1 /* $NetBSD: udp_usrreq.c,v 1.255 2018/07/15 05:16:45 maxv 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 1.255 2018/07/15 05:16:45 maxv Exp $"); 70 71 #ifdef _KERNEL_OPT 72 #include "opt_inet.h" 73 #include "opt_ipsec.h" 74 #include "opt_inet_csum.h" 75 #include "opt_mbuftrace.h" 76 #include "opt_net_mpsafe.h" 77 #endif 78 79 #include <sys/param.h> 80 #include <sys/mbuf.h> 81 #include <sys/once.h> 82 #include <sys/protosw.h> 83 #include <sys/socket.h> 84 #include <sys/socketvar.h> 85 #include <sys/systm.h> 86 #include <sys/proc.h> 87 #include <sys/domain.h> 88 #include <sys/sysctl.h> 89 90 #include <net/if.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 <netinet6/ip6_var.h> 106 #include <netinet6/ip6_private.h> 107 #include <netinet6/in6_pcb.h> 108 #include <netinet6/udp6_var.h> 109 #include <netinet6/udp6_private.h> 110 #endif 111 112 #ifndef INET6 113 #include <netinet/ip6.h> 114 #endif 115 116 #ifdef IPSEC 117 #include <netipsec/ipsec.h> 118 #include <netipsec/esp.h> 119 #endif 120 121 int udpcksum = 1; 122 int udp_do_loopback_cksum = 0; 123 124 struct inpcbtable udbtable; 125 126 percpu_t *udpstat_percpu; 127 128 #ifdef INET 129 #ifdef IPSEC 130 static int udp4_espinudp(struct mbuf **, int, struct socket *); 131 #endif 132 static void udp4_sendup(struct mbuf *, int, struct sockaddr *, 133 struct socket *); 134 static int udp4_realinput(struct sockaddr_in *, struct sockaddr_in *, 135 struct mbuf **, int); 136 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int); 137 #endif 138 #ifdef INET 139 static void udp_notify (struct inpcb *, int); 140 #endif 141 142 #ifndef UDBHASHSIZE 143 #define UDBHASHSIZE 128 144 #endif 145 int udbhashsize = UDBHASHSIZE; 146 147 /* 148 * For send - really max datagram size; for receive - 40 1K datagrams. 149 */ 150 static int udp_sendspace = 9216; 151 static int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); 152 153 #ifdef MBUFTRACE 154 struct mowner udp_mowner = MOWNER_INIT("udp", ""); 155 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx"); 156 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx"); 157 #endif 158 159 #ifdef UDP_CSUM_COUNTERS 160 #include <sys/device.h> 161 162 #if defined(INET) 163 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 164 NULL, "udp", "hwcsum bad"); 165 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 166 NULL, "udp", "hwcsum ok"); 167 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 168 NULL, "udp", "hwcsum data"); 169 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 170 NULL, "udp", "swcsum"); 171 172 EVCNT_ATTACH_STATIC(udp_hwcsum_bad); 173 EVCNT_ATTACH_STATIC(udp_hwcsum_ok); 174 EVCNT_ATTACH_STATIC(udp_hwcsum_data); 175 EVCNT_ATTACH_STATIC(udp_swcsum); 176 #endif /* defined(INET) */ 177 178 #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 179 #else 180 #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */ 181 #endif /* UDP_CSUM_COUNTERS */ 182 183 static void sysctl_net_inet_udp_setup(struct sysctllog **); 184 185 static int 186 do_udpinit(void) 187 { 188 189 in_pcbinit(&udbtable, udbhashsize, udbhashsize); 190 udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS); 191 192 MOWNER_ATTACH(&udp_tx_mowner); 193 MOWNER_ATTACH(&udp_rx_mowner); 194 MOWNER_ATTACH(&udp_mowner); 195 196 return 0; 197 } 198 199 void 200 udp_init_common(void) 201 { 202 static ONCE_DECL(doudpinit); 203 204 RUN_ONCE(&doudpinit, do_udpinit); 205 } 206 207 void 208 udp_init(void) 209 { 210 211 sysctl_net_inet_udp_setup(NULL); 212 213 udp_init_common(); 214 } 215 216 /* 217 * Checksum extended UDP header and data. 218 */ 219 int 220 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh, 221 int iphlen, int len) 222 { 223 224 switch (af) { 225 #ifdef INET 226 case AF_INET: 227 return udp4_input_checksum(m, uh, iphlen, len); 228 #endif 229 #ifdef INET6 230 case AF_INET6: 231 return udp6_input_checksum(m, uh, iphlen, len); 232 #endif 233 } 234 #ifdef DIAGNOSTIC 235 panic("udp_input_checksum: unknown af %d", af); 236 #endif 237 /* NOTREACHED */ 238 return -1; 239 } 240 241 #ifdef INET 242 243 /* 244 * Checksum extended UDP header and data. 245 */ 246 static int 247 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh, 248 int iphlen, int len) 249 { 250 251 /* 252 * XXX it's better to record and check if this mbuf is 253 * already checked. 