1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * Copyright (c) 2010-2011 Juniper Networks, Inc. 4 * Copyright (c) 2014 Kevin Lo 5 * All rights reserved. 6 * 7 * Portions of this software were developed by Robert N. M. Watson under 8 * contract to Juniper Networks, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the project nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * $KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $ 35 * $KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $ 36 */ 37 38 /*- 39 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 40 * The Regents of the University of California. 41 * All rights reserved. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 4. Neither the name of the University nor the names of its contributors 52 * may be used to endorse or promote products derived from this software 53 * without specific prior written permission. 54 * 55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 65 * SUCH DAMAGE. 66 * 67 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 68 */ 69 70 #include <sys/cdefs.h> 71 __FBSDID("$FreeBSD$"); 72 73 #include "opt_inet.h" 74 #include "opt_inet6.h" 75 #include "opt_ipfw.h" 76 #include "opt_ipsec.h" 77 #include "opt_rss.h" 78 79 #include <sys/param.h> 80 #include <sys/jail.h> 81 #include <sys/kernel.h> 82 #include <sys/lock.h> 83 #include <sys/mbuf.h> 84 #include <sys/priv.h> 85 #include <sys/proc.h> 86 #include <sys/protosw.h> 87 #include <sys/sdt.h> 88 #include <sys/signalvar.h> 89 #include <sys/socket.h> 90 #include <sys/socketvar.h> 91 #include <sys/sx.h> 92 #include <sys/sysctl.h> 93 #include <sys/syslog.h> 94 #include <sys/systm.h> 95 96 #include <net/if.h> 97 #include <net/if_var.h> 98 #include <net/if_types.h> 99 #include <net/route.h> 100 #include <net/rss_config.h> 101 102 #include <netinet/in.h> 103 #include <netinet/in_kdtrace.h> 104 #include <netinet/in_pcb.h> 105 #include <netinet/in_systm.h> 106 #include <netinet/in_var.h> 107 #include <netinet/ip.h> 108 #include <netinet/ip_icmp.h> 109 #include <netinet/ip6.h> 110 #include <netinet/icmp_var.h> 111 #include <netinet/icmp6.h> 112 #include <netinet/ip_var.h> 113 #include <netinet/udp.h> 114 #include <netinet/udp_var.h> 115 #include <netinet/udplite.h> 116 117 #include <netinet6/ip6protosw.h> 118 #include <netinet6/ip6_var.h> 119 #include <netinet6/in6_pcb.h> 120 #include <netinet6/in6_rss.h> 121 #include <netinet6/udp6_var.h> 122 #include <netinet6/scope6_var.h> 123 124 #ifdef IPSEC 125 #include <netipsec/ipsec.h> 126 #include <netipsec/ipsec6.h> 127 #endif /* IPSEC */ 128 129 #include <security/mac/mac_framework.h> 130 131 /* 132 * UDP protocol implementation. 133 * Per RFC 768, August, 1980. 134 */ 135 136 extern struct protosw inetsw[]; 137 static void udp6_detach(struct socket *so); 138 139 static int 140 udp6_append(struct inpcb *inp, struct mbuf *n, int off, 141 struct sockaddr_in6 *fromsa) 142 { 143 struct socket *so; 144 struct mbuf *opts; 145 struct udpcb *up; 146 147 INP_LOCK_ASSERT(inp); 148 149 /* 150 * Engage the tunneling protocol. 151 */ 152 up = intoudpcb(inp); 153 if (up->u_tun_func != NULL) { 154 in_pcbref(inp); 155 INP_RUNLOCK(inp); 156 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)fromsa, 157 up->u_tun_ctx); 158 INP_RLOCK(inp); 159 return (in_pcbrele_rlocked(inp)); 160 } 161 #ifdef IPSEC 162 /* Check AH/ESP integrity. */ 163 if (ipsec6_in_reject(n, inp)) { 164 m_freem(n); 165 return (0); 166 } 167 #endif /* IPSEC */ 168 #ifdef MAC 169 if (mac_inpcb_check_deliver(inp, n) != 0) { 170 m_freem(n); 171 return (0); 172 } 173 #endif 174 opts = NULL; 175 if (inp->inp_flags & INP_CONTROLOPTS || 176 inp->inp_socket->so_options & SO_TIMESTAMP) 177 ip6_savecontrol(inp, n, &opts); 178 m_adj(n, off + sizeof(struct udphdr)); 179 180 so = inp->inp_socket; 181 SOCKBUF_LOCK(&so->so_rcv); 182 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n, 183 opts) == 0) { 184 SOCKBUF_UNLOCK(&so->so_rcv); 185 m_freem(n); 186 if (opts) 187 m_freem(opts); 188 UDPSTAT_INC(udps_fullsock); 189 } else 190 sorwakeup_locked(so); 191 return (0); 192 } 193 194 int 195 udp6_input(struct mbuf **mp, int *offp, int proto) 196 { 197 