1 /* $NetBSD: udp_usrreq.c,v 1.199 2014/05/30 01:39:03 christos 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 /* 64 * UDP protocol implementation. 65 * Per RFC 768, August, 1980. 66 */ 67 68 #include <sys/cdefs.h> 69 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.199 2014/05/30 01:39:03 christos Exp $"); 70 71 #include "opt_inet.h" 72 #include "opt_compat_netbsd.h" 73 #include "opt_ipsec.h" 74 #include "opt_inet_csum.h" 75 #include "opt_ipkdb.h" 76 #include "opt_mbuftrace.h" 77 78 #include <sys/param.h> 79 #include <sys/mbuf.h> 80 #include <sys/once.h> 81 #include <sys/protosw.h> 82 #include <sys/socket.h> 83 #include <sys/socketvar.h> 84 #include <sys/systm.h> 85 #include <sys/proc.h> 86 #include <sys/domain.h> 87 #include <sys/sysctl.h> 88 89 #include <net/if.h> 90 #include <net/route.h> 91 92 #include <netinet/in.h> 93 #include <netinet/in_systm.h> 94 #include <netinet/in_var.h> 95 #include <netinet/ip.h> 96 #include <netinet/in_pcb.h> 97 #include <netinet/ip_var.h> 98 #include <netinet/ip_icmp.h> 99 #include <netinet/udp.h> 100 #include <netinet/udp_var.h> 101 #include <netinet/udp_private.h> 102 103 #ifdef INET6 104 #include <netinet/ip6.h> 105 #include <netinet/icmp6.h> 106 #include <netinet6/ip6_var.h> 107 #include <netinet6/ip6_private.h> 108 #include <netinet6/in6_pcb.h> 109 #include <netinet6/udp6_var.h> 110 #include <netinet6/udp6_private.h> 111 #endif 112 113 #ifndef INET6 114 /* always need ip6.h for IP6_EXTHDR_GET */ 115 #include <netinet/ip6.h> 116 #endif 117 118 #ifdef IPSEC 119 #include <netipsec/ipsec.h> 120 #include <netipsec/ipsec_var.h> 121 #include <netipsec/ipsec_private.h> 122 #include <netipsec/esp.h> 123 #ifdef INET6 124 #include <netipsec/ipsec6.h> 125 #endif 126 #endif /* IPSEC */ 127 128 #ifdef COMPAT_50 129 #include <compat/sys/socket.h> 130 #endif 131 132 #ifdef IPKDB 133 #include <ipkdb/ipkdb.h> 134 #endif 135 136 int udpcksum = 1; 137 int udp_do_loopback_cksum = 0; 138 139 struct inpcbtable udbtable; 140 141 percpu_t *udpstat_percpu; 142 143 #ifdef INET 144 #ifdef IPSEC 145 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *, 146 struct socket *); 147 #endif 148 static void udp4_sendup (struct mbuf *, int, struct sockaddr *, 149 struct socket *); 150 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *, 151 struct mbuf **, int); 152 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int); 153 #endif 154 #ifdef INET 155 static void udp_notify (struct inpcb *, int); 156 #endif 157 158 #ifndef UDBHASHSIZE 159 #define UDBHASHSIZE 128 160 #endif 161 int udbhashsize = UDBHASHSIZE; 162 163 /* 164 * For send - really max datagram size; for receive - 40 1K datagrams. 165 */ 166 static int udp_sendspace = 9216; 167 static int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); 168 169 #ifdef MBUFTRACE 170 struct mowner udp_mowner = MOWNER_INIT("udp", ""); 171 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx"); 172 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx"); 173 #endif 174 175 #ifdef UDP_CSUM_COUNTERS 176 #include <sys/device.h> 177 178 #if defined(INET) 179 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 180 NULL, "udp", "hwcsum bad"); 181 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 182 NULL, "udp", "hwcsum ok"); 183 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 184 NULL, "udp", "hwcsum data"); 185 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 186 NULL, "udp", "swcsum"); 187 188 EVCNT_ATTACH_STATIC(udp_hwcsum_bad); 189 EVCNT_ATTACH_STATIC(udp_hwcsum_ok); 190 EVCNT_ATTACH_STATIC(udp_hwcsum_data); 191 EVCNT_ATTACH_STATIC(udp_swcsum); 192 #endif /* defined(INET) */ 193 194 #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 195 #else 196 #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */ 197 #endif /* UDP_CSUM_COUNTERS */ 198 199 static void sysctl_net_inet_udp_setup(struct sysctllog **); 200 201 static int 202 do_udpinit(void) 203 { 204 205 in_pcbinit(&udbtable, udbhashsize, udbhashsize); 206 udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS); 207 208 MOWNER_ATTACH(&udp_tx_mowner); 209 MOWNER_ATTACH(&udp_rx_mowner); 210 MOWNER_ATTACH(&udp_mowner); 211 212 return 0; 213 } 214 215 void 216 udp_init_common(void) 217 { 218 static ONCE_DECL(doudpinit); 219 220 RUN_ONCE(&doudpinit, do_udpinit); 221 } 222 223 void 224 udp_init(void) 225 { 226 227 sysctl_net_inet_udp_setup(NULL); 228 229 udp_init_common(); 230 } 231 232 /* 233 * Checksum extended UDP header and data. 234 */ 235 236 int 237 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh, 238 int iphlen, int len) 239 { 240 241 switch (af) { 242 #ifdef INET 243 case AF_INET: 244 return udp4_input_checksum(m, uh, iphlen, len); 245 #endif 246 #ifdef INET6 247 case AF_INET6: 248 return udp6_input_checksum(m, uh, iphlen, len); 249 #endif 250 } 251 #ifdef DIAGNOSTIC 252 panic("udp_input_checksum: unknown af %d", af); 253 #endif 254 /* NOTREACHED */ 255 return -1; 256 } 257 258 #ifdef INET 259 260 /* 261 * Checksum extended UDP header and data. 262 */ 263 264 static int 265 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh, 266 int iphlen, int len) 267 { 268 269 /* 270 * XXX it's better to record and check if this mbuf is 271 * already checked. 272 */ 273 274 if (uh->uh_sum == 0) 275 return 0; 276 277 switch (m->m_pkthdr.csum_flags & 278 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) | 279 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 280 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD: 281 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad); 282 goto badcsum; 283 284 case M_CSUM_UDPv4|M_CSUM_DATA: { 285 u_int32_t hw_csum = m->m_pkthdr.csum_data; 286 287 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); 288 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { 289 const struct ip *ip = 290 mtod(m, const struct ip *); 291 292 hw_csum = in_cksum_phdr(ip->ip_src.s_addr, 293 ip->ip_dst.s_addr, 294 htons(hw_csum + len + IPPROTO_UDP)); 295 } 296 if ((hw_csum ^ 0xffff) != 0) 297 goto badcsum; 298 break; 299 } 300 301 case M_CSUM_UDPv4: 302 /* Checksum was okay. */ 303 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); 304 break; 305 306 default: 307 /* 308 * Need to compute it ourselves. Maybe skip checksum 309 * on loopback interfaces. 310 */ 311 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 312 IFF_LOOPBACK) || 313 udp_do_loopback_cksum)) { 314 UDP_CSUM_COUNTER_INCR(&udp_swcsum); 315 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) 316 goto badcsum; 317 } 318 break; 319 } 320 321 return 0; 322 323 badcsum: 324 UDP_STATINC(UDP_STAT_BADSUM); 325 return -1; 326 } 327 328 void 329 udp_input(struct mbuf *m, ...) 330 { 331 va_list ap; 332 struct sockaddr_in src, dst; 333 struct ip *ip; 334 struct udphdr *uh; 335 int iphlen; 336 int len; 337 int n; 338 u_int16_t ip_len; 339 340 va_start(ap, m); 341 iphlen = va_arg(ap, int); 342 (void)va_arg(ap, int); /* ignore value, advance ap */ 343 va_end(ap); 344 345 MCLAIM(m, &udp_rx_mowner); 346 UDP_STATINC(UDP_STAT_IPACKETS); 347 348 /* 349 * Get IP and UDP header together in first mbuf. 350 */ 351 ip = mtod(m, struct ip *); 352 IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 353 if (uh == NULL) { 354 UDP_STATINC(UDP_STAT_HDROPS); 355 return; 356 } 357 KASSERT(UDP_HDR_ALIGNED_P(uh)); 358 359 /* destination port of 0 is illegal, based on RFC768. */ 360 if (uh->uh_dport == 0) 361 goto bad; 362 363 /* 364 * Make mbuf data length reflect UDP length. 