1/* $NetBSD: udp_usrreq.c,v 1.229 2016/11/18 06:50:04 knakahara Exp $ */
2
3/*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32/*
33 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
61 */
62
63/*
64 * UDP protocol implementation.
65 * Per RFC 768, August, 1980.
66 */
67
68#include <sys/cdefs.h>
69__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.229 2016/11/18 06:50:04 knakahara Exp $");
70
71#ifdef _KERNEL_OPT
72#include "opt_inet.h"
73#include "opt_compat_netbsd.h"
74#include "opt_ipsec.h"
75#include "opt_inet_csum.h"
76#include "opt_ipkdb.h"
77#include "opt_mbuftrace.h"
78#include "opt_net_mpsafe.h"
79#endif
80
81#include <sys/param.h>
82#include <sys/mbuf.h>
83#include <sys/once.h>
84#include <sys/protosw.h>
85#include <sys/socket.h>
86#include <sys/socketvar.h>
87#include <sys/systm.h>
88#include <sys/proc.h>
89#include <sys/domain.h>
90#include <sys/sysctl.h>
91
92#include <net/if.h>
93
94#include <netinet/in.h>
95#include <netinet/in_systm.h>
96#include <netinet/in_var.h>
97#include <netinet/ip.h>
98#include <netinet/in_pcb.h>
99#include <netinet/ip_var.h>
100#include <netinet/ip_icmp.h>
101#include <netinet/udp.h>
102#include <netinet/udp_var.h>
103#include <netinet/udp_private.h>
104
105#ifdef INET6
106#include <netinet/ip6.h>
107#include <netinet/icmp6.h>
108#include <netinet6/ip6_var.h>
109#include <netinet6/ip6_private.h>
110#include <netinet6/in6_pcb.h>
111#include <netinet6/udp6_var.h>
112#include <netinet6/udp6_private.h>
113#endif
114
115#ifndef INET6
116/* always need ip6.h for IP6_EXTHDR_GET */
117#include <netinet/ip6.h>
118#endif
119
120#ifdef IPSEC
121#include <netipsec/ipsec.h>
122#include <netipsec/ipsec_var.h>
123#include <netipsec/ipsec_private.h>
124#include <netipsec/esp.h>
125#ifdef INET6
126#include <netipsec/ipsec6.h>
127#endif
128#endif /* IPSEC */
129
130#ifdef COMPAT_50
131#include <compat/sys/socket.h>
132#endif
133
134#ifdef IPKDB
135#include <ipkdb/ipkdb.h>
136#endif
137
138int udpcksum = 1;
139int udp_do_loopback_cksum = 0;
140
141struct inpcbtable udbtable;
142
143percpu_t *udpstat_percpu;
144
145#ifdef INET
146#ifdef IPSEC
147static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
148 struct socket *);
149#endif
150static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
151 struct socket *);
152static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
153 struct mbuf **, int);
154static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
155#endif
156#ifdef INET
157static void udp_notify (struct inpcb *, int);
158#endif
159
160#ifndef UDBHASHSIZE
161#define UDBHASHSIZE 128
162#endif
163int udbhashsize = UDBHASHSIZE;
164
165/*
166 * For send - really max datagram size; for receive - 40 1K datagrams.
167 */
168static int udp_sendspace = 9216;
169static int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
170
171#ifdef MBUFTRACE
172struct mowner udp_mowner = MOWNER_INIT("udp", "");
173struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
174struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
175#endif
176
177#ifdef UDP_CSUM_COUNTERS
178#include <sys/device.h>
179
180#if defined(INET)
181struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
182 NULL, "udp", "hwcsum bad");
183struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
184 NULL, "udp", "hwcsum ok");
185struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
186 NULL, "udp", "hwcsum data");
187struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
188 NULL, "udp", "swcsum");
189
190EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
191EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
192EVCNT_ATTACH_STATIC(udp_hwcsum_data);
193EVCNT_ATTACH_STATIC(udp_swcsum);
194#endif /* defined(INET) */
195
196#define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
197#else
198#define UDP_CSUM_COUNTER_INCR(ev) /* nothing */
199#endif /* UDP_CSUM_COUNTERS */
200
201static void sysctl_net_inet_udp_setup(struct sysctllog **);
202
203static int
204do_udpinit(void)
205{
206
207 in_pcbinit(&udbtable, udbhashsize, udbhashsize);
208 udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
209
210 MOWNER_ATTACH(&udp_tx_mowner);
211 MOWNER_ATTACH(&udp_rx_mowner);
212 MOWNER_ATTACH(&udp_mowner);
213
214 return 0;
215}
216
217void
218udp_init_common(void)
219{
220 static ONCE_DECL(doudpinit);
221
222 RUN_ONCE(&doudpinit, do_udpinit);
223}
224
225void
226udp_init(void)
227{
228
229 sysctl_net_inet_udp_setup(NULL);
230
231 udp_init_common();
232}
233
234/*
235 * Checksum extended UDP header and data.
