Annotation of src/sys/netinet/udp_usrreq.c, Revision 1.143
1.143 ! dsl 1: /* $NetBSD: udp_usrreq.c,v 1.142 2005/09/03 18:01:07 kleink 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>
1.143 ! dsl 64: __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.142 2005/09/03 18:01:07 kleink 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:
1.130 manu 98: #ifdef IPSEC_NAT_T
99: #include <netinet6/ipsec.h>
100: #include <netinet6/esp.h>
101: #endif
102:
1.53 itojun 103: #ifdef INET6
104: #include <netinet/ip6.h>
105: #include <netinet/icmp6.h>
106: #include <netinet6/ip6_var.h>
107: #include <netinet6/in6_pcb.h>
108: #include <netinet6/udp6_var.h>
109: #endif
110:
111: #ifndef INET6
112: /* always need ip6.h for IP6_EXTHDR_GET */
113: #include <netinet/ip6.h>
114: #endif
115:
1.76 itojun 116: #include "faith.h"
117: #if defined(NFAITH) && NFAITH > 0
118: #include <net/if_faith.h>
119: #endif
120:
1.27 christos 121: #include <machine/stdarg.h>
122:
1.105 jonathan 123: #ifdef FAST_IPSEC
124: #include <netipsec/ipsec.h>
1.121 jonathan 125: #include <netipsec/ipsec_var.h> /* XXX ipsecstat namespace */
1.105 jonathan 126: #ifdef INET6
127: #include <netipsec/ipsec6.h>
128: #endif
129: #endif /* FAST_IPSEC*/
130:
1.48 itojun 131: #ifdef IPSEC
132: #include <netinet6/ipsec.h>
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: #ifndef COMPAT_42
145: int udpcksum = 1;
146: #else
147: int udpcksum = 0; /* XXX */
148: #endif
1.141 yamt 149: int udp_do_loopback_cksum = 0;
1.93 matt 150:
151: struct inpcbtable udbtable;
152: struct udpstat udpstat;
1.8 mycroft 153:
1.72 itojun 154: #ifdef INET
1.130 manu 155: #ifdef IPSEC_NAT_T
156: static int udp4_espinudp (struct mbuf *, int, struct sockaddr *,
157: struct socket *);
158: #endif
1.119 matt 159: static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
160: struct socket *);
161: static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
162: struct mbuf *, int);
1.129 yamt 163: static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
1.72 itojun 164: #endif
1.53 itojun 165: #ifdef INET6
1.119 matt 166: static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
167: struct socket *);
168: static int udp6_realinput (int, struct sockaddr_in6 *,
169: struct sockaddr_in6 *, struct mbuf *, int);
1.129 yamt 170: static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
1.53 itojun 171: #endif
1.72 itojun 172: #ifdef INET
1.119 matt 173: static void udp_notify (struct inpcb *, int);
1.72 itojun 174: #endif
1.7 mycroft 175:
1.26 mycroft 176: #ifndef UDBHASHSIZE
177: #define UDBHASHSIZE 128
178: #endif
179: int udbhashsize = UDBHASHSIZE;
180:
1.98 matt 181: #ifdef MBUFTRACE
182: struct mowner udp_mowner = { "udp" };
183: struct mowner udp_rx_mowner = { "udp", "rx" };
184: struct mowner udp_tx_mowner = { "udp", "tx" };
185: #endif
186:
1.78 thorpej 187: #ifdef UDP_CSUM_COUNTERS
188: #include <sys/device.h>
189:
1.140 yamt 190: #if defined(INET)
1.78 thorpej 191: struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
192: NULL, "udp", "hwcsum bad");
193: struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
194: NULL, "udp", "hwcsum ok");
195: struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
196: NULL, "udp", "hwcsum data");
197: struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
198: NULL, "udp", "swcsum");
199:
1.120 matt 200: EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
201: EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
202: EVCNT_ATTACH_STATIC(udp_hwcsum_data);
203: EVCNT_ATTACH_STATIC(udp_swcsum);
1.140 yamt 204: #endif /* defined(INET) */
205:
206: #if defined(INET6)
207: struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
208: NULL, "udp6", "hwcsum bad");
209: struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
210: NULL, "udp6", "hwcsum ok");
211: struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
212: NULL, "udp6", "hwcsum data");
213: struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
214: NULL, "udp6", "swcsum");
215:
216: EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
217: EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
218: EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
219: EVCNT_ATTACH_STATIC(udp6_swcsum);
220: #endif /* defined(INET6) */
221:
222: #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
1.120 matt 223:
1.78 thorpej 224: #else
225:
226: #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */
227:
228: #endif /* UDP_CSUM_COUNTERS */
229:
1.7 mycroft 230: void
1.119 matt 231: udp_init(void)
1.1 cgd 232: {
1.18 mycroft 233:
1.35 mycroft 234: in_pcbinit(&udbtable, udbhashsize, udbhashsize);
1.78 thorpej 235:
1.98 matt 236: MOWNER_ATTACH(&udp_tx_mowner);
237: MOWNER_ATTACH(&udp_rx_mowner);
238: MOWNER_ATTACH(&udp_mowner);
1.1 cgd 239: }
240:
1.129 yamt 241: /*
242: * Checksum extended UDP header and data.
243: */
244:
245: int
246: udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
247: int iphlen, int len)
248: {
249:
250: switch (af) {
1.72 itojun 251: #ifdef INET
1.129 yamt 252: case AF_INET:
253: return udp4_input_checksum(m, uh, iphlen, len);
254: #endif
255: #ifdef INET6
256: case AF_INET6:
257: return udp6_input_checksum(m, uh, iphlen, len);
258: #endif
259: }
260: #ifdef DIAGNOSTIC
261: panic("udp_input_checksum: unknown af %d", af);
262: #endif
263: /* NOTREACHED */
264: return -1;
265: }
266:
267: #ifdef INET
268:
269: /*
270: * Checksum extended UDP header and data.
271: */
272:
273: static int
274: udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
275: int iphlen, int len)
276: {
277:
278: /*
279: * XXX it's better to record and check if this mbuf is
280: * already checked.