254 */ 255 256 if (uh->uh_sum == 0) 257 return 0; 258 259 switch (m->m_pkthdr.csum_flags & 260 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv4) | 261 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 262 case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD: 263 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad); 264 goto badcsum; 265 266 case M_CSUM_UDPv4|M_CSUM_DATA: { 267 u_int32_t hw_csum = m->m_pkthdr.csum_data; 268 269 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); 270 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { 271 const struct ip *ip = 272 mtod(m, const struct ip *); 273 274 hw_csum = in_cksum_phdr(ip->ip_src.s_addr, 275 ip->ip_dst.s_addr, 276 htons(hw_csum + len + IPPROTO_UDP)); 277 } 278 if ((hw_csum ^ 0xffff) != 0) 279 goto badcsum; 280 break; 281 } 282 283 case M_CSUM_UDPv4: 284 /* Checksum was okay. */ 285 UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); 286 break; 287 288 default: 289 /* 290 * Need to compute it ourselves. Maybe skip checksum 291 * on loopback interfaces. 292 */ 293 if (__predict_true(!(m_get_rcvif_NOMPSAFE(m)->if_flags & 294 IFF_LOOPBACK) || 295 udp_do_loopback_cksum)) { 296 UDP_CSUM_COUNTER_INCR(&udp_swcsum); 297 if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) 298 goto badcsum; 299 } 300 break; 301 } 302 303 return 0; 304 305 badcsum: 306 UDP_STATINC(UDP_STAT_BADSUM); 307 return -1; 308 } 309 310 void 311 udp_input(struct mbuf *m, ...) 312 { 313 va_list ap; 314 struct sockaddr_in src, dst; 315 struct ip *ip; 316 struct udphdr *uh; 317 int iphlen; 318 int len; 319 int n; 320 u_int16_t ip_len; 321 322 va_start(ap, m); 323 iphlen = va_arg(ap, int); 324 (void)va_arg(ap, int); /* ignore value, advance ap */ 325 va_end(ap); 326 327 MCLAIM(m, &udp_rx_mowner); 328 UDP_STATINC(UDP_STAT_IPACKETS); 329 330 /* 331 * Get IP and UDP header together in first mbuf. 332 */ 333 ip = mtod(m, struct ip *); 334 M_REGION_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 335 if (uh == NULL) { 336 UDP_STATINC(UDP_STAT_HDROPS); 337 return; 338 } 339 340 /* 341 * Enforce alignment requirements that are violated in 342 * some cases, see kern/50766 for details. 343 */ 344 if (UDP_HDR_ALIGNED_P(uh) == 0) { 345 m = m_copyup(m, iphlen + sizeof(struct udphdr), 0); 346 if (m == NULL) { 347 UDP_STATINC(UDP_STAT_HDROPS); 348 return; 349 } 350 ip = mtod(m, struct ip *); 351 uh = (struct udphdr *)(mtod(m, char *) + iphlen); 352 } 353 KASSERT(UDP_HDR_ALIGNED_P(uh)); 354 355 /* destination port of 0 is illegal, based on RFC768. */ 356 if (uh->uh_dport == 0) 357 goto bad; 358 359 /* 360 * Make mbuf data length reflect UDP length. 361 * If not enough data to reflect UDP length, drop. 362 */ 363 ip_len = ntohs(ip->ip_len); 364 len = ntohs((u_int16_t)uh->uh_ulen); 365 if (len < sizeof(struct udphdr)) { 366 UDP_STATINC(UDP_STAT_BADLEN); 367 goto bad; 368 } 369 if (ip_len != iphlen + len) { 370 if (ip_len < iphlen + len) { 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 } 398 399 ip = mtod(m, struct ip *); 400 M_REGION_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 401 if (uh == NULL) { 402 UDP_STATINC(UDP_STAT_HDROPS); 403 return; 404 } 405 /* XXX Re-enforce alignment? */ 406 407 #ifdef INET6 408 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { 409 struct sockaddr_in6 src6, dst6; 410 411 memset(&src6, 0, sizeof(src6)); 412 src6.sin6_family = AF_INET6; 413 src6.sin6_len = sizeof(struct sockaddr_in6); 414 in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr); 415 src6.sin6_port = uh->uh_sport; 416 