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