281: */
282:
283: if (uh->uh_sum == 0)
284: return 0;
285:
286: switch (m->m_pkthdr.csum_flags &
287: ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
288: M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
289: case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
290: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
291: goto badcsum;
292:
293: case M_CSUM_UDPv4|M_CSUM_DATA: {
294: u_int32_t hw_csum = m->m_pkthdr.csum_data;
295:
296: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
297: if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
298: const struct ip *ip =
299: mtod(m, const struct ip *);
300:
301: hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
302: ip->ip_dst.s_addr,
303: htons(hw_csum + len + IPPROTO_UDP));
304: }
305: if ((hw_csum ^ 0xffff) != 0)
306: goto badcsum;
307: break;
308: }
309:
310: case M_CSUM_UDPv4:
311: /* Checksum was okay. */
312: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
313: break;
314:
315: default:
316: /*
317: * Need to compute it ourselves. Maybe skip checksum
318: * on loopback interfaces.
319: */
320: if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
321: IFF_LOOPBACK) ||
322: udp_do_loopback_cksum)) {
323: UDP_CSUM_COUNTER_INCR(&udp_swcsum);
324: if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
325: goto badcsum;
326: }
327: break;
328: }
329:
330: return 0;
331:
332: badcsum:
333: udpstat.udps_badsum++;
334: return -1;
335: }
336:
1.7 mycroft 337: void
1.27 christos 338: udp_input(struct mbuf *m, ...)
1.1 cgd 339: {
1.53 itojun 340: va_list ap;
341: struct sockaddr_in src, dst;
342: struct ip *ip;
343: struct udphdr *uh;
1.97 simonb 344: int iphlen;
1.53 itojun 345: int len;
346: int n;
1.96 itojun 347: u_int16_t ip_len;
1.53 itojun 348:
349: va_start(ap, m);
350: iphlen = va_arg(ap, int);
1.97 simonb 351: (void)va_arg(ap, int); /* ignore value, advance ap */
1.53 itojun 352: va_end(ap);
353:
1.98 matt 354: MCLAIM(m, &udp_rx_mowner);
1.53 itojun 355: udpstat.udps_ipackets++;
356:
357: /*
358: * Get IP and UDP header together in first mbuf.
359: */
360: ip = mtod(m, struct ip *);
361: IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
362: if (uh == NULL) {
363: udpstat.udps_hdrops++;
364: return;
365: }
1.95 thorpej 366: KASSERT(UDP_HDR_ALIGNED_P(uh));
1.53 itojun 367:
1.57 itojun 368: /* destination port of 0 is illegal, based on RFC768. */
369: if (uh->uh_dport == 0)
370: goto bad;
371:
1.53 itojun 372: /*
373: * Make mbuf data length reflect UDP length.
374: * If not enough data to reflect UDP length, drop.
375: */
1.96 itojun 376: ip_len = ntohs(ip->ip_len);
1.53 itojun 377: len = ntohs((u_int16_t)uh->uh_ulen);
1.96 itojun 378: if (ip_len != iphlen + len) {
379: if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
1.53 itojun 380: udpstat.udps_badlen++;
381: goto bad;
382: }
1.96 itojun 383: m_adj(m, iphlen + len - ip_len);
1.53 itojun 384: }
385:
386: /*
387: * Checksum extended UDP header and data.
388: */
1.129 yamt 389: if (udp4_input_checksum(m, uh, iphlen, len))
390: goto badcsum;
1.53 itojun 391:
392: /* construct source and dst sockaddrs. */
393: bzero(&src, sizeof(src));
394: src.sin_family = AF_INET;
395: src.sin_len = sizeof(struct sockaddr_in);
396: bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
397: src.sin_port = uh->uh_sport;
398: bzero(&dst, sizeof(dst));
399: dst.sin_family = AF_INET;
400: dst.sin_len = sizeof(struct sockaddr_in);
401: bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
402: dst.sin_port = uh->uh_dport;
403:
404: n = udp4_realinput(&src, &dst, m, iphlen);
405: #ifdef INET6
406: if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
407: struct sockaddr_in6 src6, dst6;
408:
409: bzero(&src6, sizeof(src6));
410: src6.sin6_family = AF_INET6;
411: src6.sin6_len = sizeof(struct sockaddr_in6);
412: src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
413: bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
414: sizeof(ip->ip_src));
415: src6.sin6_port = uh->uh_sport;
416: bzero(&dst6, sizeof(dst6));
417: dst6.sin6_family = AF_INET6;
418: dst6.sin6_len = sizeof(struct sockaddr_in6);
419: dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
420: bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
421: sizeof(ip->ip_dst));
422: dst6.sin6_port = uh->uh_dport;
423:
424: n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
425: }
426: #endif
427:
428: if (n == 0) {
429: if (m->m_flags & (M_BCAST | M_MCAST)) {
430: udpstat.udps_noportbcast++;
431: goto bad;
432: }
1.61 itojun 433: udpstat.udps_noport++;
1.64 ws 434: #ifdef IPKDB
1.53 itojun 435: if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
436: m, iphlen + sizeof(struct udphdr),
437: m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
438: /*
439: * It was a debugger connect packet,
440: * just drop it now
441: */
442: goto bad;
443: }
444: #endif
445: icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
446: m = NULL;
447: }
448:
449: bad:
450: if (m)
451: m_freem(m);
1.78 thorpej 452: return;
453:
454: badcsum:
455: m_freem(m);
1.53 itojun 456: }
1.72 itojun 457: #endif
1.53 itojun 458:
459: #ifdef INET6
1.129 yamt 460: static int
461: udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
462: {
463:
1.140 yamt 464: /*
465: * XXX it's better to record and check if this mbuf is
466: * already checked.
467: */
468:
1.129 yamt 469: if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
470: goto good;
471: }
472: if (uh->uh_sum == 0) {
473: udp6stat.udp6s_nosum++;
474: goto bad;
475: }
1.140 yamt 476:
477: switch (m->m_pkthdr.csum_flags &
478: ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
479: M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
480: case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
481: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
1.129 yamt 482: udp6stat.udp6s_badsum++;
483: goto bad;
1.140 yamt 484:
485: #if 0 /* notyet */
486: case M_CSUM_UDPv6|M_CSUM_DATA:
487: #endif
488:
489: case M_CSUM_UDPv6:
490: /* Checksum was okay. */
491: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
492: break;
493:
494: default:
495: /*
496: * Need to compute it ourselves. Maybe skip checksum
497: * on loopback interfaces.
