1 /* $NetBSD: udp_usrreq.c,v 1.261 2021/02/19 14:51:59 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 1.261 2021/02/19 14:51:59 christos 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 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, 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_hdr *inph; 492 struct inpcb *inp; 493 struct mbuf *m = *mp; 494 495 rcvcnt = 0; 496 off += sizeof(struct udphdr); /* now, offset of payload */ 497 498 if (src->sin_family != AF_INET || dst->sin_family != AF_INET) 499 goto bad; 500 501 src4 = &src->sin_addr; 502 sport = &src->sin_port; 503 dst4 = &dst->sin_addr; 504 dport = &dst->sin_port; 505 506 if (IN_MULTICAST(dst4->s_addr) || 507 in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) { 508 /* 509 * Deliver a multicast or broadcast datagram to *all* sockets 510 * for which the local and remote addresses and ports match 511 * those of the incoming datagram. This allows more than 512 * one process to receive multi/broadcasts on the same port. 513 * (This really ought to be done for unicast datagrams as 514 * well, but that would cause problems with existing 515 * applications that open both address-specific sockets and 516 * a wildcard socket listening to the same port -- they would 517 * end up receiving duplicates of every unicast datagram. 518 * Those applications open the multiple sockets to overcome an 519 * inadequacy of the UDP socket interface, but for backwards 520 * compatibility we avoid the problem here rather than 521 * fixing the interface. Maybe 4.5BSD will remedy this?) 522 */ 523 524 /* 525 * KAME note: traditionally we dropped udpiphdr from mbuf here. 526 * we need udpiphdr for IPsec processing so we do that later. 527 */ 528 /* 529 * Locate pcb(s) for datagram. 530 */ 531 TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { 532 inp = (struct inpcb *)inph; 533 if (inp->inp_af != AF_INET) 534 continue; 535 536 if (inp->inp_lport != *dport) 537 continue; 538 if (!in_nullhost(inp->inp_laddr)) { 539 if (!in_hosteq(inp->inp_laddr, *dst4)) 540 continue; 541 } 542 if (!in_nullhost(inp->inp_faddr)) { 543 if (!in_hosteq(inp->inp_faddr, *src4) || 544 inp->inp_fport != *sport) 545 continue; 546 } 547 548 udp4_sendup(m, off, (struct sockaddr *)src, 549 inp->inp_socket); 550 rcvcnt++; 551 552 /* 553 * Don't look for additional matches if this one does 554 * not have either the SO_REUSEPORT or SO_REUSEADDR 555 * socket options set. This heuristic avoids searching 556 * through all pcbs in the common case of a non-shared 557 * port. It assumes that an application will never 558 * clear these options after setting them. 559 */ 560 if ((inp->inp_socket->so_options & 561 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 562 break; 563 } 564 } else { 565 /* 566 * Locate pcb for datagram. 567 */ 568 inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, 569 *dport, 0); 570 if (inp == 0) { 571 UDP_STATINC(UDP_STAT_PCBHASHMISS); 572 inp = in_pcblookup_bind(&udbtable, *dst4, *dport); 573 if (inp == 0) 574 return rcvcnt; 575 } 576 577 #ifdef IPSEC 578 /* Handle ESP over UDP */ 579 if (inp->inp_flags & INP_ESPINUDP) { 580 switch (udp4_espinudp(mp, off)) { 581 case -1: /* Error, m was freed */ 582 rcvcnt = -1; 583 goto bad; 584 585 case 1: /* ESP over UDP */ 586 rcvcnt++; 587 goto bad; 588 589 case 0: /* plain UDP */ 590 default: /* Unexpected */ 591 /* 592 * Normal UDP processing will take place, 593 * m may have changed. 