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 #include "opt_rss.h" 75 76 #include <sys/param.h> 77 #include <sys/systm.h> 78 #include <sys/malloc.h> 79 #include <sys/mbuf.h> 80 #include <sys/domain.h> 81 #include <sys/protosw.h> 82 #include <sys/socket.h> 83 #include <sys/socketvar.h> 84 #include <sys/sockio.h> 85 #include <sys/errno.h> 86 #include <sys/time.h> 87 #include <sys/priv.h> 88 #include <sys/proc.h> 89 #include <sys/jail.h> 90 91 #include <vm/uma.h> 92 93 #include <net/if.h> 94 #include <net/if_var.h> 95 #include <net/if_types.h> 96 #include <net/route.h> 97 98 #include <netinet/in.h> 99 #include <netinet/in_var.h> 100 #include <netinet/in_systm.h> 101 #include <netinet/tcp_var.h> 102 #include <netinet/ip6.h> 103 #include <netinet/ip_var.h> 104 105 #include <netinet6/ip6_var.h> 106 #include <netinet6/nd6.h> 107 #include <netinet/in_pcb.h> 108 #include <netinet6/in6_pcb.h> 109 #include <netinet6/scope6_var.h> 110 111 static struct inpcb *in6_pcblookup_hash_locked(struct inpcbinfo *, 112 struct in6_addr *, u_int, struct in6_addr *, u_int, int, struct ifnet *); 113 114 int 115 in6_pcbbind(register struct inpcb *inp, struct sockaddr *nam, 116 struct ucred *cred) 117 { 118 struct socket *so = inp->inp_socket; 119 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL; 120 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 121 u_short lport = 0; 122 int error, lookupflags = 0; 123 int reuseport = (so->so_options & SO_REUSEPORT); 124 125 INP_WLOCK_ASSERT(inp); 126 INP_HASH_WLOCK_ASSERT(pcbinfo); 127 128 if (TAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */ 129 return (EADDRNOTAVAIL); 130 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 131 return (EINVAL); 132 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 133 lookupflags = INPLOOKUP_WILDCARD; 134 if (nam == NULL) { 135 if ((error = prison_local_ip6(cred, &inp->in6p_laddr, 136 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 137 return (error); 138 } else { 139 sin6 = (struct sockaddr_in6 *)nam; 140 if (nam->sa_len != sizeof(*sin6)) 141 return (EINVAL); 142 /* 143 * family check. 144 */ 145 if (nam->sa_family != AF_INET6) 146 return (EAFNOSUPPORT); 147 148 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) 149 return(error); 150 151 if ((error = prison_local_ip6(cred, &sin6->sin6_addr, 152 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) 153 return (error); 154 155 lport = sin6->sin6_port; 156 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 157 /* 158 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 159 * allow compepte duplication of binding if 160 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 161 * and a multicast address is bound on both 162 * new and duplicated sockets. 163 */ 164 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0) 165 reuseport = SO_REUSEADDR|SO_REUSEPORT; 166 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 167 struct ifaddr *ifa; 168 169 sin6->sin6_port = 0; /* yech... */ 170 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) == 171 NULL && 172 (inp->inp_flags & INP_BINDANY) == 0) { 173 return (EADDRNOTAVAIL); 174 } 175 176 /* 177 * XXX: bind to an anycast address might accidentally 178 * cause sending a packet with anycast source address. 179 * We should allow to bind to a deprecated address, since 180 * the application dares to use it. 181 */ 182 if (ifa != NULL && 183 ((struct in6_ifaddr *)ifa)->ia6_flags & 184 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) { 185 ifa_free(ifa); 186 return (EADDRNOTAVAIL); 187 } 188 if (ifa != NULL) 189 ifa_free(ifa); 190 } 191 if (lport) { 192 struct inpcb *t; 193 struct tcptw *tw; 194 195 /* GROSS */ 196 if (ntohs(lport) <= V_ipport_reservedhigh && 197 ntohs(lport) >= V_ipport_reservedlow && 198 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 199 0)) 200 return (EACCES); 201 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) && 202 priv_check_cred(inp->inp_cred, 203 PRIV_NETINET_REUSEPORT, 0) != 0) { 204 t = in6_pcblookup_local(pcbinfo, 205 &sin6->sin6_addr, lport, 206 INPLOOKUP_WILDCARD, cred); 207 if (t && 208 ((inp->inp_flags2 & INP_BINDMULTI) == 0) && 209 ((t->inp_flags & INP_TIMEWAIT) == 0) && 210 (so->so_type != SOCK_STREAM || 211 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) && 212 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || 213 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) || 214 (t->inp_flags2 & INP_REUSEPORT) == 0) && 215 (inp->inp_cred->cr_uid != 216 t->inp_cred->cr_uid)) 217 return (EADDRINUSE); 218 219 /* 220 * If the socket is a BINDMULTI socket, then 221 * the credentials need to match and the 222 * original socket also has to have been bound 223 * with BINDMULTI. 