1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Copyright (c) 2010-2011 Juniper Networks, Inc.
6 * Copyright (c) 2014 Kevin Lo
7 * All rights reserved.
8 *
9 * Portions of this software were developed by Robert N. M. Watson under
10 * contract to Juniper Networks, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the project nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * $KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
37 * $KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
38 */
39
40 /*-
41 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
42 * The Regents of the University of California.
43 * All rights reserved.
44 *
45 * Redistribution and use in source and binary forms, with or without
46 * modification, are permitted provided that the following conditions
47 * are met:
48 * 1. Redistributions of source code must retain the above copyright
49 * notice, this list of conditions and the following disclaimer.
50 * 2. Redistributions in binary form must reproduce the above copyright
51 * notice, this list of conditions and the following disclaimer in the
52 * documentation and/or other materials provided with the distribution.
53 * 3. Neither the name of the University nor the names of its contributors
54 * may be used to endorse or promote products derived from this software
55 * without specific prior written permission.
56 *
57 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
58 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
59 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
60 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
61 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
62 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
63 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
64 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
65 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
66 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
67 * SUCH DAMAGE.
68 *
69 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
70 */
71
72 #include <sys/cdefs.h>
73 __FBSDID("$FreeBSD: stable/12/sys/netinet6/udp6_usrreq.c 372541 2022-09-13 05:29:03Z gbe $");
74
75 #include "opt_inet.h"
76 #include "opt_inet6.h"
77 #include "opt_ipsec.h"
78 #include "opt_rss.h"
79
80 #include <sys/param.h>
81 #include <sys/jail.h>
82 #include <sys/kernel.h>
83 #include <sys/lock.h>
84 #include <sys/mbuf.h>
85 #include <sys/priv.h>
86 #include <sys/proc.h>
87 #include <sys/protosw.h>
88 #include <sys/sdt.h>
89 #include <sys/signalvar.h>
90 #include <sys/socket.h>
91 #include <sys/socketvar.h>
92 #include <sys/sx.h>
93 #include <sys/sysctl.h>
94 #include <sys/syslog.h>
95 #include <sys/systm.h>
96
97 #include <net/if.h>
98 #include <net/if_var.h>
99 #include <net/if_types.h>
100 #include <net/route.h>
101 #include <net/rss_config.h>
102
103 #include <netinet/in.h>
104 #include <netinet/in_kdtrace.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/in_systm.h>
107 #include <netinet/in_var.h>
108 #include <netinet/ip.h>
109 #include <netinet/ip6.h>
110 #include <netinet/icmp6.h>
111 #include <netinet/ip_var.h>
112 #include <netinet/udp.h>
113 #include <netinet/udp_var.h>
114 #include <netinet/udplite.h>
115
116 #include <netinet6/ip6protosw.h>
117 #include <netinet6/ip6_var.h>
118 #include <netinet6/in6_pcb.h>
119 #include <netinet6/in6_rss.h>
120 #include <netinet6/udp6_var.h>
121 #include <netinet6/scope6_var.h>
122
123 #include <netipsec/ipsec_support.h>
124
125 #include <security/mac/mac_framework.h>
126
127 /*
128 * UDP protocol implementation.
129 * Per RFC 768, August, 1980.
130 */
131
132 extern struct protosw inetsw[];
133 static void udp6_detach(struct socket *so);
134
135 static int
udp6_append(struct inpcb * inp,struct mbuf * n,int off,struct sockaddr_in6 * fromsa)136 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
137 struct sockaddr_in6 *fromsa)
138 {
139 struct socket *so;
140 struct mbuf *opts = NULL, *tmp_opts;
141 struct udpcb *up;
142
143 INP_LOCK_ASSERT(inp);
144
145 /*
146 * Engage the tunneling protocol.
147 */
148 up = intoudpcb(inp);
149 if (up->u_tun_func != NULL) {
150 in_pcbref(inp);
151 INP_RUNLOCK(inp);
152 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0],
153 up->u_tun_ctx);
154 INP_RLOCK(inp);
155 return (in_pcbrele_rlocked(inp));
156 }
157 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
158 /* Check AH/ESP integrity. */
159 if (IPSEC_ENABLED(ipv6)) {
160 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
161 m_freem(n);
162 return (0);
163 }
164 }
165 #endif /* IPSEC */
166 #ifdef MAC
167 if (mac_inpcb_check_deliver(inp, n) != 0) {
168 m_freem(n);
169 return (0);
170 }
171 #endif
172 opts = NULL;
173 if (inp->inp_flags & INP_CONTROLOPTS ||
174 inp->inp_socket->so_options & SO_TIMESTAMP)
175 ip6_savecontrol(inp, n, &opts);
176 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
177 tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1],
178 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6);
179 if (tmp_opts) {
180 if (opts) {
181 tmp_opts->m_next = opts;
182 opts = tmp_opts;
183 } else
184 opts = tmp_opts;
185 }
186
187 }
188 m_adj(n, off + sizeof(struct udphdr));
189
190 so = inp->inp_socket;
191 SOCKBUF_LOCK(&so->so_rcv);
192 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
193 opts) == 0) {
194 soroverflow_locked(so);
195 m_freem(n);
196 if (opts)
197 m_freem(opts);
198 UDPSTAT_INC(udps_fullsock);
199 } else
200 sorwakeup_locked(so);
201 return (0);
202 }
203
204 int
udp6_input(struct mbuf ** mp,int * offp,int proto)205 udp6_input(struct mbuf **mp, int *offp, int proto)
206 {
207 struct mbuf *m = *mp;
208 struct ifnet *ifp;
209 struct ip6_hdr *ip6;
210 struct udphdr *uh;
211 struct inpcb *inp;
212 struct inpcbinfo *pcbinfo;
213 struct udpcb *up;
214 int off = *offp;
215 int cscov_partial;
216 int plen, ulen;
217 struct epoch_tracker et;
218 struct sockaddr_in6 fromsa[2];
219 struct m_tag *fwd_tag;
220 uint16_t uh_sum;
221 uint8_t nxt;
222
223 ifp = m->m_pkthdr.rcvif;
224
225 if (m->m_len < off + sizeof(struct udphdr)) {
226 m = m_pullup(m, off + sizeof(struct udphdr));
227 if (m == NULL) {
228 IP6STAT_INC(ip6s_exthdrtoolong);
229 *mp = NULL;
230 return (IPPROTO_DONE);
231 }
232 }
233 ip6 = mtod(m, struct ip6_hdr *);
234 uh = (struct udphdr *)((caddr_t)ip6 + off);
235
236 UDPSTAT_INC(udps_ipackets);
237
238 /*
239 * Destination port of 0 is illegal, based on RFC768.
