Annotation of src/sys/netinet/udp_usrreq.c, Revision 1.167
1.167 ! thorpej 1: /* $NetBSD: udp_usrreq.c,v 1.166 2008/04/12 05:58:22 thorpej 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>
1.167 ! thorpej 64: __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.166 2008/04/12 05:58:22 thorpej 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>
1.166 thorpej 97: #include <netinet/udp_private.h>
1.1 cgd 98:
1.53 itojun 99: #ifdef INET6
100: #include <netinet/ip6.h>
101: #include <netinet/icmp6.h>
102: #include <netinet6/ip6_var.h>
1.167 ! thorpej 103: #include <netinet6/ip6_private.h>
1.53 itojun 104: #include <netinet6/in6_pcb.h>
105: #include <netinet6/udp6_var.h>
1.146 rpaulo 106: #include <netinet6/scope6_var.h>
1.53 itojun 107: #endif
108:
109: #ifndef INET6
110: /* always need ip6.h for IP6_EXTHDR_GET */
111: #include <netinet/ip6.h>
112: #endif
113:
1.76 itojun 114: #include "faith.h"
115: #if defined(NFAITH) && NFAITH > 0
116: #include <net/if_faith.h>
117: #endif
118:
1.27 christos 119: #include <machine/stdarg.h>
120:
1.105 jonathan 121: #ifdef FAST_IPSEC
122: #include <netipsec/ipsec.h>
1.121 jonathan 123: #include <netipsec/ipsec_var.h> /* XXX ipsecstat namespace */
1.147 christos 124: #include <netipsec/esp.h>
1.105 jonathan 125: #ifdef INET6
126: #include <netipsec/ipsec6.h>
127: #endif
128: #endif /* FAST_IPSEC*/
129:
1.48 itojun 130: #ifdef IPSEC
131: #include <netinet6/ipsec.h>
1.147 christos 132: #include <netinet6/esp.h>
1.48 itojun 133: #include <netkey/key.h>
134: #endif /*IPSEC*/
135:
1.64 ws 136: #ifdef IPKDB
137: #include <ipkdb/ipkdb.h>
138: #endif
139:
1.8 mycroft 140: /*
141: * UDP protocol implementation.
142: * Per RFC 768, August, 1980.
143: */
144: int udpcksum = 1;
1.141 yamt 145: int udp_do_loopback_cksum = 0;
1.93 matt 146:
147: struct inpcbtable udbtable;
1.164 thorpej 148:
1.166 thorpej 149: percpu_t *udpstat_percpu;
1.8 mycroft 150:
1.72 itojun 151: #ifdef INET
1.130 manu 152: #ifdef IPSEC_NAT_T
1.144 manu 153: static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
1.130 manu 154: struct socket *);
155: #endif
1.119 matt 156: static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
157: struct socket *);
158: static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
1.144 manu 159: struct mbuf **, int);
1.129 yamt 160: static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
1.72 itojun 161: #endif
1.53 itojun 162: #ifdef INET6
1.119 matt 163: static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
164: struct socket *);
165: static int udp6_realinput (int, struct sockaddr_in6 *,
166: struct sockaddr_in6 *, struct mbuf *, int);
1.129 yamt 167: static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
1.53 itojun 168: #endif
1.72 itojun 169: #ifdef INET
1.119 matt 170: static void udp_notify (struct inpcb *, int);
1.72 itojun 171: #endif
1.7 mycroft 172:
1.26 mycroft 173: #ifndef UDBHASHSIZE
174: #define UDBHASHSIZE 128
175: #endif
176: int udbhashsize = UDBHASHSIZE;
177:
1.98 matt 178: #ifdef MBUFTRACE
1.150 dogcow 179: struct mowner udp_mowner = MOWNER_INIT("udp", "");
180: struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
181: struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
1.98 matt 182: #endif
183:
1.78 thorpej 184: #ifdef UDP_CSUM_COUNTERS
185: #include <sys/device.h>
186:
1.140 yamt 187: #if defined(INET)
1.78 thorpej 188: struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
189: NULL, "udp", "hwcsum bad");
190: struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
191: NULL, "udp", "hwcsum ok");
192: struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
193: NULL, "udp", "hwcsum data");
194: struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
195: NULL, "udp", "swcsum");
196:
1.120 matt 197: EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
198: EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
199: EVCNT_ATTACH_STATIC(udp_hwcsum_data);
200: EVCNT_ATTACH_STATIC(udp_swcsum);
1.140 yamt 201: #endif /* defined(INET) */
202:
203: #if defined(INET6)
204: struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
205: NULL, "udp6", "hwcsum bad");
206: struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
207: NULL, "udp6", "hwcsum ok");
208: struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
209: NULL, "udp6", "hwcsum data");
210: struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
211: NULL, "udp6", "swcsum");
212:
213: EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
214: EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
215: EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
216: EVCNT_ATTACH_STATIC(udp6_swcsum);
217: #endif /* defined(INET6) */
218:
219: #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
1.120 matt 220:
1.78 thorpej 221: #else
222:
223: #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */
224:
225: #endif /* UDP_CSUM_COUNTERS */
226:
1.7 mycroft 227: void
1.119 matt 228: udp_init(void)
1.1 cgd 229: {
1.18 mycroft 230:
1.35 mycroft 231: in_pcbinit(&udbtable, udbhashsize, udbhashsize);
1.78 thorpej 232:
1.98 matt 233: MOWNER_ATTACH(&udp_tx_mowner);
234: MOWNER_ATTACH(&udp_rx_mowner);
235: MOWNER_ATTACH(&udp_mowner);
1.166 thorpej 236:
237: udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
1.1 cgd 238: }
239:
1.129 yamt 240: /*
241: * Checksum extended UDP header and data.
242: */
243:
244: int
245: udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
246: int iphlen, int len)
247: {
248:
249: switch (af) {
1.72 itojun 250: #ifdef INET
1.129 yamt 251: case AF_INET:
252: return udp4_input_checksum(m, uh, iphlen, len);
253: #endif
254: #ifdef INET6
255: case AF_INET6:
256: return udp6_input_checksum(m, uh, iphlen, len);
257: #endif
258: }
259: #ifdef DIAGNOSTIC
260: panic("udp_input_checksum: unknown af %d", af);
261: #endif
262: /* NOTREACHED */
263: return -1;
264: }
265:
266: #ifdef INET
267:
268: /*
269: * Checksum extended UDP header and data.
270: */
271:
272: static int
273: udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
274: int iphlen, int len)
275: {
276:
277: /*
278: * XXX it's better to record and check if this mbuf is
279: * already checked.
