Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/netinet6/in6_pcb.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * Copyright (c) 2010-2011 Juniper Networks, Inc. 6 * All rights reserved. 7 * 8 * Portions of this software were developed by Robert N. M. Watson under 9 * contract to Juniper Networks, Inc. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. Neither the name of the project nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $ 36 */ 37 38 /*- 39 * Copyright (c) 1982, 1986, 1991, 1993 40 * The Regents of the University of California. All rights reserved. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 67 */ 68 69 #include <sys/cdefs.h> 70 __FBSDID("$FreeBSD$"); 71 72 #include "opt_inet.h" 73 #include "opt_inet6.h" 74 #include "opt_ipsec.h" 75 #include "opt_pcbgroup.h" 76 #include "opt_route.h" 77 #include "opt_rss.h" 78 79 #include <sys/param.h> 80 #include <sys/systm.h> 81 #include <sys/malloc.h> 82 #include <sys/mbuf.h> 83 #include <sys/domain.h> 84 #include <sys/protosw.h> 85 #include <sys/socket.h> 86 #include <sys/socketvar.h> 87 #include <sys/sockio.h> 88 #include <sys/errno.h> 89 #include <sys/time.h> 90 #include <sys/priv.h> 91 #include <sys/proc.h> 92 #include <sys/jail.h> 93 94 #include <vm/uma.h> 95 96 #include <net/if.h> 97 #include <net/if_var.h> 98 #include <net/if_llatbl.h> 99 #include <net/if_types.h> 100 #include <net/route.h> 101 #include <net/route/nhop.h> 102 103 #include <netinet/in.h> 104 #include <netinet/in_var.h> 105 #include <netinet/in_systm.h> 106 #include <netinet/tcp_var.h> 107 #include <netinet/ip6.h> 108 #include <netinet/ip_var.h> 109 110 #include <netinet6/ip6_var.h> 111 #include <netinet6/nd6.h> 112 #include <netinet/in_pcb.h> 113 #include <netinet6/in6_pcb.h> 114 #include <netinet6/in6_fib.h> 115 #include <netinet6/scope6_var.h> 116 117 int 118 in6_pcbbind(struct inpcb *inp, struct sockaddr *nam, 119 struct ucred *cred) 120 { 121 struct socket *so = inp->inp_socket; 122 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL; 123 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 124 u_short lport = 0; 125 int error, lookupflags = 0; 126 int reuseport = (so->so_options & SO_REUSEPORT); 127 128 /* 129 * XXX: Maybe we could let SO_REUSEPORT_LB set SO_REUSEPORT bit here 130 * so that we don't have to add to the (already messy) code below. 131 */ 132 int reuseport_lb = (so->so_options & SO_REUSEPORT_LB); 133 134 INP_WLOCK_ASSERT(inp); 135 INP_HASH_WLOCK_ASSERT(pcbinfo); 136 137 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 138 return (EINVAL); 139 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0) 140 lookupflags = INPLOOKUP_WILDCARD; 141 if (nam == NULL) { 142 if ((error = prison_local_ip6(cred, &inp->in6p_laddr, 143 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 144 return (error); 145 } else { 146 sin6 = (struct sockaddr_in6 *)nam; 147 KASSERT(sin6->sin6_family == AF_INET6, 148 ("%s: invalid address family for %p", __func__, sin6)); 149 KASSERT(sin6->sin6_len == sizeof(*sin6), 150 ("%s: invalid address length for %p", __func__, sin6)); 151 152 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) 153 return(error); 154 155 if ((error = prison_local_ip6(cred, &sin6->sin6_addr, 156 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 157 return (error); 158 159 lport = sin6->sin6_port; 160 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 161 /* 162 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 163 * allow compepte duplication of binding if 164 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 165 * and a multicast address is bound on both 166 * new and duplicated sockets. 167 */ 168 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0) 169 reuseport = SO_REUSEADDR|SO_REUSEPORT; 170 /* 171 * XXX: How to deal with SO_REUSEPORT_LB here? 172 * Treat same as SO_REUSEPORT for now. 173 */ 174 if ((so->so_options & 175 (SO_REUSEADDR|SO_REUSEPORT_LB)) != 0) 176 reuseport_lb = SO_REUSEADDR|SO_REUSEPORT_LB; 177 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 178 struct epoch_tracker et; 179 struct ifaddr *ifa; 180 181 sin6->sin6_port = 0; /* yech... */ 182 NET_EPOCH_ENTER(et); 183 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) == 184 NULL && 185 (inp->inp_flags & INP_BINDANY) == 0) { 186 NET_EPOCH_EXIT(et); 187 return (EADDRNOTAVAIL); 188 } 189 190 /* 191 * XXX: bind to an anycast address might accidentally 192 * cause sending a packet with anycast source address. 