPV4;
908 #ifdef INET
909                 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
910                         struct sockaddr_in sin;
911
912                         in6_sin6_2_sin(&sin, sin6_p);
913                         inp->inp_vflag |= INP_IPV4;
914                         inp->inp_vflag &= ~INP_IPV6;
915                         error = in_pcbbind(inp, (struct sockaddr *)&sin,
916                             td->td_ucred);
917                         goto out;
918                 }
919 #endif
920         }
921
922         error = in6_pcbbind(inp, nam, td->td_ucred);
923 #ifdef INET
924 out:
925 #endif
926         INP_HASH_WUNLOCK(&V_udbinfo);
927         INP_WUNLOCK(inp);
928         return (error);
929 }
930
931 static void
932 udp6_close(struct socket *so)
933 {
934         struct inpcb *inp;
935
936         inp = sotoinpcb(so);
937         KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
938
939 #ifdef INET
940         if (inp->inp_vflag & INP_IPV4) {
941                 struct pr_usrreqs *pru;
942
943                 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
944                 (*pru->pru_disconnect)(so);
945                 return;
946         }
947 #endif
948         INP_WLOCK(inp);
949         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
950                 INP_HASH_WLOCK(&V_udbinfo);
951                 in6_pcbdisconnect(inp);
952                 inp->in6p_laddr = in6addr_any;
953                 INP_HASH_WUNLOCK(&V_udbinfo);
954                 soisdisconnected(so);
955         }
956         INP_WUNLOCK(inp);
957 }
958
959 static int
960 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
961 {
962         struct inpcb *inp;
963         struct sockaddr_in6 *sin6;
964         int error;
965
966         inp = sotoinpcb(so);
967         sin6 = (struct sockaddr_in6 *)nam;
968         KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
969
970         /*
971          * XXXRW: Need to clarify locking of v4/v6 flags.
972          */
973         INP_WLOCK(inp);
974 #ifdef INET
975         if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
976                 struct sockaddr_in sin;
977
978                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
979                         error = EINVAL;
980                         goto out;
981                 }
982                 if (inp->inp_faddr.s_addr != INADDR_ANY) {
983                         error = EISCONN;
984                         goto out;
985                 }
986                 in6_sin6_2_sin(&sin, sin6);
987                 inp->inp_vflag |= INP_IPV4;
988                 inp->inp_vflag &= ~INP_IPV6;
989                 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
990                 if (error != 0)
991                         goto out;
992                 INP_HASH_WLOCK(&V_udbinfo);
993                 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
994                     td->td_ucred);
995                 INP_HASH_WUNLOCK(&V_udbinfo);
996                 if (error == 0)
997                         soisconnected(so);
998                 goto out;
999         }
1000 #endif
1001         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1002                 error = EISCONN;
1003                 goto out;
1004         }
1005         inp->inp_vflag &= ~INP_IPV4;
1006         inp->inp_vflag |= INP_IPV6;
1007         error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1008         if (error != 0)
1009                 goto out;
1010         INP_HASH_WLOCK(&V_udbinfo);
1011         error = in6_pcbconnect(inp, nam, td->td_ucred);
1012         INP_HASH_WUNLOCK(&V_udbinfo);
1013         if (error == 0)
1014                 soisconnected(so);
1015 out:
1016         INP_WUNLOCK(inp);
1017         return (error);
1018 }
1019
1020 static void
1021 udp6_detach(struct socket *so)
1022 {
1023         struct inpcb *inp;
1024         struct udpcb *up;
1025
1026         inp = sotoinpcb(so);
1027         KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1028
1029         INP_INFO_WLOCK(&V_udbinfo);
1030         INP_WLOCK(inp);
1031         up = intoudpcb(inp);
1032         KASSERT(up != NULL, ("%s: up == NULL", __func__));
1033         in_pcbdetach(inp);
1034         in_pcbfree(inp);
1035         INP_INFO_WUNLOCK(&V_udbinfo);
1036         udp_discardcb(up);
1037 }
1038
1039 static int
1040 udp6_disconnect(struct socket *so)
1041 {
1042         struct inpcb *inp;
1043         int error;
1044
1045         inp = sotoinpcb(so);