280: */
281:
282: if (uh->uh_sum == 0)
283: return 0;
284:
285: switch (m->m_pkthdr.csum_flags &
286: ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
287: M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
288: case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
289: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
290: goto badcsum;
291:
292: case M_CSUM_UDPv4|M_CSUM_DATA: {
293: u_int32_t hw_csum = m->m_pkthdr.csum_data;
294:
295: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
296: if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
297: const struct ip *ip =
298: mtod(m, const struct ip *);
299:
300: hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
301: ip->ip_dst.s_addr,
302: htons(hw_csum + len + IPPROTO_UDP));
303: }
304: if ((hw_csum ^ 0xffff) != 0)
305: goto badcsum;
306: break;
307: }
308:
309: case M_CSUM_UDPv4:
310: /* Checksum was okay. */
311: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
312: break;
313:
314: default:
315: /*
316: * Need to compute it ourselves. Maybe skip checksum
317: * on loopback interfaces.
318: */
319: if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
320: IFF_LOOPBACK) ||
321: udp_do_loopback_cksum)) {
322: UDP_CSUM_COUNTER_INCR(&udp_swcsum);
323: if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
324: goto badcsum;
325: }
326: break;
327: }
328:
329: return 0;
330:
331: badcsum:
1.166 thorpej 332: UDP_STATINC(UDP_STAT_BADSUM);
1.129 yamt 333: return -1;
334: }
335:
1.7 mycroft 336: void
1.27 christos 337: udp_input(struct mbuf *m, ...)
1.1 cgd 338: {
1.53 itojun 339: va_list ap;
340: struct sockaddr_in src, dst;
341: struct ip *ip;
342: struct udphdr *uh;
1.97 simonb 343: int iphlen;
1.53 itojun 344: int len;
345: int n;
1.96 itojun 346: u_int16_t ip_len;
1.53 itojun 347:
348: va_start(ap, m);
349: iphlen = va_arg(ap, int);
1.97 simonb 350: (void)va_arg(ap, int); /* ignore value, advance ap */
1.53 itojun 351: va_end(ap);
352:
1.98 matt 353: MCLAIM(m, &udp_rx_mowner);
1.166 thorpej 354: UDP_STATINC(UDP_STAT_IPACKETS);
1.53 itojun 355:
356: /*
357: * Get IP and UDP header together in first mbuf.
358: */
359: ip = mtod(m, struct ip *);
360: IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
361: if (uh == NULL) {
1.166 thorpej 362: UDP_STATINC(UDP_STAT_HDROPS);
1.53 itojun 363: return;
364: }
1.95 thorpej 365: KASSERT(UDP_HDR_ALIGNED_P(uh));
1.53 itojun 366:
1.57 itojun 367: /* destination port of 0 is illegal, based on RFC768. */
368: if (uh->uh_dport == 0)
369: goto bad;
370:
1.53 itojun 371: /*
372: * Make mbuf data length reflect UDP length.
373: * If not enough data to reflect UDP length, drop.
374: */
1.96 itojun 375: ip_len = ntohs(ip->ip_len);
1.53 itojun 376: len = ntohs((u_int16_t)uh->uh_ulen);
1.96 itojun 377: if (ip_len != iphlen + len) {
378: if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
1.166 thorpej 379: UDP_STATINC(UDP_STAT_BADLEN);
1.53 itojun 380: goto bad;
381: }
1.96 itojun 382: m_adj(m, iphlen + len - ip_len);
1.53 itojun 383: }
384:
385: /*
386: * Checksum extended UDP header and data.
387: */
1.129 yamt 388: if (udp4_input_checksum(m, uh, iphlen, len))
389: goto badcsum;
1.53 itojun 390:
391: /* construct source and dst sockaddrs. */
1.159 dyoung 392: sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
393: sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
1.53 itojun 394:
1.144 manu 395: if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
1.166 thorpej 396: UDP_STATINC(UDP_STAT_HDROPS);
1.144 manu 397: return;
398: }
1.53 itojun 399: #ifdef INET6
400: if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
401: struct sockaddr_in6 src6, dst6;
402:
403: bzero(&src6, sizeof(src6));
404: src6.sin6_family = AF_INET6;
405: src6.sin6_len = sizeof(struct sockaddr_in6);
406: src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
407: bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
408: sizeof(ip->ip_src));
409: src6.sin6_port = uh->uh_sport;
410: bzero(&dst6, sizeof(dst6));
411: dst6.sin6_family = AF_INET6;
412: dst6.sin6_len = sizeof(struct sockaddr_in6);
413: dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
414: bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
415: sizeof(ip->ip_dst));
416: dst6.sin6_port = uh->uh_dport;
417:
418: n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
419: }
420: #endif
421:
422: if (n == 0) {
423: if (m->m_flags & (M_BCAST | M_MCAST)) {
1.166 thorpej 424: UDP_STATINC(UDP_STAT_NOPORTBCAST);
1.53 itojun 425: goto bad;
426: }
1.166 thorpej 427: UDP_STATINC(UDP_STAT_NOPORT);
1.64 ws 428: #ifdef IPKDB
1.53 itojun 429: if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
430: m, iphlen + sizeof(struct udphdr),
431: m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
432: /*
433: * It was a debugger connect packet,
434: * just drop it now
435: */
436: goto bad;
437: }
438: #endif
439: icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
440: m = NULL;
441: }
442:
443: bad:
444: if (m)
445: m_freem(m);
1.78 thorpej 446: return;
447:
448: badcsum:
449: m_freem(m);
1.53 itojun 450: }
1.72 itojun 451: #endif
1.53 itojun 452:
453: #ifdef INET6
1.129 yamt 454: static int
455: udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
456: {
457:
1.140 yamt 458: /*
459: * XXX it's better to record and check if this mbuf is
460: * already checked.
461: */
462:
1.129 yamt 463: if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
464: goto good;
465: }
466: if (uh->uh_sum == 0) {
467: udp6stat.udp6s_nosum++;
468: goto bad;
469: }
1.140 yamt 470:
471: switch (m->m_pkthdr.csum_flags &
472: ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
473: M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
474: case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
475: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
1.129 yamt 476: udp6stat.udp6s_badsum++;
477: goto bad;
1.140 yamt 478:
479: #if 0 /* notyet */
480: case M_CSUM_UDPv6|M_CSUM_DATA:
481: #endif
482:
483: case M_CSUM_UDPv6:
484: /* Checksum was okay. */
485: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
486: break;
487:
488: default:
489: /*
490: * Need to compute it ourselves. Maybe skip checksum
491: * on loopback interfaces.