193 * We should allow to bind to a deprecated address, since 194 * the application dares to use it. 195 */ 196 if (ifa != NULL && 197 ((struct in6_ifaddr *)ifa)->ia6_flags & 198 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) { 199 NET_EPOCH_EXIT(et); 200 return (EADDRNOTAVAIL); 201 } 202 NET_EPOCH_EXIT(et); 203 } 204 if (lport) { 205 struct inpcb *t; 206 struct tcptw *tw; 207 208 /* GROSS */ 209 if (ntohs(lport) <= V_ipport_reservedhigh && 210 ntohs(lport) >= V_ipport_reservedlow && 211 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT)) 212 return (EACCES); 213 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) && 214 priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT) != 0) { 215 t = in6_pcblookup_local(pcbinfo, 216 &sin6->sin6_addr, lport, 217 INPLOOKUP_WILDCARD, cred); 218 if (t && 219 ((inp->inp_flags2 & INP_BINDMULTI) == 0) && 220 ((t->inp_flags & INP_TIMEWAIT) == 0) && 221 (so->so_type != SOCK_STREAM || 222 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) && 223 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || 224 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) || 225 (t->inp_flags2 & INP_REUSEPORT) || 226 (t->inp_flags2 & INP_REUSEPORT_LB) == 0) && 227 (inp->inp_cred->cr_uid != 228 t->inp_cred->cr_uid)) 229 return (EADDRINUSE); 230 231 /* 232 * If the socket is a BINDMULTI socket, then 233 * the credentials need to match and the 234 * original socket also has to have been bound 235 * with BINDMULTI. 236 */ 237 if (t && (! in_pcbbind_check_bindmulti(inp, t))) 238 return (EADDRINUSE); 239 240 #ifdef INET 241 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && 242 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 243 struct sockaddr_in sin; 244 245 in6_sin6_2_sin(&sin, sin6); 246 t = in_pcblookup_local(pcbinfo, 247 sin.sin_addr, lport, 248 INPLOOKUP_WILDCARD, cred); 249 if (t && 250 ((inp->inp_flags2 & INP_BINDMULTI) == 0) && 251 ((t->inp_flags & 252 INP_TIMEWAIT) == 0) && 253 (so->so_type != SOCK_STREAM || 254 ntohl(t->inp_faddr.s_addr) == 255 INADDR_ANY) && 256 (inp->inp_cred->cr_uid != 257 t->inp_cred->cr_uid)) 258 return (EADDRINUSE); 259 260 if (t && (! in_pcbbind_check_bindmulti(inp, t))) 261 return (EADDRINUSE); 262 } 263 #endif 264 } 265 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr, 266 lport, lookupflags, cred); 267 if (t && (t->inp_flags & INP_TIMEWAIT)) { 268 /* 269 * XXXRW: If an incpb has had its timewait 270 * state recycled, we treat the address as 271 * being in use (for now). This is better 272 * than a panic, but not desirable. 273 */ 274 tw = intotw(t); 275 if (tw == NULL || 276 ((reuseport & tw->tw_so_options) == 0 && 277 (reuseport_lb & tw->tw_so_options) == 0)) 278 return (EADDRINUSE); 279 } else if (t && (reuseport & inp_so_options(t)) == 0 && 280 (reuseport_lb & inp_so_options(t)) == 0) { 281 return (EADDRINUSE); 282 } 283 #ifdef INET 284 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && 285 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 286 struct sockaddr_in sin; 287 288 in6_sin6_2_sin(&sin, sin6); 289 t = in_pcblookup_local(pcbinfo, sin.sin_addr, 290 lport, lookupflags, cred); 291 if (t && t->inp_flags & INP_TIMEWAIT) { 292 tw = intotw(t); 293 if (tw == NULL) 294 return (EADDRINUSE); 295 if ((reuseport & tw->tw_so_options) == 0 296 && (reuseport_lb & tw->tw_so_options) == 0 297 && (ntohl(t->inp_laddr.s_addr) != 298 INADDR_ANY || ((inp->inp_vflag & 299 INP_IPV6PROTO) == 300 (t->inp_vflag & INP_IPV6PROTO)))) 301 return (EADDRINUSE); 302 } else if (t && 303 (reuseport & inp_so_options(t)) == 0 && 304 (reuseport_lb & inp_so_options(t)) == 0 && 305 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 306 (t->inp_vflag & INP_IPV6PROTO) != 0)) { 307 return (EADDRINUSE); 308 } 309 } 310 #endif 311 } 312 inp->in6p_laddr = sin6->sin6_addr; 313 } 314 if (lport == 0) { 315 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) { 316 /* Undo an address bind that may have occurred. */ 317 inp->in6p_laddr = in6addr_any; 318 return (error); 319 } 320 } else { 321 inp->inp_lport = lport; 322 if (in_pcbinshash(inp) != 0) { 323 inp->in6p_laddr = in6addr_any; 324 inp->inp_lport = 0; 325 return (EAGAIN); 326 } 327 } 328 return (0); 329 } 330 331 /* 332 * Transform old in6_pcbconnect() into an inner subroutine for new 333 * in6_pcbconnect(): Do some validity-checking on the remote 334 * address (in mbuf 'nam') and then determine local host address 335 * (i.e., which interface) to use to access that remote host. 336 * 337 * This preserves definition of in6_pcbconnect(), while supporting a 338 * slightly different version for T/TCP. (This is more than 339 * a bit of a kludge, but cleaning up the internal interfaces would 340 * have forced minor changes in every protocol). 341 */ 342 static int 343 in6_pcbladdr(struct inpcb *inp, struct sockaddr_in6 *sin6, 344 struct in6_addr *plocal_addr6) 345 { 346 int error = 0; 347 int scope_ambiguous = 0; 348 struct in6_addr in6a; 349 struct epoch_tracker et; 350 351 INP_WLOCK_ASSERT(inp); 352 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */ 353 354 if (sin6->sin6_port == 0) 355 return (EADDRNOTAVAIL); 356 357 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) 358 scope_ambiguous = 1; 359 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) 360 return(error); 361 362 if (!CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) { 363 /* 364 * If the destination address is UNSPECIFIED addr, 365 * use the loopback addr, e.g ::1. 