1 /* $NetBSD: udp_usrreq.c,v 1.93 2002/05/12 20:33:51 matt 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. All advertising materials mentioning features or use of this software 45 * must display the following acknowledgement: 46 * This product includes software developed by the University of 47 * California, Berkeley and its contributors. 48 * 4. Neither the name of the University nor the names of its contributors 49 * may be used to endorse or promote products derived from this software 50 * without specific prior written permission. 51 * 52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 62 * SUCH DAMAGE. 63 * 64 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 65 */ 66 67 #include <sys/cdefs.h> 68 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.93 2002/05/12 20:33:51 matt Exp $"); 69 70 #include "opt_inet.h" 71 #include "opt_ipsec.h" 72 #include "opt_inet_csum.h" 73 #include "opt_ipkdb.h" 74 75 #include <sys/param.h> 76 #include <sys/malloc.h> 77 #include <sys/mbuf.h> 78 #include <sys/protosw.h> 79 #include <sys/socket.h> 80 #include <sys/socketvar.h> 81 #include <sys/errno.h> 82 #include <sys/stat.h> 83 #include <sys/systm.h> 84 #include <sys/proc.h> 85 #include <sys/domain.h> 86 #include <sys/sysctl.h> 87 88 #include <net/if.h> 89 #include <net/route.h> 90 91 #include <netinet/in.h> 92 #include <netinet/in_systm.h> 93 #include <netinet/in_var.h> 94 #include <netinet/ip.h> 95 #include <netinet/in_pcb.h> 96 #include <netinet/ip_var.h> 97 #include <netinet/ip_icmp.h> 98 #include <netinet/udp.h> 99 #include <netinet/udp_var.h> 100 101 #ifdef INET6 102 #include <netinet/ip6.h> 103 #include <netinet/icmp6.h> 104 #include <netinet6/ip6_var.h> 105 #include <netinet6/in6_pcb.h> 106 #include <netinet6/udp6_var.h> 107 #endif 108 109 #ifdef PULLDOWN_TEST 110 #ifndef INET6 111 /* always need ip6.h for IP6_EXTHDR_GET */ 112 #include <netinet/ip6.h> 113 #endif 114 #endif 115 116 #include "faith.h" 117 #if defined(NFAITH) && NFAITH > 0 118 #include <net/if_faith.h> 119 #endif 120 121 #include <machine/stdarg.h> 122 123 #ifdef IPSEC 124 #include <netinet6/ipsec.h> 125 #include <netkey/key.h> 126 #endif /*IPSEC*/ 127 128 #ifdef IPKDB 129 #include <ipkdb/ipkdb.h> 130 #endif 131 132 /* 133 * UDP protocol implementation. 134 * Per RFC 768, August, 1980. 135 */ 136 #ifndef COMPAT_42 137 int udpcksum = 1; 138 #else 139 int udpcksum = 0; /* XXX */ 140 #endif 141 142 struct inpcbtable udbtable; 143 struct udpstat udpstat; 144 145 #ifdef INET 146 static void udp4_sendup __P((struct mbuf *, int, struct sockaddr *, 147 struct socket *)); 148 static int udp4_realinput __P((struct sockaddr_in *, struct sockaddr_in *, 149 struct mbuf *, int)); 150 #endif 151 #ifdef INET6 152 static void udp6_sendup __P((struct mbuf *, int, struct sockaddr *, 153 struct socket *)); 154 static int udp6_realinput __P((int, struct sockaddr_in6 *, 155 struct sockaddr_in6 *, struct mbuf *, int)); 156 #endif 157 #ifdef INET 158 static void udp_notify __P((struct inpcb *, int)); 159 #endif 160 161 #ifndef UDBHASHSIZE 162 #define UDBHASHSIZE 128 163 #endif 164 int udbhashsize = UDBHASHSIZE; 165 166 #ifdef UDP_CSUM_COUNTERS 167 #include <sys/device.h> 168 169 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 170 NULL, "udp", "hwcsum bad"); 171 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 172 NULL, "udp", "hwcsum ok"); 173 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 174 NULL, "udp", "hwcsum data"); 175 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 