492: */
493: UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
494: if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
495: udp6stat.udp6s_badsum++;
496: goto bad;
497: }
1.129 yamt 498: }
499:
500: good:
501: return 0;
502: bad:
503: return -1;
504: }
505:
1.53 itojun 506: int
1.154 yamt 507: udp6_input(struct mbuf **mp, int *offp, int proto)
1.53 itojun 508: {
509: struct mbuf *m = *mp;
510: int off = *offp;
511: struct sockaddr_in6 src, dst;
512: struct ip6_hdr *ip6;
513: struct udphdr *uh;
514: u_int32_t plen, ulen;
515:
1.76 itojun 516: ip6 = mtod(m, struct ip6_hdr *);
517:
1.53 itojun 518: #if defined(NFAITH) && 0 < NFAITH
1.76 itojun 519: if (faithprefix(&ip6->ip6_dst)) {
520: /* send icmp6 host unreach? */
521: m_freem(m);
522: return IPPROTO_DONE;
1.53 itojun 523: }
524: #endif
525:
526: udp6stat.udp6s_ipackets++;
527:
528: /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */
529: plen = m->m_pkthdr.len - off;
530: IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
531: if (uh == NULL) {
1.167 ! thorpej 532: IP6_STATINC(IP6_STAT_TOOSHORT);
1.53 itojun 533: return IPPROTO_DONE;
534: }
1.95 thorpej 535: KASSERT(UDP_HDR_ALIGNED_P(uh));
1.53 itojun 536: ulen = ntohs((u_short)uh->uh_ulen);
1.69 itojun 537: /*
538: * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
539: * iff payload length > 0xffff.
540: */
1.53 itojun 541: if (ulen == 0 && plen > 0xffff)
542: ulen = plen;
543:
544: if (plen != ulen) {
545: udp6stat.udp6s_badlen++;
1.54 itojun 546: goto bad;
547: }
548:
1.57 itojun 549: /* destination port of 0 is illegal, based on RFC768. */
550: if (uh->uh_dport == 0)
551: goto bad;
552:
1.54 itojun 553: /* Be proactive about malicious use of IPv4 mapped address */
554: if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
555: IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
556: /* XXX stat */
1.53 itojun 557: goto bad;
558: }
559:
560: /*
1.125 thorpej 561: * Checksum extended UDP header and data. Maybe skip checksum
562: * on loopback interfaces.
1.53 itojun 563: */
1.129 yamt 564: if (udp6_input_checksum(m, uh, off, ulen))
565: goto bad;
1.53 itojun 566:
567: /*
568: * Construct source and dst sockaddrs.
569: */
570: bzero(&src, sizeof(src));
571: src.sin6_family = AF_INET6;
572: src.sin6_len = sizeof(struct sockaddr_in6);
1.146 rpaulo 573: src.sin6_addr = ip6->ip6_src;
1.53 itojun 574: src.sin6_port = uh->uh_sport;
575: bzero(&dst, sizeof(dst));
576: dst.sin6_family = AF_INET6;
577: dst.sin6_len = sizeof(struct sockaddr_in6);
1.146 rpaulo 578: dst.sin6_addr = ip6->ip6_dst;
1.53 itojun 579: dst.sin6_port = uh->uh_dport;
580:
581: if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
582: if (m->m_flags & M_MCAST) {
583: udp6stat.udp6s_noportmcast++;
584: goto bad;
585: }
1.61 itojun 586: udp6stat.udp6s_noport++;
1.53 itojun 587: icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
588: m = NULL;
589: }
590:
591: bad:
592: if (m)
593: m_freem(m);
594: return IPPROTO_DONE;
595: }
596: #endif
597:
1.72 itojun 598: #ifdef INET
1.53 itojun 599: static void
1.119 matt 600: udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
601: struct sockaddr *src, struct socket *so)
1.53 itojun 602: {
603: struct mbuf *opts = NULL;
604: struct mbuf *n;
605: struct inpcb *inp = NULL;
606:
607: if (!so)
608: return;
609: switch (so->so_proto->pr_domain->dom_family) {
610: case AF_INET:
611: inp = sotoinpcb(so);
612: break;
613: #ifdef INET6
614: case AF_INET6:
615: break;
616: #endif
617: default:
618: return;
619: }
620:
1.105 jonathan 621: #if defined(IPSEC) || defined(FAST_IPSEC)
1.53 itojun 622: /* check AH/ESP integrity. */
623: if (so != NULL && ipsec4_in_reject_so(m, so)) {
624: ipsecstat.in_polvio++;
1.162 dyoung 625: if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
1.110 itojun 626: icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
627: 0, 0);
1.53 itojun 628: return;
629: }
630: #endif /*IPSEC*/
631:
1.162 dyoung 632: if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
1.53 itojun 633: if (inp && (inp->inp_flags & INP_CONTROLOPTS
634: || so->so_options & SO_TIMESTAMP)) {
635: struct ip *ip = mtod(n, struct ip *);
636: ip_savecontrol(inp, &opts, ip, n);
637: }
638:
639: m_adj(n, off);
640: if (sbappendaddr(&so->so_rcv, src, n,
641: opts) == 0) {
642: m_freem(n);
643: if (opts)
644: m_freem(opts);
1.124 darrenr 645: so->so_rcv.sb_overflowed++;
1.166 thorpej 646: UDP_STATINC(UDP_STAT_FULLSOCK);
1.53 itojun 647: } else
648: sorwakeup(so);
649: }
650: }
1.72 itojun 651: #endif
1.53 itojun 652:
653: #ifdef INET6
654: static void
1.119 matt 655: udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
656: struct sockaddr *src, struct socket *so)
1.53 itojun 657: {
658: struct mbuf *opts = NULL;
659: struct mbuf *n;
660: struct in6pcb *in6p = NULL;
661:
662: if (!so)
663: return;
664: if (so->so_proto->pr_domain->dom_family != AF_INET6)
665: return;
666: in6p = sotoin6pcb(so);
667:
1.105 jonathan 668: #if defined(IPSEC) || defined(FAST_IPSEC)
1.53 itojun 669: /* check AH/ESP integrity. */
670: if (so != NULL && ipsec6_in_reject_so(m, so)) {
671: ipsec6stat.in_polvio++;
1.162 dyoung 672: if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
1.110 itojun 673: icmp6_error(n, ICMP6_DST_UNREACH,
674: ICMP6_DST_UNREACH_ADMIN, 0);
1.53 itojun 675: return;
676: }
677: #endif /*IPSEC*/
678:
1.162 dyoung 679: if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
1.53 itojun 680: if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
681: || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
682: struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
683: ip6_savecontrol(in6p, &opts, ip6, n);
684: }
685:
686: m_adj(n, off);
687: if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
688: m_freem(n);
689: if (opts)
690: m_freem(opts);
1.124 darrenr 691: so->so_rcv.sb_overflowed++;
1.53 itojun 692: udp6stat.udp6s_fullsock++;
693: } else
694: sorwakeup(so);
695: }
696: }
697: #endif
698:
1.72 itojun 699: #ifdef INET
1.53 itojun 700: static int
1.119 matt 701: udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
1.144 manu 702: struct mbuf **mp, int off /* offset of udphdr */)
1.53 itojun 703: {
704: u_int16_t *sport, *dport;
705: int rcvcnt;
706: struct in_addr *src4, *dst4;
1.109 itojun 707: struct inpcb_hdr *inph;
1.53 itojun 708: struct inpcb *inp;
1.144 manu 709: struct mbuf *m = *mp;
1.53 itojun 710:
711: rcvcnt = 0;
712: off += sizeof(struct udphdr); /* now, offset of payload */
713:
714: if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
715: goto bad;
716:
717: src4 = &src->sin_addr;
718: sport = &src->sin_port;
719: dst4 = &dst->sin_addr;
720: dport = &dst->sin_port;
721:
1.73 itojun 722: if (IN_MULTICAST(dst4->s_addr) ||
1.53 itojun 723: in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
724: /*
725: * Deliver a multicast or broadcast datagram to *all* sockets
726: * for which the local and remote addresses and ports match
727: * those of the incoming datagram. This allows more than
728: * one process to receive multi/broadcasts on the same port.
