]> CyberLeo.Net >> Repos - FreeBSD/releng/9.2.git/blob - sys/netinet6/udp6_usrreq.c
- Copy stable/9 to releng/9.2 as part of the 9.2-RELEASE cycle.
[FreeBSD/releng/9.2.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
77 #include <sys/param.h>
78 #include <sys/jail.h>
79 #include <sys/kernel.h>
80 #include <sys/lock.h>
81 #include <sys/mbuf.h>
82 #include <sys/priv.h>
83 #include <sys/proc.h>
84 #include <sys/protosw.h>
85 #include <sys/signalvar.h>
86 #include <sys/socket.h>
87 #include <sys/socketvar.h>
88 #include <sys/sx.h>
89 #include <sys/sysctl.h>
90 #include <sys/syslog.h>
91 #include <sys/systm.h>
92
93 #include <net/if.h>
94 #include <net/if_types.h>
95 #include <net/route.h>
96
97 #include <netinet/in.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/in_systm.h>
100 #include <netinet/in_var.h>
101 #include <netinet/ip.h>
102 #include <netinet/ip_icmp.h>
103 #include <netinet/ip6.h>
104 #include <netinet/icmp_var.h>
105 #include <netinet/icmp6.h>
106 #include <netinet/ip_var.h>
107 #include <netinet/udp.h>
108 #include <netinet/udp_var.h>
109
110 #include <netinet6/ip6protosw.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet6/in6_pcb.h>
113 #include <netinet6/udp6_var.h>
114 #include <netinet6/scope6_var.h>
115
116 #ifdef IPSEC
117 #include <netipsec/ipsec.h>
118 #include <netipsec/ipsec6.h>
119 #endif /* IPSEC */
120
121 #include <security/mac/mac_framework.h>
122
123 /*
124  * UDP protocol implementation.
125  * Per RFC 768, August, 1980.
126  */
127
128 extern struct protosw   inetsw[];
129 static void             udp6_detach(struct socket *so);
130
131 static void
132 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
133     struct sockaddr_in6 *fromsa)
134 {
135         struct socket *so;
136         struct mbuf *opts;
137
138         INP_LOCK_ASSERT(inp);
139
140 #ifdef IPSEC
141         /* Check AH/ESP integrity. */
142         if (ipsec6_in_reject(n, inp)) {
143                 m_freem(n);
144                 IPSEC6STAT_INC(in_polvio);
145                 return;
146         }
147 #endif /* IPSEC */
148 #ifdef MAC
149         if (mac_inpcb_check_deliver(inp, n) != 0) {
150                 m_freem(n);
151                 return;
152         }
153 #endif
154         opts = NULL;
155         if (inp->inp_flags & INP_CONTROLOPTS ||
156             inp->inp_socket->so_options & SO_TIMESTAMP)
157                 ip6_savecontrol(inp, n, &opts);
158         m_adj(n, off + sizeof(struct udphdr));
159
160         so = inp->inp_socket;
161         SOCKBUF_LOCK(&so->so_rcv);
162         if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
163             opts) == 0) {
164                 SOCKBUF_UNLOCK(&so->so_rcv);
165                 m_freem(n);
166                 if (opts)
167                         m_freem(opts);
168                 UDPSTAT_INC(udps_fullsock);
169         } else
170                 sorwakeup_locked(so);
171 }
172
173 int
174 udp6_input(struct mbuf **mp, int *offp, int proto)
175 {
176         struct mbuf *m = *mp;
177         struct ifnet *ifp;
178         struct ip6_hdr *ip6;
179         struct udphdr *uh;
180         struct inpcb *inp;
181         struct udpcb *up;
182         int off = *offp;
183         int plen, ulen;
184         struct sockaddr_in6 fromsa;
185         struct m_tag *fwd_tag;
186         uint16_t uh_sum;
187
188         ifp = m->m_pkthdr.rcvif;
189         ip6 = mtod(m, struct ip6_hdr *);
190
191         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
192                 /* XXX send icmp6 host/port unreach? */
193                 m_freem(m);
194                 return (IPPROTO_DONE);
195         }
196
197 #ifndef PULLDOWN_TEST
198         IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
199         ip6 = mtod(m, struct ip6_hdr *);
200         uh = (struct udphdr *)((caddr_t)ip6 + off);
201 #else
202         IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
203         if (!uh)
204                 return (IPPROTO_DONE);
205 #endif
206
207         UDPSTAT_INC(udps_ipackets);
208
209         /*
210          * Destination port of 0 is illegal, based on RFC768.