366 */ 367 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 368 sin6->sin6_addr = in6addr_loopback; 369 } 370 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0) 371 return (error); 372 373 NET_EPOCH_ENTER(et); 374 error = in6_selectsrc_socket(sin6, inp->in6p_outputopts, 375 inp, inp->inp_cred, scope_ambiguous, &in6a, NULL); 376 NET_EPOCH_EXIT(et); 377 if (error) 378 return (error); 379 380 /* 381 * Do not update this earlier, in case we return with an error. 382 * 383 * XXX: this in6_selectsrc_socket result might replace the bound local 384 * address with the address specified by setsockopt(IPV6_PKTINFO). 385 * Is it the intended behavior? 386 */ 387 *plocal_addr6 = in6a; 388 389 /* 390 * Don't do pcblookup call here; return interface in 391 * plocal_addr6 392 * and exit to caller, that will do the lookup. 393 */ 394 395 return (0); 396 } 397 398 /* 399 * Outer subroutine: 400 * Connect from a socket to a specified address. 401 * Both address and port must be specified in argument sin. 402 * If don't have a local address for this socket yet, 403 * then pick one. 404 */ 405 int 406 in6_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam, 407 struct ucred *cred, struct mbuf *m, bool rehash) 408 { 409 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 410 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 411 struct sockaddr_in6 laddr6; 412 int error; 413 414 KASSERT(sin6->sin6_family == AF_INET6, 415 ("%s: invalid address family for %p", __func__, sin6)); 416 KASSERT(sin6->sin6_len == sizeof(*sin6), 417 ("%s: invalid address length for %p", __func__, sin6)); 418 419 bzero(&laddr6, sizeof(laddr6)); 420 laddr6.sin6_family = AF_INET6; 421 422 INP_WLOCK_ASSERT(inp); 423 INP_HASH_WLOCK_ASSERT(pcbinfo); 424 425 #ifdef ROUTE_MPATH 426 if (CALC_FLOWID_OUTBOUND) { 427 uint32_t hash_type, hash_val; 428 429 hash_val = fib6_calc_software_hash(&inp->in6p_laddr, 430 &sin6->sin6_addr, 0, sin6->sin6_port, 431 inp->inp_socket->so_proto->pr_protocol, &hash_type); 432 inp->inp_flowid = hash_val; 433 inp->inp_flowtype = hash_type; 434 } 435 #endif 436 /* 437 * Call inner routine, to assign local interface address. 438 * in6_pcbladdr() may automatically fill in sin6_scope_id. 439 */ 440 if ((error = in6_pcbladdr(inp, sin6, &laddr6.sin6_addr)) != 0) 441 return (error); 442 443 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr, 444 sin6->sin6_port, 445 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) 446 ? &laddr6.sin6_addr : &inp->in6p_laddr, 447 inp->inp_lport, 0, NULL, M_NODOM) != NULL) { 448 return (EADDRINUSE); 449 } 450 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 451 if (inp->inp_lport == 0) { 452 /* 453 * rehash was required to be true in the past for 454 * this case; retain that convention. However, 455 * we now call in_pcb_lport_dest rather than 456 * in6_pcbbind; the former does not insert into 457 * the hash table, the latter does. Change rehash 458 * to false to do the in_pcbinshash below. 459 */ 460 KASSERT(rehash == true, 461 ("Rehashing required for unbound inps")); 462 rehash = false; 463 error = in_pcb_lport_dest(inp, 464 (struct sockaddr *) &laddr6, &inp->inp_lport, 465 (struct sockaddr *) sin6, sin6->sin6_port, cred, 466 INPLOOKUP_WILDCARD); 467 if (error) 468 return (error); 469 } 470 inp->in6p_laddr = laddr6.sin6_addr; 471 } 472 inp->in6p_faddr = sin6->sin6_addr; 473 inp->inp_fport = sin6->sin6_port; 474 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ 475 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; 476 if (inp->inp_flags & IN6P_AUTOFLOWLABEL) 477 inp->inp_flow |= 478 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); 479 480 if (rehash) { 481 in_pcbrehash_mbuf(inp, m); 482 } else { 483 in_pcbinshash_mbuf(inp, m); 484 } 485 486 return (0); 487 } 488 489 int 490 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 491 { 492 493 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL, true)); 494 } 495 496 void 497 in6_pcbdisconnect(struct inpcb *inp) 498 { 499 500 INP_WLOCK_ASSERT(inp); 501 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); 502 503 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr)); 504 inp->inp_fport = 0; 505 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ 506 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; 507 in_pcbrehash(inp); 508 } 509 510 struct sockaddr * 511 in6_sockaddr(in_port_t port, struct in6_addr *addr_p) 512 { 513 struct sockaddr_in6 *sin6; 514 515 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK); 516 bzero(sin6, sizeof *sin6); 517 sin6->sin6_family = AF_INET6; 518 sin6->sin6_len = sizeof(*sin6); 519 sin6->sin6_port = port; 520 sin6->sin6_addr = *addr_p; 521 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */ 522 523 return (struct sockaddr *)sin6; 524 } 525 526 struct sockaddr * 527 in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p) 528 { 529 struct sockaddr_in sin; 530 struct sockaddr_in6 *sin6_p; 531 532 bzero(&sin, sizeof sin); 533 sin.sin_family = AF_INET; 534 sin.sin_len = sizeof(sin); 