1 /* $NetBSD: udp_usrreq.c,v 1.185 2012/01/09 22:26:44 liamjfoy 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 1.185 2012/01/09 22:26:44 liamjfoy 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 #include <netinet/rfc6056.h> 100 101 #ifdef INET6 102 #include <netinet/ip6.h> 103 #include <netinet/icmp6.h> 104 #include <netinet6/ip6_var.h> 105 #include <netinet6/ip6_private.h> 106 #include <netinet6/in6_pcb.h> 107 #include <netinet6/udp6_var.h> 108 #include <netinet6/udp6_private.h> 109 #include <netinet6/scope6_var.h> 110 #endif 111 112 #ifndef INET6 113 /* always need ip6.h for IP6_EXTHDR_GET */ 114 #include <netinet/ip6.h> 115 #endif 116 117 #include "faith.h" 118 #if defined(NFAITH) && NFAITH > 0 119 #include <net/if_faith.h> 120 #endif 121 122 #ifdef FAST_IPSEC 123 #include <netipsec/ipsec.h> 124 #include <netipsec/ipsec_var.h> 125 #include <netipsec/ipsec_private.h> 126 #include <netipsec/esp.h> 127 #ifdef INET6 128 #include <netipsec/ipsec6.h> 129 #endif 130 #endif /* FAST_IPSEC */ 131 132 #ifdef KAME_IPSEC 133 #include <netinet6/ipsec.h> 134 #include <netinet6/ipsec_private.h> 135 #include <netinet6/esp.h> 136 #include <netkey/key.h> 137 #endif /* KAME_IPSEC */ 138 139 #ifdef COMPAT_50 140 #include <compat/sys/socket.h> 141 #endif 142 143 #ifdef IPKDB 144 #include <ipkdb/ipkdb.h> 145 #endif 146 147 /* 148 * UDP protocol implementation. 149 * Per RFC 768, August, 1980. 150 */ 151 int udpcksum = 1; 152 int udp_do_loopback_cksum = 0; 153 154 struct inpcbtable udbtable; 155 156 percpu_t *udpstat_percpu; 157 158 #ifdef INET 159 #ifdef IPSEC_NAT_T 160 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *, 161 struct socket *); 162 #endif 163 static void udp4_sendup (struct mbuf *, int, struct sockaddr *, 164 struct socket *); 165 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *, 166 struct mbuf **, int); 167 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int); 168 #endif 169 #ifdef INET6 170 static void udp6_sendup (struct mbuf *, int, struct sockaddr *, 171 struct socket *); 172 static int udp6_realinput (int, struct sockaddr_in6 *, 173 struct sockaddr_in6 *, struct mbuf *, int); 174 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int); 175 #endif 176 #ifdef INET 177 static void udp_notify (struct inpcb *, int); 178 #endif 179 180 #ifndef UDBHASHSIZE 181 #define UDBHASHSIZE 128 182 #endif 183 int udbhashsize = UDBHASHSIZE; 184 185 #ifdef MBUFTRACE 186 struct mowner udp_mowner = MOWNER_INIT("udp", ""); 187 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx"); 188 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx"); 189 #endif 190 191 #ifdef UDP_CSUM_COUNTERS 192 #include <sys/device.h> 193 194 #if defined(INET) 195 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 196 NULL, "udp", "hwcsum bad"); 197 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 198 NULL, "udp", "hwcsum ok"); 199 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 200 NULL, "udp", "hwcsum data"); 201 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 202 NULL, "udp", "swcsum"); 203 204 EVCNT_ATTACH_STATIC(udp_hwcsum_bad); 205 EVCNT_ATTACH_STATIC(udp_hwcsum_ok); 206 EVCNT_ATTACH_STATIC(udp_hwcsum_data); 207 EVCNT_ATTACH_STATIC(udp_swcsum); 208 #endif /* defined(INET) */ 209 210 #if defined(INET6) 211 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 212 NULL, "udp6", "hwcsum bad"); 213 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 214 NULL, "udp6", "hwcsum ok"); 215 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 216 NULL, "udp6", "hwcsum data"); 217 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 218 NULL, "udp6", "swcsum"); 219 220 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad); 221 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok); 222 EVCNT_ATTACH_STATIC(udp6_hwcsum_data); 223 EVCNT_ATTACH_STATIC(udp6_swcsum); 224 #endif /* defined(INET6) */ 225 226 #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 227 228 #else 229 230 #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */ 231 232 #endif /* UDP_CSUM_COUNTERS */ 233 234 static void sysctl_net_inet_udp_setup(struct sysctllog **); 235 236 void 237 udp_init(void) 238 { 239 240 sysctl_net_inet_udp_setup(NULL); 241 242 in_pcbinit(&udbtable, udbhashsize, udbhashsize); 243 244 MOWNER_ATTACH(&udp_tx_mowner); 245 MOWNER_ATTACH(&udp_rx_mowner); 246 MOWNER_ATTACH(&udp_mowner); 247 248 #ifdef INET 249 udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS); 250 #endif 251 #ifdef INET6 252 udp6stat_percpu = percpu_alloc(sizeof(uint64_t) * UDP6_NSTATS); 253 #endif 254 } 255 256 /* 257 * Checksum extended UDP header and data. 258 */ 259 260 int 261 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh, 262 int iphlen, int len) 263 { 264 265 switch (af) { 266 #ifdef INET 267 case AF_INET: 268 return udp4_input_checksum(m, uh, iphlen, len); 269 #endif 270 #ifdef INET6 271 case AF_INET6: 272 return udp6_input_checksum(m, uh, iphlen, len); 273 #endif 274 } 275 #ifdef DIAGNOSTIC 276 panic("udp_input_checksum: unknown af %d", af); 277 #endif 278 /* NOTREACHED */ 279 return -1; 280 } 281 282 #ifdef INET 283 284 /* 285 * Checksum extended UDP header and data. 286 */ 287 288 static int 289 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh, 290 int iphlen, int len) 291 { 292 293 /* 294 * XXX it's better to record and check if this mbuf is 295 * already checked. 