Annotation of src/sys/netinet/udp_usrreq.c, Revision 1.140
1.140 ! yamt 1: /* $NetBSD: udp_usrreq.c,v 1.139 2005/08/05 09:21:26 elad 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>
1.140 ! yamt 64: __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.139 2005/08/05 09:21:26 elad 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
1.93 matt 149:
150: struct inpcbtable udbtable;
151: struct udpstat udpstat;
1.8 mycroft 152:
1.72 itojun 153: #ifdef INET
1.130 manu 154: #ifdef IPSEC_NAT_T
155: static int udp4_espinudp (struct mbuf *, int, struct sockaddr *,
156: struct socket *);
157: #endif
1.119 matt 158: static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
159: struct socket *);
160: static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
161: struct mbuf *, int);
1.129 yamt 162: static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
1.72 itojun 163: #endif
1.53 itojun 164: #ifdef INET6
1.119 matt 165: static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
166: struct socket *);
167: static int udp6_realinput (int, struct sockaddr_in6 *,
168: struct sockaddr_in6 *, struct mbuf *, int);
1.129 yamt 169: static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
1.53 itojun 170: #endif
1.72 itojun 171: #ifdef INET
1.119 matt 172: static void udp_notify (struct inpcb *, int);
1.72 itojun 173: #endif
1.7 mycroft 174:
1.26 mycroft 175: #ifndef UDBHASHSIZE
176: #define UDBHASHSIZE 128
177: #endif
178: int udbhashsize = UDBHASHSIZE;
179:
1.98 matt 180: #ifdef MBUFTRACE
181: struct mowner udp_mowner = { "udp" };
182: struct mowner udp_rx_mowner = { "udp", "rx" };
183: struct mowner udp_tx_mowner = { "udp", "tx" };
184: #endif
185:
1.78 thorpej 186: #ifdef UDP_CSUM_COUNTERS
187: #include <sys/device.h>
188:
1.140 ! yamt 189: #if defined(INET)
1.78 thorpej 190: struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
191: NULL, "udp", "hwcsum bad");
192: struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
193: NULL, "udp", "hwcsum ok");
194: struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
195: NULL, "udp", "hwcsum data");
196: struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
197: NULL, "udp", "swcsum");
198:
1.120 matt 199: EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
200: EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
201: EVCNT_ATTACH_STATIC(udp_hwcsum_data);
202: EVCNT_ATTACH_STATIC(udp_swcsum);
1.140 ! yamt 203: #endif /* defined(INET) */
! 204:
! 205: #if defined(INET6)
! 206: struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
! 207: NULL, "udp6", "hwcsum bad");
! 208: struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
! 209: NULL, "udp6", "hwcsum ok");
! 210: struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
! 211: NULL, "udp6", "hwcsum data");
! 212: struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
! 213: NULL, "udp6", "swcsum");
! 214:
! 215: EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
! 216: EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
! 217: EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
! 218: EVCNT_ATTACH_STATIC(udp6_swcsum);
! 219: #endif /* defined(INET6) */
! 220:
! 221: #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
1.120 matt 222:
1.78 thorpej 223: #else
224:
225: #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */
226:
227: #endif /* UDP_CSUM_COUNTERS */
228:
1.7 mycroft 229: void
1.119 matt 230: udp_init(void)
1.1 cgd 231: {
1.18 mycroft 232:
1.35 mycroft 233: in_pcbinit(&udbtable, udbhashsize, udbhashsize);
1.78 thorpej 234:
1.98 matt 235: MOWNER_ATTACH(&udp_tx_mowner);
236: MOWNER_ATTACH(&udp_rx_mowner);
237: MOWNER_ATTACH(&udp_mowner);
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:
332: udpstat.udps_badsum++;
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);
1.53 itojun 354: udpstat.udps_ipackets++;
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) {
362: udpstat.udps_hdrops++;
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)) {
1.53 itojun 379: udpstat.udps_badlen++;
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. */
392: bzero(&src, sizeof(src));
393: src.sin_family = AF_INET;
394: src.sin_len = sizeof(struct sockaddr_in);
395: bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
396: src.sin_port = uh->uh_sport;
397: bzero(&dst, sizeof(dst));
398: dst.sin_family = AF_INET;
399: dst.sin_len = sizeof(struct sockaddr_in);
400: bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
401: dst.sin_port = uh->uh_dport;
402:
403: n = udp4_realinput(&src, &dst, m, iphlen);
404: #ifdef INET6
405: if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
406: struct sockaddr_in6 src6, dst6;
407:
408: bzero(&src6, sizeof(src6));
409: src6.sin6_family = AF_INET6;
410: src6.sin6_len = sizeof(struct sockaddr_in6);
411: src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
412: bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
413: sizeof(ip->ip_src));
414: src6.sin6_port = uh->uh_sport;
415: bzero(&dst6, sizeof(dst6));
416: dst6.sin6_family = AF_INET6;
417: dst6.sin6_len = sizeof(struct sockaddr_in6);
418: dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
419: bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
420: sizeof(ip->ip_dst));
421: dst6.sin6_port = uh->uh_dport;
422:
423: n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
424: }
425: #endif
426:
427: if (n == 0) {
428: if (m->m_flags & (M_BCAST | M_MCAST)) {
429: udpstat.udps_noportbcast++;
430: goto bad;
431: }
1.61 itojun 432: udpstat.udps_noport++;
1.64 ws 433: #ifdef IPKDB
1.53 itojun 434: if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
435: m, iphlen + sizeof(struct udphdr),
436: m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
437: /*
438: * It was a debugger connect packet,
439: * just drop it now
440: */
441: goto bad;
442: }
443: #endif
444: icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
445: m = NULL;
446: }
447:
448: bad:
449: if (m)
450: m_freem(m);
1.78 thorpej 451: return;
452:
453: badcsum:
454: m_freem(m);
1.53 itojun 455: }
1.72 itojun 456: #endif
1.53 itojun 457:
458: #ifdef INET6
1.129 yamt 459: static int
460: udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
461: {
462:
1.140 ! yamt 463: /*
! 464: * XXX it's better to record and check if this mbuf is
! 465: * already checked.
