udp_usrreq.c revision 1.253 1 /* $NetBSD: udp_usrreq.c,v 1.253 2018/05/31 07:03:57 maxv 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 1.253 2018/05/31 07:03:57 maxv Exp $");
70
71 #ifdef _KERNEL_OPT
72 #include "opt_inet.h"
73 #include "opt_ipsec.h"
74 #include "opt_inet_csum.h"
75 #include "opt_ipkdb.h"
76 #include "opt_mbuftrace.h"
77 #include "opt_net_mpsafe.h"
78 #endif
79
80 #include <sys/param.h>
81 #include <sys/mbuf.h>
82 #include <sys/once.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/systm.h>
87 #include <sys/proc.h>
88 #include <sys/domain.h>
89 #include <sys/sysctl.h>
90
91 #include <net/if.h>
92
93 #include <netinet/in.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/in_var.h>
96 #include <netinet/ip.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/ip_var.h>
99 #include <netinet/ip_icmp.h>
100 #include <netinet/udp.h>
101 #include <netinet/udp_var.h>
102 #include <netinet/udp_private.h>
103
104 #ifdef INET6
105 #include <netinet/ip6.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet6/ip6_private.h>
108 #include <netinet6/in6_pcb.h>
109 #include <netinet6/udp6_var.h>
110 #include <netinet6/udp6_private.h>
111 #endif
112
113 #ifndef INET6
114 #include <netinet/ip6.h>
115 #endif
116
117 #ifdef IPSEC
118 #include <netipsec/ipsec.h>
119 #include <netipsec/esp.h>
120 #endif
121
122 #ifdef IPKDB
123 #include <ipkdb/ipkdb.h>
124 #endif
125
126 int udpcksum = 1;
127 int udp_do_loopback_cksum = 0;
128
129 struct inpcbtable udbtable;
130
131 percpu_t *udpstat_percpu;
132
133 #ifdef INET
134 #ifdef IPSEC
135 static int udp4_espinudp(struct mbuf **, int, struct socket *);
136 #endif
137 static void udp4_sendup(struct mbuf *, int, struct sockaddr *,
138 struct socket *);
139 static int udp4_realinput(struct sockaddr_in *, struct sockaddr_in *,
140 struct mbuf **, int);
141 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
142 #endif
143 #ifdef INET
144 static void udp_notify (struct inpcb *, int);
145 #endif
146
147 #ifndef UDBHASHSIZE
148 #define UDBHASHSIZE 128
149 #endif
150 int udbhashsize = UDBHASHSIZE;
151
152 /*
153 * For send - really max datagram size; for receive - 40 1K datagrams.
154 */
155 static int udp_sendspace = 9216;
156 static int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
157
158 #ifdef MBUFTRACE
159 struct mowner udp_mowner = MOWNER_INIT("udp", "");
160 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
161 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
162 #endif
163
164 #ifdef UDP_CSUM_COUNTERS
165 #include <sys/device.h>
166
167 #if defined(INET)
168 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
169 NULL, "udp", "hwcsum bad");
170 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
171 NULL, "udp", "hwcsum ok");
172 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
173 NULL, "udp", "hwcsum data");
174 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
175 NULL, "udp", "swcsum");
176
177 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
178 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
179 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
180 EVCNT_ATTACH_STATIC(udp_swcsum);
181 #endif /* defined(INET) */
182
183 #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
184 #else
185 #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */
186 #endif /* UDP_CSUM_COUNTERS */
187
188 static void sysctl_net_inet_udp_setup(struct sysctllog **);
189
190 static int
191 do_udpinit(void)
192 {
193
194 in_pcbinit(&udbtable, udbhashsize, udbhashsize);
195 udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
196
197 MOWNER_ATTACH(&udp_tx_mowner);
198 MOWNER_ATTACH(&udp_rx_mowner);
199 MOWNER_ATTACH(&udp_mowner);
200
201 return 0;
202 }
203
204 void
205 udp_init_common(void)
206 {
207 static ONCE_DECL(doudpinit);
208
209 RUN_ONCE(&doudpinit, do_udpinit);
210 }
211
212 void
213 udp_init(void)
214 {
215
216 sysctl_net_inet_udp_setup(NULL);
217
218 udp_init_common();
219 }
220
221 /*
222 * Checksum extended UDP header and data.
223 */
224 int
225 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
226 int iphlen, int len)
227 {
228
229 switch (af) {
230 #ifdef INET
231 case AF_INET:
232 return udp4_input_checksum(m, uh, iphlen, len);
233 #endif
234 #ifdef INET6
235 case AF_INET6:
236 return udp6_input_checksum(m, uh, iphlen, len);
237 #endif
238 }
239 #ifdef DIAGNOSTIC
240 panic("udp_input_checksum: unknown af %d", af);
241 #endif
242 /* NOTREACHED */
243 return -1;
244 }
245
246 #ifdef INET
247
248 /*
249 * Checksum extended UDP header and data.
