1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * Copyright (c) 2010-2011 Juniper Networks, Inc. 4 * All rights reserved. 5 * 6 * Portions of this software were developed by Robert N. M. Watson under 7 * contract to Juniper Networks, Inc. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. Neither the name of the project nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $ 34 */ 35 36 /*- 37 * Copyright (c) 1982, 1986, 1991, 1993 38 * The Regents of the University of California. All rights reserved. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 4. Neither the name of the University nor the names of its contributors 49 * may be used to endorse or promote products derived from this software 50 * without specific prior written permission. 51 * 52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 62 * SUCH DAMAGE. 63 * 64 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 65 */ 66 67 #include <sys/cdefs.h> 68 __FBSDID("$FreeBSD$"); 69 70 #include "opt_inet.h" 71 #include "opt_inet6.h" 72 #include "opt_ipsec.h" 73 #include "opt_pcbgroup.h" 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/malloc.h> 78 #include <sys/mbuf.h> 79 #include <sys/domain.h> 80 #include <sys/protosw.h> 81 #include <sys/socket.h> 82 #include <sys/socketvar.h> 83 #include <sys/sockio.h> 84 #include <sys/errno.h> 85 #include <sys/time.h> 86 #include <sys/priv.h> 87 #include <sys/proc.h> 88 #include <sys/jail.h> 89 90 #include <vm/uma.h> 91 92 #include <net/if.h> 93 #include <net/if_types.h> 94 #include <net/route.h> 95 96 #include <netinet/in.h> 97 #include <netinet/in_var.h> 98 #include <netinet/in_systm.h> 99 #include <netinet/tcp_var.h> 100 #include <netinet/ip6.h> 101 #include <netinet/ip_var.h> 102 103 #include <netinet6/ip6_var.h> 104 #include <netinet6/nd6.h> 105 #include <netinet/in_pcb.h> 106 #include <netinet6/in6_pcb.h> 107 #include <netinet6/scope6_var.h> 108 109 struct in6_addr zeroin6_addr; 110 111 int 112 in6_pcbbind(register struct inpcb *inp, struct sockaddr *nam, 113 struct ucred *cred) 114 { 115 struct socket *so = inp->inp_socket; 116 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL; 117 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 118 u_short lport = 0; 119 int error, lookupflags = 0; 120 int reuseport = (so->so_options & SO_REUSEPORT); 121 122 INP_WLOCK_ASSERT(inp); 123 INP_HASH_WLOCK_ASSERT(pcbinfo); 124 125 if (TAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */ 126 return (EADDRNOTAVAIL); 127 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 128 return (EINVAL); 129 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 130 lookupflags = INPLOOKUP_WILDCARD; 131 if (nam == NULL) { 132 if ((error = prison_local_ip6(cred, &inp->in6p_laddr, 133 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 134 return (error); 135 } else { 136 sin6 = (struct sockaddr_in6 *)nam; 137 if (nam->sa_len != sizeof(*sin6)) 138 return (EINVAL); 139 /* 140 * family check. 141 */ 142 if (nam->sa_family != AF_INET6) 143 return (EAFNOSUPPORT); 144 145 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) 146 return(error); 147 148 if ((error = prison_local_ip6(cred, &sin6->sin6_addr, 149 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 150 return (error); 151 152 lport = sin6->sin6_port; 153 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 154 /* 155 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 156 * allow compepte duplication of binding if 157 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 158 * and a multicast address is bound on both 159 * new and duplicated sockets. 160 */ 161 if (so->so_options & SO_REUSEADDR) 162 reuseport = SO_REUSEADDR|SO_REUSEPORT; 163 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 164 struct ifaddr *ifa; 165 166 sin6->sin6_port = 0; /* yech... */ 167 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) == 168 NULL && 169 (inp->inp_flags & INP_BINDANY) == 0) { 170 return (EADDRNOTAVAIL); 171 } 172 173 /* 174 * XXX: bind to an anycast address might accidentally 175 * cause sending a packet with anycast source address. 