! 466: */
! 467:
1.129 yamt 468: if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
469: goto good;
470: }
471: if (uh->uh_sum == 0) {
472: udp6stat.udp6s_nosum++;
473: goto bad;
474: }
1.140 ! yamt 475:
! 476: switch (m->m_pkthdr.csum_flags &
! 477: ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
! 478: M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
! 479: case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
! 480: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
1.129 yamt 481: udp6stat.udp6s_badsum++;
482: goto bad;
1.140 ! yamt 483:
! 484: #if 0 /* notyet */
! 485: case M_CSUM_UDPv6|M_CSUM_DATA:
! 486: #endif
! 487:
! 488: case M_CSUM_UDPv6:
! 489: /* Checksum was okay. */
! 490: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
! 491: break;
! 492:
! 493: default:
! 494: /*
! 495: * Need to compute it ourselves. Maybe skip checksum
! 496: * on loopback interfaces.
! 497: */
! 498: UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
! 499: if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
! 500: udp6stat.udp6s_badsum++;
! 501: goto bad;
! 502: }
1.129 yamt 503: }
504:
505: good:
506: return 0;
507: bad:
508: return -1;
509: }
510:
1.53 itojun 511: int
1.119 matt 512: udp6_input(struct mbuf **mp, int *offp, int proto)
1.53 itojun 513: {
514: struct mbuf *m = *mp;
515: int off = *offp;
516: struct sockaddr_in6 src, dst;
517: struct ip6_hdr *ip6;
518: struct udphdr *uh;
519: u_int32_t plen, ulen;
520:
1.76 itojun 521: ip6 = mtod(m, struct ip6_hdr *);
522:
1.53 itojun 523: #if defined(NFAITH) && 0 < NFAITH
1.76 itojun 524: if (faithprefix(&ip6->ip6_dst)) {
525: /* send icmp6 host unreach? */
526: m_freem(m);
527: return IPPROTO_DONE;
1.53 itojun 528: }
529: #endif
530:
531: udp6stat.udp6s_ipackets++;
532:
533: /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */
534: plen = m->m_pkthdr.len - off;
535: IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
536: if (uh == NULL) {
537: ip6stat.ip6s_tooshort++;
538: return IPPROTO_DONE;
539: }
1.95 thorpej 540: KASSERT(UDP_HDR_ALIGNED_P(uh));
1.53 itojun 541: ulen = ntohs((u_short)uh->uh_ulen);
1.69 itojun 542: /*
543: * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
544: * iff payload length > 0xffff.
545: */
1.53 itojun 546: if (ulen == 0 && plen > 0xffff)
547: ulen = plen;
548:
549: if (plen != ulen) {
550: udp6stat.udp6s_badlen++;
1.54 itojun 551: goto bad;
552: }
553:
1.57 itojun 554: /* destination port of 0 is illegal, based on RFC768. */
555: if (uh->uh_dport == 0)
556: goto bad;
557:
1.54 itojun 558: /* Be proactive about malicious use of IPv4 mapped address */
559: if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
560: IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
561: /* XXX stat */
1.53 itojun 562: goto bad;
563: }
564:
565: /*
1.125 thorpej 566: * Checksum extended UDP header and data. Maybe skip checksum
567: * on loopback interfaces.
1.53 itojun 568: */
1.129 yamt 569: if (udp6_input_checksum(m, uh, off, ulen))
570: goto bad;
1.53 itojun 571:
572: /*
573: * Construct source and dst sockaddrs.
574: * Note that ifindex (s6_addr16[1]) is already filled.