240 */
241 if (uh->uh_dport == 0)
242 goto badunlocked;
243
244 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
245 ulen = ntohs((u_short)uh->uh_ulen);
246
247 nxt = proto;
248 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
249 if (nxt == IPPROTO_UDPLITE) {
250 /* Zero means checksum over the complete packet. */
251 if (ulen == 0)
252 ulen = plen;
253 if (ulen == plen)
254 cscov_partial = 0;
255 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
256 /* XXX: What is the right UDPLite MIB counter? */
257 goto badunlocked;
258 }
259 if (uh->uh_sum == 0) {
260 /* XXX: What is the right UDPLite MIB counter? */
261 goto badunlocked;
262 }
263 } else {
264 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
265 UDPSTAT_INC(udps_badlen);
266 goto badunlocked;
267 }
268 if (uh->uh_sum == 0) {
269 UDPSTAT_INC(udps_nosum);
270 goto badunlocked;
271 }
272 }
273
274 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
275 !cscov_partial) {
276 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
277 uh_sum = m->m_pkthdr.csum_data;
278 else
279 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
280 m->m_pkthdr.csum_data);
281 uh_sum ^= 0xffff;
282 } else
283 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
284
285 if (uh_sum != 0) {
286 UDPSTAT_INC(udps_badsum);
287 goto badunlocked;
288 }
289
290 /*
291 * Construct sockaddr format source address.
292 */
293 init_sin6(&fromsa[0], m, 0);
294 fromsa[0].sin6_port = uh->uh_sport;
295 init_sin6(&fromsa[1], m, 1);
296 fromsa[1].sin6_port = uh->uh_dport;
297
298 pcbinfo = udp_get_inpcbinfo(nxt);
299 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
300 struct inpcb *last;
301 struct inpcbhead *pcblist;
302 struct ip6_moptions *imo;
303
304 INP_INFO_RLOCK_ET(pcbinfo, et);
305 /*
306 * In the event that laddr should be set to the link-local
307 * address (this happens in RIPng), the multicast address
308 * specified in the received packet will not match laddr. To
309 * handle this situation, matching is relaxed if the
310 * receiving interface is the same as one specified in the
311 * socket and if the destination multicast address matches
312 * one of the multicast groups specified in the socket.
313 */
314
315 /*
316 * KAME note: traditionally we dropped udpiphdr from mbuf
317 * here. We need udphdr for IPsec processing so we do that
318 * later.
319 */
320 pcblist = udp_get_pcblist(nxt);
321 last = NULL;
322 CK_LIST_FOREACH(inp, pcblist, inp_list) {
323 if ((inp->inp_vflag & INP_IPV6) == 0)
324 continue;
325 if (inp->inp_lport != uh->uh_dport)
326 continue;
327 if (inp->inp_fport != 0 &&
328 inp->inp_fport != uh->uh_sport)
329 continue;
330 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
331 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
332 &ip6->ip6_dst))
333 continue;
334 }
335 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
336 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
337 &ip6->ip6_src) ||
338 inp->inp_fport != uh->uh_sport)
339 continue;
340 }
341
342 /*
343 * XXXRW: Because we weren't holding either the inpcb
344 * or the hash lock when we checked for a match
345 * before, we should probably recheck now that the
346 * inpcb lock is (supposed to be) held.
347 */
348
349 /*
350 * Handle socket delivery policy for any-source
351 * and source-specific multicast. [RFC3678]
352 */
353 imo = inp->in6p_moptions;
354 if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
355 struct sockaddr_in6 mcaddr;
356 int blocked;
357
358 INP_RLOCK(inp);
359 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
360 INP_RUNLOCK(inp);
361 continue;
362 }
363
364 bzero(&mcaddr, sizeof(struct sockaddr_in6));
365 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
366 mcaddr.sin6_family = AF_INET6;
367 mcaddr.sin6_addr = ip6->ip6_dst;
368
369 blocked = im6o_mc_filter(imo, ifp,
370 (struct sockaddr *)&mcaddr,
371 (struct sockaddr *)&fromsa[0]);
372 if (blocked != MCAST_PASS) {
373 if (blocked == MCAST_NOTGMEMBER)
374 IP6STAT_INC(ip6s_notmember);
375 if (blocked == MCAST_NOTSMEMBER ||
376 blocked == MCAST_MUTED)
377 UDPSTAT_INC(udps_filtermcast);
378 INP_RUNLOCK(inp); /* XXX */
379 continue;
380 }
381
382 INP_RUNLOCK(inp);
383 }
384 if (last != NULL) {
385 struct mbuf *n;
386
387 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
388 NULL) {
389 INP_RLOCK(last);
390 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
391 if (nxt == IPPROTO_UDPLITE)
392 UDPLITE_PROBE(receive, NULL, last,
393 ip6, last, uh);
394 else
395 UDP_PROBE(receive, NULL, last,
396 ip6, last, uh);
397 if (udp6_append(last, n, off, fromsa)) {
398 /* XXX-BZ do we leak m here? */
399 *mp = NULL;
400 goto inp_lost;
401 }
402 }
403 INP_RUNLOCK(last);
404 }
405 }
406 last = inp;
407 /*
408 * Don't look for additional matches if this one does
409 * not have either the SO_REUSEPORT or SO_REUSEADDR
410 * socket options set. This heuristic avoids
411 * searching through all pcbs in the common case of a
412 * non-shared port. It assumes that an application
413 * will never clear these options after setting them.