250 */
251 static int
252 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
253 int iphlen, int len)
254 {
255
256 /*
257 * XXX it's better to record and check if this mbuf is
258 * already checked.
259 */
260
261 if (uh->uh_sum == 0)
262 return 0;
263
264 switch (m->m_pkthdr.csum_flags &
265 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv4) |
266 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
267 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
268 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
269 goto badcsum;
270
271 case M_CSUM_UDPv4|M_CSUM_DATA: {
272 u_int32_t hw_csum = m->m_pkthdr.csum_data;
273
274 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
275 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
276 const struct ip *ip =
277 mtod(m, const struct ip *);
278
279 hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
280 ip->ip_dst.s_addr,
281 htons(hw_csum + len + IPPROTO_UDP));
282 }
283 if ((hw_csum ^ 0xffff) != 0)
284 goto badcsum;
285 break;
286 }
287
288 case M_CSUM_UDPv4:
289 /* Checksum was okay. */
290 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
291 break;
292
293 default:
294 /*
295 * Need to compute it ourselves. Maybe skip checksum
296 * on loopback interfaces.
297 */
298 if (__predict_true(!(m_get_rcvif_NOMPSAFE(m)->if_flags &
299 IFF_LOOPBACK) ||
300 udp_do_loopback_cksum)) {
301 UDP_CSUM_COUNTER_INCR(&udp_swcsum);
302 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
303 goto badcsum;
304 }
305 break;
306 }
307
308 return 0;
309
310 badcsum:
311 UDP_STATINC(UDP_STAT_BADSUM);
312 return -1;
313 }
314
315 void
316 udp_input(struct mbuf *m, ...)
317 {
318 va_list ap;
319 struct sockaddr_in src, dst;
320 struct ip *ip;
321 struct udphdr *uh;
322 int iphlen;
323 int len;
324 int n;
325 u_int16_t ip_len;
326
327 va_start(ap, m);
328 iphlen = va_arg(ap, int);
329 (void)va_arg(ap, int); /* ignore value, advance ap */
330 va_end(ap);
331
332 MCLAIM(m, &udp_rx_mowner);
333 UDP_STATINC(UDP_STAT_IPACKETS);
334
335 /*
336 * Get IP and UDP header together in first mbuf.
337 */
338 ip = mtod(m, struct ip *);
339 M_REGION_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
340 if (uh == NULL) {
341 UDP_STATINC(UDP_STAT_HDROPS);
342 return;
343 }
344
345 /*
346 * Enforce alignment requirements that are violated in
347 * some cases, see kern/50766 for details.
348 */
349 if (UDP_HDR_ALIGNED_P(uh) == 0) {
350 m = m_copyup(m, iphlen + sizeof(struct udphdr), 0);
351 if (m == NULL) {
352 UDP_STATINC(UDP_STAT_HDROPS);
353 return;
354 }
355 ip = mtod(m, struct ip *);
356 uh = (struct udphdr *)(mtod(m, char *) + iphlen);
357 }
358 KASSERT(UDP_HDR_ALIGNED_P(uh));
359
360 /* destination port of 0 is illegal, based on RFC768. */
361 if (uh->uh_dport == 0)
362 goto bad;
363
364 /*
365 * Make mbuf data length reflect UDP length.
366 * If not enough data to reflect UDP length, drop.
367 */
368 ip_len = ntohs(ip->ip_len);
369 len = ntohs((u_int16_t)uh->uh_ulen);
370 if (len < sizeof(struct udphdr)) {
371 UDP_STATINC(UDP_STAT_BADLEN);
372 goto bad;
373 }
374 if (ip_len != iphlen + len) {
375 if (ip_len < iphlen + len) {
376 UDP_STATINC(UDP_STAT_BADLEN);
377 goto bad;
378 }
379 m_adj(m, iphlen + len - ip_len);
380 }
381
382 /*
383 * Checksum extended UDP header and data.
384 */
385 if (udp4_input_checksum(m, uh, iphlen, len))
386 goto badcsum;
387
388 /* construct source and dst sockaddrs. */
389 sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
390 sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
391
392 if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
393 UDP_STATINC(UDP_STAT_HDROPS);
394 return;
395 }
396 if (m == NULL) {
397 /*
398 * packet has been processed by ESP stuff -
399 * e.g. dropped NAT-T-keep-alive-packet ...