296 */ 297 298 if (uh->uh_sum == 0) 299 return 0; 300 301 switch (m->m_pkthdr.csum_flags & 302 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) | 303 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 304 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD: 305 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad); 306 goto badcsum; 307 308 case M_CSUM_UDPv4|M_CSUM_DATA: { 309 u_int32_t hw_csum = m->m_pkthdr.csum_data; 310 311 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); 312 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { 313 const struct ip *ip = 314 mtod(m, const struct ip *); 315 316 hw_csum = in_cksum_phdr(ip->ip_src.s_addr, 317 ip->ip_dst.s_addr, 318 htons(hw_csum + len + IPPROTO_UDP)); 319 } 320 if ((hw_csum ^ 0xffff) != 0) 321 goto badcsum; 322 break; 323 } 324 325 case M_CSUM_UDPv4: 326 /* Checksum was okay. */ 327 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); 328 break; 329 330 default: 331 /* 332 * Need to compute it ourselves. Maybe skip checksum 333 * on loopback interfaces. 334 */ 335 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 336 IFF_LOOPBACK) || 337 udp_do_loopback_cksum)) { 338 UDP_CSUM_COUNTER_INCR(&udp_swcsum); 339 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) 340 goto badcsum; 341 } 342 break; 343 } 344 345 return 0; 346 347 badcsum: 348 UDP_STATINC(UDP_STAT_BADSUM); 349 return -1; 350 } 351 352 void 353 udp_input(struct mbuf *m, ...) 354 { 355 va_list ap; 356 struct sockaddr_in src, dst; 357 struct ip *ip; 358 struct udphdr *uh; 359 int iphlen; 360 int len; 361 int n; 362 u_int16_t ip_len; 363 364 va_start(ap, m); 365 iphlen = va_arg(ap, int); 366 (void)va_arg(ap, int); /* ignore value, advance ap */ 367 va_end(ap); 368 369 MCLAIM(m, &udp_rx_mowner); 370 UDP_STATINC(UDP_STAT_IPACKETS); 371 372 /* 373 * Get IP and UDP header together in first mbuf. 374 */ 375 ip = mtod(m, struct ip *); 376 IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 377 if (uh == NULL) { 378 UDP_STATINC(UDP_STAT_HDROPS); 379 return; 380 } 381 KASSERT(UDP_HDR_ALIGNED_P(uh)); 382 383 /* destination port of 0 is illegal, based on RFC768. */ 384 if (uh->uh_dport == 0) 385 goto bad; 386 387 /* 388 * Make mbuf data length reflect UDP length. 389 * If not enough data to reflect UDP length, drop. 390 */ 391 ip_len = ntohs(ip->ip_len); 392 len = ntohs((u_int16_t)uh->uh_ulen); 393 if (ip_len != iphlen + len) { 394 if (ip_len < iphlen + len || len < sizeof(struct udphdr)) { 395 UDP_STATINC(UDP_STAT_BADLEN); 396 goto bad; 397 } 398 m_adj(m, iphlen + len - ip_len); 399 } 400 401 /* 402 * Checksum extended UDP header and data. 403 */ 404 if (udp4_input_checksum(m, uh, iphlen, len)) 405 goto badcsum; 406 407 /* construct source and dst sockaddrs. */ 408 sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport); 409 sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport); 410 411 if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) { 412 UDP_STATINC(UDP_STAT_HDROPS); 413 return; 414 } 415 #ifdef INET6 416 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { 417 struct sockaddr_in6 src6, dst6; 418 419 memset(&src6, 0, sizeof(src6)); 420 src6.sin6_family = AF_INET6; 421 src6.sin6_len = sizeof(struct sockaddr_in6); 422 src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff; 423 memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src, 424 sizeof(ip->ip_src)); 425 src6.sin6_port = uh->uh_sport; 426 memset(&dst6, 0, sizeof(dst6)); 427 dst6.sin6_family = AF_INET6; 428 dst6.sin6_len = sizeof(struct sockaddr_in6); 429 dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff; 430 memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst, 431 sizeof(ip->ip_dst)); 432 dst6.sin6_port = uh->uh_dport; 433 434 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen); 435 } 436 #endif 437 438 if (n == 0) { 439 if (m->m_flags & (M_BCAST | M_MCAST)) { 440 UDP_STATINC(UDP_STAT_NOPORTBCAST); 441 goto bad; 442 } 443 UDP_STATINC(UDP_STAT_NOPORT); 444 #ifdef IPKDB 445 if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport, 446 m, iphlen + sizeof(struct udphdr), 447 m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) { 448 /* 449 * It was a debugger connect packet, 450 * just drop it now 451 */ 452 goto bad; 453 } 454 #endif 455 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 456 m = NULL; 457 } 458 459 bad: 460 if (m) 461 m_freem(m); 462 return; 463 464 badcsum: 465 m_freem(m); 466 } 467 #endif 468 469 #ifdef INET6 470 static int 471 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len) 472 { 473 474 /* 475 * XXX it's better to record and check if this mbuf is 476 * already checked. 