1 /* 2 * Copyright (c) 1982, 1986, 1991, 1993, 1995 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 30 * $FreeBSD$ 31 */ 32 33 #include "opt_ipsec.h" 34 #include "opt_inet6.h" 35 #include "opt_mac.h" 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/mac.h> 40 #include <sys/malloc.h> 41 #include <sys/mbuf.h> 42 #include <sys/domain.h> 43 #include <sys/protosw.h> 44 #include <sys/socket.h> 45 #include <sys/socketvar.h> 46 #include <sys/proc.h> 47 #include <sys/jail.h> 48 #include <sys/kernel.h> 49 #include <sys/sysctl.h> 50 51 #include <vm/uma.h> 52 53 #include <net/if.h> 54 #include <net/if_types.h> 55 #include <net/route.h> 56 57 #include <netinet/in.h> 58 #include <netinet/in_pcb.h> 59 #include <netinet/in_var.h> 60 #include <netinet/ip_var.h> 61 #include <netinet/tcp_var.h> 62 #ifdef INET6 63 #include <netinet/ip6.h> 64 #include <netinet6/ip6_var.h> 65 #endif /* INET6 */ 66 67 #ifdef IPSEC 68 #include <netinet6/ipsec.h> 69 #include <netkey/key.h> 70 #endif /* IPSEC */ 71 72 #ifdef FAST_IPSEC 73 #if defined(IPSEC) || defined(IPSEC_ESP) 74 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!" 75 #endif 76 77 #include <netipsec/ipsec.h> 78 #include <netipsec/key.h> 79 #endif /* FAST_IPSEC */ 80 81 /* 82 * These configure the range of local port addresses assigned to 83 * "unspecified" outgoing connections/packets/whatever. 84 */ 85 int ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 86 int ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 87 int ipport_firstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 88 int ipport_lastauto = IPPORT_HILASTAUTO; /* 65535 */ 89 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 90 int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 91 92 /* 93 * Reserved ports accessible only to root. There are significant 94 * security considerations that must be accounted for when changing these, 95 * but the security benefits can be great. Please be careful. 96 */ 97 int ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 98 int ipport_reservedlow = 0; 99 100 /* Shall we allocate ephemeral ports in random order? */ 101 int ipport_randomized = 1; 102 103 #define RANGECHK(var, min, max) \ 104 if ((var) < (min)) { (var) = (min); } \ 105 else if ((var) > (max)) { (var) = (max); } 106 107 static int 108 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) 109 { 110 int error; 111 112 error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req); 113 if (error == 0) { 114 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); 115 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1); 116 RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); 117 RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); 118 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); 119 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); 120 } 121 return (error); 122 } 123 124 #undef RANGECHK 125 126 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); 127 128 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW, 129 &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", ""); 130 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW, 131 &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", ""); 132 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW, 133 &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", ""); 134 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW, 135 &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", ""); 136 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW, 137 &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", ""); 138 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW, 139 &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", ""); 140 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, 141 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, ""); 142 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, 143 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, ""); 144 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, 145 CTLFLAG_RW, &ipport_randomized, 0, ""); 146 147 /* 148 * in_pcb.c: manage the Protocol Control Blocks. 