memset(&dst6, 0, sizeof(dst6)); 417 dst6.sin6_family = AF_INET6; 418 dst6.sin6_len = sizeof(struct sockaddr_in6); 419 in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr); 420 dst6.sin6_port = uh->uh_dport; 421 422 n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen); 423 } 424 #endif 425 426 if (n == 0) { 427 if (m->m_flags & (M_BCAST | M_MCAST)) { 428 UDP_STATINC(UDP_STAT_NOPORTBCAST); 429 goto bad; 430 } 431 UDP_STATINC(UDP_STAT_NOPORT); 432 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 433 m = NULL; 434 } 435 436 bad: 437 if (m) 438 m_freem(m); 439 return; 440 441 badcsum: 442 m_freem(m); 443 } 444 #endif 445 446 #ifdef INET 447 static void 448 udp4_sendup(struct mbuf *m, int off /* offset of data portion */, 449 struct sockaddr *src, struct socket *so) 450 { 451 struct mbuf *opts = NULL; 452 struct mbuf *n; 453 struct inpcb *inp; 454 455 KASSERT(so != NULL); 456 KASSERT(so->so_proto->pr_domain->dom_family == AF_INET); 457 inp = sotoinpcb(so); 458 KASSERT(inp != NULL); 459 460 #if defined(IPSEC) 461 if (ipsec_used && ipsec_in_reject(m, inp)) { 462 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) 463 icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT, 464 0, 0); 465 return; 466 } 467 #endif 468 469 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { 470 if (inp->inp_flags & INP_CONTROLOPTS || 471 SOOPT_TIMESTAMP(so->so_options)) { 472 struct ip *ip = mtod(n, struct ip *); 473 ip_savecontrol(inp, &opts, ip, n); 474 } 475 476 m_adj(n, off); 477 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) { 478 m_freem(n); 479 if (opts) 480 m_freem(opts); 481 UDP_STATINC(UDP_STAT_FULLSOCK); 482 soroverflow(so); 483 } else 484 sorwakeup(so); 485 } 486 } 487 #endif 488 489 #ifdef INET 490 static int 491 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst, 492 struct mbuf **mp, int off /* offset of udphdr */) 493 { 494 u_int16_t *sport, *dport; 495 int rcvcnt; 496 struct in_addr *src4, *dst4; 497 struct inpcb_hdr *inph; 498 struct inpcb *inp; 499 struct mbuf *m = *mp; 500 501 rcvcnt = 0; 502 off += sizeof(struct udphdr); /* now, offset of payload */ 503 504 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 505 goto bad; 506 507 src4 = &src->sin_addr; 508 sport = &src->sin_port; 509 dst4 = &dst->sin_addr; 510 dport = &dst->sin_port; 511 512 if (IN_MULTICAST(dst4->s_addr) || 513 in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) { 514 /* 515 * Deliver a multicast or broadcast datagram to *all* sockets 516 * for which the local and remote addresses and ports match 517 * those of the incoming datagram. This allows more than 518 * one process to receive multi/broadcasts on the same port. 519 * (This really ought to be done for unicast datagrams as 520 * well, but that would cause problems with existing 521 * applications that open both address-specific sockets and 522 * a wildcard socket listening to the same port -- they would 523 * end up receiving duplicates of every unicast datagram. 524 * Those applications open the multiple sockets to overcome an 525 * inadequacy of the UDP socket interface, but for backwards 526 * compatibility we avoid the problem here rather than 527 * fixing the interface. Maybe 4.5BSD will remedy this?) 528 */ 529 530 /* 531 * KAME note: traditionally we dropped udpiphdr from mbuf here. 532 * we need udpiphdr for IPsec processing so we do that later. 533 */ 534 /* 535 * Locate pcb(s) for datagram. 