535 sin.sin_port = port; 536 sin.sin_addr = *addr_p; 537 538 sin6_p = malloc(sizeof *sin6_p, M_SONAME, 539 M_WAITOK); 540 in6_sin_2_v4mapsin6(&sin, sin6_p); 541 542 return (struct sockaddr *)sin6_p; 543 } 544 545 int 546 in6_getsockaddr(struct socket *so, struct sockaddr **nam) 547 { 548 struct inpcb *inp; 549 struct in6_addr addr; 550 in_port_t port; 551 552 inp = sotoinpcb(so); 553 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL")); 554 555 INP_RLOCK(inp); 556 port = inp->inp_lport; 557 addr = inp->in6p_laddr; 558 INP_RUNLOCK(inp); 559 560 *nam = in6_sockaddr(port, &addr); 561 return 0; 562 } 563 564 int 565 in6_getpeeraddr(struct socket *so, struct sockaddr **nam) 566 { 567 struct inpcb *inp; 568 struct in6_addr addr; 569 in_port_t port; 570 571 inp = sotoinpcb(so); 572 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL")); 573 574 INP_RLOCK(inp); 575 port = inp->inp_fport; 576 addr = inp->in6p_faddr; 577 INP_RUNLOCK(inp); 578 579 *nam = in6_sockaddr(port, &addr); 580 return 0; 581 } 582 583 int 584 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam) 585 { 586 struct inpcb *inp; 587 int error; 588 589 inp = sotoinpcb(so); 590 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL")); 591 592 #ifdef INET 593 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { 594 error = in_getsockaddr(so, nam); 595 if (error == 0) 596 in6_sin_2_v4mapsin6_in_sock(nam); 597 } else 598 #endif 599 { 600 /* scope issues will be handled in in6_getsockaddr(). */ 601 error = in6_getsockaddr(so, nam); 602 } 603 604 return error; 605 } 606 607 int 608 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam) 609 { 610 struct inpcb *inp; 611 int error; 612 613 inp = sotoinpcb(so); 614 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL")); 615 616 #ifdef INET 617 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { 618 error = in_getpeeraddr(so, nam); 619 if (error == 0) 620 in6_sin_2_v4mapsin6_in_sock(nam); 621 } else 622 #endif 623 /* scope issues will be handled in in6_getpeeraddr(). */ 624 error = in6_getpeeraddr(so, nam); 625 626 return error; 627 } 628 629 /* 630 * Pass some notification to all connections of a protocol 631 * associated with address dst. The local address and/or port numbers 632 * may be specified to limit the search. The "usual action" will be 633 * taken, depending on the ctlinput cmd. The caller must filter any 634 * cmds that are uninteresting (e.g., no error in the map). 635 * Call the protocol specific routine (if any) to report 636 * any errors for each matching socket. 637 */ 638 void 639 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, 640 u_int fport_arg, const struct sockaddr *src, u_int lport_arg, 641 int cmd, void *cmdarg, 642 struct inpcb *(*notify)(struct inpcb *, int)) 643 { 644 struct inpcb *inp, *inp_temp; 645 struct sockaddr_in6 sa6_src, *sa6_dst; 646 u_short fport = fport_arg, lport = lport_arg; 647 u_int32_t flowinfo; 648 int errno; 649 650 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6) 651 return; 652 653 sa6_dst = (struct sockaddr_in6 *)dst; 654 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr)) 655 return; 656 657 /* 658 * note that src can be NULL when we get notify by local fragmentation. 659 */ 660 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src; 661 flowinfo = sa6_src.sin6_flowinfo; 662 663 /* 664 * Redirects go to all references to the destination, 665 * and use in6_rtchange to invalidate the route cache. 666 * Dead host indications: also use in6_rtchange to invalidate 667 * the cache, and deliver the error to all the sockets. 668 * Otherwise, if we have knowledge of the local port and address, 669 * deliver only to that socket. 670 */ 671 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) { 672 fport = 0; 673 lport = 0; 674 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr)); 675 676 if (cmd != PRC_HOSTDEAD) 677 notify = in6_rtchange; 678 } 679 errno = inet6ctlerrmap[cmd]; 680 INP_INFO_WLOCK(pcbinfo); 681 CK_LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { 682 INP_WLOCK(inp); 683 if ((inp->inp_vflag & INP_IPV6) == 0) { 684 INP_WUNLOCK(inp); 685 continue; 686 } 687 688 /* 689 * If the error designates a new path MTU for a destination 690 * and the application (associated with this socket) wanted to 691 * know the value, notify. 692 * XXX: should we avoid to notify the value to TCP sockets? 693 */ 694 if (cmd == PRC_MSGSIZE && cmdarg != NULL) 695 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, 696 *(u_int32_t *)cmdarg); 697 698 /* 699 * Detect if we should notify the error. If no source and 700 * destination ports are specifed, but non-zero flowinfo and 701 * local address match, notify the error. This is the case 702 * when the error is delivered with an encrypted buffer 703 * by ESP. Otherwise, just compare addresses and ports 704 * as usual. 