149 * 150 * NOTE: It is assumed that most of these functions will be called at 151 * splnet(). XXX - There are, unfortunately, a few exceptions to this 152 * rule that should be fixed. 153 */ 154 155 /* 156 * Allocate a PCB and associate it with the socket. 157 */ 158 int 159 in_pcballoc(so, pcbinfo, type) 160 struct socket *so; 161 struct inpcbinfo *pcbinfo; 162 const char *type; 163 { 164 register struct inpcb *inp; 165 int error; 166 167 INP_INFO_WLOCK_ASSERT(pcbinfo); 168 error = 0; 169 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT | M_ZERO); 170 if (inp == NULL) 171 return (ENOBUFS); 172 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 173 inp->inp_pcbinfo = pcbinfo; 174 inp->inp_socket = so; 175 #ifdef MAC 176 error = mac_init_inpcb(inp, M_NOWAIT); 177 if (error != 0) 178 goto out; 179 SOCK_LOCK(so); 180 mac_create_inpcb_from_socket(so, inp); 181 SOCK_UNLOCK(so); 182 #endif 183 #if defined(IPSEC) || defined(FAST_IPSEC) 184 #ifdef FAST_IPSEC 185 error = ipsec_init_policy(so, &inp->inp_sp); 186 #else 187 error = ipsec_init_pcbpolicy(so, &inp->inp_sp); 188 #endif 189 if (error != 0) 190 goto out; 191 #endif /*IPSEC*/ 192 #if defined(INET6) 193 if (INP_SOCKAF(so) == AF_INET6) { 194 inp->inp_vflag |= INP_IPV6PROTO; 195 if (ip6_v6only) 196 inp->inp_flags |= IN6P_IPV6_V6ONLY; 197 } 198 #endif 199 LIST_INSERT_HEAD(pcbinfo->listhead, inp, inp_list); 200 pcbinfo->ipi_count++; 201 so->so_pcb = (caddr_t)inp; 202 INP_LOCK_INIT(inp, "inp", type); 203 #ifdef INET6 204 if (ip6_auto_flowlabel) 205 inp->inp_flags |= IN6P_AUTOFLOWLABEL; 206 #endif 207 #if defined(IPSEC) || defined(FAST_IPSEC) || defined(MAC) 208 out: 209 if (error != 0) 210 uma_zfree(pcbinfo->ipi_zone, inp); 211 #endif 212 return (error); 213 } 214 215 int 216 in_pcbbind(inp, nam, cred) 217 register struct inpcb *inp; 218 struct sockaddr *nam; 219 struct ucred *cred; 220 { 221 int anonport, error; 222 223 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 224 INP_LOCK_ASSERT(inp); 225 226 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) 227 return (EINVAL); 228 anonport = inp->inp_lport == 0 && (nam == NULL || 229 ((struct sockaddr_in *)nam)->sin_port == 0); 230 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, 231 &inp->inp_lport, cred); 232 if (error) 233 return (error); 234 if (in_pcbinshash(inp) != 0) { 235 inp->inp_laddr.s_addr = INADDR_ANY; 236 inp->inp_lport = 0; 237 return (EAGAIN); 238 } 239 if (anonport) 240 inp->inp_flags |= INP_ANONPORT; 241 return (0); 242 } 243 244 /* 245 * Set up a bind operation on a PCB, performing port allocation 246 * as required, but do not actually modify the PCB. Callers can 247 * either complete the bind by setting inp_laddr/inp_lport and 248 * calling in_pcbinshash(), or they can just use the resulting 249 * port and address to authorise the sending of a once-off packet. 250 * 251 * On error, the values of *laddrp and *lportp are not changed. 252 */ 253 int 254 in_pcbbind_setup(inp, nam, laddrp, lportp, cred) 255 struct inpcb *inp; 256 struct sockaddr *nam; 257 in_addr_t *laddrp; 258 u_short *lportp; 259 struct ucred *cred; 260 { 261 struct socket *so = inp->inp_socket; 262 unsigned short *lastport; 263 struct sockaddr_in *sin; 264 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 265 struct in_addr laddr; 266 u_short lport = 0; 267 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT); 268 int error, prison = 0; 269 270 INP_INFO_WLOCK_ASSERT(pcbinfo); 271 INP_LOCK_ASSERT(inp); 272 273 if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */ 274 return (EADDRNOTAVAIL); 275 laddr.s_addr = *laddrp; 276 if (nam != NULL && laddr.s_addr != INADDR_ANY) 277 return (EINVAL); 278 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 279 wild = 1; 280 if (nam) { 281 sin = (struct sockaddr_in *)nam; 282 if (nam->sa_len != sizeof (*sin)) 283 return (EINVAL); 284 #ifdef notdef 285 /* 286 * We should check the family, but old programs 287 * incorrectly fail to initialize it. 288 */ 289 if (sin->sin_family != AF_INET) 290 return (EAFNOSUPPORT); 291 #endif 292 if (sin->sin_addr.s_addr != INADDR_ANY) 293 if (prison_ip(cred, 0, &sin->sin_addr.s_addr)) 294 return(EINVAL); 295 if (sin->sin_port != *lportp) { 296 /* Don't allow the port to change. */ 297 if (*lportp != 0) 298 return (EINVAL); 299 lport = sin->sin_port; 300 } 301 /* NB: lport is left as 0 if the port isn't being changed. */ 302 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 303 /* 304 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 305 * allow complete duplication of binding if 306 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 307 * and a multicast address is bound on both 308 * new and duplicated sockets. 309 */ 310 if (so->so_options & SO_REUSEADDR) 311 reuseport = SO_REUSEADDR|SO_REUSEPORT; 312 } else if (sin->sin_addr.s_addr != INADDR_ANY) { 313 sin->sin_port = 0; /* yech... */ 314 bzero(&sin->sin_zero, sizeof(sin->sin_zero)); 315 if (ifa_ifwithaddr((struct sockaddr *)sin) == 0) 316 return (EADDRNOTAVAIL); 317 } 318 laddr = sin->sin_addr; 319 if (lport) { 320 struct inpcb *t; 321 /* GROSS */ 322 if (ntohs(lport) <= ipport_reservedhigh && 323 ntohs(lport) >= ipport_reservedlow && 324 suser_cred(cred, SUSER_ALLOWJAIL)) 325 return (EACCES); 326 if (jailed(cred)) 327 prison = 1; 328 if (so->so_cred->cr_uid != 0 && 329 !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 330 t = in_pcblookup_local(inp->inp_pcbinfo, 331 sin->sin_addr, lport, 332 prison ? 0 : INPLOOKUP_WILDCARD); 333 /* 334 * XXX 335 * This entire block sorely needs a rewrite. 336 */ 337 if (t && 338 ((t->inp_vflag & INP_TIMEWAIT) == 0) && 339 (so->so_type != SOCK_STREAM || 340 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && 341 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 342 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 343 (t->inp_socket->so_options & 344 SO_REUSEPORT) == 0) && 345 (so->so_cred->cr_uid != 346 t->inp_socket->so_cred->cr_uid)) 347 return (EADDRINUSE); 348 } 349 if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr)) 350 return (EADDRNOTAVAIL); 351 t = in_pcblookup_local(pcbinfo, sin->sin_addr, 352 lport, prison ? 0 : wild); 353 if (t && (t->inp_vflag & INP_TIMEWAIT)) { 354 if ((reuseport & intotw(t)->tw_so_options) == 0) 355 return (EADDRINUSE); 356 } else 357 if (t && 358 (reuseport & t->inp_socket->so_options) == 0) { 359 #if defined(INET6) 360 if (ntohl(sin->sin_addr.s_addr) != 361 INADDR_ANY || 362 ntohl(t->inp_laddr.s_addr) != 363 INADDR_ANY || 364 INP_SOCKAF(so) == 365 INP_SOCKAF(t->inp_socket)) 366 #endif /* defined(INET6) */ 367 return (EADDRINUSE); 368 } 369 } 370 } 371 if (*lportp != 0) 372 lport = *lportp; 373 if (lport == 0) { 374 u_short first, last; 375 int count; 376 377 if (laddr.s_addr != INADDR_ANY) 378 if (prison_ip(cred, 0, &laddr.s_addr)) 379 return (EINVAL); 380 381 if (inp->inp_flags & INP_HIGHPORT) { 382 first = ipport_hifirstauto; /* sysctl */ 383 last = ipport_hilastauto; 384 lastport = &pcbinfo->lasthi; 385 } else if (inp->inp_flags & INP_LOWPORT) { 386 if ((error = suser_cred(cred, SUSER_ALLOWJAIL)) != 0) 387 return error; 388 first = ipport_lowfirstauto; /* 1023 */ 389 last = ipport_lowlastauto; /* 600 */ 390 lastport = &pcbinfo->lastlow; 391 } else { 392 first = ipport_firstauto; /* sysctl */ 393 last = ipport_lastauto; 394 lastport = &pcbinfo->lastport; 395 } 396 /* 397 * Simple check to ensure all ports are not used up causing 398 * a deadlock here. 399 * 400 * We split the two cases (up and down) so that the direction 401 * is not being tested on each round of the loop. 402 */ 403 if (first > last) { 404 /* 405 * counting down 406 */ 407 if (ipport_randomized) 408 *lastport = first - 409 (arc4random() % (first - last)); 410 count = first - last; 411 412 do { 413 if (count-- < 0) /* completely