729: * (This really ought to be done for unicast datagrams as
730: * well, but that would cause problems with existing
731: * applications that open both address-specific sockets and
732: * a wildcard socket listening to the same port -- they would
733: * end up receiving duplicates of every unicast datagram.
734: * Those applications open the multiple sockets to overcome an
735: * inadequacy of the UDP socket interface, but for backwards
736: * compatibility we avoid the problem here rather than
737: * fixing the interface. Maybe 4.5BSD will remedy this?)
738: */
739:
740: /*
1.92 itojun 741: * KAME note: traditionally we dropped udpiphdr from mbuf here.
1.71 itojun 742: * we need udpiphdr for IPsec processing so we do that later.
1.53 itojun 743: */
744: /*
745: * Locate pcb(s) for datagram.
746: */
1.109 itojun 747: CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
748: inp = (struct inpcb *)inph;
749: if (inp->inp_af != AF_INET)
750: continue;
751:
1.53 itojun 752: if (inp->inp_lport != *dport)
753: continue;
754: if (!in_nullhost(inp->inp_laddr)) {
755: if (!in_hosteq(inp->inp_laddr, *dst4))
756: continue;
757: }
758: if (!in_nullhost(inp->inp_faddr)) {
759: if (!in_hosteq(inp->inp_faddr, *src4) ||
760: inp->inp_fport != *sport)
761: continue;
762: }
763:
764: udp4_sendup(m, off, (struct sockaddr *)src,
765: inp->inp_socket);
766: rcvcnt++;
767:
768: /*
769: * Don't look for additional matches if this one does
770: * not have either the SO_REUSEPORT or SO_REUSEADDR
771: * socket options set. This heuristic avoids searching
772: * through all pcbs in the common case of a non-shared
773: * port. It assumes that an application will never
774: * clear these options after setting them.
775: */
776: if ((inp->inp_socket->so_options &
777: (SO_REUSEPORT|SO_REUSEADDR)) == 0)
778: break;
779: }
780: } else {
781: /*
782: * Locate pcb for datagram.
783: */
784: inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
785: if (inp == 0) {
1.166 thorpej 786: UDP_STATINC(UDP_STAT_PCBHASHMISS);
1.53 itojun 787: inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
1.82 itojun 788: if (inp == 0)
1.53 itojun 789: return rcvcnt;
790: }
791:
1.130 manu 792: #ifdef IPSEC_NAT_T
793: /* Handle ESP over UDP */
794: if (inp->inp_flags & INP_ESPINUDP_ALL) {
795: struct sockaddr *sa = (struct sockaddr *)src;
796:
1.144 manu 797: switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
798: case -1: /* Error, m was freeed */
799: rcvcnt = -1;
800: goto bad;
801: break;
802:
803: case 1: /* ESP over UDP */
1.130 manu 804: rcvcnt++;
805: goto bad;
1.144 manu 806: break;
807:
808: case 0: /* plain UDP */
809: default: /* Unexpected */
810: /*
811: * Normal UDP processing will take place
812: * m may have changed.
813: */
814: m = *mp;
815: break;
1.130 manu 816: }
817: }
818: #endif
819:
1.53 itojun 820: udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
821: rcvcnt++;
822: }
823:
824: bad:
825: return rcvcnt;
826: }
1.72 itojun 827: #endif
1.53 itojun 828:
829: #ifdef INET6
830: static int
1.119 matt 831: udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
832: struct mbuf *m, int off)
1.53 itojun 833: {
1.79 itojun 834: u_int16_t sport, dport;
1.53 itojun 835: int rcvcnt;
1.146 rpaulo 836: struct in6_addr src6, *dst6;
1.79 itojun 837: const struct in_addr *dst4;
1.109 itojun 838: struct inpcb_hdr *inph;
1.53 itojun 839: struct in6pcb *in6p;
840:
841: rcvcnt = 0;
842: off += sizeof(struct udphdr); /* now, offset of payload */
843:
844: if (af != AF_INET && af != AF_INET6)
845: goto bad;
846: if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
847: goto bad;
848:
1.146 rpaulo 849: src6 = src->sin6_addr;
850: if (sa6_recoverscope(src) != 0) {
851: /* XXX: should be impossible. */
852: goto bad;
853: }
1.79 itojun 854: sport = src->sin6_port;
1.146 rpaulo 855:
1.79 itojun 856: dport = dst->sin6_port;
1.88 itojun 857: dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
1.146 rpaulo 858: dst6 = &dst->sin6_addr;
1.53 itojun 859:
1.146 rpaulo 860: if (IN6_IS_ADDR_MULTICAST(dst6) ||
1.79 itojun 861: (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
1.53 itojun 862: /*
863: * Deliver a multicast or broadcast datagram to *all* sockets
864: * for which the local and remote addresses and ports match
865: * those of the incoming datagram. This allows more than
866: * one process to receive multi/broadcasts on the same port.
867: * (This really ought to be done for unicast datagrams as
868: * well, but that would cause problems with existing
869: * applications that open both address-specific sockets and
870: * a wildcard socket listening to the same port -- they would
871: * end up receiving duplicates of every unicast datagram.
872: * Those applications open the multiple sockets to overcome an
873: * inadequacy of the UDP socket interface, but for backwards
874: * compatibility we avoid the problem here rather than
875: * fixing the interface. Maybe 4.5BSD will remedy this?)
