1 /* $NetBSD: udp_usrreq.c,v 1.264 2022/11/04 09:00:58 ozaki-r 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 1.264 2022/11/04 09:00:58 ozaki-r 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); 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 inpcb_init(&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, int off, int proto) 312 { 313 struct sockaddr_in src, dst; 314 struct ip *ip; 315 struct udphdr *uh; 316 int iphlen = off; 317 int len; 318 int n; 319 u_int16_t ip_len; 320 321 MCLAIM(m, &udp_rx_mowner); 322 UDP_STATINC(UDP_STAT_IPACKETS); 323 324 /* 325 * Get IP and UDP header together in first mbuf. 326 */ 327 ip = mtod(m, struct ip *); 328 M_REGION_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 329 if (uh == NULL) { 330 UDP_STATINC(UDP_STAT_HDROPS); 331 return; 332 } 333 334 /* 335 * Enforce alignment requirements that are violated in 336 * some cases, see kern/50766 for details. 337 */ 338 if (ACCESSIBLE_POINTER(uh, struct udphdr) == 0) { 339 m = m_copyup(m, iphlen + sizeof(struct udphdr), 0); 340 if (m == NULL) { 341 UDP_STATINC(UDP_STAT_HDROPS); 342 return; 343 } 344 ip = mtod(m, struct ip *); 345 uh = (struct udphdr *)(mtod(m, char *) + iphlen); 346 } 347 KASSERT(ACCESSIBLE_POINTER(uh, struct udphdr)); 348 349 /* destination port of 0 is illegal, based on RFC768. */ 350 if (uh->uh_dport == 0) 351 goto bad; 352 353 /* 354 * Make mbuf data length reflect UDP length. 355 * If not enough data to reflect UDP length, drop. 356 */ 357 ip_len = ntohs(ip->ip_len); 358 len = ntohs((u_int16_t)uh->uh_ulen); 359 if (len < sizeof(struct udphdr)) { 360 UDP_STATINC(UDP_STAT_BADLEN); 361 goto bad; 362 } 363 if (ip_len != iphlen + len) { 364 if (ip_len < iphlen + len) { 365 UDP_STATINC(UDP_STAT_BADLEN); 366 goto bad; 367 } 368 m_adj(m, iphlen + len - ip_len); 369 } 370 371 /* 372 * Checksum extended UDP header and data. 373 */ 374 if (udp4_input_checksum(m, uh, iphlen, len)) 375 goto badcsum; 376 377 /* construct source and dst sockaddrs. */ 378 sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport); 379 sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport); 380 381 if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) { 382 UDP_STATINC(UDP_STAT_HDROPS); 383 return; 384 } 385 if (m == NULL) { 386 /* 387 * packet has been processed by ESP stuff - 388 * e.g. dropped NAT-T-keep-alive-packet ... 389 */ 390 return; 391 } 392 393 ip = mtod(m, struct ip *); 394 M_REGION_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); 395 if (uh == NULL) { 396 UDP_STATINC(UDP_STAT_HDROPS); 397 return; 398 } 399 /* XXX Re-enforce alignment? */ 400 401 #ifdef INET6 402 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { 403 struct sockaddr_in6 src6, dst6; 404 405 memset(&src6, 0, sizeof(src6)); 406 src6.sin6_family = AF_INET6; 407 src6.sin6_len = sizeof(struct sockaddr_in6); 408 in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr); 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 in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr); 414 dst6.sin6_port = uh->uh_dport; 415 416 n += udp6_realinput(AF_INET, &src6, &dst6, &m, iphlen); 417 } 418 #endif 419 420 if (n == 0) { 421 if (m->m_flags & (M_BCAST | M_MCAST)) { 422 UDP_STATINC(UDP_STAT_NOPORTBCAST); 423 goto bad; 424 } 425 UDP_STATINC(UDP_STAT_NOPORT); 426 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 