365 * If not enough data to reflect UDP length, drop. 366 */ 367 ip_len = ntohs(ip->ip_len); 368 len = ntohs((u_int16_t)uh->uh_ulen); 369 if (ip_len != iphlen + len) { 370 if (ip_len < iphlen + len || len < sizeof(struct udphdr)) { 371 UDP_STATINC(UDP_STAT_BADLEN); 372 goto bad; 373 } 374 m_adj(m, iphlen + len - ip_len); 375 } 376 377 /* 378 * Checksum extended UDP header and data. 379 */ 380 if (udp4_input_checksum(m, uh, iphlen, len)) 381 goto badcsum; 382 383 /* construct source and dst sockaddrs. */ 384 sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport); 385 sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport); 386 387 if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) { 388 UDP_STATINC(UDP_STAT_HDROPS); 389 return; 390 } 391 if (m == NULL) { 392 /* 393 * packet has been processed by ESP stuff - 394 * e.g. dropped NAT-T-keep-alive-packet ... 395 */ 396 return; 397 } 398 ip = mtod(m, struct ip *); 399 #ifdef INET6 400 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { 401 struct sockaddr_in6 src6, dst6; 402 403 memset(&src6, 0, sizeof(src6)); 404 src6.sin6_family = AF_INET6; 405 src6.sin6_len = sizeof(struct sockaddr_in6); 406 src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff; 407 memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src, 408 sizeof(ip->ip_src)); 409 src6.sin6_port = uh->uh_sport; 410 memset(&dst6, 0, sizeof(dst6)); 411 dst6.sin6_family = AF_INET6; 412 dst6.sin6_len = sizeof(struct sockaddr_in6); 413 dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff; 414 memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst, 415 sizeof(ip->ip_dst)); 416 dst6.sin6_port = uh->uh_dport; 417 418 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen); 419 } 420 #endif 421 422 if (n == 0) { 423 if (m->m_flags & (M_BCAST | M_MCAST)) { 424 UDP_STATINC(UDP_STAT_NOPORTBCAST); 425 goto bad; 426 } 427 UDP_STATINC(UDP_STAT_NOPORT); 428 #ifdef IPKDB 429 if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport, 430 m, iphlen + sizeof(struct udphdr), 431 m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) { 432 /* 433 * It was a debugger connect packet, 434 * just drop it now 435 */ 436 goto bad; 437 } 438 #endif 439 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 440 m = NULL; 441 } 442 443 bad: 444 if (m) 445 m_freem(m); 446 return; 447 448 badcsum: 449 m_freem(m); 450 } 451 #endif 452 453 #ifdef INET 454 static void 455 udp4_sendup(struct mbuf *m, int off /* offset of data portion */, 456 struct sockaddr *src, struct socket *so) 457 { 458 struct mbuf *opts = NULL; 459 struct mbuf *n; 460 struct inpcb *inp = NULL; 461 462 if (!so) 463 return; 464 switch (so->so_proto->pr_domain->dom_family) { 465 case AF_INET: 466 inp = sotoinpcb(so); 467 break; 468 #ifdef INET6 469 case AF_INET6: 470 break; 471 #endif 472 default: 473 return; 474 } 475 476 #if defined(IPSEC) 477 /* check AH/ESP integrity. */ 478 if (ipsec_used && so != NULL && ipsec4_in_reject_so(m, so)) { 479 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); 480 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) 481 icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT, 482 0, 0); 483 return; 484 } 485 #endif /*IPSEC*/ 486 487 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { 488 if (inp && (inp->inp_flags & INP_CONTROLOPTS 489 #ifdef SO_OTIMESTAMP 490 || so->so_options & SO_OTIMESTAMP 491 #endif 492 || so->so_options & SO_TIMESTAMP)) { 493 struct ip *ip = mtod(n, struct ip *); 494 ip_savecontrol(inp, &opts, ip, n); 495 } 496 497 m_adj(n, off); 498 if (sbappendaddr(&so->so_rcv, src, n, 499 opts) == 0) { 500 m_freem(n); 501 if (opts) 502 m_freem(opts); 503 so->so_rcv.sb_overflowed++; 504 UDP_STATINC(UDP_STAT_FULLSOCK); 505 } else 506 sorwakeup(so); 507 } 508 } 509 #endif 510 511 #ifdef INET 512 static int 513 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst, 514 struct mbuf **mp, int off /* offset of udphdr */) 515 { 516 u_int16_t *sport, *dport; 517 int rcvcnt; 518 struct in_addr *src4, *dst4; 519 struct inpcb_hdr *inph; 520 struct inpcb *inp; 521 struct mbuf *m = *mp; 522 523 rcvcnt = 0; 524 off += sizeof(struct udphdr); /* now, offset of payload */ 525 526 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 527 goto bad; 528 529 src4 = &src->sin_addr; 530 sport = &src->sin_port; 531 dst4 = &dst->sin_addr; 532 dport = &dst->sin_port; 533 534 if (IN_MULTICAST(dst4->s_addr) || 535 in_broadcast(*dst4, m->m_pkthdr.rcvif)) { 536 /* 537 * Deliver a multicast or broadcast datagram to *all* sockets 538 * for which the local and remote addresses and ports match 539 * those of the incoming datagram. This allows more than 540 * one process to receive multi/broadcasts on the same port. 541 * (This really ought to be done for unicast datagrams as 542 * well, but that would cause problems with existing 543 * applications that open both address-specific sockets and 544 * a wildcard socket listening to the same port -- they would 545 * end up receiving duplicates of every unicast datagram. 546 * Those applications open the multiple sockets to overcome an 547 * inadequacy of the UDP socket interface, but for backwards 548 * compatibility we avoid the problem here rather than 549 * fixing the interface. Maybe 4.5BSD will remedy this?) 550 */ 551 552 /* 553 * KAME note: traditionally we dropped udpiphdr from mbuf here. 554 * we need udpiphdr for IPsec processing so we do that later. 555 */ 556 /* 557 * Locate pcb(s) for datagram. 558 */ 559 TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 560 inp = (struct inpcb *)inph; 561 if (inp->inp_af != AF_INET) 562 continue; 563 564 if (inp->inp_lport != *dport) 565 continue; 566 if (!in_nullhost(inp->inp_laddr)) { 567 if (!in_hosteq(inp->inp_laddr, *dst4)) 568 continue; 569 } 570 if (!in_nullhost(inp->inp_faddr)) { 571 if (!in_hosteq(inp->inp_faddr, *src4) || 572 inp->inp_fport != *sport) 573 continue; 574 } 575 576 udp4_sendup(m, off, (struct sockaddr *)src, 577 inp->inp_socket); 578 rcvcnt++; 579 580 /* 581 * Don't look for additional matches if this one does 582 * not have either the SO_REUSEPORT or SO_REUSEADDR 583 * socket options set. This heuristic avoids searching 584 * through all pcbs in the common case of a non-shared 585 * port. It assumes that an application will never 586 * clear these options after setting them. 587 */ 588 if ((inp->inp_socket->so_options & 589 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 590 break; 591 } 592 } else { 593 /* 594 * Locate pcb for datagram. 595 */ 596 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, 597 *dport, 0); 598 if (inp == 0) { 599 UDP_STATINC(UDP_STAT_PCBHASHMISS); 600 inp = in_pcblookup_bind(&udbtable, *dst4, *dport); 601 if (inp == 0) 602 return rcvcnt; 603 } 604 605 #ifdef IPSEC 606 /* Handle ESP over UDP */ 607 if (inp->inp_flags & INP_ESPINUDP_ALL) { 608 struct sockaddr *sa = (struct sockaddr *)src; 609 610 switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) { 611 case -1: /* Error, m was freeed */ 612 rcvcnt = -1; 613 goto bad; 614 break; 615 616 case 1: /* ESP over UDP */ 617 rcvcnt++; 618 goto bad; 619 break; 620 621 case 0: /* plain UDP */ 622 default: /* Unexpected */ 623 /* 624 * Normal UDP processing will take place 625 * m may have changed. 626 */ 627 m = *mp; 628 break; 629 } 630 } 631 #endif 632 633 /* 634 * Check the minimum TTL for socket. 635 */ 636 if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl) 637 goto bad; 638 639 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 640 rcvcnt++; 641 } 642 643 bad: 644 return rcvcnt; 645 } 646 #endif 647 648 #ifdef INET 649 /* 650 * Notify a udp user of an asynchronous error; 651 * just wake up so that he can collect error status. 