498: */
499: UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
500: if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
501: udp6stat.udp6s_badsum++;
502: goto bad;
503: }
1.129 yamt 504: }
505:
506: good:
507: return 0;
508: bad:
509: return -1;
510: }
511:
1.53 itojun 512: int
1.119 matt 513: udp6_input(struct mbuf **mp, int *offp, int proto)
1.53 itojun 514: {
515: struct mbuf *m = *mp;
516: int off = *offp;
517: struct sockaddr_in6 src, dst;
518: struct ip6_hdr *ip6;
519: struct udphdr *uh;
520: u_int32_t plen, ulen;
521:
1.76 itojun 522: ip6 = mtod(m, struct ip6_hdr *);
523:
1.53 itojun 524: #if defined(NFAITH) && 0 < NFAITH
1.76 itojun 525: if (faithprefix(&ip6->ip6_dst)) {
526: /* send icmp6 host unreach? */
527: m_freem(m);
528: return IPPROTO_DONE;
1.53 itojun 529: }
530: #endif
531:
532: udp6stat.udp6s_ipackets++;
533:
534: /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */
535: plen = m->m_pkthdr.len - off;
536: IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
537: if (uh == NULL) {
538: ip6stat.ip6s_tooshort++;
539: return IPPROTO_DONE;
540: }
1.95 thorpej 541: KASSERT(UDP_HDR_ALIGNED_P(uh));
1.53 itojun 542: ulen = ntohs((u_short)uh->uh_ulen);
1.69 itojun 543: /*
544: * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
545: * iff payload length > 0xffff.
546: */
1.53 itojun 547: if (ulen == 0 && plen > 0xffff)
548: ulen = plen;
549:
550: if (plen != ulen) {
551: udp6stat.udp6s_badlen++;
1.54 itojun 552: goto bad;
553: }
554:
1.57 itojun 555: /* destination port of 0 is illegal, based on RFC768. */
556: if (uh->uh_dport == 0)
557: goto bad;
558:
1.54 itojun 559: /* Be proactive about malicious use of IPv4 mapped address */
560: if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
561: IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
562: /* XXX stat */
1.53 itojun 563: goto bad;
564: }
565:
566: /*
1.125 thorpej 567: * Checksum extended UDP header and data. Maybe skip checksum
568: * on loopback interfaces.
1.53 itojun 569: */
1.129 yamt 570: if (udp6_input_checksum(m, uh, off, ulen))
571: goto bad;
1.53 itojun 572:
573: /*
574: * Construct source and dst sockaddrs.
575: * Note that ifindex (s6_addr16[1]) is already filled.
576: */
577: bzero(&src, sizeof(src));
578: src.sin6_family = AF_INET6;
579: src.sin6_len = sizeof(struct sockaddr_in6);
1.69 itojun 580: /* KAME hack: recover scopeid */
581: (void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1.53 itojun 582: src.sin6_port = uh->uh_sport;
583: bzero(&dst, sizeof(dst));
584: dst.sin6_family = AF_INET6;
585: dst.sin6_len = sizeof(struct sockaddr_in6);
1.69 itojun 586: /* KAME hack: recover scopeid */
587: (void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1.53 itojun 588: dst.sin6_port = uh->uh_dport;
589:
590: if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
591: if (m->m_flags & M_MCAST) {
592: udp6stat.udp6s_noportmcast++;
593: goto bad;
594: }
1.61 itojun 595: udp6stat.udp6s_noport++;
1.53 itojun 596: icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
597: m = NULL;
598: }
599:
600: bad:
601: if (m)
602: m_freem(m);
603: return IPPROTO_DONE;
604: }
605: #endif
606:
1.72 itojun 607: #ifdef INET
1.53 itojun 608: static void
1.119 matt 609: udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
610: struct sockaddr *src, struct socket *so)
1.53 itojun 611: {
612: struct mbuf *opts = NULL;
613: struct mbuf *n;
614: struct inpcb *inp = NULL;
615:
616: if (!so)
617: return;
618: switch (so->so_proto->pr_domain->dom_family) {
619: case AF_INET:
620: inp = sotoinpcb(so);
621: break;
622: #ifdef INET6
623: case AF_INET6:
624: break;
625: #endif
626: default:
627: return;
628: }
629:
1.105 jonathan 630: #if defined(IPSEC) || defined(FAST_IPSEC)
1.53 itojun 631: /* check AH/ESP integrity. */
632: if (so != NULL && ipsec4_in_reject_so(m, so)) {
633: ipsecstat.in_polvio++;
1.110 itojun 634: if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
635: icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
636: 0, 0);
1.53 itojun 637: return;
638: }
639: #endif /*IPSEC*/
640:
641: if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
642: if (inp && (inp->inp_flags & INP_CONTROLOPTS
643: || so->so_options & SO_TIMESTAMP)) {
644: struct ip *ip = mtod(n, struct ip *);
645: ip_savecontrol(inp, &opts, ip, n);
646: }
647:
648: m_adj(n, off);
649: if (sbappendaddr(&so->so_rcv, src, n,
650: opts) == 0) {
651: m_freem(n);
652: if (opts)
653: m_freem(opts);
1.124 darrenr 654: so->so_rcv.sb_overflowed++;
1.81 enami 655: udpstat.udps_fullsock++;
1.53 itojun 656: } else
657: sorwakeup(so);
658: }
659: }
1.72 itojun 660: #endif
1.53 itojun 661:
662: #ifdef INET6
663: static void
1.119 matt 664: udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
665: struct sockaddr *src, struct socket *so)
1.53 itojun 666: {
667: struct mbuf *opts = NULL;
668: struct mbuf *n;
669: struct in6pcb *in6p = NULL;
670:
671: if (!so)
672: return;
673: if (so->so_proto->pr_domain->dom_family != AF_INET6)
674: return;
675: in6p = sotoin6pcb(so);
676:
1.105 jonathan 677: #if defined(IPSEC) || defined(FAST_IPSEC)
1.53 itojun 678: /* check AH/ESP integrity. */
679: if (so != NULL && ipsec6_in_reject_so(m, so)) {
680: ipsec6stat.in_polvio++;
1.111 mycroft 681: if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
1.110 itojun 682: icmp6_error(n, ICMP6_DST_UNREACH,
683: ICMP6_DST_UNREACH_ADMIN, 0);
1.53 itojun 684: return;
685: }
686: #endif /*IPSEC*/
687:
688: if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