224 */ 225 if (t && (! in_pcbbind_check_bindmulti(inp, t))) 226 return (EADDRINUSE); 227 228 #ifdef INET 229 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && 230 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 231 struct sockaddr_in sin; 232 233 in6_sin6_2_sin(&sin, sin6); 234 t = in_pcblookup_local(pcbinfo, 235 sin.sin_addr, lport, 236 INPLOOKUP_WILDCARD, cred); 237 if (t && 238 ((inp->inp_flags2 & INP_BINDMULTI) == 0) && 239 ((t->inp_flags & 240 INP_TIMEWAIT) == 0) && 241 (so->so_type != SOCK_STREAM || 242 ntohl(t->inp_faddr.s_addr) == 243 INADDR_ANY) && 244 (inp->inp_cred->cr_uid != 245 t->inp_cred->cr_uid)) 246 return (EADDRINUSE); 247 248 if (t && (! in_pcbbind_check_bindmulti(inp, t))) 249 return (EADDRINUSE); 250 } 251 #endif 252 } 253 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr, 254 lport, lookupflags, cred); 255 if (t && (t->inp_flags & INP_TIMEWAIT)) { 256 /* 257 * XXXRW: If an incpb has had its timewait 258 * state recycled, we treat the address as 259 * being in use (for now). This is better 260 * than a panic, but not desirable. 261 */ 262 tw = intotw(t); 263 if (tw == NULL || 264 (reuseport & tw->tw_so_options) == 0) 265 return (EADDRINUSE); 266 } else if (t && (reuseport & inp_so_options(t)) == 0) { 267 return (EADDRINUSE); 268 } 269 #ifdef INET 270 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && 271 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 272 struct sockaddr_in sin; 273 274 in6_sin6_2_sin(&sin, sin6); 275 t = in_pcblookup_local(pcbinfo, sin.sin_addr, 276 lport, lookupflags, cred); 277 if (t && t->inp_flags & INP_TIMEWAIT) { 278 tw = intotw(t); 279 if (tw == NULL) 280 return (EADDRINUSE); 281 if ((reuseport & tw->tw_so_options) == 0 282 && (ntohl(t->inp_laddr.s_addr) != 283 INADDR_ANY || ((inp->inp_vflag & 284 INP_IPV6PROTO) == 285 (t->inp_vflag & INP_IPV6PROTO)))) 286 return (EADDRINUSE); 287 } else if (t && 288 (reuseport & inp_so_options(t)) == 0 && 289 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 290 (t->inp_vflag & INP_IPV6PROTO) != 0)) 291 return (EADDRINUSE); 292 } 293 #endif 294 } 295 inp->in6p_laddr = sin6->sin6_addr; 296 } 297 if (lport == 0) { 298 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) { 299 /* Undo an address bind that may have occurred. */ 300 inp->in6p_laddr = in6addr_any; 301 return (error); 302 } 303 } else { 304 inp->inp_lport = lport; 305 if (in_pcbinshash(inp) != 0) { 306 inp->in6p_laddr = in6addr_any; 307 inp->inp_lport = 0; 308 return (EAGAIN); 309 } 310 } 311 return (0); 312 } 313 314 /* 315 * Transform old in6_pcbconnect() into an inner subroutine for new 316 * in6_pcbconnect(): Do some validity-checking on the remote 317 * address (in mbuf 'nam') and then determine local host address 318 * (i.e., which interface) to use to access that remote host. 319 * 320 * This preserves definition of in6_pcbconnect(), while supporting a 321 * slightly different version for T/TCP. (This is more than 322 * a bit of a kludge, but cleaning up the internal interfaces would 323 * have forced minor changes in every protocol). 324 */ 325 static int 326 in6_pcbladdr(register struct inpcb *inp, struct sockaddr *nam, 327 struct in6_addr *plocal_addr6) 328 { 329 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 330 int error = 0; 331 int scope_ambiguous = 0; 332 struct in6_addr in6a; 333 334 INP_WLOCK_ASSERT(inp); 335 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */ 336 337 if (nam->sa_len != sizeof (*sin6)) 338 return (EINVAL); 339 if (sin6->sin6_family != AF_INET6) 340 return (EAFNOSUPPORT); 341 if (sin6->sin6_port == 0) 342 return (EADDRNOTAVAIL); 343 344 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) 345 scope_ambiguous = 1; 346 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) 347 return(error); 348 349 if (!TAILQ_EMPTY(&V_in6_ifaddrhead)) { 350 /* 351 * If the destination address is UNSPECIFIED addr, 352 * use the loopback addr, e.g ::1. 