575: */
576: bzero(&src, sizeof(src));
577: src.sin6_family = AF_INET6;
578: src.sin6_len = sizeof(struct sockaddr_in6);
1.69 itojun 579: /* KAME hack: recover scopeid */
580: (void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1.53 itojun 581: src.sin6_port = uh->uh_sport;
582: bzero(&dst, sizeof(dst));
583: dst.sin6_family = AF_INET6;
584: dst.sin6_len = sizeof(struct sockaddr_in6);
1.69 itojun 585: /* KAME hack: recover scopeid */
586: (void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1.53 itojun 587: dst.sin6_port = uh->uh_dport;
588:
589: if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
590: if (m->m_flags & M_MCAST) {
591: udp6stat.udp6s_noportmcast++;
592: goto bad;
593: }
1.61 itojun 594: udp6stat.udp6s_noport++;
1.53 itojun 595: icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
596: m = NULL;
597: }
598:
599: bad:
600: if (m)
601: m_freem(m);
602: return IPPROTO_DONE;
603: }
604: #endif
605:
1.72 itojun 606: #ifdef INET
1.53 itojun 607: static void
1.119 matt 608: udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
609: struct sockaddr *src, struct socket *so)
1.53 itojun 610: {
611: struct mbuf *opts = NULL;
612: struct mbuf *n;
613: struct inpcb *inp = NULL;
614:
615: if (!so)
616: return;
617: switch (so->so_proto->pr_domain->dom_family) {
618: case AF_INET:
619: inp = sotoinpcb(so);
620: break;
621: #ifdef INET6
622: case AF_INET6:
623: break;
624: #endif
625: default:
626: return;
627: }
628:
1.105 jonathan 629: #if defined(IPSEC) || defined(FAST_IPSEC)
1.53 itojun 630: /* check AH/ESP integrity. */
631: if (so != NULL && ipsec4_in_reject_so(m, so)) {
632: ipsecstat.in_polvio++;
1.110 itojun 633: if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
634: icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
635: 0, 0);
1.53 itojun 636: return;
637: }
638: #endif /*IPSEC*/
639:
640: if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
641: if (inp && (inp->inp_flags & INP_CONTROLOPTS
642: || so->so_options & SO_TIMESTAMP)) {
643: struct ip *ip = mtod(n, struct ip *);
644: ip_savecontrol(inp, &opts, ip, n);
645: }
646:
647: m_adj(n, off);
648: if (sbappendaddr(&so->so_rcv, src, n,
649: opts) == 0) {
650: m_freem(n);
651: if (opts)
652: m_freem(opts);
1.124 darrenr 653: so->so_rcv.sb_overflowed++;
1.81 enami 654: udpstat.udps_fullsock++;
1.53 itojun 655: } else
656: sorwakeup(so);
657: }
658: }
1.72 itojun 659: #endif
1.53 itojun 660:
661: #ifdef INET6
662: static void
1.119 matt 663: udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
664: struct sockaddr *src, struct socket *so)
1.53 itojun 665: {
666: struct mbuf *opts = NULL;
667: struct mbuf *n;
668: struct in6pcb *in6p = NULL;
669:
670: if (!so)
671: return;
672: if (so->so_proto->pr_domain->dom_family != AF_INET6)
673: return;
674: in6p = sotoin6pcb(so);
675:
1.105 jonathan 676: #if defined(IPSEC) || defined(FAST_IPSEC)
1.53 itojun 677: /* check AH/ESP integrity. */
678: if (so != NULL && ipsec6_in_reject_so(m, so)) {
679: ipsec6stat.in_polvio++;
1.111 mycroft 680: if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
1.110 itojun 681: icmp6_error(n, ICMP6_DST_UNREACH,
682: ICMP6_DST_UNREACH_ADMIN, 0);
1.53 itojun 683: return;
684: }
685: #endif /*IPSEC*/
686:
687: if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
688: if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
689: || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
690: struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
691: ip6_savecontrol(in6p, &opts, ip6, n);
692: }
693:
694: m_adj(n, off);
695: if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
696: m_freem(n);
697: if (opts)
698: m_freem(opts);
1.124 darrenr 699: so->so_rcv.sb_overflowed++;
1.53 itojun 700: udp6stat.udp6s_fullsock++;
701: } else
702: sorwakeup(so);
703: }
704: }
705: #endif
706:
1.72 itojun 707: #ifdef INET
1.53 itojun 708: static int
1.119 matt 709: udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
710: struct mbuf *m, int off /* offset of udphdr */)
1.53 itojun 711: {
712: u_int16_t *sport, *dport;
713: int rcvcnt;
714: struct in_addr *src4, *dst4;
1.109 itojun 715: struct inpcb_hdr *inph;
1.53 itojun 716: struct inpcb *inp;
717:
718: rcvcnt = 0;
719: off += sizeof(struct udphdr); /* now, offset of payload */
720:
721: if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
722: goto bad;
723:
724: src4 = &src->sin_addr;
725: sport = &src->sin_port;
726: dst4 = &dst->sin_addr;
727: dport = &dst->sin_port;
728:
1.73 itojun 729: if (IN_MULTICAST(dst4->s_addr) ||
1.53 itojun 730: in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
731: /*
732: * Deliver a multicast or broadcast datagram to *all* sockets
733: * for which the local and remote addresses and ports match
734: * those of the incoming datagram. This allows more than
735: * one process to receive multi/broadcasts on the same port.
736: * (This really ought to be done for unicast datagrams as
737: * well, but that would cause problems with existing
738: * applications that open both address-specific sockets and
739: * a wildcard socket listening to the same port -- they would
740: * end up receiving duplicates of every unicast datagram.
741: * Those applications open the multiple sockets to overcome an
742: * inadequacy of the UDP socket interface, but for backwards
743: * compatibility we avoid the problem here rather than
744: * fixing the interface. Maybe 4.5BSD will remedy this?)
745: */
746:
747: /*
1.92 itojun 748: * KAME note: traditionally we dropped udpiphdr from mbuf here.
1.71 itojun 749: * we need udpiphdr for IPsec processing so we do that later.
1.53 itojun 750: */
751: /*
752: * Locate pcb(s) for datagram.
753: */
1.109 itojun 754: CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
755: inp = (struct inpcb *)inph;
756: if (inp->inp_af != AF_INET)
757: continue;
758:
1.53 itojun 759: if (inp->inp_lport != *dport)
760: continue;
761: if (!in_nullhost(inp->inp_laddr)) {
762: if (!in_hosteq(inp->inp_laddr, *dst4))
763: continue;
764: }
765: if (!in_nullhost(inp->inp_faddr)) {
766: if (!in_hosteq(inp->inp_faddr, *src4) ||
767: inp->inp_fport != *sport)
768: continue;
769: }
770:
771: udp4_sendup(m, off, (struct sockaddr *)src,
772: inp->inp_socket);
773: rcvcnt++;
774:
775: /*
776: * Don't look for additional matches if this one does
777: * not have either the SO_REUSEPORT or SO_REUSEADDR
778: * socket options set. This heuristic avoids searching
779: * through all pcbs in the common case of a non-shared
780: * port. It assumes that an application will never
781: * clear these options after setting them.