used? */ 414 return (EADDRNOTAVAIL); 415 --*lastport; 416 if (*lastport > first || *lastport < last) 417 *lastport = first; 418 lport = htons(*lastport); 419 } while (in_pcblookup_local(pcbinfo, laddr, lport, 420 wild)); 421 } else { 422 /* 423 * counting up 424 */ 425 if (ipport_randomized) 426 *lastport = first + 427 (arc4random() % (last - first)); 428 count = last - first; 429 430 do { 431 if (count-- < 0) /* completely used? */ 432 return (EADDRNOTAVAIL); 433 ++*lastport; 434 if (*lastport < first || *lastport > last) 435 *lastport = first; 436 lport = htons(*lastport); 437 } while (in_pcblookup_local(pcbinfo, laddr, lport, 438 wild)); 439 } 440 } 441 if (prison_ip(cred, 0, &laddr.s_addr)) 442 return (EINVAL); 443 *laddrp = laddr.s_addr; 444 *lportp = lport; 445 return (0); 446 } 447 448 /* 449 * Connect from a socket to a specified address. 450 * Both address and port must be specified in argument sin. 451 * If don't have a local address for this socket yet, 452 * then pick one. 453 */ 454 int 455 in_pcbconnect(inp, nam, cred) 456 register struct inpcb *inp; 457 struct sockaddr *nam; 458 struct ucred *cred; 459 { 460 u_short lport, fport; 461 in_addr_t laddr, faddr; 462 int anonport, error; 463 464 lport = inp->inp_lport; 465 laddr = inp->inp_laddr.s_addr; 466 anonport = (lport == 0); 467 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, 468 NULL, cred); 469 if (error) 470 return (error); 471 472 /* Do the initial binding of the local address if required. */ 473 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { 474 inp->inp_lport = lport; 475 inp->inp_laddr.s_addr = laddr; 476 if (in_pcbinshash(inp) != 0) { 477 inp->inp_laddr.s_addr = INADDR_ANY; 478 inp->inp_lport = 0; 479 return (EAGAIN); 480 } 481 } 482 483 /* Commit the remaining changes. */ 484 inp->inp_lport = lport; 485 inp->inp_laddr.s_addr = laddr; 486 inp->inp_faddr.s_addr = faddr; 487 inp->inp_fport = fport; 488 in_pcbrehash(inp); 489 #ifdef IPSEC 490 if (inp->inp_socket->so_type == SOCK_STREAM) 491 ipsec_pcbconn(inp->inp_sp); 492 #endif 493 if (anonport) 494 inp->inp_flags |= INP_ANONPORT; 495 return (0); 496 } 497 498 /* 499 * Set up for a connect from a socket to the specified address. 500 * On entry, *laddrp and *lportp should contain the current local 501 * address and port for the PCB; these are updated to the values 502 * that should be placed in inp_laddr and inp_lport to complete 503 * the connect. 504 * 505 * On success, *faddrp and *fportp will be set to the remote address 506 * and port. These are not updated in the error case. 507 * 508 * If the operation fails because the connection already exists, 509 * *oinpp will be set to the PCB of that connection so that the 510 * caller can decide to override it. In all other cases, *oinpp 511 * is set to NULL. 512 */ 513 int 514 in_pcbconnect_setup(inp, nam, laddrp, lportp, faddrp, fportp, oinpp, cred) 515 register struct inpcb *inp; 516 struct sockaddr *nam; 517 in_addr_t *laddrp; 518 u_short *lportp; 519 in_addr_t *faddrp; 520 u_short *fportp; 521 struct inpcb **oinpp; 522 struct ucred *cred; 523 { 524 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 525 struct in_ifaddr *ia; 526 struct sockaddr_in sa; 527 struct ucred *socred; 528 struct inpcb *oinp; 529 struct in_addr laddr, faddr; 530 u_short lport, fport; 531 int error; 532 533 if (oinpp != NULL) 534 *oinpp = NULL; 535 if (nam->sa_len != sizeof (*sin)) 536 return (EINVAL); 537 if (sin->sin_family != AF_INET) 538 return (EAFNOSUPPORT); 539 if (sin->sin_port == 0) 540 return (EADDRNOTAVAIL); 541 laddr.s_addr = *laddrp; 542 lport = *lportp; 543 faddr = sin->sin_addr; 544 fport = sin->sin_port; 545 socred = inp->inp_socket->so_cred; 546 if (laddr.s_addr == INADDR_ANY && jailed(socred)) { 547 bzero(&sa, sizeof(sa)); 548 sa.sin_addr.s_addr = htonl(prison_getip(socred)); 549 sa.sin_len = sizeof(sa); 550 sa.sin_family = AF_INET; 551 error = in_pcbbind_setup(inp, (struct sockaddr *)&sa, 552 &laddr.s_addr, &lport, cred); 553 if (error) 554 return (error); 555 } 556 if (!TAILQ_EMPTY(&in_ifaddrhead)) { 557 /* 558 * If the destination address is INADDR_ANY, 559 * use the primary local address. 