236 */
237
238int
239udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
240 int iphlen, int len)
241{
242
243 switch (af) {
244#ifdef INET
245 case AF_INET:
246 return udp4_input_checksum(m, uh, iphlen, len);
247#endif
248#ifdef INET6
249 case AF_INET6:
250 return udp6_input_checksum(m, uh, iphlen, len);
251#endif
252 }
253#ifdef DIAGNOSTIC
254 panic("udp_input_checksum: unknown af %d", af);
255#endif
256 /* NOTREACHED */
257 return -1;
258}
259
260#ifdef INET
261
262/*
263 * Checksum extended UDP header and data.
264 */
265
266static int
267udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
268 int iphlen, int len)
269{
270
271 /*
272 * XXX it's better to record and check if this mbuf is
273 * already checked.
274 */
275
276 if (uh->uh_sum == 0)
277 return 0;
278
279 switch (m->m_pkthdr.csum_flags &
280 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv4) |
281 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
282 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
283 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
284 goto badcsum;
285
286 case M_CSUM_UDPv4|M_CSUM_DATA: {
287 u_int32_t hw_csum = m->m_pkthdr.csum_data;
288
289 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
290 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
291 const struct ip *ip =
292 mtod(m, const struct ip *);
293
294 hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
295 ip->ip_dst.s_addr,
296 htons(hw_csum + len + IPPROTO_UDP));
297 }
298 if ((hw_csum ^ 0xffff) != 0)
299 goto badcsum;
300 break;
301 }
302
303 case M_CSUM_UDPv4:
304 /* Checksum was okay. */
305 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
306 break;
307
308 default:
309 /*
310 * Need to compute it ourselves. Maybe skip checksum
311 * on loopback interfaces.
312 */
313 if (__predict_true(!(m_get_rcvif_NOMPSAFE(m)->if_flags &
314 IFF_LOOPBACK) ||
315 udp_do_loopback_cksum)) {
316 UDP_CSUM_COUNTER_INCR(&udp_swcsum);
317 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
318 goto badcsum;
319 }
320 break;
321 }
322
323 return 0;
324
325badcsum:
326 UDP_STATINC(UDP_STAT_BADSUM);
327 return -1;
328}
329
330void
331udp_input(struct mbuf *m, ...)
332{
333 va_list ap;
334 struct sockaddr_in src, dst;
335 struct ip *ip;
336 struct udphdr *uh;
337 int iphlen;
338 int len;
339 int n;
340 u_int16_t ip_len;
341
342 va_start(ap, m);
343 iphlen = va_arg(ap, int);
344 (void)va_arg(ap, int); /* ignore value, advance ap */
345 va_end(ap);
346
347 MCLAIM(m, &udp_rx_mowner);
348 UDP_STATINC(UDP_STAT_IPACKETS);
349
350 /*
351 * Get IP and UDP header together in first mbuf.
352 */
353 ip = mtod(m, struct ip *);
354 IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
355 if (uh == NULL) {
356 UDP_STATINC(UDP_STAT_HDROPS);
357 return;
358 }
359 /*
360 * Enforce alignment requirements that are violated in
361 * some cases, see kern/50766 for details.
362 */
363 if (UDP_HDR_ALIGNED_P(uh) == 0) {
364 m = m_copyup(m, iphlen + sizeof(struct udphdr), 0);
365 if (m == NULL) {
366 UDP_STATINC(UDP_STAT_HDROPS);
367 return;
368 }
369 ip = mtod(m, struct ip *);
370 uh = (struct udphdr *)(mtod(m, char *) + iphlen);
371 }
372 KASSERT(UDP_HDR_ALIGNED_P(uh));
373
374 /* destination port of 0 is illegal, based on RFC768. */
375 if (uh->uh_dport == 0)
376 goto bad;
377
378 /*
379 * Make mbuf data length reflect UDP length.
380 * If not enough data to reflect UDP length, drop.
381 */
382 ip_len = ntohs(ip->ip_len);
383 len = ntohs((u_int16_t)uh->uh_ulen);
384 if (ip_len != iphlen + len) {
385 if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
386 UDP_STATINC(UDP_STAT_BADLEN);
387 goto bad;
388 }
389 m_adj(m, iphlen + len - ip_len);
390 }
391
392 /*
393 * Checksum extended UDP header and data.
394 */
395 if (udp4_input_checksum(m, uh, iphlen, len))
396 goto badcsum;
397
398 /* construct source and dst sockaddrs. */
399 sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
400 sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
401
402 if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
403 UDP_STATINC(UDP_STAT_HDROPS);
404 return;
405 }
406 if (m == NULL) {
407 /*
408 * packet has been processed by ESP stuff -
409 * e.g. dropped NAT-T-keep-alive-packet ...