176 NULL, "udp", "swcsum"); 177 178 #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 179 180 #else 181 182 #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */ 183 184 #endif /* UDP_CSUM_COUNTERS */ 185 186 void 187 udp_init() 188 { 189 190 #ifdef INET 191 in_pcbinit(&udbtable, udbhashsize, udbhashsize); 192 #endif 193 194 #ifdef UDP_CSUM_COUNTERS 195 evcnt_attach_static(&udp_hwcsum_bad); 196 evcnt_attach_static(&udp_hwcsum_ok); 197 evcnt_attach_static(&udp_hwcsum_data); 198 evcnt_attach_static(&udp_swcsum); 199 #endif /* UDP_CSUM_COUNTERS */ 200 } 201 202 #ifdef INET 203 void 204 #if __STDC__ 205 udp_input(struct mbuf *m, ...) 206 #else 207 udp_input(m, va_alist) 208 struct mbuf *m; 209 va_dcl 210 #endif 211 { 212 va_list ap; 213 struct sockaddr_in src, dst; 214 struct ip *ip; 215 struct udphdr *uh; 216 int iphlen, proto; 217 int len; 218 int n; 219 220 va_start(ap, m); 221 iphlen = va_arg(ap, int); 222 proto = va_arg(ap, int); 223 va_end(ap); 224 225 udpstat.udps_ipackets++; 226 227 #ifndef PULLDOWN_TEST 228 /* 229 * Strip IP options, if any; should skip this, 230 * make available to user, and use on returned packets, 231 * but we don't yet have a way to check the checksum 232 * with options still present. 233 */ 234 if (iphlen > sizeof (struct ip)) { 235 ip_stripoptions(m, (struct mbuf *)0); 236 iphlen = sizeof(struct ip); 237 } 238 #else 239 /* 240 * we may enable the above code if we save and pass IPv4 options 241 * to the userland. 242 */ 243 #endif 244 245 /* 246 * Get IP and UDP header together in first mbuf. 247 */ 248 ip = mtod(m, struct ip *); 249 #ifndef PULLDOWN_TEST 250 if (m->m_len < iphlen + sizeof(struct udphdr)) { 251 if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) { 252 udpstat.udps_hdrops++; 253 return; 254 } 255 ip = mtod(m, struct ip *); 256 } 257 uh = (struct udphdr *)((caddr_t)ip + iphlen); 258 #else 259 IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 260 if (uh == NULL) { 261 udpstat.udps_hdrops++; 262 return; 263 } 264 #endif 265 266 /* destination port of 0 is illegal, based on RFC768. */ 267 if (uh->uh_dport == 0) 268 goto bad; 269 270 /* 271 * Make mbuf data length reflect UDP length. 272 * If not enough data to reflect UDP length, drop. 273 */ 274 len = ntohs((u_int16_t)uh->uh_ulen); 275 if (ip->ip_len != iphlen + len) { 276 if (ip->ip_len < iphlen + len || len < sizeof(struct udphdr)) { 277 udpstat.udps_badlen++; 278 goto bad; 279 } 280 m_adj(m, iphlen + len - ip->ip_len); 281 } 282 283 /* 284 * Checksum extended UDP header and data. 285 */ 286 if (uh->uh_sum) { 287 switch (m->m_pkthdr.csum_flags & 288 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) | 289 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 290 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD: 291 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad); 292 goto badcsum; 293 294 case M_CSUM_UDPv4|M_CSUM_DATA: 295 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); 296 if ((m->m_pkthdr.csum_data ^ 0xffff) != 0) 297 goto badcsum; 298 break; 299 300 case M_CSUM_UDPv4: 301 /* Checksum was okay. */ 302 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); 303 break; 304 305 default: 306 /* Need to compute it ourselves. */ 307 UDP_CSUM_COUNTER_INCR(&udp_swcsum); 308 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) 309 goto badcsum; 310 break; 311 } 312 } 313 314 /* construct source and dst sockaddrs. */ 315 bzero(&src, sizeof(src)); 316 src.sin_family = AF_INET; 