udp6_usrreq.c revision 1.155 1 /* $NetBSD: udp6_usrreq.c,v 1.155 2024/07/05 04:31:54 rin Exp $ */
2 /* $KAME: udp6_usrreq.c,v 1.86 2001/05/27 17:33:00 itojun Exp $ */
3 /* $KAME: udp6_output.c,v 1.43 2001/10/15 09:19:52 itojun Exp $ */
4
5 /*
6 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7 * All rights reserved.
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
34 /*
35 * Copyright (c) 1982, 1986, 1989, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
49 *
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
61 *
62 * @(#)udp_var.h 8.1 (Berkeley) 6/10/93
63 */
64
65 #include <sys/cdefs.h>
66 __KERNEL_RCSID(0, "$NetBSD: udp6_usrreq.c,v 1.155 2024/07/05 04:31:54 rin Exp $");
67
68 #ifdef _KERNEL_OPT
69 #include "opt_inet.h"
70 #include "opt_inet_csum.h"
71 #include "opt_ipsec.h"
72 #include "opt_net_mpsafe.h"
73 #endif
74
75 #include <sys/param.h>
76 #include <sys/mbuf.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/syslog.h>
83 #include <sys/domain.h>
84 #include <sys/sysctl.h>
85
86 #include <net/if.h>
87 #include <net/if_types.h>
88
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/in_systm.h>
92 #include <netinet/in_offload.h>
93 #include <netinet/ip.h>
94 #include <netinet/ip_var.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/udp.h>
97 #include <netinet/udp_var.h>
98 #include <netinet/udp_private.h>
99
100 #include <netinet/ip6.h>
101 #include <netinet/icmp6.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet6/ip6_private.h>
104 #include <netinet6/in6_pcb.h>
105 #include <netinet6/udp6_var.h>
106 #include <netinet6/udp6_private.h>
107 #include <netinet6/ip6protosw.h>
108 #include <netinet6/scope6_var.h>
109
110 #ifdef IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/esp.h>
113 #ifdef INET6
114 #include <netipsec/ipsec6.h>
115 #endif
116 #endif
117
118 #include "faith.h"
119 #if defined(NFAITH) && NFAITH > 0
120 #include <net/if_faith.h>
121 #endif
122
123 /*
124 * UDP protocol implementation.
125 * Per RFC 768, August, 1980.
126 */
127
128 extern struct inpcbtable udbtable;
129
130 percpu_t *udp6stat_percpu;
131
132 /* UDP on IP6 parameters */
133 static int udp6_sendspace = 9216; /* really max datagram size */
134 static int udp6_recvspace = 40 * (1024 + sizeof(struct sockaddr_in6));
135 /* 40 1K datagrams */
136
137 static void udp6_notify(struct inpcb *, int);
138 static void sysctl_net_inet6_udp6_setup(struct sysctllog **);
139 #ifdef IPSEC
140 static int udp6_espinudp(struct mbuf **, int);
141 #endif
142
143 #ifdef UDP_CSUM_COUNTERS
144 #include <sys/device.h>
145 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
146 NULL, "udp6", "hwcsum bad");
147 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
148 NULL, "udp6", "hwcsum ok");
149 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
150 NULL, "udp6", "hwcsum data");
151 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
152 NULL, "udp6", "swcsum");
153
154 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
155 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
156 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
157 EVCNT_ATTACH_STATIC(udp6_swcsum);
158
159 #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
160 #else
161 #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */
162 #endif
163
164 void
165 udp6_init(void)
166 {
167 sysctl_net_inet6_udp6_setup(NULL);
168 udp6stat_percpu = percpu_alloc(sizeof(uint64_t) * UDP6_NSTATS);
169
170 udp_init_common();
171 }
172
173 /*
174 * Notify a udp user of an asynchronous error;
175 * just wake up so that he can collect error status.
176 */
177 static void
178 udp6_notify(struct inpcb *inp, int errno)
179 {
180 inp->inp_socket->so_error = errno;
181 sorwakeup(inp->inp_socket);
182 sowwakeup(inp->inp_socket);
183 }
184
185 void *
186 udp6_ctlinput(int cmd, const struct sockaddr *sa, void *d)
187 {
188 struct udphdr uh;
189 struct ip6_hdr *ip6;
190 const struct sockaddr_in6 *sa6 = (const struct sockaddr_in6 *)sa;
191 struct mbuf *m;
192 int off;
193 void *cmdarg;
194 struct ip6ctlparam *ip6cp = NULL;
195 const struct sockaddr_in6 *sa6_src = NULL;
196 void (*notify)(struct inpcb *, int) = udp6_notify;
197 struct udp_portonly {
198 u_int16_t uh_sport;
199 u_int16_t uh_dport;
200 } *uhp;
201
202 if (sa->sa_family != AF_INET6 ||
203 sa->sa_len != sizeof(struct sockaddr_in6))
204 return NULL;
205
206 if ((unsigned)cmd >= PRC_NCMDS)
207 return NULL;
208 if (PRC_IS_REDIRECT(cmd))
209 notify = in6pcb_rtchange, d = NULL;
210 else if (cmd == PRC_HOSTDEAD)
211 d = NULL;
212 else if (cmd == PRC_MSGSIZE) {
213 /* special code is present, see below */
214 notify = in6pcb_rtchange;
215 }
216 else if (inet6ctlerrmap[cmd] == 0)
217 return NULL;
218
219 /* if the parameter is from icmp6, decode it. */
220 if (d != NULL) {
221 ip6cp = (struct ip6ctlparam *)d;
222 m = ip6cp->ip6c_m;
223 ip6 = ip6cp->ip6c_ip6;
224 off = ip6cp->ip6c_off;
225 cmdarg = ip6cp->ip6c_cmdarg;
226 sa6_src = ip6cp->ip6c_src;
227 } else {
228 m = NULL;
229 ip6 = NULL;
230 cmdarg = NULL;
231 sa6_src = &sa6_any;
232 off = 0;
233 }
234
235 if (ip6) {
236 /* check if we can safely examine src and dst ports */
237 if (m->m_pkthdr.len < off + sizeof(*uhp)) {
238 if (cmd == PRC_MSGSIZE)
239 icmp6_mtudisc_update((struct ip6ctlparam *)d, 0);
240 return NULL;
241 }
242
243 memset(&uh, 0, sizeof(uh));
244 m_copydata(m, off, sizeof(*uhp), (void *)&uh);
245
246 if (cmd == PRC_MSGSIZE) {
247 int valid = 0;
248
249 /*
250 * Check to see if we have a valid UDP socket
251 * corresponding to the address in the ICMPv6 message
252 * payload.
