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); 154 return; 155 } 156 #ifdef IPSEC 157 /* Check AH/ESP integrity. */ 158 if (ipsec6_in_reject(n, inp)) { 159 m_freem(n); 160 IPSEC6STAT_INC(ips_in_polvio); 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 if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) { 211 /* XXX send icmp6 host/port unreach? */ 212 m_freem(m); 213 return (IPPROTO_DONE); 214 } 215 216 #ifndef PULLDOWN_TEST 217 IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE); 218 ip6 = mtod(m, struct ip6_hdr *); 219 uh = (struct udphdr *)((caddr_t)ip6 + off); 220 #else 221 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh)); 222 if (!uh) 223 return (IPPROTO_DONE); 224 #endif 225 226 UDPSTAT_INC(udps_ipackets); 227 228 /* 229 * Destination port of 0 is illegal, based on RFC768. 230 */ 231 if (uh->uh_dport == 0) 232 goto badunlocked; 233 234 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6); 235 ulen = ntohs((u_short)uh->uh_ulen); 236 237 nxt = ip6->ip6_nxt; 238 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0; 239 if (nxt == IPPROTO_UDPLITE) { 240 /* Zero means checksum over the complete packet. */ 241 if (ulen == 0) 242 ulen = plen; 243 if (ulen == plen) 244 cscov_partial = 0; 245 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) { 246 /* XXX: What is the right UDPLite MIB counter? */ 247 goto badunlocked; 248 } 249 if (uh->uh_sum == 0) { 250 /* XXX: What is the right UDPLite MIB counter? */ 251 goto badunlocked; 252 } 253 } else { 254 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) { 255 UDPSTAT_INC(udps_badlen); 256 goto badunlocked; 257 } 258 if (uh->uh_sum == 0) { 259 UDPSTAT_INC(udps_nosum); 260 goto badunlocked; 261 } 262 } 263 264 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) && 265 !cscov_partial) { 266 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) 267 uh_sum = m->m_pkthdr.csum_data; 268 else 269 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt, 270 m->m_pkthdr.csum_data); 271 uh_sum ^= 0xffff; 272 } else 273 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen); 274 275 if (uh_sum != 0) { 276 UDPSTAT_INC(udps_badsum); 277 goto badunlocked; 278 } 279 280 /* 281 * Construct sockaddr format source address. 282 */ 283 init_sin6(&fromsa, m); 284 fromsa.sin6_port = uh->uh_sport; 285 286 pcbinfo = get_inpcbinfo(nxt); 287 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 288 struct inpcb *last; 289 struct inpcbhead *pcblist; 290 struct ip6_moptions *imo; 291 292 INP_INFO_RLOCK(pcbinfo); 293 /* 294 * In the event that laddr should be set to the link-local 295 * address (this happens in RIPng), the multicast address 296 * specified in the received packet will not match laddr. To 297 * handle this situation, matching is relaxed if the 298 * receiving interface is the same as one specified in the 299 * socket and if the destination multicast address matches 300 * one of the multicast groups specified in the socket. 301 */ 302 303 /* 304 * KAME note: traditionally we dropped udpiphdr from mbuf 305 * here. We need udphdr for IPsec processing so we do that 306 * later. 307 */ 308 pcblist = get_pcblist(nxt); 309 last = NULL; 310 LIST_FOREACH(inp, pcblist, inp_list) { 311 if ((inp->inp_vflag & INP_IPV6) == 0) 312 continue; 313 if (inp->inp_lport != uh->uh_dport) 314 continue; 315 if (inp->inp_fport != 0 && 316 inp->inp_fport != uh->uh_sport) 317 continue; 318 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 319 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, 320 &ip6->ip6_dst)) 321 continue; 322 } 323 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 324 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, 325 &ip6->ip6_src) || 326 inp->inp_fport != uh->uh_sport) 327 continue; 328 } 329 330 /* 331 * XXXRW: Because we weren't holding either the inpcb 332 * or the hash lock when we checked for a match 333 * before, we should probably recheck now that the 334 * inpcb lock is (supposed to be) held. 335 */ 336 337 /* 338 * Handle socket delivery policy for any-source 339 * and source-specific multicast. [RFC3678] 340 */ 341 imo = inp->in6p_moptions; 342 if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 343 struct sockaddr_in6 mcaddr; 344 int blocked; 345 346 INP_RLOCK(inp); 347 348 bzero(&mcaddr, sizeof(struct sockaddr_in6)); 349 mcaddr.sin6_len = sizeof(struct sockaddr_in6); 350 mcaddr.sin6_family = AF_INET6; 351 mcaddr.sin6_addr = ip6->ip6_dst; 352 353 blocked = im6o_mc_filter(imo, ifp, 354 (struct