400 */
401 return;
402 }
403
404 ip = mtod(m, struct ip *);
405 M_REGION_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
406 if (uh == NULL) {
407 UDP_STATINC(UDP_STAT_HDROPS);
408 return;
409 }
410 /* XXX Re-enforce alignment? */
411
412 #ifdef INET6
413 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
414 struct sockaddr_in6 src6, dst6;
415
416 memset(&src6, 0, sizeof(src6));
417 src6.sin6_family = AF_INET6;
418 src6.sin6_len = sizeof(struct sockaddr_in6);
419 in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr);
420 src6.sin6_port = uh->uh_sport;
421 memset(&dst6, 0, sizeof(dst6));
422 dst6.sin6_family = AF_INET6;
423 dst6.sin6_len = sizeof(struct sockaddr_in6);
424 in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr);
425 dst6.sin6_port = uh->uh_dport;
426
427 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
428 }
429 #endif
430
431 if (n == 0) {
432 if (m->m_flags & (M_BCAST | M_MCAST)) {
433 UDP_STATINC(UDP_STAT_NOPORTBCAST);
434 goto bad;
435 }
436 UDP_STATINC(UDP_STAT_NOPORT);
437 #ifdef IPKDB
438 if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
439 m, iphlen + sizeof(struct udphdr),
440 m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
441 /*
442 * It was a debugger connect packet,
443 * just drop it now
444 */
445 goto bad;
446 }
447 #endif
448 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
449 m = NULL;
450 }
451
452 bad:
453 if (m)
454 m_freem(m);
455 return;
456
457 badcsum:
458 m_freem(m);
459 }
460 #endif
461
462 #ifdef INET
463 static void
464 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
465 struct sockaddr *src, struct socket *so)
466 {
467 struct mbuf *opts = NULL;
468 struct mbuf *n;
469 struct inpcb *inp;
470
471 KASSERT(so != NULL);
472 KASSERT(so->so_proto->pr_domain->dom_family == AF_INET);
473 inp = sotoinpcb(so);
474 KASSERT(inp != NULL);
475
476 #if defined(IPSEC)
477 if (ipsec_used && ipsec_in_reject(m, inp)) {
478 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
479 icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
480 0, 0);
481 return;
482 }
483 #endif
484
485 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
486 if (inp->inp_flags & INP_CONTROLOPTS ||
487 SOOPT_TIMESTAMP(so->so_options)) {
488 struct ip *ip = mtod(n, struct ip *);
489 ip_savecontrol(inp, &opts, ip, n);
490 }
491
492 m_adj(n, off);
493 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
494 m_freem(n);
495 if (opts)
496 m_freem(opts);
497 UDP_STATINC(UDP_STAT_FULLSOCK);
498 soroverflow(so);
499 } else
500 sorwakeup(so);
501 }
502 }
503 #endif
504
505 #ifdef INET
506 static int
507 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
508 struct mbuf **mp, int off /* offset of udphdr */)
509 {
510 u_int16_t *sport, *dport;
511 int rcvcnt;
512 struct in_addr *src4, *dst4;
513 struct inpcb_hdr *inph;
514 struct inpcb *inp;
515 struct mbuf *m = *mp;
516
517 rcvcnt = 0;
518 off += sizeof(struct udphdr); /* now, offset of payload */
519
520 if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
521 goto bad;
522
523 src4 = &src->sin_addr;
524 sport = &src->sin_port;
525 dst4 = &dst->sin_addr;
526 dport = &dst->sin_port;
527
528 if (IN_MULTICAST(dst4->s_addr) ||
529 in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) {
530 /*
531 * Deliver a multicast or broadcast datagram to *all* sockets
532 * for which the local and remote addresses and ports match
533 * those of the incoming datagram. This allows more than
534 * one process to receive multi/broadcasts on the same port.
535 * (This really ought to be done for unicast datagrams as
536 * well, but that would cause problems with existing
537 * applications that open both address-specific sockets and
538 * a wildcard socket listening to the same port -- they would
539 * end up receiving duplicates of every unicast datagram.
540 * Those applications open the multiple sockets to overcome an
541 * inadequacy of the UDP socket interface, but for backwards
542 * compatibility we avoid the problem here rather than
543 * fixing the interface. Maybe 4.5BSD will remedy this?)
544 */
545
546 /*
547 * KAME note: traditionally we dropped udpiphdr from mbuf here.
548 * we need udpiphdr for IPsec processing so we do that later.
549 */
550 /*
551 * Locate pcb(s) for datagram.
552 */
553 TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
554 inp = (struct inpcb *)inph;
555 if (inp->inp_af != AF_INET)
556 continue;
557
558 if (inp->inp_lport != *dport)
559 continue;
560 if (!in_nullhost(inp->inp_laddr)) {
561 if (!in_hosteq(inp->inp_laddr, *dst4))
562 continue;
563 }
564 if (!in_nullhost(inp->inp_faddr)) {
565 if (!in_hosteq(inp->inp_faddr, *src4) ||
566 inp->inp_fport != *sport)
567 continue;
568 }
569
570 udp4_sendup(m, off, (struct sockaddr *)src,
571 inp->inp_socket);
572 rcvcnt++;
573
574 /*
575 * Don't look for additional matches if this one does
576 * not have either the SO_REUSEPORT or SO_REUSEADDR
577 * socket options set. This heuristic avoids searching
578 * through all pcbs in the common case of a non-shared
579 * port. It assumes that an application will never
580 * clear these options after setting them.
