udp_usrreq.c revision 1.116.2.4
1 /* $NetBSD: udp_usrreq.c,v 1.116.2.4 2004/05/28 07:24:17 tron Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 61 */ 62 63 #include <sys/cdefs.h> 64 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.116.2.4 2004/05/28 07:24:17 tron Exp $"); 65 66 #include "opt_inet.h" 67 #include "opt_ipsec.h" 68 #include "opt_inet_csum.h" 69 #include "opt_ipkdb.h" 70 #include "opt_mbuftrace.h" 71 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> 78 #include <sys/errno.h> 79 #include <sys/stat.h> 80 #include <sys/systm.h> 81 #include <sys/proc.h> 82 #include <sys/domain.h> 83 #include <sys/sysctl.h> 84 85 #include <net/if.h> 86 #include <net/route.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_systm.h> 90 #include <netinet/in_var.h> 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> 97 98 #ifdef INET6 99 #include <netinet/ip6.h> 100 #include <netinet/icmp6.h> 101 #include <netinet6/ip6_var.h> 102 #include <netinet6/in6_pcb.h> 103 #include <netinet6/udp6_var.h> 104 #endif 105 106 #ifndef INET6 107 /* always need ip6.h for IP6_EXTHDR_GET */ 108 #include <netinet/ip6.h> 109 #endif 110 111 #include "faith.h" 112 #if defined(NFAITH) && NFAITH > 0 113 #include <net/if_faith.h> 114 #endif 115 116 #include <machine/stdarg.h> 117 118 #ifdef FAST_IPSEC 119 #include <netipsec/ipsec.h> 120 #include <netipsec/ipsec_var.h> /* XXX ipsecstat namespace */ 121 #ifdef INET6 122 #include <netipsec/ipsec6.h> 123 #endif 124 #endif /* FAST_IPSEC*/ 125 126 #ifdef IPSEC 127 #include <netinet6/ipsec.h> 128 #include <netkey/key.h> 129 #endif /*IPSEC*/ 130 131 #ifdef IPKDB 132 #include <ipkdb/ipkdb.h> 133 #endif 134 135 /* 136 * UDP protocol implementation. 137 * Per RFC 768, August, 1980. 138 */ 139 #ifndef COMPAT_42 140 int udpcksum = 1; 141 #else 142 int udpcksum = 0; /* XXX */ 143 #endif 144 145 struct inpcbtable udbtable; 146 struct udpstat udpstat; 147 148 #ifdef INET 149 static void udp4_sendup __P((struct mbuf *, int, struct sockaddr *, 150 struct socket *)); 151 static int udp4_realinput __P((struct sockaddr_in *, struct sockaddr_in *, 152 struct mbuf *, int)); 153 #endif 154 #ifdef INET6 155 static void udp6_sendup __P((struct mbuf *, int, struct sockaddr *, 156 struct socket *)); 157 static int udp6_realinput __P((int, struct sockaddr_in6 *, 158 struct sockaddr_in6 *, struct mbuf *, int)); 159 #endif 160 #ifdef INET 161 static void udp_notify __P((struct inpcb *, int)); 162 #endif 163 164 #ifndef UDBHASHSIZE 165 #define UDBHASHSIZE 128 166 #endif 167 int udbhashsize = UDBHASHSIZE; 168 169 #ifdef MBUFTRACE 170 struct mowner udp_mowner = { "udp" }; 171 struct mowner udp_rx_mowner = { "udp", "rx" }; 172 struct mowner udp_tx_mowner = { "udp", "tx" }; 173 #endif 174 175 #ifdef UDP_CSUM_COUNTERS 176 #include <sys/device.h> 177 178 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 179 NULL, "udp", "hwcsum bad"); 180 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 181 NULL, "udp", "hwcsum ok"); 182 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 183 NULL, "udp", "hwcsum data"); 184 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 185 NULL, "udp", "swcsum"); 186 187 #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 188 189 #else 190 191 #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */ 192 193 #endif /* UDP_CSUM_COUNTERS */ 194 195 void 196 udp_init() 197 { 198 199 in_pcbinit(&udbtable, udbhashsize, udbhashsize); 200 201 #ifdef UDP_CSUM_COUNTERS 202 evcnt_attach_static(&udp_hwcsum_bad); 203 evcnt_attach_static(&udp_hwcsum_ok); 204 evcnt_attach_static(&udp_hwcsum_data); 205 evcnt_attach_static(&udp_swcsum); 206 #endif /* UDP_CSUM_COUNTERS */ 207 208 MOWNER_ATTACH(&udp_tx_mowner); 209 MOWNER_ATTACH(&udp_rx_mowner); 210 MOWNER_ATTACH(&udp_mowner); 211 } 212 213 #ifdef INET 214 void 215 #if __STDC__ 216 udp_input(struct mbuf *m, ...) 