sockaddr *)&mcaddr, 355 (struct sockaddr *)&fromsa); 356 if (blocked != MCAST_PASS) { 357 if (blocked == MCAST_NOTGMEMBER) 358 IP6STAT_INC(ip6s_notmember); 359 if (blocked == MCAST_NOTSMEMBER || 360 blocked == MCAST_MUTED) 361 UDPSTAT_INC(udps_filtermcast); 362 INP_RUNLOCK(inp); /* XXX */ 363 continue; 364 } 365 366 INP_RUNLOCK(inp); 367 } 368 if (last != NULL) { 369 struct mbuf *n; 370 371 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { 372 INP_RLOCK(last); 373 UDP_PROBE(receive, NULL, last, ip6, 374 last, uh); 375 udp6_append(last, n, off, &fromsa); 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 udp6_append(last, m, off, &fromsa); 407 INP_RUNLOCK(last); 408 return (IPPROTO_DONE); 409 } 410 /* 411 * Locate pcb for datagram. 412 */ 413 414 /* 415 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. 416 */ 417 if ((m->m_flags & M_IP6_NEXTHOP) && 418 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 419 struct sockaddr_in6 *next_hop6; 420 421 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1); 422 423 /* 424 * Transparently forwarded. Pretend to be the destination. 425 * Already got one like this? 426 */ 427 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src, 428 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport, 429 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m); 430 if (!inp) { 431 /* 432 * It's new. Try to find the ambushing socket. 433 * Because we've rewritten the destination address, 434 * any hardware-generated hash is ignored. 435 */ 436 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src, 437 uh->uh_sport, &next_hop6->sin6_addr, 438 next_hop6->sin6_port ? htons(next_hop6->sin6_port) : 439 uh->uh_dport, INPLOOKUP_WILDCARD | 440 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif); 441 } 442 /* Remove the tag from the packet. We don't need it anymore. */ 443 m_tag_delete(m, fwd_tag); 444 m->m_flags &= ~M_IP6_NEXTHOP; 445 } else 446 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src, 447 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport, 448 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, 449 m->m_pkthdr.rcvif, m); 450 if (inp == NULL) { 451 if (udp_log_in_vain) { 452 char ip6bufs[INET6_ADDRSTRLEN]; 453 char ip6bufd[INET6_ADDRSTRLEN]; 454 455 log(LOG_INFO, 456 "Connection attempt to UDP [%s]:%d from [%s]:%d\n", 457 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 458 ntohs(uh->uh_dport), 459 ip6_sprintf(ip6bufs, &ip6->ip6_src), 460 ntohs(uh->uh_sport)); 461 } 462 UDPSTAT_INC(udps_noport); 463 if (m->m_flags & M_MCAST) { 464 printf("UDP6: M_MCAST is set in a unicast packet.\n"); 465 UDPSTAT_INC(udps_noportmcast); 466 goto badunlocked; 467 } 468 if (V_udp_blackhole) 469 goto badunlocked; 470 if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0) 471 goto badunlocked; 472 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); 473 return (IPPROTO_DONE); 474 } 475 INP_RLOCK_ASSERT(inp); 476 up = intoudpcb(inp); 477 if (cscov_partial) { 478 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) { 479 INP_RUNLOCK(inp); 480 m_freem(m); 481 return (IPPROTO_DONE); 482 } 483 } 484 UDP_PROBE(receive, NULL, inp, ip6, inp, uh); 485 udp6_append(inp, m, off, &fromsa); 486 INP_RUNLOCK(inp); 487 return (IPPROTO_DONE); 488 489 badheadlocked: 490 INP_INFO_RUNLOCK(pcbinfo); 491 badunlocked: 492 if (m) 493 m_freem(m); 494 return (IPPROTO_DONE); 495 } 496 497 static void 498 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d, 499 struct inpcbinfo *pcbinfo) 500 { 501 struct udphdr uh; 502 struct ip6_hdr *ip6; 503 struct mbuf *m; 504 int off = 0; 505 struct ip6ctlparam *ip6cp = NULL; 506 const struct sockaddr_in6 *sa6_src = NULL; 507 void *cmdarg; 508 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify; 509 struct udp_portonly { 510 u_int16_t uh_sport; 511 u_int16_t uh_dport; 512 } *uhp; 513 514 if (sa->sa_family != AF_INET6 || 515 sa->sa_len != sizeof(struct sockaddr_in6)) 516 return; 517 518 if ((unsigned)cmd >= PRC_NCMDS) 519 return; 520 if (PRC_IS_REDIRECT(cmd)) 521 notify = in6_rtchange, d = NULL; 522 else if (cmd == PRC_HOSTDEAD) 523 d = NULL; 524 else if (inet6ctlerrmap[cmd] == 0) 525 return; 526 527 /* if the parameter is from icmp6, decode it. */ 528 if (d != NULL) { 529 ip6cp = (struct ip6ctlparam *)d; 530 m = ip6cp->ip6c_m; 531 ip6 = ip6cp->ip6c_ip6; 532 off = ip6cp->ip6c_off; 533 cmdarg = ip6cp->ip6c_cmdarg; 534 sa6_src = ip6cp->ip6c_src; 535 } else { 536 m = NULL; 537 ip6 = NULL; 538 cmdarg = NULL; 539 sa6_src = &sa6_any; 540 } 541 542 if (ip6) { 543 /* 544 * XXX: We assume that when IPV6 is non NULL, 545 * M and OFF are valid. 