581 */
582 if ((inp->inp_socket->so_options &
583 (SO_REUSEPORT|SO_REUSEADDR)) == 0)
584 break;
585 }
586 } else {
587 /*
588 * Locate pcb for datagram.
589 */
590 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
591 *dport, 0);
592 if (inp == 0) {
593 UDP_STATINC(UDP_STAT_PCBHASHMISS);
594 inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
595 if (inp == 0)
596 return rcvcnt;
597 }
598
599 #ifdef IPSEC
600 /* Handle ESP over UDP */
601 if (inp->inp_flags & INP_ESPINUDP) {
602 switch (udp4_espinudp(mp, off, inp->inp_socket)) {
603 case -1: /* Error, m was freed */
604 rcvcnt = -1;
605 goto bad;
606
607 case 1: /* ESP over UDP */
608 rcvcnt++;
609 goto bad;
610
611 case 0: /* plain UDP */
612 default: /* Unexpected */
613 /*
614 * Normal UDP processing will take place,
615 * m may have changed.
616 */
617 m = *mp;
618 break;
619 }
620 }
621 #endif
622
623 /*
624 * Check the minimum TTL for socket.
625 */
626 if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
627 goto bad;
628
629 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
630 rcvcnt++;
631 }
632
633 bad:
634 return rcvcnt;
635 }
636 #endif
637
638 #ifdef INET
639 /*
640 * Notify a udp user of an asynchronous error;
641 * just wake up so that he can collect error status.
642 */
643 static void
644 udp_notify(struct inpcb *inp, int errno)
645 {
646 inp->inp_socket->so_error = errno;
647 sorwakeup(inp->inp_socket);
648 sowwakeup(inp->inp_socket);
649 }
650
651 void *
652 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
653 {
654 struct ip *ip = v;
655 struct udphdr *uh;
656 void (*notify)(struct inpcb *, int) = udp_notify;
657 int errno;
658
659 if (sa->sa_family != AF_INET ||
660 sa->sa_len != sizeof(struct sockaddr_in))
661 return NULL;
662 if ((unsigned)cmd >= PRC_NCMDS)
663 return NULL;
664
665 errno = inetctlerrmap[cmd];
666 if (PRC_IS_REDIRECT(cmd)) {
667 notify = in_rtchange;
668 ip = NULL;
669 } else if (cmd == PRC_HOSTDEAD) {
670 ip = NULL;
671 } else if (errno == 0) {
672 return NULL;
673 }
674
675 if (ip) {
676 uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
677 in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
678 ip->ip_src, uh->uh_sport, errno, notify);
679 /* XXX mapped address case */
680 } else {
681 in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
682 notify);
683 }
684
685 return NULL;
686 }
687
688 int
689 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
690 {
691 int s;
692 int error = 0;
693 struct inpcb *inp;
694 int family;
695 int optval;
696
697 family = so->so_proto->pr_domain->dom_family;
698
699 s = splsoftnet();
700 switch (family) {
701 #ifdef INET
702 case PF_INET:
703 if (sopt->sopt_level != IPPROTO_UDP) {
704 error = ip_ctloutput(op, so, sopt);
705 goto end;
706 }
707 break;
708 #endif
709 #ifdef INET6
710 case PF_INET6:
711 if (sopt->sopt_level != IPPROTO_UDP) {
712 error = ip6_ctloutput(op, so, sopt);
713 goto end;
714 }
715 break;
716 #endif
717 default:
718 error = EAFNOSUPPORT;
719 goto end;
720 }
721
722
723 switch (op) {
724 case PRCO_SETOPT:
725 inp = sotoinpcb(so);
726
727 switch (sopt->sopt_name) {
728 case UDP_ENCAP:
729 error = sockopt_getint(sopt, &optval);
730 if (error)
731 break;
732
733 switch(optval) {
734 case 0:
735 inp->inp_flags &= ~INP_ESPINUDP;
736 break;
737
738 case UDP_ENCAP_ESPINUDP:
739 inp->inp_flags |= INP_ESPINUDP;
740 break;
741
742 default:
743 error = EINVAL;
744 break;
745 }
746 break;
747
748 default:
749 error = ENOPROTOOPT;
750 break;
751 }
752 break;
753
754 default:
755 error = EINVAL;
756 break;
757 }
758
759 end:
760 splx(s);
761 return error;
762 }
763
764 int
765 udp_output(struct mbuf *m, struct inpcb *inp, struct mbuf *control,
766 struct lwp *l)
767 {
768 struct udpiphdr *ui;
769 struct route *ro;
770 struct ip_pktopts pktopts;
771 kauth_cred_t cred;
772 int len = m->m_pkthdr.len;
773 int error, flags = 0;
774
775 MCLAIM(m, &udp_tx_mowner);
776
777 /*
778 * Calculate data length and get a mbuf
779 * for UDP and IP headers.