176 * We should allow to bind to a deprecated address, since 177 * the application dares to use it. 178 */ 179 if (ifa != NULL && 180 ((struct in6_ifaddr *)ifa)->ia6_flags & 181 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) { 182 ifa_free(ifa); 183 return (EADDRNOTAVAIL); 184 } 185 if (ifa != NULL) 186 ifa_free(ifa); 187 } 188 if (lport) { 189 struct inpcb *t; 190 struct tcptw *tw; 191 192 /* GROSS */ 193 if (ntohs(lport) <= V_ipport_reservedhigh && 194 ntohs(lport) >= V_ipport_reservedlow && 195 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 196 0)) 197 return (EACCES); 198 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) && 199 priv_check_cred(inp->inp_cred, 200 PRIV_NETINET_REUSEPORT, 0) != 0) { 201 t = in6_pcblookup_local(pcbinfo, 202 &sin6->sin6_addr, lport, 203 INPLOOKUP_WILDCARD, cred); 204 if (t && 205 ((t->inp_flags & INP_TIMEWAIT) == 0) && 206 (so->so_type != SOCK_STREAM || 207 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) && 208 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || 209 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) || 210 (t->inp_flags2 & INP_REUSEPORT) == 0) && 211 (inp->inp_cred->cr_uid != 212 t->inp_cred->cr_uid)) 213 return (EADDRINUSE); 214 #ifdef INET 215 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && 216 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 217 struct sockaddr_in sin; 218 219 in6_sin6_2_sin(&sin, sin6); 220 t = in_pcblookup_local(pcbinfo, 221 sin.sin_addr, lport, 222 INPLOOKUP_WILDCARD, cred); 223 if (t && 224 ((t->inp_flags & 225 INP_TIMEWAIT) == 0) && 226 (so->so_type != SOCK_STREAM || 227 ntohl(t->inp_faddr.s_addr) == 228 INADDR_ANY) && 229 (inp->inp_cred->cr_uid != 230 t->inp_cred->cr_uid)) 231 return (EADDRINUSE); 232 } 233 #endif 234 } 235 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr, 236 lport, lookupflags, cred); 237 if (t && (t->inp_flags & INP_TIMEWAIT)) { 238 /* 239 * XXXRW: If an incpb has had its timewait 240 * state recycled, we treat the address as 241 * being in use (for now). This is better 242 * than a panic, but not desirable. 243 */ 244 tw = intotw(t); 245 if (tw == NULL || 246 (reuseport & tw->tw_so_options) == 0) 247 return (EADDRINUSE); 248 } else if (t && (reuseport == 0 || 249 (t->inp_flags2 & INP_REUSEPORT) == 0)) { 250 return (EADDRINUSE); 251 } 252 #ifdef INET 253 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && 254 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 255 struct sockaddr_in sin; 256 257 in6_sin6_2_sin(&sin, sin6); 258 t = in_pcblookup_local(pcbinfo, sin.sin_addr, 259 lport, lookupflags, cred); 260 if (t && t->inp_flags & INP_TIMEWAIT) { 261 tw = intotw(t); 262 if (tw == NULL) 263 return (EADDRINUSE); 264 if ((reuseport & tw->tw_so_options) == 0 265 && (ntohl(t->inp_laddr.s_addr) != 266 INADDR_ANY || ((inp->inp_vflag & 267 INP_IPV6PROTO) == 268 (t->inp_vflag & INP_IPV6PROTO)))) 269 return (EADDRINUSE); 270 } else if (t && (reuseport == 0 || 271 (t->inp_flags2 & INP_REUSEPORT) == 0) && 272 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 273 (t->inp_vflag & INP_IPV6PROTO) != 0)) 274 return (EADDRINUSE); 275 } 276 #endif 277 } 278 inp->in6p_laddr = sin6->sin6_addr; 279 } 280 if (lport == 0) { 281 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) { 282 /* Undo an address bind that may have occurred. */ 283 inp->in6p_laddr = in6addr_any; 284 return (error); 285 } 286 } else { 287 inp->inp_lport = lport; 288 if (in_pcbinshash(inp) != 0) { 289 inp->in6p_laddr = in6addr_any; 290 inp->inp_lport = 0; 291 return (EAGAIN); 292 } 293 } 294 return (0); 295 } 296 297 /* 298 * Transform old in6_pcbconnect() into an inner subroutine for new 299 * in6_pcbconnect(): Do some validity-checking on the remote 300 * address (in mbuf 'nam') and then determine local host address 301 * (i.e., which interface) to use to access that remote host. 302 * 303 * This preserves definition of in6_pcbconnect(), while supporting a 304 * slightly different version for T/TCP. (This is more than 305 * a bit of a kludge, but cleaning up the internal interfaces would 306 * have forced minor changes in every protocol). 