217 #else 218 udp_input(m, va_alist) 219 struct mbuf *m; 220 va_dcl 221 #endif 222 { 223 va_list ap; 224 struct sockaddr_in src, dst; 225 struct ip *ip; 226 struct udphdr *uh; 227 int iphlen; 228 int len; 229 int n; 230 u_int16_t ip_len; 231 232 va_start(ap, m); 233 iphlen = va_arg(ap, int); 234 (void)va_arg(ap, int); /* ignore value, advance ap */ 235 va_end(ap); 236 237 MCLAIM(m, &udp_rx_mowner); 238 udpstat.udps_ipackets++; 239 240 /* 241 * Get IP and UDP header together in first mbuf. 242 */ 243 ip = mtod(m, struct ip *); 244 IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 245 if (uh == NULL) { 246 udpstat.udps_hdrops++; 247 return; 248 } 249 KASSERT(UDP_HDR_ALIGNED_P(uh)); 250 251 /* destination port of 0 is illegal, based on RFC768. */ 252 if (uh->uh_dport == 0) 253 goto bad; 254 255 /* 256 * Make mbuf data length reflect UDP length. 257 * If not enough data to reflect UDP length, drop. 258 */ 259 ip_len = ntohs(ip->ip_len); 260 len = ntohs((u_int16_t)uh->uh_ulen); 261 if (ip_len != iphlen + len) { 262 if (ip_len < iphlen + len || len < sizeof(struct udphdr)) { 263 udpstat.udps_badlen++; 264 goto bad; 265 } 266 m_adj(m, iphlen + len - ip_len); 267 } 268 269 /* 270 * Checksum extended UDP header and data. 271 */ 272 if (uh->uh_sum) { 273 switch (m->m_pkthdr.csum_flags & 274 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) | 275 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 276 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD: 277 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad); 278 goto badcsum; 279 280 case M_CSUM_UDPv4|M_CSUM_DATA: { 281 u_int32_t hw_csum = m->m_pkthdr.csum_data; 282 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); 283 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) 284 hw_csum = in_cksum_phdr(ip->ip_src.s_addr, 285 ip->ip_dst.s_addr, 286 htons(hw_csum + len + IPPROTO_UDP)); 287 if ((hw_csum ^ 0xffff) != 0) 288 goto badcsum; 289 break; 290 } 291 292 case M_CSUM_UDPv4: 293 /* Checksum was okay. */ 294 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); 295 break; 296 297 default: 298 /* Need to compute it ourselves. */ 299 UDP_CSUM_COUNTER_INCR(&udp_swcsum); 300 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) 301 goto badcsum; 302 break; 303 } 304 } 305 306 /* construct source and dst sockaddrs. */ 307 bzero(&src, sizeof(src)); 308 src.sin_family = AF_INET; 309 src.sin_len = sizeof(struct sockaddr_in); 310 bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr)); 311 src.sin_port = uh->uh_sport; 312 bzero(&dst, sizeof(dst)); 313 dst.sin_family = AF_INET; 314 dst.sin_len = sizeof(struct sockaddr_in); 315 bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr)); 316 dst.sin_port = uh->uh_dport; 317 318 n = udp4_realinput(&src, &dst, m, iphlen); 319 #ifdef INET6 320 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { 321 struct sockaddr_in6 src6, dst6; 322 323 bzero(&src6, sizeof(src6)); 324 src6.sin6_family = AF_INET6; 325 src6.sin6_len = sizeof(struct sockaddr_in6); 326 src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff; 327 bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12], 328 sizeof(ip->ip_src)); 329 src6.sin6_port = uh->uh_sport; 330 bzero(&dst6, sizeof(dst6)); 331 dst6.sin6_family = AF_INET6; 332 dst6.sin6_len = sizeof(struct