705 */ 706 if (lport == 0 && fport == 0 && flowinfo && 707 inp->inp_socket != NULL && 708 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) && 709 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr)) 710 goto do_notify; 711 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, 712 &sa6_dst->sin6_addr) || 713 inp->inp_socket == 0 || 714 (lport && inp->inp_lport != lport) || 715 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) && 716 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, 717 &sa6_src.sin6_addr)) || 718 (fport && inp->inp_fport != fport)) { 719 INP_WUNLOCK(inp); 720 continue; 721 } 722 723 do_notify: 724 if (notify) { 725 if ((*notify)(inp, errno)) 726 INP_WUNLOCK(inp); 727 } else 728 INP_WUNLOCK(inp); 729 } 730 INP_INFO_WUNLOCK(pcbinfo); 731 } 732 733 /* 734 * Lookup a PCB based on the local address and port. Caller must hold the 735 * hash lock. No inpcb locks or references are acquired. 736 */ 737 struct inpcb * 738 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr, 739 u_short lport, int lookupflags, struct ucred *cred) 740 { 741 struct inpcb *inp; 742 int matchwild = 3, wildcard; 743 744 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, 745 ("%s: invalid lookup flags %d", __func__, lookupflags)); 746 747 INP_HASH_LOCK_ASSERT(pcbinfo); 748 749 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { 750 struct inpcbhead *head; 751 /* 752 * Look for an unconnected (wildcard foreign addr) PCB that 753 * matches the local address and port we're looking for. 754 */ 755 head = &pcbinfo->ipi_hashbase[INP_PCBHASH( 756 INP6_PCBHASHKEY(&in6addr_any), lport, 0, 757 pcbinfo->ipi_hashmask)]; 758 CK_LIST_FOREACH(inp, head, inp_hash) { 759 /* XXX inp locking */ 760 if ((inp->inp_vflag & INP_IPV6) == 0) 761 continue; 762 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 763 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 764 inp->inp_lport == lport) { 765 /* Found. */ 766 if (cred == NULL || 767 prison_equal_ip6(cred->cr_prison, 768 inp->inp_cred->cr_prison)) 769 return (inp); 770 } 771 } 772 /* 773 * Not found. 774 */ 775 return (NULL); 776 } else { 777 struct inpcbporthead *porthash; 778 struct inpcbport *phd; 779 struct inpcb *match = NULL; 780 /* 781 * Best fit PCB lookup. 782 * 783 * First see if this local port is in use by looking on the 784 * port hash list. 785 */ 786 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, 787 pcbinfo->ipi_porthashmask)]; 788 CK_LIST_FOREACH(phd, porthash, phd_hash) { 789 if (phd->phd_port == lport) 790 break; 791 } 792 if (phd != NULL) { 793 /* 794 * Port is in use by one or more PCBs. Look for best 795 * fit. 796 */ 797 CK_LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 798 wildcard = 0; 799 if (cred != NULL && 800 !prison_equal_ip6(cred->cr_prison, 801 inp->inp_cred->cr_prison)) 802 continue; 803 /* XXX inp locking */ 804 if ((inp->inp_vflag & INP_IPV6) == 0) 805 continue; 806 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) 807 wildcard++; 808 if (!IN6_IS_ADDR_UNSPECIFIED( 809 &inp->in6p_laddr)) { 810 if (IN6_IS_ADDR_UNSPECIFIED(laddr)) 811 wildcard++; 812 else if (!IN6_ARE_ADDR_EQUAL( 813 &inp->in6p_laddr, laddr)) 814 continue; 815 } else { 816 if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) 817 wildcard++; 818 } 819 if (wildcard < matchwild) { 820 match = inp; 821 matchwild = wildcard; 822 if (matchwild == 0) 823 break; 824 } 825 } 826 } 827 return (match); 828 } 829 } 830 831 void 832 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) 833 { 834 struct inpcb *inp; 835 struct in6_multi *inm; 836 struct in6_mfilter *imf; 837 struct ip6_moptions *im6o; 838 839 INP_INFO_WLOCK(pcbinfo); 840 CK_LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) { 841 INP_WLOCK(inp); 842 if (__predict_false(inp->inp_flags2 & INP_FREED)) { 843 INP_WUNLOCK(inp); 844 continue; 845 } 846 im6o = inp->in6p_moptions; 847 if ((inp->inp_vflag & INP_IPV6) && im6o != NULL) { 848 /* 849 * Unselect the outgoing ifp for multicast if it 850 * is being detached. 851 */ 852 if (im6o->im6o_multicast_ifp == ifp) 853 im6o->im6o_multicast_ifp = NULL; 854 /* 855 * Drop multicast group membership if we joined 856 * through the interface being detached. 857 */ 858 restart: 859 IP6_MFILTER_FOREACH(imf, &im6o->im6o_head) { 860 if ((inm = imf->im6f_in6m) == NULL) 861 continue; 862 if (inm->in6m_ifp != ifp) 863 continue; 864 ip6_mfilter_remove(&im6o->im6o_head, imf); 865 IN6_MULTI_LOCK_ASSERT(); 866 in6_leavegroup_locked(inm, NULL); 867 ip6_mfilter_free(imf); 868 goto restart; 869 } 870 } 871 INP_WUNLOCK(inp); 872 } 873 INP_INFO_WUNLOCK(pcbinfo); 874 } 875 876 /* 877 * Check for alternatives when higher level complains 878 * about service problems. For now, invalidate cached 879 * routing information. If the route was created dynamically 880 * (by a redirect), time to try a default gateway again. 