560 * If the supplied address is INADDR_BROADCAST, 561 * and the primary interface supports broadcast, 562 * choose the broadcast address for that interface. 563 */ 564 if (faddr.s_addr == INADDR_ANY) 565 faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr; 566 else if (faddr.s_addr == (u_long)INADDR_BROADCAST && 567 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags & 568 IFF_BROADCAST)) 569 faddr = satosin(&TAILQ_FIRST( 570 &in_ifaddrhead)->ia_broadaddr)->sin_addr; 571 } 572 if (laddr.s_addr == INADDR_ANY) { 573 struct route sro; 574 575 bzero(&sro, sizeof(sro)); 576 ia = (struct in_ifaddr *)0; 577 /* 578 * If route is known our src addr is taken from the i/f, 579 * else punt. 580 */ 581 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) { 582 /* Find out route to destination */ 583 sro.ro_dst.sa_family = AF_INET; 584 sro.ro_dst.sa_len = sizeof(struct sockaddr_in); 585 ((struct sockaddr_in *)&sro.ro_dst)->sin_addr = faddr; 586 rtalloc_ign(&sro, RTF_CLONING); 587 } 588 /* 589 * If we found a route, use the address 590 * corresponding to the outgoing interface 591 * unless it is the loopback (in case a route 592 * to our address on another net goes to loopback). 593 */ 594 if (sro.ro_rt && !(sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) 595 ia = ifatoia(sro.ro_rt->rt_ifa); 596 if (sro.ro_rt) 597 RTFREE(sro.ro_rt); 598 if (ia == 0) { 599 bzero(&sa, sizeof(sa)); 600 sa.sin_addr = faddr; 601 sa.sin_len = sizeof(sa); 602 sa.sin_family = AF_INET; 603 604 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa))); 605 if (ia == 0) 606 ia = ifatoia(ifa_ifwithnet(sintosa(&sa))); 607 if (ia == 0) 608 return (ENETUNREACH); 609 } 610 /* 611 * If the destination address is multicast and an outgoing 612 * interface has been set as a multicast option, use the 613 * address of that interface as our source address. 614 */ 615 if (IN_MULTICAST(ntohl(faddr.s_addr)) && 616 inp->inp_moptions != NULL) { 617 struct ip_moptions *imo; 618 struct ifnet *ifp; 619 620 imo = inp->inp_moptions; 621 if (imo->imo_multicast_ifp != NULL) { 622 ifp = imo->imo_multicast_ifp; 623 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) 624 if (ia->ia_ifp == ifp) 625 break; 626 if (ia == 0) 627 return (EADDRNOTAVAIL); 628 } 629 } 630 laddr = ia->ia_addr.sin_addr; 631 } 632 633 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport, 634 0, NULL); 635 if (oinp != NULL) { 636 if (oinpp != NULL) 637 *oinpp = oinp; 638 return (EADDRINUSE); 639 } 640 if (lport == 0) { 641 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, 642 cred); 643 if (error) 644 return (error); 645 } 646 *laddrp = laddr.s_addr; 647 *lportp = lport; 648 *faddrp = faddr.s_addr; 649 *fportp = fport; 650 return (0); 651 } 652 653 void 654 in_pcbdisconnect(inp) 655 struct inpcb *inp; 656 { 657 INP_LOCK_ASSERT(inp); 658 659 inp->inp_faddr.s_addr = INADDR_ANY; 660 inp->inp_fport = 0; 661 in_pcbrehash(inp); 662 #ifdef IPSEC 663 ipsec_pcbdisconn(inp->inp_sp); 664 #endif 665 if (inp->inp_socket->so_state & SS_NOFDREF) 666 in_pcbdetach(inp); 667 } 668 669 void 670 in_pcbdetach(inp) 671 struct inpcb *inp; 672 { 673 struct socket *so = inp->inp_socket; 674 struct inpcbinfo *ipi = inp->inp_pcbinfo; 675 676 INP_LOCK_ASSERT(inp); 677 678 #if defined(IPSEC) || defined(FAST_IPSEC) 679 ipsec4_delete_pcbpolicy(inp); 680 #endif /*IPSEC*/ 681 inp->inp_gencnt = ++ipi->ipi_gencnt; 682 in_pcbremlists(inp); 683 if (so) { 684 SOCK_LOCK(so); 685 so->so_pcb = 0; 686 sotryfree(so); 687 } 688 if (inp->inp_options) 689 (void)m_free(inp->inp_options); 690 ip_freemoptions(inp->inp_moptions); 691 inp->inp_vflag = 0; 692 INP_LOCK_DESTROY(inp); 693 #ifdef MAC 694 mac_destroy_inpcb(inp); 695 #endif 696 uma_zfree(ipi->ipi_zone, inp); 697 } 698 699 struct sockaddr * 700 in_sockaddr(port, addr_p) 