317 src.sin_len = sizeof(struct sockaddr_in); 318 bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr)); 319 src.sin_port = uh->uh_sport; 320 bzero(&dst, sizeof(dst)); 321 dst.sin_family = AF_INET; 322 dst.sin_len = sizeof(struct sockaddr_in); 323 bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr)); 324 dst.sin_port = uh->uh_dport; 325 326 n = udp4_realinput(&src, &dst, m, iphlen); 327 #ifdef INET6 328 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { 329 struct sockaddr_in6 src6, dst6; 330 331 bzero(&src6, sizeof(src6)); 332 src6.sin6_family = AF_INET6; 333 src6.sin6_len = sizeof(struct sockaddr_in6); 334 src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff; 335 bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12], 336 sizeof(ip->ip_src)); 337 src6.sin6_port = uh->uh_sport; 338 bzero(&dst6, sizeof(dst6)); 339 dst6.sin6_family = AF_INET6; 340 dst6.sin6_len = sizeof(struct sockaddr_in6); 341 dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff; 342 bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12], 343 sizeof(ip->ip_dst)); 344 dst6.sin6_port = uh->uh_dport; 345 346 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen); 347 } 348 #endif 349 350 if (n == 0) { 351 if (m->m_flags & (M_BCAST | M_MCAST)) { 352 udpstat.udps_noportbcast++; 353 goto bad; 354 } 355 udpstat.udps_noport++; 356 #ifdef IPKDB 357 if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport, 358 m, iphlen + sizeof(struct udphdr), 359 m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) { 360 /* 361 * It was a debugger connect packet, 362 * just drop it now 363 */ 364 goto bad; 365 } 366 #endif 367 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 368 m = NULL; 369 } 370 371 bad: 372 if (m) 373 m_freem(m); 374 return; 375 376 badcsum: 377 m_freem(m); 378 udpstat.udps_badsum++; 379 } 380 #endif 381 382 #ifdef INET6 383 int 384 udp6_input(mp, offp, proto) 385 struct mbuf **mp; 386 int *offp, proto; 387 { 388 struct mbuf *m = *mp; 389 int off = *offp; 390 struct sockaddr_in6 src, dst; 391 struct ip6_hdr *ip6; 392 struct udphdr *uh; 393 u_int32_t plen, ulen; 394 395 #ifndef PULLDOWN_TEST 396 IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE); 397 #endif 398 ip6 = mtod(m, struct ip6_hdr *); 399 400 #if defined(NFAITH) && 0 < NFAITH 401 if (faithprefix(&ip6->ip6_dst)) { 402 /* send icmp6 host unreach? */ 403 m_freem(m); 404 return IPPROTO_DONE; 405 } 406 #endif 407 408 udp6stat.udp6s_ipackets++; 409 410 /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */ 411 plen = m->m_pkthdr.len - off; 412 #ifndef PULLDOWN_TEST 413 uh = (struct udphdr *)((caddr_t)ip6 + off); 414 #else 415 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr)); 416 if (uh == NULL) { 417 ip6stat.ip6s_tooshort++; 418 return IPPROTO_DONE; 419 } 420 #endif 421 ulen = ntohs((u_short)uh->uh_ulen); 422 /* 423 * RFC2675 section 4: jumbograms will have 0 in the UDP header field, 424 * iff payload length > 0xffff. 425 */ 426 if (ulen == 0 && plen > 0xffff) 427 ulen = plen; 428 429 if (plen != ulen) { 430 udp6stat.udp6s_badlen++; 431 goto bad; 432 } 433 434 /* destination port of 0 is illegal, based on RFC768. */ 435 if (uh->uh_dport == 0) 436 goto bad; 437 438 /* Be proactive about malicious use of IPv4 mapped address */ 439 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 440 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 441 /* XXX stat */ 442 goto bad; 443 } 444 445 /* 446 * Checksum extended UDP header and data. 447 */ 448 if (uh->uh_sum == 0) 449 udp6stat.udp6s_nosum++; 450 else if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) { 451 udp6stat.udp6s_badsum++; 452 goto bad; 453 } 454 455 /* 456 * Construct source and dst sockaddrs. 