sockaddr_in6); 333 dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff; 334 bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12], 335 sizeof(ip->ip_dst)); 336 dst6.sin6_port = uh->uh_dport; 337 338 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen); 339 } 340 #endif 341 342 if (n == 0) { 343 if (m->m_flags & (M_BCAST | M_MCAST)) { 344 udpstat.udps_noportbcast++; 345 goto bad; 346 } 347 udpstat.udps_noport++; 348 #ifdef IPKDB 349 if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport, 350 m, iphlen + sizeof(struct udphdr), 351 m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) { 352 /* 353 * It was a debugger connect packet, 354 * just drop it now 355 */ 356 goto bad; 357 } 358 #endif 359 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 360 m = NULL; 361 } 362 363 bad: 364 if (m) 365 m_freem(m); 366 return; 367 368 badcsum: 369 m_freem(m); 370 udpstat.udps_badsum++; 371 } 372 #endif 373 374 #ifdef INET6 375 int 376 udp6_input(mp, offp, proto) 377 struct mbuf **mp; 378 int *offp, proto; 379 { 380 struct mbuf *m = *mp; 381 int off = *offp; 382 struct sockaddr_in6 src, dst; 383 struct ip6_hdr *ip6; 384 struct udphdr *uh; 385 u_int32_t plen, ulen; 386 387 ip6 = mtod(m, struct ip6_hdr *); 388 389 #if defined(NFAITH) && 0 < NFAITH 390 if (faithprefix(&ip6->ip6_dst)) { 391 /* send icmp6 host unreach? */ 392 m_freem(m); 393 return IPPROTO_DONE; 394 } 395 #endif 396 397 udp6stat.udp6s_ipackets++; 398 399 /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */ 400 plen = m->m_pkthdr.len - off; 401 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr)); 402 if (uh == NULL) { 403 ip6stat.ip6s_tooshort++; 404 return IPPROTO_DONE; 405 } 406 KASSERT(UDP_HDR_ALIGNED_P(uh)); 407 ulen = ntohs((u_short)uh->uh_ulen); 408 /* 409 * RFC2675 section 4: jumbograms will have 0 in the UDP header field, 410 * iff payload length > 0xffff. 411 */ 412 if (ulen == 0 && plen > 0xffff) 413 ulen = plen; 414 415 if (plen != ulen) { 416 udp6stat.udp6s_badlen++; 417 goto bad; 418 } 419 420 /* destination port of 0 is illegal, based on RFC768. */ 421 if (uh->uh_dport == 0) 422 goto bad; 423 424 /* Be proactive about malicious use of IPv4 mapped address */ 425 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 426 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 427 /* XXX stat */ 428 goto bad; 429 } 430 431 /* 432 * Checksum extended UDP header and data. 433 */ 434 if (uh->uh_sum == 0) { 435 udp6stat.udp6s_nosum++; 436 goto bad; 437 } 438 if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) { 439 udp6stat.udp6s_badsum++; 440 goto bad; 441 } 442 443 /* 444 * Construct source and dst sockaddrs. 445 * Note that ifindex (s6_addr16[1]) is already filled. 446 */ 447 bzero(&src, sizeof(src)); 448 src.sin6_family = AF_INET6; 449 src.sin6_len = sizeof(struct sockaddr_in6); 450 /* KAME hack: recover scopeid */ 451 (void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif); 452 src.sin6_port = uh->uh_sport; 453 bzero(&dst, sizeof(dst)); 454 dst.sin6_family = AF_INET6; 455 dst.sin6_len = sizeof(struct sockaddr_in6); 456 /* KAME hack: recover scopeid */ 457 (void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif); 458 dst.sin6_port = uh->uh_dport; 459 460 if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) { 461 if (m->m_flags & M_MCAST) { 462 udp6stat.udp6s_noportmcast++; 463 goto bad; 464 } 465 udp6stat.udp6s_noport++; 466 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); 467 m = NULL; 468 } 469 470 bad: 471 if (m) 472 m_freem(m); 473 return IPPROTO_DONE; 474 } 475 #endif 476 477 #ifdef INET 478 static