Annotation of src/sys/netinet/udp_usrreq.c, Revision 1.164
1.164 ! thorpej 1: /* $NetBSD: udp_usrreq.c,v 1.163 2007/11/27 22:45:30 christos Exp $ */
1.48 itojun 2:
3: /*
4: * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5: * All rights reserved.
1.94 itojun 6: *
1.48 itojun 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.
1.94 itojun 18: *
1.48 itojun 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: */
1.14 cgd 31:
1.1 cgd 32: /*
1.44 thorpej 33: * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
1.13 mycroft 34: * The Regents of the University of California. All rights reserved.
1.1 cgd 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.
1.104 agc 44: * 3. Neither the name of the University nor the names of its contributors
1.1 cgd 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: *
1.44 thorpej 60: * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
1.1 cgd 61: */
1.91 lukem 62:
63: #include <sys/cdefs.h>
1.164 ! thorpej 64: __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.163 2007/11/27 22:45:30 christos Exp $");
1.50 thorpej 65:
1.77 soda 66: #include "opt_inet.h"
1.50 thorpej 67: #include "opt_ipsec.h"
1.78 thorpej 68: #include "opt_inet_csum.h"
1.64 ws 69: #include "opt_ipkdb.h"
1.101 martin 70: #include "opt_mbuftrace.h"
1.1 cgd 71:
1.5 mycroft 72: #include <sys/param.h>
73: #include <sys/malloc.h>
74: #include <sys/mbuf.h>
75: #include <sys/protosw.h>
76: #include <sys/socket.h>
77: #include <sys/socketvar.h>
1.13 mycroft 78: #include <sys/errno.h>
1.5 mycroft 79: #include <sys/stat.h>
1.27 christos 80: #include <sys/systm.h>
81: #include <sys/proc.h>
1.53 itojun 82: #include <sys/domain.h>
1.27 christos 83: #include <sys/sysctl.h>
1.1 cgd 84:
1.5 mycroft 85: #include <net/if.h>
86: #include <net/route.h>
1.1 cgd 87:
1.5 mycroft 88: #include <netinet/in.h>
89: #include <netinet/in_systm.h>
1.15 cgd 90: #include <netinet/in_var.h>
1.5 mycroft 91: #include <netinet/ip.h>
92: #include <netinet/in_pcb.h>
93: #include <netinet/ip_var.h>
94: #include <netinet/ip_icmp.h>
95: #include <netinet/udp.h>
96: #include <netinet/udp_var.h>
1.1 cgd 97:
1.53 itojun 98: #ifdef INET6
99: #include <netinet/ip6.h>
100: #include <netinet/icmp6.h>
101: #include <netinet6/ip6_var.h>
102: #include <netinet6/in6_pcb.h>
103: #include <netinet6/udp6_var.h>
1.146 rpaulo 104: #include <netinet6/scope6_var.h>
1.53 itojun 105: #endif
106:
107: #ifndef INET6
108: /* always need ip6.h for IP6_EXTHDR_GET */
109: #include <netinet/ip6.h>
110: #endif
111:
1.76 itojun 112: #include "faith.h"
113: #if defined(NFAITH) && NFAITH > 0
114: #include <net/if_faith.h>
115: #endif
116:
1.27 christos 117: #include <machine/stdarg.h>
118:
1.105 jonathan 119: #ifdef FAST_IPSEC
120: #include <netipsec/ipsec.h>
1.121 jonathan 121: #include <netipsec/ipsec_var.h> /* XXX ipsecstat namespace */
1.147 christos 122: #include <netipsec/esp.h>
1.105 jonathan 123: #ifdef INET6
124: #include <netipsec/ipsec6.h>
125: #endif
126: #endif /* FAST_IPSEC*/
127:
1.48 itojun 128: #ifdef IPSEC
129: #include <netinet6/ipsec.h>
1.147 christos 130: #include <netinet6/esp.h>
1.48 itojun 131: #include <netkey/key.h>
132: #endif /*IPSEC*/
133:
1.64 ws 134: #ifdef IPKDB
135: #include <ipkdb/ipkdb.h>
136: #endif
137:
1.8 mycroft 138: /*
139: * UDP protocol implementation.
140: * Per RFC 768, August, 1980.
141: */
142: int udpcksum = 1;
1.141 yamt 143: int udp_do_loopback_cksum = 0;
1.93 matt 144:
145: struct inpcbtable udbtable;
1.164 ! thorpej 146:
! 147: uint64_t udpstat[UDP_NSTATS];
1.8 mycroft 148:
1.72 itojun 149: #ifdef INET
1.130 manu 150: #ifdef IPSEC_NAT_T
1.144 manu 151: static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
1.130 manu 152: struct socket *);
153: #endif
1.119 matt 154: static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
155: struct socket *);
156: static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
1.144 manu 157: struct mbuf **, int);
1.129 yamt 158: static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
1.72 itojun 159: #endif
1.53 itojun 160: #ifdef INET6
1.119 matt 161: static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
162: struct socket *);
163: static int udp6_realinput (int, struct sockaddr_in6 *,
164: struct sockaddr_in6 *, struct mbuf *, int);
1.129 yamt 165: static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
1.53 itojun 166: #endif
1.72 itojun 167: #ifdef INET
1.119 matt 168: static void udp_notify (struct inpcb *, int);
1.72 itojun 169: #endif
1.7 mycroft 170:
1.26 mycroft 171: #ifndef UDBHASHSIZE
172: #define UDBHASHSIZE 128
173: #endif
174: int udbhashsize = UDBHASHSIZE;
175:
1.98 matt 176: #ifdef MBUFTRACE
1.150 dogcow 177: struct mowner udp_mowner = MOWNER_INIT("udp", "");
178: struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
179: struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
1.98 matt 180: #endif
181:
1.78 thorpej 182: #ifdef UDP_CSUM_COUNTERS
183: #include <sys/device.h>
184:
1.140 yamt 185: #if defined(INET)
1.78 thorpej 186: struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
187: NULL, "udp", "hwcsum bad");
188: struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
189: NULL, "udp", "hwcsum ok");
190: struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
191: NULL, "udp", "hwcsum data");
192: struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
193: NULL, "udp", "swcsum");
194:
1.120 matt 195: EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
196: EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
197: EVCNT_ATTACH_STATIC(udp_hwcsum_data);
198: EVCNT_ATTACH_STATIC(udp_swcsum);
1.140 yamt 199: #endif /* defined(INET) */
200:
201: #if defined(INET6)
202: struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
203: NULL, "udp6", "hwcsum bad");
204: struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
205: NULL, "udp6", "hwcsum ok");
206: struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
207: NULL, "udp6", "hwcsum data");
208: struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
209: NULL, "udp6", "swcsum");
210:
211: EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
212: EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
213: EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
214: EVCNT_ATTACH_STATIC(udp6_swcsum);
215: #endif /* defined(INET6) */
216:
217: #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
1.120 matt 218:
1.78 thorpej 219: #else
220:
221: #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */
222:
223: #endif /* UDP_CSUM_COUNTERS */
224:
1.7 mycroft 225: void
1.119 matt 226: udp_init(void)
1.1 cgd 227: {
1.18 mycroft 228:
1.35 mycroft 229: in_pcbinit(&udbtable, udbhashsize, udbhashsize);
1.78 thorpej 230:
1.98 matt 231: MOWNER_ATTACH(&udp_tx_mowner);
232: MOWNER_ATTACH(&udp_rx_mowner);
233: MOWNER_ATTACH(&udp_mowner);
1.1 cgd 234: }
235:
1.129 yamt 236: /*
237: * Checksum extended UDP header and data.