477 */ 478 479 if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) { 480 goto good; 481 } 482 if (uh->uh_sum == 0) { 483 UDP6_STATINC(UDP6_STAT_NOSUM); 484 goto bad; 485 } 486 487 switch (m->m_pkthdr.csum_flags & 488 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) | 489 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 490 case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD: 491 UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad); 492 UDP6_STATINC(UDP6_STAT_BADSUM); 493 goto bad; 494 495 #if 0 /* notyet */ 496 case M_CSUM_UDPv6|M_CSUM_DATA: 497 #endif 498 499 case M_CSUM_UDPv6: 500 /* Checksum was okay. */ 501 UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok); 502 break; 503 504 default: 505 /* 506 * Need to compute it ourselves. Maybe skip checksum 507 * on loopback interfaces. 508 */ 509 UDP_CSUM_COUNTER_INCR(&udp6_swcsum); 510 if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) { 511 UDP6_STATINC(UDP6_STAT_BADSUM); 512 goto bad; 513 } 514 } 515 516 good: 517 return 0; 518 bad: 519 return -1; 520 } 521 522 int 523 udp6_input(struct mbuf **mp, int *offp, int proto) 524 { 525 struct mbuf *m = *mp; 526 int off = *offp; 527 struct sockaddr_in6 src, dst; 528 struct ip6_hdr *ip6; 529 struct udphdr *uh; 530 u_int32_t plen, ulen; 531 532 ip6 = mtod(m, struct ip6_hdr *); 533 534 #if defined(NFAITH) && 0 < NFAITH 535 if (faithprefix(&ip6->ip6_dst)) { 536 /* send icmp6 host unreach? */ 537 m_freem(m); 538 return IPPROTO_DONE; 539 } 540 #endif 541 542 UDP6_STATINC(UDP6_STAT_IPACKETS); 543 544 /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */ 545 plen = m->m_pkthdr.len - off; 546 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr)); 547 if (uh == NULL) { 548 IP6_STATINC(IP6_STAT_TOOSHORT); 549 return IPPROTO_DONE; 550 } 551 KASSERT(UDP_HDR_ALIGNED_P(uh)); 552 ulen = ntohs((u_short)uh->uh_ulen); 553 /* 554 * RFC2675 section 4: jumbograms will have 0 in the UDP header field, 555 * iff payload length > 0xffff. 556 */ 557 if (ulen == 0 && plen > 0xffff) 558 ulen = plen; 559 560 if (plen != ulen) { 561 UDP6_STATINC(UDP6_STAT_BADLEN); 562 goto bad; 563 } 564 565 /* destination port of 0 is illegal, based on RFC768. */ 566 if (uh->uh_dport == 0) 567 goto bad; 568 569 /* Be proactive about malicious use of IPv4 mapped address */ 570 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 571 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 572 /* XXX stat */ 573 goto bad; 574 } 575 576 /* 577 * Checksum extended UDP header and data. Maybe skip checksum 578 * on loopback interfaces. 579 */ 580 if (udp6_input_checksum(m, uh, off, ulen)) 581 goto bad; 582 583 /* 584 * Construct source and dst sockaddrs. 585 */ 586 memset(&src, 0, sizeof(src)); 587 src.sin6_family = AF_INET6; 588 src.sin6_len = sizeof(struct sockaddr_in6); 589 src.sin6_addr = ip6->ip6_src; 590 src.sin6_port = uh->uh_sport; 591 memset(&dst, 0, sizeof(dst)); 592 dst.sin6_family = AF_INET6; 593 dst.sin6_len = sizeof(struct sockaddr_in6); 594 dst.sin6_addr = ip6->ip6_dst; 595 dst.sin6_port = uh->uh_dport; 596 597 if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) { 598 if (m->m_flags & M_MCAST) { 599 UDP6_STATINC(UDP6_STAT_NOPORTMCAST); 600 goto bad; 601 } 602 UDP6_STATINC(UDP6_STAT_NOPORT); 603 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); 604 m = NULL; 605 } 606 607 bad: 608 if (m) 609 m_freem(m); 610 return IPPROTO_DONE; 611 } 612 #endif 613 614 #ifdef INET 615 static void 616 udp4_sendup(struct mbuf *m, int off /* offset of data portion */, 617 struct sockaddr *src, struct socket *so) 618 { 619 struct mbuf *opts = NULL; 620 struct mbuf *n; 621 struct inpcb *inp = NULL; 622 623 if (!so) 624 return; 625 switch (so->so_proto->pr_domain->dom_family) { 626 case AF_INET: 627 inp = sotoinpcb(so); 628 break; 629 #ifdef INET6 630 case AF_INET6: 631 break; 632 #endif 633 default: 634 return; 635 } 636 637 #if defined(KAME_IPSEC) || defined(FAST_IPSEC) 638 /* check AH/ESP integrity. */ 639 if (so != NULL && ipsec4_in_reject_so(m, so)) { 640 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); 641 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) 642 icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT, 643 0, 0); 644 return; 645 } 646 #endif /*IPSEC*/ 647 648 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { 649 if (inp && (inp->inp_flags & INP_CONTROLOPTS 650 #ifdef SO_OTIMESTAMP 651 || so->so_options & SO_OTIMESTAMP 652 #endif 653 || so->so_options & SO_TIMESTAMP)) { 654 struct ip *ip = mtod(n, struct ip *); 655 ip_savecontrol(inp, &opts, ip, n); 656 } 657 658 m_adj(n, off); 659 if (sbappendaddr(&so->so_rcv, src, n, 660 opts) == 0) { 661 m_freem(n); 662 if (opts) 663 m_freem(opts); 664 so->so_rcv.sb_overflowed++; 665 UDP_STATINC(UDP_STAT_FULLSOCK); 666 } else 667 sorwakeup(so); 668 } 669 } 670 #endif 671 672 #ifdef INET6 673 static void 674 udp6_sendup(struct mbuf *m, int off /* offset of data portion */, 675 struct sockaddr *src, struct socket *so) 676 { 677 