1 /* $NetBSD: udp_usrreq.c,v 1.96 2002/08/14 00:23:36 itojun 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 1.96 2002/08/14 00:23:36 itojun 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 u_int16_t ip_len; 220 221 va_start(ap, m); 222 iphlen = va_arg(ap, int); 223 proto = va_arg(ap, int); 224 va_end(ap); 225 226 udpstat.udps_ipackets++; 227 228 #ifndef PULLDOWN_TEST 229 /* 230 * Strip IP options, if any; should skip this, 231 * make available to user, and use on returned packets, 232 * but we don't yet have a way to check the checksum 233 * with options still present. 234 */ 235 if (iphlen > sizeof (struct ip)) { 236 ip_stripoptions(m, (struct mbuf *)0); 237 iphlen = sizeof(struct ip); 238 } 239 #else 240 /* 241 * we may enable the above code if we save and pass IPv4 options 242 * to the userland. 243 */ 244 #endif 245 246 /* 247 * Get IP and UDP header together in first mbuf. 248 */ 249 ip = mtod(m, struct ip *); 250 #ifndef PULLDOWN_TEST 251 if (m->m_len < iphlen + sizeof(struct udphdr)) { 252 if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) { 253 udpstat.udps_hdrops++; 254 return; 255 } 256 ip = mtod(m, struct ip *); 257 } 258 uh = (struct udphdr *)((caddr_t)ip + iphlen); 259 #else 260 IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 261 if (uh == NULL) { 262 udpstat.udps_hdrops++; 263 return; 264 } 265 #endif 266 KASSERT(UDP_HDR_ALIGNED_P(uh)); 267 268 /* destination port of 0 is illegal, based on RFC768. */ 269 if (uh->uh_dport == 0) 270 goto bad; 271 272 /* 273 * Make mbuf data length reflect UDP length. 274 * If not enough data to reflect UDP length, drop. 275 */ 276 ip_len = ntohs(ip->ip_len); 277 len = ntohs((u_int16_t)uh->uh_ulen); 278 if (ip_len != iphlen + len) { 279 if (ip_len < iphlen + len || len < sizeof(struct udphdr)) { 280 udpstat.udps_badlen++; 281 goto bad; 282 } 283 m_adj(m, iphlen + len - ip_len); 284 } 285 286 /* 287 * Checksum extended UDP header and data. 288 */ 289 if (uh->uh_sum) { 290 switch (m->m_pkthdr.csum_flags & 291 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) | 292 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 293 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD: 294 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad); 295 goto badcsum; 296 297 case M_CSUM_UDPv4|M_CSUM_DATA: 298 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); 299 if ((m->m_pkthdr.csum_data ^ 0xffff) != 0) 300 goto badcsum; 301 break; 302 303 case M_CSUM_UDPv4: 304 /* Checksum was okay. */ 305 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); 306 break; 307 308 default: 309 /* Need to compute it ourselves. */ 310 UDP_CSUM_COUNTER_INCR(&udp_swcsum); 311 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) 312 goto badcsum; 313 break; 314 } 315 } 316 317 /* construct source and dst sockaddrs. */ 318 bzero(&src, sizeof(src)); 319 src.sin_family = AF_INET; 320 src.sin_len = sizeof(struct sockaddr_in); 321 bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr)); 322 src.sin_port = uh->uh_sport; 323 bzero(&dst, sizeof(dst)); 324 dst.sin_family = AF_INET; 325 dst.sin_len = sizeof(struct sockaddr_in); 326 bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr)); 327 dst.sin_port = uh->uh_dport; 328 329 n = udp4_realinput(&src, &dst, m, iphlen); 330 #ifdef INET6 331 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { 332 struct sockaddr_in6 src6, dst6; 333 334 bzero(&src6, sizeof(src6)); 335 src6.sin6_family = AF_INET6; 