238: */
239:
240: int
241: udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
242: int iphlen, int len)
243: {
244:
245: switch (af) {
1.72 itojun 246: #ifdef INET
1.129 yamt 247: case AF_INET:
248: return udp4_input_checksum(m, uh, iphlen, len);
249: #endif
250: #ifdef INET6
251: case AF_INET6:
252: return udp6_input_checksum(m, uh, iphlen, len);
253: #endif
254: }
255: #ifdef DIAGNOSTIC
256: panic("udp_input_checksum: unknown af %d", af);
257: #endif
258: /* NOTREACHED */
259: return -1;
260: }
261:
262: #ifdef INET
263:
264: /*
265: * Checksum extended UDP header and data.
266: */
267:
268: static int
269: udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
270: int iphlen, int len)
271: {
272:
273: /*
274: * XXX it's better to record and check if this mbuf is
275: * already checked.
276: */
277:
278: if (uh->uh_sum == 0)
279: return 0;
280:
281: switch (m->m_pkthdr.csum_flags &
282: ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
283: M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
284: case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
285: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
286: goto badcsum;
287:
288: case M_CSUM_UDPv4|M_CSUM_DATA: {
289: u_int32_t hw_csum = m->m_pkthdr.csum_data;
290:
291: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
292: if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
293: const struct ip *ip =
294: mtod(m, const struct ip *);
295:
296: hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
297: ip->ip_dst.s_addr,
298: htons(hw_csum + len + IPPROTO_UDP));
299: }
300: if ((hw_csum ^ 0xffff) != 0)
301: goto badcsum;
302: break;
303: }
304:
305: case M_CSUM_UDPv4:
306: /* Checksum was okay. */
307: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
308: break;
309:
310: default:
311: /*
312: * Need to compute it ourselves. Maybe skip checksum
313: * on loopback interfaces.
314: */
315: if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
316: IFF_LOOPBACK) ||
317: udp_do_loopback_cksum)) {
318: UDP_CSUM_COUNTER_INCR(&udp_swcsum);
319: if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
320: goto badcsum;
321: }
322: break;
323: }
324:
325: return 0;
326:
327: badcsum:
1.164 ! thorpej 328: udpstat[UDP_STAT_BADSUM]++;
1.129 yamt 329: return -1;
330: }
331:
1.7 mycroft 332: void
1.27 christos 333: udp_input(struct mbuf *m, ...)
1.1 cgd 334: {
1.53 itojun 335: va_list ap;
336: struct sockaddr_in src, dst;
337: struct ip *ip;
338: struct udphdr *uh;
1.97 simonb 339: int iphlen;
1.53 itojun 340: int len;
341: int n;
1.96 itojun 342: u_int16_t ip_len;
1.53 itojun 343:
344: va_start(ap, m);
345: iphlen = va_arg(ap, int);
1.97 simonb 346: (void)va_arg(ap, int); /* ignore value, advance ap */
1.53 itojun 347: va_end(ap);
348:
1.98 matt 349: MCLAIM(m, &udp_rx_mowner);
1.164 ! thorpej 350: udpstat[UDP_STAT_IPACKETS]++;
1.53 itojun 351:
352: /*
353: * Get IP and UDP header together in first mbuf.
354: */
355: ip = mtod(m, struct ip *);
356: IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
357: if (uh == NULL) {
1.164 ! thorpej 358: udpstat[UDP_STAT_HDROPS]++;
1.53 itojun 359: return;
360: }
1.95 thorpej 361: KASSERT(UDP_HDR_ALIGNED_P(uh));
1.53 itojun 362:
1.57 itojun 363: /* destination port of 0 is illegal, based on RFC768. */
364: if (uh->uh_dport == 0)
365: goto bad;
366:
1.53 itojun 367: /*
368: * Make mbuf data length reflect UDP length.
369: * If not enough data to reflect UDP length, drop.
370: */
1.96 itojun 371: ip_len = ntohs(ip->ip_len);
1.53 itojun 372: len = ntohs((u_int16_t)uh->uh_ulen);
1.96 itojun 373: if (ip_len != iphlen + len) {
374: if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
1.164 ! thorpej 375: udpstat[UDP_STAT_BADLEN]++;
1.53 itojun 376: goto bad;
377: }
1.96 itojun 378: m_adj(m, iphlen + len - ip_len);
1.53 itojun 379: }
380:
381: /*
382: * Checksum extended UDP header and data.
383: */
1.129 yamt 384: if (udp4_input_checksum(m, uh, iphlen, len))
385: goto badcsum;
1.53 itojun 386:
387: /* construct source and dst sockaddrs. */
1.159 dyoung 388: sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
389: sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
1.53 itojun 390:
1.144 manu 391: if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
1.164 ! thorpej 392: udpstat[UDP_STAT_HDROPS]++;
1.144 manu 393: return;
394: }
1.53 itojun 395: #ifdef INET6
396: if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
397: struct sockaddr_in6 src6, dst6;
398:
399: bzero(&src6, sizeof(src6));
400: src6.sin6_family = AF_INET6;
401: src6.sin6_len = sizeof(struct sockaddr_in6);
402: src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
403: bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
404: sizeof(ip->ip_src));
405: src6.sin6_port = uh->uh_sport;
406: bzero(&dst6, sizeof(dst6));
407: dst6.sin6_family = AF_INET6;
408: dst6.sin6_len = sizeof(struct sockaddr_in6);
409: dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
410: bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
411: sizeof(ip->ip_dst));
412: dst6.sin6_port = uh->uh_dport;
413:
414: n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
415: }
416: #endif
417:
418: if (n == 0) {
419: if (m->m_flags & (M_BCAST | M_MCAST)) {
1.164 ! thorpej 420: udpstat[UDP_STAT_NOPORTBCAST]++;
1.53 itojun 421: goto bad;
422: }
1.164 ! thorpej 423: udpstat[UDP_STAT_NOPORT]++;
1.64 ws 424: #ifdef IPKDB
1.53 itojun 425: if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
426: m, iphlen + sizeof(struct udphdr),
427: m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
428: /*
429: * It was a debugger connect packet,
430: * just drop it now
431: */
432: goto bad;
433: }
434: #endif
435: icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
436: m = NULL;
437: }
438:
439: bad:
440: if (m)
441: m_freem(m);
1.78 thorpej 442: return;
443:
444: badcsum:
445: m_freem(m);
1.53 itojun 446: }
1.72 itojun 447: #endif
1.53 itojun 448:
449: #ifdef INET6
1.129 yamt 450: static int
451: udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
452: {
453:
1.140 yamt 454: /*
455: * XXX it's better to record and check if this mbuf is
456: * already checked.