536 */ 537 TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 538 inp = (struct inpcb *)inph; 539 if (inp->inp_af != AF_INET) 540 continue; 541 542 if (inp->inp_lport != *dport) 543 continue; 544 if (!in_nullhost(inp->inp_laddr)) { 545 if (!in_hosteq(inp->inp_laddr, *dst4)) 546 continue; 547 } 548 if (!in_nullhost(inp->inp_faddr)) { 549 if (!in_hosteq(inp->inp_faddr, *src4) || 550 inp->inp_fport != *sport) 551 continue; 552 } 553 554 udp4_sendup(m, off, (struct sockaddr *)src, 555 inp->inp_socket); 556 rcvcnt++; 557 558 /* 559 * Don't look for additional matches if this one does 560 * not have either the SO_REUSEPORT or SO_REUSEADDR 561 * socket options set. This heuristic avoids searching 562 * through all pcbs in the common case of a non-shared 563 * port. It assumes that an application will never 564 * clear these options after setting them. 565 */ 566 if ((inp->inp_socket->so_options & 567 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 568 break; 569 } 570 } else { 571 /* 572 * Locate pcb for datagram. 573 */ 574 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, 575 *dport, 0); 576 if (inp == 0) { 577 UDP_STATINC(UDP_STAT_PCBHASHMISS); 578 inp = in_pcblookup_bind(&udbtable, *dst4, *dport); 579 if (inp == 0) 580 return rcvcnt; 581 } 582 583 #ifdef IPSEC 584 /* Handle ESP over UDP */ 585 if (inp->inp_flags & INP_ESPINUDP) { 586 switch (udp4_espinudp(mp, off, inp->inp_socket)) { 587 case -1: /* Error, m was freed */ 588 rcvcnt = -1; 589 goto bad; 590 591 case 1: /* ESP over UDP */ 592 rcvcnt++; 593 goto bad; 594 595 case 0: /* plain UDP */ 596 default: /* Unexpected */ 597 /* 598 * Normal UDP processing will take place, 599 * m may have changed. 600 */ 601 m = *mp; 602 break; 603 } 604 } 605 #endif 606 607 /* 608 * Check the minimum TTL for socket. 609 */ 610 if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl) 611 goto bad; 612 613 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 614 rcvcnt++; 615 } 616 617 bad: 618 return rcvcnt; 619 } 620 #endif 621 622 #ifdef INET 623 /* 624 * Notify a udp user of an asynchronous error; 625 * just wake up so that he can collect error status. 626 */ 627 static void 628 udp_notify(struct inpcb *inp, int errno) 629 { 630 inp->inp_socket->so_error = errno; 631 sorwakeup(inp->inp_socket); 632 sowwakeup(inp->inp_socket); 633 } 634 635 void * 636 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v) 637 { 638 struct ip *ip = v; 639 struct udphdr *uh; 640 void (*notify)(struct inpcb *, int) = udp_notify; 641 int errno; 642 643 if (sa->sa_family != AF_INET || 644 sa->sa_len != sizeof(struct sockaddr_in)) 645 return NULL; 646 if ((unsigned)cmd >= PRC_NCMDS) 647 return NULL; 648 649 errno = inetctlerrmap[cmd]; 650 if (PRC_IS_REDIRECT(cmd)) { 651 notify = in_rtchange; 652 ip = NULL; 653 } else if (cmd == PRC_HOSTDEAD) { 654 ip = NULL; 655 } else if (errno == 0) { 656 return NULL; 657 } 658 659 if (ip) { 660 uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2)); 661 in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport, 662 ip->ip_src, uh->uh_sport, errno, notify); 663 /* XXX mapped address case */ 664 } else { 665 in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno, 666 notify); 667 } 668 669 return NULL; 670 } 671 672 int 673 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 674 { 675 int s; 676 int error = 0; 677 struct inpcb *inp; 678 int family; 679 int optval; 680 681 family = so->so_proto->pr_domain->dom_family; 682 683 s = splsoftnet(); 684 switch (family) { 685 #ifdef INET 686 case PF_INET: 687 if (sopt->sopt_level != IPPROTO_UDP) { 688 error = ip_ctloutput(op, so, sopt); 689 goto end; 690 } 691 break; 692 #endif 693 #ifdef INET6 694 case PF_INET6: 695 if (sopt->sopt_level != IPPROTO_UDP) { 696 error = ip6_ctloutput(op, so, sopt); 697 goto end; 698 } 699 break; 700 #endif 701 default: 702 error = EAFNOSUPPORT; 703 goto end; 704 } 705 706 707 switch (op) { 708 case PRCO_SETOPT: 709 inp = sotoinpcb(so); 710 711 switch (sopt->sopt_name) { 712 case UDP_ENCAP: 713 error = sockopt_getint(sopt, &optval); 714 if (error) 715 break; 716 717 switch(optval) { 718 case 0: 719 inp->inp_flags &= ~INP_ESPINUDP; 720 break; 721 722 case UDP_ENCAP_ESPINUDP: 723 inp->inp_flags |= INP_ESPINUDP; 724 break; 725 726 default: 727 error = EINVAL; 728 break; 729 } 730 break; 731 732 default: 733 error = ENOPROTOOPT; 734 break; 735 } 736 break; 737 738 default: 739 error = EINVAL; 740 break; 741 } 742 743 end: 744 splx(s); 745 return error; 746 } 747 748 int 749 udp_output(struct mbuf *m, struct inpcb *inp, struct mbuf *control, 750 struct lwp *l) 751 { 752 struct udpiphdr *ui; 753 struct route *ro; 754 struct ip_pktopts pktopts; 755 kauth_cred_t cred; 756 int len = m->m_pkthdr.len; 757 int error, flags = 0; 758 759 MCLAIM(m, &udp_tx_mowner); 760 761 /* 762 * Calculate data length and get a mbuf 763 * for UDP and IP headers. 