1 /* $NetBSD: udp_usrreq.c,v 1.178 2009/07/19 23:17:33 minskim 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 1.178 2009/07/19 23:17:33 minskim Exp $"); 65 66 #include "opt_inet.h" 67 #include "opt_compat_netbsd.h" 68 #include "opt_ipsec.h" 69 #include "opt_inet_csum.h" 70 #include "opt_ipkdb.h" 71 #include "opt_mbuftrace.h" 72 73 #include <sys/param.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/protosw.h> 77 #include <sys/socket.h> 78 #include <sys/socketvar.h> 79 #include <sys/errno.h> 80 #include <sys/stat.h> 81 #include <sys/systm.h> 82 #include <sys/proc.h> 83 #include <sys/domain.h> 84 #include <sys/sysctl.h> 85 86 #include <net/if.h> 87 #include <net/route.h> 88 89 #include <netinet/in.h> 90 #include <netinet/in_systm.h> 91 #include <netinet/in_var.h> 92 #include <netinet/ip.h> 93 #include <netinet/in_pcb.h> 94 #include <netinet/ip_var.h> 95 #include <netinet/ip_icmp.h> 96 #include <netinet/udp.h> 97 #include <netinet/udp_var.h> 98 #include <netinet/udp_private.h> 99 100 #ifdef INET6 101 #include <netinet/ip6.h> 102 #include <netinet/icmp6.h> 103 #include <netinet6/ip6_var.h> 104 #include <netinet6/ip6_private.h> 105 #include <netinet6/in6_pcb.h> 106 #include <netinet6/udp6_var.h> 107 #include <netinet6/udp6_private.h> 108 #include <netinet6/scope6_var.h> 109 #endif 110 111 #ifndef INET6 112 /* always need ip6.h for IP6_EXTHDR_GET */ 113 #include <netinet/ip6.h> 114 #endif 115 116 #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 FAST_IPSEC 124 #include <netipsec/ipsec.h> 125 #include <netipsec/ipsec_var.h> 126 #include <netipsec/ipsec_private.h> 127 #include <netipsec/esp.h> 128 #ifdef INET6 129 #include <netipsec/ipsec6.h> 130 #endif 131 #endif /* FAST_IPSEC */ 132 133 #ifdef IPSEC 134 #include <netinet6/ipsec.h> 135 #include <netinet6/ipsec_private.h> 136 #include <netinet6/esp.h> 137 #include <netkey/key.h> 138 #endif /* IPSEC */ 139 140 #ifdef COMPAT_50 141 #include <compat/sys/socket.h> 142 #endif 143 144 #ifdef IPKDB 145 #include <ipkdb/ipkdb.h> 146 #endif 147 148 /* 149 * UDP protocol implementation. 150 * Per RFC 768, August, 1980. 151 */ 152 int udpcksum = 1; 153 int udp_do_loopback_cksum = 0; 154 155 struct inpcbtable udbtable; 156 157 percpu_t *udpstat_percpu; 158 159 #ifdef INET 160 #ifdef IPSEC_NAT_T 161 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *, 162 struct socket *); 163 #endif 164 static void udp4_sendup (struct mbuf *, int, struct sockaddr *, 165 struct socket *); 166 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *, 167 struct mbuf **, int); 168 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int); 169 #endif 170 #ifdef INET6 171 static void udp6_sendup (struct mbuf *, int, struct sockaddr *, 172 struct socket *); 173 static int udp6_realinput (int, struct sockaddr_in6 *, 174 struct sockaddr_in6 *, struct mbuf *, int); 175 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int); 176 #endif 177 #ifdef INET 178 static void udp_notify (struct inpcb *, int); 179 #endif 180 181 #ifndef UDBHASHSIZE 182 #define UDBHASHSIZE 128 183 #endif 184 int udbhashsize = UDBHASHSIZE; 185 186 #ifdef MBUFTRACE 187 struct mowner udp_mowner = MOWNER_INIT("udp", ""); 188 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx"); 189 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx"); 190 #endif 191 192 #ifdef UDP_CSUM_COUNTERS 193 #include <sys/device.h> 194 195 #if defined(INET) 196 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 197 NULL, "udp", "hwcsum bad"); 198 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 199 NULL, "udp", "hwcsum ok"); 200 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 201 NULL, "udp", "hwcsum data"); 202 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 203 NULL, "udp", "swcsum"); 204 205 EVCNT_ATTACH_STATIC(udp_hwcsum_bad); 206 EVCNT_ATTACH_STATIC(udp_hwcsum_ok); 207 EVCNT_ATTACH_STATIC(udp_hwcsum_data); 208 EVCNT_ATTACH_STATIC(udp_swcsum); 209 #endif /* defined(INET) */ 210 211 #if defined(INET6) 212 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 213 NULL, "udp6", "hwcsum bad"); 214 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 215 NULL, "udp6", "hwcsum ok"); 216 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 217 NULL, "udp6", "hwcsum data"); 218 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 219 NULL, "udp6", "swcsum"); 220 221 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad); 222 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok); 223 EVCNT_ATTACH_STATIC(udp6_hwcsum_data); 224 EVCNT_ATTACH_STATIC(udp6_swcsum); 225 #endif /* defined(INET6) */ 226 227 #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 228 229 #else 230 231 #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */ 232 233 #endif /* UDP_CSUM_COUNTERS */ 234 235 void 236 udp_init(void) 237 { 238 239 in_pcbinit(&udbtable, udbhashsize, udbhashsize); 240 241 MOWNER_ATTACH(&udp_tx_mowner); 242 MOWNER_ATTACH(&udp_rx_mowner); 243 