struct mbuf *opts = NULL; 678 struct mbuf *n; 679 struct in6pcb *in6p = NULL; 680 681 if (!so) 682 return; 683 if (so->so_proto->pr_domain->dom_family != AF_INET6) 684 return; 685 in6p = sotoin6pcb(so); 686 687 #if defined(KAME_IPSEC) || defined(FAST_IPSEC) 688 /* check AH/ESP integrity. */ 689 if (so != NULL && ipsec6_in_reject_so(m, so)) { 690 IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); 691 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) 692 icmp6_error(n, ICMP6_DST_UNREACH, 693 ICMP6_DST_UNREACH_ADMIN, 0); 694 return; 695 } 696 #endif /*IPSEC*/ 697 698 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { 699 if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS 700 #ifdef SO_OTIMESTAMP 701 || in6p->in6p_socket->so_options & SO_OTIMESTAMP 702 #endif 703 || in6p->in6p_socket->so_options & SO_TIMESTAMP)) { 704 struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *); 705 ip6_savecontrol(in6p, &opts, ip6, n); 706 } 707 708 m_adj(n, off); 709 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) { 710 m_freem(n); 711 if (opts) 712 m_freem(opts); 713 so->so_rcv.sb_overflowed++; 714 UDP6_STATINC(UDP6_STAT_FULLSOCK); 715 } else 716 sorwakeup(so); 717 } 718 } 719 #endif 720 721 #ifdef INET 722 static int 723 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst, 724 struct mbuf **mp, int off /* offset of udphdr */) 725 { 726 u_int16_t *sport, *dport; 727 int rcvcnt; 728 struct in_addr *src4, *dst4; 729 struct inpcb_hdr *inph; 730 struct inpcb *inp; 731 struct mbuf *m = *mp; 732 733 rcvcnt = 0; 734 off += sizeof(struct udphdr); /* now, offset of payload */ 735 736 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 737 goto bad; 738 739 src4 = &src->sin_addr; 740 sport = &src->sin_port; 741 dst4 = &dst->sin_addr; 742 dport = &dst->sin_port; 743 744 if (IN_MULTICAST(dst4->s_addr) || 745 in_broadcast(*dst4, m->m_pkthdr.rcvif)) { 746 /* 747 * Deliver a multicast or broadcast datagram to *all* sockets 748 * for which the local and remote addresses and ports match 749 * those of the incoming datagram. This allows more than 750 * one process to receive multi/broadcasts on the same port. 751 * (This really ought to be done for unicast datagrams as 752 * well, but that would cause problems with existing 753 * applications that open both address-specific sockets and 754 * a wildcard socket listening to the same port -- they would 755 * end up receiving duplicates of every unicast datagram. 756 * Those applications open the multiple sockets to overcome an 757 * inadequacy of the UDP socket interface, but for backwards 758 * compatibility we avoid the problem here rather than 759 * fixing the interface. Maybe 4.5BSD will remedy this?) 760 */ 761 762 /* 763 * KAME note: traditionally we dropped udpiphdr from mbuf here. 764 * we need udpiphdr for IPsec processing so we do that later. 765 */ 766 /* 767 * Locate pcb(s) for datagram. 768 */ 769 CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 770 inp = (struct inpcb *)inph; 771 if (inp->inp_af != AF_INET) 772 continue; 773 774 if (inp->inp_lport != *dport) 775 continue; 776 if (!in_nullhost(inp->inp_laddr)) { 777 if (!in_hosteq(inp->inp_laddr, *dst4)) 778 continue; 779 } 780 if (!in_nullhost(inp->inp_faddr)) { 781 if (!in_hosteq(inp->inp_faddr, *src4) || 782 inp->inp_fport != *sport) 783 continue; 784 } 785 786 udp4_sendup(m, off, (struct sockaddr *)src, 787 inp->inp_socket); 788 rcvcnt++; 789 790 /* 791 * Don't look for additional matches if this one does 792 * not have either the SO_REUSEPORT or SO_REUSEADDR 793 * socket options set. This heuristic avoids searching 794 * through all pcbs in the common case of a non-shared 795 * port. It assumes that an application will never 796 * clear these options after setting them. 797 */ 798 if ((inp->inp_socket->so_options & 799 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 800 break; 801 } 802 } else { 803 /* 804 * Locate pcb for datagram. 805 */ 806 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, 807 *dport, 0); 808 if (inp == 0) { 809 UDP_STATINC(UDP_STAT_PCBHASHMISS); 810 inp = in_pcblookup_bind(&udbtable, *dst4, *dport); 811 if (inp == 0) 812 return rcvcnt; 813 } 814 815 #ifdef IPSEC_NAT_T 816 /* Handle ESP over UDP */ 817 if (inp->inp_flags & INP_ESPINUDP_ALL) { 818 struct sockaddr *sa = (struct sockaddr *)src; 819 820 switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) { 821 case -1: /* Error, m was freeed */ 822 rcvcnt = -1; 823 goto bad; 824 break; 825 826 case 1: /* ESP over UDP */ 827 rcvcnt++; 828 goto bad; 829 break; 830 831 case 0: /* plain UDP */ 832 default: /* Unexpected */ 833 /* 834 * Normal UDP processing will take place 835 * m may have changed. 836 */ 837 m = *mp; 838 break; 839 } 840 } 841 #endif 842 843 /* 844 * Check the minimum TTL for socket. 