546 */ 547 548 /* Check if we can safely examine src and dst ports. */ 549 if (m->m_pkthdr.len < off + sizeof(*uhp)) 550 return; 551 552 bzero(&uh, sizeof(uh)); 553 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh); 554 555 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport, 556 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd, 557 cmdarg, notify); 558 } else 559 (void)in6_pcbnotify(pcbinfo, sa, 0, 560 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify); 561 } 562 563 void 564 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d) 565 { 566 567 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo)); 568 } 569 570 void 571 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d) 572 { 573 574 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo)); 575 } 576 577 static int 578 udp6_getcred(SYSCTL_HANDLER_ARGS) 579 { 580 struct xucred xuc; 581 struct sockaddr_in6 addrs[2]; 582 struct inpcb *inp; 583 int error; 584 585 error = priv_check(req->td, PRIV_NETINET_GETCRED); 586 if (error) 587 return (error); 588 589 if (req->newlen != sizeof(addrs)) 590 return (EINVAL); 591 if (req->oldlen != sizeof(struct xucred)) 592 return (EINVAL); 593 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 594 if (error) 595 return (error); 596 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 || 597 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) { 598 return (error); 599 } 600 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr, 601 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 602 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL); 603 if (inp != NULL) { 604 INP_RLOCK_ASSERT(inp); 605 if (inp->inp_socket == NULL) 606 error = ENOENT; 607 if (error == 0) 608 error = cr_canseesocket(req->td->td_ucred, 609 inp->inp_socket); 610 if (error == 0) 611 cru2x(inp->inp_cred, &xuc); 612 INP_RUNLOCK(inp); 613 } else 614 error = ENOENT; 615 if (error == 0) 616 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); 617 return (error); 618 } 619 620 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0, 621 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection"); 622 623 static int 624 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6, 625 struct mbuf *control, struct thread *td) 626 { 627 u_int32_t ulen = m->m_pkthdr.len; 628 u_int32_t plen = sizeof(struct udphdr) + ulen; 629 struct ip6_hdr *ip6; 630 struct udphdr *udp6; 631 struct in6_addr *laddr, *faddr, in6a; 632 struct sockaddr_in6 *sin6 = NULL; 633 struct ifnet *oifp = NULL; 634 int cscov_partial = 0; 635 int scope_ambiguous = 0; 636 u_short fport; 637 int error = 0; 638 uint8_t nxt; 639 uint16_t cscov = 0; 640 struct ip6_pktopts *optp, opt; 641 int af = AF_INET6, hlen = sizeof(struct ip6_hdr); 642 int flags; 643 struct sockaddr_in6 tmp; 644 645 INP_WLOCK_ASSERT(inp); 646 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); 647 648 if (addr6) { 649 /* addr6 has been validated in udp6_send(). */ 650 sin6 = (struct sockaddr_in6 *)addr6; 651 652 /* protect *sin6 from overwrites */ 653 tmp = *sin6; 654 sin6 = &tmp; 655 656 /* 657 * Application should provide a proper zone ID or the use of 658 * default zone IDs should be enabled. Unfortunately, some 659 * applications do not behave as it should, so we need a 660 * workaround. Even if an appropriate ID is not determined, 661 * we'll see if we can determine the outgoing interface. If we 662 * can, determine the zone ID based on the interface below. 663 */ 664 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) 665 scope_ambiguous = 1; 666 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) 667 return (error); 668 } 669 670 if (control) { 671 if ((error = ip6_setpktopts(control, &opt, 672 inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0) 673 goto release; 674 optp = &opt; 675 } else 676 optp = inp->in6p_outputopts; 677 678 if (sin6) { 679 faddr = &sin6->sin6_addr; 680 681 /* 682 * Since we saw no essential reason for calling in_pcbconnect, 683 * we get rid of such kind of logic, and call in6_selectsrc 684 * and in6_pcbsetport in order to fill in the local address 685 * and the local port. 686 */ 687 if (sin6->sin6_port == 0) { 688 error = EADDRNOTAVAIL; 689 goto release; 690 } 691 692 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 693 /* how about ::ffff:0.0.0.0 case? */ 694 error = EISCONN; 695 goto release; 696 } 697 698 fport = sin6->sin6_port; /* allow 0 port */ 699 700 if (IN6_IS_ADDR_V4MAPPED(faddr)) { 701 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) { 702 /* 703 * I believe we should explicitly discard the 704 * packet when mapped addresses are disabled, 705 * rather than send the packet as an IPv6 one. 706 * If we chose the latter approach, the packet 707 * might be sent out on the wire based on the 708 * default route, the situation which we'd 709 * probably want to avoid. 