353 */ 354 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 355 sin6->sin6_addr = in6addr_loopback; 356 } 357 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0) 358 return (error); 359 360 error = in6_selectsrc_socket(sin6, inp->in6p_outputopts, 361 inp, inp->inp_cred, scope_ambiguous, &in6a, NULL); 362 if (error) 363 return (error); 364 365 /* 366 * Do not update this earlier, in case we return with an error. 367 * 368 * XXX: this in6_selectsrc_socket result might replace the bound local 369 * address with the address specified by setsockopt(IPV6_PKTINFO). 370 * Is it the intended behavior? 371 */ 372 *plocal_addr6 = in6a; 373 374 /* 375 * Don't do pcblookup call here; return interface in 376 * plocal_addr6 377 * and exit to caller, that will do the lookup. 378 */ 379 380 return (0); 381 } 382 383 /* 384 * Outer subroutine: 385 * Connect from a socket to a specified address. 386 * Both address and port must be specified in argument sin. 387 * If don't have a local address for this socket yet, 388 * then pick one. 389 */ 390 int 391 in6_pcbconnect_mbuf(register struct inpcb *inp, struct sockaddr *nam, 392 struct ucred *cred, struct mbuf *m) 393 { 394 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 395 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 396 struct in6_addr addr6; 397 int error; 398 399 INP_WLOCK_ASSERT(inp); 400 INP_HASH_WLOCK_ASSERT(pcbinfo); 401 402 /* 403 * Call inner routine, to assign local interface address. 404 * in6_pcbladdr() may automatically fill in sin6_scope_id. 405 */ 406 if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0) 407 return (error); 408 409 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr, 410 sin6->sin6_port, 411 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) 412 ? &addr6 : &inp->in6p_laddr, 413 inp->inp_lport, 0, NULL) != NULL) { 414 return (EADDRINUSE); 415 } 416 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 417 if (inp->inp_lport == 0) { 418 error = in6_pcbbind(inp, (struct sockaddr *)0, cred); 419 if (error) 420 return (error); 421 } 422 inp->in6p_laddr = addr6; 423 } 424 inp->in6p_faddr = sin6->sin6_addr; 425 inp->inp_fport = sin6->sin6_port; 426 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ 427 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; 428 if (inp->inp_flags & IN6P_AUTOFLOWLABEL) 429 inp->inp_flow |= 430 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); 431 432 in_pcbrehash_mbuf(inp, m); 433 434 return (0); 435 } 436 437 int 438 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 439 { 440 441 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL)); 442 } 443 444 void 445 in6_pcbdisconnect(struct inpcb *inp) 446 { 447 448 INP_WLOCK_ASSERT(inp); 449 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); 450 451 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr)); 452 inp->inp_fport = 0; 453 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ 454 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; 455 in_pcbrehash(inp); 456 } 457 458 struct sockaddr * 459 in6_sockaddr(in_port_t port, struct in6_addr *addr_p) 460 { 461 struct sockaddr_in6 *sin6; 462 463 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK); 464 bzero(sin6, sizeof *sin6); 465 sin6->sin6_family = AF_INET6; 466 sin6->sin6_len = sizeof(*sin6); 467 sin6->sin6_port = port; 468 sin6->sin6_addr = *addr_p; 469 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */ 470 471 return (struct sockaddr *)sin6; 472 } 473 474 struct sockaddr * 475 in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p) 476 { 477 struct sockaddr_in sin; 478 struct sockaddr_in6 *sin6_p; 479 480 bzero(&sin, sizeof sin); 481 sin.sin_family = AF_INET; 482 sin.sin_len = sizeof(sin); 483 sin.sin_port = port; 484 sin.sin_addr = *addr_p; 485 486 sin6_p = malloc(sizeof *sin6_p, M_SONAME, 487 M_WAITOK); 488 in6_sin_2_v4mapsin6(&sin, sin6_p); 489 490 return (struct sockaddr *)sin6_p; 491 } 492 493 int 494 in6_getsockaddr(struct socket *so, struct sockaddr **nam) 495 { 496 register struct inpcb *inp; 497 struct in6_addr addr; 498 in_port_t port; 499 500 inp = sotoinpcb(so); 501 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL")); 502 503 INP_RLOCK(inp); 504 port = inp->inp_lport; 505 addr = inp->in6p_laddr; 506 INP_RUNLOCK(inp); 507 508 *nam = in6_sockaddr(port, &addr); 