307 */ 308 int 309 in6_pcbladdr(register struct inpcb *inp, struct sockaddr *nam, 310 struct in6_addr *plocal_addr6) 311 { 312 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 313 int error = 0; 314 struct ifnet *ifp = NULL; 315 int scope_ambiguous = 0; 316 struct in6_addr in6a; 317 318 INP_WLOCK_ASSERT(inp); 319 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */ 320 321 if (nam->sa_len != sizeof (*sin6)) 322 return (EINVAL); 323 if (sin6->sin6_family != AF_INET6) 324 return (EAFNOSUPPORT); 325 if (sin6->sin6_port == 0) 326 return (EADDRNOTAVAIL); 327 328 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) 329 scope_ambiguous = 1; 330 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) 331 return(error); 332 333 if (!TAILQ_EMPTY(&V_in6_ifaddrhead)) { 334 /* 335 * If the destination address is UNSPECIFIED addr, 336 * use the loopback addr, e.g ::1. 337 */ 338 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 339 sin6->sin6_addr = in6addr_loopback; 340 } 341 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0) 342 return (error); 343 344 error = in6_selectsrc(sin6, inp->in6p_outputopts, 345 inp, NULL, inp->inp_cred, &ifp, &in6a); 346 if (error) 347 return (error); 348 349 if (ifp && scope_ambiguous && 350 (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) { 351 return(error); 352 } 353 354 /* 355 * Do not update this earlier, in case we return with an error. 356 * 357 * XXX: this in6_selectsrc result might replace the bound local 358 * address with the address specified by setsockopt(IPV6_PKTINFO). 359 * Is it the intended behavior? 360 */ 361 *plocal_addr6 = in6a; 362 363 /* 364 * Don't do pcblookup call here; return interface in 365 * plocal_addr6 366 * and exit to caller, that will do the lookup. 367 */ 368 369 return (0); 370 } 371 372 /* 373 * Outer subroutine: 374 * Connect from a socket to a specified address. 375 * Both address and port must be specified in argument sin. 376 * If don't have a local address for this socket yet, 377 * then pick one. 378 */ 379 int 380 in6_pcbconnect_mbuf(register struct inpcb *inp, struct sockaddr *nam, 381 struct ucred *cred, struct mbuf *m) 382 { 383 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 384 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 385 struct in6_addr addr6; 386 int error; 387 388 INP_WLOCK_ASSERT(inp); 389 INP_HASH_WLOCK_ASSERT(pcbinfo); 390 391 /* 392 * Call inner routine, to assign local interface address. 393 * in6_pcbladdr() may automatically fill in sin6_scope_id. 394 */ 395 if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0) 396 return (error); 397 398 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr, 399 sin6->sin6_port, 400 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) 401 ? &addr6 : &inp->in6p_laddr, 402 inp->inp_lport, 0, NULL) != NULL) { 403 return (EADDRINUSE); 404 } 405 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 406 if (inp->inp_lport == 0) { 407 error = in6_pcbbind(inp, (struct sockaddr *)0, cred); 408 if (error) 409 return (error); 410 } 411 inp->in6p_laddr = addr6; 412 } 413 inp->in6p_faddr = sin6->sin6_addr; 414 inp->inp_fport = sin6->sin6_port; 415 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ 416 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; 417 if (inp->inp_flags & IN6P_AUTOFLOWLABEL) 418 inp->inp_flow |= 419 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); 420 421 in_pcbrehash_mbuf(inp, m); 422 423 return (0); 424 } 425 426 int 427 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 428 { 429 430 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL)); 431 } 432 433 void 434 in6_pcbdisconnect(struct inpcb *inp) 435 { 436 437 INP_WLOCK_ASSERT(inp); 438 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); 439 440 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr)); 441 inp->inp_fport = 0; 442 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ 443 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; 444 in_pcbrehash(inp); 445 } 446 447 struct sockaddr * 448 in6_sockaddr(in_port_t port, struct in6_addr *addr_p) 449 { 450 struct sockaddr_in6 *sin6; 451 452 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK); 453 bzero(sin6, sizeof *sin6); 454 sin6->sin6_family = AF_INET6; 455 sin6->sin6_len = sizeof(*sin6); 