594 */ 595 m = *mp; 596 break; 597 } 598 } 599 #endif 600 if (inp->inp_overudp_cb != NULL) { 601 int ret; 602 ret = inp->inp_overudp_cb(mp, off, inp->inp_socket, 603 sintosa(src), inp->inp_overudp_arg); 604 switch (ret) { 605 case -1: /* Error, m was freed */ 606 rcvcnt = -1; 607 goto bad; 608 609 case 1: /* Foo over UDP */ 610 KASSERT(*mp == NULL); 611 rcvcnt++; 612 goto bad; 613 614 case 0: /* plain UDP */ 615 default: /* Unexpected */ 616 /* 617 * Normal UDP processing will take place, 618 * m may have changed. 619 */ 620 m = *mp; 621 break; 622 } 623 } 624 625 /* 626 * Check the minimum TTL for socket. 627 */ 628 if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl) 629 goto bad; 630 631 udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); 632 rcvcnt++; 633 } 634 635 bad: 636 return rcvcnt; 637 } 638 #endif 639 640 #ifdef INET 641 /* 642 * Notify a udp user of an asynchronous error; 643 * just wake up so that he can collect error status. 644 */ 645 static void 646 udp_notify(struct inpcb *inp, int errno) 647 { 648 inp->inp_socket->so_error = errno; 649 sorwakeup(inp->inp_socket); 650 sowwakeup(inp->inp_socket); 651 } 652 653 void * 654 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v) 655 { 656 struct ip *ip = v; 657 struct udphdr *uh; 658 void (*notify)(struct inpcb *, int) = udp_notify; 659 int errno; 660 661 if (sa->sa_family != AF_INET || 662 sa->sa_len != sizeof(struct sockaddr_in)) 663 return NULL; 664 if ((unsigned)cmd >= PRC_NCMDS) 665 return NULL; 666 667 errno = inetctlerrmap[cmd]; 668 if (PRC_IS_REDIRECT(cmd)) { 669 notify = in_rtchange; 670 ip = NULL; 671 } else if (cmd == PRC_HOSTDEAD) { 672 ip = NULL; 673 } else if (errno == 0) { 674 return NULL; 675 } 676 677 if (ip) { 678 uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2)); 679 in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport, 680 ip->ip_src, uh->uh_sport, errno, notify); 681 /* XXX mapped address case */ 682 } else { 683 in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno, 684 notify); 685 } 686 687 return NULL; 688 } 689 690 int 691 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 692 { 693 int s; 694 int error = 0; 695 struct inpcb *inp; 696 int family; 697 int optval; 698 699 family = so->so_proto->pr_domain->dom_family; 700 701 s = splsoftnet(); 702 switch (family) { 703 #ifdef INET 704 case PF_INET: 705 if (sopt->sopt_level != IPPROTO_UDP) { 706 error = ip_ctloutput(op, so, sopt); 707 goto end; 708 } 709 break; 710 #endif 711 #ifdef INET6 712 case PF_INET6: 713 if (sopt->sopt_level != IPPROTO_UDP) { 714 error = ip6_ctloutput(op, so, sopt); 715 goto end; 716 } 717 break; 718 #endif 719 default: 720 error = EAFNOSUPPORT; 721 goto end; 722 } 723 724 725 switch (op) { 726 case PRCO_SETOPT: 727 inp = sotoinpcb(so); 728 729 switch (sopt->sopt_name) { 730 case UDP_ENCAP: 731 error = sockopt_getint(sopt, &optval); 732 if (error) 733 break; 734 735 switch(optval) { 736 case 0: 737 inp->inp_flags &= ~INP_ESPINUDP; 738 break; 739 740 case UDP_ENCAP_ESPINUDP: 741 inp->inp_flags |= INP_ESPINUDP; 742 break; 743 744 default: 745 error = EINVAL; 746 break; 747 } 748 break; 749 750 default: 751 error = ENOPROTOOPT; 752 break; 753 } 754 break; 755 756 default: 757 error = EINVAL; 758 break; 759 } 760 761 end: 762 splx(s); 763 return error; 764 } 765 766 int 767 udp_output(struct mbuf *m, struct inpcb *inp, struct mbuf *control, 768 struct lwp *l) 769 { 770 struct udpiphdr *ui; 771 struct route *ro; 772 struct ip_pktopts pktopts; 773 kauth_cred_t cred; 774 int len = m->m_pkthdr.len; 775 int error, flags = 0; 776 777 MCLAIM(m, &udp_tx_mowner); 778 779 /* 780 * Calculate data length and get a mbuf 781 * for UDP and IP headers. 782 */ 783 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 784 if (m == NULL) { 785 error = ENOBUFS; 786 goto release; 787 } 788 789 /* 790 * Compute the packet length of the IP header, and 791 * punt if the length looks bogus. 792 */ 793 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 794 error = EMSGSIZE; 795 goto release; 796 } 797 798 if (l == NULL) 799 cred = NULL; 800 else 801 cred = l->l_cred; 802 803 /* Setup IP outgoing packet options */ 804 memset(&pktopts, 0, sizeof(pktopts)); 805 error = ip_setpktopts(control, &pktopts, &flags, inp, cred); 806 if (error != 0) 807 goto release; 808 809 if (control != NULL) { 810 m_freem(control); 811 control = NULL; 812 } 813 814 /* 815 * Fill in mbuf with extended UDP header 816 * and addresses and length put into network format. 817 */ 818 ui = mtod(m, struct udpiphdr *); 819 ui->ui_pr = IPPROTO_UDP; 820 ui->ui_src = pktopts.ippo_laddr.sin_addr; 821 ui->ui_dst = inp->inp_faddr; 822 ui->ui_sport = inp->inp_lport; 823 ui->ui_dport = inp->inp_fport; 824 ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); 825 826 ro = &inp->inp_route; 827 828 /* 829 * Set up checksum and output datagram. 830 */ 831 if (udpcksum) { 832 /* 833 * XXX Cache pseudo-header checksum part for 834 * XXX "connected" UDP sockets. 835 */ 836 ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, 837 ui->ui_dst.s_addr, htons((u_int16_t)len + 838 sizeof(struct udphdr) + IPPROTO_UDP)); 839 m->m_pkthdr.csum_flags = M_CSUM_UDPv4; 840 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 841 } else 842 ui->ui_sum = 0; 843 844 ((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len); 845 ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ 846 ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ 847 UDP_STATINC(UDP_STAT_OPACKETS); 848 849 flags |= inp->inp_socket->so_options & (SO_DONTROUTE|SO_BROADCAST); 850 return ip_output(m, inp->inp_options, ro, flags, pktopts.ippo_imo, inp); 851 852 release: 853 if (control != NULL) 854 m_freem(control); 855 m_freem(m); 856 return error; 857 } 858 859 static int 860 udp_attach(struct socket *so, int proto) 861 { 862 struct inpcb *inp; 863 int error; 864 865 KASSERT(sotoinpcb(so) == NULL); 866 867 /* Assign the lock (must happen even if we will error out). */ 868 sosetlock(so); 869 870 #ifdef MBUFTRACE 871 so->so_mowner = &udp_mowner; 872 so->so_rcv.sb_mowner = &udp_rx_mowner; 873 so->so_snd.sb_mowner = &udp_tx_mowner; 874 #endif 875 