782: */
783: if ((inp->inp_socket->so_options &
784: (SO_REUSEPORT|SO_REUSEADDR)) == 0)
785: break;
786: }
787: } else {
788: /*
789: * Locate pcb for datagram.
790: */
791: inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
792: if (inp == 0) {
793: ++udpstat.udps_pcbhashmiss;
794: inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
1.82 itojun 795: if (inp == 0)
1.53 itojun 796: return rcvcnt;
797: }
798:
1.130 manu 799: #ifdef IPSEC_NAT_T
800: /* Handle ESP over UDP */
801: if (inp->inp_flags & INP_ESPINUDP_ALL) {
802: struct sockaddr *sa = (struct sockaddr *)src;
803:
804: if (udp4_espinudp(m, off, sa, inp->inp_socket) != 0) {
805: rcvcnt++;
806: goto bad;
807: }
808:
809: /* Normal UDP processing will take place */
810: }
811: #endif
812:
1.53 itojun 813: udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
814: rcvcnt++;
815: }
816:
817: bad:
818: return rcvcnt;
819: }
1.72 itojun 820: #endif
1.53 itojun 821:
822: #ifdef INET6
823: static int
1.119 matt 824: udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
825: struct mbuf *m, int off)
1.53 itojun 826: {
1.79 itojun 827: u_int16_t sport, dport;
1.53 itojun 828: int rcvcnt;
1.79 itojun 829: struct in6_addr src6, dst6;
830: const struct in_addr *dst4;
1.109 itojun 831: struct inpcb_hdr *inph;
1.53 itojun 832: struct in6pcb *in6p;
833:
834: rcvcnt = 0;
835: off += sizeof(struct udphdr); /* now, offset of payload */
836:
837: if (af != AF_INET && af != AF_INET6)
838: goto bad;
839: if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
840: goto bad;
841:
1.79 itojun 842: in6_embedscope(&src6, src, NULL, NULL);
843: sport = src->sin6_port;
844: in6_embedscope(&dst6, dst, NULL, NULL);
845: dport = dst->sin6_port;
1.88 itojun 846: dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
1.53 itojun 847:
1.79 itojun 848: if (IN6_IS_ADDR_MULTICAST(&dst6) ||
849: (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
1.53 itojun 850: /*
851: * Deliver a multicast or broadcast datagram to *all* sockets
852: * for which the local and remote addresses and ports match
853: * those of the incoming datagram. This allows more than
854: * one process to receive multi/broadcasts on the same port.
855: * (This really ought to be done for unicast datagrams as
856: * well, but that would cause problems with existing
857: * applications that open both address-specific sockets and
858: * a wildcard socket listening to the same port -- they would
859: * end up receiving duplicates of every unicast datagram.
860: * Those applications open the multiple sockets to overcome an
861: * inadequacy of the UDP socket interface, but for backwards
862: * compatibility we avoid the problem here rather than
863: * fixing the interface. Maybe 4.5BSD will remedy this?)
864: */
865:
866: /*
1.92 itojun 867: * KAME note: traditionally we dropped udpiphdr from mbuf here.
1.71 itojun 868: * we need udpiphdr for IPsec processing so we do that later.
1.53 itojun 869: */
870: /*
871: * Locate pcb(s) for datagram.
872: */
1.109 itojun 873: CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
874: in6p = (struct in6pcb *)inph;
875: if (in6p->in6p_af != AF_INET6)
876: continue;
877:
1.79 itojun 878: if (in6p->in6p_lport != dport)
1.53 itojun 879: continue;
880: if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1.90 itojun 881: if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
1.53 itojun 882: continue;
1.92 itojun 883: } else {
1.79 itojun 884: if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
1.85 itojun 885: (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1.55 itojun 886: continue;
887: }
1.53 itojun 888: if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1.79 itojun 889: if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
890: &src6) || in6p->in6p_fport != sport)
1.53 itojun 891: continue;
1.92 itojun 892: } else {
1.79 itojun 893: if (IN6_IS_ADDR_V4MAPPED(&src6) &&
1.85 itojun 894: (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1.55 itojun 895: continue;
896: }
1.53 itojun 897:
898: udp6_sendup(m, off, (struct sockaddr *)src,
899: in6p->in6p_socket);
900: rcvcnt++;
901:
902: /*
903: * Don't look for additional matches if this one does
904: * not have either the SO_REUSEPORT or SO_REUSEADDR
905: * socket options set. This heuristic avoids searching
906: * through all pcbs in the common case of a non-shared
907: * port. It assumes that an application will never
908: * clear these options after setting them.
909: */
910: if ((in6p->in6p_socket->so_options &
911: (SO_REUSEPORT|SO_REUSEADDR)) == 0)
912: break;
913: }
914: } else {
915: /*
916: * Locate pcb for datagram.
917: */
1.109 itojun 918: in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
1.79 itojun 919: &dst6, dport, 0);
1.53 itojun 920: if (in6p == 0) {
921: ++udpstat.udps_pcbhashmiss;
1.109 itojun 922: in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
1.82 itojun 923: if (in6p == 0)
1.53 itojun 924: return rcvcnt;
925: }
926:
927: udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
928: rcvcnt++;
929: }
930:
931: bad:
932: return rcvcnt;
933: }
934: #endif
1.1 cgd 935:
1.72 itojun 936: #ifdef INET
1.1 cgd 937: /*
938: * Notify a udp user of an asynchronous error;
939: * just wake up so that he can collect error status.