780 */
781 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
782 if (m == NULL) {
783 error = ENOBUFS;
784 goto release;
785 }
786
787 /*
788 * Compute the packet length of the IP header, and
789 * punt if the length looks bogus.
790 */
791 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
792 error = EMSGSIZE;
793 goto release;
794 }
795
796 if (l == NULL)
797 cred = NULL;
798 else
799 cred = l->l_cred;
800
801 /* Setup IP outgoing packet options */
802 memset(&pktopts, 0, sizeof(pktopts));
803 error = ip_setpktopts(control, &pktopts, &flags, inp, cred);
804 if (error != 0)
805 goto release;
806
807 if (control != NULL) {
808 m_freem(control);
809 control = NULL;
810 }
811
812 /*
813 * Fill in mbuf with extended UDP header
814 * and addresses and length put into network format.
815 */
816 ui = mtod(m, struct udpiphdr *);
817 ui->ui_pr = IPPROTO_UDP;
818 ui->ui_src = pktopts.ippo_laddr.sin_addr;
819 ui->ui_dst = inp->inp_faddr;
820 ui->ui_sport = inp->inp_lport;
821 ui->ui_dport = inp->inp_fport;
822 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
823
824 ro = &inp->inp_route;
825
826 /*
827 * Set up checksum and output datagram.
828 */
829 if (udpcksum) {
830 /*
831 * XXX Cache pseudo-header checksum part for
832 * XXX "connected" UDP sockets.
833 */
834 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
835 ui->ui_dst.s_addr, htons((u_int16_t)len +
836 sizeof(struct udphdr) + IPPROTO_UDP));
837 m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
838 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
839 } else
840 ui->ui_sum = 0;
841
842 ((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len);
843 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */
844 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */
845 UDP_STATINC(UDP_STAT_OPACKETS);
846
847 flags |= inp->inp_socket->so_options & (SO_DONTROUTE|SO_BROADCAST);
848 return ip_output(m, inp->inp_options, ro, flags, pktopts.ippo_imo, inp);
849
850 release:
851 if (control != NULL)
852 m_freem(control);
853 m_freem(m);
854 return error;
855 }
856
857 static int
858 udp_attach(struct socket *so, int proto)
859 {
860 struct inpcb *inp;
861 int error;
862
863 KASSERT(sotoinpcb(so) == NULL);
864
865 /* Assign the lock (must happen even if we will error out). */
866 sosetlock(so);
867
868 #ifdef MBUFTRACE
869 so->so_mowner = &udp_mowner;
870 so->so_rcv.sb_mowner = &udp_rx_mowner;
871 so->so_snd.sb_mowner = &udp_tx_mowner;
872 #endif
873 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
874 error = soreserve(so, udp_sendspace, udp_recvspace);
875 if (error) {
876 return error;
877 }
878 }
879
880 error = in_pcballoc(so, &udbtable);
881 if (error) {
882 return error;
883 }
884 inp = sotoinpcb(so);
885 inp->inp_ip.ip_ttl = ip_defttl;
886 KASSERT(solocked(so));
887
888 return error;
889 }
890
891 static void
892 udp_detach(struct socket *so)
893 {
894 struct inpcb *inp;
895
896 KASSERT(solocked(so));
897 inp = sotoinpcb(so);
898 KASSERT(inp != NULL);
899 in_pcbdetach(inp);
900 }
901
902 static int
903 udp_accept(struct socket *so, struct sockaddr *nam)
904 {
905 KASSERT(solocked(so));
906
907 panic("udp_accept");
908
909 return EOPNOTSUPP;
910 }
911
912 static int
913 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
914 {
915 struct inpcb *inp = sotoinpcb(so);
916 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
917 int error = 0;
918 int s;
919
920 KASSERT(solocked(so));
921 KASSERT(inp != NULL);
922 KASSERT(nam != NULL);
923
924 s = splsoftnet();
925 error = in_pcbbind(inp, sin, l);
926 splx(s);
927
928 return error;
929 }
930
931 static int
932 udp_listen(struct socket *so, struct lwp *l)
933 {
934 KASSERT(solocked(so));
935
936 return EOPNOTSUPP;
937 }
938
939 static int
940 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
941 {
942 struct inpcb *inp = sotoinpcb(so);
943 int error = 0;
944 int s;
945
946 KASSERT(solocked(so));
947 KASSERT(inp != NULL);
948 KASSERT(nam != NULL);
949
950 s = splsoftnet();
951 error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
952 if (! error)
953 soisconnected(so);
954 splx(s);
955 return error;
956 }
957
958 static int