457: */
458:
1.129 yamt 459: if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
460: goto good;
461: }
462: if (uh->uh_sum == 0) {
463: udp6stat.udp6s_nosum++;
464: goto bad;
465: }
1.140 yamt 466:
467: switch (m->m_pkthdr.csum_flags &
468: ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
469: M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
470: case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
471: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
1.129 yamt 472: udp6stat.udp6s_badsum++;
473: goto bad;
1.140 yamt 474:
475: #if 0 /* notyet */
476: case M_CSUM_UDPv6|M_CSUM_DATA:
477: #endif
478:
479: case M_CSUM_UDPv6:
480: /* Checksum was okay. */
481: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
482: break;
483:
484: default:
485: /*
486: * Need to compute it ourselves. Maybe skip checksum
487: * on loopback interfaces.
488: */
489: UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
490: if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
491: udp6stat.udp6s_badsum++;
492: goto bad;
493: }
1.129 yamt 494: }
495:
496: good:
497: return 0;
498: bad:
499: return -1;
500: }
501:
1.53 itojun 502: int
1.154 yamt 503: udp6_input(struct mbuf **mp, int *offp, int proto)
1.53 itojun 504: {
505: struct mbuf *m = *mp;
506: int off = *offp;
507: struct sockaddr_in6 src, dst;
508: struct ip6_hdr *ip6;
509: struct udphdr *uh;
510: u_int32_t plen, ulen;
511:
1.76 itojun 512: ip6 = mtod(m, struct ip6_hdr *);
513:
1.53 itojun 514: #if defined(NFAITH) && 0 < NFAITH
1.76 itojun 515: if (faithprefix(&ip6->ip6_dst)) {
516: /* send icmp6 host unreach? */
517: m_freem(m);
518: return IPPROTO_DONE;
1.53 itojun 519: }
520: #endif
521:
522: udp6stat.udp6s_ipackets++;
523:
524: /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */
525: plen = m->m_pkthdr.len - off;
526: IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
527: if (uh == NULL) {
528: ip6stat.ip6s_tooshort++;
529: return IPPROTO_DONE;
530: }
1.95 thorpej 531: KASSERT(UDP_HDR_ALIGNED_P(uh));
1.53 itojun 532: ulen = ntohs((u_short)uh->uh_ulen);
1.69 itojun 533: /*
534: * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
535: * iff payload length > 0xffff.
536: */
1.53 itojun 537: if (ulen == 0 && plen > 0xffff)
538: ulen = plen;
539:
540: if (plen != ulen) {
541: udp6stat.udp6s_badlen++;
1.54 itojun 542: goto bad;
543: }
544:
1.57 itojun 545: /* destination port of 0 is illegal, based on RFC768. */
546: if (uh->uh_dport == 0)
547: goto bad;
548:
1.54 itojun 549: /* Be proactive about malicious use of IPv4 mapped address */
550: if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
551: IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
552: /* XXX stat */
1.53 itojun 553: goto bad;
554: }
555:
556: /*
1.125 thorpej 557: * Checksum extended UDP header and data. Maybe skip checksum
558: * on loopback interfaces.
1.53 itojun 559: */
1.129 yamt 560: if (udp6_input_checksum(m, uh, off, ulen))
561: goto bad;
1.53 itojun 562:
563: /*
564: * Construct source and dst sockaddrs.
565: */
566: bzero(&src, sizeof(src));
567: src.sin6_family = AF_INET6;
568: src.sin6_len = sizeof(struct sockaddr_in6);
1.146 rpaulo 569: src.sin6_addr = ip6->ip6_src;
1.53 itojun 570: src.sin6_port = uh->uh_sport;
571: bzero(&dst, sizeof(dst));
572: dst.sin6_family = AF_INET6;
573: dst.sin6_len = sizeof(struct sockaddr_in6);
1.146 rpaulo 574: dst.sin6_addr = ip6->ip6_dst;
1.53 itojun 575: dst.sin6_port = uh->uh_dport;
576:
577: if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
578: if (m->m_flags & M_MCAST) {
579: udp6stat.udp6s_noportmcast++;
580: goto bad;
581: }
1.61 itojun 582: udp6stat.udp6s_noport++;
1.53 itojun 583: icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
584: m = NULL;
585: }
586:
587: bad:
588: if (m)
589: m_freem(m);
590: return IPPROTO_DONE;
591: }
592: #endif
593:
1.72 itojun 594: #ifdef INET
1.53 itojun 595: static void
1.119 matt 596: udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
597: struct sockaddr *src, struct socket *so)
1.53 itojun 598: {
599: struct mbuf *opts = NULL;
600: struct mbuf *n;
601: struct inpcb *inp = NULL;
602:
603: if (!so)
604: return;
605: switch (so->so_proto->pr_domain->dom_family) {
606: case AF_INET:
607: inp = sotoinpcb(so);
608: break;
609: #ifdef INET6
610: case AF_INET6:
611: break;
612: #endif
613: default:
614: return;
615: }
616:
1.105 jonathan 617: #if defined(IPSEC) || defined(FAST_IPSEC)
1.53 itojun 618: /* check AH/ESP integrity. */
619: if (so != NULL && ipsec4_in_reject_so(m, so)) {
620: ipsecstat.in_polvio++;
1.162 dyoung 621: if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
1.110 itojun 622: icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
623: 0, 0);
1.53 itojun 624: return;
625: }
626: #endif /*IPSEC*/
627:
1.162 dyoung 628: if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
1.53 itojun 629: if (inp && (inp->inp_flags & INP_CONTROLOPTS
630: || so->so_options & SO_TIMESTAMP)) {
631: struct ip *ip = mtod(n, struct ip *);
632: ip_savecontrol(inp, &opts, ip, n);
633: }
634:
635: m_adj(n, off);
636: if (sbappendaddr(&so->so_rcv, src, n,
637: opts) == 0) {
638: m_freem(n);
639: if (opts)
640: m_freem(opts);
1.124 darrenr 641: so->so_rcv.sb_overflowed++;
1.164 ! thorpej 642: udpstat[UDP_STAT_FULLSOCK]++;
1.53 itojun 643: } else
644: sorwakeup(so);
645: }