689: if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
690: || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
691: struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
692: ip6_savecontrol(in6p, &opts, ip6, n);
693: }
694:
695: m_adj(n, off);
696: if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
697: m_freem(n);
698: if (opts)
699: m_freem(opts);
1.124 darrenr 700: so->so_rcv.sb_overflowed++;
1.53 itojun 701: udp6stat.udp6s_fullsock++;
702: } else
703: sorwakeup(so);
704: }
705: }
706: #endif
707:
1.72 itojun 708: #ifdef INET
1.53 itojun 709: static int
1.119 matt 710: udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
711: struct mbuf *m, int off /* offset of udphdr */)
1.53 itojun 712: {
713: u_int16_t *sport, *dport;
714: int rcvcnt;
715: struct in_addr *src4, *dst4;
1.109 itojun 716: struct inpcb_hdr *inph;
1.53 itojun 717: struct inpcb *inp;
718:
719: rcvcnt = 0;
720: off += sizeof(struct udphdr); /* now, offset of payload */
721:
722: if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
723: goto bad;
724:
725: src4 = &src->sin_addr;
726: sport = &src->sin_port;
727: dst4 = &dst->sin_addr;
728: dport = &dst->sin_port;
729:
1.73 itojun 730: if (IN_MULTICAST(dst4->s_addr) ||
1.53 itojun 731: in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
732: /*
733: * Deliver a multicast or broadcast datagram to *all* sockets
734: * for which the local and remote addresses and ports match
735: * those of the incoming datagram. This allows more than
736: * one process to receive multi/broadcasts on the same port.
737: * (This really ought to be done for unicast datagrams as
738: * well, but that would cause problems with existing
739: * applications that open both address-specific sockets and
740: * a wildcard socket listening to the same port -- they would
741: * end up receiving duplicates of every unicast datagram.
742: * Those applications open the multiple sockets to overcome an
743: * inadequacy of the UDP socket interface, but for backwards
744: * compatibility we avoid the problem here rather than
745: * fixing the interface. Maybe 4.5BSD will remedy this?)
746: */
747:
748: /*
1.92 itojun 749: * KAME note: traditionally we dropped udpiphdr from mbuf here.
1.71 itojun 750: * we need udpiphdr for IPsec processing so we do that later.
1.53 itojun 751: */
752: /*
753: * Locate pcb(s) for datagram.
754: */
1.109 itojun 755: CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
756: inp = (struct inpcb *)inph;
757: if (inp->inp_af != AF_INET)
758: continue;
759:
1.53 itojun 760: if (inp->inp_lport != *dport)
761: continue;
762: if (!in_nullhost(inp->inp_laddr)) {
763: if (!in_hosteq(inp->inp_laddr, *dst4))
764: continue;
765: }
766: if (!in_nullhost(inp->inp_faddr)) {
767: if (!in_hosteq(inp->inp_faddr, *src4) ||
768: inp->inp_fport != *sport)
769: continue;
770: }
771:
772: udp4_sendup(m, off, (struct sockaddr *)src,
773: inp->inp_socket);
774: rcvcnt++;
775:
776: /*
777: * Don't look for additional matches if this one does
778: * not have either the SO_REUSEPORT or SO_REUSEADDR
779: * socket options set. This heuristic avoids searching
780: * through all pcbs in the common case of a non-shared
781: * port. It assumes that an application will never
782: * clear these options after setting them.
783: */
784: if ((inp->inp_socket->so_options &
785: (SO_REUSEPORT|SO_REUSEADDR)) == 0)
786: break;
787: }
788: } else {
789: /*
790: * Locate pcb for datagram.
791: */
792: inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
793: if (inp == 0) {
794: ++udpstat.udps_pcbhashmiss;
795: inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
1.82 itojun 796: if (inp == 0)
1.53 itojun 797: return rcvcnt;
798: }
799:
1.130 manu 800: #ifdef IPSEC_NAT_T
801: /* Handle ESP over UDP */
802: if (inp->inp_flags & INP_ESPINUDP_ALL) {
803: struct sockaddr *sa = (struct sockaddr *)src;
804:
805: if (udp4_espinudp(m, off, sa, inp->inp_socket) != 0) {
806: rcvcnt++;
807: goto bad;
808: }
809:
810: /* Normal UDP processing will take place */
811: }
812: #endif
813:
1.53 itojun 814: udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
815: rcvcnt++;
816: }
817:
818: bad:
819: return rcvcnt;
820: }
1.72 itojun 821: #endif
1.53 itojun 822:
823: #ifdef INET6
824: static int
1.119 matt 825: udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
826: struct mbuf *m, int off)
1.53 itojun 827: {
1.79 itojun 828: u_int16_t sport, dport;
1.53 itojun 829: int rcvcnt;
1.79 itojun 830: struct in6_addr src6, dst6;
831: const struct in_addr *dst4;
1.109 itojun 832: struct inpcb_hdr *inph;
1.53 itojun 833: struct in6pcb *in6p;
834:
835: rcvcnt = 0;
836: off += sizeof(struct udphdr); /* now, offset of payload */
837:
838: if (af != AF_INET && af != AF_INET6)
839: goto bad;
840: if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
841: goto bad;
842:
1.79 itojun 843: in6_embedscope(&src6, src, NULL, NULL);
844: sport = src->sin6_port;
845: in6_embedscope(&dst6, dst, NULL, NULL);
846: dport = dst->sin6_port;
1.88 itojun 847: dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
1.53 itojun 848:
1.79 itojun 849: if (IN6_IS_ADDR_MULTICAST(&dst6) ||
850: (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
1.53 itojun 851: /*
852: * Deliver a multicast or broadcast datagram to *all* sockets
853: * for which the local and remote addresses and ports match
854: * those of the incoming datagram. This allows more than
855: * one process to receive multi/broadcasts on the same port.