701 in_port_t port; 702 struct in_addr *addr_p; 703 { 704 struct sockaddr_in *sin; 705 706 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, 707 M_WAITOK | M_ZERO); 708 sin->sin_family = AF_INET; 709 sin->sin_len = sizeof(*sin); 710 sin->sin_addr = *addr_p; 711 sin->sin_port = port; 712 713 return (struct sockaddr *)sin; 714 } 715 716 /* 717 * The wrapper function will pass down the pcbinfo for this function to lock. 718 * The socket must have a valid 719 * (i.e., non-nil) PCB, but it should be impossible to get an invalid one 720 * except through a kernel programming error, so it is acceptable to panic 721 * (or in this case trap) if the PCB is invalid. (Actually, we don't trap 722 * because there actually /is/ a programming error somewhere... XXX) 723 */ 724 int 725 in_setsockaddr(so, nam, pcbinfo) 726 struct socket *so; 727 struct sockaddr **nam; 728 struct inpcbinfo *pcbinfo; 729 { 730 int s; 731 register struct inpcb *inp; 732 struct in_addr addr; 733 in_port_t port; 734 735 s = splnet(); 736 INP_INFO_RLOCK(pcbinfo); 737 inp = sotoinpcb(so); 738 if (!inp) { 739 INP_INFO_RUNLOCK(pcbinfo); 740 splx(s); 741 return ECONNRESET; 742 } 743 INP_LOCK(inp); 744 port = inp->inp_lport; 745 addr = inp->inp_laddr; 746 INP_UNLOCK(inp); 747 INP_INFO_RUNLOCK(pcbinfo); 748 splx(s); 749 750 *nam = in_sockaddr(port, &addr); 751 return 0; 752 } 753 754 /* 755 * The wrapper function will pass down the pcbinfo for this function to lock. 756 */ 757 int 758 in_setpeeraddr(so, nam, pcbinfo) 759 struct socket *so; 760 struct sockaddr **nam; 761 struct inpcbinfo *pcbinfo; 762 { 763 int s; 764 register struct inpcb *inp; 765 struct in_addr addr; 766 in_port_t port; 767 768 s = splnet(); 769 INP_INFO_RLOCK(pcbinfo); 770 inp = sotoinpcb(so); 771 if (!inp) { 772 INP_INFO_RUNLOCK(pcbinfo); 773 splx(s); 774 return ECONNRESET; 775 } 776 INP_LOCK(inp); 777 port = inp->inp_fport; 778 addr = inp->inp_faddr; 779 INP_UNLOCK(inp); 780 INP_INFO_RUNLOCK(pcbinfo); 781 splx(s); 782 783 *nam = in_sockaddr(port, &addr); 784 return 0; 785 } 786 787 void 788 in_pcbnotifyall(pcbinfo, faddr, errno, notify) 789 struct inpcbinfo *pcbinfo; 790 struct in_addr faddr; 791 int errno; 792 struct inpcb *(*notify)(struct inpcb *, int); 793 { 794 struct inpcb *inp, *ninp; 795 struct inpcbhead *head; 796 int s; 797 798 s = splnet(); 799 INP_INFO_WLOCK(pcbinfo); 800 head = pcbinfo->listhead; 801 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) { 802 INP_LOCK(inp); 803 ninp = LIST_NEXT(inp, inp_list); 804 #ifdef INET6 805 if ((inp->inp_vflag & INP_IPV4) == 0) { 806 INP_UNLOCK(inp); 807 continue; 808 } 809 #endif 810 if (inp->inp_faddr.s_addr != faddr.s_addr || 811 inp->inp_socket == NULL) { 812 INP_UNLOCK(inp); 813 continue; 814 } 815 if ((*notify)(inp, errno)) 816 INP_UNLOCK(inp); 817 } 818 INP_INFO_WUNLOCK(pcbinfo); 819 splx(s); 820 } 821 822 void 823 in_pcbpurgeif0(pcbinfo, ifp) 824 struct inpcbinfo *pcbinfo; 825 struct ifnet *ifp; 826 { 827 struct inpcb *inp; 828 struct ip_moptions *imo; 829 int i, gap; 830 831 /* why no splnet here? XXX */ 832 INP_INFO_RLOCK(pcbinfo); 833 LIST_FOREACH(inp, pcbinfo->listhead, inp_list) { 834 INP_LOCK(inp); 835 imo = inp->inp_moptions; 836 if ((inp->inp_vflag & INP_IPV4) && 837 imo != NULL) { 838 /* 839 * Unselect the outgoing interface if it is being 840 * detached. 841 */ 842 if (imo->imo_multicast_ifp == ifp) 843 imo->imo_multicast_ifp = NULL; 844 845 /* 846 * Drop multicast group membership if we joined 847 * through the interface being detached. 