427 m = NULL; 428 } 429 430 bad: 431 if (m) 432 m_freem(m); 433 return; 434 435 badcsum: 436 m_freem(m); 437 } 438 #endif 439 440 #ifdef INET 441 static void 442 udp4_sendup(struct mbuf *m, int off /* offset of data portion */, 443 struct sockaddr *src, struct socket *so) 444 { 445 struct mbuf *opts = NULL; 446 struct mbuf *n; 447 struct inpcb *inp; 448 449 KASSERT(so != NULL); 450 KASSERT(so->so_proto->pr_domain->dom_family == AF_INET); 451 inp = sotoinpcb(so); 452 KASSERT(inp != NULL); 453 454 #if defined(IPSEC) 455 if (ipsec_used && ipsec_in_reject(m, inp)) { 456 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) 457 icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT, 458 0, 0); 459 return; 460 } 461 #endif 462 463 if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { 464 if (inp->inp_flags & INP_CONTROLOPTS || 465 SOOPT_TIMESTAMP(so->so_options)) { 466 struct ip *ip = mtod(n, struct ip *); 467 ip_savecontrol(inp, &opts, ip, n); 468 } 469 470 m_adj(n, off); 471 if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) { 472 m_freem(n); 473 if (opts) 474 m_freem(opts); 475 UDP_STATINC(UDP_STAT_FULLSOCK); 476 soroverflow(so); 477 } else 478 sorwakeup(so); 479 } 480 } 481 #endif 482 483 #ifdef INET 484 static int 485 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst, 486 struct mbuf **mp, int off /* offset of udphdr */) 487 { 488 u_int16_t *sport, *dport; 489 int rcvcnt; 490 struct in_addr *src4, *dst4; 491 struct inpcb *inp; 492 struct mbuf *m = *mp; 493 494 rcvcnt = 0; 495 off += sizeof(struct udphdr); /* now, offset of payload */ 496 497 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 498 goto bad; 499 500 src4 = &src->sin_addr; 501 sport = &src->sin_port; 502 dst4 = &dst->sin_addr; 503 dport = &dst->sin_port; 504 505 if (IN_MULTICAST(dst4->s_addr) || 506 in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) { 507 /* 508 * Deliver a multicast or broadcast datagram to *all* sockets 509 * for which the local and remote addresses and ports match 510 * those of the incoming datagram. This allows more than 511 * one process to receive multi/broadcasts on the same port. 512 * (This really ought to be done for unicast datagrams as 513 * well, but that would cause problems with existing 514 * applications that open both address-specific sockets and 515 * a wildcard socket listening to the same port -- they would 516 * end up receiving duplicates of every unicast datagram. 517 * Those applications open the multiple sockets to overcome an 518 * inadequacy of the UDP socket interface, but for backwards 519 * compatibility we avoid the problem here rather than 520 * fixing the interface. Maybe 4.5BSD will remedy this?) 521 */ 522 523 /* 524 * KAME note: traditionally we dropped udpiphdr from mbuf here. 525 * we need udpiphdr for IPsec processing so we do that later. 526 */ 527 /* 528 * Locate pcb(s) for datagram. 