336 src6.sin6_len = sizeof(struct sockaddr_in6); 337 src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff; 338 bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12], 339 sizeof(ip->ip_src)); 340 src6.sin6_port = uh->uh_sport; 341 bzero(&dst6, sizeof(dst6)); 342 dst6.sin6_family = AF_INET6; 343 dst6.sin6_len = sizeof(struct sockaddr_in6); 344 dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff; 345 bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12], 346 sizeof(ip->ip_dst)); 347 dst6.sin6_port = uh->uh_dport; 348 349 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen); 350 } 351 #endif 352 353 if (n == 0) { 354 if (m->m_flags & (M_BCAST | M_MCAST)) { 355 udpstat.udps_noportbcast++; 356 goto bad; 357 } 358 udpstat.udps_noport++; 359 #ifdef IPKDB 360 if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport, 361 m, iphlen + sizeof(struct udphdr), 362 m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) { 363 /* 364 * It was a debugger connect packet, 365 * just drop it now 366 */ 367 goto bad; 368 } 369 #endif 370 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 371 m = NULL; 372 } 373 374 bad: 375 if (m) 376 m_freem(m); 377 return; 378 379 badcsum: 380 m_freem(m); 381 udpstat.udps_badsum++; 382 } 383 #endif 384 385 #ifdef INET6 386 int 387 udp6_input(mp, offp, proto) 388 struct mbuf **mp; 389 int *offp, proto; 390 { 391 struct mbuf *m = *mp; 392 int off = *offp; 393 struct sockaddr_in6 src, dst; 394 struct ip6_hdr *ip6; 395 struct udphdr *uh; 396 u_int32_t plen, ulen; 397 398 #ifndef PULLDOWN_TEST 399 IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE); 400 #endif 401 ip6 = mtod(m, struct ip6_hdr *); 402 403 #if defined(NFAITH) && 0 < NFAITH 404 if (faithprefix(&ip6->ip6_dst)) { 405 /* send icmp6 host unreach? */ 406 m_freem(m); 407 return IPPROTO_DONE; 408 } 409 #endif 410 411 udp6stat.udp6s_ipackets++; 412 413 /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */ 414 plen = m->m_pkthdr.len - off; 415 #ifndef PULLDOWN_TEST 416 uh = (struct udphdr *)((caddr_t)ip6 + off); 417 #else 418 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr)); 419 if (uh == NULL) { 420 ip6stat.ip6s_tooshort++; 421 return IPPROTO_DONE; 422 } 423 #endif 424 KASSERT(UDP_HDR_ALIGNED_P(uh)); 425 ulen = ntohs((u_short)uh->uh_ulen); 426 /* 427 * RFC2675 section 4: jumbograms will have 0 in the UDP header field, 428 * iff payload length > 0xffff. 429 */ 430 if (ulen == 0 && plen > 0xffff) 431 ulen = plen; 432 433 if (plen != ulen) { 434 udp6stat.udp6s_badlen++; 435 goto bad; 436 } 437 438 /* destination port of 0 is illegal, based on RFC768. */ 439 if (uh->uh_dport == 0) 440 goto bad; 441 442 /* Be proactive about malicious use of IPv4 mapped address */ 443 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 444 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 445 /* XXX stat */ 446 goto bad; 447 } 448 449 /* 450 * Checksum extended UDP header and data. 451 */ 452 if (uh->uh_sum == 0) 453 udp6stat.udp6s_nosum++; 454 else if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) { 455 udp6stat.udp6s_badsum++; 456 goto bad; 457 } 458 459 /* 460 * Construct source and dst sockaddrs. 461 * Note that ifindex (s6_addr16[1]) is already filled. 