253 */
254 if (in6pcb_lookup(&udbtable, &sa6->sin6_addr,
255 uh.uh_dport, (const struct in6_addr *)&sa6_src->sin6_addr,
256 uh.uh_sport, 0, 0))
257 valid++;
258 #if 0
259 /*
260 * As the use of sendto(2) is fairly popular,
261 * we may want to allow non-connected pcb too.
262 * But it could be too weak against attacks...
263 * We should at least check if the local address (= s)
264 * is really ours.
265 */
266 else if (in6pcb_lookup_bound(&udbtable, &sa6->sin6_addr,
267 uh.uh_dport, 0))
268 valid++;
269 #endif
270
271 /*
272 * Depending on the value of "valid" and routing table
273 * size (mtudisc_{hi,lo}wat), we will:
274 * - recalculate the new MTU and create the
275 * corresponding routing entry, or
276 * - ignore the MTU change notification.
277 */
278 icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
279
280 /*
281 * regardless of if we called
282 * icmp6_mtudisc_update(), we need to call
283 * in6pcb_notify(), to notify path MTU change
284 * to the userland (RFC3542), because some
285 * unconnected sockets may share the same
286 * destination and want to know the path MTU.
287 */
288 }
289
290 (void)in6pcb_notify(&udbtable, sa, uh.uh_dport,
291 sin6tocsa(sa6_src), uh.uh_sport, cmd, cmdarg,
292 notify);
293 } else {
294 (void)in6pcb_notify(&udbtable, sa, 0,
295 sin6tocsa(sa6_src), 0, cmd, cmdarg, notify);
296 }
297 return NULL;
298 }
299
300 int
301 udp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
302 {
303 int s;
304 int error = 0;
305 struct inpcb *inp;
306 int family;
307 int optval;
308
309 family = so->so_proto->pr_domain->dom_family;
310
311 s = splsoftnet();
312 switch (family) {
313 #ifdef INET
314 case PF_INET:
315 if (sopt->sopt_level != IPPROTO_UDP) {
316 error = ip_ctloutput(op, so, sopt);
317 goto end;
318 }
319 break;
320 #endif
321 #ifdef INET6
322 case PF_INET6:
323 if (sopt->sopt_level != IPPROTO_UDP) {
324 error = ip6_ctloutput(op, so, sopt);
325 goto end;
326 }
327 break;
328 #endif
329 default:
330 error = EAFNOSUPPORT;
331 goto end;
332 }
333
334 switch (op) {
335 case PRCO_SETOPT:
336 inp = sotoinpcb(so);
337
338 switch (sopt->sopt_name) {
339 case UDP_ENCAP:
340 error = sockopt_getint(sopt, &optval);
341 if (error)
342 break;
343
344 switch(optval) {
345 case 0:
346 inp->inp_flags &= ~IN6P_ESPINUDP;
347 break;
348
349 case UDP_ENCAP_ESPINUDP:
350 inp->inp_flags |= IN6P_ESPINUDP;
351 break;
352
353 default:
354 error = EINVAL;
355 break;
356 }
357 break;
358
359 default:
360 error = ENOPROTOOPT;
361 break;
362 }
363 break;
364
365 default:
366 error = EINVAL;
367 break;
368 }
369
370 end:
371 splx(s);
372 return error;
373 }
374
375 static void
376 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
377 struct sockaddr *src, struct socket *so)
378 {
379 struct mbuf *opts = NULL;
380 struct mbuf *n;
381 struct inpcb *inp;
382
383 KASSERT(so != NULL);
384 KASSERT(so->so_proto->pr_domain->dom_family == AF_INET6);
385 inp = sotoinpcb(so);
386 KASSERT(inp != NULL);
387
388 #if defined(IPSEC)
389 if (ipsec_used && ipsec_in_reject(m, inp)) {
390 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
391 icmp6_error(n, ICMP6_DST_UNREACH,
392 ICMP6_DST_UNREACH_ADMIN, 0);
393 return;
394 }
395 #endif
396
397 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
398 if (inp->inp_flags & IN6P_CONTROLOPTS ||
399 SOOPT_TIMESTAMP(inp->inp_socket->so_options)) {
400 struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
401 ip6_savecontrol(inp, &opts, ip6, n);
402 }
403
404 m_adj(n, off);
405 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
406 m_freem(n);
407 m_freem(opts);
408 UDP6_STATINC(UDP6_STAT_FULLSOCK);
409 soroverflow(so);
410 } else
411 sorwakeup(so);
412 }
413 }
414
415 int
416 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
417 struct mbuf **mp, int off)
418 {
419 u_int16_t sport, dport;
420 int rcvcnt;
421 struct in6_addr src6, *dst6;
422 const struct in_addr *dst4;
423 struct inpcb *inp;
424 struct mbuf *m = *mp;
425
426 rcvcnt = 0;
427 off += sizeof(struct udphdr); /* now, offset of payload */
428
429 if (af != AF_INET && af != AF_INET6)
430 goto bad;
431 if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
432 goto bad;
433
434 src6 = src->sin6_addr;
435 if (sa6_recoverscope(src) != 0) {
436 /* XXX: should be impossible. */
437 goto bad;
438 }
439 sport = src->sin6_port;
440
441 dport = dst->sin6_port;
442 dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
443 dst6 = &dst->sin6_addr;
444
445 if (IN6_IS_ADDR_MULTICAST(dst6) ||
446 (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
447 /*
448 * Deliver a multicast or broadcast datagram to *all* sockets
449 * for which the local and remote addresses and ports match
450 * those of the incoming datagram. This allows more than
451 * one process to receive multi/broadcasts on the same port.
452 * (This really ought to be done for unicast datagrams as
453 * well, but that would cause problems with existing
454 * applications that open both address-specific sockets and
455 * a wildcard socket listening to the same port -- they would
456 * end up receiving duplicates of every unicast datagram.