211          */
212         if (uh->uh_dport == 0)
213                 goto badunlocked;
214
215         plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
216         ulen = ntohs((u_short)uh->uh_ulen);
217
218         if (plen != ulen) {
219                 UDPSTAT_INC(udps_badlen);
220                 goto badunlocked;
221         }
222
223         /*
224          * Checksum extended UDP header and data.
225          */
226         if (uh->uh_sum == 0) {
227                 UDPSTAT_INC(udps_nosum);
228                 goto badunlocked;
229         }
230
231         if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) {
232                 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
233                         uh_sum = m->m_pkthdr.csum_data;
234                 else
235                         uh_sum = in6_cksum_pseudo(ip6, ulen,
236                             IPPROTO_UDP, m->m_pkthdr.csum_data);
237                 uh_sum ^= 0xffff;
238         } else
239                 uh_sum = in6_cksum(m, IPPROTO_UDP, off, ulen);
240
241         if (uh_sum != 0) {
242                 UDPSTAT_INC(udps_badsum);
243                 goto badunlocked;
244         }
245
246         /*
247          * Construct sockaddr format source address.
248          */
249         init_sin6(&fromsa, m);
250         fromsa.sin6_port = uh->uh_sport;
251
252         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
253                 struct inpcb *last;
254                 struct ip6_moptions *imo;
255
256                 INP_INFO_RLOCK(&V_udbinfo);
257                 /*
258                  * In the event that laddr should be set to the link-local
259                  * address (this happens in RIPng), the multicast address
260                  * specified in the received packet will not match laddr.  To
261                  * handle this situation, matching is relaxed if the
262                  * receiving interface is the same as one specified in the
263                  * socket and if the destination multicast address matches
264                  * one of the multicast groups specified in the socket.
265                  */
266
267                 /*
268                  * KAME note: traditionally we dropped udpiphdr from mbuf
269                  * here.  We need udphdr for IPsec processing so we do that
270                  * later.
271                  */
272                 last = NULL;
273                 LIST_FOREACH(inp, &V_udb, inp_list) {
274                         if ((inp->inp_vflag & INP_IPV6) == 0)
275                                 continue;
276                         if (inp->inp_lport != uh->uh_dport)
277                                 continue;
278                         if (inp->inp_fport != 0 &&
279                             inp->inp_fport != uh->uh_sport)
280                                 continue;
281                         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
282                                 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
283                                                         &ip6->ip6_dst))
284                                         continue;
285                         }
286                         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
287                                 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
288                                                         &ip6->ip6_src) ||
289                                     inp->inp_fport != uh->uh_sport)
290                                         continue;
291                         }
292
293                         /*
294                          * XXXRW: Because we weren't holding either the inpcb
295                          * or the hash lock when we checked for a match 
296                          * before, we should probably recheck now that the 
297                          * inpcb lock is (supposed to be) held.
298                          */
299
300                         /*
301                          * Handle socket delivery policy for any-source
302                          * and source-specific multicast. [RFC3678]
303                          */
304                         imo = inp->in6p_moptions;
305                         if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
306                                 struct sockaddr_in6      mcaddr;
307                                 int                      blocked;
308
309                                 INP_RLOCK(inp);
310
311                                 bzero(&mcaddr, sizeof(struct sockaddr_in6));
312                                 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
313                                 mcaddr.sin6_family = AF_INET6;
314                                 mcaddr.sin6_addr = ip6->ip6_dst;
315
316                                 blocked = im6o_mc_filter(imo, ifp,
317                                         (struct sockaddr *)&mcaddr,
318                                         (struct sockaddr *)&fromsa);
319                                 if (blocked != MCAST_PASS) {
320                                         if (blocked == MCAST_NOTGMEMBER)
321                                                 IP6STAT_INC(ip6s_notmember);
322                                         if (blocked == MCAST_NOTSMEMBER ||
323                                             blocked == MCAST_MUTED)
324                                                 UDPSTAT_INC(udps_filtermcast);
325                                         INP_RUNLOCK(inp); /* XXX */
326                                         continue;
327                                 }
328
329                                 INP_RUNLOCK(inp);
330                         }
331                         if (last != NULL) {
332                                 struct mbuf *n;
333
334                                 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
335                                         INP_RLOCK(last);
336                                         up = intoudpcb(last);
337                                         if (up->u_tun_func == NULL) {
338                                                 udp6_append(last, n, off, &fromsa);
339                                         } else {
340                                                 /*
341                                                  * Engage the tunneling
342                                                  * protocol we will have to
343                                                  * leave the info_lock up,
344                                                  * since we are hunting
345                                                  * through multiple UDP's.