456 sin6->sin6_port = port; 457 sin6->sin6_addr = *addr_p; 458 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */ 459 460 return (struct sockaddr *)sin6; 461 } 462 463 struct sockaddr * 464 in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p) 465 { 466 struct sockaddr_in sin; 467 struct sockaddr_in6 *sin6_p; 468 469 bzero(&sin, sizeof sin); 470 sin.sin_family = AF_INET; 471 sin.sin_len = sizeof(sin); 472 sin.sin_port = port; 473 sin.sin_addr = *addr_p; 474 475 sin6_p = malloc(sizeof *sin6_p, M_SONAME, 476 M_WAITOK); 477 in6_sin_2_v4mapsin6(&sin, sin6_p); 478 479 return (struct sockaddr *)sin6_p; 480 } 481 482 int 483 in6_getsockaddr(struct socket *so, struct sockaddr **nam) 484 { 485 register struct inpcb *inp; 486 struct in6_addr addr; 487 in_port_t port; 488 489 inp = sotoinpcb(so); 490 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL")); 491 492 INP_RLOCK(inp); 493 port = inp->inp_lport; 494 addr = inp->in6p_laddr; 495 INP_RUNLOCK(inp); 496 497 *nam = in6_sockaddr(port, &addr); 498 return 0; 499 } 500 501 int 502 in6_getpeeraddr(struct socket *so, struct sockaddr **nam) 503 { 504 struct inpcb *inp; 505 struct in6_addr addr; 506 in_port_t port; 507 508 inp = sotoinpcb(so); 509 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL")); 510 511 INP_RLOCK(inp); 512 port = inp->inp_fport; 513 addr = inp->in6p_faddr; 514 INP_RUNLOCK(inp); 515 516 *nam = in6_sockaddr(port, &addr); 517 return 0; 518 } 519 520 int 521 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam) 522 { 523 struct inpcb *inp; 524 int error; 525 526 inp = sotoinpcb(so); 527 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL")); 528 529 #ifdef INET 530 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { 531 error = in_getsockaddr(so, nam); 532 if (error == 0) 533 in6_sin_2_v4mapsin6_in_sock(nam); 534 } else 535 #endif 536 { 537 /* scope issues will be handled in in6_getsockaddr(). */ 538 error = in6_getsockaddr(so, nam); 539 } 540 541 return error; 542 } 543 544 int 545 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam) 546 { 547 struct inpcb *inp; 548 int error; 549 550 inp = sotoinpcb(so); 551 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL")); 552 553 #ifdef INET 554 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { 555 error = in_getpeeraddr(so, nam); 556 if (error == 0) 557 in6_sin_2_v4mapsin6_in_sock(nam); 558 } else 559 #endif 560 /* scope issues will be handled in in6_getpeeraddr(). */ 561 error = in6_getpeeraddr(so, nam); 562 563 return error; 564 } 565 566 /* 567 * Pass some notification to all connections of a protocol 568 * associated with address dst. The local address and/or port numbers 569 * may be specified to limit the search. The "usual action" will be 570 * taken, depending on the ctlinput cmd. The caller must filter any 571 * cmds that are uninteresting (e.g., no error in the map). 572 * Call the protocol specific routine (if any) to report 573 * any errors for each matching socket. 574 */ 575 void 576 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, 577 u_int fport_arg, const struct sockaddr *src, u_int lport_arg, 578 int cmd, void *cmdarg, 579 struct inpcb *(*notify)(struct inpcb *, int)) 580 { 581 struct inpcb *inp, *inp_temp; 582 struct sockaddr_in6 sa6_src, *sa6_dst; 583 u_short fport = fport_arg, lport = lport_arg; 584 u_int32_t flowinfo; 585 int errno; 586 587 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6) 588 return; 589 590 sa6_dst = (struct sockaddr_in6 *)dst; 591 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr)) 592 return; 593 594 /* 595 * note that src can be NULL when we get notify by local fragmentation. 596 */ 597 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src; 598 flowinfo = sa6_src.sin6_flowinfo; 599 600 /* 601 * Redirects go to all references to the destination, 602 * and use in6_rtchange to invalidate the route cache. 603 * Dead host indications: also use in6_rtchange to invalidate 604 * the cache, and deliver the error to all the sockets. 605 * Otherwise, if we have knowledge of the local port and address, 606 * deliver only to that socket. 