457 * Those applications open the multiple sockets to overcome an
458 * inadequacy of the UDP socket interface, but for backwards
459 * compatibility we avoid the problem here rather than
460 * fixing the interface. Maybe 4.5BSD will remedy this?)
461 */
462
463 /*
464 * KAME note: traditionally we dropped udpiphdr from mbuf here.
465 * we need udpiphdr for IPsec processing so we do that later.
466 */
467 /*
468 * Locate pcb(s) for datagram.
469 */
470 TAILQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) {
471 if (inp->inp_af != AF_INET6)
472 continue;
473
474 if (inp->inp_lport != dport)
475 continue;
476 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp))) {
477 if (!IN6_ARE_ADDR_EQUAL(&in6p_laddr(inp),
478 dst6))
479 continue;
480 } else {
481 if (IN6_IS_ADDR_V4MAPPED(dst6) &&
482 (inp->inp_flags & IN6P_IPV6_V6ONLY))
483 continue;
484 }
485 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp))) {
486 if (!IN6_ARE_ADDR_EQUAL(&in6p_faddr(inp),
487 &src6) || inp->inp_fport != sport)
488 continue;
489 } else {
490 if (IN6_IS_ADDR_V4MAPPED(&src6) &&
491 (inp->inp_flags & IN6P_IPV6_V6ONLY))
492 continue;
493 }
494
495 udp6_sendup(m, off, sin6tosa(src), inp->inp_socket);
496 rcvcnt++;
497
498 /*
499 * Don't look for additional matches if this one does
500 * not have either the SO_REUSEPORT or SO_REUSEADDR
501 * socket options set. This heuristic avoids searching
502 * through all pcbs in the common case of a non-shared
503 * port. It assumes that an application will never
504 * clear these options after setting them.
505 */
506 if ((inp->inp_socket->so_options &
507 (SO_REUSEPORT|SO_REUSEADDR)) == 0)
508 break;
509 }
510 } else {
511 /*
512 * Locate pcb for datagram.
513 */
514 inp = in6pcb_lookup(&udbtable, &src6, sport, dst6,
515 dport, 0, 0);
516 if (inp == NULL) {
517 UDP_STATINC(UDP_STAT_PCBHASHMISS);
518 inp = in6pcb_lookup_bound(&udbtable, dst6, dport, 0);
519 if (inp == NULL)
520 return rcvcnt;
521 }
522
523 #ifdef IPSEC
524 /* Handle ESP over UDP */
525 if (inp->inp_flags & IN6P_ESPINUDP) {
526 switch (udp6_espinudp(mp, off)) {
527 case -1: /* Error, m was freed */
528 rcvcnt = -1;
529 goto bad;
530
531 case 1: /* ESP over UDP */
532 rcvcnt++;
533 goto bad;
534
535 case 0: /* plain UDP */
536 default: /* Unexpected */
537 /*
538 * Normal UDP processing will take place,
539 * m may have changed.
540 */
541 m = *mp;
542 break;
543 }
544 }
545 #endif
546
547 if (inp->inp_overudp_cb != NULL) {
548 int ret;
549 ret = inp->inp_overudp_cb(mp, off, inp->inp_socket,
550 sin6tosa(src), inp->inp_overudp_arg);
551 switch (ret) {
552 case -1: /* Error, m was freed */
553 rcvcnt = -1;
554 goto bad;
555
556 case 1: /* Foo over UDP */
557 KASSERT(*mp == NULL);
558 rcvcnt++;
559 goto bad;
560
561 case 0: /* plain UDP */
562 default: /* Unexpected */
563 /*
564 * Normal UDP processing will take place,
565 * m may have changed.
566 */
567 break;
568 }
569 }
570
571 udp6_sendup(m, off, sin6tosa(src), inp->inp_socket);
572 rcvcnt++;
573 }
574
575 bad:
576 return rcvcnt;
577 }
578
579 int
580 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
581 {
582
583 /*
584 * XXX it's better to record and check if this mbuf is
585 * already checked.
586 */
587
588 if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
589 goto good;
590 }
591 if (uh->uh_sum == 0) {
592 UDP6_STATINC(UDP6_STAT_NOSUM);
593 goto bad;
594 }
595
596 switch (m->m_pkthdr.csum_flags &
597 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv6) |
598 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
599 case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
600 UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
601 UDP6_STATINC(UDP6_STAT_BADSUM);
602 goto bad;
603
604 #if 0 /* notyet */
605 case M_CSUM_UDPv6|M_CSUM_DATA:
606 #endif
607
608 case M_CSUM_UDPv6:
609 /* Checksum was okay. */
610 UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
611 break;
612
613 default:
614 /*
615 * Need to compute it ourselves. Maybe skip checksum
616 * on loopback interfaces.
617 */
618 UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
619 if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
620 UDP6_STATINC(UDP6_STAT_BADSUM);
621 goto bad;
622 }
623 }
624
625 good:
626 return 0;
627 bad:
628 return -1;
629 }
630
631 int
632 udp6_input(struct mbuf **mp, int *offp, int proto)
633 {
634 struct mbuf *m = *mp;
635 int off = *offp;
636 struct sockaddr_in6 src, dst;
637 struct ip6_hdr *ip6;
638 struct udphdr *uh;
639 u_int32_t plen, ulen;
640
641 ip6 = mtod(m, struct ip6_hdr *);
642
643 #if defined(NFAITH) && 0 < NFAITH
644 if (faithprefix(&ip6->ip6_dst)) {
645 /* send icmp6 host unreach? */
646 m_freem(m);
647 return IPPROTO_DONE;
648 }
649 #endif
650
651 UDP6_STATINC(UDP6_STAT_IPACKETS);
652
653 /* Check for jumbogram is done in ip6_input. We can trust pkthdr.len. */
654 plen = m->m_pkthdr.len - off;
655 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
656 if (uh == NULL) {
657 IP6_STATINC(IP6_STAT_TOOSHORT);
658 return IPPROTO_DONE;
659 }
660
661 /*
662 * Enforce alignment requirements that are violated in
663 * some cases, see kern/50766 for details.