346                                                  * 
347                                                  */
348                                                 (*up->u_tun_func)(n, off, last);
349                                         }
350                                         INP_RUNLOCK(last);
351                                 }
352                         }
353                         last = inp;
354                         /*
355                          * Don't look for additional matches if this one does
356                          * not have either the SO_REUSEPORT or SO_REUSEADDR
357                          * socket options set.  This heuristic avoids
358                          * searching through all pcbs in the common case of a
359                          * non-shared port.  It assumes that an application
360                          * will never clear these options after setting them.
361                          */
362                         if ((last->inp_socket->so_options &
363                              (SO_REUSEPORT|SO_REUSEADDR)) == 0)
364                                 break;
365                 }
366
367                 if (last == NULL) {
368                         /*
369                          * No matching pcb found; discard datagram.  (No need
370                          * to send an ICMP Port Unreachable for a broadcast
371                          * or multicast datgram.)
372                          */
373                         UDPSTAT_INC(udps_noport);
374                         UDPSTAT_INC(udps_noportmcast);
375                         goto badheadlocked;
376                 }
377                 INP_RLOCK(last);
378                 INP_INFO_RUNLOCK(&V_udbinfo);
379                 up = intoudpcb(last);
380                 if (up->u_tun_func == NULL) {
381                         udp6_append(last, m, off, &fromsa);
382                 } else {
383                         /*
384                          * Engage the tunneling protocol.
385                          */
386                         (*up->u_tun_func)(m, off, last);
387                 }
388                 INP_RUNLOCK(last);
389                 return (IPPROTO_DONE);
390         }
391         /*
392          * Locate pcb for datagram.
393          */
394
395         /*
396          * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
397          */
398         if ((m->m_flags & M_IP6_NEXTHOP) &&
399             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
400                 struct sockaddr_in6 *next_hop6;
401
402                 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
403
404                 /*
405                  * Transparently forwarded. Pretend to be the destination.
406                  * Already got one like this?
407                  */
408                 inp = in6_pcblookup_mbuf(&V_udbinfo,
409                     &ip6->ip6_src, uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
410                     INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
411                 if (!inp) {
412                         /*
413                          * It's new.  Try to find the ambushing socket.
414                          * Because we've rewritten the destination address,
415                          * any hardware-generated hash is ignored.
416                          */
417                         inp = in6_pcblookup(&V_udbinfo, &ip6->ip6_src,
418                             uh->uh_sport, &next_hop6->sin6_addr,
419                             next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
420                             uh->uh_dport, INPLOOKUP_WILDCARD |
421                             INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
422                 }
423                 /* Remove the tag from the packet. We don't need it anymore. */
424                 m_tag_delete(m, fwd_tag);
425                 m->m_flags &= ~M_IP6_NEXTHOP;
426         } else
427                 inp = in6_pcblookup_mbuf(&V_udbinfo, &ip6->ip6_src,
428                     uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
429                     INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
430                     m->m_pkthdr.rcvif, m);
431         if (inp == NULL) {
432                 if (udp_log_in_vain) {
433                         char ip6bufs[INET6_ADDRSTRLEN];
434                         char ip6bufd[INET6_ADDRSTRLEN];
435
436                         log(LOG_INFO,
437                             "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
438                             ip6_sprintf(ip6bufd, &ip6->ip6_dst),
439                             ntohs(uh->uh_dport),
440                             ip6_sprintf(ip6bufs, &ip6->ip6_src),
441                             ntohs(uh->uh_sport));
442                 }
443                 UDPSTAT_INC(udps_noport);
444                 if (m->m_flags & M_MCAST) {
445                         printf("UDP6: M_MCAST is set in a unicast packet.\n");
446                         UDPSTAT_INC(udps_noportmcast);
447                         goto badunlocked;
448                 }
449                 if (V_udp_blackhole)
450                         goto badunlocked;
451                 if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
452                         goto badunlocked;
453                 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
454                 return (IPPROTO_DONE);
455         }
456         INP_RLOCK_ASSERT(inp);
457         up = intoudpcb(inp);
458         if (up->u_tun_func == NULL) {
459                 udp6_append(inp, m, off, &fromsa);
460         } else {
461                 /*
462                  * Engage the tunneling protocol.