529 */ 530 TAILQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) { 531 if (inp->inp_af != AF_INET) 532 continue; 533 534 if (inp->inp_lport != *dport) 535 continue; 536 if (!in_nullhost(in4p_laddr(inp))) { 537 if (!in_hosteq(in4p_laddr(inp), *dst4)) 538 continue; 539 } 540 if (!in_nullhost(in4p_faddr(inp))) { 541 if (!in_hosteq(in4p_faddr(inp), *src4) || 542 inp->inp_fport != *sport) 543 continue; 544 } 545 546 udp4_sendup(m, off, (struct sockaddr *)src, 547 inp->inp_socket); 548 rcvcnt++; 549 550 /* 551 * Don't look for additional matches if this one does 552 * not have either the SO_REUSEPORT or SO_REUSEADDR 553 * socket options set. This heuristic avoids searching 554 * through all pcbs in the common case of a non-shared 555 * port. It assumes that an application will never 556 * clear these options after setting them. 557 */ 558 if ((inp->inp_socket->so_options & 559 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 560 break; 561 } 562 } else { 563 /* 564 * Locate pcb for datagram. 565 */ 566 inp = inpcb_lookup(&udbtable, *src4, *sport, *dst4, 567 *dport, 0); 568 if (inp == 0) { 569 UDP_STATINC(UDP_STAT_PCBHASHMISS); 570 inp = inpcb_lookup_bound(&udbtable, *dst4, *dport); 571 if (inp == 0) 572 return rcvcnt; 573 } 574 575 #ifdef IPSEC 576 /* Handle ESP over UDP */ 577 if (inp->inp_flags & INP_ESPINUDP) { 578 switch (udp4_espinudp(mp, off)) { 579 case -1: /* Error, m was freed */ 580 rcvcnt = -1; 581 goto bad; 582 583 case 1: /* ESP over UDP */ 584 rcvcnt++; 585 goto bad; 586 587 case 0: /* plain UDP */ 588 default: /* Unexpected */ 589 /* 590 * Normal UDP processing will take place, 591 * m may have changed. 592 */ 593 m = *mp; 594 break; 595 } 596 } 597 #endif 598 if (inp->inp_overudp_cb != NULL) { 599 int ret; 600 ret = inp->inp_overudp_cb(mp, off, inp->inp_socket, 601 sintosa(src), inp->inp_overudp_arg); 602 switch (ret) { 603 case -1: /* Error, m was freed */ 604 rcvcnt = -1; 605 goto bad; 606 607 case 1: /* Foo over UDP */ 608 KASSERT(*mp == NULL); 609 rcvcnt++; 610 goto bad; 611 612 case 0: /* plain UDP */ 613 default: /* Unexpected */ 614 /* 615 * Normal UDP processing will take place, 616 * m may have changed. 617 */ 618 m = *mp; 619 break; 620 } 621 } 622 623 /* 624 * Check the minimum TTL for socket. 625 */ 626 if (mtod(m, struct ip *)->ip_ttl < in4p_ip_minttl(inp)) 627 goto bad; 628 629 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 630 rcvcnt++; 631 } 632 633 bad: 634 return rcvcnt; 635 } 636 #endif 637 638 #ifdef INET 639 /* 640 * Notify a udp user of an asynchronous error; 641 * just wake up so that he can collect error status. 642 */ 643 static void 644 udp_notify(struct inpcb *inp, int errno) 645 { 646 inp->inp_socket->so_error = errno; 647 sorwakeup(inp->inp_socket); 648 sowwakeup(inp->inp_socket); 649 } 650 651 void * 652 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v) 653 { 654 struct ip *ip = v; 655 struct udphdr *uh; 656 void (*notify)(struct inpcb *, int) = udp_notify; 657 int errno; 658 659 if (sa->sa_family != AF_INET || 660 sa->sa_len != sizeof(struct sockaddr_in)) 661 return NULL; 662 if ((unsigned)cmd >= PRC_NCMDS) 663 return NULL; 664 665 errno = inetctlerrmap[cmd]; 666 if (PRC_IS_REDIRECT(cmd)) { 667 notify = inpcb_rtchange; 668 ip = NULL; 669 } else if (cmd == PRC_HOSTDEAD) { 670 ip = NULL; 671 } else if (errno == 0) { 672 return NULL; 673 } 674 675 if (ip) { 676 uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2)); 677 inpcb_notify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport, 678 ip->ip_src, uh->uh_sport, errno, notify); 679 /* XXX mapped address case */ 680 } else { 681 inpcb_notifyall(&udbtable, satocsin(sa)->sin_addr, errno, 682 notify); 683 } 684 685 return NULL; 686 } 687 688 int 689 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 690 { 