845 */ 846 if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl) 847 goto bad; 848 849 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 850 rcvcnt++; 851 } 852 853 bad: 854 return rcvcnt; 855 } 856 #endif 857 858 #ifdef INET6 859 static int 860 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst, 861 struct mbuf *m, int off) 862 { 863 u_int16_t sport, dport; 864 int rcvcnt; 865 struct in6_addr src6, *dst6; 866 const struct in_addr *dst4; 867 struct inpcb_hdr *inph; 868 struct in6pcb *in6p; 869 870 rcvcnt = 0; 871 off += sizeof(struct udphdr); /* now, offset of payload */ 872 873 if (af != AF_INET && af != AF_INET6) 874 goto bad; 875 if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6) 876 goto bad; 877 878 src6 = src->sin6_addr; 879 if (sa6_recoverscope(src) != 0) { 880 /* XXX: should be impossible. */ 881 goto bad; 882 } 883 sport = src->sin6_port; 884 885 dport = dst->sin6_port; 886 dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12]; 887 dst6 = &dst->sin6_addr; 888 889 if (IN6_IS_ADDR_MULTICAST(dst6) || 890 (af == AF_INET && IN_MULTICAST(dst4->s_addr))) { 891 /* 892 * Deliver a multicast or broadcast datagram to *all* sockets 893 * for which the local and remote addresses and ports match 894 * those of the incoming datagram. This allows more than 895 * one process to receive multi/broadcasts on the same port. 896 * (This really ought to be done for unicast datagrams as 897 * well, but that would cause problems with existing 898 * applications that open both address-specific sockets and 899 * a wildcard socket listening to the same port -- they would 900 * end up receiving duplicates of every unicast datagram. 901 * Those applications open the multiple sockets to overcome an 902 * inadequacy of the UDP socket interface, but for backwards 903 * compatibility we avoid the problem here rather than 904 * fixing the interface. Maybe 4.5BSD will remedy this?) 905 */ 906 907 /* 908 * KAME note: traditionally we dropped udpiphdr from mbuf here. 909 * we need udpiphdr for IPsec processing so we do that later. 910 */ 911 /* 912 * Locate pcb(s) for datagram. 913 */ 914 CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 915 in6p = (struct in6pcb *)inph; 916 if (in6p->in6p_af != AF_INET6) 917 continue; 918 919 if (in6p->in6p_lport != dport) 920 continue; 921 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 922 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, 923 dst6)) 924 continue; 925 } else { 926 if (IN6_IS_ADDR_V4MAPPED(dst6) && 927 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 928 continue; 929 } 930 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { 931 if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, 932 &src6) || in6p->in6p_fport != sport) 933 continue; 934 } else { 935 if (IN6_IS_ADDR_V4MAPPED(&src6) && 936 (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) 937 continue; 938 } 939 940 udp6_sendup(m, off, (struct sockaddr *)src, 941 in6p->in6p_socket); 942 rcvcnt++; 943 944 /* 945 * Don't look for additional matches if this one does 946 * not have either the SO_REUSEPORT or SO_REUSEADDR 947 * socket options set. This heuristic avoids searching 948 * through all pcbs in the common case of a non-shared 949 * port. It assumes that an application will never 950 * clear these options after setting them. 951 */ 952 if ((in6p->in6p_socket->so_options & 953 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 954 break; 955 } 956 } else { 957 /* 958 * Locate pcb for datagram. 959 */ 960 in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6, 961 dport, 0, 0); 962 if (in6p == 0) { 963 UDP_STATINC(UDP_STAT_PCBHASHMISS); 964 in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0); 965 if (in6p == 0) 966 return rcvcnt; 967 } 968 969 udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket); 970 rcvcnt++; 971 } 972 973 bad: 974 return rcvcnt; 975 } 976 #endif 977 978 #ifdef INET 979 /* 980 * Notify a udp user of an asynchronous error; 981 * just wake up so that he can collect error status. 982 */ 983 static void 984 udp_notify(struct inpcb *inp, int errno) 985 { 986 inp->inp_socket->so_error = errno; 987 sorwakeup(inp->inp_socket); 988 sowwakeup(inp->inp_socket); 989 } 990 991 void * 992 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v) 993 { 994 struct ip *ip = v; 995 struct udphdr *uh; 996 void (*notify)(struct inpcb *, int) = udp_notify; 997 int errno; 998 999 if (sa->sa_family != AF_INET 1000 || sa->sa_len != sizeof(struct sockaddr_in)) 1001 return NULL; 1002 if ((unsigned)cmd >= PRC_NCMDS) 1003 return NULL; 1004 errno = inetctlerrmap[cmd]; 1005 if (PRC_IS_REDIRECT(cmd)) 1006 notify = in_rtchange, ip = 0; 1007 else if (cmd == PRC_HOSTDEAD) 1008 ip = 0; 1009 else if (errno == 0) 1010 return NULL; 1011 if (ip) { 1012 uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2)); 1013 in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport, 1014 ip->ip_src, uh->uh_sport, errno, notify); 1015 1016 /* XXX mapped address case */ 1017 } else 1018 in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno, 1019 notify); 1020 return NULL; 1021 } 1022 1023 int 1024 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1025 { 1026 int s; 1027 