457 * Note that ifindex (s6_addr16[1]) is already filled. 458 */ 459 bzero(&src, sizeof(src)); 460 src.sin6_family = AF_INET6; 461 src.sin6_len = sizeof(struct sockaddr_in6); 462 /* KAME hack: recover scopeid */ 463 (void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif); 464 src.sin6_port = uh->uh_sport; 465 bzero(&dst, sizeof(dst)); 466 dst.sin6_family = AF_INET6; 467 dst.sin6_len = sizeof(struct sockaddr_in6); 468 /* KAME hack: recover scopeid */ 469 (void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif); 470 dst.sin6_port = uh->uh_dport; 471 472 if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) { 473 if (m->m_flags & M_MCAST) { 474 udp6stat.udp6s_noportmcast++; 475 goto bad; 476 } 477 udp6stat.udp6s_noport++; 478 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); 479 m = NULL; 480 } 481 482 bad: 483 if (m) 484 m_freem(m); 485 return IPPROTO_DONE; 486 } 487 #endif 488 489 #ifdef INET 490 static void 491 udp4_sendup(m, off, src, so) 492 struct mbuf *m; 493 int off; /* offset of data portion */ 494 struct sockaddr *src; 495 struct socket *so; 496 { 497 struct mbuf *opts = NULL; 498 struct mbuf *n; 499 struct inpcb *inp = NULL; 500 #ifdef INET6 501 struct in6pcb *in6p = NULL; 502 #endif 503 504 if (!so) 505 return; 506 switch (so->so_proto->pr_domain->dom_family) { 507 case AF_INET: 508 inp = sotoinpcb(so); 509 break; 510 #ifdef INET6 511 case AF_INET6: 512 in6p = sotoin6pcb(so); 513 break; 514 #endif 515 default: 516 return; 517 } 518 519 #ifdef IPSEC 520 /* check AH/ESP integrity. */ 521 if (so != NULL && ipsec4_in_reject_so(m, so)) { 522 ipsecstat.in_polvio++; 523 return; 524 } 525 #endif /*IPSEC*/ 526 527 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { 528 if (inp && (inp->inp_flags & INP_CONTROLOPTS 529 || so->so_options & SO_TIMESTAMP)) { 530 struct ip *ip = mtod(n, struct ip *); 531 ip_savecontrol(inp, &opts, ip, n); 532 } 533 534 m_adj(n, off); 535 if (sbappendaddr(&so->so_rcv, src, n, 536 opts) == 0) { 537 m_freem(n); 538 if (opts) 539 m_freem(opts); 540 udpstat.udps_fullsock++; 541 } else 542 sorwakeup(so); 543 } 544 } 545 #endif 546 547 #ifdef INET6 548 static void 549 udp6_sendup(m, off, src, so) 550 struct mbuf *m; 551 int off; /* offset of data portion */ 552 struct sockaddr *src; 553 struct socket *so; 554 { 555 struct mbuf *opts = NULL; 556 struct mbuf *n; 557 struct in6pcb *in6p = NULL; 558 559 if (!so) 560 return; 561 if (so->so_proto->pr_domain->dom_family != AF_INET6) 562 return; 563 in6p = sotoin6pcb(so); 564 565 #ifdef IPSEC 566 /* check AH/ESP integrity. */ 567 if (so != NULL && ipsec6_in_reject_so(m, so)) { 568 ipsec6stat.in_polvio++; 569 return; 570 } 571 #endif /*IPSEC*/ 572 573 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { 574 if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS 575 || in6p->in6p_socket->so_options & SO_TIMESTAMP)) { 576 struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *); 577 ip6_savecontrol(in6p, &opts, ip6, n); 578 } 579 580 m_adj(n, off); 581 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) { 582 m_freem(n); 583 if (opts) 584 m_freem(opts); 585 udp6stat.udp6s_fullsock++; 586 } else 587 sorwakeup(so); 588 } 589 } 590 #endif 591 592 #ifdef INET 593 static int 594 udp4_realinput(src, dst, m, off) 595 struct