410 */
411 return;
412 }
413 ip = mtod(m, struct ip *);
414#ifdef INET6
415 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
416 struct sockaddr_in6 src6, dst6;
417
418 memset(&src6, 0, sizeof(src6));
419 src6.sin6_family = AF_INET6;
420 src6.sin6_len = sizeof(struct sockaddr_in6);
421 in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr);
422 src6.sin6_port = uh->uh_sport;
423 memset(&dst6, 0, sizeof(dst6));
424 dst6.sin6_family = AF_INET6;
425 dst6.sin6_len = sizeof(struct sockaddr_in6);
426 in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr);
427 dst6.sin6_port = uh->uh_dport;
428
429 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
430 }
431#endif
432
433 if (n == 0) {
434 if (m->m_flags & (M_BCAST | M_MCAST)) {
435 UDP_STATINC(UDP_STAT_NOPORTBCAST);
436 goto bad;
437 }
438 UDP_STATINC(UDP_STAT_NOPORT);
439#ifdef IPKDB
440 if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
441 m, iphlen + sizeof(struct udphdr),
442 m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
443 /*
444 * It was a debugger connect packet,
445 * just drop it now
446 */
447 goto bad;
448 }
449#endif
450 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
451 m = NULL;
452 }
453
454bad:
455 if (m)
456 m_freem(m);
457 return;
458
459badcsum:
460 m_freem(m);
461}
462#endif
463
464#ifdef INET
465static void
466udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
467 struct sockaddr *src, struct socket *so)
468{
469 struct mbuf *opts = NULL;
470 struct mbuf *n;
471 struct inpcb *inp = NULL;
472
473 if (!so)
474 return;
475 switch (so->so_proto->pr_domain->dom_family) {
476 case AF_INET:
477 inp = sotoinpcb(so);
478 break;
479#ifdef INET6
480 case AF_INET6:
481 break;
482#endif
483 default:
484 return;
485 }
486
487#if defined(IPSEC)
488 /* check AH/ESP integrity. */
489 if (ipsec_used && so != NULL && ipsec4_in_reject_so(m, so)) {
490 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
491 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
492 icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
493 0, 0);
494 return;
495 }
496#endif /*IPSEC*/
497
498 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
499 if (inp && (inp->inp_flags & INP_CONTROLOPTS
500#ifdef SO_OTIMESTAMP
501 || so->so_options & SO_OTIMESTAMP
502#endif
503 || so->so_options & SO_TIMESTAMP)) {
504 struct ip *ip = mtod(n, struct ip *);
505 ip_savecontrol(inp, &opts, ip, n);
506 }
507
508 m_adj(n, off);
509 if (sbappendaddr(&so->so_rcv, src, n,
510 opts) == 0) {
511 m_freem(n);
512 if (opts)
513 m_freem(opts);
514 so->so_rcv.sb_overflowed++;
515 UDP_STATINC(UDP_STAT_FULLSOCK);
516 } else
517 sorwakeup(so);
518 }
519}
520#endif
521
522#ifdef INET
523static int
524udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
525 struct mbuf **mp, int off /* offset of udphdr */)
526{
527 u_int16_t *sport, *dport;
528 int rcvcnt;
529 struct in_addr *src4, *dst4;
530 struct inpcb_hdr *inph;
531 struct inpcb *inp;
532 struct mbuf *m = *mp;
533
534 rcvcnt = 0;
535 off += sizeof(struct udphdr); /* now, offset of payload */
536
537 if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
538 goto bad;
539
540 src4 = &src->sin_addr;
541 sport = &src->sin_port;
542 dst4 = &dst->sin_addr;
543 dport = &dst->sin_port;
544
545 if (IN_MULTICAST(dst4->s_addr) ||
546 in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) {
547 /*
548 * Deliver a multicast or broadcast datagram to *all* sockets
549 * for which the local and remote addresses and ports match
550 * those of the incoming datagram. This allows more than
551 * one process to receive multi/broadcasts on the same port.
552 * (This really ought to be done for unicast datagrams as
553 * well, but that would cause problems with existing
554 * applications that open both address-specific sockets and
555 * a wildcard socket listening to the same port -- they would
556 * end up receiving duplicates of every unicast datagram.
557 * Those applications open the multiple sockets to overcome an
558 * inadequacy of the UDP socket interface, but for backwards
559 * compatibility we avoid the problem here rather than
560 * fixing the interface. Maybe 4.5BSD will remedy this?)
561 */
562
563 /*
564 * KAME note: traditionally we dropped udpiphdr from mbuf here.
565 * we need udpiphdr for IPsec processing so we do that later.
566 */
567 /*
568 * Locate pcb(s) for datagram.