881 */ 882 void 883 in6_losing(struct inpcb *inp) 884 { 885 886 RO_INVALIDATE_CACHE(&inp->inp_route6); 887 } 888 889 /* 890 * After a routing change, flush old routing 891 * and allocate a (hopefully) better one. 892 */ 893 struct inpcb * 894 in6_rtchange(struct inpcb *inp, int errno __unused) 895 { 896 897 RO_INVALIDATE_CACHE(&inp->inp_route6); 898 return inp; 899 } 900 901 static struct inpcb * 902 in6_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo, 903 const struct in6_addr *laddr, uint16_t lport, const struct in6_addr *faddr, 904 uint16_t fport, int lookupflags, uint8_t numa_domain) 905 { 906 struct inpcb *local_wild, *numa_wild; 907 const struct inpcblbgrouphead *hdr; 908 struct inpcblbgroup *grp; 909 uint32_t idx; 910 911 INP_HASH_LOCK_ASSERT(pcbinfo); 912 913 hdr = &pcbinfo->ipi_lbgrouphashbase[ 914 INP_PCBPORTHASH(lport, pcbinfo->ipi_lbgrouphashmask)]; 915 916 /* 917 * Order of socket selection: 918 * 1. non-wild. 919 * 2. wild (if lookupflags contains INPLOOKUP_WILDCARD). 920 * 921 * NOTE: 922 * - Load balanced group does not contain jailed sockets. 923 * - Load balanced does not contain IPv4 mapped INET6 wild sockets. 924 */ 925 local_wild = NULL; 926 numa_wild = NULL; 927 CK_LIST_FOREACH(grp, hdr, il_list) { 928 #ifdef INET 929 if (!(grp->il_vflag & INP_IPV6)) 930 continue; 931 #endif 932 if (grp->il_lport != lport) 933 continue; 934 935 idx = INP_PCBLBGROUP_PKTHASH(INP6_PCBHASHKEY(faddr), lport, 936 fport) % grp->il_inpcnt; 937 if (IN6_ARE_ADDR_EQUAL(&grp->il6_laddr, laddr)) { 938 if (numa_domain == M_NODOM || 939 grp->il_numa_domain == numa_domain) { 940 return (grp->il_inp[idx]); 941 } 942 else 943 numa_wild = grp->il_inp[idx]; 944 } 945 if (IN6_IS_ADDR_UNSPECIFIED(&grp->il6_laddr) && 946 (lookupflags & INPLOOKUP_WILDCARD) != 0 && 947 (local_wild == NULL || numa_domain == M_NODOM || 948 grp->il_numa_domain == numa_domain)) { 949 local_wild = grp->il_inp[idx]; 950 } 951 } 952 if (numa_wild != NULL) 953 return (numa_wild); 954 return (local_wild); 955 } 956 957 #ifdef PCBGROUP 958 /* 959 * Lookup PCB in hash list, using pcbgroup tables. 960 */ 961 static struct inpcb * 962 in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup, 963 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr, 964 u_int lport_arg, int lookupflags, struct ifnet *ifp) 965 { 966 struct inpcbhead *head; 967 struct inpcb *inp, *tmpinp; 968 u_short fport = fport_arg, lport = lport_arg; 969 bool locked; 970 971 /* 972 * First look for an exact match. 973 */ 974 tmpinp = NULL; 975 INP_GROUP_LOCK(pcbgroup); 976 head = &pcbgroup->ipg_hashbase[INP_PCBHASH( 977 INP6_PCBHASHKEY(faddr), lport, fport, pcbgroup->ipg_hashmask)]; 978 CK_LIST_FOREACH(inp, head, inp_pcbgrouphash) { 979 /* XXX inp locking */ 980 if ((inp->inp_vflag & INP_IPV6) == 0) 981 continue; 982 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && 983 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 984 inp->inp_fport == fport && 985 inp->inp_lport == lport) { 986 /* 987 * XXX We should be able to directly return 988 * the inp here, without any checks. 989 * Well unless both bound with SO_REUSEPORT? 990 */ 991 if (prison_flag(inp->inp_cred, PR_IP6)) 992 goto found; 993 if (tmpinp == NULL) 994 tmpinp = inp; 995 } 996 } 997 if (tmpinp != NULL) { 998 inp = tmpinp; 999 goto found; 1000 } 1001 1002 /* 1003 * Then look for a wildcard match in the pcbgroup. 1004 */ 1005 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 1006 struct inpcb *local_wild = NULL, *local_exact = NULL; 1007 struct inpcb *jail_wild = NULL; 1008 int injail; 1009 1010 /* 1011 * Order of socket selection - we always prefer jails. 1012 * 1. jailed, non-wild. 1013 * 2. jailed, wild. 1014 * 3. non-jailed, non-wild. 1015 * 4. non-jailed, wild. 1016 */ 1017 head = &pcbgroup->ipg_hashbase[ 1018 INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)]; 1019 CK_LIST_FOREACH(inp, head, inp_pcbgrouphash) { 1020 /* XXX inp locking */ 1021 if ((inp->inp_vflag & INP_IPV6) == 0) 1022 continue; 1023 1024 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 1025 inp->inp_lport != lport) { 1026 continue; 1027 } 1028 1029 injail = prison_flag(inp->inp_cred, PR_IP6); 1030 if (injail) { 1031 if (prison_check_ip6(inp->inp_cred, 1032 laddr) != 0) 1033 continue; 1034 } else { 1035 if (local_exact != NULL) 1036 continue; 1037 } 1038 1039 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { 1040 if (injail) 1041 goto found; 1042 else 1043 local_exact = inp; 1044 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 1045 if (injail) 1046 jail_wild = inp; 1047 else 1048 local_wild = inp; 1049 } 1050 } /* LIST_FOREACH */ 1051 1052 inp = jail_wild; 1053 if (inp == NULL) 1054 inp = jail_wild; 1055 if (inp == NULL) 1056 inp = local_exact; 1057 if (inp == NULL) 1058 inp = local_wild; 1059 if (inp != NULL) 1060 goto found; 1061 } 1062 1063 /* 1064 * Then look for a wildcard match, if requested. 1065 */ 1066 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 1067 struct inpcb *local_wild = NULL, *local_exact = NULL; 1068 struct inpcb *jail_wild = NULL; 1069 int injail; 1070 1071 /* 1072 * Order of socket selection - we always prefer jails. 