void 479 udp4_sendup(m, off, src, so) 480 struct mbuf *m; 481 int off; /* offset of data portion */ 482 struct sockaddr *src; 483 struct socket *so; 484 { 485 struct mbuf *opts = NULL; 486 struct mbuf *n; 487 struct inpcb *inp = NULL; 488 489 if (!so) 490 return; 491 switch (so->so_proto->pr_domain->dom_family) { 492 case AF_INET: 493 inp = sotoinpcb(so); 494 break; 495 #ifdef INET6 496 case AF_INET6: 497 break; 498 #endif 499 default: 500 return; 501 } 502 503 #if defined(IPSEC) || defined(FAST_IPSEC) 504 /* check AH/ESP integrity. */ 505 if (so != NULL && ipsec4_in_reject_so(m, so)) { 506 ipsecstat.in_polvio++; 507 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) 508 icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT, 509 0, 0); 510 return; 511 } 512 #endif /*IPSEC*/ 513 514 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { 515 if (inp && (inp->inp_flags & INP_CONTROLOPTS 516 || so->so_options & SO_TIMESTAMP)) { 517 struct ip *ip = mtod(n, struct ip *); 518 ip_savecontrol(inp, &opts, ip, n); 519 } 520 521 m_adj(n, off); 522 if (sbappendaddr(&so->so_rcv, src, n, 523 opts) == 0) { 524 m_freem(n); 525 if (opts) 526 m_freem(opts); 527 udpstat.udps_fullsock++; 528 } else 529 sorwakeup(so); 530 } 531 } 532 #endif 533 534 #ifdef INET6 535 static void 536 udp6_sendup(m, off, src, so) 537 struct mbuf *m; 538 int off; /* offset of data portion */ 539 struct sockaddr *src; 540 struct socket *so; 541 { 542 struct mbuf *opts = NULL; 543 struct mbuf *n; 544 struct in6pcb *in6p = NULL; 545 546 if (!so) 547 return; 548 if (so->so_proto->pr_domain->dom_family != AF_INET6) 549 return; 550 in6p = sotoin6pcb(so); 551 552 #if defined(IPSEC) || defined(FAST_IPSEC) 553 /* check AH/ESP integrity. */ 554 if (so != NULL && ipsec6_in_reject_so(m, so)) { 555 ipsec6stat.in_polvio++; 556 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) 557 icmp6_error(n, ICMP6_DST_UNREACH, 558 ICMP6_DST_UNREACH_ADMIN, 0); 559 return; 560 } 561 #endif /*IPSEC*/ 562 563 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { 564 if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS 565 || in6p->in6p_socket->so_options & SO_TIMESTAMP)) { 566 struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *); 567 ip6_savecontrol(in6p, &opts, ip6, n); 568 } 569 570 m_adj(n, off); 571 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) { 572 m_freem(n); 573 if (opts) 574 m_freem(opts); 575 udp6stat.udp6s_fullsock++; 576 } else 577 sorwakeup(so); 578 } 579 } 580 #endif 581 582 #ifdef INET 583 static int 584 udp4_realinput(src, dst, m, off) 585 struct sockaddr_in *src; 586 struct sockaddr_in *dst; 587 struct mbuf *m; 588 int off; /* offset of udphdr */ 589 { 590 u_int16_t *sport, *dport; 591 int rcvcnt; 592 struct in_addr *src4, *dst4; 593 struct inpcb_hdr *inph; 594 struct inpcb *inp; 595 596 rcvcnt = 0; 597 off += sizeof(struct udphdr); /* now, offset of payload */ 598 599 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 600 goto bad; 601 602 src4 = &src->sin_addr; 603 sport = &src->sin_port; 604 dst4 = &dst->sin_addr; 605 dport = &dst->sin_port; 606 607 if (IN_MULTICAST(dst4->s_addr) || 608 in_broadcast(*dst4, m->m_pkthdr.rcvif)) { 609 /* 610 * Deliver a multicast or broadcast datagram to *all* sockets 611 * for which the local and remote addresses and ports match 612 * those of the incoming datagram. This allows more than 613 * one process to receive multi/broadcasts on the same port. 614 * (This really ought to be done for unicast datagrams as 615 * well, but that would cause problems with existing 616 * applications that open both address-specific sockets and 617 * a wildcard socket listening to the same port -- they would 618 * end up receiving duplicates of every unicast datagram. 