940: */
1.7 mycroft 941: static void
1.119 matt 942: udp_notify(struct inpcb *inp, int errno)
1.1 cgd 943: {
944: inp->inp_socket->so_error = errno;
945: sorwakeup(inp->inp_socket);
946: sowwakeup(inp->inp_socket);
947: }
948:
1.27 christos 949: void *
1.119 matt 950: udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
1.1 cgd 951: {
1.66 augustss 952: struct ip *ip = v;
953: struct udphdr *uh;
1.119 matt 954: void (*notify)(struct inpcb *, int) = udp_notify;
1.21 mycroft 955: int errno;
1.1 cgd 956:
1.53 itojun 957: if (sa->sa_family != AF_INET
958: || sa->sa_len != sizeof(struct sockaddr_in))
1.51 itojun 959: return NULL;
1.20 mycroft 960: if ((unsigned)cmd >= PRC_NCMDS)
1.27 christos 961: return NULL;
1.20 mycroft 962: errno = inetctlerrmap[cmd];
1.18 mycroft 963: if (PRC_IS_REDIRECT(cmd))
1.19 mycroft 964: notify = in_rtchange, ip = 0;
1.18 mycroft 965: else if (cmd == PRC_HOSTDEAD)
1.19 mycroft 966: ip = 0;
1.23 cgd 967: else if (errno == 0)
1.27 christos 968: return NULL;
1.19 mycroft 969: if (ip) {
1.1 cgd 970: uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
1.34 mycroft 971: in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
972: ip->ip_src, uh->uh_sport, errno, notify);
1.53 itojun 973:
974: /* XXX mapped address case */
1.19 mycroft 975: } else
1.34 mycroft 976: in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
977: notify);
1.27 christos 978: return NULL;
1.1 cgd 979: }
980:
1.7 mycroft 981: int
1.131 perry 982: udp_ctloutput(op, so, level, optname, mp)
1.130 manu 983: int op;
984: struct socket *so;
985: int level, optname;
986: struct mbuf **mp;
987: {
988: int s;
989: int error = 0;
990: struct mbuf *m;
991: struct inpcb *inp;
992: int family;
993:
994: family = so->so_proto->pr_domain->dom_family;
995:
996: s = splsoftnet();
997: switch (family) {
998: #ifdef INET
999: case PF_INET:
1000: if (level != IPPROTO_UDP) {
1001: error = ip_ctloutput(op, so, level, optname, mp);
1002: goto end;
1003: }
1004: break;
1005: #endif
1006: #ifdef INET6
1007: case PF_INET6:
1008: if (level != IPPROTO_UDP) {
1009: error = ip6_ctloutput(op, so, level, optname, mp);
1010: goto end;
1011: }
1012: break;
1013: #endif
1014: default:
1015: error = EAFNOSUPPORT;
1016: goto end;
1017: break;
1018: }
1019:
1020:
1021: switch (op) {
1022: case PRCO_SETOPT:
1023: m = *mp;
1024: inp = sotoinpcb(so);
1025:
1026: switch (optname) {
1027: case UDP_ENCAP:
1028: if (m == NULL || m->m_len < sizeof (int)) {
1029: error = EINVAL;
1030: goto end;
1031: }
1.131 perry 1032:
1.130 manu 1033: switch(*mtod(m, int *)) {
1034: #ifdef IPSEC_NAT_T
1035: case 0:
1036: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1037: break;
1038:
1039: case UDP_ENCAP_ESPINUDP:
1040: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1041: inp->inp_flags |= INP_ESPINUDP;
1042: break;
1.131 perry 1043:
1.130 manu 1044: case UDP_ENCAP_ESPINUDP_NON_IKE:
1045: inp->inp_flags &= ~INP_ESPINUDP_ALL;
1046: inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1047: break;
1048: #endif
1049: default:
1050: error = EINVAL;
1051: goto end;
1052: break;
1053: }
1054: break;
1055:
1056: default:
1057: error = ENOPROTOOPT;
1058: goto end;
1059: break;
1060: }
1061: break;
1062:
1063: default:
1064: error = EINVAL;
1065: goto end;
1066: break;
1.131 perry 1067: }
1068:
1.130 manu 1069: end:
1070: splx(s);
1071: return error;
1072: }
1.131 perry 1073:
1.130 manu 1074:
1075: int
1.27 christos 1076: udp_output(struct mbuf *m, ...)
1077: {
1.66 augustss 1078: struct inpcb *inp;
1079: struct udpiphdr *ui;
1.125 thorpej 1080: struct route *ro;
1.66 augustss 1081: int len = m->m_pkthdr.len;
1.31 mycroft 1082: int error = 0;
1.27 christos 1083: va_list ap;
1084:
1.98 matt 1085: MCLAIM(m, &udp_tx_mowner);
1.27 christos 1086: va_start(ap, m);
1087: inp = va_arg(ap, struct inpcb *);
1088: va_end(ap);
1.1 cgd 1089:
1090: /*
1091: * Calculate data length and get a mbuf
1092: * for UDP and IP headers.
1093: */
1.13 mycroft 1094: M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1095: if (m == 0) {
1096: error = ENOBUFS;
1.39 thorpej 1097: goto release;
1098: }
1099:
1100: /*
1101: * Compute the packet length of the IP header, and
1102: * punt if the length looks bogus.