691 int s; 692 int error = 0; 693 struct inpcb *inp; 694 int family; 695 int optval; 696 697 family = so->so_proto->pr_domain->dom_family; 698 699 s = splsoftnet(); 700 switch (family) { 701 #ifdef INET 702 case PF_INET: 703 if (sopt->sopt_level != IPPROTO_UDP) { 704 error = ip_ctloutput(op, so, sopt); 705 goto end; 706 } 707 break; 708 #endif 709 #ifdef INET6 710 case PF_INET6: 711 if (sopt->sopt_level != IPPROTO_UDP) { 712 error = ip6_ctloutput(op, so, sopt); 713 goto end; 714 } 715 break; 716 #endif 717 default: 718 error = EAFNOSUPPORT; 719 goto end; 720 } 721 722 723 switch (op) { 724 case PRCO_SETOPT: 725 inp = sotoinpcb(so); 726 727 switch (sopt->sopt_name) { 728 case UDP_ENCAP: 729 error = sockopt_getint(sopt, &optval); 730 if (error) 731 break; 732 733 switch(optval) { 734 case 0: 735 inp->inp_flags &= ~INP_ESPINUDP; 736 break; 737 738 case UDP_ENCAP_ESPINUDP: 739 inp->inp_flags |= INP_ESPINUDP; 740 break; 741 742 default: 743 error = EINVAL; 744 break; 745 } 746 break; 747 748 default: 749 error = ENOPROTOOPT; 750 break; 751 } 752 break; 753 754 default: 755 error = EINVAL; 756 break; 757 } 758 759 end: 760 splx(s); 761 return error; 762 } 763 764 int 765 udp_output(struct mbuf *m, struct inpcb *inp, struct mbuf *control, 766 struct lwp *l) 767 { 768 struct udpiphdr *ui; 769 struct route *ro; 770 struct ip_pktopts pktopts; 771 kauth_cred_t cred; 772 int len = m->m_pkthdr.len; 773 int error, flags = 0; 774 775 MCLAIM(m, &udp_tx_mowner); 776 777 /* 778 * Calculate data length and get a mbuf 779 * for UDP and IP headers. 780 */ 781 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 782 if (m == NULL) { 783 error = ENOBUFS; 784 goto release; 785 } 786 787 /* 788 * Compute the packet length of the IP header, and 789 * punt if the length looks bogus. 790 */ 791 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 792 error = EMSGSIZE; 793 goto release; 794 } 795 796 if (l == NULL) 797 cred = NULL; 798 else 799 cred = l->l_cred; 800 801 /* Setup IP outgoing packet options */ 802 memset(&pktopts, 0, sizeof(pktopts)); 803 error = ip_setpktopts(control, &pktopts, &flags, inp, cred); 804 if (error != 0) 805 goto release; 806 807 if (control != NULL) { 808 m_freem(control); 809 control = NULL; 810 } 811 812 /* 813 * Fill in mbuf with extended UDP header 814 * and addresses and length put into network format. 815 */ 816 ui = mtod(m, struct udpiphdr *); 817 ui->ui_pr = IPPROTO_UDP; 818 ui->ui_src = pktopts.ippo_laddr.sin_addr; 819 ui->ui_dst = in4p_faddr(inp); 820 ui->ui_sport = inp->inp_lport; 821 ui->ui_dport = inp->inp_fport; 822 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 823 824 ro = &inp->inp_route; 825 826 /* 827 * Set up checksum and output datagram. 828 */ 829 if (udpcksum) { 830 /* 831 * XXX Cache pseudo-header checksum part for 832 * XXX "connected" UDP sockets. 833 */ 834 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 835 ui->ui_dst.s_addr, htons((u_int16_t)len + 836 sizeof(struct udphdr) + IPPROTO_UDP)); 837 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 838 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 839 } else 840 ui->ui_sum = 0; 841 842 ((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len); 843 ((struct ip *)ui)->ip_ttl = in4p_ip(inp).ip_ttl; /* XXX */ 844 ((struct ip *)ui)->ip_tos = in4p_ip(inp).ip_tos; /* XXX */ 845 