1046         KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1047
1048 #ifdef INET
1049         if (inp->inp_vflag & INP_IPV4) {
1050                 struct pr_usrreqs *pru;
1051
1052                 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1053                 (void)(*pru->pru_disconnect)(so);
1054                 return (0);
1055         }
1056 #endif
1057
1058         INP_WLOCK(inp);
1059
1060         if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1061                 error = ENOTCONN;
1062                 goto out;
1063         }
1064
1065         INP_HASH_WLOCK(&V_udbinfo);
1066         in6_pcbdisconnect(inp);
1067         inp->in6p_laddr = in6addr_any;
1068         INP_HASH_WUNLOCK(&V_udbinfo);
1069         SOCK_LOCK(so);
1070         so->so_state &= ~SS_ISCONNECTED;                /* XXX */
1071         SOCK_UNLOCK(so);
1072 out:
1073         INP_WUNLOCK(inp);
1074         return (0);
1075 }
1076
1077 static int
1078 udp6_send(struct socket *so, int flags, struct mbuf *m,
1079     struct sockaddr *addr, struct mbuf *control, struct thread *td)
1080 {
1081         struct inpcb *inp;
1082         int error = 0;
1083
1084         inp = sotoinpcb(so);
1085         KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1086
1087         INP_WLOCK(inp);
1088         if (addr) {
1089                 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1090                         error = EINVAL;
1091                         goto bad;
1092                 }
1093                 if (addr->sa_family != AF_INET6) {
1094                         error = EAFNOSUPPORT;
1095                         goto bad;
1096                 }
1097         }
1098
1099 #ifdef INET
1100         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1101                 int hasv4addr;
1102                 struct sockaddr_in6 *sin6 = 0;
1103
1104                 if (addr == 0)
1105                         hasv4addr = (inp->inp_vflag & INP_IPV4);
1106                 else {
1107                         sin6 = (struct sockaddr_in6 *)addr;
1108                         hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1109                             ? 1 : 0;
1110                 }
1111                 if (hasv4addr) {
1112                         struct pr_usrreqs *pru;
1113
1114                         /*
1115                          * XXXRW: We release UDP-layer locks before calling
1116                          * udp_send() in order to avoid recursion.  However,
1117                          * this does mean there is a short window where inp's
1118                          * fields are unstable.  Could this lead to a
1119                          * potential race in which the factors causing us to
1120                          * select the UDPv4 output routine are invalidated?
1121                          */
1122                         INP_WUNLOCK(inp);
1123                         if (sin6)
1124                                 in6_sin6_2_sin_in_sock(addr);
1125                         pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1126                         /* addr will just be freed in sendit(). */
1127                         return ((*pru->pru_send)(so, flags, m, addr, control,
1128                             td));
1129                 }
1130         }
1131 #endif
1132 #ifdef MAC
1133         mac_inpcb_create_mbuf(inp, m);
1134 #endif
1135         INP_HASH_WLOCK(&V_udbinfo);
1136         error = udp6_output(inp, m, addr, control, td);
1137         INP_HASH_WUNLOCK(&V_udbinfo);
1138 #ifdef INET
1139 #endif  
1140         INP_WUNLOCK(inp);
1141         return (error);
1142
1143 bad:
1144         INP_WUNLOCK(inp);
1145         m_freem(m);
1146         return (error);
1147 }
1148
1149 struct pr_usrreqs udp6_usrreqs = {
1150         .pru_abort =            udp6_abort,
1151         .pru_attach =           udp6_attach,
1152         .pru_bind =             udp6_bind,
1153         .pru_connect =          udp6_connect,
1154         .pru_control =          in6_control,
1155         .pru_detach =           udp6_detach,
1156         .pru_disconnect =       udp6_disconnect,
1157         .pru_peeraddr =         in6_mapped_peeraddr,
1158         .pru_send =             udp6_send,
1159         .pru_shutdown =         udp_shutdown,
1160         .pru_sockaddr =         in6_mapped_sockaddr,
1161         .pru_soreceive =        soreceive_dgram,
1162         .pru_sosend =           sosend_dgram,
1163         .pru_sosetlabel =       in_pcbsosetlabel,
1164         .pru_close =            udp6_close
1165 };