MOWNER_ATTACH(&udp_mowner); 244 245 #ifdef INET 246 udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS); 247 #endif 248 #ifdef INET6 249 udp6stat_percpu = percpu_alloc(sizeof(uint64_t) * UDP6_NSTATS); 250 #endif 251 } 252 253 /* 254 * Checksum extended UDP header and data. 255 */ 256 257 int 258 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh, 259 int iphlen, int len) 260 { 261 262 switch (af) { 263 #ifdef INET 264 case AF_INET: 265 return udp4_input_checksum(m, uh, iphlen, len); 266 #endif 267 #ifdef INET6 268 case AF_INET6: 269 return udp6_input_checksum(m, uh, iphlen, len); 270 #endif 271 } 272 #ifdef DIAGNOSTIC 273 panic("udp_input_checksum: unknown af %d", af); 274 #endif 275 /* NOTREACHED */ 276 return -1; 277 } 278 279 #ifdef INET 280 281 /* 282 * Checksum extended UDP header and data. 283 */ 284 285 static int 286 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh, 287 int iphlen, int len) 288 { 289 290 /* 291 * XXX it's better to record and check if this mbuf is 292 * already checked. 293 */ 294 295 if (uh->uh_sum == 0) 296 return 0; 297 298 switch (m->m_pkthdr.csum_flags & 299 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) | 300 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 301 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD: 302 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad); 303 goto badcsum; 304 305 case M_CSUM_UDPv4|M_CSUM_DATA: { 306 u_int32_t hw_csum = m->m_pkthdr.csum_data; 307 308 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); 309 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { 310 const struct ip *ip = 311 mtod(m, const struct ip *); 312 313 hw_csum = in_cksum_phdr(ip->ip_src.s_addr, 314 ip->ip_dst.s_addr, 315 htons(hw_csum + len + IPPROTO_UDP)); 316 } 317 if ((hw_csum ^ 0xffff) != 0) 318 goto badcsum; 319 break; 320 } 321 322 case M_CSUM_UDPv4: 323 /* Checksum was okay. */ 324 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); 325 break; 326 327 default: 328 /* 329 * Need to compute it ourselves. Maybe skip checksum 330 * on loopback interfaces. 331 */ 332 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 333 IFF_LOOPBACK) || 334 udp_do_loopback_cksum)) { 335 UDP_CSUM_COUNTER_INCR(&udp_swcsum); 336 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) 337 goto badcsum; 338 } 339 break; 340 } 341 342 return 0; 343 344 badcsum: 345 UDP_STATINC(UDP_STAT_BADSUM); 346 return -1; 347 } 348 349 void 350 udp_input(struct mbuf *m, ...) 351 { 352 va_list ap; 353 struct sockaddr_in src, dst; 354 struct ip *ip; 355 struct udphdr *uh; 356 int iphlen; 357 int len; 358 int n; 359 u_int16_t ip_len; 360 361 va_start(ap, m); 362 iphlen = va_arg(ap, int); 363 (void)va_arg(ap, int); /* ignore value, advance ap */ 364 va_end(ap); 365 366 MCLAIM(m, &udp_rx_mowner); 367 UDP_STATINC(UDP_STAT_IPACKETS); 368 369 /* 370 * Get IP and UDP header together in first mbuf. 371 */ 372 ip = mtod(m, struct ip *); 373 IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 374 if (uh == NULL) { 375 UDP_STATINC(UDP_STAT_HDROPS); 376 return; 377 } 378 KASSERT(UDP_HDR_ALIGNED_P(uh)); 379 380 /* destination port of 0 is illegal, based on RFC768. */ 381 if (uh->uh_dport == 0) 382 goto bad; 383 384 /* 385 * Make mbuf data length reflect UDP length. 386 * If not enough data to reflect UDP length, drop. 387 */ 388 ip_len = ntohs(ip->ip_len); 389 len = ntohs((u_int16_t)uh->uh_ulen); 390 if (ip_len != iphlen + len) { 391 if (ip_len < iphlen + len || len < sizeof(struct udphdr)) { 392 UDP_STATINC(UDP_STAT_BADLEN); 393 goto bad; 394 } 395 m_adj(m, iphlen + len - ip_len); 396 } 397 398 /* 399 * Checksum extended UDP header and data. 400 */ 401 if (udp4_input_checksum(m, uh, iphlen, len)) 402 goto badcsum; 403 404 /* construct source and dst sockaddrs. */ 405 sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport); 406 sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport); 407 408 if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) { 409 UDP_STATINC(UDP_STAT_HDROPS); 410 return; 411 } 412 #ifdef INET6 413 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { 414 struct sockaddr_in6 src6, dst6; 415 416 memset(&src6, 0, sizeof(src6)); 417 src6.sin6_family = AF_INET6; 418 src6.sin6_len = sizeof(struct sockaddr_in6); 419 src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff; 420 memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src, 421 sizeof(ip->ip_src)); 422 src6.sin6_port = uh->uh_sport; 423 memset(&dst6, 0, sizeof(dst6)); 424 dst6.sin6_family = AF_INET6; 425 dst6.sin6_len = sizeof(struct sockaddr_in6); 426 dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff; 427 memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst, 428 sizeof(ip->ip_dst)); 429 dst6.sin6_port = uh->uh_dport; 430 431 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen); 432 } 433 #endif 434 435 if (n == 0) { 436 if (m->m_flags & (M_BCAST | M_MCAST)) { 437 UDP_STATINC(UDP_STAT_NOPORTBCAST); 438 goto bad; 439 } 440 UDP_STATINC(UDP_STAT_NOPORT); 441 #ifdef IPKDB 442 if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport, 443 m, iphlen + sizeof(struct udphdr), 444 m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) { 445 /* 446 * It was a debugger connect packet, 447 * just drop it now 448 */ 449 goto bad; 450 } 451 #endif 452 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 453 m = NULL; 454 } 455 456 bad: 457 if (m) 458 m_freem(m); 459 return; 460 461 badcsum: 462 m_freem(m); 463 } 464 #endif 465 466 #ifdef INET6 467 static int 468 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len) 469 { 470 471 /* 472 * XXX it's better to record and check if this mbuf is 473 * already checked. 