414 */
415 if ((last->inp_socket->so_options &
416 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0)
417 break;
418 }
419
420 if (last == NULL) {
421 /*
422 * No matching pcb found; discard datagram. (No need
423 * to send an ICMP Port Unreachable for a broadcast
424 * or multicast datgram.)
425 */
426 UDPSTAT_INC(udps_noport);
427 UDPSTAT_INC(udps_noportmcast);
428 goto badheadlocked;
429 }
430 INP_RLOCK(last);
431 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
432 if (nxt == IPPROTO_UDPLITE)
433 UDPLITE_PROBE(receive, NULL, last, ip6, last, uh);
434 else
435 UDP_PROBE(receive, NULL, last, ip6, last, uh);
436 if (udp6_append(last, m, off, fromsa) == 0)
437 INP_RUNLOCK(last);
438 } else
439 INP_RUNLOCK(last);
440 inp_lost:
441 INP_INFO_RUNLOCK_ET(pcbinfo, et);
442 *mp = NULL;
443 return (IPPROTO_DONE);
444 }
445 /*
446 * Locate pcb for datagram.
447 */
448
449 /*
450 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
451 */
452 if ((m->m_flags & M_IP6_NEXTHOP) &&
453 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
454 struct sockaddr_in6 *next_hop6;
455
456 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
457
458 /*
459 * Transparently forwarded. Pretend to be the destination.
460 * Already got one like this?
461 */
462 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
463 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
464 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
465 if (!inp) {
466 /*
467 * It's new. Try to find the ambushing socket.
468 * Because we've rewritten the destination address,
469 * any hardware-generated hash is ignored.
470 */
471 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
472 uh->uh_sport, &next_hop6->sin6_addr,
473 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
474 uh->uh_dport, INPLOOKUP_WILDCARD |
475 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
476 }
477 /* Remove the tag from the packet. We don't need it anymore. */
478 m_tag_delete(m, fwd_tag);
479 m->m_flags &= ~M_IP6_NEXTHOP;
480 } else
481 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
482 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
483 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
484 m->m_pkthdr.rcvif, m);
485 if (inp == NULL) {
486 if (V_udp_log_in_vain) {
487 char ip6bufs[INET6_ADDRSTRLEN];
488 char ip6bufd[INET6_ADDRSTRLEN];
489
490 log(LOG_INFO,
491 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
492 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
493 ntohs(uh->uh_dport),
494 ip6_sprintf(ip6bufs, &ip6->ip6_src),
495 ntohs(uh->uh_sport));
496 }
497 if (nxt == IPPROTO_UDPLITE)
498 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
499 else
500 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
501 UDPSTAT_INC(udps_noport);
502 if (m->m_flags & M_MCAST) {
503 printf("UDP6: M_MCAST is set in a unicast packet.\n");
504 UDPSTAT_INC(udps_noportmcast);
505 goto badunlocked;
506 }
507 if (V_udp_blackhole)
508 goto badunlocked;
509 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
510 *mp = NULL;
511 return (IPPROTO_DONE);
512 }
513 INP_RLOCK_ASSERT(inp);
514 up = intoudpcb(inp);
515 if (cscov_partial) {
516 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
517 INP_RUNLOCK(inp);
518 m_freem(m);
519 *mp = NULL;
520 return (IPPROTO_DONE);
521 }
522 }
523 if (nxt == IPPROTO_UDPLITE)
524 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
525 else
526 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
527 if (udp6_append(inp, m, off, fromsa) == 0)
528 INP_RUNLOCK(inp);
529 *mp = NULL;
530 return (IPPROTO_DONE);
531
532 badheadlocked:
533 INP_INFO_RUNLOCK_ET(pcbinfo, et);
534 badunlocked:
535 if (m)
536 m_freem(m);
537 *mp = NULL;
538 return (IPPROTO_DONE);
539 }
540
541 static void
udp6_common_ctlinput(int cmd,struct sockaddr * sa,void * d,struct inpcbinfo * pcbinfo)542 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
543 struct inpcbinfo *pcbinfo)
544 {
545 struct udphdr uh;
546 struct ip6_hdr *ip6;
547 struct mbuf *m;
548 int off = 0;
549 struct ip6ctlparam *ip6cp = NULL;
550 const struct sockaddr_in6 *sa6_src = NULL;
551 void *cmdarg;
552 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
553 struct udp_portonly {
554 u_int16_t uh_sport;
555 u_int16_t uh_dport;
556 } *uhp;
557
558 if (sa->sa_family != AF_INET6 ||
559 sa->sa_len != sizeof(struct sockaddr_in6))
560 return;
561
562 if ((unsigned)cmd >= PRC_NCMDS)
563 return;
564 if (PRC_IS_REDIRECT(cmd))
565 notify = in6_rtchange, d = NULL;
566 else if (cmd == PRC_HOSTDEAD)
567 d = NULL;
568 else if (inet6ctlerrmap[cmd] == 0)
569 return;
570
571 /* if the parameter is from icmp6, decode it. */
572 if (d != NULL) {
573 ip6cp = (struct ip6ctlparam *)d;
574 m = ip6cp->ip6c_m;
575 ip6 = ip6cp->ip6c_ip6;
576 off = ip6cp->ip6c_off;
577 cmdarg = ip6cp->ip6c_cmdarg;
578 sa6_src = ip6cp->ip6c_src;
579 } else {
580 m = NULL;
581 ip6 = NULL;
582 cmdarg = NULL;
583 sa6_src = &sa6_any;
584 }
585
586 if (ip6) {
587 /*
588 * XXX: We assume that when IPV6 is non NULL,
589 * M and OFF are valid.