1103: */
1.96 itojun 1104: if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1.39 thorpej 1105: error = EMSGSIZE;
1.13 mycroft 1106: goto release;
1107: }
1.1 cgd 1108:
1109: /*
1110: * Fill in mbuf with extended UDP header
1111: * and addresses and length put into network format.
1112: */
1113: ui = mtod(m, struct udpiphdr *);
1114: ui->ui_pr = IPPROTO_UDP;
1115: ui->ui_src = inp->inp_laddr;
1116: ui->ui_dst = inp->inp_faddr;
1117: ui->ui_sport = inp->inp_lport;
1118: ui->ui_dport = inp->inp_fport;
1.78 thorpej 1119: ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1.1 cgd 1120:
1.125 thorpej 1121: ro = &inp->inp_route;
1122:
1.1 cgd 1123: /*
1.78 thorpej 1124: * Set up checksum and output datagram.
1.1 cgd 1125: */
1126: if (udpcksum) {
1.78 thorpej 1127: /*
1128: * XXX Cache pseudo-header checksum part for
1129: * XXX "connected" UDP sockets.
1130: */
1131: ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1132: ui->ui_dst.s_addr, htons((u_int16_t)len +
1133: sizeof(struct udphdr) + IPPROTO_UDP));
1.135 yamt 1134: m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1.78 thorpej 1135: m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1136: } else
1137: ui->ui_sum = 0;
1.96 itojun 1138: ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1.1 cgd 1139: ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */
1140: ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */
1141: udpstat.udps_opackets++;
1.48 itojun 1142:
1.125 thorpej 1143: return (ip_output(m, inp->inp_options, ro,
1.12 mycroft 1144: inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1.107 itojun 1145: inp->inp_moptions, inp->inp_socket));
1.1 cgd 1146:
1147: release:
1148: m_freem(m);
1149: return (error);
1150: }
1151:
1.42 thorpej 1152: int udp_sendspace = 9216; /* really max datagram size */
1153: int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1.1 cgd 1154: /* 40 1K datagrams */
1155:
1156: /*ARGSUSED*/
1.7 mycroft 1157: int
1.119 matt 1158: udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1159: struct mbuf *control, struct proc *p)
1.1 cgd 1160: {
1.66 augustss 1161: struct inpcb *inp;
1.1 cgd 1162: int s;
1.66 augustss 1163: int error = 0;
1.1 cgd 1164:
1165: if (req == PRU_CONTROL)
1.31 mycroft 1166: return (in_control(so, (long)m, (caddr_t)nam,
1.30 mycroft 1167: (struct ifnet *)control, p));
1.59 thorpej 1168:
1.60 thorpej 1169: if (req == PRU_PURGEIF) {
1.80 itojun 1170: in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1.60 thorpej 1171: in_purgeif((struct ifnet *)control);
1172: in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1.59 thorpej 1173: return (0);
1174: }
1.31 mycroft 1175:
1176: s = splsoftnet();
1177: inp = sotoinpcb(so);
1.32 mycroft 1178: #ifdef DIAGNOSTIC
1179: if (req != PRU_SEND && req != PRU_SENDOOB && control)
1180: panic("udp_usrreq: unexpected control mbuf");
1181: #endif
1.31 mycroft 1182: if (inp == 0 && req != PRU_ATTACH) {
1183: error = EINVAL;
1184: goto release;
1185: }
1186:
1.1 cgd 1187: /*
1188: * Note: need to block udp_input while changing
1189: * the udp pcb queue and/or pcb addresses.
1190: */
1191: switch (req) {
1192:
1193: case PRU_ATTACH:
1.31 mycroft 1194: if (inp != 0) {
1195: error = EISCONN;
1.1 cgd 1196: break;
1197: }
1.98 matt 1198: #ifdef MBUFTRACE
1199: so->so_mowner = &udp_mowner;
1200: so->so_rcv.sb_mowner = &udp_rx_mowner;
1201: so->so_snd.sb_mowner = &udp_tx_mowner;
1202: #endif
1.31 mycroft 1203: if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1204: error = soreserve(so, udp_sendspace, udp_recvspace);
1205: if (error)
1206: break;
1207: }
1.18 mycroft 1208: error = in_pcballoc(so, &udbtable);
1.1 cgd 1209: if (error)
1210: break;
1.31 mycroft 1211: inp = sotoinpcb(so);
1212: inp->inp_ip.ip_ttl = ip_defttl;
1.1 cgd 1213: break;
1214:
1215: case PRU_DETACH:
1.31 mycroft 1216: in_pcbdetach(inp);
1.1 cgd 1217: break;
1218:
1219: case PRU_BIND:
1.31 mycroft 1220: error = in_pcbbind(inp, nam, p);
1.1 cgd 1221: break;
1222:
1223: case PRU_LISTEN:
1224: error = EOPNOTSUPP;
1225: break;
1226:
1227: case PRU_CONNECT:
1.31 mycroft 1228: error = in_pcbconnect(inp, nam);
1229: if (error)
1.1 cgd 1230: break;
1.31 mycroft 1231: soisconnected(so);
1.1 cgd 1232: break;
1233:
1234: case PRU_CONNECT2:
1235: error = EOPNOTSUPP;
1236: break;
1237:
1238: case PRU_DISCONNECT:
1.31 mycroft 1239: /*soisdisconnected(so);*/
1240: so->so_state &= ~SS_ISCONNECTED; /* XXX */
1.1 cgd 1241: in_pcbdisconnect(inp);