876: */
877:
878: /*
1.92 itojun 879: * KAME note: traditionally we dropped udpiphdr from mbuf here.
1.71 itojun 880: * we need udpiphdr for IPsec processing so we do that later.
1.53 itojun 881: */
882: /*
883: * Locate pcb(s) for datagram.
884: */
1.109 itojun 885: CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
886: in6p = (struct in6pcb *)inph;
887: if (in6p->in6p_af != AF_INET6)
888: continue;
889:
1.79 itojun 890: if (in6p->in6p_lport != dport)
1.53 itojun 891: continue;
892: if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1.146 rpaulo 893: if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
894: dst6))
1.53 itojun 895: continue;
1.92 itojun 896: } else {
1.146 rpaulo 897: if (IN6_IS_ADDR_V4MAPPED(dst6) &&
1.85 itojun 898: (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1.55 itojun 899: continue;
900: }
1.53 itojun 901: if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1.79 itojun 902: if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
903: &src6) || in6p->in6p_fport != sport)
1.53 itojun 904: continue;
1.92 itojun 905: } else {
1.79 itojun 906: if (IN6_IS_ADDR_V4MAPPED(&src6) &&
1.85 itojun 907: (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1.55 itojun 908: continue;
909: }
1.53 itojun 910:
911: udp6_sendup(m, off, (struct sockaddr *)src,
912: in6p->in6p_socket);
913: rcvcnt++;
914:
915: /*
916: * Don't look for additional matches if this one does
917: * not have either the SO_REUSEPORT or SO_REUSEADDR
918: * socket options set. This heuristic avoids searching
919: * through all pcbs in the common case of a non-shared
920: * port. It assumes that an application will never
921: * clear these options after setting them.
922: */
923: if ((in6p->in6p_socket->so_options &
924: (SO_REUSEPORT|SO_REUSEADDR)) == 0)
925: break;
926: }
927: } else {
928: /*
929: * Locate pcb for datagram.
930: */
1.146 rpaulo 931: in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
932: dport, 0);
1.53 itojun 933: if (in6p == 0) {
1.166 thorpej 934: UDP_STATINC(UDP_STAT_PCBHASHMISS);
1.146 rpaulo 935: in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
1.82 itojun 936: if (in6p == 0)
1.53 itojun 937: return rcvcnt;
938: }
939:
940: udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
941: rcvcnt++;
942: }
943:
944: bad:
945: return rcvcnt;
946: }
947: #endif
1.1 cgd 948:
1.72 itojun 949: #ifdef INET
1.1 cgd 950: /*
951: * Notify a udp user of an asynchronous error;
952: * just wake up so that he can collect error status.
953: */
1.7 mycroft 954: static void
1.119 matt 955: udp_notify(struct inpcb *inp, int errno)
1.1 cgd 956: {
957: inp->inp_socket->so_error = errno;
958: sorwakeup(inp->inp_socket);
959: sowwakeup(inp->inp_socket);
960: }
961:
1.27 christos 962: void *
1.157 dyoung 963: udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
1.1 cgd 964: {
1.66 augustss 965: struct ip *ip = v;
966: struct udphdr *uh;
1.119 matt 967: void (*notify)(struct inpcb *, int) = udp_notify;
1.21 mycroft 968: int errno;
1.1 cgd 969:
1.53 itojun 970: if (sa->sa_family != AF_INET
971: || sa->sa_len != sizeof(struct sockaddr_in))
1.51 itojun 972: return NULL;
1.20 mycroft 973: if ((unsigned)cmd >= PRC_NCMDS)
1.27 christos 974: return NULL;
1.20 mycroft 975: errno = inetctlerrmap[cmd];
1.18 mycroft 976: if (PRC_IS_REDIRECT(cmd))
1.19 mycroft 977: notify = in_rtchange, ip = 0;
1.18 mycroft 978: else if (cmd == PRC_HOSTDEAD)
1.19 mycroft 979: ip = 0;
1.23 cgd 980: else if (errno == 0)
1.27 christos 981: return NULL;
1.19 mycroft 982: if (ip) {
1.158 christos 983: uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
1.157 dyoung 984: in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
1.34 mycroft 985: ip->ip_src, uh->uh_sport, errno, notify);
1.53 itojun 986:
987: /* XXX mapped address case */
1.19 mycroft 988: } else
1.157 dyoung 989: in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
1.34 mycroft 990: notify);
1.27 christos 991: return NULL;
1.1 cgd 992: }
993:
1.7 mycroft 994: int
1.152 yamt 995: udp_ctloutput(int op, struct socket *so, int level, int optname,
996: struct mbuf **mp)
1.130 manu 997: {
998: int s;
999: int error = 0;
1000: struct mbuf *m;
1001: struct inpcb *inp;
1002: int family;
1003:
1004: family = so->so_proto->pr_domain->dom_family;
1005:
1006: s = splsoftnet();
1007: switch (family) {
1008: #ifdef INET
1009: case PF_INET:
1010: if (level != IPPROTO_UDP) {
1011: error = ip_ctloutput(op, so, level, optname, mp);
1012: goto end;
1013: }
1014: break;
1015: #endif
1016: #ifdef INET6
1017: case PF_INET6:
1018: if (level != IPPROTO_UDP) {
1019: error = ip6_ctloutput(op, so, level, optname, mp);
1020: goto end;
1021: }
1022: break;
1023: #endif
1024: default:
1025: error = EAFNOSUPPORT;
1026: goto end;
1027: }
1028:
1029:
1030: switch (op) {
1031: case PRCO_SETOPT:
1032: m = *mp;
1033: inp = sotoinpcb(so);
1034:
1035: switch (optname) {
1036: case UDP_ENCAP:
1.163 christos 1037: if (m == NULL || m->m_len != sizeof(int)) {
1.130 manu 1038: error = EINVAL;
1.153 yamt 1039: break;
1.130 manu 1040: }
1.131 perry 1041:
1.130 manu 1042: switch(*mtod(m, int *)) {
1043: #ifdef IPSEC_NAT_T
1044: case 0:
1045: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1046: break;
1047:
1048: case UDP_ENCAP_ESPINUDP:
1049: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1050: inp->inp_flags |= INP_ESPINUDP;
1051: break;
1.131 perry 1052:
1.130 manu 1053: case UDP_ENCAP_ESPINUDP_NON_IKE:
1054: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1055: inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1056: break;
1057: #endif
1058: default:
1059: error = EINVAL;
1060: break;
1061: }
1062: break;
1063:
1064: default:
1065: error = ENOPROTOOPT;
1066: break;
1067: }
1.155 yamt 1068: if (m != NULL) {
1069: m_free(m);
1070: }
1.130 manu 1071: break;
1072:
1073: default:
1074: error = EINVAL;
1075: break;
1.131 perry 1076: }
1077:
1.130 manu 1078: end:
1079: splx(s);
1080: return error;
1081: }
1.131 perry 1082:
1.130 manu 1083:
1084: int
1.27 christos 1085: udp_output(struct mbuf *m, ...)