664 */
665 if (ACCESSIBLE_POINTER(uh, struct udphdr) == 0) {
666 m = m_copyup(m, off + sizeof(struct udphdr), 0);
667 if (m == NULL) {
668 IP6_STATINC(IP6_STAT_TOOSHORT);
669 return IPPROTO_DONE;
670 }
671 ip6 = mtod(m, struct ip6_hdr *);
672 uh = (struct udphdr *)(mtod(m, char *) + off);
673 }
674 KASSERT(ACCESSIBLE_POINTER(uh, struct udphdr));
675 ulen = ntohs((u_short)uh->uh_ulen);
676
677 /*
678 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
679 * iff payload length > 0xffff.
680 */
681 if (ulen == 0 && plen > 0xffff)
682 ulen = plen;
683
684 if (plen != ulen) {
685 UDP6_STATINC(UDP6_STAT_BADLEN);
686 goto bad;
687 }
688
689 /* destination port of 0 is illegal, based on RFC768. */
690 if (uh->uh_dport == 0)
691 goto bad;
692
693 /*
694 * Checksum extended UDP header and data. Maybe skip checksum
695 * on loopback interfaces.
696 */
697 if (udp6_input_checksum(m, uh, off, ulen))
698 goto bad;
699
700 /*
701 * Construct source and dst sockaddrs.
702 */
703 memset(&src, 0, sizeof(src));
704 src.sin6_family = AF_INET6;
705 src.sin6_len = sizeof(struct sockaddr_in6);
706 src.sin6_addr = ip6->ip6_src;
707 src.sin6_port = uh->uh_sport;
708 memset(&dst, 0, sizeof(dst));
709 dst.sin6_family = AF_INET6;
710 dst.sin6_len = sizeof(struct sockaddr_in6);
711 dst.sin6_addr = ip6->ip6_dst;
712 dst.sin6_port = uh->uh_dport;
713
714 if (udp6_realinput(AF_INET6, &src, &dst, &m, off) == 0) {
715 if (m->m_flags & M_MCAST) {
716 UDP6_STATINC(UDP6_STAT_NOPORTMCAST);
717 goto bad;
718 }
719 UDP6_STATINC(UDP6_STAT_NOPORT);
720 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
721 m = NULL;
722 }
723
724 bad:
725 m_freem(m);
726 return IPPROTO_DONE;
727 }
728
729 int
730 udp6_output(struct inpcb * const inp, struct mbuf *m,
731 struct sockaddr_in6 * const addr6, struct mbuf * const control,
732 struct lwp * const l)
733 {
734 u_int32_t ulen = m->m_pkthdr.len;
735 u_int32_t plen = sizeof(struct udphdr) + ulen;
736 struct ip6_hdr *ip6;
737 struct udphdr *udp6;
738 struct in6_addr _laddr, *laddr, *faddr;
739 struct in6_addr laddr_mapped; /* XXX ugly */
740 struct sockaddr_in6 *sin6 = NULL;
741 struct ifnet *oifp = NULL;
742 int scope_ambiguous = 0;
743 u_int16_t fport;
744 int error = 0;
745 struct ip6_pktopts *optp = NULL;
746 struct ip6_pktopts opt;
747 int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
748 #ifdef INET
749 struct ip *ip;
750 struct udpiphdr *ui;
751 int flags = 0;
752 #endif
753 struct sockaddr_in6 tmp;
754
755 if (addr6) {
756 sin6 = addr6;
757 if (sin6->sin6_len != sizeof(*sin6)) {
758 error = EINVAL;
759 goto release;
760 }
761 if (sin6->sin6_family != AF_INET6) {
762 error = EAFNOSUPPORT;
763 goto release;
764 }
765
766 /* protect *sin6 from overwrites */
767 tmp = *sin6;
768 sin6 = &tmp;
769
770 /*
771 * Application should provide a proper zone ID or the use of
772 * default zone IDs should be enabled. Unfortunately, some
773 * applications do not behave as it should, so we need a
774 * workaround. Even if an appropriate ID is not determined,
775 * we'll see if we can determine the outgoing interface. If we
776 * can, determine the zone ID based on the interface below.
777 */
778 if (sin6->sin6_scope_id == 0 && !ip6_use_defzone)
779 scope_ambiguous = 1;
780 if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
781 goto release;
782 }
783
784 if (control) {
785 if (__predict_false(l == NULL)) {
786 panic("%s: control but no lwp", __func__);
787 }
788 if ((error = ip6_setpktopts(control, &opt,
789 in6p_outputopts(inp), l->l_cred, IPPROTO_UDP)) != 0)
790 goto release;
791 optp = &opt;
792 } else
793 optp = in6p_outputopts(inp);
794
795
796 if (sin6) {
797 /*
798 * Slightly different than v4 version in that we call
799 * in6_selectsrc and in6pcb_set_port to fill in the local
800 * address and port rather than inpcb_connect. inpcb_connect
801 * sets inp_faddr which causes EISCONN below to be hit on
802 * subsequent sendto.
803 */
804 if (sin6->sin6_port == 0) {
805 error = EADDRNOTAVAIL;
806 goto release;
807 }
808
809 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp))) {
810 /* how about ::ffff:0.0.0.0 case? */
811 error = EISCONN;
812 goto release;
813 }
814
815 faddr = &sin6->sin6_addr;
816 fport = sin6->sin6_port; /* allow 0 port */
817
818 if (IN6_IS_ADDR_V4MAPPED(faddr)) {
819 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
820 /*
821 * I believe we should explicitly discard the
822 * packet when mapped addresses are disabled,
823 * rather than send the packet as an IPv6 one.
824 * If we chose the latter approach, the packet
825 * might be sent out on the wire based on the
826 * default route, the situation which we'd
827 * probably want to avoid.
828 * (20010421 jinmei (at) kame.net)
829 */
830 error = EINVAL;
831 goto release;
832 }
833 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp)) &&
834 !IN6_IS_ADDR_V4MAPPED(&in6p_laddr(inp))) {
835 /*
836 * when remote addr is an IPv4-mapped address,
837 * local addr should not be an IPv6 address,
838 * since you cannot determine how to map IPv6
839 * source address to IPv4.