646: }
1.72 itojun 647: #endif
1.53 itojun 648:
649: #ifdef INET6
650: static void
1.119 matt 651: udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
652: struct sockaddr *src, struct socket *so)
1.53 itojun 653: {
654: struct mbuf *opts = NULL;
655: struct mbuf *n;
656: struct in6pcb *in6p = NULL;
657:
658: if (!so)
659: return;
660: if (so->so_proto->pr_domain->dom_family != AF_INET6)
661: return;
662: in6p = sotoin6pcb(so);
663:
1.105 jonathan 664: #if defined(IPSEC) || defined(FAST_IPSEC)
1.53 itojun 665: /* check AH/ESP integrity. */
666: if (so != NULL && ipsec6_in_reject_so(m, so)) {
667: ipsec6stat.in_polvio++;
1.162 dyoung 668: if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
1.110 itojun 669: icmp6_error(n, ICMP6_DST_UNREACH,
670: ICMP6_DST_UNREACH_ADMIN, 0);
1.53 itojun 671: return;
672: }
673: #endif /*IPSEC*/
674:
1.162 dyoung 675: if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
1.53 itojun 676: if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
677: || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
678: struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
679: ip6_savecontrol(in6p, &opts, ip6, n);
680: }
681:
682: m_adj(n, off);
683: if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
684: m_freem(n);
685: if (opts)
686: m_freem(opts);
1.124 darrenr 687: so->so_rcv.sb_overflowed++;
1.53 itojun 688: udp6stat.udp6s_fullsock++;
689: } else
690: sorwakeup(so);
691: }
692: }
693: #endif
694:
1.72 itojun 695: #ifdef INET
1.53 itojun 696: static int
1.119 matt 697: udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
1.144 manu 698: struct mbuf **mp, int off /* offset of udphdr */)
1.53 itojun 699: {
700: u_int16_t *sport, *dport;
701: int rcvcnt;
702: struct in_addr *src4, *dst4;
1.109 itojun 703: struct inpcb_hdr *inph;
1.53 itojun 704: struct inpcb *inp;
1.144 manu 705: struct mbuf *m = *mp;
1.53 itojun 706:
707: rcvcnt = 0;
708: off += sizeof(struct udphdr); /* now, offset of payload */
709:
710: if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
711: goto bad;
712:
713: src4 = &src->sin_addr;
714: sport = &src->sin_port;
715: dst4 = &dst->sin_addr;
716: dport = &dst->sin_port;
717:
1.73 itojun 718: if (IN_MULTICAST(dst4->s_addr) ||
1.53 itojun 719: in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
720: /*
721: * Deliver a multicast or broadcast datagram to *all* sockets
722: * for which the local and remote addresses and ports match
723: * those of the incoming datagram. This allows more than
724: * one process to receive multi/broadcasts on the same port.
725: * (This really ought to be done for unicast datagrams as
726: * well, but that would cause problems with existing
727: * applications that open both address-specific sockets and
728: * a wildcard socket listening to the same port -- they would
729: * end up receiving duplicates of every unicast datagram.
730: * Those applications open the multiple sockets to overcome an
731: * inadequacy of the UDP socket interface, but for backwards
732: * compatibility we avoid the problem here rather than
733: * fixing the interface. Maybe 4.5BSD will remedy this?)
734: */
735:
736: /*
1.92 itojun 737: * KAME note: traditionally we dropped udpiphdr from mbuf here.
1.71 itojun 738: * we need udpiphdr for IPsec processing so we do that later.
1.53 itojun 739: */
740: /*
741: * Locate pcb(s) for datagram.
742: */
1.109 itojun 743: CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
744: inp = (struct inpcb *)inph;
745: if (inp->inp_af != AF_INET)
746: continue;
747:
1.53 itojun 748: if (inp->inp_lport != *dport)
749: continue;
750: if (!in_nullhost(inp->inp_laddr)) {
751: if (!in_hosteq(inp->inp_laddr, *dst4))
752: continue;
753: }
754: if (!in_nullhost(inp->inp_faddr)) {
755: if (!in_hosteq(inp->inp_faddr, *src4) ||
756: inp->inp_fport != *sport)
757: continue;
758: }
759:
760: udp4_sendup(m, off, (struct sockaddr *)src,
761: inp->inp_socket);
762: rcvcnt++;
763:
764: /*
765: * Don't look for additional matches if this one does
766: * not have either the SO_REUSEPORT or SO_REUSEADDR
767: * socket options set. This heuristic avoids searching
768: * through all pcbs in the common case of a non-shared
769: * port. It assumes that an application will never
770: * clear these options after setting them.
771: */
772: if ((inp->inp_socket->so_options &
773: (SO_REUSEPORT|SO_REUSEADDR)) == 0)
774: break;
775: }
776: } else {
777: /*
778: * Locate pcb for datagram.
779: */
780: inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
781: if (inp == 0) {
1.164 ! thorpej 782: udpstat[UDP_STAT_PCBHASHMISS]++;
1.53 itojun 783: inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
1.82 itojun 784: if (inp == 0)
1.53 itojun 785: return rcvcnt;
786: }
787:
1.130 manu 788: #ifdef IPSEC_NAT_T
789: /* Handle ESP over UDP */
790: if (inp->inp_flags & INP_ESPINUDP_ALL) {
791: struct sockaddr *sa = (struct sockaddr *)src;
792:
1.144 manu 793: switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
794: case -1: /* Error, m was freeed */
795: rcvcnt = -1;
796: goto bad;
797: break;
798:
799: case 1: /* ESP over UDP */
1.130 manu 800: rcvcnt++;
801: goto bad;
1.144 manu 802: break;
803:
804: case 0: /* plain UDP */
805: default: /* Unexpected */
806: /*
807: * Normal UDP processing will take place
808: * m may have changed.