462 */ 463 bzero(&src, sizeof(src)); 464 src.sin6_family = AF_INET6; 465 src.sin6_len = sizeof(struct sockaddr_in6); 466 /* KAME hack: recover scopeid */ 467 (void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif); 468 src.sin6_port = uh->uh_sport; 469 bzero(&dst, sizeof(dst)); 470 dst.sin6_family = AF_INET6; 471 dst.sin6_len = sizeof(struct sockaddr_in6); 472 /* KAME hack: recover scopeid */ 473 (void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif); 474 dst.sin6_port = uh->uh_dport; 475 476 if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) { 477 if (m->m_flags & M_MCAST) { 478 udp6stat.udp6s_noportmcast++; 479 goto bad; 480 } 481 udp6stat.udp6s_noport++; 482 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); 483 m = NULL; 484 } 485 486 bad: 487 if (m) 488 m_freem(m); 489 return IPPROTO_DONE; 490 } 491 #endif 492 493 #ifdef INET 494 static void 495 udp4_sendup(m, off, src, so) 496 struct mbuf *m; 497 int off; /* offset of data portion */ 498 struct sockaddr *src; 499 struct socket *so; 500 { 501 struct mbuf *opts = NULL; 502 struct mbuf *n; 503 struct inpcb *inp = NULL; 504 #ifdef INET6 505 struct in6pcb *in6p = NULL; 506 #endif 507 508 if (!so) 509 return; 510 switch (so->so_proto->pr_domain->dom_family) { 511 case AF_INET: 512 inp = sotoinpcb(so); 513 break; 514 #ifdef INET6 515 case AF_INET6: 516 in6p = sotoin6pcb(so); 517 break; 518 #endif 519 default: 520 return; 521 } 522 523 #ifdef IPSEC 524 /* check AH/ESP integrity. */ 525 if (so != NULL && ipsec4_in_reject_so(m, so)) { 526 ipsecstat.in_polvio++; 527 return; 528 } 529 #endif /*IPSEC*/ 530 531 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { 532 if (inp && (inp->inp_flags & INP_CONTROLOPTS 533 || so->so_options & SO_TIMESTAMP)) { 534 struct ip *ip = mtod(n, struct ip *); 535 ip_savecontrol(inp, &opts, ip, n); 536 } 537 538 m_adj(n, off); 539 if (sbappendaddr(&so->so_rcv, src, n, 540 opts) == 0) { 541 m_freem(n); 542 if (opts) 543 m_freem(opts); 544 udpstat.udps_fullsock++; 545 } else 546 sorwakeup(so); 547 } 548 } 549 #endif 550 551 #ifdef INET6 552 static void 553 udp6_sendup(m, off, src, so) 554 struct mbuf *m; 555 int off; /* offset of data portion */ 556 struct sockaddr *src; 557 struct socket *so; 558 { 559 struct mbuf *opts = NULL; 560 struct mbuf *n; 561 struct in6pcb *in6p = NULL; 562 563 if (!so) 564 return; 565 if (so->so_proto->pr_domain->dom_family != AF_INET6) 566 return; 567 in6p = sotoin6pcb(so); 568 569 #ifdef IPSEC 570 /* check AH/ESP integrity. */ 571 if (so != NULL && ipsec6_in_reject_so(m, so)) { 572 ipsec6stat.in_polvio++; 573 return; 574 } 575 #endif /*IPSEC*/ 576 577 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { 578 if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS 579 || in6p->in6p_socket->so_options & SO_TIMESTAMP)) { 580 struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *); 581 ip6_savecontrol(in6p, &opts, ip6, n); 582 } 583 584 m_adj(n, off); 585 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) { 586 m_freem(n); 587 if (opts) 588 m_freem(opts); 589 udp6stat.udp6s_fullsock++; 590 } else 591 sorwakeup(so); 592 } 593 } 594 #endif 595 596 #ifdef INET 597 static int 598 udp4_realinput(src, dst, m, off) 599 struct sockaddr_in *src; 600 struct sockaddr_in *dst; 601 struct mbuf *m; 602 int off; /* offset of udphdr */ 603 { 604 u_int16_t *sport, *dport; 605 int rcvcnt; 606 struct in_addr *src4, *dst4; 607 struct inpcb *inp; 608 609 rcvcnt = 0; 610 off += sizeof(struct udphdr); /* now, offset of payload */ 611 612 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 613 goto bad; 614 615 src4 = &src->sin_addr; 616 sport = &src->sin_port; 617 dst4 = &dst->sin_addr; 618 dport = &dst->sin_port; 