struct mbuf *m = *mp; 198 struct ifnet *ifp; 199 struct ip6_hdr *ip6; 200 struct udphdr *uh; 201 struct inpcb *inp; 202 struct inpcbinfo *pcbinfo; 203 struct udpcb *up; 204 int off = *offp; 205 int cscov_partial; 206 int plen, ulen; 207 struct sockaddr_in6 fromsa; 208 struct m_tag *fwd_tag; 209 uint16_t uh_sum; 210 uint8_t nxt; 211 212 ifp = m->m_pkthdr.rcvif; 213 ip6 = mtod(m, struct ip6_hdr *); 214 215 #ifndef PULLDOWN_TEST 216 IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE); 217 ip6 = mtod(m, struct ip6_hdr *); 218 uh = (struct udphdr *)((caddr_t)ip6 + off); 219 #else 220 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh)); 221 if (!uh) 222 return (IPPROTO_DONE); 223 #endif 224 225 UDPSTAT_INC(udps_ipackets); 226 227 /* 228 * Destination port of 0 is illegal, based on RFC768. 229 */ 230 if (uh->uh_dport == 0) 231 goto badunlocked; 232 233 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6); 234 ulen = ntohs((u_short)uh->uh_ulen); 235 236 nxt = ip6->ip6_nxt; 237 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0; 238 if (nxt == IPPROTO_UDPLITE) { 239 /* Zero means checksum over the complete packet. */ 240 if (ulen == 0) 241 ulen = plen; 242 if (ulen == plen) 243 cscov_partial = 0; 244 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) { 245 /* XXX: What is the right UDPLite MIB counter? */ 246 goto badunlocked; 247 } 248 if (uh->uh_sum == 0) { 249 /* XXX: What is the right UDPLite MIB counter? */ 250 goto badunlocked; 251 } 252 } else { 253 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) { 254 UDPSTAT_INC(udps_badlen); 255 goto badunlocked; 256 } 257 if (uh->uh_sum == 0) { 258 UDPSTAT_INC(udps_nosum); 259 goto badunlocked; 260 } 261 } 262 263 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) && 264 !cscov_partial) { 265 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) 266 uh_sum = m->m_pkthdr.csum_data; 267 else 268 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt, 269 m->m_pkthdr.csum_data); 270 uh_sum ^= 0xffff; 271 } else 272 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen); 273 274 if (uh_sum != 0) { 275 UDPSTAT_INC(udps_badsum); 276 goto badunlocked; 277 } 278 279 /* 280 * Construct sockaddr format source address. 281 */ 282 init_sin6(&fromsa, m); 283 fromsa.sin6_port = uh->uh_sport; 284 285 pcbinfo = get_inpcbinfo(nxt); 286 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 287 struct inpcb *last; 288 struct inpcbhead *pcblist; 289 struct ip6_moptions *imo; 290 291 INP_INFO_RLOCK(pcbinfo); 292 /* 293 * In the event that laddr should be set to the link-local 294 * address (this happens in RIPng), the multicast address 295 * specified in the received packet will not match laddr. To 296 * handle this situation, matching is relaxed if the 297 * receiving interface is the same as one specified in the 298 * socket and if the destination multicast address matches 299 * one of the multicast groups specified in the socket. 300 */ 301 302 /* 303 * KAME note: traditionally we dropped udpiphdr from mbuf 304 * here. We need udphdr for IPsec processing so we do that 305 * later. 306 */ 307 pcblist = get_pcblist(nxt); 308 last = NULL; 309 LIST_FOREACH(inp, pcblist, inp_list) { 310 if ((inp->inp_vflag & INP_IPV6) == 0) 311 continue; 312 if (inp->inp_lport != uh->uh_dport) 313 continue; 314 if (inp->inp_fport != 0 && 315 inp->inp_fport != uh->uh_sport) 316 continue; 317 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 318 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, 319 &ip6->ip6_dst)) 320 continue; 321 } 322 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 323 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, 324 &ip6->ip6_src) || 325 inp->inp_fport != uh->uh_sport) 326 continue; 327 } 328 329 /* 330 * XXXRW: Because we weren't holding either the inpcb 331 * or the hash lock when we checked for a match 332 * before, we should probably recheck now that the 333 * inpcb lock is (supposed to be) held. 334 */ 335 336 /* 337 * Handle socket delivery policy for any-source 338 * and source-specific multicast. [RFC3678] 339 */ 340 imo = inp->in6p_moptions; 341 if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 