607 */ 608 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) { 609 fport = 0; 610 lport = 0; 611 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr)); 612 613 if (cmd != PRC_HOSTDEAD) 614 notify = in6_rtchange; 615 } 616 errno = inet6ctlerrmap[cmd]; 617 INP_INFO_WLOCK(pcbinfo); 618 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { 619 INP_WLOCK(inp); 620 if ((inp->inp_vflag & INP_IPV6) == 0) { 621 INP_WUNLOCK(inp); 622 continue; 623 } 624 625 /* 626 * If the error designates a new path MTU for a destination 627 * and the application (associated with this socket) wanted to 628 * know the value, notify. Note that we notify for all 629 * disconnected sockets if the corresponding application 630 * wanted. This is because some UDP applications keep sending 631 * sockets disconnected. 632 * XXX: should we avoid to notify the value to TCP sockets? 633 */ 634 if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 && 635 (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 636 IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) { 637 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, 638 (u_int32_t *)cmdarg); 639 } 640 641 /* 642 * Detect if we should notify the error. If no source and 643 * destination ports are specifed, but non-zero flowinfo and 644 * local address match, notify the error. This is the case 645 * when the error is delivered with an encrypted buffer 646 * by ESP. Otherwise, just compare addresses and ports 647 * as usual. 648 */ 649 if (lport == 0 && fport == 0 && flowinfo && 650 inp->inp_socket != NULL && 651 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) && 652 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr)) 653 goto do_notify; 654 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, 655 &sa6_dst->sin6_addr) || 656 inp->inp_socket == 0 || 657 (lport && inp->inp_lport != lport) || 658 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) && 659 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, 660 &sa6_src.sin6_addr)) || 661 (fport && inp->inp_fport != fport)) { 662 INP_WUNLOCK(inp); 663 continue; 664 } 665 666 do_notify: 667 if (notify) { 668 if ((*notify)(inp, errno)) 669 INP_WUNLOCK(inp); 670 } else 671 INP_WUNLOCK(inp); 672 } 673 INP_INFO_WUNLOCK(pcbinfo); 674 } 675 676 /* 677 * Lookup a PCB based on the local address and port. Caller must hold the 678 * hash lock. No inpcb locks or references are acquired. 679 */ 680 struct inpcb * 681 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr, 682 u_short lport, int lookupflags, struct ucred *cred) 683 { 684 register struct inpcb *inp; 685 int matchwild = 3, wildcard; 686 687 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, 688 ("%s: invalid lookup flags %d", __func__, lookupflags)); 689 690 INP_HASH_WLOCK_ASSERT(pcbinfo); 691 692 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { 693 struct inpcbhead *head; 694 /* 695 * Look for an unconnected (wildcard foreign addr) PCB that 696 * matches the local address and port we're looking for. 697 */ 698 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 699 0, pcbinfo->ipi_hashmask)]; 700 LIST_FOREACH(inp, head, inp_hash) { 701 /* XXX inp locking */ 702 if ((inp->inp_vflag & INP_IPV6) == 0) 703 continue; 704 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 705 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 706 inp->inp_lport == lport) { 707 /* Found. */ 708 if (cred == NULL || 709 prison_equal_ip6(cred->cr_prison, 710 inp->inp_cred->cr_prison)) 711 return (inp); 712 } 713 } 714 /* 715 * Not found. 716 */ 717 return (NULL); 718 } else { 719 struct inpcbporthead *porthash; 720 struct inpcbport *phd; 721 struct inpcb *match = NULL; 722 /* 723 * Best fit PCB lookup. 