509 return 0; 510 } 511 512 int 513 in6_getpeeraddr(struct socket *so, struct sockaddr **nam) 514 { 515 struct inpcb *inp; 516 struct in6_addr addr; 517 in_port_t port; 518 519 inp = sotoinpcb(so); 520 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL")); 521 522 INP_RLOCK(inp); 523 port = inp->inp_fport; 524 addr = inp->in6p_faddr; 525 INP_RUNLOCK(inp); 526 527 *nam = in6_sockaddr(port, &addr); 528 return 0; 529 } 530 531 int 532 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam) 533 { 534 struct inpcb *inp; 535 int error; 536 537 inp = sotoinpcb(so); 538 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL")); 539 540 #ifdef INET 541 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { 542 error = in_getsockaddr(so, nam); 543 if (error == 0) 544 in6_sin_2_v4mapsin6_in_sock(nam); 545 } else 546 #endif 547 { 548 /* scope issues will be handled in in6_getsockaddr(). */ 549 error = in6_getsockaddr(so, nam); 550 } 551 552 return error; 553 } 554 555 int 556 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam) 557 { 558 struct inpcb *inp; 559 int error; 560 561 inp = sotoinpcb(so); 562 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL")); 563 564 #ifdef INET 565 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { 566 error = in_getpeeraddr(so, nam); 567 if (error == 0) 568 in6_sin_2_v4mapsin6_in_sock(nam); 569 } else 570 #endif 571 /* scope issues will be handled in in6_getpeeraddr(). */ 572 error = in6_getpeeraddr(so, nam); 573 574 return error; 575 } 576 577 /* 578 * Pass some notification to all connections of a protocol 579 * associated with address dst. The local address and/or port numbers 580 * may be specified to limit the search. The "usual action" will be 581 * taken, depending on the ctlinput cmd. The caller must filter any 582 * cmds that are uninteresting (e.g., no error in the map). 583 * Call the protocol specific routine (if any) to report 584 * any errors for each matching socket. 585 */ 586 void 587 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, 588 u_int fport_arg, const struct sockaddr *src, u_int lport_arg, 589 int cmd, void *cmdarg, 590 struct inpcb *(*notify)(struct inpcb *, int)) 591 { 592 struct inpcb *inp, *inp_temp; 593 struct sockaddr_in6 sa6_src, *sa6_dst; 594 u_short fport = fport_arg, lport = lport_arg; 595 u_int32_t flowinfo; 596 int errno; 597 598 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6) 599 return; 600 601 sa6_dst = (struct sockaddr_in6 *)dst; 602 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr)) 603 return; 604 605 /* 606 * note that src can be NULL when we get notify by local fragmentation. 607 */ 608 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src; 609 flowinfo = sa6_src.sin6_flowinfo; 610 611 /* 612 * Redirects go to all references to the destination, 613 * and use in6_rtchange to invalidate the route cache. 614 * Dead host indications: also use in6_rtchange to invalidate 615 * the cache, and deliver the error to all the sockets. 616 * Otherwise, if we have knowledge of the local port and address, 617 * deliver only to that socket. 618 */ 619 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) { 620 fport = 0; 621 lport = 0; 622 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr)); 623 624 if (cmd != PRC_HOSTDEAD) 625 notify = in6_rtchange; 626 } 627 errno = inet6ctlerrmap[cmd]; 628 INP_INFO_WLOCK(pcbinfo); 629 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { 630 INP_WLOCK(inp); 631 if ((inp->inp_vflag & INP_IPV6) == 0) { 632 INP_WUNLOCK(inp); 633 continue; 634 } 635 636 /* 637 * If the error designates a new path MTU for a destination 638 * and the application (associated with this socket) wanted to 639 * know the value, notify. 640 * XXX: should we avoid to notify the value to TCP sockets? 641 */ 642 if (cmd == PRC_MSGSIZE && cmdarg != NULL) 643 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, 644 *(u_int32_t *)cmdarg); 645 646 /* 647 * Detect if we should notify the error. If no source and 648 * destination ports are specifed, but non-zero flowinfo and 649 * local address match, notify the error. This is the case 650 * when the error is delivered with an encrypted buffer 651 * by ESP. Otherwise, just compare addresses and ports 652 * as usual. 