1086: {
1.66 augustss 1087: struct inpcb *inp;
1088: struct udpiphdr *ui;
1.125 thorpej 1089: struct route *ro;
1.66 augustss 1090: int len = m->m_pkthdr.len;
1.31 mycroft 1091: int error = 0;
1.27 christos 1092: va_list ap;
1093:
1.98 matt 1094: MCLAIM(m, &udp_tx_mowner);
1.27 christos 1095: va_start(ap, m);
1096: inp = va_arg(ap, struct inpcb *);
1097: va_end(ap);
1.1 cgd 1098:
1099: /*
1100: * Calculate data length and get a mbuf
1101: * for UDP and IP headers.
1102: */
1.13 mycroft 1103: M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1104: if (m == 0) {
1105: error = ENOBUFS;
1.39 thorpej 1106: goto release;
1107: }
1108:
1109: /*
1110: * Compute the packet length of the IP header, and
1111: * punt if the length looks bogus.
1112: */
1.96 itojun 1113: if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1.39 thorpej 1114: error = EMSGSIZE;
1.13 mycroft 1115: goto release;
1116: }
1.1 cgd 1117:
1118: /*
1119: * Fill in mbuf with extended UDP header
1120: * and addresses and length put into network format.
1121: */
1122: ui = mtod(m, struct udpiphdr *);
1123: ui->ui_pr = IPPROTO_UDP;
1124: ui->ui_src = inp->inp_laddr;
1125: ui->ui_dst = inp->inp_faddr;
1126: ui->ui_sport = inp->inp_lport;
1127: ui->ui_dport = inp->inp_fport;
1.78 thorpej 1128: ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1.1 cgd 1129:
1.125 thorpej 1130: ro = &inp->inp_route;
1131:
1.1 cgd 1132: /*
1.78 thorpej 1133: * Set up checksum and output datagram.
1.1 cgd 1134: */
1135: if (udpcksum) {
1.78 thorpej 1136: /*
1137: * XXX Cache pseudo-header checksum part for
1138: * XXX "connected" UDP sockets.
1139: */
1140: ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1141: ui->ui_dst.s_addr, htons((u_int16_t)len +
1142: sizeof(struct udphdr) + IPPROTO_UDP));
1.135 yamt 1143: m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1.78 thorpej 1144: m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1145: } else
1146: ui->ui_sum = 0;
1.96 itojun 1147: ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1.1 cgd 1148: ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */
1149: ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */
1.166 thorpej 1150: UDP_STATINC(UDP_STAT_OPACKETS);
1.48 itojun 1151:
1.125 thorpej 1152: return (ip_output(m, inp->inp_options, ro,
1.12 mycroft 1153: inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1.107 itojun 1154: inp->inp_moptions, inp->inp_socket));
1.1 cgd 1155:
1156: release:
1157: m_freem(m);
1158: return (error);
1159: }
1160:
1.42 thorpej 1161: int udp_sendspace = 9216; /* really max datagram size */
1162: int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1.1 cgd 1163: /* 40 1K datagrams */
1164:
1165: /*ARGSUSED*/
1.7 mycroft 1166: int
1.119 matt 1167: udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1.145 christos 1168: struct mbuf *control, struct lwp *l)
1.1 cgd 1169: {
1.66 augustss 1170: struct inpcb *inp;
1.1 cgd 1171: int s;
1.66 augustss 1172: int error = 0;
1.1 cgd 1173:
1174: if (req == PRU_CONTROL)
1.158 christos 1175: return (in_control(so, (long)m, (void *)nam,
1.148 ad 1176: (struct ifnet *)control, l));
1.59 thorpej 1177:
1.149 tls 1178: s = splsoftnet();
1179:
1.60 thorpej 1180: if (req == PRU_PURGEIF) {
1.80 itojun 1181: in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1.60 thorpej 1182: in_purgeif((struct ifnet *)control);
1183: in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1.149 tls 1184: splx(s);
1.59 thorpej 1185: return (0);
1186: }
1.31 mycroft 1187:
1188: inp = sotoinpcb(so);
1.32 mycroft 1189: #ifdef DIAGNOSTIC
1190: if (req != PRU_SEND && req != PRU_SENDOOB && control)
1191: panic("udp_usrreq: unexpected control mbuf");
1192: #endif
1.31 mycroft 1193: if (inp == 0 && req != PRU_ATTACH) {
1194: error = EINVAL;
1195: goto release;
1196: }
1197:
1.1 cgd 1198: /*
1199: * Note: need to block udp_input while changing
1200: * the udp pcb queue and/or pcb addresses.