569 */
570 TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
571 inp = (struct inpcb *)inph;
572 if (inp->inp_af != AF_INET)
573 continue;
574
575 if (inp->inp_lport != *dport)
576 continue;
577 if (!in_nullhost(inp->inp_laddr)) {
578 if (!in_hosteq(inp->inp_laddr, *dst4))
579 continue;
580 }
581 if (!in_nullhost(inp->inp_faddr)) {
582 if (!in_hosteq(inp->inp_faddr, *src4) ||
583 inp->inp_fport != *sport)
584 continue;
585 }
586
587 udp4_sendup(m, off, (struct sockaddr *)src,
588 inp->inp_socket);
589 rcvcnt++;
590
591 /*
592 * Don't look for additional matches if this one does
593 * not have either the SO_REUSEPORT or SO_REUSEADDR
594 * socket options set. This heuristic avoids searching
595 * through all pcbs in the common case of a non-shared
596 * port. It assumes that an application will never
597 * clear these options after setting them.
598 */
599 if ((inp->inp_socket->so_options &
600 (SO_REUSEPORT|SO_REUSEADDR)) == 0)
601 break;
602 }
603 } else {
604 /*
605 * Locate pcb for datagram.
606 */
607 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
608 *dport, 0);
609 if (inp == 0) {
610 UDP_STATINC(UDP_STAT_PCBHASHMISS);
611 inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
612 if (inp == 0)
613 return rcvcnt;
614 }
615
616#ifdef IPSEC
617 /* Handle ESP over UDP */
618 if (inp->inp_flags & INP_ESPINUDP_ALL) {
619 struct sockaddr *sa = (struct sockaddr *)src;
620
621 switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
622 case -1: /* Error, m was freeed */
623 rcvcnt = -1;
624 goto bad;
625 break;
626
627 case 1: /* ESP over UDP */
628 rcvcnt++;
629 goto bad;
630 break;
631
632 case 0: /* plain UDP */
633 default: /* Unexpected */
634 /*
635 * Normal UDP processing will take place
636 * m may have changed.
637 */
638 m = *mp;
639 break;
640 }
641 }
642#endif
643
644 /*
645 * Check the minimum TTL for socket.
646 */
647 if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
648 goto bad;
649
650 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
651 rcvcnt++;
652 }
653
654bad:
655 return rcvcnt;
656}
657#endif
658
659#ifdef INET
660/*
661 * Notify a udp user of an asynchronous error;
662 * just wake up so that he can collect error status.
663 */
664static void
665udp_notify(struct inpcb *inp, int errno)
666{
667 inp->inp_socket->so_error = errno;
668 sorwakeup(inp->inp_socket);
669 sowwakeup(inp->inp_socket);
670}
671
672void *
673udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
674{
675 struct ip *ip = v;
676 struct udphdr *uh;
677 void (*notify)(struct inpcb *, int) = udp_notify;
678 int errno;
679
680 if (sa->sa_family != AF_INET
681 || sa->sa_len != sizeof(struct sockaddr_in))
682 return NULL;
683 if ((unsigned)cmd >= PRC_NCMDS)
684 return NULL;
685 errno = inetctlerrmap[cmd];
686 if (PRC_IS_REDIRECT(cmd))
687 notify = in_rtchange, ip = 0;
688 else if (cmd == PRC_HOSTDEAD)
689 ip = 0;
690 else if (errno == 0)
691 return NULL;
692 if (ip) {
693 uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
694 in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
695 ip->ip_src, uh->uh_sport, errno, notify);
696
697 /* XXX mapped address case */
698 } else
699 in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
700 notify);
701 return NULL;
702}
703
704int
705udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
706{
707 int s;
708 int error = 0;
709 struct inpcb *inp;
710 int family;
711 int optval;
712
713 family = so->so_proto->pr_domain->dom_family;
714
715 s = splsoftnet();
716 switch (family) {
717#ifdef INET
718 case PF_INET:
719 if (sopt->sopt_level != IPPROTO_UDP) {
720 error = ip_ctloutput(op, so, sopt);
721 goto end;
722 }
723 break;
724#endif
725#ifdef INET6
726 case PF_INET6:
727 if (sopt->sopt_level != IPPROTO_UDP) {
728 error = ip6_ctloutput(op, so, sopt);
729 goto end;
730 }
731 break;
732#endif
733 default:
734 error = EAFNOSUPPORT;
735 goto end;
736 }
737
738
739 switch (op) {
740 case PRCO_SETOPT:
741 inp = sotoinpcb(so);
742
743 switch (sopt->sopt_name) {
744 case UDP_ENCAP:
745 error = sockopt_getint(sopt, &optval);
746 if (error)
747 break;
748
749 switch(optval) {
750 case 0:
751 inp->inp_flags &= ~INP_ESPINUDP_ALL;
752 break;
753
754 case UDP_ENCAP_ESPINUDP:
755 inp->inp_flags &= ~INP_ESPINUDP_ALL;
756 inp->inp_flags |= INP_ESPINUDP;
757 break;
758
759 case UDP_ENCAP_ESPINUDP_NON_IKE:
760 inp->inp_flags &= ~INP_ESPINUDP_ALL;
761 inp->inp_flags |= INP_ESPINUDP_NON_IKE;
762 break;
763 default:
764 error = EINVAL;
765 break;
766 }
767 break;
768
769 default:
770 error = ENOPROTOOPT;
771 break;
772 }
773 break;
774
775 default:
776 error = EINVAL;
777 break;
778 }
779
780end:
781 splx(s);
782 return error;
783}
784
785
786int
787udp_output(struct mbuf *m, struct inpcb *inp)
788{
789 struct udpiphdr *ui;
790 struct route *ro;
791 int len = m->m_pkthdr.len;
792 int error = 0;
793
794 MCLAIM(m, &udp_tx_mowner);
795
796 /*
797 * Calculate data length and get a mbuf
798 * for UDP and IP headers.