652 */ 653 static void 654 udp_notify(struct inpcb *inp, int errno) 655 { 656 inp->inp_socket->so_error = errno; 657 sorwakeup(inp->inp_socket); 658 sowwakeup(inp->inp_socket); 659 } 660 661 void * 662 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v) 663 { 664 struct ip *ip = v; 665 struct udphdr *uh; 666 void (*notify)(struct inpcb *, int) = udp_notify; 667 int errno; 668 669 if (sa->sa_family != AF_INET 670 || sa->sa_len != sizeof(struct sockaddr_in)) 671 return NULL; 672 if ((unsigned)cmd >= PRC_NCMDS) 673 return NULL; 674 errno = inetctlerrmap[cmd]; 675 if (PRC_IS_REDIRECT(cmd)) 676 notify = in_rtchange, ip = 0; 677 else if (cmd == PRC_HOSTDEAD) 678 ip = 0; 679 else if (errno == 0) 680 return NULL; 681 if (ip) { 682 uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2)); 683 in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport, 684 ip->ip_src, uh->uh_sport, errno, notify); 685 686 /* XXX mapped address case */ 687 } else 688 in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno, 689 notify); 690 return NULL; 691 } 692 693 int 694 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 695 { 696 int s; 697 int error = 0; 698 struct inpcb *inp; 699 int family; 700 int optval; 701 702 family = so->so_proto->pr_domain->dom_family; 703 704 s = splsoftnet(); 705 switch (family) { 706 #ifdef INET 707 case PF_INET: 708 if (sopt->sopt_level != IPPROTO_UDP) { 709 error = ip_ctloutput(op, so, sopt); 710 goto end; 711 } 712 break; 713 #endif 714 #ifdef INET6 715 case PF_INET6: 716 if (sopt->sopt_level != IPPROTO_UDP) { 717 error = ip6_ctloutput(op, so, sopt); 718 goto end; 719 } 720 break; 721 #endif 722 default: 723 error = EAFNOSUPPORT; 724 goto end; 725 } 726 727 728 switch (op) { 729 case PRCO_SETOPT: 730 inp = sotoinpcb(so); 731 732 switch (sopt->sopt_name) { 733 case UDP_ENCAP: 734 error = sockopt_getint(sopt, &optval); 735 if (error) 736 break; 737 738 switch(optval) { 739 case 0: 740 inp->inp_flags &= ~INP_ESPINUDP_ALL; 741 break; 742 743 case UDP_ENCAP_ESPINUDP: 744 inp->inp_flags &= ~INP_ESPINUDP_ALL; 745 inp->inp_flags |= INP_ESPINUDP; 746 break; 747 748 case UDP_ENCAP_ESPINUDP_NON_IKE: 749 inp->inp_flags &= ~INP_ESPINUDP_ALL; 750 inp->inp_flags |= INP_ESPINUDP_NON_IKE; 751 break; 752 default: 753 error = EINVAL; 754 break; 755 } 756 break; 757 758 default: 759 error = ENOPROTOOPT; 760 break; 761 } 762 break; 763 764 default: 765 error = EINVAL; 766 break; 767 } 768 769 end: 770 splx(s); 771 return error; 772 } 773 774 775 int 776 udp_output(struct mbuf *m, ...) 777 { 778 struct inpcb *inp; 779 struct udpiphdr *ui; 780 struct route *ro; 781 int len = m->m_pkthdr.len; 782 int error = 0; 783 va_list ap; 784 785 MCLAIM(m, &udp_tx_mowner); 786 va_start(ap, m); 787 inp = va_arg(ap, struct inpcb *); 788 va_end(ap); 789 790 /* 791 * Calculate data length and get a mbuf 792 * for UDP and IP headers. 793 */ 794 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 795 if (m == 0) { 796 error = ENOBUFS; 797 goto release; 798 } 799 800 /* 801 * Compute the packet length of the IP header, and 802 * punt if the length looks bogus. 803 */ 804 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 805 error = EMSGSIZE; 806 goto release; 807 } 808 809 /* 810 * Fill in mbuf with extended UDP header 811 * and addresses and length put into network format. 812 */ 813 ui = mtod(m, struct udpiphdr *); 814 ui->ui_pr = IPPROTO_UDP; 815 ui->ui_src = inp->inp_laddr; 816 ui->ui_dst = inp->inp_faddr; 817 ui->ui_sport = inp->inp_lport; 818 ui->ui_dport = inp->inp_fport; 819 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 820 821 ro = &inp->inp_route; 822 823 /* 824 * Set up checksum and output datagram. 825 */ 826 if (udpcksum) { 827 /* 828 * XXX Cache pseudo-header checksum part for 829 * XXX "connected" UDP sockets. 