UDP_STATINC(UDP_STAT_OPACKETS); 846 847 flags |= inp->inp_socket->so_options & (SO_DONTROUTE|SO_BROADCAST); 848 return ip_output(m, inp->inp_options, ro, flags, pktopts.ippo_imo, inp); 849 850 release: 851 if (control != NULL) 852 m_freem(control); 853 m_freem(m); 854 return error; 855 } 856 857 static int 858 udp_attach(struct socket *so, int proto) 859 { 860 struct inpcb *inp; 861 int error; 862 863 KASSERT(sotoinpcb(so) == NULL); 864 865 /* Assign the lock (must happen even if we will error out). */ 866 sosetlock(so); 867 868 #ifdef MBUFTRACE 869 so->so_mowner = &udp_mowner; 870 so->so_rcv.sb_mowner = &udp_rx_mowner; 871 so->so_snd.sb_mowner = &udp_tx_mowner; 872 #endif 873 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 874 error = soreserve(so, udp_sendspace, udp_recvspace); 875 if (error) { 876 return error; 877 } 878 } 879 880 error = inpcb_create(so, &udbtable); 881 if (error) { 882 return error; 883 } 884 inp = sotoinpcb(so); 885 in4p_ip(inp).ip_ttl = ip_defttl; 886 KASSERT(solocked(so)); 887 888 return error; 889 } 890 891 static void 892 udp_detach(struct socket *so) 893 { 894 struct inpcb *inp; 895 896 KASSERT(solocked(so)); 897 inp = sotoinpcb(so); 898 KASSERT(inp != NULL); 899 inpcb_destroy(inp); 900 } 901 902 static int 903 udp_accept(struct socket *so, struct sockaddr *nam) 904 { 905 KASSERT(solocked(so)); 906 907 panic("udp_accept"); 908 909 return EOPNOTSUPP; 910 } 911 912 static int 913 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l) 914 { 915 struct inpcb *inp = sotoinpcb(so); 916 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 917 int error = 0; 918 int s; 919 920 KASSERT(solocked(so)); 921 KASSERT(inp != NULL); 922 KASSERT(nam != NULL); 923 924 s = splsoftnet(); 925 error = inpcb_bind(inp, sin, l); 926 splx(s); 927 928 return error; 929 } 930 931 static int 932 udp_listen(struct socket *so, struct lwp *l) 933 { 934 KASSERT(solocked(so)); 935 936 return EOPNOTSUPP; 937 } 938 939 static int 940 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l) 941 { 942 struct inpcb *inp = sotoinpcb(so); 943 int error = 0; 944 int s; 945 946 KASSERT(solocked(so)); 947 KASSERT(inp != NULL); 948 KASSERT(nam != NULL); 949 950 s = splsoftnet(); 951 error = inpcb_connect(inp, (struct sockaddr_in *)nam, l); 952 if (! error) 953 soisconnected(so); 954 splx(s); 955 return error; 956 } 957 958 static int 959 udp_connect2(struct socket *so, struct socket *so2) 960 { 961 KASSERT(solocked(so)); 962 963 return EOPNOTSUPP; 964 } 965 966 static int 967 udp_disconnect(struct socket *so) 968 { 969 struct inpcb *inp = sotoinpcb(so); 970 int s; 971 972 KASSERT(solocked(so)); 973 KASSERT(inp != NULL); 974 975 s = splsoftnet(); 976 /*soisdisconnected(so);*/ 977 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 978 inpcb_disconnect(inp); 979 in4p_laddr(inp) = zeroin_addr; /* XXX */ 980 inpcb_set_state(inp, INP_BOUND); /* XXX */ 981 splx(s); 982 983 return 0; 984 } 985 986 static int 987 udp_shutdown(struct socket *so) 988 { 989 int s; 990 991 KASSERT(solocked(so)); 992 993 s = splsoftnet(); 994 socantsendmore(so); 995 splx(s); 996 997 return 0; 998 } 999 1000 static int 1001 udp_abort(struct socket *so) 1002 { 1003 KASSERT(solocked(so)); 1004 1005 panic("udp_abort"); 1006 1007 return EOPNOTSUPP; 1008 } 1009 1010 static int 1011 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp) 