653 */ 654 if (lport == 0 && fport == 0 && flowinfo && 655 inp->inp_socket != NULL && 656 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) && 657 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr)) 658 goto do_notify; 659 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, 660 &sa6_dst->sin6_addr) || 661 inp->inp_socket == 0 || 662 (lport && inp->inp_lport != lport) || 663 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) && 664 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, 665 &sa6_src.sin6_addr)) || 666 (fport && inp->inp_fport != fport)) { 667 INP_WUNLOCK(inp); 668 continue; 669 } 670 671 do_notify: 672 if (notify) { 673 if ((*notify)(inp, errno)) 674 INP_WUNLOCK(inp); 675 } else 676 INP_WUNLOCK(inp); 677 } 678 INP_INFO_WUNLOCK(pcbinfo); 679 } 680 681 /* 682 * Lookup a PCB based on the local address and port. Caller must hold the 683 * hash lock. No inpcb locks or references are acquired. 684 */ 685 struct inpcb * 686 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr, 687 u_short lport, int lookupflags, struct ucred *cred) 688 { 689 register struct inpcb *inp; 690 int matchwild = 3, wildcard; 691 692 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, 693 ("%s: invalid lookup flags %d", __func__, lookupflags)); 694 695 INP_HASH_WLOCK_ASSERT(pcbinfo); 696 697 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { 698 struct inpcbhead *head; 699 /* 700 * Look for an unconnected (wildcard foreign addr) PCB that 701 * matches the local address and port we're looking for. 702 */ 703 head = &pcbinfo->ipi_hashbase[INP_PCBHASH( 704 INP6_PCBHASHKEY(&in6addr_any), lport, 0, 705 pcbinfo->ipi_hashmask)]; 706 LIST_FOREACH(inp, head, inp_hash) { 707 /* XXX inp locking */ 708 if ((inp->inp_vflag & INP_IPV6) == 0) 709 continue; 710 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 711 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 712 inp->inp_lport == lport) { 713 /* Found. */ 714 if (cred == NULL || 715 prison_equal_ip6(cred->cr_prison, 716 inp->inp_cred->cr_prison)) 717 return (inp); 718 } 719 } 720 /* 721 * Not found. 722 */ 723 return (NULL); 724 } else { 725 struct inpcbporthead *porthash; 726 struct inpcbport *phd; 727 struct inpcb *match = NULL; 728 /* 729 * Best fit PCB lookup. 730 * 731 * First see if this local port is in use by looking on the 732 * port hash list. 733 */ 734 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, 735 pcbinfo->ipi_porthashmask)]; 736 LIST_FOREACH(phd, porthash, phd_hash) { 737 if (phd->phd_port == lport) 738 break; 739 } 740 if (phd != NULL) { 741 /* 742 * Port is in use by one or more PCBs. Look for best 743 * fit. 744 */ 745 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 746 wildcard = 0; 747 if (cred != NULL && 748 !prison_equal_ip6(cred->cr_prison, 749 inp->inp_cred->cr_prison)) 750 continue; 751 /* XXX inp locking */ 752 if ((inp->inp_vflag & INP_IPV6) == 0) 753 continue; 754 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) 755 wildcard++; 756 if (!IN6_IS_ADDR_UNSPECIFIED( 757 &inp->in6p_laddr)) { 758 if (IN6_IS_ADDR_UNSPECIFIED(laddr)) 759 wildcard++; 760 else if (!IN6_ARE_ADDR_EQUAL( 761 &inp->in6p_laddr, laddr)) 762 continue; 763 } else { 764 if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) 765 wildcard++; 766 } 767 if (wildcard < matchwild) { 768 match = inp; 769 matchwild = wildcard; 770 if (matchwild == 0) 771 break; 772 } 773 } 774 } 775 return (match); 776 } 777 } 778 779 void 780 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) 781 { 782 struct inpcb *in6p; 783 struct ip6_moptions *im6o; 784 int i, gap; 785 786 INP_INFO_WLOCK(pcbinfo); 787 LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) { 788 INP_WLOCK(in6p); 789 im6o = in6p->in6p_moptions; 790 if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) { 791 /* 792 * Unselect the outgoing ifp for multicast if it 793 * is being detached. 794 */ 795 if (im6o->im6o_multicast_ifp == ifp) 796 im6o->im6o_multicast_ifp = NULL; 797 /* 798 * Drop multicast group membership if we joined 799 * through the interface being detached. 