sockaddr_in *src; 596 struct sockaddr_in *dst; 597 struct mbuf *m; 598 int off; /* offset of udphdr */ 599 { 600 u_int16_t *sport, *dport; 601 int rcvcnt; 602 struct in_addr *src4, *dst4; 603 struct inpcb *inp; 604 605 rcvcnt = 0; 606 off += sizeof(struct udphdr); /* now, offset of payload */ 607 608 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 609 goto bad; 610 611 src4 = &src->sin_addr; 612 sport = &src->sin_port; 613 dst4 = &dst->sin_addr; 614 dport = &dst->sin_port; 615 616 if (IN_MULTICAST(dst4->s_addr) || 617 in_broadcast(*dst4, m->m_pkthdr.rcvif)) { 618 struct inpcb *last; 619 /* 620 * Deliver a multicast or broadcast datagram to *all* sockets 621 * for which the local and remote addresses and ports match 622 * those of the incoming datagram. This allows more than 623 * one process to receive multi/broadcasts on the same port. 624 * (This really ought to be done for unicast datagrams as 625 * well, but that would cause problems with existing 626 * applications that open both address-specific sockets and 627 * a wildcard socket listening to the same port -- they would 628 * end up receiving duplicates of every unicast datagram. 629 * Those applications open the multiple sockets to overcome an 630 * inadequacy of the UDP socket interface, but for backwards 631 * compatibility we avoid the problem here rather than 632 * fixing the interface. Maybe 4.5BSD will remedy this?) 633 */ 634 635 /* 636 * KAME note: traditionally we dropped udpiphdr from mbuf here. 637 * we need udpiphdr for IPsec processing so we do that later. 638 */ 639 /* 640 * Locate pcb(s) for datagram. 641 */ 642 CIRCLEQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) { 643 if (inp->inp_lport != *dport) 644 continue; 645 if (!in_nullhost(inp->inp_laddr)) { 646 if (!in_hosteq(inp->inp_laddr, *dst4)) 647 continue; 648 } 649 if (!in_nullhost(inp->inp_faddr)) { 650 if (!in_hosteq(inp->inp_faddr, *src4) || 651 inp->inp_fport != *sport) 652 continue; 653 } 654 655 last = inp; 656 udp4_sendup(m, off, (struct sockaddr *)src, 657 inp->inp_socket); 658 rcvcnt++; 659 660 /* 661 * Don't look for additional matches if this one does 662 * not have either the SO_REUSEPORT or SO_REUSEADDR 663 * socket options set. This heuristic avoids searching 664 * through all pcbs in the common case of a non-shared 665 * port. It assumes that an application will never 666 * clear these options after setting them. 667 */ 668 if ((inp->inp_socket->so_options & 669 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 670 break; 671 } 672 } else { 673 /* 674 * Locate pcb for datagram. 675 */ 676 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport); 677 if (inp == 0) { 678 ++udpstat.udps_pcbhashmiss; 679 inp = in_pcblookup_bind(&udbtable, *dst4, *dport); 680 if (inp == 0) 681 return rcvcnt; 682 } 683 684 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 685 rcvcnt++; 686 } 687 688 bad: 689 return rcvcnt; 690 } 691 #endif 692 693 #ifdef INET6 694 static int 695 udp6_realinput(af, src, dst, m, off) 696 int af; /* af on packet */ 697 struct sockaddr_in6 *src; 698 struct sockaddr_in6 *dst; 699 struct mbuf *m; 700 int off; /* offset of udphdr */ 701 { 702 u_int16_t sport, dport; 703 int rcvcnt; 704 struct in6_addr src6, dst6; 705 const struct in_addr *dst4; 706 struct in6pcb *in6p; 707 708 rcvcnt = 0; 709 off += sizeof(struct udphdr); /* now, offset of payload */ 710 711 if (af != AF_INET && af != AF_INET6) 712 goto bad; 713 if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6) 