1012 { 1013 return in_control(so, cmd, nam, ifp); 1014 } 1015 1016 static int 1017 udp_stat(struct socket *so, struct stat *ub) 1018 { 1019 KASSERT(solocked(so)); 1020 1021 /* stat: don't bother with a blocksize. */ 1022 return 0; 1023 } 1024 1025 static int 1026 udp_peeraddr(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 inpcb_fetch_peeraddr(sotoinpcb(so), (struct sockaddr_in *)nam); 1036 splx(s); 1037 1038 return 0; 1039 } 1040 1041 static int 1042 udp_sockaddr(struct socket *so, struct sockaddr *nam) 1043 { 1044 int s; 1045 1046 KASSERT(solocked(so)); 1047 KASSERT(sotoinpcb(so) != NULL); 1048 KASSERT(nam != NULL); 1049 1050 s = splsoftnet(); 1051 inpcb_fetch_sockaddr(sotoinpcb(so), (struct sockaddr_in *)nam); 1052 splx(s); 1053 1054 return 0; 1055 } 1056 1057 static int 1058 udp_rcvd(struct socket *so, int flags, struct lwp *l) 1059 { 1060 KASSERT(solocked(so)); 1061 1062 return EOPNOTSUPP; 1063 } 1064 1065 static int 1066 udp_recvoob(struct socket *so, struct mbuf *m, int flags) 1067 { 1068 KASSERT(solocked(so)); 1069 1070 return EOPNOTSUPP; 1071 } 1072 1073 int 1074 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam, 1075 struct mbuf *control, struct lwp *l) 1076 { 1077 struct inpcb *inp = sotoinpcb(so); 1078 int error = 0; 1079 struct in_addr laddr; /* XXX */ 1080 int s; 1081 1082 KASSERT(solocked(so)); 1083 KASSERT(inp != NULL); 1084 KASSERT(m != NULL); 1085 1086 memset(&laddr, 0, sizeof laddr); 1087 1088 s = splsoftnet(); 1089 if (nam) { 1090 laddr = in4p_laddr(inp); /* XXX */ 1091 if ((so->so_state & SS_ISCONNECTED) != 0) { 1092 error = EISCONN; 1093 goto die; 1094 } 1095 error = inpcb_connect(inp, (struct sockaddr_in *)nam, l); 1096 if (error) 1097 goto die; 1098 } else { 1099 if ((so->so_state & SS_ISCONNECTED) == 0) { 1100 error = ENOTCONN; 1101 goto die; 1102 } 1103 } 1104 error = udp_output(m, inp, control, l); 1105 m = NULL; 1106 control = NULL; 1107 if (nam) { 1108 inpcb_disconnect(inp); 1109 in4p_laddr(inp) = laddr; /* XXX */ 1110 inpcb_set_state(inp, INP_BOUND); /* XXX */ 1111 } 1112 die: 1113 if (m != NULL) 1114 m_freem(m); 1115 if (control != NULL) 1116 m_freem(control); 1117 1118 splx(s); 1119 return error; 1120 } 1121 1122 static int 1123 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control) 1124 { 1125 KASSERT(solocked(so)); 1126 1127 m_freem(m); 1128 m_freem(control); 1129 1130 return EOPNOTSUPP; 1131 } 1132 1133 static int 1134 udp_purgeif(struct socket *so, struct ifnet *ifp) 1135 { 1136 int s; 1137 1138 s = splsoftnet(); 1139 mutex_enter(softnet_lock); 1140 inpcb_purgeif0(&udbtable, ifp); 1141 #ifdef NET_MPSAFE 1142 mutex_exit(softnet_lock); 1143 #endif 1144 in_purgeif(ifp); 1145 #ifdef NET_MPSAFE 1146 mutex_enter(softnet_lock); 1147 #endif 1148 inpcb_purgeif(&udbtable, ifp); 1149 mutex_exit(softnet_lock); 1150 splx(s); 1151 1152 return 0; 1153 } 1154 1155 static int 1156 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) 1157 { 1158 1159 return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS)); 1160 } 1161 1162 /* 1163 * Sysctl for udp variables. 