1.34 mycroft 1242: inp->inp_laddr = zeroin_addr; /* XXX */
1.35 mycroft 1243: in_pcbstate(inp, INP_BOUND); /* XXX */
1.1 cgd 1244: break;
1245:
1246: case PRU_SHUTDOWN:
1247: socantsendmore(so);
1248: break;
1249:
1.31 mycroft 1250: case PRU_RCVD:
1251: error = EOPNOTSUPP;
1.1 cgd 1252: break;
1253:
1.31 mycroft 1254: case PRU_SEND:
1.32 mycroft 1255: if (control && control->m_len) {
1256: m_freem(control);
1257: m_freem(m);
1258: error = EINVAL;
1259: break;
1260: }
1.31 mycroft 1261: {
1.35 mycroft 1262: struct in_addr laddr; /* XXX */
1.1 cgd 1263:
1.31 mycroft 1264: if (nam) {
1.35 mycroft 1265: laddr = inp->inp_laddr; /* XXX */
1.31 mycroft 1266: if ((so->so_state & SS_ISCONNECTED) != 0) {
1267: error = EISCONN;
1.32 mycroft 1268: goto die;
1.31 mycroft 1269: }
1270: error = in_pcbconnect(inp, nam);
1.100 matt 1271: if (error)
1272: goto die;
1.31 mycroft 1273: } else {
1274: if ((so->so_state & SS_ISCONNECTED) == 0) {
1275: error = ENOTCONN;
1.32 mycroft 1276: goto die;
1.31 mycroft 1277: }
1278: }
1.33 mycroft 1279: error = udp_output(m, inp);
1.100 matt 1280: m = NULL;
1.31 mycroft 1281: if (nam) {
1282: in_pcbdisconnect(inp);
1.35 mycroft 1283: inp->inp_laddr = laddr; /* XXX */
1284: in_pcbstate(inp, INP_BOUND); /* XXX */
1.31 mycroft 1285: }
1.100 matt 1286: die:
1287: if (m)
1288: m_freem(m);
1.31 mycroft 1289: }
1.1 cgd 1290: break;
1291:
1292: case PRU_SENSE:
1293: /*
1294: * stat: don't bother with a blocksize.
1295: */
1.31 mycroft 1296: splx(s);
1.1 cgd 1297: return (0);
1298:
1.31 mycroft 1299: case PRU_RCVOOB:
1300: error = EOPNOTSUPP;
1301: break;
1302:
1.1 cgd 1303: case PRU_SENDOOB:
1.32 mycroft 1304: m_freem(control);
1.31 mycroft 1305: m_freem(m);
1.1 cgd 1306: error = EOPNOTSUPP;
1307: break;
1308:
1.31 mycroft 1309: case PRU_SOCKADDR:
1310: in_setsockaddr(inp, nam);
1311: break;
1312:
1313: case PRU_PEERADDR:
1314: in_setpeeraddr(inp, nam);
1315: break;
1.1 cgd 1316:
1317: default:
1318: panic("udp_usrreq");
1319: }
1320:
1321: release:
1.31 mycroft 1322: splx(s);
1.1 cgd 1323: return (error);
1.13 mycroft 1324: }
1325:
1326: /*
1327: * Sysctl for udp variables.
1328: */
1.114 atatat 1329: SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1.13 mycroft 1330: {
1.114 atatat 1331:
1.116 atatat 1332: sysctl_createv(clog, 0, NULL, NULL,
1333: CTLFLAG_PERMANENT,
1.114 atatat 1334: CTLTYPE_NODE, "net", NULL,
1335: NULL, 0, NULL, 0,
1336: CTL_NET, CTL_EOL);
1.116 atatat 1337: sysctl_createv(clog, 0, NULL, NULL,
1338: CTLFLAG_PERMANENT,
1.114 atatat 1339: CTLTYPE_NODE, "inet", NULL,
1340: NULL, 0, NULL, 0,
1341: CTL_NET, PF_INET, CTL_EOL);
1.116 atatat 1342: sysctl_createv(clog, 0, NULL, NULL,
1343: CTLFLAG_PERMANENT,
1.122 atatat 1344: CTLTYPE_NODE, "udp",
1345: SYSCTL_DESCR("UDPv4 related settings"),
1.114 atatat 1346: NULL, 0, NULL, 0,
1347: CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1348:
1.116 atatat 1349: sysctl_createv(clog, 0, NULL, NULL,
1350: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1351: CTLTYPE_INT, "checksum",
1.123 heas 1352: SYSCTL_DESCR("Compute UDP checksums"),
1.114 atatat 1353: NULL, 0, &udpcksum, 0,
1354: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1355: CTL_EOL);
1.116 atatat 1356: sysctl_createv(clog, 0, NULL, NULL,
1357: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1358: CTLTYPE_INT, "sendspace",
1359: SYSCTL_DESCR("Default UDP send buffer size"),
1.114 atatat 1360: NULL, 0, &udp_sendspace, 0,
1361: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1362: CTL_EOL);
1.116 atatat 1363: sysctl_createv(clog, 0, NULL, NULL,
1364: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.122 atatat 1365: CTLTYPE_INT, "recvspace",
1366: SYSCTL_DESCR("Default UDP receive buffer size"),
1.114 atatat 1367: NULL, 0, &udp_recvspace, 0,
1368: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1369: CTL_EOL);
1.125 thorpej 1370: sysctl_createv(clog, 0, NULL, NULL,
1371: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1372: CTLTYPE_INT, "do_loopback_cksum",
1373: SYSCTL_DESCR("Perform UDP checksum on loopback"),
1374: NULL, 0, &udp_do_loopback_cksum, 0,
1375: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1376: CTL_EOL);
1.132 atatat 1377: sysctl_createv(clog, 0, NULL, NULL,
1378: CTLFLAG_PERMANENT,
1.134 atatat 1379: CTLTYPE_STRUCT, "pcblist",
1.132 atatat 1380: SYSCTL_DESCR("UDP protocol control block list"),
1381: sysctl_inpcblist, 0, &udbtable, 0,
1382: CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1383: CTL_EOL);