474 */ 475 476 if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) { 477 goto good; 478 } 479 if (uh->uh_sum == 0) { 480 UDP6_STATINC(UDP6_STAT_NOSUM); 481 goto bad; 482 } 483 484 switch (m->m_pkthdr.csum_flags & 485 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) | 486 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 487 case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD: 488 UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad); 489 UDP6_STATINC(UDP6_STAT_BADSUM); 490 goto bad; 491 492 #if 0 /* notyet */ 493 case M_CSUM_UDPv6|M_CSUM_DATA: 494 #endif 495 496 case M_CSUM_UDPv6: 497 /* Checksum was okay. */ 498 UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok); 499 break; 500 501 default: 502 /* 503 * Need to compute it ourselves. Maybe skip checksum 504 * on loopback interfaces. 505 */ 506 UDP_CSUM_COUNTER_INCR(&udp6_swcsum); 507 if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) { 508 UDP6_STATINC(UDP6_STAT_BADSUM); 509 goto bad; 510 } 511 } 512 513 good: 514 return 0; 515 bad: 516 return -1; 517 } 518 519 int 520 udp6_input(struct mbuf **mp, int *offp, int proto) 521 { 522 struct mbuf *m = *mp; 523 int off = *offp; 524 struct sockaddr_in6 src, dst; 525 struct ip6_hdr *ip6; 526 struct udphdr *uh; 527 u_int32_t plen, ulen; 528 529 ip6 = mtod(m, struct ip6_hdr *); 530 531 #if defined(NFAITH) && 0 < NFAITH 532 if (faithprefix(&ip6->ip6_dst)) { 533 /* send icmp6 host unreach? */ 534 m_freem(m); 535 return IPPROTO_DONE; 536 } 537 #endif 538 539 UDP6_STATINC(UDP6_STAT_IPACKETS); 540 541 /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */ 542 plen = m->m_pkthdr.len - off; 543 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr)); 544 if (uh == NULL) { 545 IP6_STATINC(IP6_STAT_TOOSHORT); 546 return IPPROTO_DONE; 547 } 548 KASSERT(UDP_HDR_ALIGNED_P(uh)); 549 ulen = ntohs((u_short)uh->uh_ulen); 550 /* 551 * RFC2675 section 4: jumbograms will have 0 in the UDP header field, 552 * iff payload length > 0xffff. 553 */ 554 if (ulen == 0 && plen > 0xffff) 555 ulen = plen; 556 557 if (plen != ulen) { 558 UDP6_STATINC(UDP6_STAT_BADLEN); 559 goto bad; 560 } 561 562 /* destination port of 0 is illegal, based on RFC768. */ 563 if (uh->uh_dport == 0) 564 goto bad; 565 566 /* Be proactive about malicious use of IPv4 mapped address */ 567 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 568 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 569 /* XXX stat */ 570 goto bad; 571 } 572 573 /* 574 * Checksum extended UDP header and data. Maybe skip checksum 575 * on loopback interfaces. 576 */ 577 if (udp6_input_checksum(m, uh, off, ulen)) 578 goto bad; 579 580 /* 581 * Construct source and dst sockaddrs. 582 */ 583 memset(&src, 0, sizeof(src)); 584 src.sin6_family = AF_INET6; 585 src.sin6_len = sizeof(struct sockaddr_in6); 586 src.sin6_addr = ip6->ip6_src; 587 src.sin6_port = uh->uh_sport; 588 memset(&dst, 0, sizeof(dst)); 589 dst.sin6_family = AF_INET6; 590 dst.sin6_len = sizeof(struct sockaddr_in6); 591 dst.sin6_addr = ip6->ip6_dst; 592 dst.sin6_port = uh->uh_dport; 593 594 if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) { 595 if (m->m_flags & M_MCAST) { 596 UDP6_STATINC(UDP6_STAT_NOPORTMCAST); 597 goto bad; 598 } 599 UDP6_STATINC(UDP6_STAT_NOPORT); 600 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); 601 m = NULL; 602 } 603 604 bad: 605 if (m) 606 m_freem(m); 607 return IPPROTO_DONE; 608 } 609 #endif 610 611 #ifdef INET 612 static void 613 udp4_sendup(struct mbuf *m, int off /* offset of data portion */, 614 struct sockaddr *src, struct socket *so) 615 { 616 struct mbuf *opts = NULL; 617 struct mbuf *n; 618 struct inpcb *inp = NULL; 619 620 if (!so) 621 return; 622 switch (so->so_proto->pr_domain->dom_family) { 623 case AF_INET: 624 inp = sotoinpcb(so); 625 break; 626 #ifdef INET6 627 case AF_INET6: 628 break; 629 #endif 630 default: 631 return; 632 } 633 634 #if defined(IPSEC) || defined(FAST_IPSEC) 635 /* check AH/ESP integrity. */ 636 if (so != NULL && ipsec4_in_reject_so(m, so)) { 637 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); 638 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) 639 icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT, 640 0, 