830 */ 831 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 832 ui->ui_dst.s_addr, htons((u_int16_t)len + 833 sizeof(struct udphdr) + IPPROTO_UDP)); 834 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 835 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 836 } else 837 ui->ui_sum = 0; 838 ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len); 839 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ 840 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ 841 UDP_STATINC(UDP_STAT_OPACKETS); 842 843 return (ip_output(m, inp->inp_options, ro, 844 inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), 845 inp->inp_moptions, inp->inp_socket)); 846 847 release: 848 m_freem(m); 849 return (error); 850 } 851 852 static int 853 udp_attach(struct socket *so, int proto) 854 { 855 struct inpcb *inp; 856 int error; 857 858 KASSERT(sotoinpcb(so) == NULL); 859 860 /* Assign the lock (must happen even if we will error out). */ 861 sosetlock(so); 862 863 #ifdef MBUFTRACE 864 so->so_mowner = &udp_mowner; 865 so->so_rcv.sb_mowner = &udp_rx_mowner; 866 so->so_snd.sb_mowner = &udp_tx_mowner; 867 #endif 868 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 869 error = soreserve(so, udp_sendspace, udp_recvspace); 870 if (error) { 871 return error; 872 } 873 } 874 875 error = in_pcballoc(so, &udbtable); 876 if (error) { 877 return error; 878 } 879 inp = sotoinpcb(so); 880 inp->inp_ip.ip_ttl = ip_defttl; 881 KASSERT(solocked(so)); 882 883 return error; 884 } 885 886 static void 887 udp_detach(struct socket *so) 888 { 889 struct inpcb *inp; 890 891 KASSERT(solocked(so)); 892 inp = sotoinpcb(so); 893 KASSERT(inp != NULL); 894 in_pcbdetach(inp); 895 } 896 897 static int 898 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, 899 struct mbuf *control, struct lwp *l) 900 { 901 struct inpcb *inp; 902 int s, error = 0; 903 904 KASSERT(req != PRU_ATTACH); 905 KASSERT(req != PRU_DETACH); 906 907 if (req == PRU_CONTROL) { 908 return in_control(so, (long)m, (void *)nam, 909 (struct ifnet *)control, l); 910 } 911 s = splsoftnet(); 912 if (req == PRU_PURGEIF) { 913 mutex_enter(softnet_lock); 914 in_pcbpurgeif0(&udbtable, (struct ifnet *)control); 915 in_purgeif((struct ifnet *)control); 916 in_pcbpurgeif(&udbtable, (struct ifnet *)control); 917 mutex_exit(softnet_lock); 918 splx(s); 919 return 0; 920 } 921 922 KASSERT(solocked(so)); 923 inp = sotoinpcb(so); 924 925 KASSERT(!control || (req == PRU_SEND || req == PRU_SENDOOB)); 926 if (inp == NULL) { 927 splx(s); 928 return EINVAL; 929 } 930 931 /* 932 * Note: need to block udp_input while changing 933 * the udp pcb queue and/or pcb addresses. 934 */ 935 switch (req) { 936 case PRU_BIND: 937 error = in_pcbbind(inp, nam, l); 938 break; 939 940 case PRU_LISTEN: 941 error = EOPNOTSUPP; 942 break; 943 944 case PRU_CONNECT: 945 error = in_pcbconnect(inp, nam, l); 946 if (error) 947 break; 948 soisconnected(so); 949 break; 950 951 case PRU_CONNECT2: 952 error = EOPNOTSUPP; 953 break; 954 955 case PRU_DISCONNECT: 956 /*soisdisconnected(so);*/ 957 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 958 in_pcbdisconnect(inp); 959 inp->inp_laddr = zeroin_addr; /* XXX */ 960 in_pcbstate(inp, INP_BOUND); /* XXX */ 961 break; 962 963 case PRU_SHUTDOWN: 964 socantsendmore(so); 965 break; 966 967 case PRU_RCVD: 968 error = EOPNOTSUPP; 969 break; 970 971 case PRU_SEND: 972 if (control && control->m_len) { 973 m_freem(control); 974 m_freem(m); 975 error = EINVAL; 976 break; 977 } 978 { 979 struct in_addr laddr; /* XXX */ 980 981 memset(&laddr, 0, sizeof laddr); 982 if (nam) { 983 laddr = inp->inp_laddr; /* XXX */ 984 if ((so->so_state & SS_ISCONNECTED) != 0) { 985 error = EISCONN; 986 goto die; 987 } 988 error = in_pcbconnect(inp, nam, l); 989 if (error) 990 goto die; 991 } else { 992 if ((so->so_state & SS_ISCONNECTED) == 0) { 993 error = ENOTCONN; 994 goto die; 995 } 996 } 997 error = udp_output(m, inp); 998 m = NULL; 999 if (nam) { 1000 in_pcbdisconnect(inp); 1001 inp->inp_laddr = laddr; /* XXX */ 1002 in_pcbstate(inp, INP_BOUND); /* XXX */ 1003 } 1004 die: 1005 if (m) 1006 m_freem(m); 1007 } 1008 break; 1009 1010 case PRU_SENSE: 1011 /* 1012 * stat: don't bother with a blocksize. 