800 */ 801 gap = 0; 802 for (i = 0; i < im6o->im6o_num_memberships; i++) { 803 if (im6o->im6o_membership[i]->in6m_ifp == 804 ifp) { 805 in6_mc_leave(im6o->im6o_membership[i], 806 NULL); 807 gap++; 808 } else if (gap != 0) { 809 im6o->im6o_membership[i - gap] = 810 im6o->im6o_membership[i]; 811 } 812 } 813 im6o->im6o_num_memberships -= gap; 814 } 815 INP_WUNLOCK(in6p); 816 } 817 INP_INFO_WUNLOCK(pcbinfo); 818 } 819 820 /* 821 * Check for alternatives when higher level complains 822 * about service problems. For now, invalidate cached 823 * routing information. If the route was created dynamically 824 * (by a redirect), time to try a default gateway again. 825 */ 826 void 827 in6_losing(struct inpcb *in6p) 828 { 829 830 if (in6p->inp_route6.ro_rt) { 831 RTFREE(in6p->inp_route6.ro_rt); 832 in6p->inp_route6.ro_rt = (struct rtentry *)NULL; 833 } 834 return; 835 } 836 837 /* 838 * After a routing change, flush old routing 839 * and allocate a (hopefully) better one. 840 */ 841 struct inpcb * 842 in6_rtchange(struct inpcb *inp, int errno) 843 { 844 845 if (inp->inp_route6.ro_rt) { 846 RTFREE(inp->inp_route6.ro_rt); 847 inp->inp_route6.ro_rt = (struct rtentry *)NULL; 848 } 849 return inp; 850 } 851 852 #ifdef PCBGROUP 853 /* 854 * Lookup PCB in hash list, using pcbgroup tables. 855 */ 856 static struct inpcb * 857 in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup, 858 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr, 859 u_int lport_arg, int lookupflags, struct ifnet *ifp) 860 { 861 struct inpcbhead *head; 862 struct inpcb *inp, *tmpinp; 863 u_short fport = fport_arg, lport = lport_arg; 864 865 /* 866 * First look for an exact match. 867 */ 868 tmpinp = NULL; 869 INP_GROUP_LOCK(pcbgroup); 870 head = &pcbgroup->ipg_hashbase[INP_PCBHASH( 871 INP6_PCBHASHKEY(faddr), lport, fport, pcbgroup->ipg_hashmask)]; 872 LIST_FOREACH(inp, head, inp_pcbgrouphash) { 873 /* XXX inp locking */ 874 if ((inp->inp_vflag & INP_IPV6) == 0) 875 continue; 876 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && 877 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 878 inp->inp_fport == fport && 879 inp->inp_lport == lport) { 880 /* 881 * XXX We should be able to directly return 882 * the inp here, without any checks. 883 * Well unless both bound with SO_REUSEPORT? 884 */ 885 if (prison_flag(inp->inp_cred, PR_IP6)) 886 goto found; 887 if (tmpinp == NULL) 888 tmpinp = inp; 889 } 890 } 891 if (tmpinp != NULL) { 892 inp = tmpinp; 893 goto found; 894 } 895 896 /* 897 * Then look for a wildcard match in the pcbgroup. 898 */ 899 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 900 struct inpcb *local_wild = NULL, *local_exact = NULL; 901 struct inpcb *jail_wild = NULL; 902 int injail; 903 904 /* 905 * Order of socket selection - we always prefer jails. 906 * 1. jailed, non-wild. 907 * 2. jailed, wild. 908 * 3. non-jailed, non-wild. 909 * 4. non-jailed, wild. 910 */ 911 head = &pcbgroup->ipg_hashbase[ 912 INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)]; 913 LIST_FOREACH(inp, head, inp_pcbgrouphash) { 914 /* XXX inp locking */ 915 if ((inp->inp_vflag & INP_IPV6) == 0) 916 continue; 917 918 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 919 inp->inp_lport != lport) { 920 continue; 921 } 922 923 injail = prison_flag(inp->inp_cred, PR_IP6); 924 if (injail) { 925 if (prison_check_ip6(inp->inp_cred, 926 laddr) != 0) 927 continue; 928 } else { 929 if (local_exact != NULL) 930 continue; 931 } 932 933 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { 934 if (injail) 935 goto found; 936 else 937 local_exact = inp; 938 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 939 if (injail) 940 jail_wild = inp; 941 else 942 local_wild = inp; 943 } 944 } /* LIST_FOREACH */ 945 946 inp = jail_wild; 947 if (inp == NULL) 948 inp = jail_wild; 949 if (inp == NULL) 950 inp = local_exact; 951 if (inp == NULL) 952 inp = local_wild; 953 if (inp != NULL) 954 goto found; 955 } 956 957 /* 958 * Then look for a wildcard match, if requested. 959 */ 960 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 961 struct inpcb *local_wild = NULL, *local_exact = NULL; 962 struct inpcb *jail_wild = NULL; 963 int injail; 964 965 /* 966 * Order of socket selection - we always prefer jails. 