342 struct sockaddr_in6 mcaddr; 343 int blocked; 344 345 INP_RLOCK(inp); 346 347 bzero(&mcaddr, sizeof(struct sockaddr_in6)); 348 mcaddr.sin6_len = sizeof(struct sockaddr_in6); 349 mcaddr.sin6_family = AF_INET6; 350 mcaddr.sin6_addr = ip6->ip6_dst; 351 352 blocked = im6o_mc_filter(imo, ifp, 353 (struct sockaddr *)&mcaddr, 354 (struct sockaddr *)&fromsa); 355 if (blocked != MCAST_PASS) { 356 if (blocked == MCAST_NOTGMEMBER) 357 IP6STAT_INC(ip6s_notmember); 358 if (blocked == MCAST_NOTSMEMBER || 359 blocked == MCAST_MUTED) 360 UDPSTAT_INC(udps_filtermcast); 361 INP_RUNLOCK(inp); /* XXX */ 362 continue; 363 } 364 365 INP_RUNLOCK(inp); 366 } 367 if (last != NULL) { 368 struct mbuf *n; 369 370 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { 371 INP_RLOCK(last); 372 UDP_PROBE(receive, NULL, last, ip6, 373 last, uh); 374 if (udp6_append(last, n, off, &fromsa)) 375 goto inp_lost; 376 INP_RUNLOCK(last); 377 } 378 } 379 last = inp; 380 /* 381 * Don't look for additional matches if this one does 382 * not have either the SO_REUSEPORT or SO_REUSEADDR 383 * socket options set. This heuristic avoids 384 * searching through all pcbs in the common case of a 385 * non-shared port. It assumes that an application 386 * will never clear these options after setting them. 387 */ 388 if ((last->inp_socket->so_options & 389 (SO_REUSEPORT|SO_REUSEADDR)) == 0) 390 break; 391 } 392 393 if (last == NULL) { 394 /* 395 * No matching pcb found; discard datagram. (No need 396 * to send an ICMP Port Unreachable for a broadcast 397 * or multicast datgram.) 398 */ 399 UDPSTAT_INC(udps_noport); 400 UDPSTAT_INC(udps_noportmcast); 401 goto badheadlocked; 402 } 403 INP_RLOCK(last); 404 INP_INFO_RUNLOCK(pcbinfo); 405 UDP_PROBE(receive, NULL, last, ip6, last, uh); 406 if (udp6_append(last, m, off, &fromsa) == 0) 407 INP_RUNLOCK(last); 408 inp_lost: 409 return (IPPROTO_DONE); 410 } 411 /* 412 * Locate pcb for datagram. 413 */ 414 415 /* 416 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. 417 */ 418 if ((m->m_flags & M_IP6_NEXTHOP) && 419 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 420 struct sockaddr_in6 *next_hop6; 421 422 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1); 423 424 /* 425 * Transparently forwarded. Pretend to be the destination. 426 * Already got one like this? 427 */ 428 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src, 429 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport, 430 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m); 431 if (!inp) { 432 /* 433 * It's new. Try to find the ambushing socket. 434 * Because we've rewritten the destination address, 435 * any hardware-generated hash is ignored. 436 */ 437 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src, 438 uh->uh_sport, &next_hop6->sin6_addr, 439 next_hop6->sin6_port ? htons(next_hop6->sin6_port) : 440 uh->uh_dport, INPLOOKUP_WILDCARD | 441 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif); 442 } 443 /* Remove the tag from the packet. We don't need it anymore. */ 444 m_tag_delete(m, fwd_tag); 445 m->m_flags &= ~M_IP6_NEXTHOP; 446 } else 447 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src, 448 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport, 449 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, 450 m->m_pkthdr.rcvif, m); 451 if (inp == NULL) { 452 if (udp_log_in_vain) { 453 char ip6bufs[INET6_ADDRSTRLEN]; 454 char ip6bufd[INET6_ADDRSTRLEN]; 455 456 log(LOG_INFO, 457 "Connection attempt to UDP [%s]:%d from [%s]:%d\n", 458 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 459 ntohs(uh->uh_dport), 460 ip6_sprintf(ip6bufs, &ip6->ip6_src), 461 ntohs(uh->uh_sport)); 462 } 463 UDPSTAT_INC(udps_noport); 464 if (m->m_flags & M_MCAST) { 465 printf("UDP6: M_MCAST is set in a unicast packet.