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 876 error = soreserve(so, udp_sendspace, udp_recvspace); 877 if (error) { 878 return error; 879 } 880 } 881 882 error = in_pcballoc(so, &udbtable); 883 if (error) { 884 return error; 885 } 886 inp = sotoinpcb(so); 887 inp->inp_ip.ip_ttl = ip_defttl; 888 KASSERT(solocked(so)); 889 890 return error; 891 } 892 893 static void 894 udp_detach(struct socket *so) 895 { 896 struct inpcb *inp; 897 898 KASSERT(solocked(so)); 899 inp = sotoinpcb(so); 900 KASSERT(inp != NULL); 901 in_pcbdetach(inp); 902 } 903 904 static int 905 udp_accept(struct socket *so, struct sockaddr *nam) 906 { 907 KASSERT(solocked(so)); 908 909 panic("udp_accept"); 910 911 return EOPNOTSUPP; 912 } 913 914 static int 915 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l) 916 { 917 struct inpcb *inp = sotoinpcb(so); 918 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 919 int error = 0; 920 int s; 921 922 KASSERT(solocked(so)); 923 KASSERT(inp != NULL); 924 KASSERT(nam != NULL); 925 926 s = splsoftnet(); 927 error = in_pcbbind(inp, sin, l); 928 splx(s); 929 930 return error; 931 } 932 933 static int 934 udp_listen(struct socket *so, struct lwp *l) 935 { 936 KASSERT(solocked(so)); 937 938 return EOPNOTSUPP; 939 } 940 941 static int 942 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l) 943 { 944 struct inpcb *inp = sotoinpcb(so); 945 int error = 0; 946 int s; 947 948 KASSERT(solocked(so)); 949 KASSERT(inp != NULL); 950 KASSERT(nam != NULL); 951 952 s = splsoftnet(); 953 error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l); 954 if (! error) 955 soisconnected(so); 956 splx(s); 957 return error; 958 } 959 960 static int 961 udp_connect2(struct socket *so, struct socket *so2) 962 { 963 KASSERT(solocked(so)); 964 965 return EOPNOTSUPP; 966 } 967 968 static int 969 udp_disconnect(struct socket *so) 970 { 971 struct inpcb *inp = sotoinpcb(so); 972 int s; 973 974 KASSERT(solocked(so)); 975 KASSERT(inp != NULL); 976 977 s = splsoftnet(); 978 /*soisdisconnected(so);*/ 979 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 980 in_pcbdisconnect(inp); 981 inp->inp_laddr = zeroin_addr; /* XXX */ 982 in_pcbstate(inp, INP_BOUND); /* XXX */ 983 splx(s); 984 985 return 0; 986 } 987 988 static int 989 udp_shutdown(struct socket *so) 990 { 991 int s; 992 993 KASSERT(solocked(so)); 994 995 s = splsoftnet(); 996 socantsendmore(so); 997 splx(s); 998 999 return 0; 1000 } 1001 1002 static int 1003 udp_abort(struct socket *so) 1004 { 1005 KASSERT(solocked(so)); 1006 1007 panic("udp_abort"); 1008 1009 return EOPNOTSUPP; 1010 } 1011 1012 static int 1013 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp) 1014 { 1015 return in_control(so, cmd, nam, ifp); 1016 } 1017 1018 static int 1019 udp_stat(struct socket *so, struct stat *ub) 1020 { 1021 KASSERT(solocked(so)); 1022 1023 /* stat: don't bother with a blocksize. */ 1024 return 0; 1025 } 1026 1027 static int 1028 udp_peeraddr(struct socket *so, struct sockaddr *nam) 1029 { 1030 int s; 1031 1032 KASSERT(solocked(so)); 1033 KASSERT(sotoinpcb(so) != NULL); 1034 KASSERT(nam != NULL); 1035 1036 s = splsoftnet(); 1037 in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam); 1038 splx(s); 1039 1040 return 