619 620 if (IN_MULTICAST(dst4->s_addr) || 621 in_broadcast(*dst4, m->m_pkthdr.rcvif)) { 622 struct inpcb *last; 623 /* 624 * Deliver a multicast or broadcast datagram to *all* sockets 625 * for which the local and remote addresses and ports match 626 * those of the incoming datagram. This allows more than 627 * one process to receive multi/broadcasts on the same port. 628 * (This really ought to be done for unicast datagrams as 629 * well, but that would cause problems with existing 630 * applications that open both address-specific sockets and 631 * a wildcard socket listening to the same port -- they would 632 * end up receiving duplicates of every unicast datagram. 633 * Those applications open the multiple sockets to overcome an 634 * inadequacy of the UDP socket interface, but for backwards 635 * compatibility we avoid the problem here rather than 636 * fixing the interface. Maybe 4.5BSD will remedy this?) 637 */ 638 639 /* 640 * KAME note: traditionally we dropped udpiphdr from mbuf here. 641 * we need udpiphdr for IPsec processing so we do that later. 642 */ 643 /* 644 * Locate pcb(s) for datagram. 645 */ 646 CIRCLEQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) { 647 if (inp->inp_lport != *dport) 648 continue; 649 if (!in_nullhost(inp->inp_laddr)) { 650 if (!in_hosteq(inp->inp_laddr, *dst4)) 651 continue; 652 } 653 if (!in_nullhost(inp->inp_faddr)) { 654 if (!in_hosteq(inp->inp_faddr, *src4) || 655 inp->inp_fport != *sport) 656 continue; 657 } 658 659 last = inp; 660 udp4_sendup(m, off, (struct sockaddr *)src, 661 inp->inp_socket); 662 rcvcnt++; 663 664 /* 665 * Don't look for additional matches if this one does 666 * not have either the SO_REUSEPORT or SO_REUSEADDR 667 * socket options set. This heuristic avoids searching 668 * through all pcbs in the common case of a non-shared 669 * port. It assumes that an application will never 670 * clear these options after setting them. 671 */ 672 if ((inp->inp_socket->so_options & 673 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 674 break; 675 } 676 } else { 677 /* 678 * Locate pcb for datagram. 679 */ 680 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport); 681 if (inp == 0) { 682 ++udpstat.udps_pcbhashmiss; 683 inp = in_pcblookup_bind(&udbtable, *dst4, *dport); 684 if (inp == 0) 685 return rcvcnt; 686 } 687 688 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 689 rcvcnt++; 690 } 691 692 bad: 693 return rcvcnt; 694 } 695 #endif 696 697 #ifdef INET6 698 static int 699 udp6_realinput(af, src, dst, m, off) 700 int af; /* af on packet */ 701 struct sockaddr_in6 *src; 702 struct sockaddr_in6 *dst; 703 struct mbuf *m; 704 int off; /* offset of udphdr */ 705 { 706 u_int16_t sport, dport; 707 int rcvcnt; 708 struct in6_addr src6, dst6; 709 const struct in_addr *dst4; 710 struct in6pcb *in6p; 711 712 rcvcnt = 0; 713 off += sizeof(struct udphdr); /* now, offset of payload */ 714 715 if (af != AF_INET && af != AF_INET6) 716 goto bad; 717 if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6) 718 goto bad; 719 720 in6_embedscope(&src6, src, NULL, NULL); 721 sport = src->sin6_port; 722 in6_embedscope(&dst6, dst, NULL, NULL); 723 dport = dst->sin6_port; 724 dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12]; 725 726 if (IN6_IS_ADDR_MULTICAST(&dst6) || 727 (af == AF_INET && IN_MULTICAST(dst4->s_addr))) { 728 struct in6pcb *last; 729 /* 730 * Deliver a multicast or broadcast datagram to *all* sockets 731 * for which the local and remote addresses and ports match 732 * those of the incoming datagram. This allows more than 733 * one process to receive multi/broadcasts on the same port. 734 * (This really ought to be done for unicast datagrams as 735 * well, but that would cause problems with existing 736 * applications that open both address-specific sockets and 737 * a wildcard socket listening to the same port -- they would 738 * end up receiving duplicates of every unicast datagram. 