764 */ 765 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 766 if (m == NULL) { 767 error = ENOBUFS; 768 goto release; 769 } 770 771 /* 772 * Compute the packet length of the IP header, and 773 * punt if the length looks bogus. 774 */ 775 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 776 error = EMSGSIZE; 777 goto release; 778 } 779 780 if (l == NULL) 781 cred = NULL; 782 else 783 cred = l->l_cred; 784 785 /* Setup IP outgoing packet options */ 786 memset(&pktopts, 0, sizeof(pktopts)); 787 error = ip_setpktopts(control, &pktopts, &flags, inp, cred); 788 if (error != 0) 789 goto release; 790 791 if (control != NULL) { 792 m_freem(control); 793 control = NULL; 794 } 795 796 /* 797 * Fill in mbuf with extended UDP header 798 * and addresses and length put into network format. 799 */ 800 ui = mtod(m, struct udpiphdr *); 801 ui->ui_pr = IPPROTO_UDP; 802 ui->ui_src = pktopts.ippo_laddr.sin_addr; 803 ui->ui_dst = inp->inp_faddr; 804 ui->ui_sport = inp->inp_lport; 805 ui->ui_dport = inp->inp_fport; 806 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 807 808 ro = &inp->inp_route; 809 810 /* 811 * Set up checksum and output datagram. 812 */ 813 if (udpcksum) { 814 /* 815 * XXX Cache pseudo-header checksum part for 816 * XXX "connected" UDP sockets. 817 */ 818 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 819 ui->ui_dst.s_addr, htons((u_int16_t)len + 820 sizeof(struct udphdr) + IPPROTO_UDP)); 821 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 822 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 823 } else 824 ui->ui_sum = 0; 825 826 ((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len); 827 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ 828 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ 829 UDP_STATINC(UDP_STAT_OPACKETS); 830 831 flags |= inp->inp_socket->so_options & (SO_DONTROUTE|SO_BROADCAST); 832 return ip_output(m, inp->inp_options, ro, flags, pktopts.ippo_imo, inp); 833 834 release: 835 if (control != NULL) 836 m_freem(control); 837 m_freem(m); 838 return error; 839 } 840 841 static int 842 udp_attach(struct socket *so, int proto) 843 { 844 struct inpcb *inp; 845 int error; 846 847 KASSERT(sotoinpcb(so) == NULL); 848 849 /* Assign the lock (must happen even if we will error out). */ 850 sosetlock(so); 851 852 #ifdef MBUFTRACE 853 so->so_mowner = &udp_mowner; 854 so->so_rcv.sb_mowner = &udp_rx_mowner; 855 so->so_snd.sb_mowner = &udp_tx_mowner; 856 #endif 857 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 858 error = soreserve(so, udp_sendspace, udp_recvspace); 859 if (error) { 860 return error; 861 } 862 } 863 864 error = in_pcballoc(so, &udbtable); 865 if (error) { 866 return error; 867 } 868 inp = sotoinpcb(so); 869 inp->inp_ip.ip_ttl = ip_defttl; 870 KASSERT(solocked(so)); 871 872 return error; 873 } 874 875 static void 876 udp_detach(struct socket *so) 877 { 878 struct inpcb *inp; 879 880 KASSERT(solocked(so)); 881 inp = sotoinpcb(so); 882 KASSERT(inp != NULL); 883 in_pcbdetach(inp); 884 } 885 886 static int 887 udp_accept(struct socket *so, struct sockaddr *nam) 888 { 889 KASSERT(solocked(so)); 890 891 panic("udp_accept"); 892 893 return EOPNOTSUPP; 894 } 895 896 static int 897 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l) 898 { 899 struct inpcb *inp = sotoinpcb(so); 900 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 901 int error = 0; 902 int s; 903 904 KASSERT(solocked(so)); 905 KASSERT(inp != NULL); 906 KASSERT(nam != NULL); 907 908 s = splsoftnet(); 