1013 */ 1014 splx(s); 1015 return (0); 1016 1017 case PRU_RCVOOB: 1018 error = EOPNOTSUPP; 1019 break; 1020 1021 case PRU_SENDOOB: 1022 m_freem(control); 1023 m_freem(m); 1024 error = EOPNOTSUPP; 1025 break; 1026 1027 case PRU_SOCKADDR: 1028 in_setsockaddr(inp, nam); 1029 break; 1030 1031 case PRU_PEERADDR: 1032 in_setpeeraddr(inp, nam); 1033 break; 1034 1035 default: 1036 panic("udp_usrreq"); 1037 } 1038 splx(s); 1039 1040 return error; 1041 } 1042 1043 static int 1044 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) 1045 { 1046 1047 return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS)); 1048 } 1049 1050 /* 1051 * Sysctl for udp variables. 1052 */ 1053 static void 1054 sysctl_net_inet_udp_setup(struct sysctllog **clog) 1055 { 1056 1057 sysctl_createv(clog, 0, NULL, NULL, 1058 CTLFLAG_PERMANENT, 1059 CTLTYPE_NODE, "inet", NULL, 1060 NULL, 0, NULL, 0, 1061 CTL_NET, PF_INET, CTL_EOL); 1062 sysctl_createv(clog, 0, NULL, NULL, 1063 CTLFLAG_PERMANENT, 1064 CTLTYPE_NODE, "udp", 1065 SYSCTL_DESCR("UDPv4 related settings"), 1066 NULL, 0, NULL, 0, 1067 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL); 1068 1069 sysctl_createv(clog, 0, NULL, NULL, 1070 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1071 CTLTYPE_INT, "checksum", 1072 SYSCTL_DESCR("Compute UDP checksums"), 1073 NULL, 0, &udpcksum, 0, 1074 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM, 1075 CTL_EOL); 1076 sysctl_createv(clog, 0, NULL, NULL, 1077 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1078 CTLTYPE_INT, "sendspace", 1079 SYSCTL_DESCR("Default UDP send buffer size"), 1080 NULL, 0, &udp_sendspace, 0, 1081 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE, 1082 CTL_EOL); 1083 sysctl_createv(clog, 0, NULL, NULL, 1084 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1085 CTLTYPE_INT, "recvspace", 1086 SYSCTL_DESCR("Default UDP receive buffer size"), 1087 NULL, 0, &udp_recvspace, 0, 1088 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE, 1089 CTL_EOL); 1090 sysctl_createv(clog, 0, NULL, NULL, 1091 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1092 CTLTYPE_INT, "do_loopback_cksum", 1093 SYSCTL_DESCR("Perform UDP checksum on loopback"), 1094 NULL, 0, &udp_do_loopback_cksum, 0, 1095 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM, 1096 CTL_EOL); 1097 sysctl_createv(clog, 0, NULL, NULL, 1098 CTLFLAG_PERMANENT, 1099 CTLTYPE_STRUCT, "pcblist", 1100 SYSCTL_DESCR("UDP protocol control block list"), 1101 sysctl_inpcblist, 0, &udbtable, 0, 1102 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE, 1103 CTL_EOL); 1104 sysctl_createv(clog, 0, NULL, NULL, 1105 CTLFLAG_PERMANENT, 1106 CTLTYPE_STRUCT, "stats", 1107 SYSCTL_DESCR("UDP statistics"), 1108 sysctl_net_inet_udp_stats, 0, NULL, 0, 1109 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS, 1110 CTL_EOL); 1111 } 1112 #endif 1113 1114 void 1115 udp_statinc(u_int stat) 1116 { 1117 1118 KASSERT(stat < UDP_NSTATS); 1119 UDP_STATINC(stat); 1120 } 1121 1122 #if defined(INET) && defined(IPSEC) 1123 /* 1124 * Returns: 1125 * 1 if the packet was processed 1126 * 0 if normal UDP processing should take place 1127 * -1 if an error occurent and m was freed 1128 */ 1129 static int 1130 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src, 1131 struct socket *so) 1132 { 1133 size_t