\n"); 466 UDPSTAT_INC(udps_noportmcast); 467 goto badunlocked; 468 } 469 if (V_udp_blackhole) 470 goto badunlocked; 471 if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0) 472 goto badunlocked; 473 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); 474 return (IPPROTO_DONE); 475 } 476 INP_RLOCK_ASSERT(inp); 477 up = intoudpcb(inp); 478 if (cscov_partial) { 479 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) { 480 INP_RUNLOCK(inp); 481 m_freem(m); 482 return (IPPROTO_DONE); 483 } 484 } 485 UDP_PROBE(receive, NULL, inp, ip6, inp, uh); 486 if (udp6_append(inp, m, off, &fromsa) == 0) 487 INP_RUNLOCK(inp); 488 return (IPPROTO_DONE); 489 490 badheadlocked: 491 INP_INFO_RUNLOCK(pcbinfo); 492 badunlocked: 493 if (m) 494 m_freem(m); 495 return (IPPROTO_DONE); 496 } 497 498 static void 499 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d, 500 struct inpcbinfo *pcbinfo) 501 { 502 struct udphdr uh; 503 struct ip6_hdr *ip6; 504 struct mbuf *m; 505 int off = 0; 506 struct ip6ctlparam *ip6cp = NULL; 507 const struct sockaddr_in6 *sa6_src = NULL; 508 void *cmdarg; 509 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify; 510 struct udp_portonly { 511 u_int16_t uh_sport; 512 u_int16_t uh_dport; 513 } *uhp; 514 515 if (sa->sa_family != AF_INET6 || 516 sa->sa_len != sizeof(struct sockaddr_in6)) 517 return; 518 519 if ((unsigned)cmd >= PRC_NCMDS) 520 return; 521 if (PRC_IS_REDIRECT(cmd)) 522 notify = in6_rtchange, d = NULL; 523 else if (cmd == PRC_HOSTDEAD) 524 d = NULL; 525 else if (inet6ctlerrmap[cmd] == 0) 526 return; 527 528 /* if the parameter is from icmp6, decode it. */ 529 if (d != NULL) { 530 ip6cp = (struct ip6ctlparam *)d; 531 m = ip6cp->ip6c_m; 532 ip6 = ip6cp->ip6c_ip6; 533 off = ip6cp->ip6c_off; 534 cmdarg = ip6cp->ip6c_cmdarg; 535 sa6_src = ip6cp->ip6c_src; 536 } else { 537 m = NULL; 538 ip6 = NULL; 539 cmdarg = NULL; 540 sa6_src = &sa6_any; 541 } 542 543 if (ip6) { 544 /* 545 * XXX: We assume that when IPV6 is non NULL, 546 * M and OFF are valid. 547 */ 548 549 /* Check if we can safely examine src and dst ports. */ 550 if (m->m_pkthdr.len < off + sizeof(*uhp)) 551 return; 552 553 bzero(&uh, sizeof(uh)); 554 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh); 555 556 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport, 557 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd, 558 cmdarg, notify); 559 } else 560 (void)in6_pcbnotify(pcbinfo, sa, 0, 561 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify); 562 } 563 564 void 565 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d) 566 { 567 568 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo)); 569 } 570 571 void 572 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d) 573 { 574 575 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo)); 576 } 577 578 static int 579 udp6_getcred(SYSCTL_HANDLER_ARGS) 580 { 581 struct xucred xuc; 582 struct sockaddr_in6 addrs[2]; 583 struct inpcb *inp; 584 int error; 585 586 error = priv_check(req->td, PRIV_NETINET_GETCRED); 587 if (error) 588 return (error); 589 590 if (req->newlen != sizeof(addrs)) 591 return (EINVAL); 592 if (req->oldlen != sizeof(struct xucred)) 593 return (EINVAL); 594 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 595 if (error) 596 return (error); 597 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 || 598 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) { 599 return (error); 600 } 601 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr, 602 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 603 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL); 604 if (inp != NULL) { 605 INP_RLOCK_ASSERT(inp); 606 if (inp->inp_socket == NULL) 607 error = ENOENT; 608 if (error == 0) 609 error = cr_canseesocket(req->td->td_ucred, 610 inp->inp_socket); 611 if (error == 0) 612 cru2x(inp->inp_cred, &xuc); 613 INP_RUNLOCK(inp); 614 } else 615 error = ENOENT; 616 if (error == 0) 617 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); 618 return (error); 619 } 620 621 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0, 622 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection"); 623 624 static int 625 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6, 626 struct mbuf *control, struct thread *td) 627 { 628 u_int32_t ulen = m->m_pkthdr.len; 629 u_int32_t plen = sizeof(struct udphdr) + ulen; 630 struct ip6_hdr *ip6; 631 struct udphdr *udp6; 632 struct in6_addr *laddr, *faddr, in6a; 633 struct sockaddr_in6 *sin6 = NULL; 634 struct ifnet *oifp = NULL; 635 int cscov_partial = 0; 636 int scope_ambiguous = 0; 637 u_short fport; 638 int error = 0; 639 uint8_t nxt; 640 uint16_t cscov = 0; 641 struct ip6_pktopts *optp, opt; 642 int af = AF_INET6, hlen = sizeof(struct ip6_hdr); 643 int flags; 644 struct sockaddr_in6 tmp; 645 646 INP_WLOCK_ASSERT(inp); 647 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); 648 649 if (addr6) { 650 /* addr6 has been validated in udp6_send(). */ 651 sin6 = (struct sockaddr_in6 *)addr6; 652 653 /* protect *sin6 from overwrites */ 654 tmp = *sin6; 655 sin6 = &tmp; 656 657 /* 658 * Application should provide a proper zone ID or the use of 659 * default zone IDs should be enabled. Unfortunately, some 660 * applications do not behave as it should, so we need a 661 * workaround. Even if an appropriate ID is not determined, 662 * we'll see if we can determine the outgoing interface. If we 663 * can, determine the zone ID based on the interface below. 664 */ 665 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) 666 scope_ambiguous = 1; 667 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) 668 return (error); 669 } 670 671 if (control) { 672 if ((error = ip6_setpktopts(control, &opt, 673 inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0) 674 goto release; 675 optp = &opt; 676 } else 677 optp = inp->in6p_outputopts; 678 679 if (sin6) { 680 faddr = &sin6->sin6_addr; 681 682 /* 683 * Since we saw no essential reason for calling in_pcbconnect, 684 * we get rid of such kind of logic, and call in6_selectsrc 685 * and in6_pcbsetport in order to fill in the local address 686 * and the local port. 687 */ 688 if (sin6->sin6_port == 0) { 689 error = EADDRNOTAVAIL; 690 goto release; 691 } 692 693 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 694 /* how about ::ffff:0.0.0.0 case? */ 695 error = EISCONN; 696 goto release; 697 } 698 699 fport = sin6->sin6_port; /* allow 0 port */ 700 701 if (IN6_IS_ADDR_V4MAPPED(faddr)) { 702 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) { 703 /* 704 * I believe we should explicitly discard the 705 * packet when mapped addresses are disabled, 706 * rather than send the packet as an IPv6 one. 707 * If we chose the latter approach, the packet 708 * might be sent out on the wire based on the 709 * default route, the situation which we'd 710 * probably want to avoid. 711 * (20010421 jinmei@kame.net) 712 */ 713 error = EINVAL; 714 goto release; 715 } 716 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) && 717 !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) { 718 /* 719 * when remote addr is an IPv4-mapped address, 720 * local addr should not be an IPv6 address, 721 * since you cannot determine how to map IPv6 722 * source address to IPv4. 723 */ 724 error = EINVAL; 725 goto release; 726 } 727 728 af = AF_INET; 729 } 730 731 if (!IN6_IS_ADDR_V4MAPPED(faddr)) { 732 error = in6_selectsrc(sin6, optp, inp, NULL, 733 td->td_ucred, &oifp, &in6a); 734 if (error) 735 goto release; 736 if (oifp && scope_ambiguous && 737 (error = in6_setscope(&sin6->sin6_addr, 738 oifp, NULL))) { 739 goto release; 740 } 741 laddr = &in6a; 742 } else 743 laddr = &inp->in6p_laddr; /* XXX */ 744 if (laddr == NULL) { 745 if (error == 0) 746 error = EADDRNOTAVAIL; 747 goto release; 748 } 749 if (inp->inp_lport == 0 && 750 (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) { 751 /* Undo an address bind that may have occurred. */ 752 inp->in6p_laddr = in6addr_any; 753 goto release; 754 } 755 } else { 756 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 757 error = ENOTCONN; 758 goto release; 759 } 760 if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) { 761 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) { 762 /* 763 * XXX: this case would happen when the 764 * application sets the V6ONLY flag after 765 * connecting the foreign address. 766 * Such applications should be fixed, 767 * so we bark here. 768 */ 769 log(LOG_INFO, "udp6_output: IPV6_V6ONLY " 770 "option was set for a connected socket\n"); 771 error = EINVAL; 772 goto release; 773 } else 774 af = AF_INET; 775 } 776 laddr = &inp->in6p_laddr; 777 faddr = &inp->in6p_faddr; 778 fport = inp->inp_fport; 779 } 780 781 if (af == AF_INET) 782 hlen = sizeof(struct ip); 783 784 /* 785 * Calculate data length and get a mbuf 786 * for UDP and IP6 headers. 