840 */
841 error = EINVAL;
842 goto release;
843 }
844
845 af = AF_INET;
846 }
847
848 if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
849 struct psref psref;
850 int bound = curlwp_bind();
851
852 error = in6_selectsrc(sin6, optp,
853 in6p_moptions(inp),
854 &inp->inp_route,
855 &in6p_laddr(inp), &oifp, &psref, &_laddr);
856 if (error)
857 laddr = NULL;
858 else
859 laddr = &_laddr;
860 if (oifp && scope_ambiguous &&
861 (error = in6_setscope(&sin6->sin6_addr,
862 oifp, NULL))) {
863 if_put(oifp, &psref);
864 curlwp_bindx(bound);
865 goto release;
866 }
867 if_put(oifp, &psref);
868 curlwp_bindx(bound);
869 } else {
870 /*
871 * XXX: freebsd[34] does not have in_selectsrc, but
872 * we can omit the whole part because freebsd4 calls
873 * udp_output() directly in this case, and thus we'll
874 * never see this path.
875 */
876 if (IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp))) {
877 struct sockaddr_in sin_dst;
878 struct in_addr ina;
879 struct in_ifaddr *ia4;
880 struct psref _psref;
881 int bound;
882
883 memcpy(&ina, &faddr->s6_addr[12], sizeof(ina));
884 sockaddr_in_init(&sin_dst, &ina, 0);
885 bound = curlwp_bind();
886 ia4 = in_selectsrc(&sin_dst, &inp->inp_route,
887 inp->inp_socket->so_options, NULL,
888 &error, &_psref);
889 if (ia4 == NULL) {
890 curlwp_bindx(bound);
891 if (error == 0)
892 error = EADDRNOTAVAIL;
893 goto release;
894 }
895 memset(&laddr_mapped, 0, sizeof(laddr_mapped));
896 laddr_mapped.s6_addr16[5] = 0xffff; /* ugly */
897 memcpy(&laddr_mapped.s6_addr[12],
898 &IA_SIN(ia4)->sin_addr,
899 sizeof(IA_SIN(ia4)->sin_addr));
900 ia4_release(ia4, &_psref);
901 curlwp_bindx(bound);
902 laddr = &laddr_mapped;
903 } else
904 {
905 laddr = &in6p_laddr(inp); /* XXX */
906 }
907 }
908 if (laddr == NULL) {
909 if (error == 0)
910 error = EADDRNOTAVAIL;
911 goto release;
912 }
913 if (inp->inp_lport == 0) {
914 /*
915 * Craft a sockaddr_in6 for the local endpoint. Use the
916 * "any" as a base, set the address, and recover the
917 * scope.
918 */
919 struct sockaddr_in6 lsin6 =
920 *((const struct sockaddr_in6 *)inp->inp_socket->so_proto->pr_domain->dom_sa_any);
921 lsin6.sin6_addr = *laddr;
922 error = sa6_recoverscope(&lsin6);
923 if (error)
924 goto release;
925
926 error = in6pcb_set_port(&lsin6, inp, l);
927
928 if (error) {
929 in6p_laddr(inp) = in6addr_any;
930 goto release;
931 }
932 }
933 } else {
934 if (IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp))) {
935 error = ENOTCONN;
936 goto release;
937 }
938 if (IN6_IS_ADDR_V4MAPPED(&in6p_faddr(inp))) {
939 if ((inp->inp_flags & IN6P_IPV6_V6ONLY))
940 {
941 /*
942 * XXX: this case would happen when the
943 * application sets the V6ONLY flag after
944 * connecting the foreign address.
945 * Such applications should be fixed,
946 * so we bark here.
947 */
948 log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
949 "option was set for a connected socket\n");
950 error = EINVAL;
951 goto release;
952 } else
953 af = AF_INET;
954 }
955 laddr = &in6p_laddr(inp);
956 faddr = &in6p_faddr(inp);
957 fport = inp->inp_fport;
958 }
959
960 if (af == AF_INET)
961 hlen = sizeof(struct ip);
962
963 /*
964 * Calculate data length and get a mbuf
965 * for UDP and IP6 headers.
966 */
967 M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
968 if (m == NULL) {
969 error = ENOBUFS;
970 goto release;
971 }
972
973 /*
974 * Stuff checksum and output datagram.
975 */
976 udp6 = (struct udphdr *)(mtod(m, char *) + hlen);
977 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
978 udp6->uh_dport = fport;
979 if (plen <= 0xffff)
980 udp6->uh_ulen = htons((u_int16_t)plen);
981 else
982 udp6->uh_ulen = 0;
983 udp6->uh_sum = 0;
984
985 switch (af) {
986 case AF_INET6:
987 ip6 = mtod(m, struct ip6_hdr *);
988 ip6->ip6_flow = in6p_flowinfo(inp) & IPV6_FLOWINFO_MASK;
989 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
990 ip6->ip6_vfc |= IPV6_VERSION;
991 #if 0 /* ip6_plen will be filled in ip6_output. */
992 ip6->ip6_plen = htons((u_int16_t)plen);
993 #endif
994 ip6->ip6_nxt = IPPROTO_UDP;
995 ip6->ip6_hlim = in6pcb_selecthlim_rt(inp);
996 ip6->ip6_src = *laddr;
997 ip6->ip6_dst = *faddr;
998
999 udp6->uh_sum = in6_cksum_phdr(laddr, faddr,
1000 htonl(plen), htonl(IPPROTO_UDP));
1001 m->m_pkthdr.csum_flags = M_CSUM_UDPv6;
1002 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1003
1004 UDP6_STATINC(UDP6_STAT_OPACKETS);
1005 error = ip6_output(m, optp, &inp->inp_route, 0,
1006 in6p_moptions(inp), inp, NULL);
1007 break;
1008 case AF_INET:
1009 #ifdef INET
1010 /* can't transmit jumbogram over IPv4 */
1011 if (plen > 0xffff) {
1012 error = EMSGSIZE;
1013 goto release;
1014 }
1015
1016 ip = mtod(m, struct ip *);
1017 ui = (struct udpiphdr *)ip;
1018 memset(ui->ui_x1, 0, sizeof(ui->ui_x1));
1019 ui->ui_pr = IPPROTO_UDP;