0); 641 return; 642 } 643 #endif /*IPSEC*/ 644 645 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { 646 if (inp && (inp->inp_flags & INP_CONTROLOPTS 647 #ifdef SO_OTIMESTAMP 648 || so->so_options & SO_OTIMESTAMP 649 #endif 650 || so->so_options & SO_TIMESTAMP)) { 651 struct ip *ip = mtod(n, struct ip *); 652 ip_savecontrol(inp, &opts, ip, n); 653 } 654 655 m_adj(n, off); 656 if (sbappendaddr(&so->so_rcv, src, n, 657 opts) == 0) { 658 m_freem(n); 659 if (opts) 660 m_freem(opts); 661 so->so_rcv.sb_overflowed++; 662 UDP_STATINC(UDP_STAT_FULLSOCK); 663 } else 664 sorwakeup(so); 665 } 666 } 667 #endif 668 669 #ifdef INET6 670 static void 671 udp6_sendup(struct mbuf *m, int off /* offset of data portion */, 672 struct sockaddr *src, struct socket *so) 673 { 674 struct mbuf *opts = NULL; 675 struct mbuf *n; 676 struct in6pcb *in6p = NULL; 677 678 if (!so) 679 return; 680 if (so->so_proto->pr_domain->dom_family != AF_INET6) 681 return; 682 in6p = sotoin6pcb(so); 683 684 #if defined(IPSEC) || defined(FAST_IPSEC) 685 /* check AH/ESP integrity. */ 686 if (so != NULL && ipsec6_in_reject_so(m, so)) { 687 IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); 688 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) 689 icmp6_error(n, ICMP6_DST_UNREACH, 690 ICMP6_DST_UNREACH_ADMIN, 0); 691 return; 692 } 693 #endif /*IPSEC*/ 694 695 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { 696 if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS 697 #ifdef SO_OTIMESTAMP 698 || in6p->in6p_socket->so_options & SO_OTIMESTAMP 699 #endif 700 || in6p->in6p_socket->so_options & SO_TIMESTAMP)) { 701 struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *); 702 ip6_savecontrol(in6p, &opts, ip6, n); 703 } 704 705 m_adj(n, off); 706 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) { 707 m_freem(n); 708 if (opts) 709 m_freem(opts); 710 so->so_rcv.sb_overflowed++; 711 UDP6_STATINC(UDP6_STAT_FULLSOCK); 712 } else 713 sorwakeup(so); 714 } 715 } 716 #endif 717 718 #ifdef INET 719 static int 720 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst, 721 struct mbuf **mp, int off /* offset of udphdr */) 722 { 723 u_int16_t *sport, *dport; 724 int rcvcnt; 725 struct in_addr *src4, *dst4; 726 struct inpcb_hdr *inph; 727 struct inpcb *inp; 728 struct mbuf *m = *mp; 729 730 rcvcnt = 0; 731 off += sizeof(struct udphdr); /* now, offset of payload */ 732 733 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 734 goto bad; 735 736 src4 = &src->sin_addr; 737 sport = &src->sin_port; 738 dst4 = &dst->sin_addr; 739 dport = &dst->sin_port; 740 741 if (IN_MULTICAST(dst4->s_addr) || 742 in_broadcast(*dst4, m->m_pkthdr.rcvif)) { 743 /* 744 * Deliver a multicast or broadcast datagram to *all* sockets 745 * for which the local and remote addresses and ports match 746 * those of the incoming datagram. This allows more than 747 * one process to receive multi/broadcasts on the same port. 748 * (This really ought to be done for unicast datagrams as 749 * well, but that would cause problems with existing 750 * applications that open both address-specific sockets and 751 * a wildcard socket listening to the same port -- they would 752 * end up receiving duplicates of every unicast datagram. 753 * Those applications open the multiple sockets to overcome an 754 * inadequacy of the UDP socket interface, but for backwards 755 * compatibility we avoid the problem here rather than 756 * fixing the interface. Maybe 4.5BSD will remedy this?) 757 */ 758 759 /* 760 * KAME note: traditionally we dropped udpiphdr from mbuf here. 761 * we need udpiphdr for IPsec processing so we do that later. 762 */ 763 /* 764 * Locate pcb(s) for datagram. 765 */ 766 CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 767 inp = (struct inpcb *)inph; 768 if (inp->inp_af != AF_INET) 769 continue; 770 771 if (inp->inp_lport != *dport) 772 continue; 773 if (!in_nullhost(inp->inp_laddr)) { 774 if (!in_hosteq(inp->inp_laddr, *dst4)) 775 continue; 776 } 777 if (!in_nullhost(inp->inp_faddr)) { 778 if (!in_hosteq(inp->inp_faddr, *src4) || 779 inp->inp_fport != *sport) 780 continue; 781 } 782 783 udp4_sendup(m, off, (struct sockaddr *)src, 784 inp->inp_socket); 785 rcvcnt++; 786 787 /* 788 * Don't look for additional matches if this one does 789 * not have either the SO_REUSEPORT or SO_REUSEADDR 790 * socket options set. This heuristic avoids searching 791 * through all pcbs in the common case of a non-shared 792 * port. It assumes that an application will never 793 * clear these options after setting them. 794 */ 795 if ((inp->inp_socket->so_options & 796 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 797 break; 798 } 799 } else { 800 /* 801 * Locate pcb for datagram. 