739 * Those applications open the multiple sockets to overcome an 740 * inadequacy of the UDP socket interface, but for backwards 741 * compatibility we avoid the problem here rather than 742 * fixing the interface. Maybe 4.5BSD will remedy this?) 743 */ 744 745 /* 746 * KAME note: traditionally we dropped udpiphdr from mbuf here. 747 * we need udpiphdr for IPsec processing so we do that later. 748 */ 749 /* 750 * Locate pcb(s) for datagram. 751 */ 752 for (in6p = udb6.in6p_next; in6p != &udb6; 753 in6p = in6p->in6p_next) { 754 if (in6p->in6p_lport != dport) 755 continue; 756 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 757 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6)) 758 continue; 759 } else { 760 if (IN6_IS_ADDR_V4MAPPED(&dst6) && 761 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 762 continue; 763 } 764 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 765 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, 766 &src6) || in6p->in6p_fport != sport) 767 continue; 768 } else { 769 if (IN6_IS_ADDR_V4MAPPED(&src6) && 770 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 771 continue; 772 } 773 774 last = in6p; 775 udp6_sendup(m, off, (struct sockaddr *)src, 776 in6p->in6p_socket); 777 rcvcnt++; 778 779 /* 780 * Don't look for additional matches if this one does 781 * not have either the SO_REUSEPORT or SO_REUSEADDR 782 * socket options set. This heuristic avoids searching 783 * through all pcbs in the common case of a non-shared 784 * port. It assumes that an application will never 785 * clear these options after setting them. 786 */ 787 if ((in6p->in6p_socket->so_options & 788 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 789 break; 790 } 791 } else { 792 /* 793 * Locate pcb for datagram. 794 */ 795 in6p = in6_pcblookup_connect(&udb6, &src6, sport, 796 &dst6, dport, 0); 797 if (in6p == 0) { 798 ++udpstat.udps_pcbhashmiss; 799 in6p = in6_pcblookup_bind(&udb6, &dst6, dport, 0); 800 if (in6p == 0) 801 return rcvcnt; 802 } 803 804 udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket); 805 rcvcnt++; 806 } 807 808 bad: 809 return rcvcnt; 810 } 811 #endif 812 813 #ifdef INET 814 /* 815 * Notify a udp user of an asynchronous error; 816 * just wake up so that he can collect error status. 817 */ 818 static void 819 udp_notify(inp, errno) 820 struct inpcb *inp; 821 int errno; 822 { 823 824 inp->inp_socket->so_error = errno; 825 sorwakeup(inp->inp_socket); 826 sowwakeup(inp->inp_socket); 827 } 828 829 void * 830 udp_ctlinput(cmd, sa, v) 831 int cmd; 832 struct sockaddr *sa; 833 void *v; 834 { 835 struct ip *ip = v; 836 struct udphdr *uh; 837 void (*notify) __P((struct inpcb *, int)) = udp_notify; 838 int errno; 839 840 if (sa->sa_family != AF_INET 841 || sa->sa_len != sizeof(struct sockaddr_in)) 842 return NULL; 843 if ((unsigned)cmd >= PRC_NCMDS) 844 return NULL; 845 errno = inetctlerrmap[cmd]; 846 if (PRC_IS_REDIRECT(cmd)) 847 notify = in_rtchange, ip = 0; 848 else if (cmd == PRC_HOSTDEAD) 849 ip = 0; 850 else if (errno == 0) 851 return NULL; 852 if (ip) { 853 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 854 in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport, 855 ip->ip_src, uh->uh_sport, errno, notify); 856 857 /* XXX mapped address case */ 858 } else 859 in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno, 860 notify); 861 return NULL; 862 } 863 864 int 865 #if __STDC__ 866 udp_output(struct mbuf *m, ...) 