714 goto bad; 715 716 in6_embedscope(&src6, src, NULL, NULL); 717 sport = src->sin6_port; 718 in6_embedscope(&dst6, dst, NULL, NULL); 719 dport = dst->sin6_port; 720 dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12]; 721 722 if (IN6_IS_ADDR_MULTICAST(&dst6) || 723 (af == AF_INET && IN_MULTICAST(dst4->s_addr))) { 724 struct in6pcb *last; 725 /* 726 * Deliver a multicast or broadcast datagram to *all* sockets 727 * for which the local and remote addresses and ports match 728 * those of the incoming datagram. This allows more than 729 * one process to receive multi/broadcasts on the same port. 730 * (This really ought to be done for unicast datagrams as 731 * well, but that would cause problems with existing 732 * applications that open both address-specific sockets and 733 * a wildcard socket listening to the same port -- they would 734 * end up receiving duplicates of every unicast datagram. 735 * Those applications open the multiple sockets to overcome an 736 * inadequacy of the UDP socket interface, but for backwards 737 * compatibility we avoid the problem here rather than 738 * fixing the interface. Maybe 4.5BSD will remedy this?) 739 */ 740 741 /* 742 * KAME note: traditionally we dropped udpiphdr from mbuf here. 743 * we need udpiphdr for IPsec processing so we do that later. 744 */ 745 /* 746 * Locate pcb(s) for datagram. 747 */ 748 for (in6p = udb6.in6p_next; in6p != &udb6; 749 in6p = in6p->in6p_next) { 750 if (in6p->in6p_lport != dport) 751 continue; 752 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 753 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6)) 754 continue; 755 } else { 756 if (IN6_IS_ADDR_V4MAPPED(&dst6) && 757 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 758 continue; 759 } 760 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 761 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, 762 &src6) || in6p->in6p_fport != sport) 763 continue; 764 } else { 765 if (IN6_IS_ADDR_V4MAPPED(&src6) && 766 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 767 continue; 768 } 769 770 last = in6p; 771 udp6_sendup(m, off, (struct sockaddr *)src, 772 in6p->in6p_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 ((in6p->in6p_socket->so_options & 784 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 785 break; 786 } 787 } else { 788 /* 789 * Locate pcb for datagram. 790 */ 791 in6p = in6_pcblookup_connect(&udb6, &src6, sport, 792 &dst6, dport, 0); 793 if (in6p == 0) { 794 ++udpstat.udps_pcbhashmiss; 795 in6p = in6_pcblookup_bind(&udb6, &dst6, dport, 0); 796 if (in6p == 0) 797 return rcvcnt; 798 } 799 800 udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket); 801 rcvcnt++; 802 } 803 804 bad: 805 return rcvcnt; 806 } 807 #endif 808 809 #ifdef INET 810 /* 811 * Notify a udp user of an asynchronous error; 812 * just wake up so that he can collect error status. 813 */ 814 static void 815 udp_notify(inp, errno) 816 struct inpcb *inp; 817 int errno; 818 { 819 820 inp->inp_socket->so_error = errno; 821 sorwakeup(inp->inp_socket); 822 sowwakeup(inp->inp_socket); 823 } 824 825 void * 826 udp_ctlinput(cmd, sa, v) 827 int cmd; 828 struct sockaddr *sa; 829 void *v; 830 { 831 struct ip *ip = v; 832 struct udphdr *uh; 833 void (*notify) __P((struct inpcb *, int)) = udp_notify; 834 int errno; 835 836 if (sa->sa_family != AF_INET 837 || sa->sa_len != sizeof(struct sockaddr_in)) 838 return NULL; 839 if ((unsigned)cmd >= PRC_NCMDS) 840 return NULL; 841 errno = inetctlerrmap[cmd]; 