856: * (This really ought to be done for unicast datagrams as
857: * well, but that would cause problems with existing
858: * applications that open both address-specific sockets and
859: * a wildcard socket listening to the same port -- they would
860: * end up receiving duplicates of every unicast datagram.
861: * Those applications open the multiple sockets to overcome an
862: * inadequacy of the UDP socket interface, but for backwards
863: * compatibility we avoid the problem here rather than
864: * fixing the interface. Maybe 4.5BSD will remedy this?)
865: */
866:
867: /*
1.92 itojun 868: * KAME note: traditionally we dropped udpiphdr from mbuf here.
1.71 itojun 869: * we need udpiphdr for IPsec processing so we do that later.
1.53 itojun 870: */
871: /*
872: * Locate pcb(s) for datagram.
873: */
1.109 itojun 874: CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
875: in6p = (struct in6pcb *)inph;
876: if (in6p->in6p_af != AF_INET6)
877: continue;
878:
1.79 itojun 879: if (in6p->in6p_lport != dport)
1.53 itojun 880: continue;
881: if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1.90 itojun 882: if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
1.53 itojun 883: continue;
1.92 itojun 884: } else {
1.79 itojun 885: if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
1.85 itojun 886: (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1.55 itojun 887: continue;
888: }
1.53 itojun 889: if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1.79 itojun 890: if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
891: &src6) || in6p->in6p_fport != sport)
1.53 itojun 892: continue;
1.92 itojun 893: } else {
1.79 itojun 894: if (IN6_IS_ADDR_V4MAPPED(&src6) &&
1.85 itojun 895: (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1.55 itojun 896: continue;
897: }
1.53 itojun 898:
899: udp6_sendup(m, off, (struct sockaddr *)src,
900: in6p->in6p_socket);
901: rcvcnt++;
902:
903: /*
904: * Don't look for additional matches if this one does
905: * not have either the SO_REUSEPORT or SO_REUSEADDR
906: * socket options set. This heuristic avoids searching
907: * through all pcbs in the common case of a non-shared
908: * port. It assumes that an application will never
909: * clear these options after setting them.
910: */
911: if ((in6p->in6p_socket->so_options &
912: (SO_REUSEPORT|SO_REUSEADDR)) == 0)
913: break;
914: }
915: } else {
916: /*
917: * Locate pcb for datagram.
918: */
1.109 itojun 919: in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
1.79 itojun 920: &dst6, dport, 0);
1.53 itojun 921: if (in6p == 0) {
922: ++udpstat.udps_pcbhashmiss;
1.109 itojun 923: in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
1.82 itojun 924: if (in6p == 0)
1.53 itojun 925: return rcvcnt;
926: }
927:
928: udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
929: rcvcnt++;
930: }
931:
932: bad:
933: return rcvcnt;
934: }
935: #endif
1.1 cgd 936:
1.72 itojun 937: #ifdef INET
1.1 cgd 938: /*
939: * Notify a udp user of an asynchronous error;
940: * just wake up so that he can collect error status.
941: */
1.7 mycroft 942: static void
1.119 matt 943: udp_notify(struct inpcb *inp, int errno)
1.1 cgd 944: {
945: inp->inp_socket->so_error = errno;
946: sorwakeup(inp->inp_socket);
947: sowwakeup(inp->inp_socket);
948: }
949:
1.27 christos 950: void *
1.119 matt 951: udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
1.1 cgd 952: {
1.66 augustss 953: struct ip *ip = v;
954: struct udphdr *uh;
1.119 matt 955: void (*notify)(struct inpcb *, int) = udp_notify;
1.21 mycroft 956: int errno;
1.1 cgd 957:
1.53 itojun 958: if (sa->sa_family != AF_INET
959: || sa->sa_len != sizeof(struct sockaddr_in))
1.51 itojun 960: return NULL;
1.20 mycroft 961: if ((unsigned)cmd >= PRC_NCMDS)
1.27 christos 962: return NULL;
1.20 mycroft 963: errno = inetctlerrmap[cmd];
1.18 mycroft 964: if (PRC_IS_REDIRECT(cmd))
1.19 mycroft 965: notify = in_rtchange, ip = 0;
1.18 mycroft 966: else if (cmd == PRC_HOSTDEAD)
1.19 mycroft 967: ip = 0;
1.23 cgd 968: else if (errno == 0)
1.27 christos 969: return NULL;
1.19 mycroft 970: if (ip) {
1.1 cgd 971: uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
1.34 mycroft 972: in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
973: ip->ip_src, uh->uh_sport, errno, notify);
1.53 itojun 974:
975: /* XXX mapped address case */
1.19 mycroft 976: } else
1.34 mycroft 977: in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
978: notify);
1.27 christos 979: return NULL;
1.1 cgd 980: }
981:
1.7 mycroft 982: int
1.131 perry 983: udp_ctloutput(op, so, level, optname, mp)
1.130 manu 984: int op;
985: struct socket *so;
986: int level, optname;
987: struct mbuf **mp;
988: {
989: int s;
990: int error = 0;
991: struct mbuf *m;
992: struct inpcb *inp;
993: int family;
994:
995: family = so->so_proto->pr_domain->dom_family;
996:
997: s = splsoftnet();
998: switch (family) {
999: #ifdef INET
1000: case PF_INET:
1001: if (level != IPPROTO_UDP) {
1002: error = ip_ctloutput(op, so, level, optname, mp);
1003: goto end;
1004: }
1005: break;
1006: #endif
1007: #ifdef INET6
1008: case PF_INET6:
1009: if (level != IPPROTO_UDP) {
1010: error = ip6_ctloutput(op, so, level, optname, mp);
1011: goto end;
1012: }
1013: break;
1014: #endif
1015: default:
1016: error = EAFNOSUPPORT;
1017: goto end;
1018: break;
1019: }
1020:
1021:
1022: switch (op) {
1023: case PRCO_SETOPT:
1024: m = *mp;
1025: inp = sotoinpcb(so);
1026:
1027: switch (optname) {
1028: case UDP_ENCAP:
1029: if (m == NULL || m->m_len < sizeof (int)) {
1030: error = EINVAL;
1031: goto end;
1032: }
1.131 perry 1033:
1.130 manu 1034: switch(*mtod(m, int *)) {
1035: #ifdef IPSEC_NAT_T
1036: case 0:
1037: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1038: break;
1039:
1040: case UDP_ENCAP_ESPINUDP:
1041: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1042: inp->inp_flags |= INP_ESPINUDP;
1043: break;
1.131 perry 1044:
1.130 manu 1045: case UDP_ENCAP_ESPINUDP_NON_IKE:
1046: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1047: inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1048: break;
1049: #endif
1050: default:
1051: error = EINVAL;
1052: goto end;
1053: break;
1054: }
1055: break;
1056:
1057: default:
1058: error = ENOPROTOOPT;
1059: goto end;
1060: break;
1061: }
1062: break;
1063:
1064: default:
1065: error = EINVAL;
1066: goto end;
1067: break;
1.131 perry 1068: }
1069:
1.130 manu 1070: end:
1071: splx(s);
1072: return error;
1073: }
1.131 perry 1074:
1.130 manu 1075:
1076: int
1.27 christos 1077: udp_output(struct mbuf *m, ...)