802 */ 803 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport); 804 if (inp == 0) { 805 UDP_STATINC(UDP_STAT_PCBHASHMISS); 806 inp = in_pcblookup_bind(&udbtable, *dst4, *dport); 807 if (inp == 0) 808 return rcvcnt; 809 } 810 811 #ifdef IPSEC_NAT_T 812 /* Handle ESP over UDP */ 813 if (inp->inp_flags & INP_ESPINUDP_ALL) { 814 struct sockaddr *sa = (struct sockaddr *)src; 815 816 switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) { 817 case -1: /* Error, m was freeed */ 818 rcvcnt = -1; 819 goto bad; 820 break; 821 822 case 1: /* ESP over UDP */ 823 rcvcnt++; 824 goto bad; 825 break; 826 827 case 0: /* plain UDP */ 828 default: /* Unexpected */ 829 /* 830 * Normal UDP processing will take place 831 * m may have changed. 832 */ 833 m = *mp; 834 break; 835 } 836 } 837 #endif 838 839 /* 840 * Check the minimum TTL for socket. 841 */ 842 if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl) 843 goto bad; 844 845 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 846 rcvcnt++; 847 } 848 849 bad: 850 return rcvcnt; 851 } 852 #endif 853 854 #ifdef INET6 855 static int 856 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst, 857 struct mbuf *m, int off) 858 { 859 u_int16_t sport, dport; 860 int rcvcnt; 861 struct in6_addr src6, *dst6; 862 const struct in_addr *dst4; 863 struct inpcb_hdr *inph; 864 struct in6pcb *in6p; 865 866 rcvcnt = 0; 867 off += sizeof(struct udphdr); /* now, offset of payload */ 868 869 if (af != AF_INET && af != AF_INET6) 870 goto bad; 871 if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6) 872 goto bad; 873 874 src6 = src->sin6_addr; 875 if (sa6_recoverscope(src) != 0) { 876 /* XXX: should be impossible. */ 877 goto bad; 878 } 879 sport = src->sin6_port; 880 881 dport = dst->sin6_port; 882 dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12]; 883 dst6 = &dst->sin6_addr; 884 885 if (IN6_IS_ADDR_MULTICAST(dst6) || 886 (af == AF_INET && IN_MULTICAST(dst4->s_addr))) { 887 /* 888 * Deliver a multicast or broadcast datagram to *all* sockets 889 * for which the local and remote addresses and ports match 890 * those of the incoming datagram. This allows more than 891 * one process to receive multi/broadcasts on the same port. 892 * (This really ought to be done for unicast datagrams as 893 * well, but that would cause problems with existing 894 * applications that open both address-specific sockets and 895 * a wildcard socket listening to the same port -- they would 896 * end up receiving duplicates of every unicast datagram. 897 * Those applications open the multiple sockets to overcome an 898 * inadequacy of the UDP socket interface, but for backwards 899 * compatibility we avoid the problem here rather than 900 * fixing the interface. Maybe 4.5BSD will remedy this?) 901 */ 902 903 /* 904 * KAME note: traditionally we dropped udpiphdr from mbuf here. 905 * we need udpiphdr for IPsec processing so we do that later. 906 */ 907 /* 908 * Locate pcb(s) for datagram. 909 */ 910 CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 911 in6p = (struct in6pcb *)inph; 912 if (in6p->in6p_af != AF_INET6) 913 continue; 914 915 if (in6p->in6p_lport != dport) 916 continue; 917 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 918 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, 919 dst6)) 920 continue; 921 } else { 922 if (IN6_IS_ADDR_V4MAPPED(dst6) && 923 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 924 continue; 925 } 926 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 927 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, 928 &src6) || in6p->in6p_fport != sport) 929 continue; 930 } else { 931 if (IN6_IS_ADDR_V4MAPPED(&src6) && 932 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 933 continue; 934 } 935 936 udp6_sendup(m, off, (struct sockaddr *)src, 937 in6p->in6p_socket); 938 rcvcnt++; 939 940 /* 941 * Don't look for additional matches if this one does 942 * not have either the SO_REUSEPORT or SO_REUSEADDR 943 * socket options set. This heuristic avoids searching 944 * through all pcbs in the common case of a non-shared 945 * port. It assumes that an application will never 946 * clear these options after setting them. 947 */ 948 if ((in6p->in6p_socket->so_options & 949 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 950 break; 951 } 952 } else { 953 /* 954 * Locate pcb for datagram. 955 */ 956 in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6, 957 dport, 0); 958 if (in6p == 0) { 959 UDP_STATINC(UDP_STAT_PCBHASHMISS); 960 in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0); 961 if (in6p == 0) 962 return rcvcnt; 963 } 964 965 udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket); 966 rcvcnt++; 967 } 968 969 bad: 970 return rcvcnt; 971 } 972 #endif 973 974 #ifdef INET 975 /* 976 * Notify a udp user of an asynchronous error; 977 * just wake up so that he can collect error status. 