842 if (PRC_IS_REDIRECT(cmd)) 843 notify = in_rtchange, ip = 0; 844 else if (cmd == PRC_HOSTDEAD) 845 ip = 0; 846 else if (errno == 0) 847 return NULL; 848 if (ip) { 849 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 850 in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport, 851 ip->ip_src, uh->uh_sport, errno, notify); 852 853 /* XXX mapped address case */ 854 } else 855 in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno, 856 notify); 857 return NULL; 858 } 859 860 int 861 #if __STDC__ 862 udp_output(struct mbuf *m, ...) 863 #else 864 udp_output(m, va_alist) 865 struct mbuf *m; 866 va_dcl 867 #endif 868 { 869 struct inpcb *inp; 870 struct udpiphdr *ui; 871 int len = m->m_pkthdr.len; 872 int error = 0; 873 va_list ap; 874 875 va_start(ap, m); 876 inp = va_arg(ap, struct inpcb *); 877 va_end(ap); 878 879 /* 880 * Calculate data length and get a mbuf 881 * for UDP and IP headers. 882 */ 883 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 884 if (m == 0) { 885 error = ENOBUFS; 886 goto release; 887 } 888 889 /* 890 * Compute the packet length of the IP header, and 891 * punt if the length looks bogus. 892 */ 893 if ((len + sizeof(struct udpiphdr)) > IP_MAXPACKET) { 894 error = EMSGSIZE; 895 goto release; 896 } 897 898 /* 899 * Fill in mbuf with extended UDP header 900 * and addresses and length put into network format. 901 */ 902 ui = mtod(m, struct udpiphdr *); 903 ui->ui_pr = IPPROTO_UDP; 904 ui->ui_src = inp->inp_laddr; 905 ui->ui_dst = inp->inp_faddr; 906 ui->ui_sport = inp->inp_lport; 907 ui->ui_dport = inp->inp_fport; 908 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 909 910 /* 911 * Set up checksum and output datagram. 912 */ 913 if (udpcksum) { 914 /* 915 * XXX Cache pseudo-header checksum part for 916 * XXX "connected" UDP sockets. 917 */ 918 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 919 ui->ui_dst.s_addr, htons((u_int16_t)len + 920 sizeof(struct udphdr) + IPPROTO_UDP)); 921 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 922 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 923 } else 924 ui->ui_sum = 0; 925 ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len; 926 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ 927 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ 928 udpstat.udps_opackets++; 929 930 #ifdef IPSEC 931 if (ipsec_setsocket(m, inp->inp_socket) != 0) { 932 error = ENOBUFS; 933 goto release; 934 } 935 #endif /*IPSEC*/ 936 937 return (ip_output(m, inp->inp_options, &inp->inp_route, 938 inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), 939 inp->inp_moptions)); 940 941 release: 942 m_freem(m); 943 return (error); 944 } 945 946 int udp_sendspace = 9216; /* really max datagram size */ 947 int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); 948 /* 40 1K datagrams */ 949 950 /*ARGSUSED*/ 951 int 952 udp_usrreq(so, req, m, nam, control, p) 953 struct socket *so; 954 int req; 955 struct mbuf *m, *nam, *control; 956 struct proc *p; 957 { 958 struct inpcb *inp; 959 int s; 960 int error = 0; 961 962 if (req == PRU_CONTROL) 963 return (in_control(so, (long)m, (caddr_t)nam, 964 (struct ifnet *)control, p)); 965 966 if (req == PRU_PURGEIF) { 967 in_pcbpurgeif0(&udbtable, (struct ifnet *)control); 968 in_purgeif((struct ifnet *)control); 969 in_pcbpurgeif(&udbtable, (struct ifnet *)control); 970 return (0); 971 } 972 973 s = splsoftnet(); 974 inp = sotoinpcb(so); 975 #ifdef DIAGNOSTIC 976 if (req != PRU_SEND && req != PRU_SENDOOB && control) 977 panic("udp_usrreq: unexpected control mbuf"); 978 #endif 979 if (inp == 0 && req != PRU_ATTACH) { 980 error = EINVAL; 981 goto release; 982 } 983 984 /* 985 * Note: need to block udp_input while changing 986 * the udp pcb queue and/or pcb addresses. 