619 * Those applications open the multiple sockets to overcome an 620 * inadequacy of the UDP socket interface, but for backwards 621 * compatibility we avoid the problem here rather than 622 * fixing the interface. Maybe 4.5BSD will remedy this?) 623 */ 624 625 /* 626 * KAME note: traditionally we dropped udpiphdr from mbuf here. 627 * we need udpiphdr for IPsec processing so we do that later. 628 */ 629 /* 630 * Locate pcb(s) for datagram. 631 */ 632 CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 633 inp = (struct inpcb *)inph; 634 if (inp->inp_af != AF_INET) 635 continue; 636 637 if (inp->inp_lport != *dport) 638 continue; 639 if (!in_nullhost(inp->inp_laddr)) { 640 if (!in_hosteq(inp->inp_laddr, *dst4)) 641 continue; 642 } 643 if (!in_nullhost(inp->inp_faddr)) { 644 if (!in_hosteq(inp->inp_faddr, *src4) || 645 inp->inp_fport != *sport) 646 continue; 647 } 648 649 udp4_sendup(m, off, (struct sockaddr *)src, 650 inp->inp_socket); 651 rcvcnt++; 652 653 /* 654 * Don't look for additional matches if this one does 655 * not have either the SO_REUSEPORT or SO_REUSEADDR 656 * socket options set. This heuristic avoids searching 657 * through all pcbs in the common case of a non-shared 658 * port. It assumes that an application will never 659 * clear these options after setting them. 660 */ 661 if ((inp->inp_socket->so_options & 662 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 663 break; 664 } 665 } else { 666 /* 667 * Locate pcb for datagram. 668 */ 669 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport); 670 if (inp == 0) { 671 ++udpstat.udps_pcbhashmiss; 672 inp = in_pcblookup_bind(&udbtable, *dst4, *dport); 673 if (inp == 0) 674 return rcvcnt; 675 } 676 677 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 678 rcvcnt++; 679 } 680 681 bad: 682 return rcvcnt; 683 } 684 #endif 685 686 #ifdef INET6 687 static int 688 udp6_realinput(af, src, dst, m, off) 689 int af; /* af on packet */ 690 struct sockaddr_in6 *src; 691 struct sockaddr_in6 *dst; 692 struct mbuf *m; 693 int off; /* offset of udphdr */ 694 { 695 u_int16_t sport, dport; 696 int rcvcnt; 697 struct in6_addr src6, dst6; 698 const struct in_addr *dst4; 699 struct inpcb_hdr *inph; 700 struct in6pcb *in6p; 701 702 rcvcnt = 0; 703 off += sizeof(struct udphdr); /* now, offset of payload */ 704 705 if (af != AF_INET && af != AF_INET6) 706 goto bad; 707 if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6) 708 goto bad; 709 710 in6_embedscope(&src6, src, NULL, NULL); 711 sport = src->sin6_port; 712 in6_embedscope(&dst6, dst, NULL, NULL); 713 dport = dst->sin6_port; 714 dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12]; 715 716 if (IN6_IS_ADDR_MULTICAST(&dst6) || 717 (af == AF_INET && IN_MULTICAST(dst4->s_addr))) { 718 /* 719 * Deliver a multicast or broadcast datagram to *all* sockets 720 * for which the local and remote addresses and ports match 721 * those of the incoming datagram. This allows more than 722 * one process to receive multi/broadcasts on the same port. 723 * (This really ought to be done for unicast datagrams as 724 * well, but that would cause problems with existing 725 * applications that open both address-specific sockets and 726 * a wildcard socket listening to the same port -- they would 727 * end up receiving duplicates of every unicast datagram. 