978 */ 979 static void 980 udp_notify(struct inpcb *inp, int errno) 981 { 982 inp->inp_socket->so_error = errno; 983 sorwakeup(inp->inp_socket); 984 sowwakeup(inp->inp_socket); 985 } 986 987 void * 988 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v) 989 { 990 struct ip *ip = v; 991 struct udphdr *uh; 992 void (*notify)(struct inpcb *, int) = udp_notify; 993 int errno; 994 995 if (sa->sa_family != AF_INET 996 || sa->sa_len != sizeof(struct sockaddr_in)) 997 return NULL; 998 if ((unsigned)cmd >= PRC_NCMDS) 999 return NULL; 1000 errno = inetctlerrmap[cmd]; 1001 if (PRC_IS_REDIRECT(cmd)) 1002 notify = in_rtchange, ip = 0; 1003 else if (cmd == PRC_HOSTDEAD) 1004 ip = 0; 1005 else if (errno == 0) 1006 return NULL; 1007 if (ip) { 1008 uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2)); 1009 in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport, 1010 ip->ip_src, uh->uh_sport, errno, notify); 1011 1012 /* XXX mapped address case */ 1013 } else 1014 in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno, 1015 notify); 1016 return NULL; 1017 } 1018 1019 int 1020 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1021 { 1022 int s; 1023 int error = 0; 1024 struct inpcb *inp; 1025 int family; 1026 int optval; 1027 1028 family = so->so_proto->pr_domain->dom_family; 1029 1030 s = splsoftnet(); 1031 switch (family) { 1032 #ifdef INET 1033 case PF_INET: 1034 if (sopt->sopt_level != IPPROTO_UDP) { 1035 error = ip_ctloutput(op, so, sopt); 1036 goto end; 1037 } 1038 break; 1039 #endif 1040 #ifdef INET6 1041 case PF_INET6: 1042 if (sopt->sopt_level != IPPROTO_UDP) { 1043 error = ip6_ctloutput(op, so, sopt); 1044 goto end; 1045 } 1046 break; 1047 #endif 1048 default: 1049 error = EAFNOSUPPORT; 1050 goto end; 1051 } 1052 1053 1054 switch (op) { 1055 case PRCO_SETOPT: 1056 inp = sotoinpcb(so); 1057 1058 switch (sopt->sopt_name) { 1059 case UDP_ENCAP: 1060 error = sockopt_getint(sopt, &optval); 1061 if (error) 1062 break; 1063 1064 switch(optval) { 1065 #ifdef IPSEC_NAT_T 1066 case 0: 1067 inp->inp_flags &= ~INP_ESPINUDP_ALL; 1068 break; 1069 1070 case UDP_ENCAP_ESPINUDP: 1071 inp->inp_flags &= ~INP_ESPINUDP_ALL; 1072 inp->inp_flags |= INP_ESPINUDP; 1073 break; 1074 1075 case UDP_ENCAP_ESPINUDP_NON_IKE: 1076 inp->inp_flags &= ~INP_ESPINUDP_ALL; 1077 inp->inp_flags |= INP_ESPINUDP_NON_IKE; 1078 break; 1079 #endif 1080 default: 1081 error = EINVAL; 1082 break; 1083 } 1084 break; 1085 1086 default: 1087 error = ENOPROTOOPT; 1088 break; 1089 } 1090 break; 1091 1092 default: 1093 error = EINVAL; 1094 break; 1095 } 1096 1097 end: 1098 splx(s); 1099 return error; 1100 } 1101 1102 1103 int 1104 udp_output(struct mbuf *m, ...) 1105 { 1106 struct inpcb *inp; 1107 struct udpiphdr *ui; 1108 struct route *ro; 1109 int len = m->m_pkthdr.len; 1110 int error = 0; 1111 va_list ap; 1112 1113 MCLAIM(m, &udp_tx_mowner); 1114 va_start(ap, m); 1115 inp = va_arg(ap, struct inpcb *); 1116 va_end(ap); 1117 1118 /* 1119 * Calculate data length and get a mbuf 1120 * for UDP and IP headers. 1121 */ 1122 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 1123 if (m == 0) { 1124 error = ENOBUFS; 1125 goto release; 1126 } 1127 1128 /* 1129 * Compute the packet length of the IP header, and 1130 * punt if the length looks bogus. 1131 */ 1132 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 1133 error = EMSGSIZE; 1134 goto release; 1135 } 1136 1137 /* 1138 * Fill in mbuf with extended UDP header 1139 * and addresses and length put into network format. 1140 */ 1141 ui = mtod(m, struct udpiphdr *); 1142 ui->ui_pr = IPPROTO_UDP; 1143 ui->ui_src = inp->inp_laddr; 1144 ui->ui_dst = inp->inp_faddr; 1145 ui->ui_sport = inp->inp_lport; 1146 ui->ui_dport = inp->inp_fport; 1147 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 1148 1149 ro = &inp->inp_route; 1150 1151 /* 1152 * Set up checksum and output datagram. 1153 */ 1154 if (udpcksum) { 1155 /* 1156 * XXX Cache pseudo-header checksum part for 1157 * XXX "connected" UDP sockets. 1158 */ 1159 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 1160 ui->ui_dst.s_addr, htons((u_int16_t)len + 1161 sizeof(struct udphdr) + IPPROTO_UDP)); 1162 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 1163 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 1164 } else 1165 ui->ui_sum = 0; 1166 ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len); 1167 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ 1168 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ 1169 UDP_STATINC(UDP_STAT_OPACKETS); 1170 1171 return (ip_output(m, inp->inp_options, ro, 1172 inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), 1173 inp->inp_moptions, inp->inp_socket)); 1174 1175 release: 1176 m_freem(m); 1177 return (error); 1178 } 1179 1180 int udp_sendspace = 9216; /* really max datagram size */ 1181 int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); 1182 /* 40 1K datagrams */ 1183 1184 /*ARGSUSED*/ 1185 int 1186 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, 1187 struct mbuf *control, struct lwp *l) 1188 { 1189 struct inpcb *inp; 1190 int s; 1191 int error = 0; 1192 1193 if (req == PRU_CONTROL) 1194 return (in_control(so, (long)m, (void *)nam, 1195 (struct ifnet *)control, l)); 1196 1197 s = splsoftnet(); 1198 1199 if (req == PRU_PURGEIF) { 1200 mutex_enter(softnet_lock); 1201 in_pcbpurgeif0(&udbtable, (struct ifnet *)control); 1202 in_purgeif((struct ifnet *)control); 1203 in_pcbpurgeif(&udbtable, (struct ifnet *)control); 1204 mutex_exit(softnet_lock); 1205 splx(s); 1206 return (0); 1207 } 1208 1209 inp = sotoinpcb(so); 1210 #ifdef DIAGNOSTIC 1211 if (req != PRU_SEND && req != PRU_SENDOOB && control) 1212 panic("udp_usrreq: unexpected control mbuf"); 1213 #endif 1214 if (req == PRU_ATTACH) { 1215 sosetlock(so); 1216 } else if (inp == 0) { 1217 error = EINVAL; 1218 goto release; 1219 } 1220 1221 /* 1222 * Note: need to block udp_input while changing 1223 * the udp pcb queue and/or pcb addresses. 