1078: {
1.66 augustss 1079: struct inpcb *inp;
1080: struct udpiphdr *ui;
1.125 thorpej 1081: struct route *ro;
1.66 augustss 1082: int len = m->m_pkthdr.len;
1.31 mycroft 1083: int error = 0;
1.27 christos 1084: va_list ap;
1085:
1.98 matt 1086: MCLAIM(m, &udp_tx_mowner);
1.27 christos 1087: va_start(ap, m);
1088: inp = va_arg(ap, struct inpcb *);
1089: va_end(ap);
1.1 cgd 1090:
1091: /*
1092: * Calculate data length and get a mbuf
1093: * for UDP and IP headers.
1094: */
1.13 mycroft 1095: M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1096: if (m == 0) {
1097: error = ENOBUFS;
1.39 thorpej 1098: goto release;
1099: }
1100:
1101: /*
1102: * Compute the packet length of the IP header, and
1103: * punt if the length looks bogus.
1104: */
1.96 itojun 1105: if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1.39 thorpej 1106: error = EMSGSIZE;
1.13 mycroft 1107: goto release;
1108: }
1.1 cgd 1109:
1110: /*
1111: * Fill in mbuf with extended UDP header
1112: * and addresses and length put into network format.
1113: */
1114: ui = mtod(m, struct udpiphdr *);
1115: ui->ui_pr = IPPROTO_UDP;
1116: ui->ui_src = inp->inp_laddr;
1117: ui->ui_dst = inp->inp_faddr;
1118: ui->ui_sport = inp->inp_lport;
1119: ui->ui_dport = inp->inp_fport;
1.78 thorpej 1120: ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1.1 cgd 1121:
1.125 thorpej 1122: ro = &inp->inp_route;
1123:
1.1 cgd 1124: /*
1.78 thorpej 1125: * Set up checksum and output datagram.
1.1 cgd 1126: */
1127: if (udpcksum) {
1.78 thorpej 1128: /*
1129: * XXX Cache pseudo-header checksum part for
1130: * XXX "connected" UDP sockets.
1131: */
1132: ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1133: ui->ui_dst.s_addr, htons((u_int16_t)len +
1134: sizeof(struct udphdr) + IPPROTO_UDP));
1.135 yamt 1135: m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1.78 thorpej 1136: m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1137: } else
1138: ui->ui_sum = 0;
1.96 itojun 1139: ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1.1 cgd 1140: ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */
1141: ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */
1142: udpstat.udps_opackets++;
1.48 itojun 1143:
1.125 thorpej 1144: return (ip_output(m, inp->inp_options, ro,
1.12 mycroft 1145: inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1.107 itojun 1146: inp->inp_moptions, inp->inp_socket));
1.1 cgd 1147:
1148: release:
1149: m_freem(m);
1150: return (error);
1151: }
1152:
1.42 thorpej 1153: int udp_sendspace = 9216; /* really max datagram size */
1154: int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1.1 cgd 1155: /* 40 1K datagrams */
1156:
1157: /*ARGSUSED*/
1.7 mycroft 1158: int
1.119 matt 1159: udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1160: struct mbuf *control, struct proc *p)
1.1 cgd 1161: {
1.66 augustss 1162: struct inpcb *inp;
1.1 cgd 1163: int s;
1.66 augustss 1164: int error = 0;
1.1 cgd 1165:
1166: if (req == PRU_CONTROL)
1.31 mycroft 1167: return (in_control(so, (long)m, (caddr_t)nam,
1.30 mycroft 1168: (struct ifnet *)control, p));
1.59 thorpej 1169:
1.60 thorpej 1170: if (req == PRU_PURGEIF) {
1.80 itojun 1171: in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1.60 thorpej 1172: in_purgeif((struct ifnet *)control);
1173: in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1.59 thorpej 1174: return (0);
1175: }
1.31 mycroft 1176:
1177: s = splsoftnet();
1178: inp = sotoinpcb(so);
1.32 mycroft 1179: #ifdef DIAGNOSTIC
1180: if (req != PRU_SEND && req != PRU_SENDOOB && control)
1181: panic("udp_usrreq: unexpected control mbuf");
1182: #endif
1.31 mycroft 1183: if (inp == 0 && req != PRU_ATTACH) {
1184: error = EINVAL;
1185: goto release;
1186: }
1187:
1.1 cgd 1188: /*
1189: * Note: need to block udp_input while changing
1190: * the udp pcb queue and/or pcb addresses.