909 error = in_pcbbind(inp, sin, l); 910 splx(s); 911 912 return error; 913 } 914 915 static int 916 udp_listen(struct socket *so, struct lwp *l) 917 { 918 KASSERT(solocked(so)); 919 920 return EOPNOTSUPP; 921 } 922 923 static int 924 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l) 925 { 926 struct inpcb *inp = sotoinpcb(so); 927 int error = 0; 928 int s; 929 930 KASSERT(solocked(so)); 931 KASSERT(inp != NULL); 932 KASSERT(nam != NULL); 933 934 s = splsoftnet(); 935 error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l); 936 if (! error) 937 soisconnected(so); 938 splx(s); 939 return error; 940 } 941 942 static int 943 udp_connect2(struct socket *so, struct socket *so2) 944 { 945 KASSERT(solocked(so)); 946 947 return EOPNOTSUPP; 948 } 949 950 static int 951 udp_disconnect(struct socket *so) 952 { 953 struct inpcb *inp = sotoinpcb(so); 954 int s; 955 956 KASSERT(solocked(so)); 957 KASSERT(inp != NULL); 958 959 s = splsoftnet(); 960 /*soisdisconnected(so);*/ 961 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 962 in_pcbdisconnect(inp); 963 inp->inp_laddr = zeroin_addr; /* XXX */ 964 in_pcbstate(inp, INP_BOUND); /* XXX */ 965 splx(s); 966 967 return 0; 968 } 969 970 static int 971 udp_shutdown(struct socket *so) 972 { 973 int s; 974 975 KASSERT(solocked(so)); 976 977 s = splsoftnet(); 978 socantsendmore(so); 979 splx(s); 980 981 return 0; 982 } 983 984 static int 985 udp_abort(struct socket *so) 986 { 987 KASSERT(solocked(so)); 988 989 panic("udp_abort"); 990 991 return EOPNOTSUPP; 992 } 993 994 static int 995 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp) 996 { 997 return in_control(so, cmd, nam, ifp); 998 } 999 1000 static int 1001 udp_stat(struct socket *so, struct stat *ub) 1002 { 1003 KASSERT(solocked(so)); 1004 1005 /* stat: don't bother with a blocksize. */ 1006 return 0; 1007 } 1008 1009 static int 1010 udp_peeraddr(struct socket *so, struct sockaddr *nam) 1011 { 1012 int s; 1013 1014 KASSERT(solocked(so)); 1015 KASSERT(sotoinpcb(so) != NULL); 1016 KASSERT(nam != NULL); 1017 1018 s = splsoftnet(); 1019 in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam); 1020 splx(s); 1021 1022 return 0; 1023 } 1024 1025 static int 1026 udp_sockaddr(struct socket *so, struct sockaddr *nam) 1027 { 1028 int s; 1029 1030 KASSERT(solocked(so)); 1031 KASSERT(sotoinpcb(so) != NULL); 1032 KASSERT(nam != NULL); 1033 1034 s = splsoftnet(); 1035 in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam); 1036 splx(s); 1037 1038 return 0; 1039 } 1040 1041 static int 1042 udp_rcvd(struct socket *so, int flags, struct lwp *l) 1043 { 1044 KASSERT(solocked(so)); 1045 1046 return EOPNOTSUPP; 1047 } 1048 1049 static int 1050 udp_recvoob(struct socket *so, struct mbuf *m, int flags) 1051 { 1052 KASSERT(solocked(so)); 1053 1054 return EOPNOTSUPP; 1055 } 1056 1057 static int 1058 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam, 1059 struct mbuf *control, struct lwp *l) 1060 { 1061 struct inpcb *inp = sotoinpcb(so); 1062 int error = 0; 1063 struct in_addr laddr; /* XXX */ 1064 int s; 1065 1066 KASSERT(solocked(so)); 1067 KASSERT(inp != NULL); 1068 KASSERT(m != NULL); 1069 1070 memset(&laddr, 0, sizeof laddr); 1071 1072 s = splsoftnet(); 1073 if (nam) { 1074 laddr = inp->inp_laddr; /* XXX */ 1075 if ((so->so_state & SS_ISCONNECTED) != 0) { 1076 error = EISCONN; 1077 goto die; 1078 } 1079 error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l); 1080 if (error) 1081 goto die; 1082 } else { 1083 if ((so->so_state & SS_ISCONNECTED) == 0) { 1084 error = ENOTCONN; 1085 goto die; 1086 } 1087 } 1088 error = udp_output(m, inp, control, l); 1089 m = NULL; 1090 control = NULL; 1091 if (nam) { 1092 in_pcbdisconnect(inp); 1093 inp->inp_laddr = laddr; /* XXX */ 1094 in_pcbstate(inp, INP_BOUND); /* XXX */ 1095 } 1096 die: 1097 if (m != NULL) 1098 m_freem(m); 1099 if (control != NULL) 1100 m_freem(control); 1101 1102 splx(s); 1103 return error; 1104 } 1105 