1164 */ 1165 static void 1166 sysctl_net_inet_udp_setup(struct sysctllog **clog) 1167 { 1168 1169 sysctl_createv(clog, 0, NULL, NULL, 1170 CTLFLAG_PERMANENT, 1171 CTLTYPE_NODE, "inet", NULL, 1172 NULL, 0, NULL, 0, 1173 CTL_NET, PF_INET, CTL_EOL); 1174 sysctl_createv(clog, 0, NULL, NULL, 1175 CTLFLAG_PERMANENT, 1176 CTLTYPE_NODE, "udp", 1177 SYSCTL_DESCR("UDPv4 related settings"), 1178 NULL, 0, NULL, 0, 1179 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL); 1180 1181 sysctl_createv(clog, 0, NULL, NULL, 1182 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1183 CTLTYPE_INT, "checksum", 1184 SYSCTL_DESCR("Compute UDP checksums"), 1185 NULL, 0, &udpcksum, 0, 1186 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM, 1187 CTL_EOL); 1188 sysctl_createv(clog, 0, NULL, NULL, 1189 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1190 CTLTYPE_INT, "sendspace", 1191 SYSCTL_DESCR("Default UDP send buffer size"), 1192 NULL, 0, &udp_sendspace, 0, 1193 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE, 1194 CTL_EOL); 1195 sysctl_createv(clog, 0, NULL, NULL, 1196 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1197 CTLTYPE_INT, "recvspace", 1198 SYSCTL_DESCR("Default UDP receive buffer size"), 1199 NULL, 0, &udp_recvspace, 0, 1200 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE, 1201 CTL_EOL); 1202 sysctl_createv(clog, 0, NULL, NULL, 1203 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1204 CTLTYPE_INT, "do_loopback_cksum", 1205 SYSCTL_DESCR("Perform UDP checksum on loopback"), 1206 NULL, 0, &udp_do_loopback_cksum, 0, 1207 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM, 1208 CTL_EOL); 1209 sysctl_createv(clog, 0, NULL, NULL, 1210 CTLFLAG_PERMANENT, 1211 CTLTYPE_STRUCT, "pcblist", 1212 SYSCTL_DESCR("UDP protocol control block list"), 1213 sysctl_inpcblist, 0, &udbtable, 0, 1214 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE, 1215 CTL_EOL); 1216 sysctl_createv(clog, 0, NULL, NULL, 1217 CTLFLAG_PERMANENT, 1218 CTLTYPE_STRUCT, "stats", 1219 SYSCTL_DESCR("UDP statistics"), 1220 sysctl_net_inet_udp_stats, 0, NULL, 0, 1221 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS, 1222 CTL_EOL); 1223 } 1224 #endif 1225 1226 void 1227 udp_statinc(u_int stat) 1228 { 1229 1230 KASSERT(stat < UDP_NSTATS); 1231 UDP_STATINC(stat); 1232 } 1233 1234 #if defined(INET) && defined(IPSEC) 1235 /* 1236 * Handle ESP-in-UDP packets (RFC3948). 1237 * 1238 * We need to distinguish between ESP packets and IKE packets. We do so by 1239 * looking at the Non-ESP marker. If IKE, we process the UDP packet as usual. 1240 * Otherwise, ESP, we invoke IPsec. 1241 * 1242 * Returns: 1243 * 1 if the packet was processed 1244 * 0 if normal UDP processing should take place 1245 * -1 if an error occurred and m was freed 1246 */ 1247 static int 1248 udp4_espinudp(struct mbuf **mp, int off) 1249 { 1250 const size_t skip = sizeof(struct udphdr); 1251 size_t len; 1252 uint8_t *data; 1253 size_t minlen; 1254 size_t iphdrlen; 1255 struct ip *ip; 1256 struct m_tag *tag; 1257 struct udphdr *udphdr; 1258 u_int16_t sport, dport; 1259 struct mbuf *m = *mp; 1260 uint32_t *marker; 1261 1262 minlen = off + sizeof(struct esp); 1263 if (minlen > m->m_pkthdr.len) 1264 minlen = m->m_pkthdr.len; 1265 1266 if (m->m_len < minlen) { 1267 if ((*mp = m_pullup(m, minlen)) == NULL) { 1268 return -1; 1269 } 1270 m = *mp; 1271 } 1272 1273 len = m->m_len - off; 1274 data = mtod(m, uint8_t *) + off; 1275 1276 /* Ignore keepalive packets. */ 1277 if ((len == 1) && (*data == 0xff)) { 1278 m_freem(m); 1279 *mp = NULL; /* avoid any further processing by caller */ 1280 return 1; 1281 } 1282 1283 /* Handle Non-ESP marker (32bit). If zero, then IKE. */ 1284 marker = (uint32_t *)data; 1285 if (len <= sizeof(uint32_t)) 1286 return 0; 1287 if (marker[0] == 0) 1288 return 0; 1289 1290 /* 1291 * Get the UDP ports. They are handled in network order 1292 * everywhere in the IPSEC_NAT_T code. 1293 */ 1294 udphdr = (struct udphdr *)((char *)data - skip); 1295 sport = udphdr->uh_sport; 1296 dport = udphdr->uh_dport; 1297 1298 /* 1299 * Remove the UDP header, plus a possible marker. IP header 1300 * length is iphdrlen. 1301 * 1302 * Before: 1303 * <--- off ---> 1304 * +----+------+-----+ 1305 * | IP | UDP | ESP | 1306 * +----+------+-----+ 1307 * <-skip-> 1308 * After: 1309 * +----+-----+ 1310 * | IP | ESP | 1311 * +----+-----+ 1312 * <-skip-> 1313 */ 1314 iphdrlen = off - sizeof(struct udphdr); 1315 memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen); 1316 m_adj(m, skip); 1317 1318 ip = mtod(m, struct ip *); 1319 ip->ip_len = htons(ntohs(ip->ip_len) - skip); 1320 ip->ip_p = IPPROTO_ESP; 1321 1322 /* 1323 * We have modified the packet - it is now ESP, so we should not 1324 * return to UDP processing. 1325 * 1326 * Add a PACKET_TAG_IPSEC_NAT_T_PORTS tag to remember the source 1327 * UDP port. This is required if we want to select the right SPD 1328 * for multiple hosts behind same NAT. 1329 */ 1330 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, 1331 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) { 1332 m_freem(m); 1333 return -1; 1334 } 1335 ((u_int16_t *)(tag + 1))[0] = sport; 1336 ((u_int16_t *)(tag + 1))[1] = dport; 1337 m_tag_prepend(m, tag); 1338 1339 if (ipsec_used) 1340 ipsec4_common_input(m, iphdrlen, IPPROTO_ESP); 1341 else 1342 m_freem(m); 1343 1344 /* We handled it, it shouldn't be handled by UDP */ 1345 *mp = NULL; /* avoid free by caller ... */ 1346 return 1; 1347 } 1348 #endif 1349 1350 PR_WRAP_USRREQS(udp) 1351 #define udp_attach udp_attach_wrapper 1352 #define udp_detach udp_detach_wrapper 1353 #define udp_accept udp_accept_wrapper 1354 #define udp_bind udp_bind_wrapper 1355 #define udp_listen udp_listen_wrapper 1356 #define udp_connect udp_connect_wrapper 1357 #define udp_connect2 udp_connect2_wrapper 1358 #define udp_disconnect udp_disconnect_wrapper 1359 #define udp_shutdown udp_shutdown_wrapper 1360 #define udp_abort udp_abort_wrapper 1361 #define udp_ioctl udp_ioctl_wrapper 1362 #define udp_stat udp_stat_wrapper 1363 #define udp_peeraddr udp_peeraddr_wrapper 1364 #define udp_sockaddr udp_sockaddr_wrapper 1365 #define udp_rcvd udp_rcvd_wrapper 1366 #define udp_recvoob udp_recvoob_wrapper 1367 #define udp_send udp_send_wrapper 1368 #define udp_sendoob udp_sendoob_wrapper 1369 #define udp_purgeif udp_purgeif_wrapper 1370 1371 const struct pr_usrreqs udp_usrreqs = { 1372 .pr_attach = udp_attach, 1373 .pr_detach = udp_detach, 1374 .pr_accept = udp_accept, 1375 .pr_bind = udp_bind, 1376 .pr_listen = udp_listen, 1377 .pr_connect = udp_connect, 1378 .pr_connect2 = udp_connect2, 1379 .pr_disconnect = udp_disconnect, 1380 .pr_shutdown = udp_shutdown, 1381 .pr_abort = udp_abort, 1382 .pr_ioctl = udp_ioctl, 1383 .pr_stat = udp_stat, 1384 .pr_peeraddr = udp_peeraddr, 1385 .pr_sockaddr = udp_sockaddr, 1386 .pr_rcvd = udp_rcvd, 1387 .pr_recvoob = udp_recvoob, 1388 .pr_send = udp_send, 1389 .pr_sendoob = udp_sendoob, 1390 .pr_purgeif = udp_purgeif, 1391 }; 1392