1191: */
1192: switch (req) {
1193:
1194: case PRU_ATTACH:
1.31 mycroft 1195: if (inp != 0) {
1196: error = EISCONN;
1.1 cgd 1197: break;
1198: }
1.98 matt 1199: #ifdef MBUFTRACE
1200: so->so_mowner = &udp_mowner;
1201: so->so_rcv.sb_mowner = &udp_rx_mowner;
1202: so->so_snd.sb_mowner = &udp_tx_mowner;
1203: #endif
1.31 mycroft 1204: if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1205: error = soreserve(so, udp_sendspace, udp_recvspace);
1206: if (error)
1207: break;
1208: }
1.18 mycroft 1209: error = in_pcballoc(so, &udbtable);
1.1 cgd 1210: if (error)
1211: break;
1.31 mycroft 1212: inp = sotoinpcb(so);
1213: inp->inp_ip.ip_ttl = ip_defttl;
1.1 cgd 1214: break;
1215:
1216: case PRU_DETACH:
1.31 mycroft 1217: in_pcbdetach(inp);
1.1 cgd 1218: break;
1219:
1220: case PRU_BIND:
1.31 mycroft 1221: error = in_pcbbind(inp, nam, p);
1.1 cgd 1222: break;
1223:
1224: case PRU_LISTEN:
1225: error = EOPNOTSUPP;
1226: break;
1227:
1228: case PRU_CONNECT:
1.143 ! dsl 1229: error = in_pcbconnect(inp, nam, p);
1.31 mycroft 1230: if (error)
1.1 cgd 1231: break;
1.31 mycroft 1232: soisconnected(so);
1.1 cgd 1233: break;
1234:
1235: case PRU_CONNECT2:
1236: error = EOPNOTSUPP;
1237: break;
1238:
1239: case PRU_DISCONNECT:
1.31 mycroft 1240: /*soisdisconnected(so);*/
1241: so->so_state &= ~SS_ISCONNECTED; /* XXX */
1.1 cgd 1242: in_pcbdisconnect(inp);
1.34 mycroft 1243: inp->inp_laddr = zeroin_addr; /* XXX */
1.35 mycroft 1244: in_pcbstate(inp, INP_BOUND); /* XXX */
1.1 cgd 1245: break;
1246:
1247: case PRU_SHUTDOWN:
1248: socantsendmore(so);
1249: break;
1250:
1.31 mycroft 1251: case PRU_RCVD:
1252: error = EOPNOTSUPP;
1.1 cgd 1253: break;
1254:
1.31 mycroft 1255: case PRU_SEND:
1.32 mycroft 1256: if (control && control->m_len) {
1257: m_freem(control);
1258: m_freem(m);
1259: error = EINVAL;
1260: break;
1261: }
1.31 mycroft 1262: {
1.35 mycroft 1263: struct in_addr laddr; /* XXX */
1.1 cgd 1264:
1.31 mycroft 1265: if (nam) {
1.35 mycroft 1266: laddr = inp->inp_laddr; /* XXX */
1.31 mycroft 1267: if ((so->so_state & SS_ISCONNECTED) != 0) {
1268: error = EISCONN;
1.32 mycroft 1269: goto die;
1.31 mycroft 1270: }
1.143 ! dsl 1271: error = in_pcbconnect(inp, nam, p);
1.100 matt 1272: if (error)
1273: goto die;
1.31 mycroft 1274: } else {
1275: if ((so->so_state & SS_ISCONNECTED) == 0) {
1276: error = ENOTCONN;
1.32 mycroft 1277: goto die;
1.31 mycroft 1278: }
1279: }
1.33 mycroft 1280: error = udp_output(m, inp);
1.100 matt 1281: m = NULL;
1.31 mycroft 1282: if (nam) {
1283: in_pcbdisconnect(inp);
1.35 mycroft 1284: inp->inp_laddr = laddr; /* XXX */
1285: in_pcbstate(inp, INP_BOUND); /* XXX */
1.31 mycroft 1286: }
1.100 matt 1287: die:
1288: if (m)
1289: m_freem(m);
1.31 mycroft 1290: }
1.1 cgd 1291: break;
1292:
1293: case PRU_SENSE:
1294: /*
1295: * stat: don't bother with a blocksize.
1296: */
1.31 mycroft 1297: splx(s);
1.1 cgd 1298: return (0);
1299:
1.31 mycroft 1300: case PRU_RCVOOB:
1301: error = EOPNOTSUPP;
1302: break;
1303:
1.1 cgd 1304: case PRU_SENDOOB:
1.32 mycroft 1305: m_freem(control);
1.31 mycroft 1306: m_freem(m);
1.1 cgd 1307: error = EOPNOTSUPP;
1308: break;
1309:
1.31 mycroft 1310: case PRU_SOCKADDR:
1311: in_setsockaddr(inp, nam);
1312: break;
1313:
1314: case PRU_PEERADDR:
1315: in_setpeeraddr(inp, nam);
1316: break;
1.1 cgd 1317:
1318: default:
1319: panic("udp_usrreq");
1320: }
1321:
1322: release:
1.31 mycroft 1323: splx(s);
1.1 cgd 1324: return (error);
1.13 mycroft 1325: }
1326:
1327: /*
1328: * Sysctl for udp variables.
1329: */
1.114 atatat 1330: SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1.13 mycroft 1331: {
1.114 atatat 1332:
1.116 atatat 1333: sysctl_createv(clog, 0, NULL, NULL,
1334: CTLFLAG_PERMANENT,
1.114 atatat 1335: CTLTYPE_NODE, "net", NULL,
1336: NULL, 0, NULL, 0,
1337: CTL_NET, CTL_EOL);
1.116 atatat 1338: sysctl_createv(clog, 0, NULL, NULL,
1339: CTLFLAG_PERMANENT,
1.114 atatat 1340: CTLTYPE_NODE, "inet", NULL,
1341: NULL, 0, NULL, 0,
1342: CTL_NET, PF_INET, CTL_EOL);
1.116 atatat 1343: sysctl_createv(clog, 0, NULL, NULL,
1344: CTLFLAG_PERMANENT,
1.122 atatat 1345: CTLTYPE_NODE, "udp",
1346: SYSCTL_DESCR("UDPv4 related settings"),
1.114 atatat 1347: NULL, 0, NULL, 0,
1348: CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1349:
1.116 atatat 1350: sysctl_createv(clog, 0, NULL, NULL,
1351: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1352: CTLTYPE_INT, "checksum",
1.123 heas 1353: SYSCTL_DESCR("Compute UDP checksums"),
1.114 atatat 1354: NULL, 0, &udpcksum, 0,
1355: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1356: CTL_EOL);
1.116 atatat 1357: sysctl_createv(clog, 0, NULL, NULL,
1358: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1359: CTLTYPE_INT, "sendspace",
1360: SYSCTL_DESCR("Default UDP send buffer size"),
1.114 atatat 1361: NULL, 0, &udp_sendspace, 0,
1362: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1363: CTL_EOL);
1.116 atatat 1364: sysctl_createv(clog, 0, NULL, NULL,
1365: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1366: CTLTYPE_INT, "recvspace",
1367: SYSCTL_DESCR("Default UDP receive buffer size"),
1.114 atatat 1368: NULL, 0, &udp_recvspace, 0,