809: */
810: m = *mp;
811: break;
1.130 manu 812: }
813: }
814: #endif
815:
1.53 itojun 816: udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
817: rcvcnt++;
818: }
819:
820: bad:
821: return rcvcnt;
822: }
1.72 itojun 823: #endif
1.53 itojun 824:
825: #ifdef INET6
826: static int
1.119 matt 827: udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
828: struct mbuf *m, int off)
1.53 itojun 829: {
1.79 itojun 830: u_int16_t sport, dport;
1.53 itojun 831: int rcvcnt;
1.146 rpaulo 832: struct in6_addr src6, *dst6;
1.79 itojun 833: const struct in_addr *dst4;
1.109 itojun 834: struct inpcb_hdr *inph;
1.53 itojun 835: struct in6pcb *in6p;
836:
837: rcvcnt = 0;
838: off += sizeof(struct udphdr); /* now, offset of payload */
839:
840: if (af != AF_INET && af != AF_INET6)
841: goto bad;
842: if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
843: goto bad;
844:
1.146 rpaulo 845: src6 = src->sin6_addr;
846: if (sa6_recoverscope(src) != 0) {
847: /* XXX: should be impossible. */
848: goto bad;
849: }
1.79 itojun 850: sport = src->sin6_port;
1.146 rpaulo 851:
1.79 itojun 852: dport = dst->sin6_port;
1.88 itojun 853: dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
1.146 rpaulo 854: dst6 = &dst->sin6_addr;
1.53 itojun 855:
1.146 rpaulo 856: if (IN6_IS_ADDR_MULTICAST(dst6) ||
1.79 itojun 857: (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
1.53 itojun 858: /*
859: * Deliver a multicast or broadcast datagram to *all* sockets
860: * for which the local and remote addresses and ports match
861: * those of the incoming datagram. This allows more than
862: * one process to receive multi/broadcasts on the same port.
863: * (This really ought to be done for unicast datagrams as
864: * well, but that would cause problems with existing
865: * applications that open both address-specific sockets and
866: * a wildcard socket listening to the same port -- they would
867: * end up receiving duplicates of every unicast datagram.
868: * Those applications open the multiple sockets to overcome an
869: * inadequacy of the UDP socket interface, but for backwards
870: * compatibility we avoid the problem here rather than
871: * fixing the interface. Maybe 4.5BSD will remedy this?)
872: */
873:
874: /*
1.92 itojun 875: * KAME note: traditionally we dropped udpiphdr from mbuf here.
1.71 itojun 876: * we need udpiphdr for IPsec processing so we do that later.
1.53 itojun 877: */
878: /*
879: * Locate pcb(s) for datagram.
880: */
1.109 itojun 881: CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
882: in6p = (struct in6pcb *)inph;
883: if (in6p->in6p_af != AF_INET6)
884: continue;
885:
1.79 itojun 886: if (in6p->in6p_lport != dport)
1.53 itojun 887: continue;
888: if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1.146 rpaulo 889: if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
890: dst6))
1.53 itojun 891: continue;
1.92 itojun 892: } else {
1.146 rpaulo 893: if (IN6_IS_ADDR_V4MAPPED(dst6) &&
1.85 itojun 894: (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1.55 itojun 895: continue;
896: }
1.53 itojun 897: if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1.79 itojun 898: if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
899: &src6) || in6p->in6p_fport != sport)
1.53 itojun 900: continue;
1.92 itojun 901: } else {
1.79 itojun 902: if (IN6_IS_ADDR_V4MAPPED(&src6) &&
1.85 itojun 903: (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1.55 itojun 904: continue;
905: }
1.53 itojun 906:
907: udp6_sendup(m, off, (struct sockaddr *)src,
908: in6p->in6p_socket);
909: rcvcnt++;
910:
911: /*
912: * Don't look for additional matches if this one does
913: * not have either the SO_REUSEPORT or SO_REUSEADDR
914: * socket options set. This heuristic avoids searching
915: * through all pcbs in the common case of a non-shared
916: * port. It assumes that an application will never
917: * clear these options after setting them.
918: */
919: if ((in6p->in6p_socket->so_options &
920: (SO_REUSEPORT|SO_REUSEADDR)) == 0)
921: break;
922: }
923: } else {
924: /*
925: * Locate pcb for datagram.
926: */
1.146 rpaulo 927: in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
928: dport, 0);
1.53 itojun 929: if (in6p == 0) {
1.164 ! thorpej 930: udpstat[UDP_STAT_PCBHASHMISS]++;
1.146 rpaulo 931: in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
1.82 itojun 932: if (in6p == 0)
1.53 itojun 933: return rcvcnt;
934: }
935:
936: udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
937: rcvcnt++;
938: }
939:
940: bad:
941: return rcvcnt;
942: }
943: #endif
1.1 cgd 944:
1.72 itojun 945: #ifdef INET
1.1 cgd 946: /*
947: * Notify a udp user of an asynchronous error;
948: * just wake up so that he can collect error status.
949: */
1.7 mycroft 950: static void
1.119 matt 951: udp_notify(struct inpcb *inp, int errno)
1.1 cgd 952: {
953: inp->inp_socket->so_error = errno;
954: sorwakeup(inp->inp_socket);
955: sowwakeup(inp->inp_socket);
956: }
957:
1.27 christos 958: void *
1.157 dyoung 959: udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
1.1 cgd 960: {
1.66 augustss 961: struct ip *ip = v;
962: struct udphdr *uh;
1.119 matt 963: void (*notify)(struct inpcb *, int) = udp_notify;
1.21 mycroft 964: int errno;
1.1 cgd 965:
1.53 itojun 966: if (sa->sa_family != AF_INET
967: || sa->sa_len != sizeof(struct sockaddr_in))
1.51 itojun 968: return NULL;
1.20 mycroft 969: if ((unsigned)cmd >= PRC_NCMDS)
1.27 christos 970: return NULL;
1.20 mycroft 971: errno = inetctlerrmap[cmd];
1.18 mycroft 972: if (PRC_IS_REDIRECT(cmd))
1.19 mycroft 973: notify = in_rtchange, ip = 0;
1.18 mycroft 974: else if (cmd == PRC_HOSTDEAD)
1.19 mycroft 975: ip = 0;
1.23 cgd 976: else if (errno == 0)
1.27 christos 977: return NULL;
1.19 mycroft 978: if (ip) {
1.158 christos 979: uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
1.157 dyoung 980: in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
1.34 mycroft 981: ip->ip_src, uh->uh_sport, errno, notify);
1.53 itojun 982:
983: /* XXX mapped address case */
1.19 mycroft 984: } else
1.157 dyoung 985: in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
1.34 mycroft 986: notify);
1.27 christos 987: return NULL;
1.1 cgd 988: }
989:
1.7 mycroft 990: int
1.152 yamt 991: udp_ctloutput(int op, struct socket *so, int level, int optname,
992: struct mbuf **mp)
1.130 manu 993: {
994: int s;
995: int error = 0;
996: struct mbuf *m;
997: struct inpcb *inp;
998: int family;
999:
1000: family = so->so_proto->pr_domain->dom_family;
1001:
1002: s = splsoftnet();
1003: switch (family) {
1004: #ifdef INET
1005: case PF_INET:
1006: if (level != IPPROTO_UDP) {
1007: error = ip_ctloutput(op, so, level, optname, mp);
1008: goto end;
1009: }
1010: break;
1011: #endif
1012: #ifdef INET6
1013: case PF_INET6:
1014: if (level != IPPROTO_UDP) {
1015: error = ip6_ctloutput(op, so, level, optname, mp);
1016: goto end;
1017: }
1018: break;
1019: #endif
1020: default:
1021: error = EAFNOSUPPORT;
1022: goto end;
1023: }
1024:
1025:
1026: switch (op) {
1027: case PRCO_SETOPT:
1028: m = *mp;
1029: inp = sotoinpcb(so);
1030:
1031: switch (optname) {
1032: case UDP_ENCAP:
1.163 christos 1033: if (m == NULL || m->m_len != sizeof(int)) {
1.130 manu 1034: error = EINVAL;
1.153 yamt 1035: break;
1.130 manu 1036: }
1.131 perry 1037:
1.130 manu 1038: switch(*mtod(m, int *)) {
1039: #ifdef IPSEC_NAT_T
1040: case 0:
1041: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1042: break;
1043:
1044: case UDP_ENCAP_ESPINUDP:
1045: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1046: inp->inp_flags |= INP_ESPINUDP;
1047: break;
1.131 perry 1048:
1.130 manu 1049: case UDP_ENCAP_ESPINUDP_NON_IKE:
1050: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1051: inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1052: break;
1053: #endif
1054: default:
1055: error = EINVAL;
1056: break;
1057: }
1058: break;
1059:
1060: default:
1061: error = ENOPROTOOPT;
1062: break;
1063: }
1.155 yamt 1064: if (m != NULL) {
1065: m_free(m);
1066: }
1.130 manu 1067: break;
1068:
1069: default:
1070: error = EINVAL;
1071: break;
1.131 perry 1072: }
1073:
1.130 manu 1074: end:
1075: splx(s);
1076: return error;
1077: }
1.131 perry 1078:
1.130 manu 1079:
1080: int
1.27 christos 1081: udp_output(struct mbuf *m, ...)