463                  */
464
465                 (*up->u_tun_func)(m, off, inp);
466         }
467         INP_RUNLOCK(inp);
468         return (IPPROTO_DONE);
469
470 badheadlocked:
471         INP_INFO_RUNLOCK(&V_udbinfo);
472 badunlocked:
473         if (m)
474                 m_freem(m);
475         return (IPPROTO_DONE);
476 }
477
478 void
479 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
480 {
481         struct udphdr uh;
482         struct ip6_hdr *ip6;
483         struct mbuf *m;
484         int off = 0;
485         struct ip6ctlparam *ip6cp = NULL;
486         const struct sockaddr_in6 *sa6_src = NULL;
487         void *cmdarg;
488         struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
489         struct udp_portonly {
490                 u_int16_t uh_sport;
491                 u_int16_t uh_dport;
492         } *uhp;
493
494         if (sa->sa_family != AF_INET6 ||
495             sa->sa_len != sizeof(struct sockaddr_in6))
496                 return;
497
498         if ((unsigned)cmd >= PRC_NCMDS)
499                 return;
500         if (PRC_IS_REDIRECT(cmd))
501                 notify = in6_rtchange, d = NULL;
502         else if (cmd == PRC_HOSTDEAD)
503                 d = NULL;
504         else if (inet6ctlerrmap[cmd] == 0)
505                 return;
506
507         /* if the parameter is from icmp6, decode it. */
508         if (d != NULL) {
509                 ip6cp = (struct ip6ctlparam *)d;
510                 m = ip6cp->ip6c_m;
511                 ip6 = ip6cp->ip6c_ip6;
512                 off = ip6cp->ip6c_off;
513                 cmdarg = ip6cp->ip6c_cmdarg;
514                 sa6_src = ip6cp->ip6c_src;
515         } else {
516                 m = NULL;
517                 ip6 = NULL;
518                 cmdarg = NULL;
519                 sa6_src = &sa6_any;
520         }
521
522         if (ip6) {
523                 /*
524                  * XXX: We assume that when IPV6 is non NULL,
525                  * M and OFF are valid.
526                  */
527
528                 /* Check if we can safely examine src and dst ports. */
529                 if (m->m_pkthdr.len < off + sizeof(*uhp))
530                         return;
531
532                 bzero(&uh, sizeof(uh));
533                 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
534
535                 (void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
536                     (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
537                     cmdarg, notify);
538         } else
539                 (void) in6_pcbnotify(&V_udbinfo, sa, 0,
540                     (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
541 }
542
543 static int
544 udp6_getcred(SYSCTL_HANDLER_ARGS)
545 {
546         struct xucred xuc;
547         struct sockaddr_in6 addrs[2];
548         struct inpcb *inp;
549         int error;
550
551         error = priv_check(req->td, PRIV_NETINET_GETCRED);
552         if (error)
553                 return (error);
554
555         if (req->newlen != sizeof(addrs))
556                 return (EINVAL);
557         if (req->oldlen != sizeof(struct xucred))
558                 return (EINVAL);
559         error = SYSCTL_IN(req, addrs, sizeof(addrs));
560         if (error)
561                 return (error);
562         if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
563             (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
564                 return (error);
565         }
566         inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
567             addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
568             INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
569         if (inp != NULL) {
570                 INP_RLOCK_ASSERT(inp);
571                 if (inp->inp_socket == NULL)
572                         error = ENOENT;
573                 if (error == 0)
574                         error = cr_canseesocket(req->td->td_ucred,
575                             inp->inp_socket);
576                 if (error == 0)
577                         cru2x(inp->inp_cred, &xuc);
578                 INP_RUNLOCK(inp);
579         } else
580                 error = ENOENT;
581         if (error == 0)
582                 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
583         return (error);
584 }
585
586 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
587     0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
588
589 static int
590 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
591     struct mbuf *control, struct thread *td)
592 {
593         u_int32_t ulen = m->m_pkthdr.len;
594         u_int32_t plen = sizeof(struct udphdr) + ulen;
595         struct ip6_hdr *ip6;
596         struct udphdr *udp6;
597         struct in6_addr *laddr, *faddr, in6a;
598         struct sockaddr_in6 *sin6 = NULL;
599         struct ifnet *oifp = NULL;
600         int scope_ambiguous = 0;
601         u_short fport;
602         int error = 0;
603         struct ip6_pktopts *optp, opt;
604         int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
605         int flags;
606         struct sockaddr_in6 tmp;
607
608         INP_WLOCK_ASSERT(inp);
609         INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
610
611         if (addr6) {
612                 /* addr6 has been validated in udp6_send(). */
613                 sin6 = (struct sockaddr_in6 *)addr6;
614
615                 /* protect *sin6 from overwrites */
616                 tmp = *sin6;
617                 sin6 = &tmp;
618
619                 /*
620                  * Application should provide a proper zone ID or the use of
621                  * default zone IDs should be enabled.  Unfortunately, some
622                  * applications do not behave as it should, so we need a
623                  * workaround.  Even if an appropriate ID is not determined,
624                  * we'll see if we can determine the outgoing interface.  If we
625                  * can, determine the zone ID based on the interface below.