724 * 725 * First see if this local port is in use by looking on the 726 * port hash list. 727 */ 728 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, 729 pcbinfo->ipi_porthashmask)]; 730 LIST_FOREACH(phd, porthash, phd_hash) { 731 if (phd->phd_port == lport) 732 break; 733 } 734 if (phd != NULL) { 735 /* 736 * Port is in use by one or more PCBs. Look for best 737 * fit. 738 */ 739 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 740 wildcard = 0; 741 if (cred != NULL && 742 !prison_equal_ip6(cred->cr_prison, 743 inp->inp_cred->cr_prison)) 744 continue; 745 /* XXX inp locking */ 746 if ((inp->inp_vflag & INP_IPV6) == 0) 747 continue; 748 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) 749 wildcard++; 750 if (!IN6_IS_ADDR_UNSPECIFIED( 751 &inp->in6p_laddr)) { 752 if (IN6_IS_ADDR_UNSPECIFIED(laddr)) 753 wildcard++; 754 else if (!IN6_ARE_ADDR_EQUAL( 755 &inp->in6p_laddr, laddr)) 756 continue; 757 } else { 758 if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) 759 wildcard++; 760 } 761 if (wildcard < matchwild) { 762 match = inp; 763 matchwild = wildcard; 764 if (matchwild == 0) 765 break; 766 } 767 } 768 } 769 return (match); 770 } 771 } 772 773 void 774 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) 775 { 776 struct inpcb *in6p; 777 struct ip6_moptions *im6o; 778 int i, gap; 779 780 INP_INFO_RLOCK(pcbinfo); 781 LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) { 782 INP_WLOCK(in6p); 783 im6o = in6p->in6p_moptions; 784 if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) { 785 /* 786 * Unselect the outgoing ifp for multicast if it 787 * is being detached. 788 */ 789 if (im6o->im6o_multicast_ifp == ifp) 790 im6o->im6o_multicast_ifp = NULL; 791 /* 792 * Drop multicast group membership if we joined 793 * through the interface being detached. 794 */ 795 gap = 0; 796 for (i = 0; i < im6o->im6o_num_memberships; i++) { 797 if (im6o->im6o_membership[i]->in6m_ifp == 798 ifp) { 799 in6_mc_leave(im6o->im6o_membership[i], 800 NULL); 801 gap++; 802 } else if (gap != 0) { 803 im6o->im6o_membership[i - gap] = 804 im6o->im6o_membership[i]; 805 } 806 } 807 im6o->im6o_num_memberships -= gap; 808 } 809 INP_WUNLOCK(in6p); 810 } 811 INP_INFO_RUNLOCK(pcbinfo); 812 } 813 814 /* 815 * Check for alternatives when higher level complains 816 * about service problems. For now, invalidate cached 817 * routing information. If the route was created dynamically 818 * (by a redirect), time to try a default gateway again. 819 */ 820 void 821 in6_losing(struct inpcb *in6p) 822 { 823 824 /* 825 * We don't store route pointers in the routing table anymore 826 */ 827 return; 828 } 829 830 /* 831 * After a routing change, flush old routing 832 * and allocate a (hopefully) better one. 833 */ 834 struct inpcb * 835 in6_rtchange(struct inpcb *inp, int errno) 836 { 837 /* 838 * We don't store route pointers in the routing table anymore 839 */ 840 return inp; 841 } 842 843 #ifdef PCBGROUP 844 /* 845 * Lookup PCB in hash list, using pcbgroup tables. 846 */ 847 static struct inpcb * 848 in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup, 849 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr, 850 u_int lport_arg, int lookupflags, struct ifnet *ifp) 851 { 852 struct inpcbhead *head; 853 struct inpcb *inp, *tmpinp; 854 u_short fport = fport_arg, lport = lport_arg; 855 int faith; 856 857 if (faithprefix_p != NULL) 858 faith = (*faithprefix_p)(laddr); 859 else 860 faith = 0; 861 862 /* 863 * First look for an exact match. 864 */ 865 tmpinp = NULL; 866 INP_GROUP_LOCK(pcbgroup); 867 head = &pcbgroup->ipg_hashbase[ 868 INP_PCBHASH(faddr->s6_addr32[3] /* XXX */, lport, fport, 869 pcbgroup->ipg_hashmask)]; 870 LIST_FOREACH(inp, head, inp_pcbgrouphash) { 871 /* XXX inp locking */ 872 if ((inp->inp_vflag & INP_IPV6) == 0) 873 continue; 874 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && 875 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 876 inp->inp_fport == fport && 877 inp->inp_lport == lport) { 878 /* 879 * XXX We should be able to directly return 880 * the inp here, without any checks. 881 * Well unless both bound with SO_REUSEPORT? 882 */ 883 if (prison_flag(inp->inp_cred, PR_IP6)) 884 goto found; 885 if (tmpinp == NULL) 886 tmpinp = inp; 887 } 888 } 889 if (tmpinp != NULL) { 890 inp = tmpinp; 891 goto found; 892 } 893 894 /* 895 * Then look for a wildcard match, if requested. 896 */ 897 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 898 struct inpcb *local_wild = NULL, *local_exact = NULL; 899 struct inpcb *jail_wild = NULL; 900 int injail; 901 902 /* 903 * Order of socket selection - we always prefer jails. 