int error = 0; 1028 struct inpcb *inp; 1029 int family; 1030 int optval; 1031 1032 family = so->so_proto->pr_domain->dom_family; 1033 1034 s = splsoftnet(); 1035 switch (family) { 1036 #ifdef INET 1037 case PF_INET: 1038 if (sopt->sopt_level != IPPROTO_UDP) { 1039 error = ip_ctloutput(op, so, sopt); 1040 goto end; 1041 } 1042 break; 1043 #endif 1044 #ifdef INET6 1045 case PF_INET6: 1046 if (sopt->sopt_level != IPPROTO_UDP) { 1047 error = ip6_ctloutput(op, so, sopt); 1048 goto end; 1049 } 1050 break; 1051 #endif 1052 default: 1053 error = EAFNOSUPPORT; 1054 goto end; 1055 } 1056 1057 1058 switch (op) { 1059 case PRCO_SETOPT: 1060 inp = sotoinpcb(so); 1061 1062 switch (sopt->sopt_name) { 1063 case UDP_ENCAP: 1064 error = sockopt_getint(sopt, &optval); 1065 if (error) 1066 break; 1067 1068 switch(optval) { 1069 #ifdef IPSEC_NAT_T 1070 case 0: 1071 inp->inp_flags &= ~INP_ESPINUDP_ALL; 1072 break; 1073 1074 case UDP_ENCAP_ESPINUDP: 1075 inp->inp_flags &= ~INP_ESPINUDP_ALL; 1076 inp->inp_flags |= INP_ESPINUDP; 1077 break; 1078 1079 case UDP_ENCAP_ESPINUDP_NON_IKE: 1080 inp->inp_flags &= ~INP_ESPINUDP_ALL; 1081 inp->inp_flags |= INP_ESPINUDP_NON_IKE; 1082 break; 1083 #endif 1084 default: 1085 error = EINVAL; 1086 break; 1087 } 1088 break; 1089 1090 case UDP_RFC6056ALGO: 1091 error = sockopt_getint(sopt, &optval); 1092 if (error) 1093 break; 1094 1095 error = rfc6056_algo_index_select( 1096 (struct inpcb_hdr *)inp, optval); 1097 break; 1098 1099 default: 1100 error = ENOPROTOOPT; 1101 break; 1102 } 1103 break; 1104 1105 default: 1106 error = EINVAL; 1107 break; 1108 } 1109 1110 end: 1111 splx(s); 1112 return error; 1113 } 1114 1115 1116 int 1117 udp_output(struct mbuf *m, ...) 1118 { 1119 struct inpcb *inp; 1120 struct udpiphdr *ui; 1121 struct route *ro; 1122 int len = m->m_pkthdr.len; 1123 int error = 0; 1124 va_list ap; 1125 1126 MCLAIM(m, &udp_tx_mowner); 1127 va_start(ap, m); 1128 inp = va_arg(ap, struct inpcb *); 1129 va_end(ap); 1130 1131 /* 1132 * Calculate data length and get a mbuf 1133 * for UDP and IP headers. 1134 */ 1135 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 1136 if (m == 0) { 1137 error = ENOBUFS; 1138 goto release; 1139 } 1140 1141 /* 1142 * Compute the packet length of the IP header, and 1143 * punt if the length looks bogus. 1144 */ 1145 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 1146 error = EMSGSIZE; 1147 goto release; 1148 } 1149 1150 /* 1151 * Fill in mbuf with extended UDP header 1152 * and addresses and length put into network format. 1153 */ 1154 ui = mtod(m, struct udpiphdr *); 1155 ui->ui_pr = IPPROTO_UDP; 1156 ui->ui_src = inp->inp_laddr; 1157 ui->ui_dst = inp->inp_faddr; 1158 ui->ui_sport = inp->inp_lport; 1159 ui->ui_dport = inp->inp_fport; 1160 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 1161 1162 ro = &inp->inp_route; 1163 1164 /* 1165 * Set up checksum and output datagram. 1166 */ 1167 if (udpcksum) { 1168 /* 1169 * XXX Cache pseudo-header checksum part for 1170 * XXX "connected" UDP sockets. 1171 */ 1172 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 1173 ui->ui_dst.s_addr, htons((u_int16_t)len + 1174 sizeof(struct udphdr) + IPPROTO_UDP)); 1175 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 1176 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 1177 } else 1178 ui->ui_sum = 0; 1179 ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len); 1180 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ 1181 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ 1182 UDP_STATINC(UDP_STAT_OPACKETS); 1183 1184 return (ip_output(m, inp->inp_options, ro, 1185 inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), 1186 inp->inp_moptions, inp->inp_socket)); 1187 1188 release: 1189 m_freem(m); 1190 return (error); 1191 } 1192 1193 int udp_sendspace = 9216; /* really max datagram size */ 1194 int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); 1195 /* 40 1K datagrams */ 1196 1197 /*ARGSUSED*/ 1198 int 1199 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, 1200 struct mbuf *control, struct lwp *l) 1201 { 1202 struct inpcb *inp; 1203 int s; 1204 int error = 0; 1205 1206 if (req == PRU_CONTROL) 1207 return (in_control(so, (long)m, (void *)nam, 1208 (struct ifnet *)control, l)); 1209 1210 s = splsoftnet(); 1211 1212 if (req == PRU_PURGEIF) { 1213 mutex_enter(softnet_lock); 1214 in_pcbpurgeif0(&udbtable, (struct ifnet *)control); 1215 in_purgeif((struct ifnet *)control); 1216 in_pcbpurgeif(&udbtable, (struct ifnet *)control); 1217 mutex_exit(softnet_lock); 1218 splx(s); 1219 return (0); 1220 } 1221 1222 inp = sotoinpcb(so); 1223 #ifdef DIAGNOSTIC 1224 if (req != PRU_SEND && req != PRU_SENDOOB && control) 1225 panic("udp_usrreq: unexpected control mbuf"); 1226 #endif 1227 if (req == PRU_ATTACH) { 1228 sosetlock(so); 1229 } else if (inp == 0) { 1230 error = EINVAL; 1231 goto release; 1232 } 1233 1234 /* 1235 * Note: need to block udp_input while changing 1236 * the udp pcb queue and/or pcb addresses. 