1020 ui->ui_len = htons(plen);
1021 memcpy(&ui->ui_src, &laddr->s6_addr[12], sizeof(ui->ui_src));
1022 ui->ui_ulen = ui->ui_len;
1023
1024 flags = (inp->inp_socket->so_options &
1025 (SO_DONTROUTE | SO_BROADCAST));
1026 memcpy(&ui->ui_dst, &faddr->s6_addr[12], sizeof(ui->ui_dst));
1027
1028 udp6->uh_sum = in_cksum(m, hlen + plen);
1029 if (udp6->uh_sum == 0)
1030 udp6->uh_sum = 0xffff;
1031
1032 ip->ip_len = htons(hlen + plen);
1033 ip->ip_ttl = in6pcb_selecthlim(inp, NULL); /* XXX */
1034 ip->ip_tos = 0; /* XXX */
1035
1036 UDP_STATINC(UDP_STAT_OPACKETS);
1037 error = ip_output(m, NULL, &inp->inp_route, flags /* XXX */,
1038 inp->inp_moptions, NULL);
1039 break;
1040 #else
1041 error = EAFNOSUPPORT;
1042 goto release;
1043 #endif
1044 }
1045 goto releaseopt;
1046
1047 release:
1048 m_freem(m);
1049
1050 releaseopt:
1051 if (control) {
1052 if (optp == &opt)
1053 ip6_clearpktopts(&opt, -1);
1054 m_freem(control);
1055 }
1056 return (error);
1057 }
1058
1059 static int
1060 udp6_attach(struct socket *so, int proto)
1061 {
1062 struct inpcb *inp;
1063 int s, error;
1064
1065 KASSERT(sotoinpcb(so) == NULL);
1066 sosetlock(so);
1067
1068 error = soreserve(so, udp6_sendspace, udp6_recvspace);
1069 if (error) {
1070 return error;
1071 }
1072
1073 /*
1074 * MAPPED_ADDR implementation spec:
1075 * Always attach for IPv6, and only when necessary for IPv4.
1076 */
1077 s = splsoftnet();
1078 error = inpcb_create(so, &udbtable);
1079 splx(s);
1080 if (error) {
1081 return error;
1082 }
1083
1084 inp = sotoinpcb(so);
1085 in6p_cksum(inp) = -1; /* just to be sure */
1086
1087 KASSERT(solocked(so));
1088 return 0;
1089 }
1090
1091 static void
1092 udp6_detach(struct socket *so)
1093 {
1094 struct inpcb *inp = sotoinpcb(so);
1095 int s;
1096
1097 KASSERT(solocked(so));
1098 KASSERT(inp != NULL);
1099
1100 s = splsoftnet();
1101 inpcb_destroy(inp);
1102 splx(s);
1103 }
1104
1105 static int
1106 udp6_accept(struct socket *so, struct sockaddr *nam)
1107 {
1108 KASSERT(solocked(so));
1109
1110 return EOPNOTSUPP;
1111 }
1112
1113 static int
1114 udp6_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
1115 {
1116 struct inpcb *inp = sotoinpcb(so);
1117 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1118 int error = 0;
1119 int s;
1120
1121 KASSERT(solocked(so));
1122 KASSERT(inp != NULL);
1123
1124 s = splsoftnet();
1125 error = in6pcb_bind(inp, sin6, l);
1126 splx(s);
1127 return error;
1128 }
1129
1130 static int
1131 udp6_listen(struct socket *so, struct lwp *l)
1132 {
1133 KASSERT(solocked(so));
1134
1135 return EOPNOTSUPP;
1136 }
1137
1138 static int
1139 udp6_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
1140 {
1141 struct inpcb *inp = sotoinpcb(so);
1142 int error = 0;
1143 int s;
1144
1145 KASSERT(solocked(so));
1146 KASSERT(inp != NULL);
1147
1148 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp)))
1149 return EISCONN;
1150 s = splsoftnet();
1151 error = in6pcb_connect(inp, (struct sockaddr_in6 *)nam, l);
1152 splx(s);
1153 if (error == 0)
1154 soisconnected(so);
1155
1156 return error;
1157 }
1158
1159 static int
1160 udp6_connect2(struct socket *so, struct socket *so2)
1161 {
1162 KASSERT(solocked(so));
1163
1164 return EOPNOTSUPP;
1165 }
1166
1167 static int
1168 udp6_disconnect(struct socket *so)
1169 {
1170 struct inpcb *inp = sotoinpcb(so);
1171 int s;
1172
1173 KASSERT(solocked(so));
1174 KASSERT(inp != NULL);
1175
1176 if (IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp)))
1177 return ENOTCONN;
1178
1179 s = splsoftnet();
1180 in6pcb_disconnect(inp);
1181 memset((void *)&in6p_laddr(inp), 0, sizeof(in6p_laddr(inp)));
1182 splx(s);
1183
1184 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1185 in6pcb_set_state(inp, INP_BOUND); /* XXX */
1186 return 0;
1187 }
1188
1189 static int
1190 udp6_shutdown(struct socket *so)
1191 {
1192 int s;
1193
1194 s = splsoftnet();
1195 socantsendmore(so);
1196 splx(s);
1197
1198 return 0;
1199 }
1200
1201 static int
1202 udp6_abort(struct socket *so)
1203 {
1204 int s;
1205
1206 KASSERT(solocked(so));
1207 KASSERT(sotoinpcb(so) != NULL);
1208
1209 s = splsoftnet();
1210 soisdisconnected(so);
1211 inpcb_destroy(sotoinpcb(so));
1212 splx(s);
1213
1214 return 0;
1215 }
1216
1217 static int
1218 udp6_ioctl(struct socket *so, u_long cmd, void *addr6, struct ifnet *ifp)
1219 {
1220 /*
1221 * MAPPED_ADDR implementation info:
1222 * Mapped addr support for PRU_CONTROL is not necessary.
1223 * Because typical user of PRU_CONTROL is such as ifconfig,
1224 * and they don't associate any addr to their socket. Then
1225 * socket family is only hint about the PRU_CONTROL'ed address
1226 * family, especially when getting addrs from kernel.
1227 * So AF_INET socket need to be used to control AF_INET addrs,
1228 * and AF_INET6 socket for AF_INET6 addrs.