967 * 1. jailed, non-wild. 968 * 2. jailed, wild. 969 * 3. non-jailed, non-wild. 970 * 4. non-jailed, wild. 971 */ 972 head = &pcbinfo->ipi_wildbase[INP_PCBHASH( 973 INP6_PCBHASHKEY(&in6addr_any), lport, 0, 974 pcbinfo->ipi_wildmask)]; 975 LIST_FOREACH(inp, head, inp_pcbgroup_wild) { 976 /* XXX inp locking */ 977 if ((inp->inp_vflag & INP_IPV6) == 0) 978 continue; 979 980 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 981 inp->inp_lport != lport) { 982 continue; 983 } 984 985 injail = prison_flag(inp->inp_cred, PR_IP6); 986 if (injail) { 987 if (prison_check_ip6(inp->inp_cred, 988 laddr) != 0) 989 continue; 990 } else { 991 if (local_exact != NULL) 992 continue; 993 } 994 995 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { 996 if (injail) 997 goto found; 998 else 999 local_exact = inp; 1000 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 1001 if (injail) 1002 jail_wild = inp; 1003 else 1004 local_wild = inp; 1005 } 1006 } /* LIST_FOREACH */ 1007 1008 inp = jail_wild; 1009 if (inp == NULL) 1010 inp = jail_wild; 1011 if (inp == NULL) 1012 inp = local_exact; 1013 if (inp == NULL) 1014 inp = local_wild; 1015 if (inp != NULL) 1016 goto found; 1017 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ 1018 INP_GROUP_UNLOCK(pcbgroup); 1019 return (NULL); 1020 1021 found: 1022 in_pcbref(inp); 1023 INP_GROUP_UNLOCK(pcbgroup); 1024 if (lookupflags & INPLOOKUP_WLOCKPCB) { 1025 INP_WLOCK(inp); 1026 if (in_pcbrele_wlocked(inp)) 1027 return (NULL); 1028 } else if (lookupflags & INPLOOKUP_RLOCKPCB) { 1029 INP_RLOCK(inp); 1030 if (in_pcbrele_rlocked(inp)) 1031 return (NULL); 1032 } else 1033 panic("%s: locking buf", __func__); 1034 return (inp); 1035 } 1036 #endif /* PCBGROUP */ 1037 1038 /* 1039 * Lookup PCB in hash list. 1040 */ 1041 static struct inpcb * 1042 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 1043 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg, 1044 int lookupflags, struct ifnet *ifp) 1045 { 1046 struct inpcbhead *head; 1047 struct inpcb *inp, *tmpinp; 1048 u_short fport = fport_arg, lport = lport_arg; 1049 1050 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, 1051 ("%s: invalid lookup flags %d", __func__, lookupflags)); 1052 1053 INP_HASH_LOCK_ASSERT(pcbinfo); 1054 1055 /* 1056 * First look for an exact match. 1057 */ 1058 tmpinp = NULL; 1059 head = &pcbinfo->ipi_hashbase[INP_PCBHASH( 1060 INP6_PCBHASHKEY(faddr), lport, fport, pcbinfo->ipi_hashmask)]; 1061 LIST_FOREACH(inp, head, inp_hash) { 1062 /* XXX inp locking */ 1063 if ((inp->inp_vflag & INP_IPV6) == 0) 1064 continue; 1065 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && 1066 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && 1067 inp->inp_fport == fport && 1068 inp->inp_lport == lport) { 1069 /* 1070 * XXX We should be able to directly return 1071 * the inp here, without any checks. 1072 * Well unless both bound with SO_REUSEPORT? 1073 */ 1074 if (prison_flag(inp->inp_cred, PR_IP6)) 1075 return (inp); 1076 if (tmpinp == NULL) 1077 tmpinp = inp; 1078 } 1079 } 1080 if (tmpinp != NULL) 1081 return (tmpinp); 1082 1083 /* 1084 * Then look for a wildcard match, if requested. 1085 */ 1086 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 1087 struct inpcb *local_wild = NULL, *local_exact = NULL; 1088 struct inpcb *jail_wild = NULL; 1089 int injail; 1090 1091 /* 1092 * Order of socket selection - we always prefer jails. 1093 * 1. jailed, non-wild. 1094 * 2. jailed, wild. 1095 * 3. non-jailed, non-wild. 1096 * 4. non-jailed, wild. 1097 */ 1098 head = &pcbinfo->ipi_hashbase[INP_PCBHASH( 1099 INP6_PCBHASHKEY(&in6addr_any), lport, 0, 1100 pcbinfo->ipi_hashmask)]; 1101 LIST_FOREACH(inp, head, inp_hash) { 1102 /* XXX inp locking */ 1103 if ((inp->inp_vflag & INP_IPV6) == 0) 1104 continue; 1105 1106 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || 1107 inp->inp_lport != lport) { 1108 continue; 1109 } 1110 1111 injail = prison_flag(inp->inp_cred, PR_IP6); 1112 if (injail) { 1113 if (prison_check_ip6(inp->inp_cred, 1114 laddr) != 0) 1115 continue; 1116 } else { 1117 if (local_exact != NULL) 1118 continue; 1119 } 1120 1121 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { 1122 if (injail) 1123 return (inp); 1124 else 1125 local_exact = inp; 1126 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { 1127 if (injail) 1128 jail_wild = inp; 1129 else 1130 local_wild = inp; 1131 } 1132 } /* LIST_FOREACH */ 1133 1134 if (jail_wild != NULL) 1135 return (jail_wild); 1136 if (local_exact != NULL) 1137 return (local_exact); 1138 if (local_wild != NULL) 1139 return (local_wild); 1140 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ 1141 1142 /* 1143 * Not found. 