626                  */
627                 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
628                         scope_ambiguous = 1;
629                 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
630                         return (error);
631         }
632
633         if (control) {
634                 if ((error = ip6_setpktopts(control, &opt,
635                     inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
636                         goto release;
637                 optp = &opt;
638         } else
639                 optp = inp->in6p_outputopts;
640
641         if (sin6) {
642                 faddr = &sin6->sin6_addr;
643
644                 /*
645                  * IPv4 version of udp_output calls in_pcbconnect in this case,
646                  * which needs splnet and affects performance.
647                  * Since we saw no essential reason for calling in_pcbconnect,
648                  * we get rid of such kind of logic, and call in6_selectsrc
649                  * and in6_pcbsetport in order to fill in the local address
650                  * and the local port.
651                  */
652                 if (sin6->sin6_port == 0) {
653                         error = EADDRNOTAVAIL;
654                         goto release;
655                 }
656
657                 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
658                         /* how about ::ffff:0.0.0.0 case? */
659                         error = EISCONN;
660                         goto release;
661                 }
662
663                 fport = sin6->sin6_port; /* allow 0 port */
664
665                 if (IN6_IS_ADDR_V4MAPPED(faddr)) {
666                         if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
667                                 /*
668                                  * I believe we should explicitly discard the
669                                  * packet when mapped addresses are disabled,
670                                  * rather than send the packet as an IPv6 one.
671                                  * If we chose the latter approach, the packet
672                                  * might be sent out on the wire based on the
673                                  * default route, the situation which we'd
674                                  * probably want to avoid.
675                                  * (20010421 jinmei@kame.net)
676                                  */
677                                 error = EINVAL;
678                                 goto release;
679                         }
680                         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
681                             !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
682                                 /*
683                                  * when remote addr is an IPv4-mapped address,
684                                  * local addr should not be an IPv6 address,
685                                  * since you cannot determine how to map IPv6
686                                  * source address to IPv4.
687                                  */
688                                 error = EINVAL;
689                                 goto release;
690                         }
691
692                         af = AF_INET;
693                 }
694
695                 if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
696                         error = in6_selectsrc(sin6, optp, inp, NULL,
697                             td->td_ucred, &oifp, &in6a);
698                         if (error)
699                                 goto release;
700                         if (oifp && scope_ambiguous &&
701                             (error = in6_setscope(&sin6->sin6_addr,
702                             oifp, NULL))) {
703                                 goto release;
704                         }
705                         laddr = &in6a;
706                 } else
707                         laddr = &inp->in6p_laddr;       /* XXX */
708                 if (laddr == NULL) {
709                         if (error == 0)
710                                 error = EADDRNOTAVAIL;
711                         goto release;
712                 }
713                 if (inp->inp_lport == 0 &&
714                     (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
715                         /* Undo an address bind that may have occurred. */
716                         inp->in6p_laddr = in6addr_any;
717                         goto release;
718                 }
719         } else {
720                 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
721                         error = ENOTCONN;
722                         goto release;
723                 }
724                 if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
725                         if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
726                                 /*
727                                  * XXX: this case would happen when the
728                                  * application sets the V6ONLY flag after
729                                  * connecting the foreign address.
730                                  * Such applications should be fixed,
731                                  * so we bark here.
732                                  */
733                                 log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
734                                     "option was set for a connected socket\n");
735                                 error = EINVAL;
736                                 goto release;
737                         } else
738                                 af = AF_INET;
739                 }
740                 laddr = &inp->in6p_laddr;
741                 faddr = &inp->in6p_faddr;
742                 fport = inp->inp_fport;
743         }
744
745         if (af == AF_INET)
746                 hlen = sizeof(struct ip);
747
748         /*
749          * Calculate data length and get a mbuf
750          * for UDP and IP6 headers.