904 * 1. jailed, non-wild. 905 * 2. jailed, wild. 906 * 3. non-jailed, non-wild. 907 * 4. non-jailed, wild. 908 */ 909 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport, 910 0, pcbinfo->ipi_wildmask)]; 911 LIST_FOREACH(inp, head, inp_pcbgroup_wild) { 912 /* XXX inp locking */ 913 if ((inp->inp_vflag & INP_IPV6) == 0) 914 continue; 915 916 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 917 inp->inp_lport != lport) { 918 continue; 919 } 920 921 /* XXX inp locking */ 922 if (faith && (inp->inp_flags & INP_FAITH) == 0) 923 continue; 924 925 injail = prison_flag(inp->inp_cred, PR_IP6); 926 if (injail) { 927 if (prison_check_ip6(inp->inp_cred, 928 laddr) != 0) 929 continue; 930 } else { 931 if (local_exact != NULL) 932 continue; 933 } 934 935 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { 936 if (injail) 937 goto found; 938 else 939 local_exact = inp; 940 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 941 if (injail) 942 jail_wild = inp; 943 else 944 local_wild = inp; 945 } 946 } /* LIST_FOREACH */ 947 948 inp = jail_wild; 949 if (inp == NULL) 950 inp = jail_wild; 951 if (inp == NULL) 952 inp = local_exact; 953 if (inp == NULL) 954 inp = local_wild; 955 if (inp != NULL) 956 goto found; 957 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ 958 INP_GROUP_UNLOCK(pcbgroup); 959 return (NULL); 960 961 found: 962 in_pcbref(inp); 963 INP_GROUP_UNLOCK(pcbgroup); 964 if (lookupflags & INPLOOKUP_WLOCKPCB) { 965 INP_WLOCK(inp); 966 if (in_pcbrele_wlocked(inp)) 967 return (NULL); 968 } else if (lookupflags & INPLOOKUP_RLOCKPCB) { 969 INP_RLOCK(inp); 970 if (in_pcbrele_rlocked(inp)) 971 return (NULL); 972 } else 973 panic("%s: locking buf", __func__); 974 return (inp); 975 } 976 #endif /* PCBGROUP */ 977 978 /* 979 * Lookup PCB in hash list. 980 */ 981 struct inpcb * 982 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 983 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg, 984 int lookupflags, struct ifnet *ifp) 985 { 986 struct inpcbhead *head; 987 struct inpcb *inp, *tmpinp; 988 u_short fport = fport_arg, lport = lport_arg; 989 int faith; 990 991 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, 992 ("%s: invalid lookup flags %d", __func__, lookupflags)); 993 994 INP_HASH_LOCK_ASSERT(pcbinfo); 995 996 if (faithprefix_p != NULL) 997 faith = (*faithprefix_p)(laddr); 998 else 999 faith = 0; 1000 1001 /* 1002 * First look for an exact match. 1003 */ 1004 tmpinp = NULL; 1005 head = &pcbinfo->ipi_hashbase[ 1006 INP_PCBHASH(faddr->s6_addr32[3] /* XXX */, lport, fport, 1007 pcbinfo->ipi_hashmask)]; 1008 LIST_FOREACH(inp, head, inp_hash) { 1009 /* XXX inp locking */ 1010 if ((inp->inp_vflag & INP_IPV6) == 0) 1011 continue; 1012 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && 1013 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 1014 inp->inp_fport == fport && 1015 inp->inp_lport == lport) { 1016 /* 1017 * XXX We should be able to directly return 1018 * the inp here, without any checks. 1019 * Well unless both bound with SO_REUSEPORT? 1020 */ 1021 if (prison_flag(inp->inp_cred, PR_IP6)) 1022 return (inp); 1023 if (tmpinp == NULL) 1024 tmpinp = inp; 1025 } 1026 } 1027 if (tmpinp != NULL) 1028 return (tmpinp); 1029 1030 /* 1031 * Then look for a wildcard match, if requested. 1032 */ 1033 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 1034 struct inpcb *local_wild = NULL, *local_exact = NULL; 1035 struct inpcb *jail_wild = NULL; 1036 int injail; 1037 1038 /* 1039 * Order of socket selection - we always prefer jails. 1040 * 1. jailed, non-wild. 1041 * 2. jailed, wild. 1042 * 3. non-jailed, non-wild. 1043 * 4. non-jailed, wild. 1044 */ 1045 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 1046 0, pcbinfo->ipi_hashmask)]; 1047 LIST_FOREACH(inp, head, inp_hash) { 1048 /* XXX inp locking */ 1049 if ((inp->inp_vflag & INP_IPV6) == 0) 1050 continue; 1051 1052 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 1053 inp->inp_lport != lport) { 1054 continue; 1055 } 1056 1057 /* XXX inp locking */ 1058 if (faith && (inp->inp_flags & INP_FAITH) == 0) 1059 continue; 1060 1061 injail = prison_flag(inp->inp_cred, PR_IP6); 1062 if (injail) { 1063 if (prison_check_ip6(inp->inp_cred, 