1201: */
1202: switch (req) {
1203:
1204: case PRU_ATTACH:
1.31 mycroft 1205: if (inp != 0) {
1206: error = EISCONN;
1.1 cgd 1207: break;
1208: }
1.98 matt 1209: #ifdef MBUFTRACE
1210: so->so_mowner = &udp_mowner;
1211: so->so_rcv.sb_mowner = &udp_rx_mowner;
1212: so->so_snd.sb_mowner = &udp_tx_mowner;
1213: #endif
1.31 mycroft 1214: if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1215: error = soreserve(so, udp_sendspace, udp_recvspace);
1216: if (error)
1217: break;
1218: }
1.18 mycroft 1219: error = in_pcballoc(so, &udbtable);
1.1 cgd 1220: if (error)
1221: break;
1.31 mycroft 1222: inp = sotoinpcb(so);
1223: inp->inp_ip.ip_ttl = ip_defttl;
1.1 cgd 1224: break;
1225:
1226: case PRU_DETACH:
1.31 mycroft 1227: in_pcbdetach(inp);
1.1 cgd 1228: break;
1229:
1230: case PRU_BIND:
1.148 ad 1231: error = in_pcbbind(inp, nam, l);
1.1 cgd 1232: break;
1233:
1234: case PRU_LISTEN:
1235: error = EOPNOTSUPP;
1236: break;
1237:
1238: case PRU_CONNECT:
1.148 ad 1239: error = in_pcbconnect(inp, nam, l);
1.31 mycroft 1240: if (error)
1.1 cgd 1241: break;
1.31 mycroft 1242: soisconnected(so);
1.1 cgd 1243: break;
1244:
1245: case PRU_CONNECT2:
1246: error = EOPNOTSUPP;
1247: break;
1248:
1249: case PRU_DISCONNECT:
1.31 mycroft 1250: /*soisdisconnected(so);*/
1251: so->so_state &= ~SS_ISCONNECTED; /* XXX */
1.1 cgd 1252: in_pcbdisconnect(inp);
1.34 mycroft 1253: inp->inp_laddr = zeroin_addr; /* XXX */
1.35 mycroft 1254: in_pcbstate(inp, INP_BOUND); /* XXX */
1.1 cgd 1255: break;
1256:
1257: case PRU_SHUTDOWN:
1258: socantsendmore(so);
1259: break;
1260:
1.31 mycroft 1261: case PRU_RCVD:
1262: error = EOPNOTSUPP;
1.1 cgd 1263: break;
1264:
1.31 mycroft 1265: case PRU_SEND:
1.32 mycroft 1266: if (control && control->m_len) {
1267: m_freem(control);
1268: m_freem(m);
1269: error = EINVAL;
1270: break;
1271: }
1.31 mycroft 1272: {
1.35 mycroft 1273: struct in_addr laddr; /* XXX */
1.1 cgd 1274:
1.31 mycroft 1275: if (nam) {
1.35 mycroft 1276: laddr = inp->inp_laddr; /* XXX */
1.31 mycroft 1277: if ((so->so_state & SS_ISCONNECTED) != 0) {
1278: error = EISCONN;
1.32 mycroft 1279: goto die;
1.31 mycroft 1280: }
1.148 ad 1281: error = in_pcbconnect(inp, nam, l);
1.100 matt 1282: if (error)
1283: goto die;
1.31 mycroft 1284: } else {
1285: if ((so->so_state & SS_ISCONNECTED) == 0) {
1286: error = ENOTCONN;
1.32 mycroft 1287: goto die;
1.31 mycroft 1288: }
1289: }
1.33 mycroft 1290: error = udp_output(m, inp);
1.100 matt 1291: m = NULL;
1.31 mycroft 1292: if (nam) {
1293: in_pcbdisconnect(inp);
1.35 mycroft 1294: inp->inp_laddr = laddr; /* XXX */
1295: in_pcbstate(inp, INP_BOUND); /* XXX */
1.31 mycroft 1296: }
1.100 matt 1297: die:
1298: if (m)
1299: m_freem(m);
1.31 mycroft 1300: }
1.1 cgd 1301: break;
1302:
1303: case PRU_SENSE:
1304: /*
1305: * stat: don't bother with a blocksize.
1306: */
1.31 mycroft 1307: splx(s);
1.1 cgd 1308: return (0);
1309:
1.31 mycroft 1310: case PRU_RCVOOB:
1311: error = EOPNOTSUPP;
1312: break;
1313:
1.1 cgd 1314: case PRU_SENDOOB:
1.32 mycroft 1315: m_freem(control);
1.31 mycroft 1316: m_freem(m);
1.1 cgd 1317: error = EOPNOTSUPP;
1318: break;
1319:
1.31 mycroft 1320: case PRU_SOCKADDR:
1321: in_setsockaddr(inp, nam);
1322: break;
1323:
1324: case PRU_PEERADDR:
1325: in_setpeeraddr(inp, nam);
1326: break;
1.1 cgd 1327:
1328: default:
1329: panic("udp_usrreq");
1330: }
1331:
1332: release:
1.31 mycroft 1333: splx(s);
1.1 cgd 1334: return (error);
1.13 mycroft 1335: }
1336:
1.166 thorpej 1337: static void
1338: udpstat_convert_to_user_cb(void *v1, void *v2, struct cpu_info *ci)
1339: {
1340: uint64_t *udpsc = v1;
1341: uint64_t *udps = v2;
1342: u_int i;
1343:
1344: for (i = 0; i < UDP_NSTATS; i++)
1345: udps[i] += udpsc[i];
1346: }
1347:
1348: static void
1349: udpstat_convert_to_user(uint64_t *udps)
1350: {
1351:
1352: memset(udps, 0, sizeof(uint64_t) * UDP_NSTATS);
1353: percpu_foreach(udpstat_percpu, udpstat_convert_to_user_cb, udps);
1354: }
1355:
1356: static int
1357: sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1358: {
1359: struct sysctlnode node;
1360: uint64_t udps[UDP_NSTATS];
1361:
1362: udpstat_convert_to_user(udps);
1363: node = *rnode;
1364: node.sysctl_data = udps;
1365: node.sysctl_size = sizeof(udps);
1366: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
1367: }
1368:
1.13 mycroft 1369: /*
1370: * Sysctl for udp variables.
1371: */
1.114 atatat 1372: SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1.13 mycroft 1373: {
1.114 atatat 1374:
1.116 atatat 1375: sysctl_createv(clog, 0, NULL, NULL,
1376: CTLFLAG_PERMANENT,
1.114 atatat 1377: CTLTYPE_NODE, "net", NULL,
1378: NULL, 0, NULL, 0,
1379: CTL_NET, CTL_EOL);
1.116 atatat 1380: sysctl_createv(clog, 0, NULL, NULL,
1381: CTLFLAG_PERMANENT,
1.114 atatat 1382: CTLTYPE_NODE, "inet", NULL,
1383: NULL, 0, NULL, 0,
1384: CTL_NET, PF_INET, CTL_EOL);
1.116 atatat 1385: sysctl_createv(clog, 0, NULL, NULL,
1386: CTLFLAG_PERMANENT,
1.122 atatat 1387: CTLTYPE_NODE, "udp",
1388: SYSCTL_DESCR("UDPv4 related settings"),
1.114 atatat 1389: NULL, 0, NULL, 0,
1390: CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1391:
1.116 atatat 1392: sysctl_createv(clog, 0, NULL, NULL,
1393: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1394: CTLTYPE_INT, "checksum",
1.123 heas 1395: SYSCTL_DESCR("Compute UDP checksums"),
1.114 atatat 1396: NULL, 0, &udpcksum, 0,
1397: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1398: CTL_EOL);
1.116 atatat 1399: sysctl_createv(clog, 0, NULL, NULL,
1400: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1401: CTLTYPE_INT, "sendspace",
1402: SYSCTL_DESCR("Default UDP send buffer size"),
1.114 atatat 1403: NULL, 0, &udp_sendspace, 0,
1404: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1405: CTL_EOL);
1.116 atatat 1406: sysctl_createv(clog, 0, NULL, NULL,
1407: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1408: CTLTYPE_INT, "recvspace",
1409: SYSCTL_DESCR("Default UDP receive buffer size"),