787 */ 788 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT); 789 if (m == 0) { 790 error = ENOBUFS; 791 goto release; 792 } 793 794 /* 795 * Stuff checksum and output datagram. 796 */ 797 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ? 798 IPPROTO_UDP : IPPROTO_UDPLITE; 799 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen); 800 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */ 801 udp6->uh_dport = fport; 802 if (nxt == IPPROTO_UDPLITE) { 803 struct udpcb *up; 804 805 up = intoudpcb(inp); 806 cscov = up->u_txcslen; 807 if (cscov >= plen) 808 cscov = 0; 809 udp6->uh_ulen = htons(cscov); 810 /* 811 * For UDP-Lite, checksum coverage length of zero means 812 * the entire UDPLite packet is covered by the checksum. 813 */ 814 cscov_partial = (cscov == 0) ? 0 : 1; 815 } else if (plen <= 0xffff) 816 udp6->uh_ulen = htons((u_short)plen); 817 else 818 udp6->uh_ulen = 0; 819 udp6->uh_sum = 0; 820 821 switch (af) { 822 case AF_INET6: 823 ip6 = mtod(m, struct ip6_hdr *); 824 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK; 825 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 826 ip6->ip6_vfc |= IPV6_VERSION; 827 ip6->ip6_plen = htons((u_short)plen); 828 ip6->ip6_nxt = nxt; 829 ip6->ip6_hlim = in6_selecthlim(inp, NULL); 830 ip6->ip6_src = *laddr; 831 ip6->ip6_dst = *faddr; 832 833 if (cscov_partial) { 834 if ((udp6->uh_sum = in6_cksum_partial(m, nxt, 835 sizeof(struct ip6_hdr), plen, cscov)) == 0) 836 udp6->uh_sum = 0xffff; 837 } else { 838 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0); 839 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6; 840 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 841 } 842 843 /* 844 * XXX for now assume UDP is 2-tuple. 845 * Later on this may become configurable as 4-tuple; 846 * we should support that. 847 * 848 * XXX .. and we should likely cache this in the inpcb. 849 */ 850 #ifdef RSS 851 m->m_pkthdr.flowid = rss_hash_ip6_2tuple(faddr, laddr); 852 M_HASHTYPE_SET(m, M_HASHTYPE_RSS_IPV6); 853 #endif 854 flags = 0; 855 856 #ifdef RSS 857 /* 858 * Don't override with the inp cached flowid. 859 * 860 * Until the whole UDP path is vetted, it may actually 861 * be incorrect. 862 */ 863 flags |= IP_NODEFAULTFLOWID; 864 #endif 865 866 UDP_PROBE(send, NULL, inp, ip6, inp, udp6); 867 UDPSTAT_INC(udps_opackets); 868 error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions, 869 NULL, inp); 870 break; 871 case AF_INET: 872 error = EAFNOSUPPORT; 873 goto release; 874 } 875 goto releaseopt; 876 877 release: 878 m_freem(m); 879 880 releaseopt: 881 if (control) { 882 ip6_clearpktopts(&opt, -1); 883 m_freem(control); 884 } 885 return (error); 886 } 887 888 static void 889 udp6_abort(struct socket *so) 890 { 891 struct inpcb *inp; 892 struct inpcbinfo *pcbinfo; 893 894 pcbinfo = get_inpcbinfo(so->so_proto->pr_protocol); 895 inp = sotoinpcb(so); 896 KASSERT(inp != NULL, ("udp6_abort: inp == NULL")); 897 898 #ifdef INET 899 if (inp->inp_vflag & INP_IPV4) { 900 struct pr_usrreqs *pru; 901 902 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs; 903 (*pru->pru_abort)(so); 904 return; 905 } 906 #endif 907 908 INP_WLOCK(inp); 909 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 910 INP_HASH_WLOCK(pcbinfo); 911 in6_pcbdisconnect(inp); 912 inp->in6p_laddr = in6addr_any; 913 INP_HASH_WUNLOCK(pcbinfo); 914 soisdisconnected(so); 915 } 916 INP_WUNLOCK(inp); 917 } 918 919 static int 920 udp6_attach(struct socket *so, int proto, struct thread *td) 921 { 922 struct inpcb *inp; 923 struct inpcbinfo *pcbinfo; 924 int error; 925 926 pcbinfo = get_inpcbinfo(so->so_proto->pr_protocol); 927 inp = sotoinpcb(so); 928 KASSERT(inp == NULL, ("udp6_attach: inp != NULL")); 929 930 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 931 error = soreserve(so, udp_sendspace, udp_recvspace); 932 if (error) 933 return (error); 934 } 935 INP_INFO_WLOCK(pcbinfo); 936 error = in_pcballoc(so, pcbinfo); 937 if (error) { 938 INP_INFO_WUNLOCK(pcbinfo); 939 return (error); 940 } 941 inp = (struct inpcb *)so->so_pcb; 942 inp->inp_vflag |= INP_IPV6; 943 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 944 inp->inp_vflag |= INP_IPV4; 945 inp->in6p_hops = -1; /* use kernel default */ 946 inp->in6p_cksum = -1; /* just to be sure */ 947 /* 948 * XXX: ugly!! 949 * IPv4 TTL initialization is necessary for an IPv6 socket as well, 950 * because the socket may be bound to an IPv6 wildcard address, 951 * which may match an IPv4-mapped IPv6 address. 