1073 * 1. jailed, non-wild. 1074 * 2. jailed, wild. 1075 * 3. non-jailed, non-wild. 1076 * 4. non-jailed, wild. 1077 */ 1078 head = &pcbinfo->ipi_wildbase[INP_PCBHASH( 1079 INP6_PCBHASHKEY(&in6addr_any), lport, 0, 1080 pcbinfo->ipi_wildmask)]; 1081 CK_LIST_FOREACH(inp, head, inp_pcbgroup_wild) { 1082 /* XXX inp locking */ 1083 if ((inp->inp_vflag & INP_IPV6) == 0) 1084 continue; 1085 1086 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 1087 inp->inp_lport != lport) { 1088 continue; 1089 } 1090 1091 injail = prison_flag(inp->inp_cred, PR_IP6); 1092 if (injail) { 1093 if (prison_check_ip6(inp->inp_cred, 1094 laddr) != 0) 1095 continue; 1096 } else { 1097 if (local_exact != NULL) 1098 continue; 1099 } 1100 1101 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { 1102 if (injail) 1103 goto found; 1104 else 1105 local_exact = inp; 1106 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 1107 if (injail) 1108 jail_wild = inp; 1109 else 1110 local_wild = inp; 1111 } 1112 } /* LIST_FOREACH */ 1113 1114 inp = jail_wild; 1115 if (inp == NULL) 1116 inp = jail_wild; 1117 if (inp == NULL) 1118 inp = local_exact; 1119 if (inp == NULL) 1120 inp = local_wild; 1121 if (inp != NULL) 1122 goto found; 1123 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ 1124 INP_GROUP_UNLOCK(pcbgroup); 1125 return (NULL); 1126 1127 found: 1128 if (lookupflags & INPLOOKUP_WLOCKPCB) 1129 locked = INP_TRY_WLOCK(inp); 1130 else if (lookupflags & INPLOOKUP_RLOCKPCB) 1131 locked = INP_TRY_RLOCK(inp); 1132 else 1133 panic("%s: locking buf", __func__); 1134 if (!locked) 1135 in_pcbref(inp); 1136 INP_GROUP_UNLOCK(pcbgroup); 1137 if (!locked) { 1138 if (lookupflags & INPLOOKUP_WLOCKPCB) { 1139 INP_WLOCK(inp); 1140 if (in_pcbrele_wlocked(inp)) 1141 return (NULL); 1142 } else { 1143 INP_RLOCK(inp); 1144 if (in_pcbrele_rlocked(inp)) 1145 return (NULL); 1146 } 1147 } 1148 #ifdef INVARIANTS 1149 if (lookupflags & INPLOOKUP_WLOCKPCB) 1150 INP_WLOCK_ASSERT(inp); 1151 else 1152 INP_RLOCK_ASSERT(inp); 1153 #endif 1154 return (inp); 1155 } 1156 #endif /* PCBGROUP */ 1157 1158 /* 1159 * Lookup PCB in hash list. Used in in_pcb.c as well as here. 1160 */ 1161 struct inpcb * 1162 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 1163 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg, 1164 int lookupflags, struct ifnet *ifp, uint8_t numa_domain) 1165 { 1166 struct inpcbhead *head; 1167 struct inpcb *inp, *tmpinp; 1168 u_short fport = fport_arg, lport = lport_arg; 1169 1170 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, 1171 ("%s: invalid lookup flags %d", __func__, lookupflags)); 1172 1173 INP_HASH_LOCK_ASSERT(pcbinfo); 1174 1175 /* 1176 * First look for an exact match. 1177 */ 1178 tmpinp = NULL; 1179 head = &pcbinfo->ipi_hashbase[INP_PCBHASH( 1180 INP6_PCBHASHKEY(faddr), lport, fport, pcbinfo->ipi_hashmask)]; 1181 CK_LIST_FOREACH(inp, head, inp_hash) { 1182 /* XXX inp locking */ 1183 if ((inp->inp_vflag & INP_IPV6) == 0) 1184 continue; 1185 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && 1186 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 1187 inp->inp_fport == fport && 1188 inp->inp_lport == lport) { 1189 /* 1190 * XXX We should be able to directly return 1191 * the inp here, without any checks. 1192 * Well unless both bound with SO_REUSEPORT? 1193 */ 1194 if (prison_flag(inp->inp_cred, PR_IP6)) 1195 return (inp); 1196 if (tmpinp == NULL) 1197 tmpinp = inp; 1198 } 1199 } 1200 if (tmpinp != NULL) 1201 return (tmpinp); 1202 1203 /* 1204 * Then look in lb group (for wildcard match). 1205 */ 1206 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 1207 inp = in6_pcblookup_lbgroup(pcbinfo, laddr, lport, faddr, 1208 fport, lookupflags, numa_domain); 1209 if (inp != NULL) 1210 return (inp); 1211 } 1212 1213 /* 1214 * Then look for a wildcard match, if requested. 1215 */ 1216 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 1217 struct inpcb *local_wild = NULL, *local_exact = NULL; 1218 struct inpcb *jail_wild = NULL; 1219 int injail; 1220 1221 /* 1222 * Order of socket selection - we always prefer jails. 1223 * 1. jailed, non-wild. 1224 * 2. jailed, wild. 1225 * 3. non-jailed, non-wild. 1226 * 4. non-jailed, wild. 1227 */ 1228 head = &pcbinfo->ipi_hashbase[INP_PCBHASH( 1229 INP6_PCBHASHKEY(&in6addr_any), lport, 0, 1230 pcbinfo->ipi_hashmask)]; 1231 CK_LIST_FOREACH(inp, head, inp_hash) { 1232 /* XXX inp locking */ 1233 if ((inp->inp_vflag & INP_IPV6) == 0) 1234 continue; 1235 1236 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 1237 inp->inp_lport != lport) { 1238 continue; 1239 } 1240 1241 injail = prison_flag(inp->inp_cred, PR_IP6); 1242 if (injail) { 1243 if (prison_check_ip6(inp->inp_cred, 1244 laddr) != 0) 1245 continue; 1246 } else { 1247 if (local_exact != NULL) 1248 continue; 1249 } 1250 1251 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { 1252 if (injail) 1253 return (inp); 1254 else 1255 local_exact = inp; 1256 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 1257 if (injail) 1258 jail_wild = inp; 1259 else 1260 local_wild = inp; 1261 } 1262 } /* LIST_FOREACH */ 1263 1264 if (jail_wild != NULL) 1265 return (jail_wild); 1266 if (local_exact != NULL) 1267 return (local_exact); 1268 if (local_wild != NULL) 1269 return (local_wild); 1270 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ 1271 1272 /* 1273 * Not found. 