1229 */
1230 return in6_control(so, cmd, addr6, ifp);
1231 }
1232
1233 static int
1234 udp6_stat(struct socket *so, struct stat *ub)
1235 {
1236 KASSERT(solocked(so));
1237
1238 /* stat: don't bother with a blocksize */
1239 return 0;
1240 }
1241
1242 static int
1243 udp6_peeraddr(struct socket *so, struct sockaddr *nam)
1244 {
1245 KASSERT(solocked(so));
1246 KASSERT(sotoinpcb(so) != NULL);
1247 KASSERT(nam != NULL);
1248
1249 in6pcb_fetch_peeraddr(sotoinpcb(so), (struct sockaddr_in6 *)nam);
1250 return 0;
1251 }
1252
1253 static int
1254 udp6_sockaddr(struct socket *so, struct sockaddr *nam)
1255 {
1256 KASSERT(solocked(so));
1257 KASSERT(sotoinpcb(so) != NULL);
1258 KASSERT(nam != NULL);
1259
1260 in6pcb_fetch_sockaddr(sotoinpcb(so), (struct sockaddr_in6 *)nam);
1261 return 0;
1262 }
1263
1264 static int
1265 udp6_rcvd(struct socket *so, int flags, struct lwp *l)
1266 {
1267 KASSERT(solocked(so));
1268
1269 return EOPNOTSUPP;
1270 }
1271
1272 static int
1273 udp6_recvoob(struct socket *so, struct mbuf *m, int flags)
1274 {
1275 KASSERT(solocked(so));
1276
1277 return EOPNOTSUPP;
1278 }
1279
1280 static int
1281 udp6_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1282 struct mbuf *control, struct lwp *l)
1283 {
1284 struct inpcb *inp = sotoinpcb(so);
1285 int error = 0;
1286 int s;
1287
1288 KASSERT(solocked(so));
1289 KASSERT(inp != NULL);
1290 KASSERT(m != NULL);
1291
1292 s = splsoftnet();
1293 error = udp6_output(inp, m, (struct sockaddr_in6 *)nam, control, l);
1294 splx(s);
1295
1296 return error;
1297 }
1298
1299 static int
1300 udp6_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1301 {
1302 KASSERT(solocked(so));
1303
1304 m_freem(m);
1305 m_freem(control);
1306
1307 return EOPNOTSUPP;
1308 }
1309
1310 static int
1311 udp6_purgeif(struct socket *so, struct ifnet *ifp)
1312 {
1313
1314 mutex_enter(softnet_lock);
1315 in6pcb_purgeif0(&udbtable, ifp);
1316 #ifdef NET_MPSAFE
1317 mutex_exit(softnet_lock);
1318 #endif
1319 in6_purgeif(ifp);
1320 #ifdef NET_MPSAFE
1321 mutex_enter(softnet_lock);
1322 #endif
1323 in6pcb_purgeif(&udbtable, ifp);
1324 mutex_exit(softnet_lock);
1325
1326 return 0;
1327 }
1328
1329 static int
1330 sysctl_net_inet6_udp6_stats(SYSCTLFN_ARGS)
1331 {
1332
1333 return (NETSTAT_SYSCTL(udp6stat_percpu, UDP6_NSTATS));
1334 }
1335
1336 static void
1337 sysctl_net_inet6_udp6_setup(struct sysctllog **clog)
1338 {
1339
1340 sysctl_createv(clog, 0, NULL, NULL,
1341 CTLFLAG_PERMANENT,
1342 CTLTYPE_NODE, "inet6", NULL,
1343 NULL, 0, NULL, 0,
1344 CTL_NET, PF_INET6, CTL_EOL);
1345 sysctl_createv(clog, 0, NULL, NULL,
1346 CTLFLAG_PERMANENT,
1347 CTLTYPE_NODE, "udp6",
1348 SYSCTL_DESCR("UDPv6 related settings"),
1349 NULL, 0, NULL, 0,
1350 CTL_NET, PF_INET6, IPPROTO_UDP, CTL_EOL);
1351
1352 sysctl_createv(clog, 0, NULL, NULL,
1353 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1354 CTLTYPE_INT, "sendspace",
1355 SYSCTL_DESCR("Default UDP send buffer size"),
1356 NULL, 0, &udp6_sendspace, 0,
1357 CTL_NET, PF_INET6, IPPROTO_UDP, UDP6CTL_SENDSPACE,
1358 CTL_EOL);
1359 sysctl_createv(clog, 0, NULL, NULL,
1360 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1361 CTLTYPE_INT, "recvspace",
1362 SYSCTL_DESCR("Default UDP receive buffer size"),
1363 NULL, 0, &udp6_recvspace, 0,
1364 CTL_NET, PF_INET6, IPPROTO_UDP, UDP6CTL_RECVSPACE,
1365 CTL_EOL);
1366 sysctl_createv(clog, 0, NULL, NULL,
1367 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1368 CTLTYPE_INT, "do_loopback_cksum",
1369 SYSCTL_DESCR("Perform UDP checksum on loopback"),
1370 NULL, 0, &udp_do_loopback_cksum, 0,
1371 CTL_NET, PF_INET6, IPPROTO_UDP, UDP6CTL_LOOPBACKCKSUM,
1372 CTL_EOL);
1373 sysctl_createv(clog, 0, NULL, NULL,
1374 CTLFLAG_PERMANENT,
1375 CTLTYPE_STRUCT, "pcblist",
1376 SYSCTL_DESCR("UDP protocol control block list"),
1377 sysctl_inpcblist, 0, &udbtable, 0,
1378 CTL_NET, PF_INET6, IPPROTO_UDP, CTL_CREATE,
1379 CTL_EOL);
1380 sysctl_createv(clog, 0, NULL, NULL,
1381 CTLFLAG_PERMANENT,
1382 CTLTYPE_STRUCT, "stats",
1383 SYSCTL_DESCR("UDPv6 statistics"),
1384 sysctl_net_inet6_udp6_stats, 0, NULL, 0,
1385 CTL_NET, PF_INET6, IPPROTO_UDP, UDP6CTL_STATS,
1386 CTL_EOL);
1387 }
1388
1389 void
1390 udp6_statinc(u_int stat)
1391 {
1392
1393 KASSERT(stat < UDP6_NSTATS);
1394 UDP6_STATINC(stat);
1395 }
1396
1397 #ifdef IPSEC
1398 /*
1399 * Returns:
1400 * 1 if the packet was processed
1401 * 0 if normal UDP processing should take place
1402 * -1 if an error occurred and m was freed
1403 */