590 */
591
592 /* Check if we can safely examine src and dst ports. */
593 if (m->m_pkthdr.len < off + sizeof(*uhp))
594 return;
595
596 bzero(&uh, sizeof(uh));
597 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
598
599 if (!PRC_IS_REDIRECT(cmd)) {
600 /* Check to see if its tunneled */
601 struct inpcb *inp;
602 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
603 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
604 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
605 m->m_pkthdr.rcvif, m);
606 if (inp != NULL) {
607 struct udpcb *up;
608
609 up = intoudpcb(inp);
610 if (up->u_icmp_func) {
611 /* Yes it is. */
612 INP_RUNLOCK(inp);
613 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
614 d, up->u_tun_ctx);
615 return;
616 } else {
617 /* Can't find it. */
618 INP_RUNLOCK(inp);
619 }
620 }
621 }
622 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
623 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
624 cmdarg, notify);
625 } else
626 (void)in6_pcbnotify(pcbinfo, sa, 0,
627 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
628 }
629
630 void
udp6_ctlinput(int cmd,struct sockaddr * sa,void * d)631 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
632 {
633
634 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
635 }
636
637 void
udplite6_ctlinput(int cmd,struct sockaddr * sa,void * d)638 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
639 {
640
641 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
642 }
643
644 static int
udp6_getcred(SYSCTL_HANDLER_ARGS)645 udp6_getcred(SYSCTL_HANDLER_ARGS)
646 {
647 struct xucred xuc;
648 struct sockaddr_in6 addrs[2];
649 struct inpcb *inp;
650 int error;
651
652 error = priv_check(req->td, PRIV_NETINET_GETCRED);
653 if (error)
654 return (error);
655
656 if (req->newlen != sizeof(addrs))
657 return (EINVAL);
658 if (req->oldlen != sizeof(struct xucred))
659 return (EINVAL);
660 error = SYSCTL_IN(req, addrs, sizeof(addrs));
661 if (error)
662 return (error);
663 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
664 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
665 return (error);
666 }
667 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
668 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
669 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
670 if (inp != NULL) {
671 INP_RLOCK_ASSERT(inp);
672 if (inp->inp_socket == NULL)
673 error = ENOENT;
674 if (error == 0)
675 error = cr_canseesocket(req->td->td_ucred,
676 inp->inp_socket);
677 if (error == 0)
678 cru2x(inp->inp_cred, &xuc);
679 INP_RUNLOCK(inp);
680 } else
681 error = ENOENT;
682 if (error == 0)
683 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
684 return (error);
685 }
686
687 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
688 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
689
690 #define UH_WLOCKED 2
691 #define UH_RLOCKED 1
692 #define UH_UNLOCKED 0
693 static int
udp6_output(struct socket * so,int flags_arg,struct mbuf * m,struct sockaddr * addr6,struct mbuf * control,struct thread * td)694 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
695 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
696 {
697 struct inpcbinfo *pcbinfo;
698 struct inpcb *inp;
699 struct ip6_hdr *ip6;
700 struct udphdr *udp6;
701 struct in6_addr *laddr, *faddr, in6a;
702 struct ip6_pktopts *optp, opt;
703 struct sockaddr_in6 *sin6, tmp;
704 struct epoch_tracker et;
705 int cscov_partial, error, flags, hlen, scope_ambiguous;
706 u_int32_t ulen, plen;
707 uint16_t cscov;
708 u_short fport;
709 uint8_t nxt, unlock_inp, unlock_udbinfo;
710
711 /* addr6 has been validated in udp6_send(). */
712 sin6 = (struct sockaddr_in6 *)addr6;
713
714 /*
715 * In contrast to IPv4 we do not validate the max. packet length
716 * here due to IPv6 Jumbograms (RFC2675).
717 */
718
719 scope_ambiguous = 0;
720 if (sin6) {
721 /* Protect *addr6 from overwrites. */
722 tmp = *sin6;
723 sin6 = &tmp;
724
725 /*
726 * Application should provide a proper zone ID or the use of
727 * default zone IDs should be enabled. Unfortunately, some
728 * applications do not behave as it should, so we need a
729 * workaround. Even if an appropriate ID is not determined,
730 * we'll see if we can determine the outgoing interface. If we
731 * can, determine the zone ID based on the interface below.