959 udp_connect2(struct socket *so, struct socket *so2)
960 {
961 KASSERT(solocked(so));
962
963 return EOPNOTSUPP;
964 }
965
966 static int
967 udp_disconnect(struct socket *so)
968 {
969 struct inpcb *inp = sotoinpcb(so);
970 int s;
971
972 KASSERT(solocked(so));
973 KASSERT(inp != NULL);
974
975 s = splsoftnet();
976 /*soisdisconnected(so);*/
977 so->so_state &= ~SS_ISCONNECTED; /* XXX */
978 in_pcbdisconnect(inp);
979 inp->inp_laddr = zeroin_addr; /* XXX */
980 in_pcbstate(inp, INP_BOUND); /* XXX */
981 splx(s);
982
983 return 0;
984 }
985
986 static int
987 udp_shutdown(struct socket *so)
988 {
989 int s;
990
991 KASSERT(solocked(so));
992
993 s = splsoftnet();
994 socantsendmore(so);
995 splx(s);
996
997 return 0;
998 }
999
1000 static int
1001 udp_abort(struct socket *so)
1002 {
1003 KASSERT(solocked(so));
1004
1005 panic("udp_abort");
1006
1007 return EOPNOTSUPP;
1008 }
1009
1010 static int
1011 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
1012 {
1013 return in_control(so, cmd, nam, ifp);
1014 }
1015
1016 static int
1017 udp_stat(struct socket *so, struct stat *ub)
1018 {
1019 KASSERT(solocked(so));
1020
1021 /* stat: don't bother with a blocksize. */
1022 return 0;
1023 }
1024
1025 static int
1026 udp_peeraddr(struct socket *so, struct sockaddr *nam)
1027 {
1028 int s;
1029
1030 KASSERT(solocked(so));
1031 KASSERT(sotoinpcb(so) != NULL);
1032 KASSERT(nam != NULL);
1033
1034 s = splsoftnet();
1035 in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1036 splx(s);
1037
1038 return 0;
1039 }
1040
1041 static int
1042 udp_sockaddr(struct socket *so, struct sockaddr *nam)
1043 {
1044 int s;
1045
1046 KASSERT(solocked(so));
1047 KASSERT(sotoinpcb(so) != NULL);
1048 KASSERT(nam != NULL);
1049
1050 s = splsoftnet();
1051 in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1052 splx(s);
1053
1054 return 0;
1055 }
1056
1057 static int
1058 udp_rcvd(struct socket *so, int flags, struct lwp *l)
1059 {
1060 KASSERT(solocked(so));
1061
1062 return EOPNOTSUPP;
1063 }
1064
1065 static int
1066 udp_recvoob(struct socket *so, struct mbuf *m, int flags)
1067 {
1068 KASSERT(solocked(so));
1069
1070 return EOPNOTSUPP;
1071 }
1072
1073 static int
1074 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1075 struct mbuf *control, struct lwp *l)
1076 {
1077 struct inpcb *inp = sotoinpcb(so);
1078 int error = 0;
1079 struct in_addr laddr; /* XXX */
1080 int s;
1081
1082 KASSERT(solocked(so));
1083 KASSERT(inp != NULL);
1084 KASSERT(m != NULL);
1085
1086 memset(&laddr, 0, sizeof laddr);
1087
1088 s = splsoftnet();
1089 if (nam) {
1090 laddr = inp->inp_laddr; /* XXX */
1091 if ((so->so_state & SS_ISCONNECTED) != 0) {
1092 error = EISCONN;
1093 goto die;
1094 }
1095 error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
1096 if (error)
1097 goto die;
1098 } else {
1099 if ((so->so_state & SS_ISCONNECTED) == 0) {
1100 error = ENOTCONN;
1101 goto die;
1102 }
1103 }
1104 error = udp_output(m, inp, control, l);
1105 m = NULL;
1106 control = NULL;
1107 if (nam) {
1108 in_pcbdisconnect(inp);
1109 inp->inp_laddr = laddr; /* XXX */
1110 in_pcbstate(inp, INP_BOUND); /* XXX */
1111 }
1112 die:
1113 if (m != NULL)
1114 m_freem(m);
1115 if (control != NULL)
1116 m_freem(control);
1117
1118 splx(s);
1119 return error;
1120 }
1121
1122 static int
1123 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1124 {
1125 KASSERT(solocked(so));
1126
1127 m_freem(m);
1128 m_freem(control);
1129
1130 return EOPNOTSUPP;
1131 }
1132
1133 static int
1134 udp_purgeif(struct socket *so, struct ifnet *ifp)
1135 {
1136 int s;
1137
1138 s = splsoftnet();
1139 mutex_enter(softnet_lock);
1140 in_pcbpurgeif0(&udbtable, ifp);
1141 #ifdef NET_MPSAFE
1142 mutex_exit(softnet_lock);
1143 #endif
1144 in_purgeif(ifp);
1145 #ifdef NET_MPSAFE
1146 mutex_enter(softnet_lock);
1147 #endif
1148 in_pcbpurgeif(&udbtable, ifp);
1149 mutex_exit(softnet_lock);
1150 splx(s);
1151
1152 return 0;
1153 }
1154
1155 static int
1156 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1157 {
1158
1159 return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1160 }
1161
1162 /*
1163 * Sysctl for udp variables.