751          */
752         M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
753         if (m == 0) {
754                 error = ENOBUFS;
755                 goto release;
756         }
757
758         /*
759          * Stuff checksum and output datagram.
760          */
761         udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
762         udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
763         udp6->uh_dport = fport;
764         if (plen <= 0xffff)
765                 udp6->uh_ulen = htons((u_short)plen);
766         else
767                 udp6->uh_ulen = 0;
768         udp6->uh_sum = 0;
769
770         switch (af) {
771         case AF_INET6:
772                 ip6 = mtod(m, struct ip6_hdr *);
773                 ip6->ip6_flow   = inp->inp_flow & IPV6_FLOWINFO_MASK;
774                 ip6->ip6_vfc    &= ~IPV6_VERSION_MASK;
775                 ip6->ip6_vfc    |= IPV6_VERSION;
776 #if 0                           /* ip6_plen will be filled in ip6_output. */
777                 ip6->ip6_plen   = htons((u_short)plen);
778 #endif
779                 ip6->ip6_nxt    = IPPROTO_UDP;
780                 ip6->ip6_hlim   = in6_selecthlim(inp, NULL);
781                 ip6->ip6_src    = *laddr;
782                 ip6->ip6_dst    = *faddr;
783
784                 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, IPPROTO_UDP, 0);
785                 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
786                 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
787
788                 flags = 0;
789
790                 UDPSTAT_INC(udps_opackets);
791                 error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
792                     NULL, inp);
793                 break;
794         case AF_INET:
795                 error = EAFNOSUPPORT;
796                 goto release;
797         }
798         goto releaseopt;
799
800 release:
801         m_freem(m);
802
803 releaseopt:
804         if (control) {
805                 ip6_clearpktopts(&opt, -1);
806                 m_freem(control);
807         }
808         return (error);
809 }
810
811 static void
812 udp6_abort(struct socket *so)
813 {
814         struct inpcb *inp;
815
816         inp = sotoinpcb(so);
817         KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
818
819 #ifdef INET
820         if (inp->inp_vflag & INP_IPV4) {
821                 struct pr_usrreqs *pru;
822
823                 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
824                 (*pru->pru_abort)(so);
825                 return;
826         }
827 #endif
828
829         INP_WLOCK(inp);
830         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
831                 INP_HASH_WLOCK(&V_udbinfo);
832                 in6_pcbdisconnect(inp);
833                 inp->in6p_laddr = in6addr_any;
834                 INP_HASH_WUNLOCK(&V_udbinfo);
835                 soisdisconnected(so);
836         }
837         INP_WUNLOCK(inp);
838 }
839
840 static int
841 udp6_attach(struct socket *so, int proto, struct thread *td)
842 {
843         struct inpcb *inp;
844         int error;
845
846         inp = sotoinpcb(so);
847         KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
848
849         if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
850                 error = soreserve(so, udp_sendspace, udp_recvspace);
851                 if (error)
852                         return (error);
853         }
854         INP_INFO_WLOCK(&V_udbinfo);
855         error = in_pcballoc(so, &V_udbinfo);
856         if (error) {
857                 INP_INFO_WUNLOCK(&V_udbinfo);
858                 return (error);
859         }
860         inp = (struct inpcb *)so->so_pcb;
861         inp->inp_vflag |= INP_IPV6;
862         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
863                 inp->inp_vflag |= INP_IPV4;
864         inp->in6p_hops = -1;    /* use kernel default */
865         inp->in6p_cksum = -1;   /* just to be sure */
866         /*
867          * XXX: ugly!!
868          * IPv4 TTL initialization is necessary for an IPv6 socket as well,
869          * because the socket may be bound to an IPv6 wildcard address,
870          * which may match an IPv4-mapped IPv6 address.