732 */
733 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
734 scope_ambiguous = 1;
735 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
736 if (control)
737 m_freem(control);
738 m_freem(m);
739 return (error);
740 }
741 }
742
743 inp = sotoinpcb(so);
744 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
745 /*
746 * In the following cases we want a write lock on the inp for either
747 * local operations or for possible route cache updates in the IPv6
748 * output path:
749 * - on connected sockets (sin6 is NULL) for route cache updates,
750 * - when we are not bound to an address and source port (it is
751 * in6_pcbsetport() which will require the write lock).
752 */
753 retry:
754 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
755 inp->inp_lport == 0)) {
756 INP_WLOCK(inp);
757 /*
758 * In case we lost a race and another thread bound addr/port
759 * on the inp we cannot keep the wlock (which still would be
760 * fine) as further down, based on these values we make
761 * decisions for the pcbinfo lock. If the locks are not in
762 * synch the assertions on unlock will fire, hence we go for
763 * one retry loop.
764 */
765 if (sin6 != NULL &&
766 (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ||
767 inp->inp_lport != 0)) {
768 INP_WUNLOCK(inp);
769 goto retry;
770 }
771 unlock_inp = UH_WLOCKED;
772 } else {
773 INP_RLOCK(inp);
774 unlock_inp = UH_RLOCKED;
775 }
776 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
777 IPPROTO_UDP : IPPROTO_UDPLITE;
778
779 #ifdef INET
780 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
781 int hasv4addr;
782
783 if (sin6 == NULL)
784 hasv4addr = (inp->inp_vflag & INP_IPV4);
785 else
786 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
787 ? 1 : 0;
788 if (hasv4addr) {
789 struct pr_usrreqs *pru;
790
791 /*
792 * XXXRW: We release UDP-layer locks before calling
793 * udp_send() in order to avoid recursion. However,
794 * this does mean there is a short window where inp's
795 * fields are unstable. Could this lead to a
796 * potential race in which the factors causing us to
797 * select the UDPv4 output routine are invalidated?
798 */
799 if (unlock_inp == UH_WLOCKED)
800 INP_WUNLOCK(inp);
801 else
802 INP_RUNLOCK(inp);
803 if (sin6)
804 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
805 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
806 /* addr will just be freed in sendit(). */
807 return ((*pru->pru_send)(so, flags_arg | PRUS_IPV6, m,
808 (struct sockaddr *)sin6, control, td));
809 }
810 } else
811 #endif
812 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
813 /*
814 * Given this is either an IPv6-only socket or no INET is
815 * supported we will fail the send if the given destination
816 * address is a v4mapped address.
817 */
818 if (unlock_inp == UH_WLOCKED)
819 INP_WUNLOCK(inp);
820 else
821 INP_RUNLOCK(inp);
822 return (EINVAL);
823 }
824
825 if (control) {
826 if ((error = ip6_setpktopts(control, &opt,
827 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
828 if (unlock_inp == UH_WLOCKED)
829 INP_WUNLOCK(inp);
830 else
831 INP_RUNLOCK(inp);
832 ip6_clearpktopts(&opt, -1);
833 if (control)
834 m_freem(control);
835 m_freem(m);
836 return (error);
837 }
838 optp = &opt;
839 } else
840 optp = inp->in6p_outputopts;
841
842 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
843 if (sin6 != NULL &&
844 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) && inp->inp_lport == 0) {
845 INP_HASH_WLOCK(pcbinfo);
846 unlock_udbinfo = UH_WLOCKED;
847 } else if (sin6 != NULL &&
848 (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
849 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ||
850 inp->inp_lport == 0)) {
851 INP_HASH_RLOCK_ET(pcbinfo, et);
852 unlock_udbinfo = UH_RLOCKED;
853 } else
854 unlock_udbinfo = UH_UNLOCKED;
855
856 if (sin6) {
857
858 /*
859 * Since we saw no essential reason for calling in_pcbconnect,
860 * we get rid of such kind of logic, and call in6_selectsrc
861 * and in6_pcbsetport in order to fill in the local address
862 * and the local port.
863 */
864 if (sin6->sin6_port == 0) {
865 error = EADDRNOTAVAIL;
866 goto release;
867 }
868
869 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
870 /* how about ::ffff:0.0.0.0 case? */
871 error = EISCONN;
872 goto release;
873 }
874
875 /*
876 * Given we handle the v4mapped case in the INET block above
877 * assert here that it must not happen anymore.
878 */
879 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
880 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
881 "should have handled.", __func__, sin6));
882
883 /* This only requires read-locking. */
884 error = in6_selectsrc_socket(sin6, optp, inp,
885 td->td_ucred, scope_ambiguous, &in6a, NULL);
886 if (error)
887 goto release;
888 laddr = &in6a;
889
890 if (inp->inp_lport == 0) {
891
892 INP_WLOCK_ASSERT(inp);
893 error = in6_pcbsetport(laddr, inp, td->td_ucred);
894 if (error != 0) {
895 /* Undo an address bind that may have occurred. */
896 inp->in6p_laddr = in6addr_any;
897 goto release;
898 }
899 }
900 faddr = &sin6->sin6_addr;
901 fport = sin6->sin6_port; /* allow 0 port */
902
903 } else {
904 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
905 error = ENOTCONN;
906 goto release;
907 }
908 laddr = &inp->in6p_laddr;
909 faddr = &inp->in6p_faddr;
910 fport = inp->inp_fport;
911 }
912
913 ulen = m->m_pkthdr.len;
914 plen = sizeof(struct udphdr) + ulen;
915 hlen = sizeof(struct ip6_hdr);
916
917 /*
918 * Calculate data length and get a mbuf
919 * for UDP and IP6 headers.
920 */
921 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
922 if (m == NULL) {
923 error = ENOBUFS;
924 goto release;
925 }
926
927 /*
928 * Stuff checksum and output datagram.