1237 */ 1238 switch (req) { 1239 1240 case PRU_ATTACH: 1241 if (inp != 0) { 1242 error = EISCONN; 1243 break; 1244 } 1245 #ifdef MBUFTRACE 1246 so->so_mowner = &udp_mowner; 1247 so->so_rcv.sb_mowner = &udp_rx_mowner; 1248 so->so_snd.sb_mowner = &udp_tx_mowner; 1249 #endif 1250 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1251 error = soreserve(so, udp_sendspace, udp_recvspace); 1252 if (error) 1253 break; 1254 } 1255 error = in_pcballoc(so, &udbtable); 1256 if (error) 1257 break; 1258 inp = sotoinpcb(so); 1259 inp->inp_ip.ip_ttl = ip_defttl; 1260 break; 1261 1262 case PRU_DETACH: 1263 in_pcbdetach(inp); 1264 break; 1265 1266 case PRU_BIND: 1267 error = in_pcbbind(inp, nam, l); 1268 break; 1269 1270 case PRU_LISTEN: 1271 error = EOPNOTSUPP; 1272 break; 1273 1274 case PRU_CONNECT: 1275 error = in_pcbconnect(inp, nam, l); 1276 if (error) 1277 break; 1278 soisconnected(so); 1279 break; 1280 1281 case PRU_CONNECT2: 1282 error = EOPNOTSUPP; 1283 break; 1284 1285 case PRU_DISCONNECT: 1286 /*soisdisconnected(so);*/ 1287 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 1288 in_pcbdisconnect(inp); 1289 inp->inp_laddr = zeroin_addr; /* XXX */ 1290 in_pcbstate(inp, INP_BOUND); /* XXX */ 1291 break; 1292 1293 case PRU_SHUTDOWN: 1294 socantsendmore(so); 1295 break; 1296 1297 case PRU_RCVD: 1298 error = EOPNOTSUPP; 1299 break; 1300 1301 case PRU_SEND: 1302 if (control && control->m_len) { 1303 m_freem(control); 1304 m_freem(m); 1305 error = EINVAL; 1306 break; 1307 } 1308 { 1309 struct in_addr laddr; /* XXX */ 1310 1311 memset(&laddr, 0, sizeof laddr); 1312 if (nam) { 1313 laddr = inp->inp_laddr; /* XXX */ 1314 if ((so->so_state & SS_ISCONNECTED) != 0) { 1315 error = EISCONN; 1316 goto die; 1317 } 1318 error = in_pcbconnect(inp, nam, l); 1319 if (error) 1320 goto die; 1321 } else { 1322 if ((so->so_state & SS_ISCONNECTED) == 0) { 1323 error = ENOTCONN; 1324 goto die; 1325 } 1326 } 1327 error = udp_output(m, inp); 1328 m = NULL; 1329 if (nam) { 1330 in_pcbdisconnect(inp); 1331 inp->inp_laddr = laddr; /* XXX */ 1332 in_pcbstate(inp, INP_BOUND); /* XXX */ 1333 } 1334 die: 1335 if (m) 1336 m_freem(m); 1337 } 1338 break; 1339 1340 case PRU_SENSE: 1341 /* 1342 * stat: don't bother with a blocksize. 1343 */ 1344 splx(s); 1345 return (0); 1346 1347 case PRU_RCVOOB: 1348 error = EOPNOTSUPP; 1349 break; 1350 1351 case PRU_SENDOOB: 1352 m_freem(control); 1353 m_freem(m); 1354 error = EOPNOTSUPP; 1355 break; 1356 1357 case PRU_SOCKADDR: 1358 in_setsockaddr(inp, nam); 1359 break; 1360 1361 case PRU_PEERADDR: 1362 in_setpeeraddr(inp, nam); 1363 break; 1364 1365 default: 1366 panic("udp_usrreq"); 1367 } 1368 1369 release: 1370 splx(s); 1371 return (error); 1372 } 1373 1374 static int 1375 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) 1376 { 1377 1378 return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS)); 1379 } 1380 1381 /* 1382 * Sysctl for udp variables. 1383 */ 1384 static void 1385 sysctl_net_inet_udp_setup(struct sysctllog **clog) 1386 { 1387 const struct sysctlnode *rfc6056_node; 1388 1389 sysctl_createv(clog, 0, NULL, NULL, 1390 CTLFLAG_PERMANENT, 1391 CTLTYPE_NODE, "net", NULL, 1392 NULL, 0, NULL, 0, 1393 CTL_NET, CTL_EOL); 1394 sysctl_createv(clog, 0, NULL, NULL, 1395 CTLFLAG_PERMANENT, 1396 CTLTYPE_NODE, "inet", NULL, 1397 NULL, 0, NULL, 0, 1398 CTL_NET, PF_INET, CTL_EOL); 1399 sysctl_createv(clog, 0, NULL, NULL, 1400 CTLFLAG_PERMANENT, 1401 CTLTYPE_NODE, "udp", 1402 SYSCTL_DESCR("UDPv4 related settings"), 1403 NULL, 0, NULL, 0, 1404 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL); 1405 1406 sysctl_createv(clog, 0, NULL, NULL, 1407 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1408 CTLTYPE_INT, "checksum", 1409 SYSCTL_DESCR("Compute UDP checksums"), 1410 NULL, 0, &udpcksum, 0, 1411 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM, 1412 CTL_EOL); 1413 sysctl_createv(clog, 0, NULL, NULL, 1414 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1415 CTLTYPE_INT, "sendspace", 1416 SYSCTL_DESCR("Default UDP send buffer size"), 1417 NULL, 0, &udp_sendspace, 0, 1418 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE, 1419 CTL_EOL); 1420 sysctl_createv(clog, 0, NULL, NULL, 1421 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1422 CTLTYPE_INT, "recvspace", 1423 SYSCTL_DESCR("Default UDP receive buffer size"), 1424 NULL, 0, &udp_recvspace, 0, 1425 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE, 1426 CTL_EOL); 1427 sysctl_createv(clog, 0, NULL, NULL, 1428 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1429 CTLTYPE_INT, "do_loopback_cksum", 1430 SYSCTL_DESCR("Perform UDP checksum on loopback"), 1431 NULL, 0, &udp_do_loopback_cksum, 0, 1432 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM, 1433 CTL_EOL); 1434 sysctl_createv(clog, 0, NULL, NULL, 1435 CTLFLAG_PERMANENT, 1436 CTLTYPE_STRUCT, "pcblist", 1437 SYSCTL_DESCR("UDP protocol control block list"), 1438 sysctl_inpcblist, 0, &udbtable, 0, 1439 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE, 1440 CTL_EOL); 1441 