1.114 atatat 1410: NULL, 0, &udp_recvspace, 0,
1411: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1412: CTL_EOL);
1.125 thorpej 1413: sysctl_createv(clog, 0, NULL, NULL,
1414: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1415: CTLTYPE_INT, "do_loopback_cksum",
1416: SYSCTL_DESCR("Perform UDP checksum on loopback"),
1417: NULL, 0, &udp_do_loopback_cksum, 0,
1418: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1419: CTL_EOL);
1.132 atatat 1420: sysctl_createv(clog, 0, NULL, NULL,
1421: CTLFLAG_PERMANENT,
1.134 atatat 1422: CTLTYPE_STRUCT, "pcblist",
1.132 atatat 1423: SYSCTL_DESCR("UDP protocol control block list"),
1424: sysctl_inpcblist, 0, &udbtable, 0,
1425: CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1426: CTL_EOL);
1.139 elad 1427: sysctl_createv(clog, 0, NULL, NULL,
1428: CTLFLAG_PERMANENT,
1429: CTLTYPE_STRUCT, "stats",
1430: SYSCTL_DESCR("UDP statistics"),
1.166 thorpej 1431: sysctl_net_inet_udp_stats, 0, NULL, 0,
1.139 elad 1432: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1433: CTL_EOL);
1.1 cgd 1434: }
1.72 itojun 1435: #endif
1.130 manu 1436:
1.166 thorpej 1437: void
1438: udp_statinc(u_int stat)
1439: {
1440:
1441: KASSERT(stat < UDP_NSTATS);
1442: UDP_STATINC(stat);
1443: }
1444:
1.130 manu 1445: #if (defined INET && defined IPSEC_NAT_T)
1446: /*
1447: * Returns:
1448: * 1 if the packet was processed
1449: * 0 if normal UDP processing should take place
1.144 manu 1450: * -1 if an error occurent and m was freed
1.130 manu 1451: */
1452: static int
1.154 yamt 1453: udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1.151 christos 1454: struct socket *so)
1.130 manu 1455: {
1456: size_t len;
1.158 christos 1457: void *data;
1.130 manu 1458: struct inpcb *inp;
1459: size_t skip = 0;
1460: size_t minlen;
1461: size_t iphdrlen;
1462: struct ip *ip;
1463: struct mbuf *n;
1.136 manu 1464: struct m_tag *tag;
1465: struct udphdr *udphdr;
1466: u_int16_t sport, dport;
1.144 manu 1467: struct mbuf *m = *mp;
1.130 manu 1468:
1.131 perry 1469: /*
1.130 manu 1470: * Collapse the mbuf chain if the first mbuf is too short
1471: * The longest case is: UDP + non ESP marker + ESP
1472: */
1473: minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1474: if (minlen > m->m_pkthdr.len)
1475: minlen = m->m_pkthdr.len;
1476:
1477: if (m->m_len < minlen) {
1.144 manu 1478: if ((*mp = m_pullup(m, minlen)) == NULL) {
1.130 manu 1479: printf("udp4_espinudp: m_pullup failed\n");
1.144 manu 1480: return -1;
1.130 manu 1481: }
1.144 manu 1482: m = *mp;
1.130 manu 1483: }
1484:
1.131 perry 1485: len = m->m_len - off;
1.158 christos 1486: data = mtod(m, char *) + off;
1.130 manu 1487: inp = sotoinpcb(so);
1488:
1489: /* Ignore keepalive packets */
1.158 christos 1490: if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1.130 manu 1491: return 1;
1492: }
1493:
1.131 perry 1494: /*
1495: * Check that the payload is long enough to hold
1.130 manu 1496: * an ESP header and compute the length of encapsulation
1.131 perry 1497: * header to remove
1.130 manu 1498: */
1499: if (inp->inp_flags & INP_ESPINUDP) {
1500: u_int32_t *st = (u_int32_t *)data;
1501:
1502: if ((len <= sizeof(struct esp)) || (*st == 0))
1503: return 0; /* Normal UDP processing */
1504:
1505: skip = sizeof(struct udphdr);
1506: }
1507:
1508: if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1.142 kleink 1509: u_int32_t *st = (u_int32_t *)data;
1.130 manu 1510:
1511: if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1.142 kleink 1512: || ((st[0] | st[1]) != 0))
1.130 manu 1513: return 0; /* Normal UDP processing */
1.131 perry 1514:
1.130 manu 1515: skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1516: }
1517:
1518: /*
1.136 manu 1519: * Get the UDP ports. They are handled in network
1520: * order everywhere in IPSEC_NAT_T code.
1521: */
1.158 christos 1522: udphdr = (struct udphdr *)((char *)data - skip);
1.136 manu 1523: sport = udphdr->uh_sport;
1524: dport = udphdr->uh_dport;
1525:
1526: /*
1.130 manu 1527: * Remove the UDP header (and possibly the non ESP marker)
1528: * IP header lendth is iphdrlen
1.131 perry 1529: * Before:
1.130 manu 1530: * <--- off --->
1531: * +----+------+-----+
1532: * | IP | UDP | ESP |
1533: * +----+------+-----+
1534: * <-skip->
1535: * After:
1536: * +----+-----+
1537: * | IP | ESP |
1538: * +----+-----+
1539: * <-skip->
1540: */
1541: iphdrlen = off - sizeof(struct udphdr);
1.158 christos 1542: memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1.130 manu 1543: m_adj(m, skip);
1544:
1545: ip = mtod(m, struct ip *);
1546: ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1547: ip->ip_p = IPPROTO_ESP;
1548:
1549: /*
1.131 perry 1550: * Copy the mbuf to avoid multiple free, as both
1551: * esp4_input (which we call) and udp_input (which
1.130 manu 1552: * called us) free the mbuf.
1553: */
1554: if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1555: printf("udp4_espinudp: m_dup failed\n");
1556: return 0;
1557: }
1558:
1.136 manu 1559: /*
1560: * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1561: * the source UDP port. This is required if we want
1562: * to select the right SPD for multiple hosts behind
1563: * same NAT
1564: */
1.137 manu 1565: if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1566: sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1567: printf("udp4_espinudp: m_tag_get failed\n");
1.138 manu 1568: m_freem(n);
1.137 manu 1569: return 0;
1570: }
1.136 manu 1571: ((u_int16_t *)(tag + 1))[0] = sport;
1572: ((u_int16_t *)(tag + 1))[1] = dport;
1573: m_tag_prepend(n, tag);
1574:
1.147 christos 1575: #ifdef FAST_IPSEC
1.160 degroote 1576: ipsec4_common_input(n, iphdrlen, IPPROTO_ESP);
1.147 christos 1577: #else
1.130 manu 1578: esp4_input(n, iphdrlen);
1.147 christos 1579: #endif
1.130 manu 1580:
1581: /* We handled it, it shoudln't be handled by UDP */
1582: return 1;
1583: }
1584: #endif
CVSweb <webmaster@jp.NetBSD.org>