1224 */ 1225 switch (req) { 1226 1227 case PRU_ATTACH: 1228 if (inp != 0) { 1229 error = EISCONN; 1230 break; 1231 } 1232 #ifdef MBUFTRACE 1233 so->so_mowner = &udp_mowner; 1234 so->so_rcv.sb_mowner = &udp_rx_mowner; 1235 so->so_snd.sb_mowner = &udp_tx_mowner; 1236 #endif 1237 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1238 error = soreserve(so, udp_sendspace, udp_recvspace); 1239 if (error) 1240 break; 1241 } 1242 error = in_pcballoc(so, &udbtable); 1243 if (error) 1244 break; 1245 inp = sotoinpcb(so); 1246 inp->inp_ip.ip_ttl = ip_defttl; 1247 break; 1248 1249 case PRU_DETACH: 1250 in_pcbdetach(inp); 1251 break; 1252 1253 case PRU_BIND: 1254 error = in_pcbbind(inp, nam, l); 1255 break; 1256 1257 case PRU_LISTEN: 1258 error = EOPNOTSUPP; 1259 break; 1260 1261 case PRU_CONNECT: 1262 error = in_pcbconnect(inp, nam, l); 1263 if (error) 1264 break; 1265 soisconnected(so); 1266 break; 1267 1268 case PRU_CONNECT2: 1269 error = EOPNOTSUPP; 1270 break; 1271 1272 case PRU_DISCONNECT: 1273 /*soisdisconnected(so);*/ 1274 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 1275 in_pcbdisconnect(inp); 1276 inp->inp_laddr = zeroin_addr; /* XXX */ 1277 in_pcbstate(inp, INP_BOUND); /* XXX */ 1278 break; 1279 1280 case PRU_SHUTDOWN: 1281 socantsendmore(so); 1282 break; 1283 1284 case PRU_RCVD: 1285 error = EOPNOTSUPP; 1286 break; 1287 1288 case PRU_SEND: 1289 if (control && control->m_len) { 1290 m_freem(control); 1291 m_freem(m); 1292 error = EINVAL; 1293 break; 1294 } 1295 { 1296 struct in_addr laddr; /* XXX */ 1297 1298 if (nam) { 1299 laddr = inp->inp_laddr; /* XXX */ 1300 if ((so->so_state & SS_ISCONNECTED) != 0) { 1301 error = EISCONN; 1302 goto die; 1303 } 1304 error = in_pcbconnect(inp, nam, l); 1305 if (error) 1306 goto die; 1307 } else { 1308 if ((so->so_state & SS_ISCONNECTED) == 0) { 1309 error = ENOTCONN; 1310 goto die; 1311 } 1312 } 1313 error = udp_output(m, inp); 1314 m = NULL; 1315 if (nam) { 1316 in_pcbdisconnect(inp); 1317 inp->inp_laddr = laddr; /* XXX */ 1318 in_pcbstate(inp, INP_BOUND); /* XXX */ 1319 } 1320 die: 1321 if (m) 1322 m_freem(m); 1323 } 1324 break; 1325 1326 case PRU_SENSE: 1327 /* 1328 * stat: don't bother with a blocksize. 1329 */ 1330 splx(s); 1331 return (0); 1332 1333 case PRU_RCVOOB: 1334 error = EOPNOTSUPP; 1335 break; 1336 1337 case PRU_SENDOOB: 1338 m_freem(control); 1339 m_freem(m); 1340 error = EOPNOTSUPP; 1341 break; 1342 1343 case PRU_SOCKADDR: 1344 in_setsockaddr(inp, nam); 1345 break; 1346 1347 case PRU_PEERADDR: 1348 in_setpeeraddr(inp, nam); 1349 break; 1350 1351 default: 1352 panic("udp_usrreq"); 1353 } 1354 1355 release: 1356 splx(s); 1357 return (error); 1358 } 1359 1360 static int 1361 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) 1362 { 1363 1364 return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS)); 1365 } 1366 1367 /* 1368 * Sysctl for udp variables. 1369 */ 1370 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup") 1371 { 1372 1373 sysctl_createv(clog, 0, NULL, NULL, 1374 CTLFLAG_PERMANENT, 1375 CTLTYPE_NODE, "net", NULL, 1376 NULL, 0, NULL, 0, 1377 CTL_NET, CTL_EOL); 1378 sysctl_createv(clog, 0, NULL, NULL, 1379 CTLFLAG_PERMANENT, 1380 CTLTYPE_NODE, "inet", NULL, 1381 NULL, 0, NULL, 0, 1382 CTL_NET, PF_INET, CTL_EOL); 1383 sysctl_createv(clog, 0, NULL, NULL, 1384 CTLFLAG_PERMANENT, 1385 CTLTYPE_NODE, "udp", 1386 SYSCTL_DESCR("UDPv4 related settings"), 1387 NULL, 0, NULL, 0, 1388 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL); 1389 1390 sysctl_createv(clog, 0, NULL, NULL, 1391 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1392 CTLTYPE_INT, "checksum", 1393 SYSCTL_DESCR("Compute UDP checksums"), 1394 NULL, 0, &udpcksum, 0, 1395 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM, 1396 CTL_EOL); 1397 sysctl_createv(clog, 0, NULL, NULL, 1398 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1399 CTLTYPE_INT, "sendspace", 1400 SYSCTL_DESCR("Default UDP send buffer size"), 1401 NULL, 0, &udp_sendspace, 0, 1402 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE, 1403 CTL_EOL); 1404 sysctl_createv(clog, 0, NULL, NULL, 1405 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1406 CTLTYPE_INT, "recvspace", 1407 SYSCTL_DESCR("Default UDP receive