1064 laddr) != 0) 1065 continue; 1066 } else { 1067 if (local_exact != NULL) 1068 continue; 1069 } 1070 1071 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { 1072 if (injail) 1073 return (inp); 1074 else 1075 local_exact = inp; 1076 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 1077 if (injail) 1078 jail_wild = inp; 1079 else 1080 local_wild = inp; 1081 } 1082 } /* LIST_FOREACH */ 1083 1084 if (jail_wild != NULL) 1085 return (jail_wild); 1086 if (local_exact != NULL) 1087 return (local_exact); 1088 if (local_wild != NULL) 1089 return (local_wild); 1090 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ 1091 1092 /* 1093 * Not found. 1094 */ 1095 return (NULL); 1096 } 1097 1098 /* 1099 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the 1100 * hash list lock, and will return the inpcb locked (i.e., requires 1101 * INPLOOKUP_LOCKPCB). 1102 */ 1103 static struct inpcb * 1104 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 1105 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, 1106 struct ifnet *ifp) 1107 { 1108 struct inpcb *inp; 1109 1110 INP_HASH_RLOCK(pcbinfo); 1111 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, 1112 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp); 1113 if (inp != NULL) { 1114 in_pcbref(inp); 1115 INP_HASH_RUNLOCK(pcbinfo); 1116 if (lookupflags & INPLOOKUP_WLOCKPCB) { 1117 INP_WLOCK(inp); 1118 if (in_pcbrele_wlocked(inp)) 1119 return (NULL); 1120 } else if (lookupflags & INPLOOKUP_RLOCKPCB) { 1121 INP_RLOCK(inp); 1122 if (in_pcbrele_rlocked(inp)) 1123 return (NULL); 1124 } else 1125 panic("%s: locking bug", __func__); 1126 } else 1127 INP_HASH_RUNLOCK(pcbinfo); 1128 return (inp); 1129 } 1130 1131 /* 1132 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf 1133 * from which a pre-calculated hash value may be extracted. 1134 * 1135 * Possibly more of this logic should be in in6_pcbgroup.c. 1136 */ 1137 struct inpcb * 1138 in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport, 1139 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp) 1140 { 1141 #if defined(PCBGROUP) 1142 struct inpcbgroup *pcbgroup; 1143 #endif 1144 1145 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, 1146 ("%s: invalid lookup flags %d", __func__, lookupflags)); 1147 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, 1148 ("%s: LOCKPCB not set", __func__)); 1149 1150 #if defined(PCBGROUP) 1151 if (in_pcbgroup_enabled(pcbinfo)) { 1152 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, 1153 fport); 1154 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, 1155 laddr, lport, lookupflags, ifp)); 1156 } 1157 #endif 1158 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, 1159 lookupflags, ifp)); 1160 } 1161 1162 struct inpcb * 1163 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 1164 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, 1165 struct ifnet *ifp, struct mbuf *m) 1166 { 1167 #ifdef PCBGROUP 1168 struct inpcbgroup *pcbgroup; 1169 #endif 1170 1171 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, 1172 ("%s: invalid lookup flags %d", __func__, lookupflags)); 1173 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, 1174 ("%s: LOCKPCB not set", __func__)); 1175 1176 #ifdef PCBGROUP 1177 if (in_pcbgroup_enabled(pcbinfo)) { 1178 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m), 1179 m->m_pkthdr.flowid); 1180 if (pcbgroup != NULL) 1181 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, 1182 fport, laddr, lport, lookupflags, ifp)); 1183 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, 1184 fport); 1185 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, 1186 laddr, lport, lookupflags, ifp)); 1187 } 1188 #endif 1189 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, 1190 lookupflags, ifp)); 1191 } 1192 1193 void 1194 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m) 1195 { 1196 struct ip6_hdr *ip; 1197 1198 ip = mtod(m, struct ip6_hdr *); 1199 bzero(sin6, sizeof(*sin6)); 1200 sin6->sin6_len = sizeof(*sin6); 1201 sin6->sin6_family = AF_INET6; 1202 sin6->sin6_addr = ip->ip6_src; 1203 1204 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */ 1205 1206 return; 1207 } 1208