867 #else 868 udp_output(m, va_alist) 869 struct mbuf *m; 870 va_dcl 871 #endif 872 { 873 struct inpcb *inp; 874 struct udpiphdr *ui; 875 int len = m->m_pkthdr.len; 876 int error = 0; 877 va_list ap; 878 879 va_start(ap, m); 880 inp = va_arg(ap, struct inpcb *); 881 va_end(ap); 882 883 /* 884 * Calculate data length and get a mbuf 885 * for UDP and IP headers. 886 */ 887 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 888 if (m == 0) { 889 error = ENOBUFS; 890 goto release; 891 } 892 893 /* 894 * Compute the packet length of the IP header, and 895 * punt if the length looks bogus. 896 */ 897 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 898 error = EMSGSIZE; 899 goto release; 900 } 901 902 /* 903 * Fill in mbuf with extended UDP header 904 * and addresses and length put into network format. 905 */ 906 ui = mtod(m, struct udpiphdr *); 907 ui->ui_pr = IPPROTO_UDP; 908 ui->ui_src = inp->inp_laddr; 909 ui->ui_dst = inp->inp_faddr; 910 ui->ui_sport = inp->inp_lport; 911 ui->ui_dport = inp->inp_fport; 912 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 913 914 /* 915 * Set up checksum and output datagram. 916 */ 917 if (udpcksum) { 918 /* 919 * XXX Cache pseudo-header checksum part for 920 * XXX "connected" UDP sockets. 921 */ 922 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 923 ui->ui_dst.s_addr, htons((u_int16_t)len + 924 sizeof(struct udphdr) + IPPROTO_UDP)); 925 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 926 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 927 } else 928 ui->ui_sum = 0; 929 ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len); 930 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ 931 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ 932 udpstat.udps_opackets++; 933 934 #ifdef IPSEC 935 if (ipsec_setsocket(m, inp->inp_socket) != 0) { 936 error = ENOBUFS; 937 goto release; 938 } 939 #endif /*IPSEC*/ 940 941 return (ip_output(m, inp->inp_options, &inp->inp_route, 942 inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), 943 inp->inp_moptions)); 944 945 release: 946 m_freem(m); 947 return (error); 948 } 949 950 int udp_sendspace = 9216; /* really max datagram size */ 951 int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); 952 /* 40 1K datagrams */ 953 954 /*ARGSUSED*/ 955 int 956 udp_usrreq(so, req, m, nam, control, p) 957 struct socket *so; 958 int req; 959 struct mbuf *m, *nam, *control; 960 struct proc *p; 961 { 962 struct inpcb *inp; 963 int s; 964 int error = 0; 965 966 if (req == PRU_CONTROL) 967 return (in_control(so, (long)m, (caddr_t)nam, 968 (struct ifnet *)control, p)); 969 970 if (req == PRU_PURGEIF) { 971 in_pcbpurgeif0(&udbtable, (struct ifnet *)control); 972 in_purgeif((struct ifnet *)control); 973 in_pcbpurgeif(&udbtable, (struct ifnet *)control); 974 return (0); 975 } 976 977 s = splsoftnet(); 978 inp = sotoinpcb(so); 979 #ifdef DIAGNOSTIC 980 if (req != PRU_SEND && req != PRU_SENDOOB && control) 981 panic("udp_usrreq: unexpected control mbuf"); 982 #endif 983 if (inp == 0 && req != PRU_ATTACH) { 984 error = EINVAL; 985 goto release; 986 } 987 988 /* 989 * Note: need to block udp_input while changing 990 * the udp pcb queue and/or pcb addresses. 