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