799 */
800 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
801 if (m == 0) {
802 error = ENOBUFS;
803 goto release;
804 }
805
806 /*
807 * Compute the packet length of the IP header, and
808 * punt if the length looks bogus.
809 */
810 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
811 error = EMSGSIZE;
812 goto release;
813 }
814
815 /*
816 * Fill in mbuf with extended UDP header
817 * and addresses and length put into network format.
818 */
819 ui = mtod(m, struct udpiphdr *);
820 ui->ui_pr = IPPROTO_UDP;
821 ui->ui_src = inp->inp_laddr;
822 ui->ui_dst = inp->inp_faddr;
823 ui->ui_sport = inp->inp_lport;
824 ui->ui_dport = inp->inp_fport;
825 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
826
827 ro = &inp->inp_route;
828
829 /*
830 * Set up checksum and output datagram.
831 */
832 if (udpcksum) {
833 /*
834 * XXX Cache pseudo-header checksum part for
835 * XXX "connected" UDP sockets.
836 */
837 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
838 ui->ui_dst.s_addr, htons((u_int16_t)len +
839 sizeof(struct udphdr) + IPPROTO_UDP));
840 m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
841 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
842 } else
843 ui->ui_sum = 0;
844 ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
845 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */
846 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */
847 UDP_STATINC(UDP_STAT_OPACKETS);
848
849 return (ip_output(m, inp->inp_options, ro,
850 inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
851 inp->inp_moptions, inp->inp_socket));
852
853release:
854 m_freem(m);
855 return (error);
856}
857
858static int
859udp_attach(struct socket *so, int proto)
860{
861 struct inpcb *inp;
862 int error;
863
864 KASSERT(sotoinpcb(so) == NULL);
865
866 /* Assign the lock (must happen even if we will error out). */
867 sosetlock(so);
868
869#ifdef MBUFTRACE
870 so->so_mowner = &udp_mowner;
871 so->so_rcv.sb_mowner = &udp_rx_mowner;
872 so->so_snd.sb_mowner = &udp_tx_mowner;
873#endif
874 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
875 error = soreserve(so, udp_sendspace, udp_recvspace);
876 if (error) {
877 return error;
878 }
879 }
880
881 error = in_pcballoc(so, &udbtable);
882 if (error) {
883 return error;
884 }
885 inp = sotoinpcb(so);
886 inp->inp_ip.ip_ttl = ip_defttl;
887 KASSERT(solocked(so));
888
889 return error;
890}
891
892static void
893udp_detach(struct socket *so)
894{
895 struct inpcb *inp;
896
897 KASSERT(solocked(so));
898 inp = sotoinpcb(so);
899 KASSERT(inp != NULL);
900 in_pcbdetach(inp);
901}
902
903static int
904udp_accept(struct socket *so, struct sockaddr *nam)
905{
906 KASSERT(solocked(so));
907
908 panic("udp_accept");
909
910 return EOPNOTSUPP;
911}
912
913static int
914udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
915{
916 struct inpcb *inp = sotoinpcb(so);
917 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
918 int error = 0;
919 int s;
920
921 KASSERT(solocked(so));
922 KASSERT(inp != NULL);
923 KASSERT(nam != NULL);
924
925 s = splsoftnet();
926 error = in_pcbbind(inp, sin, l);
927 splx(s);
928
929 return error;
930}
931
932static int
933udp_listen(struct socket *so, struct lwp *l)
934{
935 KASSERT(solocked(so));
936
937 return EOPNOTSUPP;
938}
939
940static int
941udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
942{
943 struct inpcb *inp = sotoinpcb(so);
944 int error = 0;
945 int s;
946
947 KASSERT(solocked(so));
948 KASSERT(inp != NULL);
949 KASSERT(nam != NULL);
950
951 s = splsoftnet();
952 error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
953 if (! error)
954 soisconnected(so);
955 splx(s);
956 return error;
957}
958
959static int
960udp_connect2(struct socket *so, struct socket *so2)
961{
962 KASSERT(solocked(so));
963
964 return EOPNOTSUPP;
965}
966
967static int
968udp_disconnect(struct socket *so)