991 */ 992 switch (req) { 993 994 case PRU_ATTACH: 995 if (inp != 0) { 996 error = EISCONN; 997 break; 998 } 999 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1000 error = soreserve(so, udp_sendspace, udp_recvspace); 1001 if (error) 1002 break; 1003 } 1004 error = in_pcballoc(so, &udbtable); 1005 if (error) 1006 break; 1007 inp = sotoinpcb(so); 1008 inp->inp_ip.ip_ttl = ip_defttl; 1009 break; 1010 1011 case PRU_DETACH: 1012 in_pcbdetach(inp); 1013 break; 1014 1015 case PRU_BIND: 1016 error = in_pcbbind(inp, nam, p); 1017 break; 1018 1019 case PRU_LISTEN: 1020 error = EOPNOTSUPP; 1021 break; 1022 1023 case PRU_CONNECT: 1024 error = in_pcbconnect(inp, nam); 1025 if (error) 1026 break; 1027 soisconnected(so); 1028 break; 1029 1030 case PRU_CONNECT2: 1031 error = EOPNOTSUPP; 1032 break; 1033 1034 case PRU_DISCONNECT: 1035 /*soisdisconnected(so);*/ 1036 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 1037 in_pcbdisconnect(inp); 1038 inp->inp_laddr = zeroin_addr; /* XXX */ 1039 in_pcbstate(inp, INP_BOUND); /* XXX */ 1040 break; 1041 1042 case PRU_SHUTDOWN: 1043 socantsendmore(so); 1044 break; 1045 1046 case PRU_RCVD: 1047 error = EOPNOTSUPP; 1048 break; 1049 1050 case PRU_SEND: 1051 if (control && control->m_len) { 1052 m_freem(control); 1053 m_freem(m); 1054 error = EINVAL; 1055 break; 1056 } 1057 { 1058 struct in_addr laddr; /* XXX */ 1059 1060 if (nam) { 1061 laddr = inp->inp_laddr; /* XXX */ 1062 if ((so->so_state & SS_ISCONNECTED) != 0) { 1063 error = EISCONN; 1064 goto die; 1065 } 1066 error = in_pcbconnect(inp, nam); 1067 if (error) { 1068 die: 1069 m_freem(m); 1070 break; 1071 } 1072 } else { 1073 if ((so->so_state & SS_ISCONNECTED) == 0) { 1074 error = ENOTCONN; 1075 goto die; 1076 } 1077 } 1078 error = udp_output(m, inp); 1079 if (nam) { 1080 in_pcbdisconnect(inp); 1081 inp->inp_laddr = laddr; /* XXX */ 1082 in_pcbstate(inp, INP_BOUND); /* XXX */ 1083 } 1084 } 1085 break; 1086 1087 case PRU_SENSE: 1088 /* 1089 * stat: don't bother with a blocksize. 1090 */ 1091 splx(s); 1092 return (0); 1093 1094 case PRU_RCVOOB: 1095 error = EOPNOTSUPP; 1096 break; 1097 1098 case PRU_SENDOOB: 1099 m_freem(control); 1100 m_freem(m); 1101 error = EOPNOTSUPP; 1102 break; 1103 1104 case PRU_SOCKADDR: 1105 in_setsockaddr(inp, nam); 1106 break; 1107 1108 case PRU_PEERADDR: 1109 in_setpeeraddr(inp, nam); 1110 break; 1111 1112 default: 1113 panic("udp_usrreq"); 1114 } 1115 1116 release: 1117 splx(s); 1118 return (error); 1119 } 1120 1121 /* 1122 * Sysctl for udp variables. 1123 */ 1124 int 1125 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1126 int *name; 1127 u_int namelen; 1128 void *oldp; 1129 size_t *oldlenp; 1130 void *newp; 1131 size_t newlen; 1132 { 1133 /* All sysctl names at this level are terminal. */ 1134 if (namelen != 1) 1135 return (ENOTDIR); 1136 1137 switch (name[0]) { 1138 case UDPCTL_CHECKSUM: 1139 return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum)); 1140 case UDPCTL_SENDSPACE: 1141 return (sysctl_int(oldp, oldlenp, newp, newlen, 1142 &udp_sendspace)); 1143 case UDPCTL_RECVSPACE: 1144 return (sysctl_int(oldp, oldlenp, newp, newlen, 1145 &udp_recvspace)); 1146 default: 1147 return (ENOPROTOOPT); 1148 } 1149 /* NOTREACHED */ 1150 } 1151 #endif 1152