1404 static int
1405 udp6_espinudp(struct mbuf **mp, int off)
1406 {
1407 const size_t skip = sizeof(struct udphdr);
1408 size_t len;
1409 void *data;
1410 size_t minlen;
1411 int ip6hdrlen;
1412 struct ip6_hdr *ip6;
1413 struct m_tag *tag;
1414 struct udphdr *udphdr;
1415 u_int16_t sport, dport;
1416 struct mbuf *m = *mp;
1417 uint32_t *marker;
1418
1419 /*
1420 * Collapse the mbuf chain if the first mbuf is too short
1421 * The longest case is: UDP + non ESP marker + ESP
1422 */
1423 minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1424 if (minlen > m->m_pkthdr.len)
1425 minlen = m->m_pkthdr.len;
1426
1427 if (m->m_len < minlen) {
1428 if ((*mp = m_pullup(m, minlen)) == NULL) {
1429 return -1;
1430 }
1431 m = *mp;
1432 }
1433
1434 len = m->m_len - off;
1435 data = mtod(m, char *) + off;
1436
1437 /* Ignore keepalive packets */
1438 if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1439 m_freem(m);
1440 *mp = NULL; /* avoid any further processing by caller ... */
1441 return 1;
1442 }
1443
1444 /* Handle Non-ESP marker (32bit). If zero, then IKE. */
1445 marker = (uint32_t *)data;
1446 if (len <= sizeof(uint32_t))
1447 return 0;
1448 if (marker[0] == 0)
1449 return 0;
1450
1451 /*
1452 * Get the UDP ports. They are handled in network
1453 * order everywhere in IPSEC_NAT_T code.
1454 */
1455 udphdr = (struct udphdr *)((char *)data - skip);
1456 sport = udphdr->uh_sport;
1457 dport = udphdr->uh_dport;
1458
1459 /*
1460 * Remove the UDP header (and possibly the non ESP marker)
1461 * IPv6 header length is ip6hdrlen
1462 * Before:
1463 * <---- off --->
1464 * +-----+------+-----+
1465 * | IP6 | UDP | ESP |
1466 * +-----+------+-----+
1467 * <-skip->
1468 * After:
1469 * +-----+-----+
1470 * | IP6 | ESP |
1471 * +-----+-----+
1472 * <-skip->
1473 */
1474 ip6hdrlen = off - sizeof(struct udphdr);
1475 memmove(mtod(m, char *) + skip, mtod(m, void *), ip6hdrlen);
1476 m_adj(m, skip);
1477
1478 ip6 = mtod(m, struct ip6_hdr *);
1479 ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) - skip);
1480 ip6->ip6_nxt = IPPROTO_ESP;
1481
1482 /*
1483 * We have modified the packet - it is now ESP, so we should not
1484 * return to UDP processing ...
1485 *
1486 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1487 * the source UDP port. This is required if we want
1488 * to select the right SPD for multiple hosts behind
1489 * same NAT
1490 */
1491 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1492 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1493 m_freem(m);
1494 return -1;
1495 }
1496 ((u_int16_t *)(tag + 1))[0] = sport;
1497 ((u_int16_t *)(tag + 1))[1] = dport;
1498 m_tag_prepend(m, tag);
1499
1500 if (ipsec_used)
1501 ipsec6_common_input(&m, &ip6hdrlen, IPPROTO_ESP);
1502 else
1503 m_freem(m);
1504
1505 /* We handled it, it shouldn't be handled by UDP */
1506 *mp = NULL; /* avoid free by caller ... */
1507 return 1;
1508 }
1509 #endif /* IPSEC */
1510
1511 PR_WRAP_USRREQS(udp6)
1512 #define udp6_attach udp6_attach_wrapper
1513 #define udp6_detach udp6_detach_wrapper
1514 #define udp6_accept udp6_accept_wrapper
1515 #define udp6_bind udp6_bind_wrapper
1516 #define udp6_listen udp6_listen_wrapper
1517 #define udp6_connect udp6_connect_wrapper
1518 #define udp6_connect2 udp6_connect2_wrapper
1519 #define udp6_disconnect udp6_disconnect_wrapper
1520 #define udp6_shutdown udp6_shutdown_wrapper
1521 #define udp6_abort udp6_abort_wrapper
1522 #define udp6_ioctl udp6_ioctl_wrapper
1523 #define udp6_stat udp6_stat_wrapper
1524 #define udp6_peeraddr udp6_peeraddr_wrapper
1525 #define udp6_sockaddr udp6_sockaddr_wrapper
1526 #define udp6_rcvd udp6_rcvd_wrapper
1527 #define udp6_recvoob udp6_recvoob_wrapper
1528 #define udp6_send udp6_send_wrapper
1529 #define udp6_sendoob udp6_sendoob_wrapper
1530 #define udp6_purgeif udp6_purgeif_wrapper
1531
1532 const struct pr_usrreqs udp6_usrreqs = {
1533 .pr_attach = udp6_attach,
1534 .pr_detach = udp6_detach,
1535 .pr_accept = udp6_accept,
1536 .pr_bind = udp6_bind,
1537 .pr_listen = udp6_listen,
1538 .pr_connect = udp6_connect,
1539 .pr_connect2 = udp6_connect2,
1540 .pr_disconnect = udp6_disconnect,
1541 .pr_shutdown = udp6_shutdown,
1542 .pr_abort = udp6_abort,
1543 .pr_ioctl = udp6_ioctl,
1544 .pr_stat = udp6_stat,
1545 .pr_peeraddr = udp6_peeraddr,
1546 .pr_sockaddr = udp6_sockaddr,
1547 .pr_rcvd = udp6_rcvd,
1548 .pr_recvoob = udp6_recvoob,
1549 .pr_send = udp6_send,
1550 .pr_sendoob = udp6_sendoob,
1551 .pr_purgeif = udp6_purgeif,
1552 };
1553