0; 1041 } 1042 1043 static int 1044 udp_sockaddr(struct socket *so, struct sockaddr *nam) 1045 { 1046 int s; 1047 1048 KASSERT(solocked(so)); 1049 KASSERT(sotoinpcb(so) != NULL); 1050 KASSERT(nam != NULL); 1051 1052 s = splsoftnet(); 1053 in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam); 1054 splx(s); 1055 1056 return 0; 1057 } 1058 1059 static int 1060 udp_rcvd(struct socket *so, int flags, struct lwp *l) 1061 { 1062 KASSERT(solocked(so)); 1063 1064 return EOPNOTSUPP; 1065 } 1066 1067 static int 1068 udp_recvoob(struct socket *so, struct mbuf *m, int flags) 1069 { 1070 KASSERT(solocked(so)); 1071 1072 return EOPNOTSUPP; 1073 } 1074 1075 int 1076 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam, 1077 struct mbuf *control, struct lwp *l) 1078 { 1079 struct inpcb *inp = sotoinpcb(so); 1080 int error = 0; 1081 struct in_addr laddr; /* XXX */ 1082 int s; 1083 1084 KASSERT(solocked(so)); 1085 KASSERT(inp != NULL); 1086 KASSERT(m != NULL); 1087 1088 memset(&laddr, 0, sizeof laddr); 1089 1090 s = splsoftnet(); 1091 if (nam) { 1092 laddr = inp->inp_laddr; /* XXX */ 1093 if ((so->so_state & SS_ISCONNECTED) != 0) { 1094 error = EISCONN; 1095 goto die; 1096 } 1097 error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l); 1098 if (error) 1099 goto die; 1100 } else { 1101 if ((so->so_state & SS_ISCONNECTED) == 0) { 1102 error = ENOTCONN; 1103 goto die; 1104 } 1105 } 1106 error = udp_output(m, inp, control, l); 1107 m = NULL; 1108 control = NULL; 1109 if (nam) { 1110 in_pcbdisconnect(inp); 1111 inp->inp_laddr = laddr; /* XXX */ 1112 in_pcbstate(inp, INP_BOUND); /* XXX */ 1113 } 1114 die: 1115 if (m != NULL) 1116 m_freem(m); 1117 if (control != NULL) 1118 m_freem(control); 1119 1120 splx(s); 1121 return error; 1122 } 1123 1124 static int 1125 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control) 1126 { 1127 KASSERT(solocked(so)); 1128 1129 m_freem(m); 1130 m_freem(control); 1131 1132 return EOPNOTSUPP; 1133 } 1134 1135 static int 1136 udp_purgeif(struct socket *so, struct ifnet *ifp) 1137 { 1138 int s; 1139 1140 s = splsoftnet(); 1141 mutex_enter(softnet_lock); 1142 in_pcbpurgeif0(&udbtable, ifp); 1143 #ifdef NET_MPSAFE 1144 mutex_exit(softnet_lock); 1145 #endif 1146 in_purgeif(ifp); 1147 #ifdef NET_MPSAFE 1148 mutex_enter(softnet_lock); 1149 #endif 1150 in_pcbpurgeif(&udbtable, ifp); 1151 mutex_exit(softnet_lock); 1152 splx(s); 1153 1154 return 0; 1155 } 1156 1157 static int 1158 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) 1159 { 1160 1161 return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS)); 1162 } 1163 1164 /* 1165 * Sysctl for udp variables. 1166 */ 1167 static void 1168 sysctl_net_inet_udp_setup(struct sysctllog **clog) 1169 { 1170 1171 sysctl_createv(clog, 0, NULL, NULL, 1172 CTLFLAG_PERMANENT, 1173 CTLTYPE_NODE, "inet", NULL, 1174 NULL, 0, NULL, 0, 1175 CTL_NET, PF_INET, CTL_EOL); 1176 sysctl_createv(clog, 0, NULL, NULL, 1177 CTLFLAG_PERMANENT, 1178 CTLTYPE_NODE, "udp", 1179 SYSCTL_DESCR("UDPv4 related settings"), 1180 NULL, 0, NULL, 0, 1181 CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL); 1182 1183 sysctl_createv(clog, 0, NULL, NULL, 1184 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1185 