1164 */
1165 static void
1166 sysctl_net_inet_udp_setup(struct sysctllog **clog)
1167 {
1168
1169 sysctl_createv(clog, 0, NULL, NULL,
1170 CTLFLAG_PERMANENT,
1171 CTLTYPE_NODE, "inet", NULL,
1172 NULL, 0, NULL, 0,
1173 CTL_NET, PF_INET, CTL_EOL);
1174 sysctl_createv(clog, 0, NULL, NULL,
1175 CTLFLAG_PERMANENT,
1176 CTLTYPE_NODE, "udp",
1177 SYSCTL_DESCR("UDPv4 related settings"),
1178 NULL, 0, NULL, 0,
1179 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1180
1181 sysctl_createv(clog, 0, NULL, NULL,
1182 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1183 CTLTYPE_INT, "checksum",
1184 SYSCTL_DESCR("Compute UDP checksums"),
1185 NULL, 0, &udpcksum, 0,
1186 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1187 CTL_EOL);
1188 sysctl_createv(clog, 0, NULL, NULL,
1189 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1190 CTLTYPE_INT, "sendspace",
1191 SYSCTL_DESCR("Default UDP send buffer size"),
1192 NULL, 0, &udp_sendspace, 0,
1193 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1194 CTL_EOL);
1195 sysctl_createv(clog, 0, NULL, NULL,
1196 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1197 CTLTYPE_INT, "recvspace",
1198 SYSCTL_DESCR("Default UDP receive buffer size"),
1199 NULL, 0, &udp_recvspace, 0,
1200 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1201 CTL_EOL);
1202 sysctl_createv(clog, 0, NULL, NULL,
1203 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1204 CTLTYPE_INT, "do_loopback_cksum",
1205 SYSCTL_DESCR("Perform UDP checksum on loopback"),
1206 NULL, 0, &udp_do_loopback_cksum, 0,
1207 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1208 CTL_EOL);
1209 sysctl_createv(clog, 0, NULL, NULL,
1210 CTLFLAG_PERMANENT,
1211 CTLTYPE_STRUCT, "pcblist",
1212 SYSCTL_DESCR("UDP protocol control block list"),
1213 sysctl_inpcblist, 0, &udbtable, 0,
1214 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1215 CTL_EOL);
1216 sysctl_createv(clog, 0, NULL, NULL,
1217 CTLFLAG_PERMANENT,
1218 CTLTYPE_STRUCT, "stats",
1219 SYSCTL_DESCR("UDP statistics"),
1220 sysctl_net_inet_udp_stats, 0, NULL, 0,
1221 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1222 CTL_EOL);
1223 }
1224 #endif
1225
1226 void
1227 udp_statinc(u_int stat)
1228 {
1229
1230 KASSERT(stat < UDP_NSTATS);
1231 UDP_STATINC(stat);
1232 }
1233
1234 #if defined(INET) && defined(IPSEC)
1235 /*
1236 * Handle ESP-in-UDP packets (RFC3948).
1237 *
1238 * We need to distinguish between ESP packets and IKE packets. We do so by
1239 * looking at the Non-ESP marker. If IKE, we process the UDP packet as usual.
1240 * Otherwise, ESP, we invoke IPsec.
1241 *
1242 * Returns:
1243 * 1 if the packet was processed
1244 * 0 if normal UDP processing should take place
1245 * -1 if an error occurred and m was freed
1246 */
1247 static int
1248 udp4_espinudp(struct mbuf **mp, int off, struct socket *so)
1249 {
1250 const size_t skip = sizeof(struct udphdr);
1251 size_t len;
1252 uint8_t *data;
1253 size_t minlen;
1254 size_t iphdrlen;
1255 struct ip *ip;
1256 struct m_tag *tag;
1257 struct udphdr *udphdr;
1258 u_int16_t sport, dport;
1259 struct mbuf *m = *mp;
1260 uint32_t *marker;
1261
1262 /*
1263 * Collapse the mbuf chain if the first mbuf is too short.
1264 * The longest case is: UDP + max(Non-ESP, Non-IKE) + ESP.
