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