sysctl_createv(clog, 0, NULL, NULL, 1442 CTLFLAG_PERMANENT, 1443 CTLTYPE_STRUCT, "stats", 1444 SYSCTL_DESCR("UDP statistics"), 1445 sysctl_net_inet_udp_stats, 0, NULL, 0, 1446 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS, 1447 CTL_EOL); 1448 /* RFC6056 subtree */ 1449 sysctl_createv(clog, 0, NULL, &rfc6056_node, 1450 CTLFLAG_PERMANENT, 1451 CTLTYPE_NODE, "rfc6056", 1452 SYSCTL_DESCR("RFC 6056"), 1453 NULL, 0, NULL, 0, 1454 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE, CTL_EOL); 1455 sysctl_createv(clog, 0, &rfc6056_node, NULL, 1456 CTLFLAG_PERMANENT, 1457 CTLTYPE_STRING, "available", 1458 SYSCTL_DESCR("RFC 6056 available algorithms"), 1459 sysctl_rfc6056_available, 0, NULL, RFC6056_MAXLEN, 1460 CTL_CREATE, CTL_EOL); 1461 sysctl_createv(clog, 0, &rfc6056_node, NULL, 1462 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1463 CTLTYPE_STRING, "selected", 1464 SYSCTL_DESCR("RFC 6056 selected algorithm"), 1465 sysctl_rfc6056_selected, 0, NULL, RFC6056_MAXLEN, 1466 CTL_CREATE, CTL_EOL); 1467 } 1468 #endif 1469 1470 void 1471 udp_statinc(u_int stat) 1472 { 1473 1474 KASSERT(stat < UDP_NSTATS); 1475 UDP_STATINC(stat); 1476 } 1477 1478 #if (defined INET && defined IPSEC_NAT_T) 1479 /* 1480 * Returns: 1481 * 1 if the packet was processed 1482 * 0 if normal UDP processing should take place 1483 * -1 if an error occurent and m was freed 1484 */ 1485 static int 1486 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src, 1487 struct socket *so) 1488 { 1489 size_t len; 1490 void *data; 1491 struct inpcb *inp; 1492 size_t skip = 0; 1493 size_t minlen; 1494 size_t iphdrlen; 1495 struct ip *ip; 1496 struct mbuf *n; 1497 struct m_tag *tag; 1498 struct udphdr *udphdr; 1499 u_int16_t sport, dport; 1500 struct mbuf *m = *mp; 1501 1502 /* 1503 * Collapse the mbuf chain if the first mbuf is too short 1504 * The longest case is: UDP + non ESP marker + ESP 1505 */ 1506 minlen = off + sizeof(u_int64_t) + sizeof(struct esp); 1507 if (minlen > m->m_pkthdr.len) 1508 minlen = m->m_pkthdr.len; 1509 1510 if (m->m_len < minlen) { 1511 if ((*mp = m_pullup(m, minlen)) == NULL) { 1512 printf("udp4_espinudp: m_pullup failed\n"); 1513 return -1; 1514 } 1515 m = *mp; 1516 } 1517 1518 len = m->m_len - off; 1519 data = mtod(m, char *) + off; 1520 inp = sotoinpcb(so); 1521 1522 /* Ignore keepalive packets */ 1523 if ((len == 1) && (*(unsigned char *)data == 0xff)) { 1524 return 1; 1525 } 1526 1527 /* 1528 * Check that the payload is long enough to hold 1529 * an ESP header and compute the length of encapsulation 1530 * header to remove 1531 */ 1532 if (inp->inp_flags & INP_ESPINUDP) { 1533 u_int32_t *st = (u_int32_t *)data; 1534 1535 if ((len <= sizeof(struct esp)) || (*st == 0)) 1536 return 0; /* Normal UDP processing */ 1537 1538 skip = sizeof(struct udphdr); 1539 } 1540 1541 if (inp->inp_flags & INP_ESPINUDP_NON_IKE) { 1542 u_int32_t *st = (u_int32_t *)data; 1543 1544 if ((len <= sizeof(u_int64_t) + sizeof(struct esp)) 1545 || ((st[0] | st[1]) != 0)) 1546 return 0; /* Normal UDP processing */ 1547 1548 skip = sizeof(struct udphdr) + sizeof(u_int64_t); 1549 } 1550 1551 /* 1552 * Get the UDP ports. They are handled in network 1553 * order everywhere in IPSEC_NAT_T code. 1554 */ 1555 udphdr = (struct udphdr *)((char *)data - skip); 1556 sport = udphdr->uh_sport; 1557 dport = udphdr->uh_dport; 1558 1559 /* 1560 * Remove the UDP header (and possibly the non ESP marker) 1561 * IP header lendth is iphdrlen 1562 * Before: 1563 * <--- off ---> 1564 * +----+------+-----+ 1565 * | IP | UDP | ESP | 1566 * +----+------+-----+ 1567 * <-skip-> 1568 * After: 1569 * +----+-----+ 1570 * | IP | ESP | 1571 * +----+-----+ 1572 * <-skip-> 1573 */ 1574 iphdrlen = off - sizeof(struct udphdr); 1575 memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen); 1576 m_adj(m, skip); 1577 1578 ip = mtod(m, struct ip *); 1579 ip->ip_len = htons(ntohs(ip->ip_len) - skip); 1580 ip->ip_p = IPPROTO_ESP; 1581 1582 /* 1583 * Copy the mbuf to avoid multiple free, as both 1584 * esp4_input (which we call) and udp_input (which 1585 * called us) free the mbuf. 1586 */ 1587 if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 1588 printf("udp4_espinudp: m_dup failed\n"); 1589 return 0; 1590 } 1591 1592 /* 1593 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember 1594 * the source UDP port. This is required if we want 1595 * to select the right SPD for multiple hosts behind 1596 * same NAT 1597 */ 1598 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, 1599 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) { 1600 printf("udp4_espinudp: m_tag_get failed\n"); 1601 m_freem(n); 1602 return 0; 1603 } 1604 ((u_int16_t *)(tag + 1))[0] = sport; 1605 ((u_int16_t *)(tag + 1))[1] = dport; 1606 m_tag_prepend(n, tag); 1607 1608 #ifdef FAST_IPSEC 1609 ipsec4_common_input(n, iphdrlen, IPPROTO_ESP); 1610 #else 1611 esp4_input(n, iphdrlen); 1612 #endif 1613 1614 /* We handled it, it shouldn't be handled by UDP */ 1615 return 1; 1616 } 1617 #endif 1618