848 */ 849 for (i = 0, gap = 0; i < imo->imo_num_memberships; 850 i++) { 851 if (imo->imo_membership[i]->inm_ifp == ifp) { 852 in_delmulti(imo->imo_membership[i]); 853 gap++; 854 } else if (gap != 0) 855 imo->imo_membership[i - gap] = 856 imo->imo_membership[i]; 857 } 858 imo->imo_num_memberships -= gap; 859 } 860 INP_UNLOCK(inp); 861 } 862 INP_INFO_RUNLOCK(pcbinfo); 863 } 864 865 /* 866 * Lookup a PCB based on the local address and port. 867 */ 868 struct inpcb * 869 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay) 870 struct inpcbinfo *pcbinfo; 871 struct in_addr laddr; 872 u_int lport_arg; 873 int wild_okay; 874 { 875 register struct inpcb *inp; 876 int matchwild = 3, wildcard; 877 u_short lport = lport_arg; 878 879 INP_INFO_WLOCK_ASSERT(pcbinfo); 880 881 if (!wild_okay) { 882 struct inpcbhead *head; 883 /* 884 * Look for an unconnected (wildcard foreign addr) PCB that 885 * matches the local address and port we're looking for. 886 */ 887 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 888 LIST_FOREACH(inp, head, inp_hash) { 889 #ifdef INET6 890 if ((inp->inp_vflag & INP_IPV4) == 0) 891 continue; 892 #endif 893 if (inp->inp_faddr.s_addr == INADDR_ANY && 894 inp->inp_laddr.s_addr == laddr.s_addr && 895 inp->inp_lport == lport) { 896 /* 897 * Found. 898 */ 899 return (inp); 900 } 901 } 902 /* 903 * Not found. 904 */ 905 return (NULL); 906 } else { 907 struct inpcbporthead *porthash; 908 struct inpcbport *phd; 909 struct inpcb *match = NULL; 910 /* 911 * Best fit PCB lookup. 912 * 913 * First see if this local port is in use by looking on the 914 * port hash list. 915 */ 916 retrylookup: 917 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport, 918 pcbinfo->porthashmask)]; 919 LIST_FOREACH(phd, porthash, phd_hash) { 920 if (phd->phd_port == lport) 921 break; 922 } 923 if (phd != NULL) { 924 /* 925 * Port is in use by one or more PCBs. Look for best 926 * fit. 927 */ 928 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 929 wildcard = 0; 930 #ifdef INET6 931 if ((inp->inp_vflag & INP_IPV4) == 0) 932 continue; 933 #endif 934 /* 935 * Clean out old time_wait sockets if they 936 * are clogging up needed local ports. 937 */ 938 if ((inp->inp_vflag & INP_TIMEWAIT) != 0) { 939 if (tcp_twrecycleable((struct tcptw *)inp->inp_ppcb)) { 940 INP_LOCK(inp); 941 tcp_twclose((struct tcptw *)inp->inp_ppcb, 0); 942 match = NULL; 943 goto retrylookup; 944 } 945 } 946 if (inp->inp_faddr.s_addr != INADDR_ANY) 947 wildcard++; 948 if (inp->inp_laddr.s_addr != INADDR_ANY) { 949 if (laddr.s_addr == INADDR_ANY) 950 wildcard++; 951 else if (inp->inp_laddr.s_addr != laddr.s_addr) 952 continue; 953 } else { 954 if (laddr.s_addr != INADDR_ANY) 955 wildcard++; 956 } 957 if (wildcard < matchwild) { 958 match = inp; 959 matchwild = wildcard; 960 if (matchwild == 0) { 961 break; 962 } 963 } 964 } 965 } 966 return (match); 967 } 968 } 969 970 /* 971 * Lookup PCB in hash list. 972 */ 973 struct inpcb * 974 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, 975 ifp) 976 struct inpcbinfo *pcbinfo; 977 struct in_addr faddr, laddr; 978 u_int fport_arg, lport_arg; 979 int wildcard; 980 struct ifnet *ifp; 981 { 982 struct inpcbhead *head; 983 register struct inpcb *inp; 984 u_short fport = fport_arg, lport = lport_arg; 985 986 INP_INFO_RLOCK_ASSERT(pcbinfo); 987 /* 988 * First look for an exact match. 989 */ 990 head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)]; 991 LIST_FOREACH(inp, head, inp_hash) { 992 #ifdef INET6 993 if ((inp->inp_vflag & INP_IPV4) == 0) 994 continue; 995 #endif 996 if (inp->inp_faddr.s_addr == faddr.s_addr && 997 inp->inp_laddr.s_addr == laddr.s_addr && 998 inp->inp_fport == fport && 999 inp->inp_lport == lport) { 1000 /* 1001 * Found. 1002 */ 1003 return (inp); 1004 } 1005 } 1006 if (wildcard) { 1007 struct inpcb *local_wild = NULL; 1008 #if defined(INET6) 1009 struct inpcb *local_wild_mapped = NULL; 1010 #endif /* defined(INET6) */ 1011 1012 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 1013 LIST_FOREACH(inp, head, inp_hash) { 1014 #ifdef INET6 1015 if ((inp->inp_vflag & INP_IPV4) == 0) 1016 continue; 1017 #endif 1018 if (inp->inp_faddr.s_addr == INADDR_ANY && 1019 inp->inp_lport == lport) { 1020 if (ifp && ifp->if_type == IFT_FAITH && 1021 (inp->inp_flags & INP_FAITH) == 0) 1022 continue; 1023 if (inp->inp_laddr.s_addr == laddr.s_addr) 1024 return (inp); 1025 else if (inp->inp_laddr.s_addr == INADDR_ANY) { 1026 #if defined(INET6) 1027 if (INP_CHECK_SOCKAF(inp->inp_socket, 1028 AF_INET6)) 1029 local_wild_mapped = inp; 1030 else 1031 #endif /* defined(INET6) */ 1032 local_wild = inp; 1033 } 1034 } 1035 } 1036 #if defined(INET6) 1037 if (local_wild == NULL) 1038 return (local_wild_mapped); 1039 #endif /* defined(INET6) */ 1040 return (local_wild); 1041 } 1042 1043 /* 1044 * Not found. 