1144 */ 1145 return (NULL); 1146 } 1147 1148 /* 1149 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the 1150 * hash list lock, and will return the inpcb locked (i.e., requires 1151 * INPLOOKUP_LOCKPCB). 1152 */ 1153 static struct inpcb * 1154 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 1155 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, 1156 struct ifnet *ifp) 1157 { 1158 struct inpcb *inp; 1159 1160 INP_HASH_RLOCK(pcbinfo); 1161 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, 1162 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp); 1163 if (inp != NULL) { 1164 in_pcbref(inp); 1165 INP_HASH_RUNLOCK(pcbinfo); 1166 if (lookupflags & INPLOOKUP_WLOCKPCB) { 1167 INP_WLOCK(inp); 1168 if (in_pcbrele_wlocked(inp)) 1169 return (NULL); 1170 } else if (lookupflags & INPLOOKUP_RLOCKPCB) { 1171 INP_RLOCK(inp); 1172 if (in_pcbrele_rlocked(inp)) 1173 return (NULL); 1174 } else 1175 panic("%s: locking bug", __func__); 1176 } else 1177 INP_HASH_RUNLOCK(pcbinfo); 1178 return (inp); 1179 } 1180 1181 /* 1182 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf 1183 * from which a pre-calculated hash value may be extracted. 1184 * 1185 * Possibly more of this logic should be in in6_pcbgroup.c. 1186 */ 1187 struct inpcb * 1188 in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport, 1189 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp) 1190 { 1191 #if defined(PCBGROUP) && !defined(RSS) 1192 struct inpcbgroup *pcbgroup; 1193 #endif 1194 1195 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, 1196 ("%s: invalid lookup flags %d", __func__, lookupflags)); 1197 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, 1198 ("%s: LOCKPCB not set", __func__)); 1199 1200 /* 1201 * When not using RSS, use connection groups in preference to the 1202 * reservation table when looking up 4-tuples. When using RSS, just 1203 * use the reservation table, due to the cost of the Toeplitz hash 1204 * in software. 1205 * 1206 * XXXRW: This policy belongs in the pcbgroup code, as in principle 1207 * we could be doing RSS with a non-Toeplitz hash that is affordable 1208 * in software. 1209 */ 1210 #if defined(PCBGROUP) && !defined(RSS) 1211 if (in_pcbgroup_enabled(pcbinfo)) { 1212 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, 1213 fport); 1214 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, 1215 laddr, lport, lookupflags, ifp)); 1216 } 1217 #endif 1218 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, 1219 lookupflags, ifp)); 1220 } 1221 1222 struct inpcb * 1223 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, 1224 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, 1225 struct ifnet *ifp, struct mbuf *m) 1226 { 1227 #ifdef PCBGROUP 1228 struct inpcbgroup *pcbgroup; 1229 #endif 1230 1231 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, 1232 ("%s: invalid lookup flags %d", __func__, lookupflags)); 1233 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, 1234 ("%s: LOCKPCB not set", __func__)); 1235 1236 #ifdef PCBGROUP 1237 /* 1238 * If we can use a hardware-generated hash to look up the connection 1239 * group, use that connection group to find the inpcb. Otherwise 1240 * fall back on a software hash -- or the reservation table if we're 1241 * using RSS. 1242 * 1243 * XXXRW: As above, that policy belongs in the pcbgroup code. 1244 */ 1245 if (in_pcbgroup_enabled(pcbinfo) && 1246 M_HASHTYPE_TEST(m, M_HASHTYPE_NONE) == 0) { 1247 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m), 1248 m->m_pkthdr.flowid); 1249 if (pcbgroup != NULL) 1250 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, 1251 fport, laddr, lport, lookupflags, ifp)); 1252 #ifndef RSS 1253 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, 1254 fport); 1255 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, 1256 laddr, lport, lookupflags, ifp)); 1257 #endif 1258 } 1259 #endif 1260 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, 1261 lookupflags, ifp)); 1262 } 1263 1264 void 1265 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m) 1266 { 1267 struct ip6_hdr *ip; 1268 1269 ip = mtod(m, struct ip6_hdr *); 1270 bzero(sin6, sizeof(*sin6)); 1271 sin6->sin6_len = sizeof(*sin6); 1272 sin6->sin6_family = AF_INET6; 1273 sin6->sin6_addr = ip->ip6_src; 1274 1275 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */ 1276 1277 return; 1278 } 1279