1082: {
1.66 augustss 1083: struct inpcb *inp;
1084: struct udpiphdr *ui;
1.125 thorpej 1085: struct route *ro;
1.66 augustss 1086: int len = m->m_pkthdr.len;
1.31 mycroft 1087: int error = 0;
1.27 christos 1088: va_list ap;
1089:
1.98 matt 1090: MCLAIM(m, &udp_tx_mowner);
1.27 christos 1091: va_start(ap, m);
1092: inp = va_arg(ap, struct inpcb *);
1093: va_end(ap);
1.1 cgd 1094:
1095: /*
1096: * Calculate data length and get a mbuf
1097: * for UDP and IP headers.
1098: */
1.13 mycroft 1099: M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1100: if (m == 0) {
1101: error = ENOBUFS;
1.39 thorpej 1102: goto release;
1103: }
1104:
1105: /*
1106: * Compute the packet length of the IP header, and
1107: * punt if the length looks bogus.
1108: */
1.96 itojun 1109: if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1.39 thorpej 1110: error = EMSGSIZE;
1.13 mycroft 1111: goto release;
1112: }
1.1 cgd 1113:
1114: /*
1115: * Fill in mbuf with extended UDP header
1116: * and addresses and length put into network format.
1117: */
1118: ui = mtod(m, struct udpiphdr *);
1119: ui->ui_pr = IPPROTO_UDP;
1120: ui->ui_src = inp->inp_laddr;
1121: ui->ui_dst = inp->inp_faddr;
1122: ui->ui_sport = inp->inp_lport;
1123: ui->ui_dport = inp->inp_fport;
1.78 thorpej 1124: ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1.1 cgd 1125:
1.125 thorpej 1126: ro = &inp->inp_route;
1127:
1.1 cgd 1128: /*
1.78 thorpej 1129: * Set up checksum and output datagram.
1.1 cgd 1130: */
1131: if (udpcksum) {
1.78 thorpej 1132: /*
1133: * XXX Cache pseudo-header checksum part for
1134: * XXX "connected" UDP sockets.
1135: */
1136: ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1137: ui->ui_dst.s_addr, htons((u_int16_t)len +
1138: sizeof(struct udphdr) + IPPROTO_UDP));
1.135 yamt 1139: m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1.78 thorpej 1140: m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1141: } else
1142: ui->ui_sum = 0;
1.96 itojun 1143: ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1.1 cgd 1144: ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */
1145: ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */
1.164 ! thorpej 1146: udpstat[UDP_STAT_OPACKETS]++;
1.48 itojun 1147:
1.125 thorpej 1148: return (ip_output(m, inp->inp_options, ro,
1.12 mycroft 1149: inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1.107 itojun 1150: inp->inp_moptions, inp->inp_socket));
1.1 cgd 1151:
1152: release:
1153: m_freem(m);
1154: return (error);
1155: }
1156:
1.42 thorpej 1157: int udp_sendspace = 9216; /* really max datagram size */
1158: int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1.1 cgd 1159: /* 40 1K datagrams */
1160:
1161: /*ARGSUSED*/
1.7 mycroft 1162: int
1.119 matt 1163: udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1.145 christos 1164: struct mbuf *control, struct lwp *l)
1.1 cgd 1165: {
1.66 augustss 1166: struct inpcb *inp;
1.1 cgd 1167: int s;
1.66 augustss 1168: int error = 0;
1.1 cgd 1169:
1170: if (req == PRU_CONTROL)
1.158 christos 1171: return (in_control(so, (long)m, (void *)nam,
1.148 ad 1172: (struct ifnet *)control, l));
1.59 thorpej 1173:
1.149 tls 1174: s = splsoftnet();
1175:
1.60 thorpej 1176: if (req == PRU_PURGEIF) {
1.80 itojun 1177: in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1.60 thorpej 1178: in_purgeif((struct ifnet *)control);
1179: in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1.149 tls 1180: splx(s);
1.59 thorpej 1181: return (0);
1182: }
1.31 mycroft 1183:
1184: inp = sotoinpcb(so);
1.32 mycroft 1185: #ifdef DIAGNOSTIC
1186: if (req != PRU_SEND && req != PRU_SENDOOB && control)
1187: panic("udp_usrreq: unexpected control mbuf");
1188: #endif
1.31 mycroft 1189: if (inp == 0 && req != PRU_ATTACH) {
1190: error = EINVAL;
1191: goto release;
1192: }
1193:
1.1 cgd 1194: /*
1195: * Note: need to block udp_input while changing
1196: * the udp pcb queue and/or pcb addresses.