1045 */ 1046 return (NULL); 1047 } 1048 1049 /* 1050 * Insert PCB onto various hash lists. 1051 */ 1052 int 1053 in_pcbinshash(inp) 1054 struct inpcb *inp; 1055 { 1056 struct inpcbhead *pcbhash; 1057 struct inpcbporthead *pcbporthash; 1058 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1059 struct inpcbport *phd; 1060 u_int32_t hashkey_faddr; 1061 1062 INP_INFO_WLOCK_ASSERT(pcbinfo); 1063 #ifdef INET6 1064 if (inp->inp_vflag & INP_IPV6) 1065 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1066 else 1067 #endif /* INET6 */ 1068 hashkey_faddr = inp->inp_faddr.s_addr; 1069 1070 pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1071 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1072 1073 pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport, 1074 pcbinfo->porthashmask)]; 1075 1076 /* 1077 * Go through port list and look for a head for this lport. 1078 */ 1079 LIST_FOREACH(phd, pcbporthash, phd_hash) { 1080 if (phd->phd_port == inp->inp_lport) 1081 break; 1082 } 1083 /* 1084 * If none exists, malloc one and tack it on. 1085 */ 1086 if (phd == NULL) { 1087 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT); 1088 if (phd == NULL) { 1089 return (ENOBUFS); /* XXX */ 1090 } 1091 phd->phd_port = inp->inp_lport; 1092 LIST_INIT(&phd->phd_pcblist); 1093 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 1094 } 1095 inp->inp_phd = phd; 1096 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 1097 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 1098 return (0); 1099 } 1100 1101 /* 1102 * Move PCB to the proper hash bucket when { faddr, fport } have been 1103 * changed. NOTE: This does not handle the case of the lport changing (the 1104 * hashed port list would have to be updated as well), so the lport must 1105 * not change after in_pcbinshash() has been called. 1106 */ 1107 void 1108 in_pcbrehash(inp) 1109 struct inpcb *inp; 1110 { 1111 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1112 struct inpcbhead *head; 1113 u_int32_t hashkey_faddr; 1114 1115 INP_INFO_WLOCK_ASSERT(pcbinfo); 1116 /* XXX? INP_LOCK_ASSERT(inp); */ 1117 #ifdef INET6 1118 if (inp->inp_vflag & INP_IPV6) 1119 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1120 else 1121 #endif /* INET6 */ 1122 hashkey_faddr = inp->inp_faddr.s_addr; 1123 1124 head = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1125 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1126 1127 LIST_REMOVE(inp, inp_hash); 1128 LIST_INSERT_HEAD(head, inp, inp_hash); 1129 } 1130 1131 /* 1132 * Remove PCB from various lists. 1133 */ 1134 void 1135 in_pcbremlists(inp) 1136 struct inpcb *inp; 1137 { 1138 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1139 1140 INP_INFO_WLOCK_ASSERT(pcbinfo); 1141 INP_LOCK_ASSERT(inp); 1142 1143 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 1144 if (inp->inp_lport) { 1145 struct inpcbport *phd = inp->inp_phd; 1146 1147 LIST_REMOVE(inp, inp_hash); 1148 LIST_REMOVE(inp, inp_portlist); 1149 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 1150 LIST_REMOVE(phd, phd_hash); 1151 free(phd, M_PCB); 1152 } 1153 } 1154 LIST_REMOVE(inp, inp_list); 1155 pcbinfo->ipi_count--; 1156 } 1157 1158 /* 1159 * A set label operation has occurred at the socket layer, propagate the 1160 * label change into the in_pcb for the socket. 1161 */ 1162 void 1163 in_pcbsosetlabel(so) 1164 struct socket *so; 1165 { 1166 #ifdef MAC 1167 struct inpcb *inp; 1168 1169 inp = (struct inpcb *)so->so_pcb; 1170 INP_LOCK(inp); 1171 SOCK_LOCK(so); 1172 mac_inpcb_sosetlabel(so, inp); 1173 SOCK_UNLOCK(so); 1174 INP_UNLOCK(inp); 1175 #endif 1176 } 1177