871          */
872         inp->inp_ip_ttl = V_ip_defttl;
873
874         error = udp_newudpcb(inp);
875         if (error) {
876                 in_pcbdetach(inp);
877                 in_pcbfree(inp);
878                 INP_INFO_WUNLOCK(&V_udbinfo);
879                 return (error);
880         }
881         INP_WUNLOCK(inp);
882         INP_INFO_WUNLOCK(&V_udbinfo);
883         return (0);
884 }
885
886 static int
887 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
888 {
889         struct inpcb *inp;
890         int error;
891
892         inp = sotoinpcb(so);
893         KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
894
895         INP_WLOCK(inp);
896         INP_HASH_WLOCK(&V_udbinfo);
897         inp->inp_vflag &= ~INP_IPV4;
898         inp->inp_vflag |= INP_IPV6;
899         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
900                 struct sockaddr_in6 *sin6_p;
901
902                 sin6_p = (struct sockaddr_in6 *)nam;
903
904                 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
905                         inp->inp_vflag |= INP_IPV4;
906 #ifdef INET
907                 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
908                         struct sockaddr_in sin;
909
910                         in6_sin6_2_sin(&sin, sin6_p);
911                         inp->inp_vflag |= INP_IPV4;
912                         inp->inp_vflag &= ~INP_IPV6;
913                         error = in_pcbbind(inp, (struct sockaddr *)&sin,
914                             td->td_ucred);
915                         goto out;
916                 }
917 #endif
918         }
919
920         error = in6_pcbbind(inp, nam, td->td_ucred);
921 #ifdef INET
922 out:
923 #endif
924         INP_HASH_WUNLOCK(&V_udbinfo);
925         INP_WUNLOCK(inp);
926         return (error);
927 }
928
929 static void
930 udp6_close(struct socket *so)
931 {
932         struct inpcb *inp;
933
934         inp = sotoinpcb(so);
935         KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
936
937 #ifdef INET
938         if (inp->inp_vflag & INP_IPV4) {
939                 struct pr_usrreqs *pru;
940
941                 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
942                 (*pru->pru_disconnect)(so);
943                 return;
944         }
945 #endif
946         INP_WLOCK(inp);
947         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
948                 INP_HASH_WLOCK(&V_udbinfo);
949                 in6_pcbdisconnect(inp);
950                 inp->in6p_laddr = in6addr_any;
951                 INP_HASH_WUNLOCK(&V_udbinfo);
952                 soisdisconnected(so);
953         }
954         INP_WUNLOCK(inp);
955 }
956
957 static int
958 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
959 {
960         struct inpcb *inp;
961         struct sockaddr_in6 *sin6;
962         int error;
963
964         inp = sotoinpcb(so);
965         sin6 = (struct sockaddr_in6 *)nam;
966         KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
967
968         /*
969          * XXXRW: Need to clarify locking of v4/v6 flags.
970          */
971         INP_WLOCK(inp);
972 #ifdef INET
973         if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
974                 struct sockaddr_in sin;
975
976                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
977                         error = EINVAL;
978                         goto out;
979                 }
980                 if (inp->inp_faddr.s_addr != INADDR_ANY) {
981                         error = EISCONN;
982                         goto out;
983                 }
984                 in6_sin6_2_sin(&sin, sin6);
985                 inp->inp_vflag |= INP_IPV4;
986                 inp->inp_vflag &= ~INP_IPV6;
987                 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
988                 if (error != 0)
989                         goto out;
990                 INP_HASH_WLOCK(&V_udbinfo);
991                 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
992                     td->td_ucred);
993                 INP_HASH_WUNLOCK(&V_udbinfo);
994                 if (error == 0)
995                         soisconnected(so);
996                 goto out;
997         }
998 #endif
999         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1000                 error = EISCONN;
1001                 goto out;
1002         }
1003         inp->inp_vflag &= ~INP_IPV4;
1004         inp->inp_vflag |= INP_IPV6;
1005         error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1006         if (error != 0)
1007                 goto out;
1008         INP_HASH_WLOCK(&V_udbinfo);
1009         error = in6_pcbconnect(inp, nam, td->td_ucred);
1010         INP_HASH_WUNLOCK(&V_udbinfo);
1011         if (error == 0)
1012                 soisconnected(so);
1013 out:
1014         INP_WUNLOCK(inp);
1015         return (error);
1016 }
1017
1018 static void
1019 udp6_detach(struct socket *so)
1020 {
1021         struct inpcb *inp;
1022         struct udpcb *up;
1023
1024         inp = sotoinpcb(so);