buffer size"), 1408 NULL, 0, &udp_recvspace, 0, 1409 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE, 1410 CTL_EOL); 1411 sysctl_createv(clog, 0, NULL, NULL, 1412 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1413 CTLTYPE_INT, "do_loopback_cksum", 1414 SYSCTL_DESCR("Perform UDP checksum on loopback"), 1415 NULL, 0, &udp_do_loopback_cksum, 0, 1416 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM, 1417 CTL_EOL); 1418 sysctl_createv(clog, 0, NULL, NULL, 1419 CTLFLAG_PERMANENT, 1420 CTLTYPE_STRUCT, "pcblist", 1421 SYSCTL_DESCR("UDP protocol control block list"), 1422 sysctl_inpcblist, 0, &udbtable, 0, 1423 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE, 1424 CTL_EOL); 1425 sysctl_createv(clog, 0, NULL, NULL, 1426 CTLFLAG_PERMANENT, 1427 CTLTYPE_STRUCT, "stats", 1428 SYSCTL_DESCR("UDP statistics"), 1429 sysctl_net_inet_udp_stats, 0, NULL, 0, 1430 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS, 1431 CTL_EOL); 1432 } 1433 #endif 1434 1435 void 1436 udp_statinc(u_int stat) 1437 { 1438 1439 KASSERT(stat < UDP_NSTATS); 1440 UDP_STATINC(stat); 1441 } 1442 1443 #if (defined INET && defined IPSEC_NAT_T) 1444 /* 1445 * Returns: 1446 * 1 if the packet was processed 1447 * 0 if normal UDP processing should take place 1448 * -1 if an error occurent and m was freed 1449 */ 1450 static int 1451 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src, 1452 struct socket *so) 1453 { 1454 size_t len; 1455 void *data; 1456 struct inpcb *inp; 1457 size_t skip = 0; 1458 size_t minlen; 1459 size_t iphdrlen; 1460 struct ip *ip; 1461 struct mbuf *n; 1462 struct m_tag *tag; 1463 struct udphdr *udphdr; 1464 u_int16_t sport, dport; 1465 struct mbuf *m = *mp; 1466 1467 /* 1468 * Collapse the mbuf chain if the first mbuf is too short 1469 * The longest case is: UDP + non ESP marker + ESP 1470 */ 1471 minlen = off + sizeof(u_int64_t) + sizeof(struct esp); 1472 if (minlen > m->m_pkthdr.len) 1473 minlen = m->m_pkthdr.len; 1474 1475 if (m->m_len < minlen) { 1476 if ((*mp = m_pullup(m, minlen)) == NULL) { 1477 printf("udp4_espinudp: m_pullup failed\n"); 1478 return -1; 1479 } 1480 m = *mp; 1481 } 1482 1483 len = m->m_len - off; 1484 data = mtod(m, char *) + off; 1485 inp = sotoinpcb(so); 1486 1487 /* Ignore keepalive packets */ 1488 if ((len == 1) && (*(unsigned char *)data == 0xff)) { 1489 return 1; 1490 } 1491 1492 /* 1493 * Check that the payload is long enough to hold 1494 * an ESP header and compute the length of encapsulation 1495 * header to remove 1496 */ 1497 if (inp->inp_flags & INP_ESPINUDP) { 1498 u_int32_t *st = (u_int32_t *)data; 1499 1500 if ((len <= sizeof(struct esp)) || (*st == 0)) 1501 return 0; /* Normal UDP processing */ 1502 1503 skip = sizeof(struct udphdr); 1504 } 1505 1506 if (inp->inp_flags & INP_ESPINUDP_NON_IKE) { 1507 u_int32_t *st = (u_int32_t *)data; 1508 1509 if ((len <= sizeof(u_int64_t) + sizeof(struct esp)) 1510 || ((st[0] | st[1]) != 0)) 1511 return 0; /* Normal UDP processing */ 1512 1513 skip = sizeof(struct udphdr) + sizeof(u_int64_t); 1514 } 1515 1516 /* 1517 * Get the UDP ports. They are handled in network 1518 * order everywhere in IPSEC_NAT_T code. 1519 */ 1520 udphdr = (struct udphdr *)((char *)data - skip); 1521 sport = udphdr->uh_sport; 1522 dport = udphdr->uh_dport; 1523 1524 /* 1525 * Remove the UDP header (and possibly the non ESP marker) 1526 * IP header lendth is iphdrlen 1527 * Before: 1528 * <--- off ---> 1529 * +----+------+-----+ 1530 * | IP | UDP | ESP | 1531 * +----+------+-----+ 1532 * <-skip-> 1533 * After: 1534 * +----+-----+ 1535 * | IP | ESP | 1536 * +----+-----+ 1537 * <-skip-> 1538 */ 1539 iphdrlen = off - sizeof(struct udphdr); 1540 memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen); 1541 m_adj(m, skip); 1542 1543 ip = mtod(m, struct ip *); 1544 ip->ip_len = htons(ntohs(ip->ip_len) - skip); 1545 ip->ip_p = IPPROTO_ESP; 1546 1547 /* 1548 * Copy the mbuf to avoid multiple free, as both 1549 * esp4_input (which we call) and udp_input (which 1550 * called us) free the mbuf. 1551 */ 1552 if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 1553 printf("udp4_espinudp: m_dup failed\n"); 1554 return 0; 1555 } 1556 1557 /* 1558 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember 1559 * the source UDP port. This is required if we want 1560 * to select the right SPD for multiple hosts behind 1561 * same NAT 1562 */ 1563 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, 1564 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) { 1565 printf("udp4_espinudp: m_tag_get failed\n"); 1566 m_freem(n); 1567 return 0; 1568 } 1569 ((u_int16_t *)(tag + 1))[0] = sport; 1570 ((u_int16_t *)(tag + 1))[1] = dport; 1571 m_tag_prepend(n, tag); 1572 1573 #ifdef FAST_IPSEC 1574 ipsec4_common_input(n, iphdrlen, IPPROTO_ESP); 1575 #else 1576 esp4_input(n, iphdrlen); 1577 #endif 1578 1579 /* We handled it, it shoudln't be handled by UDP */ 1580 return 1; 1581 } 1582 #endif 1583