1106 static int 1107 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control) 1108 { 1109 KASSERT(solocked(so)); 1110 1111 m_freem(m); 1112 m_freem(control); 1113 1114 return EOPNOTSUPP; 1115 } 1116 1117 static int 1118 udp_purgeif(struct socket *so, struct ifnet *ifp) 1119 { 1120 int s; 1121 1122 s = splsoftnet(); 1123 mutex_enter(softnet_lock); 1124 in_pcbpurgeif0(&udbtable, ifp); 1125 #ifdef NET_MPSAFE 1126 mutex_exit(softnet_lock); 1127 #endif 1128 in_purgeif(ifp); 1129 #ifdef NET_MPSAFE 1130 mutex_enter(softnet_lock); 1131 #endif 1132 in_pcbpurgeif(&udbtable, ifp); 1133 mutex_exit(softnet_lock); 1134 splx(s); 1135 1136 return 0; 1137 } 1138 1139 static int 1140 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) 1141 { 1142 1143 return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS)); 1144 } 1145 1146 /* 1147 * Sysctl for udp variables. 1148 */ 1149 static void 1150 sysctl_net_inet_udp_setup(struct sysctllog **clog) 1151 { 1152 1153 sysctl_createv(clog, 0, NULL, NULL, 1154 CTLFLAG_PERMANENT, 1155 CTLTYPE_NODE, "inet", NULL, 1156 NULL, 0, NULL, 0, 1157 CTL_NET, PF_INET, CTL_EOL); 1158 sysctl_createv(clog, 0, NULL, NULL, 1159 CTLFLAG_PERMANENT, 1160 CTLTYPE_NODE, "udp", 1161 SYSCTL_DESCR("UDPv4 related settings"), 1162 NULL, 0, NULL, 0, 1163 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL); 1164 1165 sysctl_createv(clog, 0, NULL, NULL, 1166 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1167 CTLTYPE_INT, "checksum", 1168 SYSCTL_DESCR("Compute UDP checksums"), 1169 NULL, 0, &udpcksum, 0, 1170 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM, 1171 CTL_EOL); 1172 sysctl_createv(clog, 0, NULL, NULL, 1173 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1174 CTLTYPE_INT, "sendspace", 1175 SYSCTL_DESCR("Default UDP send buffer size"), 1176 NULL, 0, &udp_sendspace, 0, 1177 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE, 1178 CTL_EOL); 1179 sysctl_createv(clog, 0, NULL, NULL, 1180 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1181 CTLTYPE_INT, "recvspace", 1182 SYSCTL_DESCR("Default UDP receive buffer size"), 1183 NULL, 0, &udp_recvspace, 0, 1184 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE, 1185 CTL_EOL); 1186 sysctl_createv(clog, 0, NULL, NULL, 1187 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1188 CTLTYPE_INT, "do_loopback_cksum", 1189 SYSCTL_DESCR("Perform UDP checksum on loopback"), 1190 NULL, 0, &udp_do_loopback_cksum, 0, 1191 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM, 1192 CTL_EOL); 1193 sysctl_createv(clog, 0, NULL, NULL, 1194 CTLFLAG_PERMANENT, 1195 CTLTYPE_STRUCT, "pcblist", 1196 SYSCTL_DESCR("UDP protocol control block list"), 1197 sysctl_inpcblist, 0, &udbtable, 0, 1198 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE, 1199 CTL_EOL); 1200 sysctl_createv(clog, 0, NULL, NULL, 1201 CTLFLAG_PERMANENT, 1202 CTLTYPE_STRUCT, "stats", 1203 SYSCTL_DESCR("UDP statistics"), 1204 sysctl_net_inet_udp_stats, 0, NULL, 0, 1205 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS, 1206 CTL_EOL); 1207 } 1208 #endif 1209 1210 void 1211 udp_statinc(u_int stat) 1212 { 1213 1214 KASSERT(stat < UDP_NSTATS); 1215 UDP_STATINC(stat); 1216 } 1217 1218 #if defined(INET) && defined(IPSEC) 1219 /* 1220 * Handle ESP-in-UDP packets (RFC3948). 1221 * 1222 * We need to distinguish between ESP packets and IKE packets. We do so by 1223 * looking at the Non-ESP marker. If IKE, we process the UDP packet as usual. 1224 * Otherwise, ESP, we invoke IPsec. 