987 */ 988 switch (req) { 989 990 case PRU_ATTACH: 991 if (inp != 0) { 992 error = EISCONN; 993 break; 994 } 995 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 996 error = soreserve(so, udp_sendspace, udp_recvspace); 997 if (error) 998 break; 999 } 1000 error = in_pcballoc(so, &udbtable); 1001 if (error) 1002 break; 1003 inp = sotoinpcb(so); 1004 inp->inp_ip.ip_ttl = ip_defttl; 1005 break; 1006 1007 case PRU_DETACH: 1008 in_pcbdetach(inp); 1009 break; 1010 1011 case PRU_BIND: 1012 error = in_pcbbind(inp, nam, p); 1013 break; 1014 1015 case PRU_LISTEN: 1016 error = EOPNOTSUPP; 1017 break; 1018 1019 case PRU_CONNECT: 1020 error = in_pcbconnect(inp, nam); 1021 if (error) 1022 break; 1023 soisconnected(so); 1024 break; 1025 1026 case PRU_CONNECT2: 1027 error = EOPNOTSUPP; 1028 break; 1029 1030 case PRU_DISCONNECT: 1031 /*soisdisconnected(so);*/ 1032 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 1033 in_pcbdisconnect(inp); 1034 inp->inp_laddr = zeroin_addr; /* XXX */ 1035 in_pcbstate(inp, INP_BOUND); /* XXX */ 1036 break; 1037 1038 case PRU_SHUTDOWN: 1039 socantsendmore(so); 1040 break; 1041 1042 case PRU_RCVD: 1043 error = EOPNOTSUPP; 1044 break; 1045 1046 case PRU_SEND: 1047 if (control && control->m_len) { 1048 m_freem(control); 1049 m_freem(m); 1050 error = EINVAL; 1051 break; 1052 } 1053 { 1054 struct in_addr laddr; /* XXX */ 1055 1056 if (nam) { 1057 laddr = inp->inp_laddr; /* XXX */ 1058 if ((so->so_state & SS_ISCONNECTED) != 0) { 1059 error = EISCONN; 1060 goto die; 1061 } 1062 error = in_pcbconnect(inp, nam); 1063 if (error) { 1064 die: 1065 m_freem(m); 1066 break; 1067 } 1068 } else { 1069 if ((so->so_state & SS_ISCONNECTED) == 0) { 1070 error = ENOTCONN; 1071 goto die; 1072 } 1073 } 1074 error = udp_output(m, inp); 1075 if (nam) { 1076 in_pcbdisconnect(inp); 1077 inp->inp_laddr = laddr; /* XXX */ 1078 in_pcbstate(inp, INP_BOUND); /* XXX */ 1079 } 1080 } 1081 break; 1082 1083 case PRU_SENSE: 1084 /* 1085 * stat: don't bother with a blocksize. 1086 */ 1087 splx(s); 1088 return (0); 1089 1090 case PRU_RCVOOB: 1091 error = EOPNOTSUPP; 1092 break; 1093 1094 case PRU_SENDOOB: 1095 m_freem(control); 1096 m_freem(m); 1097 error = EOPNOTSUPP; 1098 break; 1099 1100 case PRU_SOCKADDR: 1101 in_setsockaddr(inp, nam); 1102 break; 1103 1104 case PRU_PEERADDR: 1105 in_setpeeraddr(inp, nam); 1106 break; 1107 1108 default: 1109 panic("udp_usrreq"); 1110 } 1111 1112 release: 1113 splx(s); 1114 return (error); 1115 } 1116 1117 /* 1118 * Sysctl for udp variables. 1119 */ 1120 int 1121 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1122 int *name; 1123 u_int namelen; 1124 void *oldp; 1125 size_t *oldlenp; 1126 void *newp; 1127 size_t newlen; 1128 { 1129 /* All sysctl names at this level are terminal. */ 1130 if (namelen != 1) 1131 return (ENOTDIR); 1132 1133 switch (name[0]) { 1134 case UDPCTL_CHECKSUM: 1135 return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum)); 1136 case UDPCTL_SENDSPACE: 1137 return (sysctl_int(oldp, oldlenp, newp, newlen, 1138 &udp_sendspace)); 1139 case UDPCTL_RECVSPACE: 1140 return (sysctl_int(oldp, oldlenp, newp, newlen, 1141 &udp_recvspace)); 1142 default: 1143 return (ENOPROTOOPT); 1144 } 1145 /* NOTREACHED */ 1146 } 1147 #endif 1148