1197: */
1198: switch (req) {
1199:
1200: case PRU_ATTACH:
1.31 mycroft 1201: if (inp != 0) {
1202: error = EISCONN;
1.1 cgd 1203: break;
1204: }
1.98 matt 1205: #ifdef MBUFTRACE
1206: so->so_mowner = &udp_mowner;
1207: so->so_rcv.sb_mowner = &udp_rx_mowner;
1208: so->so_snd.sb_mowner = &udp_tx_mowner;
1209: #endif
1.31 mycroft 1210: if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1211: error = soreserve(so, udp_sendspace, udp_recvspace);
1212: if (error)
1213: break;
1214: }
1.18 mycroft 1215: error = in_pcballoc(so, &udbtable);
1.1 cgd 1216: if (error)
1217: break;
1.31 mycroft 1218: inp = sotoinpcb(so);
1219: inp->inp_ip.ip_ttl = ip_defttl;
1.1 cgd 1220: break;
1221:
1222: case PRU_DETACH:
1.31 mycroft 1223: in_pcbdetach(inp);
1.1 cgd 1224: break;
1225:
1226: case PRU_BIND:
1.148 ad 1227: error = in_pcbbind(inp, nam, l);
1.1 cgd 1228: break;
1229:
1230: case PRU_LISTEN:
1231: error = EOPNOTSUPP;
1232: break;
1233:
1234: case PRU_CONNECT:
1.148 ad 1235: error = in_pcbconnect(inp, nam, l);
1.31 mycroft 1236: if (error)
1.1 cgd 1237: break;
1.31 mycroft 1238: soisconnected(so);
1.1 cgd 1239: break;
1240:
1241: case PRU_CONNECT2:
1242: error = EOPNOTSUPP;
1243: break;
1244:
1245: case PRU_DISCONNECT:
1.31 mycroft 1246: /*soisdisconnected(so);*/
1247: so->so_state &= ~SS_ISCONNECTED; /* XXX */
1.1 cgd 1248: in_pcbdisconnect(inp);
1.34 mycroft 1249: inp->inp_laddr = zeroin_addr; /* XXX */
1.35 mycroft 1250: in_pcbstate(inp, INP_BOUND); /* XXX */
1.1 cgd 1251: break;
1252:
1253: case PRU_SHUTDOWN:
1254: socantsendmore(so);
1255: break;
1256:
1.31 mycroft 1257: case PRU_RCVD:
1258: error = EOPNOTSUPP;
1.1 cgd 1259: break;
1260:
1.31 mycroft 1261: case PRU_SEND:
1.32 mycroft 1262: if (control && control->m_len) {
1263: m_freem(control);
1264: m_freem(m);
1265: error = EINVAL;
1266: break;
1267: }
1.31 mycroft 1268: {
1.35 mycroft 1269: struct in_addr laddr; /* XXX */
1.1 cgd 1270:
1.31 mycroft 1271: if (nam) {
1.35 mycroft 1272: laddr = inp->inp_laddr; /* XXX */
1.31 mycroft 1273: if ((so->so_state & SS_ISCONNECTED) != 0) {
1274: error = EISCONN;
1.32 mycroft 1275: goto die;
1.31 mycroft 1276: }
1.148 ad 1277: error = in_pcbconnect(inp, nam, l);
1.100 matt 1278: if (error)
1279: goto die;
1.31 mycroft 1280: } else {
1281: if ((so->so_state & SS_ISCONNECTED) == 0) {
1282: error = ENOTCONN;
1.32 mycroft 1283: goto die;
1.31 mycroft 1284: }
1285: }
1.33 mycroft 1286: error = udp_output(m, inp);
1.100 matt 1287: m = NULL;
1.31 mycroft 1288: if (nam) {
1289: in_pcbdisconnect(inp);
1.35 mycroft 1290: inp->inp_laddr = laddr; /* XXX */
1291: in_pcbstate(inp, INP_BOUND); /* XXX */
1.31 mycroft 1292: }
1.100 matt 1293: die:
1294: if (m)
1295: m_freem(m);
1.31 mycroft 1296: }
1.1 cgd 1297: break;
1298:
1299: case PRU_SENSE:
1300: /*
1301: * stat: don't bother with a blocksize.
1302: */
1.31 mycroft 1303: splx(s);
1.1 cgd 1304: return (0);
1305:
1.31 mycroft 1306: case PRU_RCVOOB:
1307: error = EOPNOTSUPP;
1308: break;
1309:
1.1 cgd 1310: case PRU_SENDOOB:
1.32 mycroft 1311: m_freem(control);
1.31 mycroft 1312: m_freem(m);
1.1 cgd 1313: error = EOPNOTSUPP;
1314: break;
1315:
1.31 mycroft 1316: case PRU_SOCKADDR:
1317: in_setsockaddr(inp, nam);
1318: break;
1319:
1320: case PRU_PEERADDR:
1321: in_setpeeraddr(inp, nam);
1322: break;
1.1 cgd 1323:
1324: default:
1325: panic("udp_usrreq");
1326: }
1327:
1328: release:
1.31 mycroft 1329: splx(s);
1.1 cgd 1330: return (error);
1.13 mycroft 1331: }
1332:
1333: /*
1334: * Sysctl for udp variables.