CTLTYPE_INT, "checksum", 1186 SYSCTL_DESCR("Compute UDP checksums"), 1187 NULL, 0, &udpcksum, 0, 1188 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM, 1189 CTL_EOL); 1190 sysctl_createv(clog, 0, NULL, NULL, 1191 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1192 CTLTYPE_INT, "sendspace", 1193 SYSCTL_DESCR("Default UDP send buffer size"), 1194 NULL, 0, &udp_sendspace, 0, 1195 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE, 1196 CTL_EOL); 1197 sysctl_createv(clog, 0, NULL, NULL, 1198 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1199 CTLTYPE_INT, "recvspace", 1200 SYSCTL_DESCR("Default UDP receive buffer size"), 1201 NULL, 0, &udp_recvspace, 0, 1202 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE, 1203 CTL_EOL); 1204 sysctl_createv(clog, 0, NULL, NULL, 1205 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1206 CTLTYPE_INT, "do_loopback_cksum", 1207 SYSCTL_DESCR("Perform UDP checksum on loopback"), 1208 NULL, 0, &udp_do_loopback_cksum, 0, 1209 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM, 1210 CTL_EOL); 1211 sysctl_createv(clog, 0, NULL, NULL, 1212 CTLFLAG_PERMANENT, 1213 CTLTYPE_STRUCT, "pcblist", 1214 SYSCTL_DESCR("UDP protocol control block list"), 1215 sysctl_inpcblist, 0, &udbtable, 0, 1216 CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE, 1217 CTL_EOL); 1218 sysctl_createv(clog, 0, NULL, NULL, 1219 CTLFLAG_PERMANENT, 1220 CTLTYPE_STRUCT, "stats", 1221 SYSCTL_DESCR("UDP statistics"), 1222 sysctl_net_inet_udp_stats, 0, NULL, 0, 1223 CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS, 1224 CTL_EOL); 1225 } 1226 #endif 1227 1228 void 1229 udp_statinc(u_int stat) 1230 { 1231 1232 KASSERT(stat < UDP_NSTATS); 1233 UDP_STATINC(stat); 1234 } 1235 1236 #if defined(INET) && defined(IPSEC) 1237 /* 1238 * Handle ESP-in-UDP packets (RFC3948). 1239 * 1240 * We need to distinguish between ESP packets and IKE packets. We do so by 1241 * looking at the Non-ESP marker. If IKE, we process the UDP packet as usual. 1242 * Otherwise, ESP, we invoke IPsec. 1243 * 1244 * Returns: 1245 * 1 if the packet was processed 1246 * 0 if normal UDP processing should take place 1247 * -1 if an error occurred and m was freed 1248 */ 1249 static int 1250 udp4_espinudp(struct mbuf **mp, int off) 1251 { 1252 const size_t skip = sizeof(struct udphdr); 1253 size_t len; 1254 uint8_t *data; 1255 size_t minlen; 1256 size_t iphdrlen; 1257 struct ip *ip; 1258 struct m_tag *tag; 1259 struct udphdr *udphdr; 1260 u_int16_t sport, dport; 1261 struct mbuf *m = *mp; 1262 uint32_t *marker; 1263 1264 minlen = off + sizeof(struct esp); 1265 if (minlen > m->m_pkthdr.len) 1266 minlen = m->m_pkthdr.len; 1267 1268 if (m->m_len < minlen) { 1269 if ((*mp = m_pullup(m, minlen)) == NULL) { 1270 return -1; 1271 } 1272 m = *mp; 1273 } 1274 1275 len = m->m_len - off; 1276 data = mtod(m, uint8_t *) + off; 1277 1278 /* Ignore keepalive packets. */ 1279 if ((len == 1) && (*data == 0xff)) { 1280 m_freem(m); 1281 *mp = NULL; /* avoid any further processing by caller */ 1282 return 1; 1283 } 1284 1285 /* Handle Non-ESP marker (32bit). If zero, then IKE. */ 1286 marker = (uint32_t *)data; 1287 if (len <= sizeof(uint32_t)) 1288 return 0; 1289 if (marker[0] == 0) 1290 return 0; 1291 1292 /* 1293 * Get the UDP ports. They are handled in network order 1294 * everywhere in the IPSEC_NAT_T code. 1295 */ 1296 udphdr = (struct udphdr *)((char *)data - skip); 1297 sport = udphdr->uh_sport; 1298 dport = udphdr->uh_dport; 1299 1300 /* 1301 * Remove the UDP header, plus a possible marker. IP header 1302 * length is iphdrlen. 