1.139 elad 1384: sysctl_createv(clog, 0, NULL, NULL,
1385: CTLFLAG_PERMANENT,
1386: CTLTYPE_STRUCT, "stats",
1387: SYSCTL_DESCR("UDP statistics"),
1388: NULL, 0, &udpstat, sizeof(udpstat),
1389: CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1390: CTL_EOL);
1.1 cgd 1391: }
1.72 itojun 1392: #endif
1.130 manu 1393:
1394: #if (defined INET && defined IPSEC_NAT_T)
1395: /*
1396: * Returns:
1397: * 1 if the packet was processed
1398: * 0 if normal UDP processing should take place
1399: */
1400: static int
1401: udp4_espinudp(m, off, src, so)
1402: struct mbuf *m;
1403: int off;
1404: struct sockaddr *src;
1405: struct socket *so;
1406: {
1407: size_t len;
1408: caddr_t data;
1409: struct inpcb *inp;
1410: size_t skip = 0;
1411: size_t minlen;
1412: size_t iphdrlen;
1413: struct ip *ip;
1414: struct mbuf *n;
1.136 manu 1415: struct m_tag *tag;
1416: struct udphdr *udphdr;
1417: u_int16_t sport, dport;
1.130 manu 1418:
1.131 perry 1419: /*
1.130 manu 1420: * Collapse the mbuf chain if the first mbuf is too short
1421: * The longest case is: UDP + non ESP marker + ESP
1422: */
1423: minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1424: if (minlen > m->m_pkthdr.len)
1425: minlen = m->m_pkthdr.len;
1426:
1427: if (m->m_len < minlen) {
1428: if ((m = m_pullup(m, minlen)) == NULL) {
1429: printf("udp4_espinudp: m_pullup failed\n");
1430: return 0;
1431: }
1432: }
1433:
1.131 perry 1434: len = m->m_len - off;
1.130 manu 1435: data = mtod(m, caddr_t) + off;
1436: inp = sotoinpcb(so);
1437:
1438: /* Ignore keepalive packets */
1439: if ((len == 1) && (data[0] == '\xff')) {
1440: return 1;
1441: }
1442:
1.131 perry 1443: /*
1444: * Check that the payload is long enough to hold
1.130 manu 1445: * an ESP header and compute the length of encapsulation
1.131 perry 1446: * header to remove
1.130 manu 1447: */
1448: if (inp->inp_flags & INP_ESPINUDP) {
1449: u_int32_t *st = (u_int32_t *)data;
1450:
1451: if ((len <= sizeof(struct esp)) || (*st == 0))
1452: return 0; /* Normal UDP processing */
1453:
1454: skip = sizeof(struct udphdr);
1455: }
1456:
1457: if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1458: u_int64_t *st = (u_int64_t *)data;
1459:
1460: if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1461: || (*st != 0))
1462: return 0; /* Normal UDP processing */
1.131 perry 1463:
1.130 manu 1464: skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1465: }
1466:
1467: /*
1.136 manu 1468: * Get the UDP ports. They are handled in network
1469: * order everywhere in IPSEC_NAT_T code.
1470: */
1471: udphdr = (struct udphdr *)(data - skip);
1472: sport = udphdr->uh_sport;
1473: dport = udphdr->uh_dport;
1474:
1475: /*
1.130 manu 1476: * Remove the UDP header (and possibly the non ESP marker)
1477: * IP header lendth is iphdrlen
1.131 perry 1478: * Before:
1.130 manu 1479: * <--- off --->
1480: * +----+------+-----+
1481: * | IP | UDP | ESP |
1482: * +----+------+-----+
1483: * <-skip->
1484: * After:
1485: * +----+-----+
1486: * | IP | ESP |
1487: * +----+-----+
1488: * <-skip->
1489: */
1490: iphdrlen = off - sizeof(struct udphdr);
1491: memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen);
1492: m_adj(m, skip);
1493:
1494: ip = mtod(m, struct ip *);
1495: ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1496: ip->ip_p = IPPROTO_ESP;
1497:
1498: /*
1.131 perry 1499: * Copy the mbuf to avoid multiple free, as both
1500: * esp4_input (which we call) and udp_input (which
1.130 manu 1501: * called us) free the mbuf.
1502: */
1503: if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1504: printf("udp4_espinudp: m_dup failed\n");
1505: return 0;
1506: }
1507:
1.136 manu 1508: /*
1509: * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1510: * the source UDP port. This is required if we want
1511: * to select the right SPD for multiple hosts behind
1512: * same NAT
1513: */
1.137 manu 1514: if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1515: sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1516: printf("udp4_espinudp: m_tag_get failed\n");
1.138 manu 1517: m_freem(n);
1.137 manu 1518: return 0;
1519: }
1.136 manu 1520: ((u_int16_t *)(tag + 1))[0] = sport;
1521: ((u_int16_t *)(tag + 1))[1] = dport;
1522: m_tag_prepend(n, tag);
1523:
1.130 manu 1524: esp4_input(n, iphdrlen);
1525:
1526: /* We handled it, it shoudln't be handled by UDP */
1527: return 1;
1528: }
1529: #endif
CVSweb <webmaster@jp.NetBSD.org>