1369: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1370: CTL_EOL);
1.125 thorpej 1371: sysctl_createv(clog, 0, NULL, NULL,
1372: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1373: CTLTYPE_INT, "do_loopback_cksum",
1374: SYSCTL_DESCR("Perform UDP checksum on loopback"),
1375: NULL, 0, &udp_do_loopback_cksum, 0,
1376: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1377: CTL_EOL);
1.132 atatat 1378: sysctl_createv(clog, 0, NULL, NULL,
1379: CTLFLAG_PERMANENT,
1.134 atatat 1380: CTLTYPE_STRUCT, "pcblist",
1.132 atatat 1381: SYSCTL_DESCR("UDP protocol control block list"),
1382: sysctl_inpcblist, 0, &udbtable, 0,
1383: CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1384: CTL_EOL);
1.139 elad 1385: sysctl_createv(clog, 0, NULL, NULL,
1386: CTLFLAG_PERMANENT,
1387: CTLTYPE_STRUCT, "stats",
1388: SYSCTL_DESCR("UDP statistics"),
1389: NULL, 0, &udpstat, sizeof(udpstat),
1390: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1391: CTL_EOL);
1.1 cgd 1392: }
1.72 itojun 1393: #endif
1.130 manu 1394:
1395: #if (defined INET && defined IPSEC_NAT_T)
1396: /*
1397: * Returns:
1398: * 1 if the packet was processed
1399: * 0 if normal UDP processing should take place
1400: */
1401: static int
1402: udp4_espinudp(m, off, src, so)
1403: struct mbuf *m;
1404: int off;
1405: struct sockaddr *src;
1406: struct socket *so;
1407: {
1408: size_t len;
1409: caddr_t data;
1410: struct inpcb *inp;
1411: size_t skip = 0;
1412: size_t minlen;
1413: size_t iphdrlen;
1414: struct ip *ip;
1415: struct mbuf *n;
1.136 manu 1416: struct m_tag *tag;
1417: struct udphdr *udphdr;
1418: u_int16_t sport, dport;
1.130 manu 1419:
1.131 perry 1420: /*
1.130 manu 1421: * Collapse the mbuf chain if the first mbuf is too short
1422: * The longest case is: UDP + non ESP marker + ESP
1423: */
1424: minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1425: if (minlen > m->m_pkthdr.len)
1426: minlen = m->m_pkthdr.len;
1427:
1428: if (m->m_len < minlen) {
1429: if ((m = m_pullup(m, minlen)) == NULL) {
1430: printf("udp4_espinudp: m_pullup failed\n");
1431: return 0;
1432: }
1433: }
1434:
1.131 perry 1435: len = m->m_len - off;
1.130 manu 1436: data = mtod(m, caddr_t) + off;
1437: inp = sotoinpcb(so);
1438:
1439: /* Ignore keepalive packets */
1440: if ((len == 1) && (data[0] == '\xff')) {
1441: return 1;
1442: }
1443:
1.131 perry 1444: /*
1445: * Check that the payload is long enough to hold
1.130 manu 1446: * an ESP header and compute the length of encapsulation
1.131 perry 1447: * header to remove
1.130 manu 1448: */
1449: if (inp->inp_flags & INP_ESPINUDP) {
1450: u_int32_t *st = (u_int32_t *)data;
1451:
1452: if ((len <= sizeof(struct esp)) || (*st == 0))
1453: return 0; /* Normal UDP processing */
1454:
1455: skip = sizeof(struct udphdr);
1456: }
1457:
1458: if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1.142 kleink 1459: u_int32_t *st = (u_int32_t *)data;
1.130 manu 1460:
1461: if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1.142 kleink 1462: || ((st[0] | st[1]) != 0))
1.130 manu 1463: return 0; /* Normal UDP processing */
1.131 perry 1464:
1.130 manu 1465: skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1466: }
1467:
1468: /*
1.136 manu 1469: * Get the UDP ports. They are handled in network
1470: * order everywhere in IPSEC_NAT_T code.
1471: */
1472: udphdr = (struct udphdr *)(data - skip);
1473: sport = udphdr->uh_sport;
1474: dport = udphdr->uh_dport;
1475:
1476: /*
1.130 manu 1477: * Remove the UDP header (and possibly the non ESP marker)
1478: * IP header lendth is iphdrlen
1.131 perry 1479: * Before:
1.130 manu 1480: * <--- off --->
1481: * +----+------+-----+
1482: * | IP | UDP | ESP |
1483: * +----+------+-----+
1484: * <-skip->
1485: * After:
1486: * +----+-----+
1487: * | IP | ESP |
1488: * +----+-----+
1489: * <-skip->
1490: */
1491: iphdrlen = off - sizeof(struct udphdr);
1492: memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen);
1493: m_adj(m, skip);
1494:
1495: ip = mtod(m, struct ip *);
1496: ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1497: ip->ip_p = IPPROTO_ESP;
1498:
1499: /*
1.131 perry 1500: * Copy the mbuf to avoid multiple free, as both
1501: * esp4_input (which we call) and udp_input (which
1.130 manu 1502: * called us) free the mbuf.
1503: */
1504: if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1505: printf("udp4_espinudp: m_dup failed\n");
1506: return 0;
1507: }
1508:
1.136 manu 1509: /*
1510: * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1511: * the source UDP port. This is required if we want
1512: * to select the right SPD for multiple hosts behind
1513: * same NAT
1514: */
1.137 manu 1515: if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1516: sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1517: printf("udp4_espinudp: m_tag_get failed\n");
1.138 manu 1518: m_freem(n);
1.137 manu 1519: return 0;
1520: }
1.136 manu 1521: ((u_int16_t *)(tag + 1))[0] = sport;
1522: ((u_int16_t *)(tag + 1))[1] = dport;
1523: m_tag_prepend(n, tag);
1524:
1.130 manu 1525: esp4_input(n, iphdrlen);
1526:
1527: /* We handled it, it shoudln't be handled by UDP */
1528: return 1;
1529: }
1530: #endif
CVSweb <webmaster@jp.NetBSD.org>