1265 */
1266 minlen = off + 2 * sizeof(uint32_t) + sizeof(struct esp);
1267 if (minlen > m->m_pkthdr.len)
1268 minlen = m->m_pkthdr.len;
1269
1270 if (m->m_len < minlen) {
1271 if ((*mp = m_pullup(m, minlen)) == NULL) {
1272 return -1;
1273 }
1274 m = *mp;
1275 }
1276
1277 len = m->m_len - off;
1278 data = mtod(m, uint8_t *) + off;
1279
1280 /* Ignore keepalive packets. */
1281 if ((len == 1) && (*data == 0xff)) {
1282 m_freem(m);
1283 *mp = NULL; /* avoid any further processing by caller */
1284 return 1;
1285 }
1286
1287 /* Handle Non-ESP marker (32bit). If zero, then IKE. */
1288 marker = (uint32_t *)data;
1289 if (len <= sizeof(uint32_t))
1290 return 0;
1291 if (marker[0] == 0)
1292 return 0;
1293
1294 /*
1295 * Get the UDP ports. They are handled in network order
1296 * everywhere in the IPSEC_NAT_T code.
1297 */
1298 udphdr = (struct udphdr *)((char *)data - skip);
1299 sport = udphdr->uh_sport;
1300 dport = udphdr->uh_dport;
1301
1302 /*
1303 * Remove the UDP header, plus a possible marker. IP header
1304 * length is iphdrlen.
1305 *
1306 * Before:
1307 * <--- off --->
1308 * +----+------+-----+
1309 * | IP | UDP | ESP |
1310 * +----+------+-----+
1311 * <-skip->
1312 * After:
1313 * +----+-----+
1314 * | IP | ESP |
1315 * +----+-----+
1316 * <-skip->
1317 */
1318 iphdrlen = off - sizeof(struct udphdr);
1319 memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1320 m_adj(m, skip);
1321
1322 ip = mtod(m, struct ip *);
1323 ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1324 ip->ip_p = IPPROTO_ESP;
1325
1326 /*
1327 * We have modified the packet - it is now ESP, so we should not
1328 * return to UDP processing.
1329 *
1330 * Add a PACKET_TAG_IPSEC_NAT_T_PORTS tag to remember the source
1331 * UDP port. This is required if we want to select the right SPD
1332 * for multiple hosts behind same NAT.
1333 */
1334 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1335 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1336 m_freem(m);
1337 return -1;
1338 }
1339 ((u_int16_t *)(tag + 1))[0] = sport;
1340 ((u_int16_t *)(tag + 1))[1] = dport;
1341 m_tag_prepend(m, tag);
1342
1343 if (ipsec_used)
1344 ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
1345 else
1346 m_freem(m);
1347
1348 /* We handled it, it shouldn't be handled by UDP */
1349 *mp = NULL; /* avoid free by caller ... */
1350 return 1;
1351 }
1352 #endif
1353
1354 PR_WRAP_USRREQS(udp)
1355 #define udp_attach udp_attach_wrapper
1356 #define udp_detach udp_detach_wrapper
1357 #define udp_accept udp_accept_wrapper
1358 #define udp_bind udp_bind_wrapper
1359 #define udp_listen udp_listen_wrapper
1360 #define udp_connect udp_connect_wrapper
1361 #define udp_connect2 udp_connect2_wrapper
1362 #define udp_disconnect udp_disconnect_wrapper
1363 #define udp_shutdown udp_shutdown_wrapper
1364 #define udp_abort udp_abort_wrapper
1365 #define udp_ioctl udp_ioctl_wrapper
1366 #define udp_stat udp_stat_wrapper
1367 #define udp_peeraddr udp_peeraddr_wrapper
1368 #define udp_sockaddr udp_sockaddr_wrapper
1369 #define udp_rcvd udp_rcvd_wrapper
1370 #define udp_recvoob udp_recvoob_wrapper
1371 #define udp_send udp_send_wrapper
1372 #define udp_sendoob udp_sendoob_wrapper
1373 #define udp_purgeif udp_purgeif_wrapper
1374
1375 const struct pr_usrreqs udp_usrreqs = {
1376 .pr_attach = udp_attach,
1377 .pr_detach = udp_detach,
1378 .pr_accept = udp_accept,
1379 .pr_bind = udp_bind,
1380 .pr_listen = udp_listen,
1381 .pr_connect = udp_connect,
1382 .pr_connect2 = udp_connect2,
1383 .pr_disconnect = udp_disconnect,
1384 .pr_shutdown = udp_shutdown,
1385 .pr_abort = udp_abort,
1386 .pr_ioctl = udp_ioctl,
1387 .pr_stat = udp_stat,
1388 .pr_peeraddr = udp_peeraddr,
1389 .pr_sockaddr = udp_sockaddr,
1390 .pr_rcvd = udp_rcvd,
1391 .pr_recvoob = udp_recvoob,
1392 .pr_send = udp_send,
1393 .pr_sendoob = udp_sendoob,
1394 .pr_purgeif = udp_purgeif,
1395 };
1396