1025         KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1026
1027         INP_INFO_WLOCK(&V_udbinfo);
1028         INP_WLOCK(inp);
1029         up = intoudpcb(inp);
1030         KASSERT(up != NULL, ("%s: up == NULL", __func__));
1031         in_pcbdetach(inp);
1032         in_pcbfree(inp);
1033         INP_INFO_WUNLOCK(&V_udbinfo);
1034         udp_discardcb(up);
1035 }
1036
1037 static int
1038 udp6_disconnect(struct socket *so)
1039 {
1040         struct inpcb *inp;
1041         int error;
1042
1043         inp = sotoinpcb(so);
1044         KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1045
1046 #ifdef INET
1047         if (inp->inp_vflag & INP_IPV4) {
1048                 struct pr_usrreqs *pru;
1049
1050                 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1051                 (void)(*pru->pru_disconnect)(so);
1052                 return (0);
1053         }
1054 #endif
1055
1056         INP_WLOCK(inp);
1057
1058         if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1059                 error = ENOTCONN;
1060                 goto out;
1061         }
1062
1063         INP_HASH_WLOCK(&V_udbinfo);
1064         in6_pcbdisconnect(inp);
1065         inp->in6p_laddr = in6addr_any;
1066         INP_HASH_WUNLOCK(&V_udbinfo);
1067         SOCK_LOCK(so);
1068         so->so_state &= ~SS_ISCONNECTED;                /* XXX */
1069         SOCK_UNLOCK(so);
1070 out:
1071         INP_WUNLOCK(inp);
1072         return (0);
1073 }
1074
1075 static int
1076 udp6_send(struct socket *so, int flags, struct mbuf *m,
1077     struct sockaddr *addr, struct mbuf *control, struct thread *td)
1078 {
1079         struct inpcb *inp;
1080         int error = 0;
1081
1082         inp = sotoinpcb(so);
1083         KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1084
1085         INP_WLOCK(inp);
1086         if (addr) {
1087                 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1088                         error = EINVAL;
1089                         goto bad;
1090                 }
1091                 if (addr->sa_family != AF_INET6) {
1092                         error = EAFNOSUPPORT;
1093                         goto bad;
1094                 }
1095         }
1096
1097 #ifdef INET
1098         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1099                 int hasv4addr;
1100                 struct sockaddr_in6 *sin6 = 0;
1101
1102                 if (addr == 0)
1103                         hasv4addr = (inp->inp_vflag & INP_IPV4);
1104                 else {
1105                         sin6 = (struct sockaddr_in6 *)addr;
1106                         hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1107                             ? 1 : 0;
1108                 }
1109                 if (hasv4addr) {
1110                         struct pr_usrreqs *pru;
1111
1112                         /*
1113                          * XXXRW: We release UDP-layer locks before calling
1114                          * udp_send() in order to avoid recursion.  However,
1115                          * this does mean there is a short window where inp's
1116                          * fields are unstable.  Could this lead to a
1117                          * potential race in which the factors causing us to
1118                          * select the UDPv4 output routine are invalidated?
1119                          */
1120                         INP_WUNLOCK(inp);
1121                         if (sin6)
1122                                 in6_sin6_2_sin_in_sock(addr);
1123                         pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1124                         /* addr will just be freed in sendit(). */
1125                         return ((*pru->pru_send)(so, flags, m, addr, control,
1126                             td));
1127                 }
1128         }
1129 #endif
1130 #ifdef MAC
1131         mac_inpcb_create_mbuf(inp, m);
1132 #endif
1133         INP_HASH_WLOCK(&V_udbinfo);
1134         error = udp6_output(inp, m, addr, control, td);
1135         INP_HASH_WUNLOCK(&V_udbinfo);
1136 #ifdef INET
1137 #endif  
1138         INP_WUNLOCK(inp);
1139         return (error);
1140
1141 bad:
1142         INP_WUNLOCK(inp);
1143         m_freem(m);
1144         return (error);
1145 }
1146
1147 struct pr_usrreqs udp6_usrreqs = {
1148         .pru_abort =            udp6_abort,
1149         .pru_attach =           udp6_attach,
1150         .pru_bind =             udp6_bind,
1151         .pru_connect =          udp6_connect,
1152         .pru_control =          in6_control,
1153         .pru_detach =           udp6_detach,
1154         .pru_disconnect =       udp6_disconnect,
1155         .pru_peeraddr =         in6_mapped_peeraddr,
1156         .pru_send =             udp6_send,
1157         .pru_shutdown =         udp_shutdown,
1158         .pru_sockaddr =         in6_mapped_sockaddr,
1159         .pru_soreceive =        soreceive_dgram,
1160         .pru_sosend =           sosend_dgram,
1161         .pru_sosetlabel =       in_pcbsosetlabel,
1162         .pru_close =            udp6_close
1163 };