1303 * 1304 * Before: 1305 * <--- off ---> 1306 * +----+------+-----+ 1307 * | IP | UDP | ESP | 1308 * +----+------+-----+ 1309 * <-skip-> 1310 * After: 1311 * +----+-----+ 1312 * | IP | ESP | 1313 * +----+-----+ 1314 * <-skip-> 1315 */ 1316 iphdrlen = off - sizeof(struct udphdr); 1317 memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen); 1318 m_adj(m, skip); 1319 1320 ip = mtod(m, struct ip *); 1321 ip->ip_len = htons(ntohs(ip->ip_len) - skip); 1322 ip->ip_p = IPPROTO_ESP; 1323 1324 /* 1325 * We have modified the packet - it is now ESP, so we should not 1326 * return to UDP processing. 1327 * 1328 * Add a PACKET_TAG_IPSEC_NAT_T_PORTS tag to remember the source 1329 * UDP port. This is required if we want to select the right SPD 1330 * for multiple hosts behind same NAT. 1331 */ 1332 if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, 1333 sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) { 1334 m_freem(m); 1335 return -1; 1336 } 1337 ((u_int16_t *)(tag + 1))[0] = sport; 1338 ((u_int16_t *)(tag + 1))[1] = dport; 1339 m_tag_prepend(m, tag); 1340 1341 if (ipsec_used) 1342 ipsec4_common_input(m, iphdrlen, IPPROTO_ESP); 1343 else 1344 m_freem(m); 1345 1346 /* We handled it, it shouldn't be handled by UDP */ 1347 *mp = NULL; /* avoid free by caller ... */ 1348 return 1; 1349 } 1350 #endif 1351 1352 PR_WRAP_USRREQS(udp) 1353 #define udp_attach udp_attach_wrapper 1354 #define udp_detach udp_detach_wrapper 1355 #define udp_accept udp_accept_wrapper 1356 #define udp_bind udp_bind_wrapper 1357 #define udp_listen udp_listen_wrapper 1358 #define udp_connect udp_connect_wrapper 1359 #define udp_connect2 udp_connect2_wrapper 1360 #define udp_disconnect udp_disconnect_wrapper 1361 #define udp_shutdown udp_shutdown_wrapper 1362 #define udp_abort udp_abort_wrapper 1363 #define udp_ioctl udp_ioctl_wrapper 1364 #define udp_stat udp_stat_wrapper 1365 #define udp_peeraddr udp_peeraddr_wrapper 1366 #define udp_sockaddr udp_sockaddr_wrapper 1367 #define udp_rcvd udp_rcvd_wrapper 1368 #define udp_recvoob udp_recvoob_wrapper 1369 #define udp_send udp_send_wrapper 1370 #define udp_sendoob udp_sendoob_wrapper 1371 #define udp_purgeif udp_purgeif_wrapper 1372 1373 const struct pr_usrreqs udp_usrreqs = { 1374 .pr_attach = udp_attach, 1375 .pr_detach = udp_detach, 1376 .pr_accept = udp_accept, 1377 .pr_bind = udp_bind, 1378 .pr_listen = udp_listen, 1379 .pr_connect = udp_connect, 1380 .pr_connect2 = udp_connect2, 1381 .pr_disconnect = udp_disconnect, 1382 .pr_shutdown = udp_shutdown, 1383 .pr_abort = udp_abort, 1384 .pr_ioctl = udp_ioctl, 1385 .pr_stat = udp_stat, 1386 .pr_peeraddr = udp_peeraddr, 1387 .pr_sockaddr = udp_sockaddr, 1388 .pr_rcvd = udp_rcvd, 1389 .pr_recvoob = udp_recvoob, 1390 .pr_send = udp_send, 1391 .pr_sendoob = udp_sendoob, 1392 .pr_purgeif = udp_purgeif, 1393 }; 1394