728 * Those applications open the multiple sockets to overcome an 729 * inadequacy of the UDP socket interface, but for backwards 730 * compatibility we avoid the problem here rather than 731 * fixing the interface. Maybe 4.5BSD will remedy this?) 732 */ 733 734 /* 735 * KAME note: traditionally we dropped udpiphdr from mbuf here. 736 * we need udpiphdr for IPsec processing so we do that later. 737 */ 738 /* 739 * Locate pcb(s) for datagram. 740 */ 741 CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 742 in6p = (struct in6pcb *)inph; 743 if (in6p->in6p_af != AF_INET6) 744 continue; 745 746 if (in6p->in6p_lport != dport) 747 continue; 748 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 749 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6)) 750 continue; 751 } else { 752 if (IN6_IS_ADDR_V4MAPPED(&dst6) && 753 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 754 continue; 755 } 756 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 757 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, 758 &src6) || in6p->in6p_fport != sport) 759 continue; 760 } else { 761 if (IN6_IS_ADDR_V4MAPPED(&src6) && 762 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 763 continue; 764 } 765 766 udp6_sendup(m, off, (struct sockaddr *)src, 767 in6p->in6p_socket); 768 rcvcnt++; 769 770 /* 771 * Don't look for additional matches if this one does 772 * not have either the SO_REUSEPORT or SO_REUSEADDR 773 * socket options set. This heuristic avoids searching 774 * through all pcbs in the common case of a non-shared 775 * port. It assumes that an application will never 776 * clear these options after setting them. 777 */ 778 if ((in6p->in6p_socket->so_options & 779 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 780 break; 781 } 782 } else { 783 /* 784 * Locate pcb for datagram. 785 */ 786 in6p = in6_pcblookup_connect(&udbtable, &src6, sport, 787 &dst6, dport, 0); 788 if (in6p == 0) { 789 ++udpstat.udps_pcbhashmiss; 790 in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0); 791 if (in6p == 0) 792 return rcvcnt; 793 } 794 795 udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket); 796 rcvcnt++; 797 } 798 799 bad: 800 return rcvcnt; 801 } 802 #endif 803 804 #ifdef INET 805 /* 806 * Notify a udp user of an asynchronous error; 807 * just wake up so that he can collect error status. 808 */ 809 static void 810 udp_notify(inp, errno) 811 struct inpcb *inp; 812 int errno; 813 { 814 815 inp->inp_socket->so_error = errno; 816 sorwakeup(inp->inp_socket); 817 sowwakeup(inp->inp_socket); 818 } 819 820 void * 821 udp_ctlinput(cmd, sa, v) 822 int cmd; 823 struct sockaddr *sa; 824 void *v; 825 { 826 struct ip *ip = v; 827 struct udphdr *uh; 828 void (*notify) __P((struct inpcb *, int)) = udp_notify; 829 int errno; 830 831 if (sa->sa_family != AF_INET 832 || sa->sa_len != sizeof(struct sockaddr_in)) 833 return NULL; 834 if ((unsigned)cmd >= PRC_NCMDS) 835 return NULL; 836 errno = inetctlerrmap[cmd]; 837 if (PRC_IS_REDIRECT(cmd)) 838 notify = in_rtchange, ip = 0; 839 else if (cmd == PRC_HOSTDEAD) 840 ip = 0; 841 else if (errno == 0) 842 return NULL; 843 if (ip) { 844 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 845 in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport, 846 ip->ip_src, uh->uh_sport, errno, notify); 847 848 /* XXX mapped address case */ 849 } else 850 in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno, 851 notify); 852 return NULL; 853 } 854 855 int 856 #if __STDC__ 857 udp_output(struct mbuf *m, ...) 