929 */
930 cscov = cscov_partial = 0;
931 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
932 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
933 udp6->uh_dport = fport;
934 if (nxt == IPPROTO_UDPLITE) {
935 struct udpcb *up;
936
937 up = intoudpcb(inp);
938 cscov = up->u_txcslen;
939 if (cscov >= plen)
940 cscov = 0;
941 udp6->uh_ulen = htons(cscov);
942 /*
943 * For UDP-Lite, checksum coverage length of zero means
944 * the entire UDPLite packet is covered by the checksum.
945 */
946 cscov_partial = (cscov == 0) ? 0 : 1;
947 } else if (plen <= 0xffff)
948 udp6->uh_ulen = htons((u_short)plen);
949 else
950 udp6->uh_ulen = 0;
951 udp6->uh_sum = 0;
952
953 ip6 = mtod(m, struct ip6_hdr *);
954 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
955 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
956 ip6->ip6_vfc |= IPV6_VERSION;
957 ip6->ip6_plen = htons((u_short)plen);
958 ip6->ip6_nxt = nxt;
959 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
960 ip6->ip6_src = *laddr;
961 ip6->ip6_dst = *faddr;
962
963 #ifdef MAC
964 mac_inpcb_create_mbuf(inp, m);
965 #endif
966
967 if (cscov_partial) {
968 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
969 sizeof(struct ip6_hdr), plen, cscov)) == 0)
970 udp6->uh_sum = 0xffff;
971 } else {
972 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
973 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
974 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
975 }
976
977 flags = 0;
978 #ifdef RSS
979 {
980 uint32_t hash_val, hash_type;
981 uint8_t pr;
982
983 pr = inp->inp_socket->so_proto->pr_protocol;
984 /*
985 * Calculate an appropriate RSS hash for UDP and
986 * UDP Lite.
987 *
988 * The called function will take care of figuring out
989 * whether a 2-tuple or 4-tuple hash is required based
990 * on the currently configured scheme.
991 *
992 * Later later on connected socket values should be
993 * cached in the inpcb and reused, rather than constantly
994 * re-calculating it.
995 *
996 * UDP Lite is a different protocol number and will
997 * likely end up being hashed as a 2-tuple until
998 * RSS / NICs grow UDP Lite protocol awareness.
999 */
1000 if (rss_proto_software_hash_v6(faddr, laddr, fport,
1001 inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
1002 m->m_pkthdr.flowid = hash_val;
1003 M_HASHTYPE_SET(m, hash_type);
1004 }
1005
1006 /*
1007 * Don't override with the inp cached flowid.
1008 *
1009 * Until the whole UDP path is vetted, it may actually
1010 * be incorrect.
1011 */
1012 flags |= IP_NODEFAULTFLOWID;
1013 }
1014 #endif
1015
1016 UDPSTAT_INC(udps_opackets);
1017 if (unlock_udbinfo == UH_WLOCKED)
1018 INP_HASH_WUNLOCK(pcbinfo);
1019 else if (unlock_udbinfo == UH_RLOCKED)
1020 INP_HASH_RUNLOCK_ET(pcbinfo, et);
1021 if (nxt == IPPROTO_UDPLITE)
1022 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
1023 else
1024 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
1025 error = ip6_output(m, optp,
1026 (unlock_inp == UH_WLOCKED) ? &inp->inp_route6 : NULL, flags,
1027 inp->in6p_moptions, NULL, inp);
1028 if (unlock_inp == UH_WLOCKED)
1029 INP_WUNLOCK(inp);
1030 else
1031 INP_RUNLOCK(inp);
1032
1033 if (control) {
1034 ip6_clearpktopts(&opt, -1);
1035 m_freem(control);
1036 }
1037 return (error);
1038
1039 release:
1040 if (unlock_udbinfo == UH_WLOCKED) {
1041 KASSERT(unlock_inp == UH_WLOCKED, ("%s: excl udbinfo lock, "
1042 "non-excl inp lock: pcbinfo %p %#x inp %p %#x",
1043 __func__, pcbinfo, unlock_udbinfo, inp, unlock_inp));
1044 INP_HASH_WUNLOCK(pcbinfo);
1045 INP_WUNLOCK(inp);
1046 } else if (unlock_udbinfo == UH_RLOCKED) {
1047 KASSERT(unlock_inp == UH_RLOCKED, ("%s: non-excl udbinfo lock, "
1048 "excl inp lock: pcbinfo %p %#x inp %p %#x",
1049 __func__, pcbinfo, unlock_udbinfo, inp, unlock_inp));
1050 INP_HASH_RUNLOCK_ET(pcbinfo, et);
1051 INP_RUNLOCK(inp);
1052 } else if (unlock_inp == UH_WLOCKED)
1053 INP_WUNLOCK(inp);
1054 else
1055 INP_RUNLOCK(inp);
1056 if (control) {
1057 ip6_clearpktopts(&opt, -1);
1058 m_freem(control);
1059 }
1060 m_freem(m);
1061
1062 return (error);
1063 }
1064
1065 static void
udp6_abort(struct socket * so)1066 udp6_abort(struct socket *so)
1067 {
1068 struct inpcb *inp;
1069 struct inpcbinfo *pcbinfo;
1070
1071 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1072 inp = sotoinpcb(so);
1073 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1074
1075 INP_WLOCK(inp);
1076 #ifdef INET
1077 if (inp->inp_vflag & INP_IPV4) {
1078 struct pr_usrreqs *pru;
1079 uint8_t nxt;
1080
1081 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1082 IPPROTO_UDP : IPPROTO_UDPLITE;
1083 INP_WUNLOCK(inp);
1084 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1085 (*pru->pru_abort)(so);
1086 return;
1087 }
1088 #endif
1089
1090 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1091 INP_HASH_WLOCK(pcbinfo);
1092 in6_pcbdisconnect(inp);
1093 inp->in6p_laddr = in6addr_any;
1094 INP_HASH_WUNLOCK(pcbinfo);
1095 soisdisconnected(so);
1096 }
1097 INP_WUNLOCK(inp);
1098 }
1099
1100 static int
udp6_attach(struct socket * so,int proto,struct thread * td)1101 udp6_attach(struct socket *so, int proto, struct thread *td)
1102 {
1103 struct inpcb *inp;
1104 struct inpcbinfo *pcbinfo;
1105 int error;
1106
1107 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1108 inp = sotoinpcb(so);
1109 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1110
1111 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1112 error = soreserve(so, udp_sendspace, udp_recvspace);
1113 if (error)
1114 return (error);
1115 }
1116 INP_INFO_WLOCK(pcbinfo);
1117 error = in_pcballoc(so, pcbinfo);
1118 if (error) {
1119 INP_INFO_WUNLOCK(pcbinfo);
1120 return (error);
1121 }
1122 inp = (struct inpcb *)so->so_pcb;
1123 inp->inp_vflag |= INP_IPV6;
1124 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1125 inp->inp_vflag |= INP_IPV4;
1126 inp->in6p_hops = -1; /* use kernel default */
1127 inp->in6p_cksum = -1; /* just to be sure */
1128 /*
1129 * XXX: ugly!!