969{
970 struct inpcb *inp = sotoinpcb(so);
971 int s;
972
973 KASSERT(solocked(so));
974 KASSERT(inp != NULL);
975
976 s = splsoftnet();
977 /*soisdisconnected(so);*/
978 so->so_state &= ~SS_ISCONNECTED; /* XXX */
979 in_pcbdisconnect(inp);
980 inp->inp_laddr = zeroin_addr; /* XXX */
981 in_pcbstate(inp, INP_BOUND); /* XXX */
982 splx(s);
983
984 return 0;
985}
986
987static int
988udp_shutdown(struct socket *so)
989{
990 int s;
991
992 KASSERT(solocked(so));
993
994 s = splsoftnet();
995 socantsendmore(so);
996 splx(s);
997
998 return 0;
999}
1000
1001static int
1002udp_abort(struct socket *so)
1003{
1004 KASSERT(solocked(so));
1005
1006 panic("udp_abort");
1007
1008 return EOPNOTSUPP;
1009}
1010
1011static int
1012udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
1013{
1014 return in_control(so, cmd, nam, ifp);
1015}
1016
1017static int
1018udp_stat(struct socket *so, struct stat *ub)
1019{
1020 KASSERT(solocked(so));
1021
1022 /* stat: don't bother with a blocksize. */
1023 return 0;
1024}
1025
1026static int
1027udp_peeraddr(struct socket *so, struct sockaddr *nam)
1028{
1029 int s;
1030
1031 KASSERT(solocked(so));
1032 KASSERT(sotoinpcb(so) != NULL);
1033 KASSERT(nam != NULL);
1034
1035 s = splsoftnet();
1036 in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1037 splx(s);
1038
1039 return 0;
1040}
1041
1042static int
1043udp_sockaddr(struct socket *so, struct sockaddr *nam)
1044{
1045 int s;
1046
1047 KASSERT(solocked(so));
1048 KASSERT(sotoinpcb(so) != NULL);
1049 KASSERT(nam != NULL);
1050
1051 s = splsoftnet();
1052 in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1053 splx(s);
1054
1055 return 0;
1056}
1057
1058static int
1059udp_rcvd(struct socket *so, int flags, struct lwp *l)
1060{
1061 KASSERT(solocked(so));
1062
1063 return EOPNOTSUPP;
1064}
1065
1066static int
1067udp_recvoob(struct socket *so, struct mbuf *m, int flags)
1068{
1069 KASSERT(solocked(so));
1070
1071 return EOPNOTSUPP;
1072}
1073
1074static int
1075udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1076 struct mbuf *control, struct lwp *l)
1077{
1078 struct inpcb *inp = sotoinpcb(so);
1079 int error = 0;
1080 struct in_addr laddr; /* XXX */
1081 int s;
1082
1083 KASSERT(solocked(so));
1084 KASSERT(inp != NULL);
1085 KASSERT(m != NULL);
1086
1087 if (control && control->m_len) {
1088 m_freem(control);
1089 m_freem(m);
1090 return EINVAL;
1091 }
1092
1093 memset(&laddr, 0, sizeof laddr);
1094
1095 s = splsoftnet();
1096 if (nam) {
1097 laddr = inp->inp_laddr; /* XXX */
1098 if ((so->so_state & SS_ISCONNECTED) != 0) {
1099 error = EISCONN;
1100 goto die;
1101 }
1102 error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
1103 if (error)
1104 goto die;
1105 } else {
1106 if ((so->so_state & SS_ISCONNECTED) == 0) {
1107 error = ENOTCONN;
1108 goto die;
1109 }
1110 }
1111 error = udp_output(m, inp);
1112 m = NULL;
1113 if (nam) {
1114 in_pcbdisconnect(inp);
1115 inp->inp_laddr = laddr; /* XXX */
1116 in_pcbstate(inp, INP_BOUND); /* XXX */
1117 }
1118 die:
1119 if (m)
1120 m_freem(m);
1121
1122 splx(s);
1123 return error;
1124}
1125
1126static int
1127udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1128{
1129 KASSERT(solocked(so));
1130
1131 m_freem(m);
1132 m_freem(control);
1133
1134 return EOPNOTSUPP;
1135}
1136
1137static int
1138udp_purgeif(struct socket *so, struct ifnet *ifp)
1139{
1140 int s;
1141
1142 s = splsoftnet();
1143#ifndef NET_MPSAFE
1144 mutex_enter(softnet_lock);
1145#endif
1146 in_pcbpurgeif0(&udbtable, ifp);
1147 in_purgeif(ifp);
1148 in_pcbpurgeif(&udbtable, ifp);
1149#ifndef NET_MPSAFE
1150 mutex_exit(softnet_lock);
1151#endif
1152 splx(s);
1153
1154 return 0;
1155}
1156
1157static int
1158sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1159{
1160
1161 return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1162}
1163
1164/*
1165 * Sysctl for udp variables.