952 */ 953 inp->inp_ip_ttl = V_ip_defttl; 954 955 error = udp_newudpcb(inp); 956 if (error) { 957 in_pcbdetach(inp); 958 in_pcbfree(inp); 959 INP_INFO_WUNLOCK(pcbinfo); 960 return (error); 961 } 962 INP_WUNLOCK(inp); 963 INP_INFO_WUNLOCK(pcbinfo); 964 return (0); 965 } 966 967 static int 968 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 969 { 970 struct inpcb *inp; 971 struct inpcbinfo *pcbinfo; 972 int error; 973 974 pcbinfo = get_inpcbinfo(so->so_proto->pr_protocol); 975 inp = sotoinpcb(so); 976 KASSERT(inp != NULL, ("udp6_bind: inp == NULL")); 977 978 INP_WLOCK(inp); 979 INP_HASH_WLOCK(pcbinfo); 980 inp->inp_vflag &= ~INP_IPV4; 981 inp->inp_vflag |= INP_IPV6; 982 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 983 struct sockaddr_in6 *sin6_p; 984 985 sin6_p = (struct sockaddr_in6 *)nam; 986 987 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr)) 988 inp->inp_vflag |= INP_IPV4; 989 #ifdef INET 990 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) { 991 struct sockaddr_in sin; 992 993 in6_sin6_2_sin(&sin, sin6_p); 994 inp->inp_vflag |= INP_IPV4; 995 inp->inp_vflag &= ~INP_IPV6; 996 error = in_pcbbind(inp, (struct sockaddr *)&sin, 997 td->td_ucred); 998 goto out; 999 } 1000 #endif 1001 } 1002 1003 error = in6_pcbbind(inp, nam, td->td_ucred); 1004 #ifdef INET 1005 out: 1006 #endif 1007 INP_HASH_WUNLOCK(pcbinfo); 1008 INP_WUNLOCK(inp); 1009 return (error); 1010 } 1011 1012 static void 1013 udp6_close(struct socket *so) 1014 { 1015 struct inpcb *inp; 1016 struct inpcbinfo *pcbinfo; 1017 1018 pcbinfo = get_inpcbinfo(so->so_proto->pr_protocol); 1019 inp = sotoinpcb(so); 1020 KASSERT(inp != NULL, ("udp6_close: inp == NULL")); 1021 1022 #ifdef INET 1023 if (inp->inp_vflag & INP_IPV4) { 1024 struct pr_usrreqs *pru; 1025 1026 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs; 1027 (*pru->pru_disconnect)(so); 1028 return; 1029 } 1030 #endif 1031 INP_WLOCK(inp); 1032 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 1033 INP_HASH_WLOCK(pcbinfo); 1034 in6_pcbdisconnect(inp); 1035 inp->in6p_laddr = in6addr_any; 1036 INP_HASH_WUNLOCK(pcbinfo); 1037 soisdisconnected(so); 1038 } 1039 INP_WUNLOCK(inp); 1040 } 1041 1042 static int 1043 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 1044 { 1045 struct inpcb *inp; 1046 struct inpcbinfo *pcbinfo; 1047 struct sockaddr_in6 *sin6; 1048 int error; 1049 1050 pcbinfo = get_inpcbinfo(so->so_proto->pr_protocol); 1051 inp = sotoinpcb(so); 1052 sin6 = (struct sockaddr_in6 *)nam; 1053 KASSERT(inp != NULL, ("udp6_connect: inp == NULL")); 1054 1055 /* 1056 * XXXRW: Need to clarify locking of v4/v6 flags. 1057 */ 1058 INP_WLOCK(inp); 1059 #ifdef INET 1060 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 1061 struct sockaddr_in sin; 1062 1063 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 1064 error = EINVAL; 1065 goto out; 1066 } 1067 if (inp->inp_faddr.s_addr != INADDR_ANY) { 1068 error = EISCONN; 1069 goto out; 1070 } 1071 in6_sin6_2_sin(&sin, sin6); 1072 inp->inp_vflag |= INP_IPV4; 1073 inp->inp_vflag &= ~INP_IPV6; 1074 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr); 1075 if (error != 0) 1076 goto out; 1077 INP_HASH_WLOCK(pcbinfo); 1078 error = in_pcbconnect(inp, (struct sockaddr *)&sin, 1079 td->td_ucred); 1080 INP_HASH_WUNLOCK(pcbinfo); 1081 if (error == 0) 1082 soisconnected(so); 1083 goto out; 1084 } 1085 #endif 1086 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 1087 error = EISCONN; 1088 goto out; 1089 } 1090 inp->inp_vflag &= ~INP_IPV4; 1091 inp->inp_vflag |= INP_IPV6; 1092 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr); 1093 if (error != 0) 1094 goto out; 1095 INP_HASH_WLOCK(pcbinfo); 1096 error = in6_pcbconnect(inp, nam, td->td_ucred); 1097 