710 * (20010421 jinmei@kame.net) 711 */ 712 error = EINVAL; 713 goto release; 714 } 715 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) && 716 !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) { 717 /* 718 * when remote addr is an IPv4-mapped address, 719 * local addr should not be an IPv6 address, 720 * since you cannot determine how to map IPv6 721 * source address to IPv4. 722 */ 723 error = EINVAL; 724 goto release; 725 } 726 727 af = AF_INET; 728 } 729 730 if (!IN6_IS_ADDR_V4MAPPED(faddr)) { 731 error = in6_selectsrc(sin6, optp, inp, NULL, 732 td->td_ucred, &oifp, &in6a); 733 if (error) 734 goto release; 735 if (oifp && scope_ambiguous && 736 (error = in6_setscope(&sin6->sin6_addr, 737 oifp, NULL))) { 738 goto release; 739 } 740 laddr = &in6a; 741 } else 742 laddr = &inp->in6p_laddr; /* XXX */ 743 if (laddr == NULL) { 744 if (error == 0) 745 error = EADDRNOTAVAIL; 746 goto release; 747 } 748 if (inp->inp_lport == 0 && 749 (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) { 750 /* Undo an address bind that may have occurred. */ 751 inp->in6p_laddr = in6addr_any; 752 goto release; 753 } 754 } else { 755 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 756 error = ENOTCONN; 757 goto release; 758 } 759 if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) { 760 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) { 761 /* 762 * XXX: this case would happen when the 763 * application sets the V6ONLY flag after 764 * connecting the foreign address. 765 * Such applications should be fixed, 766 * so we bark here. 767 */ 768 log(LOG_INFO, "udp6_output: IPV6_V6ONLY " 769 "option was set for a connected socket\n"); 770 error = EINVAL; 771 goto release; 772 } else 773 af = AF_INET; 774 } 775 laddr = &inp->in6p_laddr; 776 faddr = &inp->in6p_faddr; 777 fport = inp->inp_fport; 778 } 779 780 if (af == AF_INET) 781 hlen = sizeof(struct ip); 782 783 /* 784 * Calculate data length and get a mbuf 785 * for UDP and IP6 headers. 786 */ 787 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT); 788 if (m == 0) { 789 error = ENOBUFS; 790 goto release; 791 } 792 793 /* 794 * Stuff checksum and output datagram. 795 */ 796 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ? 797 IPPROTO_UDP : IPPROTO_UDPLITE; 798 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen); 799 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */ 800 udp6->uh_dport = fport; 801 if (nxt == IPPROTO_UDPLITE) { 802 struct udpcb *up; 803 804 up = intoudpcb(inp); 805 cscov = up->u_txcslen; 806 if (cscov >= plen) 807 cscov = 0; 808 udp6->uh_ulen = htons(cscov); 809 /* 810 * For UDP-Lite, checksum coverage length of zero means 811 * the entire UDPLite packet is covered by the checksum. 812 */ 813 cscov_partial = (cscov == 0) ? 0 : 1; 814 } else if (plen <= 0xffff) 815 udp6->uh_ulen = htons((u_short)plen); 816 else 817 udp6->uh_ulen = 0; 818 udp6->uh_sum = 0; 819 820 switch (af) { 821 case AF_INET6: 822 ip6 = mtod(m, struct ip6_hdr *); 823 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK; 824 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 825 ip6->ip6_vfc |= IPV6_VERSION; 826 ip6->ip6_plen = htons((u_short)plen); 827 ip6->ip6_nxt = nxt; 828 ip6->ip6_hlim = in6_selecthlim(inp, NULL); 829 ip6->ip6_src = *laddr; 830 ip6->ip6_dst = *faddr; 831 832 if (cscov_partial) { 833 if ((udp6->uh_sum = in6_cksum_partial(m, nxt, 834 sizeof(struct ip6_hdr), plen, cscov)) == 0) 835 udp6->uh_sum = 0xffff; 836 } else { 837 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0); 838 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6; 839 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 840 } 841 842 /* 843 * XXX for now assume UDP is 2-tuple. 844 * Later on this may become configurable as 4-tuple; 845 * we should support that. 846 * 847 * XXX .. and we should likely cache this in the inpcb. 848 */ 849 #ifdef RSS 850 m->m_pkthdr.flowid = rss_hash_ip6_2tuple(*faddr, *laddr); 851 m->m_flags |= M_FLOWID; 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