1225 * 1226 * Returns: 1227 * 1 if the packet was processed 1228 * 0 if normal UDP processing should take place 1229 * -1 if an error occurred and m was freed 1230 */ 1231 static int 1232 udp4_espinudp(struct mbuf **mp, int off, struct socket *so) 1233 { 1234 const size_t skip = sizeof(struct udphdr); 1235 size_t len; 1236 uint8_t *data; 1237 size_t minlen; 1238 size_t iphdrlen; 1239 struct ip *ip; 1240 struct m_tag *tag; 1241 struct udphdr *udphdr; 1242 u_int16_t sport, dport; 1243 struct mbuf *m = *mp; 1244 uint32_t *marker; 1245 1246 minlen = off + sizeof(struct esp); 1247 if (minlen > m->m_pkthdr.len) 1248 minlen = m->m_pkthdr.len; 1249 1250 if (m->m_len < minlen) { 1251 if ((*mp = m_pullup(m, minlen)) == NULL) { 1252 return -1; 1253 } 1254 m = *mp; 1255 } 1256 1257 len = m->m_len - off; 1258 data = mtod(m, uint8_t *) + off; 1259 1260 /* Ignore keepalive packets. */ 1261 if ((len == 1) && (*data == 0xff)) { 1262 m_freem(m); 1263 *mp = NULL; /* avoid any further processing by caller */ 1264 return 1; 1265 } 1266 1267 /* Handle Non-ESP marker (32bit). If zero, then IKE. */ 1268 marker = (uint32_t *)data; 1269 if (len <= sizeof(uint32_t)) 1270 return 0; 1271 if (marker[0] == 0) 1272 return 0; 1273 1274 /* 1275 * Get the UDP ports. They are handled in network order 1276 * everywhere in the IPSEC_NAT_T code. 1277 */ 1278 udphdr = (struct udphdr *)((char *)data - skip); 1279 sport = udphdr->uh_sport; 1280 dport = udphdr->uh_dport; 1281 1282 /* 1283 * Remove the UDP header, plus a possible marker. IP header 1284 * length is iphdrlen. 1285 * 1286 * Before: 1287 * <--- off ---> 1288 * +----+------+-----+ 1289 * | IP | UDP | ESP | 1290 * +----+------+-----+ 1291 * <-skip-> 1292 * After: 1293 * +----+-----+ 1294 * | IP | ESP | 1295 * +----+-----+ 1296 * <-skip-> 1297 */ 1298 iphdrlen = off - sizeof(struct udphdr); 1299 memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen); 1300 m_adj(m, skip); 1301 1302 ip = mtod(m, struct ip *); 1303 ip->ip_len = htons(ntohs(ip->ip_len) - skip); 1304 ip->ip_p = IPPROTO_ESP; 1305 1306 /* 1307 * We have modified the packet - it is now ESP, so we should not 1308 * return to UDP processing. 1309 * 1310 * Add a PACKET_TAG_IPSEC_NAT_T_PORTS tag to remember the source 1311 * UDP port. This is required if we want to select the right SPD 1312 * for multiple hosts behind same NAT. 1313 */ 1314 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, 1315 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) { 1316 m_freem(m); 1317 return -1; 1318 } 1319 ((u_int16_t *)(tag + 1))[0] = sport; 1320 ((u_int16_t *)(tag + 1))[1] = dport; 1321 m_tag_prepend(m, tag); 1322 1323 if (ipsec_used) 1324 ipsec4_common_input(m, iphdrlen, IPPROTO_ESP); 1325 else 1326 m_freem(m); 1327 1328 /* We handled it, it shouldn't be handled by UDP */ 1329 *mp = NULL; /* avoid free by caller ... */ 1330 return 1; 1331 } 1332 #endif 1333 1334 PR_WRAP_USRREQS(udp) 1335 #define udp_attach udp_attach_wrapper 1336 #define udp_detach udp_detach_wrapper 1337 #define udp_accept udp_accept_wrapper 1338 #define udp_bind udp_bind_wrapper 1339 #define udp_listen udp_listen_wrapper 1340 #define udp_connect udp_connect_wrapper 1341 #define udp_connect2 udp_connect2_wrapper 1342 #define udp_disconnect udp_disconnect_wrapper 1343 #define udp_shutdown udp_shutdown_wrapper 1344 #define udp_abort udp_abort_wrapper 1345 #define udp_ioctl udp_ioctl_wrapper 1346 #define udp_stat udp_stat_wrapper 1347 #define udp_peeraddr udp_peeraddr_wrapper 1348 #define udp_sockaddr udp_sockaddr_wrapper 1349 #define udp_rcvd udp_rcvd_wrapper 1350 #define udp_recvoob udp_recvoob_wrapper 1351 #define udp_send udp_send_wrapper 1352 #define udp_sendoob udp_sendoob_wrapper 1353 #define udp_purgeif udp_purgeif_wrapper 1354 1355 const struct pr_usrreqs udp_usrreqs = { 1356 .pr_attach = udp_attach, 1357 .pr_detach = udp_detach, 1358 .pr_accept = udp_accept, 1359 .pr_bind = udp_bind, 1360 .pr_listen = udp_listen, 1361 .pr_connect = udp_connect, 1362 .pr_connect2 = udp_connect2, 1363 .pr_disconnect = udp_disconnect, 1364 .pr_shutdown = udp_shutdown, 1365 .pr_abort = udp_abort, 1366 .pr_ioctl = udp_ioctl, 1367 .pr_stat = udp_stat, 1368 .pr_peeraddr = udp_peeraddr, 1369 .pr_sockaddr = udp_sockaddr, 1370 .pr_rcvd = udp_rcvd, 1371 .pr_recvoob = udp_recvoob, 1372 .pr_send = udp_send, 1373 .pr_sendoob = udp_sendoob, 1374 .pr_purgeif = udp_purgeif, 1375 }; 1376