858 #else 859 udp_output(m, va_alist) 860 struct mbuf *m; 861 va_dcl 862 #endif 863 { 864 struct inpcb *inp; 865 struct udpiphdr *ui; 866 int len = m->m_pkthdr.len; 867 int error = 0; 868 va_list ap; 869 870 MCLAIM(m, &udp_tx_mowner); 871 va_start(ap, m); 872 inp = va_arg(ap, struct inpcb *); 873 va_end(ap); 874 875 /* 876 * Calculate data length and get a mbuf 877 * for UDP and IP headers. 878 */ 879 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 880 if (m == 0) { 881 error = ENOBUFS; 882 goto release; 883 } 884 885 /* 886 * Compute the packet length of the IP header, and 887 * punt if the length looks bogus. 888 */ 889 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 890 error = EMSGSIZE; 891 goto release; 892 } 893 894 /* 895 * Fill in mbuf with extended UDP header 896 * and addresses and length put into network format. 897 */ 898 ui = mtod(m, struct udpiphdr *); 899 ui->ui_pr = IPPROTO_UDP; 900 ui->ui_src = inp->inp_laddr; 901 ui->ui_dst = inp->inp_faddr; 902 ui->ui_sport = inp->inp_lport; 903 ui->ui_dport = inp->inp_fport; 904 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 905 906 /* 907 * Set up checksum and output datagram. 908 */ 909 if (udpcksum) { 910 /* 911 * XXX Cache pseudo-header checksum part for 912 * XXX "connected" UDP sockets. 913 */ 914 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 915 ui->ui_dst.s_addr, htons((u_int16_t)len + 916 sizeof(struct udphdr) + IPPROTO_UDP)); 917 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 918 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 919 } else 920 ui->ui_sum = 0; 921 ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len); 922 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ 923 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ 924 udpstat.udps_opackets++; 925 926 return (ip_output(m, inp->inp_options, &inp->inp_route, 927 inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), 928 inp->inp_moptions, inp->inp_socket)); 929 930 release: 931 m_freem(m); 932 return (error); 933 } 934 935 int udp_sendspace = 9216; /* really max datagram size */ 936 int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); 937 /* 40 1K datagrams */ 938 939 /*ARGSUSED*/ 940 int 941 udp_usrreq(so, req, m, nam, control, p) 942 struct socket *so; 943 int req; 944 struct mbuf *m, *nam, *control; 945 struct proc *p; 946 { 947 struct inpcb *inp; 948 int s; 949 int error = 0; 950 951 if (req == PRU_CONTROL) 952 return (in_control(so, (long)m, (caddr_t)nam, 953 (struct ifnet *)control, p)); 954 955 if (req == PRU_PURGEIF) { 956 in_pcbpurgeif0(&udbtable, (struct ifnet *)control); 957 in_purgeif((struct ifnet *)control); 958 in_pcbpurgeif(&udbtable, (struct ifnet *)control); 959 return (0); 960 } 961 962 s = splsoftnet(); 963 inp = sotoinpcb(so); 964 #ifdef DIAGNOSTIC 965 if (req != PRU_SEND && req != PRU_SENDOOB && control) 966 panic("udp_usrreq: unexpected control mbuf"); 967 #endif 968 if (inp == 0 && req != PRU_ATTACH) { 969 error = EINVAL; 970 goto release; 971 } 972 973 /* 974 * Note: need to block udp_input while changing 975 * the udp pcb queue and/or pcb addresses. 976 */ 977 switch (req) { 978 979 case PRU_ATTACH: 980 if (inp != 0) { 981 error = EISCONN; 982 break; 983 } 984 #ifdef MBUFTRACE 985 so->so_mowner = &udp_mowner; 986 so->so_rcv.sb_mowner = &udp_rx_mowner; 987 so->so_snd.sb_mowner = &udp_tx_mowner; 988 #endif 989 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 990 error = soreserve(so, udp_sendspace, udp_recvspace); 991 if (error) 992 break; 993 } 994 error = in_pcballoc(so, &udbtable); 995 if (error) 996 break; 997 inp = sotoinpcb(so); 998 inp->inp_ip.ip_ttl = ip_defttl; 999 break; 1000 1001 case PRU_DETACH: 1002 in_pcbdetach(inp); 1003 break; 1004 1005 case PRU_BIND: 1006 error = in_pcbbind(inp, nam, p); 1007 break; 1008 1009 case PRU_LISTEN: 1010 error = EOPNOTSUPP; 1011 break; 1012 1013 case PRU_CONNECT: 1014 error = in_pcbconnect(inp, nam); 1015 if (error) 1016 break; 1017 soisconnected(so); 1018 break; 1019 1020 case PRU_CONNECT2: 1021 error = EOPNOTSUPP; 1022 break; 1023 1024 case PRU_DISCONNECT: 1025 /*soisdisconnected(so);*/ 1026 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 1027 in_pcbdisconnect(inp); 1028 inp->inp_laddr = zeroin_addr; /* XXX */ 1029 in_pcbstate(inp, INP_BOUND); /* XXX */ 1030 break; 1031 1032 case PRU_SHUTDOWN: 1033 socantsendmore(so); 1034 break; 1035 1036 case PRU_RCVD: 1037 error = EOPNOTSUPP; 1038 break; 1039 1040 case PRU_SEND: 1041 if (control && control->m_len) { 1042 m_freem(control); 1043 m_freem(m); 1044 error = EINVAL; 1045 break; 1046 } 1047 { 1048 struct in_addr laddr; /* XXX */ 1049 1050 if (nam) { 1051 laddr = inp->inp_laddr; /* XXX */ 1052 if ((so->so_state & SS_ISCONNECTED) != 0) { 1053 error = EISCONN; 1054 goto die; 1055 } 1056 error = in_pcbconnect(inp, nam); 1057 if (error) 1058 goto die; 1059 } else { 1060 if ((so->so_state & SS_ISCONNECTED) == 0) { 1061 error = ENOTCONN; 1062 goto die; 1063 } 1064 } 1065 error = udp_output(m, inp); 1066 m = NULL; 1067 if (nam) { 1068 in_pcbdisconnect(inp); 1069 inp->inp_laddr = laddr; /* XXX */ 1070 in_pcbstate(inp, INP_BOUND); /* XXX */ 1071 } 1072 die: 1073 if (m) 1074 m_freem(m); 1075 } 1076 break; 1077 1078 case PRU_SENSE: 1079 /* 1080 * stat: don't bother with a blocksize. 1081 */ 1082 splx(s); 1083 return (0); 1084 1085 case PRU_RCVOOB: 1086 error = EOPNOTSUPP; 1087 break; 1088 1089 case PRU_SENDOOB: 1090 m_freem(control); 1091 m_freem(m); 1092 error = EOPNOTSUPP; 1093 break; 1094 1095 case PRU_SOCKADDR: 1096 in_setsockaddr(inp, nam); 1097 break; 1098 1099 case PRU_PEERADDR: 1100 in_setpeeraddr(inp, nam); 1101 break; 1102 1103 default: 1104 panic("udp_usrreq"); 1105 } 1106 1107 release: 1108 splx(s); 1109 return (error); 1110 } 1111 1112 /* 1113 * Sysctl for udp variables. 1114 */ 1115 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup") 1116 { 1117 1118 sysctl_createv(clog, 0, NULL, NULL, 1119 CTLFLAG_PERMANENT, 1120 CTLTYPE_NODE, "net", NULL, 1121 NULL, 0, NULL, 0, 1122 CTL_NET, CTL_EOL); 1123 sysctl_createv(clog, 0, NULL, NULL, 1124 CTLFLAG_PERMANENT, 1125 CTLTYPE_NODE, "inet", NULL, 1126 NULL, 0, NULL, 0, 1127 CTL_NET, PF_INET, CTL_EOL); 1128 sysctl_createv(clog, 0, NULL, NULL, 1129 CTLFLAG_PERMANENT, 1130 CTLTYPE_NODE, "udp", 1131 SYSCTL_DESCR("UDPv4 related settings"), 1132 NULL, 0, NULL, 0, 1133 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL); 1134 1135 sysctl_createv(clog, 0, NULL, NULL, 1136 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1137 CTLTYPE_INT, "checksum", 1138 SYSCTL_DESCR("Compute and check UDP checksums"), 1139 NULL, 0, &udpcksum, 0, 1140 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM, 1141 CTL_EOL); 1142 sysctl_createv(clog, 0, NULL, NULL, 1143 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1144 CTLTYPE_INT, "sendspace", 1145 SYSCTL_DESCR("Default UDP send buffer size"), 1146 NULL, 0, &udp_sendspace, 0, 1147 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE, 1148 CTL_EOL); 1149 sysctl_createv(clog, 0, NULL, NULL, 1150 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1151 CTLTYPE_INT, "recvspace", 1152 SYSCTL_DESCR("Default UDP receive buffer size"), 1153 NULL, 0, &udp_recvspace, 0, 1154 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE, 1155 CTL_EOL); 1156 } 1157 #endif 1158
Indexes created Fri Sep 26 20:09:58 GMT 2025