INP_HASH_WUNLOCK(pcbinfo); 1098 if (error == 0) 1099 soisconnected(so); 1100 out: 1101 INP_WUNLOCK(inp); 1102 return (error); 1103 } 1104 1105 static void 1106 udp6_detach(struct socket *so) 1107 { 1108 struct inpcb *inp; 1109 struct inpcbinfo *pcbinfo; 1110 struct udpcb *up; 1111 1112 pcbinfo = get_inpcbinfo(so->so_proto->pr_protocol); 1113 inp = sotoinpcb(so); 1114 KASSERT(inp != NULL, ("udp6_detach: inp == NULL")); 1115 1116 INP_INFO_WLOCK(pcbinfo); 1117 INP_WLOCK(inp); 1118 up = intoudpcb(inp); 1119 KASSERT(up != NULL, ("%s: up == NULL", __func__)); 1120 in_pcbdetach(inp); 1121 in_pcbfree(inp); 1122 INP_INFO_WUNLOCK(pcbinfo); 1123 udp_discardcb(up); 1124 } 1125 1126 static int 1127 udp6_disconnect(struct socket *so) 1128 { 1129 struct inpcb *inp; 1130 struct inpcbinfo *pcbinfo; 1131 int error; 1132 1133 pcbinfo = get_inpcbinfo(so->so_proto->pr_protocol); 1134 inp = sotoinpcb(so); 1135 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL")); 1136 1137 #ifdef INET 1138 if (inp->inp_vflag & INP_IPV4) { 1139 struct pr_usrreqs *pru; 1140 1141 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs; 1142 (void)(*pru->pru_disconnect)(so); 1143 return (0); 1144 } 1145 #endif 1146 1147 INP_WLOCK(inp); 1148 1149 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 1150 error = ENOTCONN; 1151 goto out; 1152 } 1153 1154 INP_HASH_WLOCK(pcbinfo); 1155 in6_pcbdisconnect(inp); 1156 inp->in6p_laddr = in6addr_any; 1157 INP_HASH_WUNLOCK(pcbinfo); 1158 SOCK_LOCK(so); 1159 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 1160 SOCK_UNLOCK(so); 1161 out: 1162 INP_WUNLOCK(inp); 1163 return (0); 1164 } 1165 1166 static int 1167 udp6_send(struct socket *so, int flags, struct mbuf *m, 1168 struct sockaddr *addr, struct mbuf *control, struct thread *td) 1169 { 1170 struct inpcb *inp; 1171 struct inpcbinfo *pcbinfo; 1172 int error = 0; 1173 1174 pcbinfo = get_inpcbinfo(so->so_proto->pr_protocol); 1175 inp = sotoinpcb(so); 1176 KASSERT(inp != NULL, ("udp6_send: inp == NULL")); 1177 1178 INP_WLOCK(inp); 1179 if (addr) { 1180 if (addr->sa_len != sizeof(struct sockaddr_in6)) { 1181 error = EINVAL; 1182 goto bad; 1183 } 1184 if (addr->sa_family != AF_INET6) { 1185 error = EAFNOSUPPORT; 1186 goto bad; 1187 } 1188 } 1189 1190 #ifdef INET 1191 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 1192 int hasv4addr; 1193 struct sockaddr_in6 *sin6 = 0; 1194 1195 if (addr == 0) 1196 hasv4addr = (inp->inp_vflag & INP_IPV4); 1197 else { 1198 sin6 = (struct sockaddr_in6 *)addr; 1199 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) 1200 ? 1 : 0; 1201 } 1202 if (hasv4addr) { 1203 struct pr_usrreqs *pru; 1204 1205 /* 1206 * XXXRW: We release UDP-layer locks before calling 1207 * udp_send() in order to avoid recursion. However, 1208 * this does mean there is a short window where inp's 1209 * fields are unstable. Could this lead to a 1210 * potential race in which the factors causing us to 1211 * select the UDPv4 output routine are invalidated? 1212 */ 1213 INP_WUNLOCK(inp); 1214 if (sin6) 1215 in6_sin6_2_sin_in_sock(addr); 1216 pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs; 1217 /* addr will just be freed in sendit(). */ 1218 return ((*pru->pru_send)(so, flags, m, addr, control, 1219 td)); 1220 } 1221 } 1222 #endif 1223 #ifdef MAC 1224 mac_inpcb_create_mbuf(inp, m); 1225 #endif 1226 INP_HASH_WLOCK(pcbinfo); 1227 error = udp6_output(inp, m, addr, control, td); 1228 INP_HASH_WUNLOCK(pcbinfo); 1229 #ifdef INET 1230 #endif 1231 INP_WUNLOCK(inp); 1232 return (error); 1233 1234 bad: 1235 INP_WUNLOCK(inp); 1236 m_freem(m); 1237 return (error); 1238 } 1239 1240 struct pr_usrreqs udp6_usrreqs = { 1241 .pru_abort = udp6_abort, 1242 .pru_attach = udp6_attach, 1243 .pru_bind = udp6_bind, 1244 .pru_connect = udp6_connect, 1245 .pru_control = in6_control, 1246 .pru_detach = udp6_detach, 1247 .pru_disconnect = udp6_disconnect, 1248 .pru_peeraddr = in6_mapped_peeraddr, 1249 .pru_send = udp6_send, 1250 .pru_shutdown = udp_shutdown, 1251 .pru_sockaddr = in6_mapped_sockaddr, 1252 .pru_soreceive = soreceive_dgram, 1253 .pru_sosend = sosend_dgram, 1254 .pru_sosetlabel = in_pcbsosetlabel, 1255 .pru_close = udp6_close 1256 }; 1257