1130 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1131 * because the socket may be bound to an IPv6 wildcard address,
1132 * which may match an IPv4-mapped IPv6 address.
1133 */
1134 inp->inp_ip_ttl = V_ip_defttl;
1135
1136 error = udp_newudpcb(inp);
1137 if (error) {
1138 in_pcbdetach(inp);
1139 in_pcbfree(inp);
1140 INP_INFO_WUNLOCK(pcbinfo);
1141 return (error);
1142 }
1143 INP_WUNLOCK(inp);
1144 INP_INFO_WUNLOCK(pcbinfo);
1145 return (0);
1146 }
1147
1148 static int
udp6_bind(struct socket * so,struct sockaddr * nam,struct thread * td)1149 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1150 {
1151 struct inpcb *inp;
1152 struct inpcbinfo *pcbinfo;
1153 int error;
1154 u_char vflagsav;
1155
1156 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1157 inp = sotoinpcb(so);
1158 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1159
1160 INP_WLOCK(inp);
1161 INP_HASH_WLOCK(pcbinfo);
1162 vflagsav = inp->inp_vflag;
1163 inp->inp_vflag &= ~INP_IPV4;
1164 inp->inp_vflag |= INP_IPV6;
1165 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1166 struct sockaddr_in6 *sin6_p;
1167
1168 sin6_p = (struct sockaddr_in6 *)nam;
1169
1170 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1171 inp->inp_vflag |= INP_IPV4;
1172 #ifdef INET
1173 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1174 struct sockaddr_in sin;
1175
1176 in6_sin6_2_sin(&sin, sin6_p);
1177 inp->inp_vflag |= INP_IPV4;
1178 inp->inp_vflag &= ~INP_IPV6;
1179 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1180 td->td_ucred);
1181 goto out;
1182 }
1183 #endif
1184 }
1185
1186 error = in6_pcbbind(inp, nam, td->td_ucred);
1187 #ifdef INET
1188 out:
1189 #endif
1190 if (error != 0)
1191 inp->inp_vflag = vflagsav;
1192 INP_HASH_WUNLOCK(pcbinfo);
1193 INP_WUNLOCK(inp);
1194 return (error);
1195 }
1196
1197 static void
udp6_close(struct socket * so)1198 udp6_close(struct socket *so)
1199 {
1200 struct inpcb *inp;
1201 struct inpcbinfo *pcbinfo;
1202
1203 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1204 inp = sotoinpcb(so);
1205 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1206
1207 INP_WLOCK(inp);
1208 #ifdef INET
1209 if (inp->inp_vflag & INP_IPV4) {
1210 struct pr_usrreqs *pru;
1211 uint8_t nxt;
1212
1213 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1214 IPPROTO_UDP : IPPROTO_UDPLITE;
1215 INP_WUNLOCK(inp);
1216 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1217 (*pru->pru_disconnect)(so);
1218 return;
1219 }
1220 #endif
1221 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1222 INP_HASH_WLOCK(pcbinfo);
1223 in6_pcbdisconnect(inp);
1224 inp->in6p_laddr = in6addr_any;
1225 INP_HASH_WUNLOCK(pcbinfo);
1226 soisdisconnected(so);
1227 }
1228 INP_WUNLOCK(inp);
1229 }
1230
1231 static int
udp6_connect(struct socket * so,struct sockaddr * nam,struct thread * td)1232 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1233 {
1234 struct inpcb *inp;
1235 struct inpcbinfo *pcbinfo;
1236 struct sockaddr_in6 *sin6;
1237 int error;
1238 u_char vflagsav;
1239
1240 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1241 inp = sotoinpcb(so);
1242 sin6 = (struct sockaddr_in6 *)nam;
1243 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1244
1245 /*
1246 * XXXRW: Need to clarify locking of v4/v6 flags.