1166 */
1167static void
1168sysctl_net_inet_udp_setup(struct sysctllog **clog)
1169{
1170
1171 sysctl_createv(clog, 0, NULL, NULL,
1172 CTLFLAG_PERMANENT,
1173 CTLTYPE_NODE, "inet", NULL,
1174 NULL, 0, NULL, 0,
1175 CTL_NET, PF_INET, CTL_EOL);
1176 sysctl_createv(clog, 0, NULL, NULL,
1177 CTLFLAG_PERMANENT,
1178 CTLTYPE_NODE, "udp",
1179 SYSCTL_DESCR("UDPv4 related settings"),
1180 NULL, 0, NULL, 0,
1181 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1182
1183 sysctl_createv(clog, 0, NULL, NULL,
1184 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1185 CTLTYPE_INT, "checksum",
1186 SYSCTL_DESCR("Compute UDP checksums"),
1187 NULL, 0, &udpcksum, 0,
1188 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1189 CTL_EOL);
1190 sysctl_createv(clog, 0, NULL, NULL,
1191 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1192 CTLTYPE_INT, "sendspace",
1193 SYSCTL_DESCR("Default UDP send buffer size"),
1194 NULL, 0, &udp_sendspace, 0,
1195 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1196 CTL_EOL);
1197 sysctl_createv(clog, 0, NULL, NULL,
1198 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1199 CTLTYPE_INT, "recvspace",
1200 SYSCTL_DESCR("Default UDP receive buffer size"),
1201 NULL, 0, &udp_recvspace, 0,
1202 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1203 CTL_EOL);
1204 sysctl_createv(clog, 0, NULL, NULL,
1205 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1206 CTLTYPE_INT, "do_loopback_cksum",
1207 SYSCTL_DESCR("Perform UDP checksum on loopback"),
1208 NULL, 0, &udp_do_loopback_cksum, 0,
1209 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1210 CTL_EOL);
1211 sysctl_createv(clog, 0, NULL, NULL,
1212 CTLFLAG_PERMANENT,
1213 CTLTYPE_STRUCT, "pcblist",
1214 SYSCTL_DESCR("UDP protocol control block list"),
1215 sysctl_inpcblist, 0, &udbtable, 0,
1216 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1217 CTL_EOL);
1218 sysctl_createv(clog, 0, NULL, NULL,
1219 CTLFLAG_PERMANENT,
1220 CTLTYPE_STRUCT, "stats",
1221 SYSCTL_DESCR("UDP statistics"),
1222 sysctl_net_inet_udp_stats, 0, NULL, 0,
1223 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1224 CTL_EOL);
1225}
1226#endif
1227
1228void
1229udp_statinc(u_int stat)
1230{
1231
1232 KASSERT(stat < UDP_NSTATS);
1233 UDP_STATINC(stat);
1234}
1235
1236#if defined(INET) && defined(IPSEC)
1237/*
1238 * Returns:
1239 * 1 if the packet was processed
1240 * 0 if normal UDP processing should take place
1241 * -1 if an error occurent and m was freed
1242 */
1243static int
1244udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1245 struct socket *so)
1246{
1247 size_t len;
1248 void *data;
1249 struct inpcb *inp;
1250 size_t skip = 0;
1251 size_t minlen;
1252 size_t iphdrlen;
1253 struct ip *ip;
1254 struct m_tag *tag;
1255 struct udphdr *udphdr;
1256 u_int16_t sport, dport;
1257 struct mbuf *m = *mp;
1258
1259 /*
1260 * Collapse the mbuf chain if the first mbuf is too short
1261 * The longest case is: UDP + non ESP marker + ESP
1262 */
1263 minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1264 if (minlen > m->m_pkthdr.len)
1265 minlen = m->m_pkthdr.len;
1266
1267 if (m->m_len < minlen) {
1268 if ((*mp = m_pullup(m, minlen)) == NULL) {
1269 printf("udp4_espinudp: m_pullup failed\n");
1270 return -1;
1271 }
1272 m = *mp;
1273 }
1274
1275 len = m->m_len - off;
1276 data = mtod(m, char *) + off;
1277 inp = sotoinpcb(so);
1278
1279 /* Ignore keepalive packets */
1280 if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1281 m_free(m);
1282 *mp = NULL; /* avoid any further processiong by caller ... */
1283 return 1;
1284 }
1285
1286 /*
1287 * Check that the payload is long enough to hold
1288 * an ESP header and compute the length of encapsulation
1289 * header to remove
1290 */
1291 if (inp->inp_flags & INP_ESPINUDP) {
1292 u_int32_t *st = (u_int32_t *)data;
1293
1294 if ((len <= sizeof(struct esp)) || (*st == 0))
1295 return 0; /* Normal UDP processing */
1296
1297 skip = sizeof(struct udphdr);
1298 }
1299
1300 if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1301 u_int32_t *st = (u_int32_t *)data;
1302
1303 if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1304 || ((st[0] | st[1]) != 0))
1305 return 0; /* Normal UDP processing */
1306
1307 skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1308 }
1309
1310 /*
1311 * Get the UDP ports. They are handled in network
1312 * order everywhere in IPSEC_NAT_T code.