1274 */ 1275 return (NULL); 1276 } 1277 1278 /* 1279 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the 1280 * hash list lock, and will return the inpcb locked (i.e., requires 1281 * INPLOOKUP_LOCKPCB). 1282 */ 1283 static struct inpcb * 1284 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 1285 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, 1286 struct ifnet *ifp, uint8_t numa_domain) 1287 { 1288 struct inpcb *inp; 1289 1290 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, 1291 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp, 1292 numa_domain); 1293 if (inp != NULL) { 1294 if (lookupflags & INPLOOKUP_WLOCKPCB) { 1295 INP_WLOCK(inp); 1296 if (__predict_false(inp->inp_flags2 & INP_FREED)) { 1297 INP_WUNLOCK(inp); 1298 inp = NULL; 1299 } 1300 } else if (lookupflags & INPLOOKUP_RLOCKPCB) { 1301 INP_RLOCK(inp); 1302 if (__predict_false(inp->inp_flags2 & INP_FREED)) { 1303 INP_RUNLOCK(inp); 1304 inp = NULL; 1305 } 1306 } else 1307 panic("%s: locking bug", __func__); 1308 #ifdef INVARIANTS 1309 if (inp != NULL) { 1310 if (lookupflags & INPLOOKUP_WLOCKPCB) 1311 INP_WLOCK_ASSERT(inp); 1312 else 1313 INP_RLOCK_ASSERT(inp); 1314 } 1315 #endif 1316 } 1317 return (inp); 1318 } 1319 1320 /* 1321 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf 1322 * from which a pre-calculated hash value may be extracted. 1323 * 1324 * Possibly more of this logic should be in in6_pcbgroup.c. 1325 */ 1326 struct inpcb * 1327 in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport, 1328 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp) 1329 { 1330 #if defined(PCBGROUP) && !defined(RSS) 1331 struct inpcbgroup *pcbgroup; 1332 #endif 1333 1334 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, 1335 ("%s: invalid lookup flags %d", __func__, lookupflags)); 1336 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, 1337 ("%s: LOCKPCB not set", __func__)); 1338 1339 /* 1340 * When not using RSS, use connection groups in preference to the 1341 * reservation table when looking up 4-tuples. When using RSS, just 1342 * use the reservation table, due to the cost of the Toeplitz hash 1343 * in software. 1344 * 1345 * XXXRW: This policy belongs in the pcbgroup code, as in principle 1346 * we could be doing RSS with a non-Toeplitz hash that is affordable 1347 * in software. 1348 */ 1349 #if defined(PCBGROUP) && !defined(RSS) 1350 if (in_pcbgroup_enabled(pcbinfo)) { 1351 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, 1352 fport); 1353 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, 1354 laddr, lport, lookupflags, ifp)); 1355 } 1356 #endif 1357 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, 1358 lookupflags, ifp, M_NODOM)); 1359 } 1360 1361 struct inpcb * 1362 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 1363 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, 1364 struct ifnet *ifp, struct mbuf *m) 1365 { 1366 #ifdef PCBGROUP 1367 struct inpcbgroup *pcbgroup; 1368 #endif 1369 1370 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, 1371 ("%s: invalid lookup flags %d", __func__, lookupflags)); 1372 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, 1373 ("%s: LOCKPCB not set", __func__)); 1374 1375 #ifdef PCBGROUP 1376 /* 1377 * If we can use a hardware-generated hash to look up the connection 1378 * group, use that connection group to find the inpcb. Otherwise 1379 * fall back on a software hash -- or the reservation table if we're 1380 * using RSS. 1381 * 1382 * XXXRW: As above, that policy belongs in the pcbgroup code. 1383 */ 1384 if (in_pcbgroup_enabled(pcbinfo) && 1385 M_HASHTYPE_TEST(m, M_HASHTYPE_NONE) == 0) { 1386 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m), 1387 m->m_pkthdr.flowid); 1388 if (pcbgroup != NULL) 1389 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, 1390 fport, laddr, lport, lookupflags, ifp)); 1391 #ifndef RSS 1392 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, 1393 fport); 1394 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, 1395 laddr, lport, lookupflags, ifp)); 1396 #endif 1397 } 1398 #endif 1399 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, 1400 lookupflags, ifp, m->m_pkthdr.numa_domain)); 1401 } 1402 1403 void 1404 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m, int srcordst) 1405 { 1406 struct ip6_hdr *ip; 1407 1408 ip = mtod(m, struct ip6_hdr *); 1409 bzero(sin6, sizeof(*sin6)); 1410 sin6->sin6_len = sizeof(*sin6); 1411 sin6->sin6_family = AF_INET6; 1412 sin6->sin6_addr = srcordst ? ip->ip6_dst : ip->ip6_src; 1413 1414 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */ 1415 1416 return; 1417 }
Cache object: 9ead06e309fe75403ff1996ef2cc116c
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]