len; 1134 void *data; 1135 struct inpcb *inp; 1136 size_t skip = 0; 1137 size_t minlen; 1138 size_t iphdrlen; 1139 struct ip *ip; 1140 struct m_tag *tag; 1141 struct udphdr *udphdr; 1142 u_int16_t sport, dport; 1143 struct mbuf *m = *mp; 1144 1145 /* 1146 * Collapse the mbuf chain if the first mbuf is too short 1147 * The longest case is: UDP + non ESP marker + ESP 1148 */ 1149 minlen = off + sizeof(u_int64_t) + sizeof(struct esp); 1150 if (minlen > m->m_pkthdr.len) 1151 minlen = m->m_pkthdr.len; 1152 1153 if (m->m_len < minlen) { 1154 if ((*mp = m_pullup(m, minlen)) == NULL) { 1155 printf("udp4_espinudp: m_pullup failed\n"); 1156 return -1; 1157 } 1158 m = *mp; 1159 } 1160 1161 len = m->m_len - off; 1162 data = mtod(m, char *) + off; 1163 inp = sotoinpcb(so); 1164 1165 /* Ignore keepalive packets */ 1166 if ((len == 1) && (*(unsigned char *)data == 0xff)) { 1167 m_free(m); 1168 *mp = NULL; /* avoid any further processiong by caller ... */ 1169 return 1; 1170 } 1171 1172 /* 1173 * Check that the payload is long enough to hold 1174 * an ESP header and compute the length of encapsulation 1175 * header to remove 1176 */ 1177 if (inp->inp_flags & INP_ESPINUDP) { 1178 u_int32_t *st = (u_int32_t *)data; 1179 1180 if ((len <= sizeof(struct esp)) || (*st == 0)) 1181 return 0; /* Normal UDP processing */ 1182 1183 skip = sizeof(struct udphdr); 1184 } 1185 1186 if (inp->inp_flags & INP_ESPINUDP_NON_IKE) { 1187 u_int32_t *st = (u_int32_t *)data; 1188 1189 if ((len <= sizeof(u_int64_t) + sizeof(struct esp)) 1190 || ((st[0] | st[1]) != 0)) 1191 return 0; /* Normal UDP processing */ 1192 1193 skip = sizeof(struct udphdr) + sizeof(u_int64_t); 1194 } 1195 1196 /* 1197 * Get the UDP ports. They are handled in network 1198 * order everywhere in IPSEC_NAT_T code. 1199 */ 1200 udphdr = (struct udphdr *)((char *)data - skip); 1201 sport = udphdr->uh_sport; 1202 dport = udphdr->uh_dport; 1203 1204 /* 1205 * Remove the UDP header (and possibly the non ESP marker) 1206 * IP header lendth is iphdrlen 1207 * Before: 1208 * <--- off ---> 1209 * +----+------+-----+ 1210 * | IP | UDP | ESP | 1211 * +----+------+-----+ 1212 * <-skip-> 1213 * After: 1214 * +----+-----+ 1215 * | IP | ESP | 1216 * +----+-----+ 1217 * <-skip-> 1218 */ 1219 iphdrlen = off - sizeof(struct udphdr); 1220 memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen); 1221 m_adj(m, skip); 1222 1223 ip = mtod(m, struct ip *); 1224 ip->ip_len = htons(ntohs(ip->ip_len) - skip); 1225 ip->ip_p = IPPROTO_ESP; 1226 1227 /* 1228 * We have modified the packet - it is now ESP, so we should not 1229 * return to UDP processing ... 1230 * 1231 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember 1232 * the source UDP port. This is required if we want 1233 * to select the right SPD for multiple hosts behind 1234 * same NAT 1235 */ 1236 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, 1237 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) { 1238 printf("udp4_espinudp: m_tag_get failed\n"); 1239 m_freem(m); 1240 return -1; 1241 } 1242 ((u_int16_t *)(tag + 1))[0] = sport; 1243 ((u_int16_t *)(tag + 1))[1] = dport; 1244 m_tag_prepend(m, tag); 1245 1246 #ifdef IPSEC 1247 if (ipsec_used) 1248 ipsec4_common_input(m, iphdrlen, IPPROTO_ESP); 1249 /* XXX: else */ 1250 #else 1251 esp4_input(m, iphdrlen); 1252 #endif 1253 1254 /* We handled it, it shouldn't be handled by UDP */ 1255 *mp = NULL; /* avoid free by caller ... */ 1256 return 1; 1257 } 1258 #endif 1259 1260 PR_WRAP_USRREQS(udp) 1261 #define udp_attach udp_attach_wrapper 1262 #define udp_detach udp_detach_wrapper 1263 #define udp_usrreq udp_usrreq_wrapper 1264 1265 const struct pr_usrreqs udp_usrreqs = { 1266 .pr_attach = udp_attach, 1267 .pr_detach = udp_detach, 1268 .pr_generic = udp_usrreq, 1269 }; 1270