1313 */
1314 udphdr = (struct udphdr *)((char *)data - skip);
1315 sport = udphdr->uh_sport;
1316 dport = udphdr->uh_dport;
1317
1318 /*
1319 * Remove the UDP header (and possibly the non ESP marker)
1320 * IP header lendth is iphdrlen
1321 * Before:
1322 * <--- off --->
1323 * +----+------+-----+
1324 * | IP | UDP | ESP |
1325 * +----+------+-----+
1326 * <-skip->
1327 * After:
1328 * +----+-----+
1329 * | IP | ESP |
1330 * +----+-----+
1331 * <-skip->
1332 */
1333 iphdrlen = off - sizeof(struct udphdr);
1334 memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1335 m_adj(m, skip);
1336
1337 ip = mtod(m, struct ip *);
1338 ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1339 ip->ip_p = IPPROTO_ESP;
1340
1341 /*
1342 * We have modified the packet - it is now ESP, so we should not
1343 * return to UDP processing ...
1344 *
1345 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1346 * the source UDP port. This is required if we want
1347 * to select the right SPD for multiple hosts behind
1348 * same NAT
1349 */
1350 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1351 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1352 printf("udp4_espinudp: m_tag_get failed\n");
1353 m_freem(m);
1354 return -1;
1355 }
1356 ((u_int16_t *)(tag + 1))[0] = sport;
1357 ((u_int16_t *)(tag + 1))[1] = dport;
1358 m_tag_prepend(m, tag);
1359
1360 if (ipsec_used)
1361 ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
1362 /* XXX: else */
1363
1364 /* We handled it, it shouldn't be handled by UDP */
1365 *mp = NULL; /* avoid free by caller ... */
1366 return 1;
1367}
1368#endif
1369
1370PR_WRAP_USRREQS(udp)
1371#define udp_attach udp_attach_wrapper
1372#define udp_detach udp_detach_wrapper
1373#define udp_accept udp_accept_wrapper
1374#define udp_bind udp_bind_wrapper
1375#define udp_listen udp_listen_wrapper
1376#define udp_connect udp_connect_wrapper
1377#define udp_connect2 udp_connect2_wrapper
1378#define udp_disconnect udp_disconnect_wrapper
1379#define udp_shutdown udp_shutdown_wrapper
1380#define udp_abort udp_abort_wrapper
1381#define udp_ioctl udp_ioctl_wrapper
1382#define udp_stat udp_stat_wrapper
1383#define udp_peeraddr udp_peeraddr_wrapper
1384#define udp_sockaddr udp_sockaddr_wrapper
1385#define udp_rcvd udp_rcvd_wrapper
1386#define udp_recvoob udp_recvoob_wrapper
1387#define udp_send udp_send_wrapper
1388#define udp_sendoob udp_sendoob_wrapper
1389#define udp_purgeif udp_purgeif_wrapper
1390
1391const struct pr_usrreqs udp_usrreqs = {
1392 .pr_attach = udp_attach,
1393 .pr_detach = udp_detach,
1394 .pr_accept = udp_accept,
1395 .pr_bind = udp_bind,
1396 .pr_listen = udp_listen,
1397 .pr_connect = udp_connect,
1398 .pr_connect2 = udp_connect2,
1399 .pr_disconnect = udp_disconnect,
1400 .pr_shutdown = udp_shutdown,
1401 .pr_abort = udp_abort,
1402 .pr_ioctl = udp_ioctl,
1403 .pr_stat = udp_stat,
1404 .pr_peeraddr = udp_peeraddr,
1405 .pr_sockaddr = udp_sockaddr,
1406 .pr_rcvd = udp_rcvd,
1407 .pr_recvoob = udp_recvoob,
1408 .pr_send = udp_send,
1409 .pr_sendoob = udp_sendoob,
1410 .pr_purgeif = udp_purgeif,
1411};
1412