1335: */
1.114 atatat 1336: SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1.13 mycroft 1337: {
1.114 atatat 1338:
1.116 atatat 1339: sysctl_createv(clog, 0, NULL, NULL,
1340: CTLFLAG_PERMANENT,
1.114 atatat 1341: CTLTYPE_NODE, "net", NULL,
1342: NULL, 0, NULL, 0,
1343: CTL_NET, CTL_EOL);
1.116 atatat 1344: sysctl_createv(clog, 0, NULL, NULL,
1345: CTLFLAG_PERMANENT,
1.114 atatat 1346: CTLTYPE_NODE, "inet", NULL,
1347: NULL, 0, NULL, 0,
1348: CTL_NET, PF_INET, CTL_EOL);
1.116 atatat 1349: sysctl_createv(clog, 0, NULL, NULL,
1350: CTLFLAG_PERMANENT,
1.122 atatat 1351: CTLTYPE_NODE, "udp",
1352: SYSCTL_DESCR("UDPv4 related settings"),
1.114 atatat 1353: NULL, 0, NULL, 0,
1354: CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1355:
1.116 atatat 1356: sysctl_createv(clog, 0, NULL, NULL,
1357: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1358: CTLTYPE_INT, "checksum",
1.123 heas 1359: SYSCTL_DESCR("Compute UDP checksums"),
1.114 atatat 1360: NULL, 0, &udpcksum, 0,
1361: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1362: CTL_EOL);
1.116 atatat 1363: sysctl_createv(clog, 0, NULL, NULL,
1364: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1365: CTLTYPE_INT, "sendspace",
1366: SYSCTL_DESCR("Default UDP send buffer size"),
1.114 atatat 1367: NULL, 0, &udp_sendspace, 0,
1368: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1369: CTL_EOL);
1.116 atatat 1370: sysctl_createv(clog, 0, NULL, NULL,
1371: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1372: CTLTYPE_INT, "recvspace",
1373: SYSCTL_DESCR("Default UDP receive buffer size"),
1.114 atatat 1374: NULL, 0, &udp_recvspace, 0,
1375: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1376: CTL_EOL);
1.125 thorpej 1377: sysctl_createv(clog, 0, NULL, NULL,
1378: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1379: CTLTYPE_INT, "do_loopback_cksum",
1380: SYSCTL_DESCR("Perform UDP checksum on loopback"),
1381: NULL, 0, &udp_do_loopback_cksum, 0,
1382: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1383: CTL_EOL);
1.132 atatat 1384: sysctl_createv(clog, 0, NULL, NULL,
1385: CTLFLAG_PERMANENT,
1.134 atatat 1386: CTLTYPE_STRUCT, "pcblist",
1.132 atatat 1387: SYSCTL_DESCR("UDP protocol control block list"),
1388: sysctl_inpcblist, 0, &udbtable, 0,
1389: CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1390: CTL_EOL);
1.139 elad 1391: sysctl_createv(clog, 0, NULL, NULL,
1392: CTLFLAG_PERMANENT,
1393: CTLTYPE_STRUCT, "stats",
1394: SYSCTL_DESCR("UDP statistics"),
1.164 ! thorpej 1395: NULL, 0, udpstat, sizeof(udpstat),
1.139 elad 1396: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1397: CTL_EOL);
1.1 cgd 1398: }
1.72 itojun 1399: #endif
1.130 manu 1400:
1401: #if (defined INET && defined IPSEC_NAT_T)
1402: /*
1403: * Returns:
1404: * 1 if the packet was processed
1405: * 0 if normal UDP processing should take place
1.144 manu 1406: * -1 if an error occurent and m was freed
1.130 manu 1407: */
1408: static int
1.154 yamt 1409: udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1.151 christos 1410: struct socket *so)
1.130 manu 1411: {
1412: size_t len;
1.158 christos 1413: void *data;
1.130 manu 1414: struct inpcb *inp;
1415: size_t skip = 0;
1416: size_t minlen;
1417: size_t iphdrlen;
1418: struct ip *ip;
1419: struct mbuf *n;
1.136 manu 1420: struct m_tag *tag;
1421: struct udphdr *udphdr;
1422: u_int16_t sport, dport;
1.144 manu 1423: struct mbuf *m = *mp;
1.130 manu 1424:
1.131 perry 1425: /*
1.130 manu 1426: * Collapse the mbuf chain if the first mbuf is too short
1427: * The longest case is: UDP + non ESP marker + ESP
1428: */
1429: minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1430: if (minlen > m->m_pkthdr.len)
1431: minlen = m->m_pkthdr.len;
1432:
1433: if (m->m_len < minlen) {
1.144 manu 1434: if ((*mp = m_pullup(m, minlen)) == NULL) {
1.130 manu 1435: printf("udp4_espinudp: m_pullup failed\n");
1.144 manu 1436: return -1;
1.130 manu 1437: }
1.144 manu 1438: m = *mp;
1.130 manu 1439: }
1440:
1.131 perry 1441: len = m->m_len - off;
1.158 christos 1442: data = mtod(m, char *) + off;
1.130 manu 1443: inp = sotoinpcb(so);
1444:
1445: /* Ignore keepalive packets */
1.158 christos 1446: if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1.130 manu 1447: return 1;
1448: }
1449:
1.131 perry 1450: /*
1451: * Check that the payload is long enough to hold
1.130 manu 1452: * an ESP header and compute the length of encapsulation
1.131 perry 1453: * header to remove
1.130 manu 1454: */
1455: if (inp->inp_flags & INP_ESPINUDP) {
1456: u_int32_t *st = (u_int32_t *)data;
1457:
1458: if ((len <= sizeof(struct esp)) || (*st == 0))
1459: return 0; /* Normal UDP processing */
1460:
1461: skip = sizeof(struct udphdr);
1462: }
1463:
1464: if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1.142 kleink 1465: u_int32_t *st = (u_int32_t *)data;
1.130 manu 1466:
1467: if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1.142 kleink 1468: || ((st[0] | st[1]) != 0))
1.130 manu 1469: return 0; /* Normal UDP processing */
1.131 perry 1470:
1.130 manu 1471: skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1472: }
1473:
1474: /*
1.136 manu 1475: * Get the UDP ports. They are handled in network
1476: * order everywhere in IPSEC_NAT_T code.
1477: */
1.158 christos 1478: udphdr = (struct udphdr *)((char *)data - skip);
1.136 manu 1479: sport = udphdr->uh_sport;
1480: dport = udphdr->uh_dport;
1481:
1482: /*
1.130 manu 1483: * Remove the UDP header (and possibly the non ESP marker)
1484: * IP header lendth is iphdrlen
1.131 perry 1485: * Before:
1.130 manu 1486: * <--- off --->
1487: * +----+------+-----+
1488: * | IP | UDP | ESP |
1489: * +----+------+-----+
1490: * <-skip->
1491: * After:
1492: * +----+-----+
1493: * | IP | ESP |
1494: * +----+-----+
1495: * <-skip->
1496: */
1497: iphdrlen = off - sizeof(struct udphdr);
1.158 christos 1498: memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1.130 manu 1499: m_adj(m, skip);
1500:
1501: ip = mtod(m, struct ip *);
1502: ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1503: ip->ip_p = IPPROTO_ESP;
1504:
1505: /*
1.131 perry 1506: * Copy the mbuf to avoid multiple free, as both
1507: * esp4_input (which we call) and udp_input (which
1.130 manu 1508: * called us) free the mbuf.
1509: */
1510: if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1511: printf("udp4_espinudp: m_dup failed\n");
1512: return 0;
1513: }
1514:
1.136 manu 1515: /*
1516: * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1517: * the source UDP port. This is required if we want
1518: * to select the right SPD for multiple hosts behind
1519: * same NAT
1520: */
1.137 manu 1521: if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1522: sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1523: printf("udp4_espinudp: m_tag_get failed\n");
1.138 manu 1524: m_freem(n);
1.137 manu 1525: return 0;
1526: }
1.136 manu 1527: ((u_int16_t *)(tag + 1))[0] = sport;
1528: ((u_int16_t *)(tag + 1))[1] = dport;
1529: m_tag_prepend(n, tag);
1530:
1.147 christos 1531: #ifdef FAST_IPSEC
1.160 degroote 1532: ipsec4_common_input(n, iphdrlen, IPPROTO_ESP);
1.147 christos 1533: #else
1.130 manu 1534: esp4_input(n, iphdrlen);
1.147 christos 1535: #endif
1.130 manu 1536:
1537: /* We handled it, it shoudln't be handled by UDP */
1538: return 1;
1539: }
1540: #endif
CVSweb <webmaster@jp.NetBSD.org>