1247 */
1248 INP_WLOCK(inp);
1249 #ifdef INET
1250 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1251 struct sockaddr_in sin;
1252
1253 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1254 error = EINVAL;
1255 goto out;
1256 }
1257 if ((inp->inp_vflag & INP_IPV4) == 0) {
1258 error = EAFNOSUPPORT;
1259 goto out;
1260 }
1261 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1262 error = EISCONN;
1263 goto out;
1264 }
1265 in6_sin6_2_sin(&sin, sin6);
1266 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1267 if (error != 0)
1268 goto out;
1269 vflagsav = inp->inp_vflag;
1270 inp->inp_vflag |= INP_IPV4;
1271 inp->inp_vflag &= ~INP_IPV6;
1272 INP_HASH_WLOCK(pcbinfo);
1273 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1274 td->td_ucred);
1275 INP_HASH_WUNLOCK(pcbinfo);
1276 /*
1277 * If connect succeeds, mark socket as connected. If
1278 * connect fails and socket is unbound, reset inp_vflag
1279 * field.
1280 */
1281 if (error == 0)
1282 soisconnected(so);
1283 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1284 inp->inp_lport == 0)
1285 inp->inp_vflag = vflagsav;
1286 goto out;
1287 } else {
1288 if ((inp->inp_vflag & INP_IPV6) == 0) {
1289 error = EAFNOSUPPORT;
1290 goto out;
1291 }
1292 }
1293 #endif
1294 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1295 error = EISCONN;
1296 goto out;
1297 }
1298 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1299 if (error != 0)
1300 goto out;
1301 vflagsav = inp->inp_vflag;
1302 inp->inp_vflag &= ~INP_IPV4;
1303 inp->inp_vflag |= INP_IPV6;
1304 INP_HASH_WLOCK(pcbinfo);
1305 error = in6_pcbconnect(inp, nam, td->td_ucred);
1306 INP_HASH_WUNLOCK(pcbinfo);
1307 /*
1308 * If connect succeeds, mark socket as connected. If
1309 * connect fails and socket is unbound, reset inp_vflag
1310 * field.
1311 */
1312 if (error == 0)
1313 soisconnected(so);
1314 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1315 inp->inp_lport == 0)
1316 inp->inp_vflag = vflagsav;
1317 out:
1318 INP_WUNLOCK(inp);
1319 return (error);
1320 }
1321
1322 static void
udp6_detach(struct socket * so)1323 udp6_detach(struct socket *so)
1324 {
1325 struct inpcb *inp;
1326 struct inpcbinfo *pcbinfo;
1327 struct udpcb *up;
1328
1329 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1330 inp = sotoinpcb(so);
1331 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1332
1333 INP_INFO_WLOCK(pcbinfo);
1334 INP_WLOCK(inp);
1335 up = intoudpcb(inp);
1336 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1337 in_pcbdetach(inp);
1338 in_pcbfree(inp);
1339 INP_INFO_WUNLOCK(pcbinfo);
1340 udp_discardcb(up);
1341 }
1342
1343 static int
udp6_disconnect(struct socket * so)1344 udp6_disconnect(struct socket *so)
1345 {
1346 struct inpcb *inp;
1347 struct inpcbinfo *pcbinfo;
1348
1349 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1350 inp = sotoinpcb(so);
1351 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1352
1353 INP_WLOCK(inp);
1354 #ifdef INET
1355 if (inp->inp_vflag & INP_IPV4) {
1356 struct pr_usrreqs *pru;
1357 uint8_t nxt;
1358
1359 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1360 IPPROTO_UDP : IPPROTO_UDPLITE;
1361 INP_WUNLOCK(inp);
1362 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1363 (void)(*pru->pru_disconnect)(so);
1364 return (0);
1365 }
1366 #endif
1367
1368 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1369 INP_WUNLOCK(inp);
1370 return (ENOTCONN);
1371 }
1372
1373 INP_HASH_WLOCK(pcbinfo);
1374 in6_pcbdisconnect(inp);
1375 inp->in6p_laddr = in6addr_any;
1376 INP_HASH_WUNLOCK(pcbinfo);
1377 SOCK_LOCK(so);
1378 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1379 SOCK_UNLOCK(so);
1380 INP_WUNLOCK(inp);
1381 return (0);
1382 }
1383
1384 static int
udp6_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * addr,struct mbuf * control,struct thread * td)1385 udp6_send(struct socket *so, int flags, struct mbuf *m,
1386 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1387 {
1388 int error;
1389
1390 if (addr) {
1391 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1392 error = EINVAL;
1393 goto bad;
1394 }
1395 if (addr->sa_family != AF_INET6) {
1396 error = EAFNOSUPPORT;
1397 goto bad;
1398 }
1399 }
1400
1401 return (udp6_output(so, flags, m, addr, control, td));
1402
1403 bad:
1404 if (control)
1405 m_freem(control);
1406 m_freem(m);
1407 return (error);
1408 }
1409
1410 struct pr_usrreqs udp6_usrreqs = {
1411 .pru_abort = udp6_abort,
1412 .pru_attach = udp6_attach,
1413 .pru_bind = udp6_bind,
1414 .pru_connect = udp6_connect,
1415 .pru_control = in6_control,
1416 .pru_detach = udp6_detach,
1417 .pru_disconnect = udp6_disconnect,
1418 .pru_peeraddr = in6_mapped_peeraddr,
1419 .pru_send = udp6_send,
1420 .pru_shutdown = udp_shutdown,
1421 .pru_sockaddr = in6_mapped_sockaddr,
1422 .pru_soreceive = soreceive_dgram,
1423 .pru_sosend = sosend_dgram,
1424 .pru_sosetlabel = in_pcbsosetlabel,
1425 .pru_close = udp6_close
1426 };
1427