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 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 465 INP_LOCK_ASSERT(inp); 466 467 lport = inp->inp_lport; 468 laddr = inp->inp_laddr.s_addr; 469 anonport = (lport == 0); 470 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, 471 NULL, cred); 472 if (error) 473 return (error); 474 475 /* Do the initial binding of the local address if required. */ 476 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { 477 inp->inp_lport = lport; 478 inp->inp_laddr.s_addr = laddr; 479 if (in_pcbinshash(inp) != 0) { 480 inp->inp_laddr.s_addr = INADDR_ANY; 481 inp->inp_lport = 0; 482 return (EAGAIN); 483 } 484 } 485 486 /* Commit the remaining changes. */ 487 inp->inp_lport = lport; 488 inp->inp_laddr.s_addr = laddr; 489 inp->inp_faddr.s_addr = faddr; 490 inp->inp_fport = fport; 491 in_pcbrehash(inp); 492 #ifdef IPSEC 493 if (inp->inp_socket->so_type == SOCK_STREAM) 494 ipsec_pcbconn(inp->inp_sp); 495 #endif 496 if (anonport) 497 inp->inp_flags |= INP_ANONPORT; 498 return (0); 499 } 500 501 /* 502 * Set up for a connect from a socket to the specified address. 503 * On entry, *laddrp and *lportp should contain the current local 504 * address and port for the PCB; these are updated to the values 505 * that should be placed in inp_laddr and inp_lport to complete 506 * the connect. 507 * 508 * On success, *faddrp and *fportp will be set to the remote address 509 * and port. These are not updated in the error case. 510 * 511 * If the operation fails because the connection already exists, 512 * *oinpp will be set to the PCB of that connection so that the 513 * caller can decide to override it. In all other cases, *oinpp 514 * is set to NULL. 515 */ 516 int 517 in_pcbconnect_setup(inp, nam, laddrp, lportp, faddrp, fportp, oinpp, cred) 518 register struct inpcb *inp; 519 struct sockaddr *nam; 520 in_addr_t *laddrp; 521 u_short *lportp; 522 in_addr_t *faddrp; 523 u_short *fportp; 524 struct inpcb **oinpp; 525 struct ucred *cred; 526 { 527 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 528 struct in_ifaddr *ia; 529 struct sockaddr_in sa; 530 struct ucred *socred; 531 struct inpcb *oinp; 532 struct in_addr laddr, faddr; 533 u_short lport, fport; 534 int error; 535 536 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 537 INP_LOCK_ASSERT(inp); 538 539 if (oinpp != NULL) 540 *oinpp = NULL; 541 if (nam->sa_len != sizeof (*sin)) 542 return (EINVAL); 543 if (sin->sin_family != AF_INET) 544 return (EAFNOSUPPORT); 545 if (sin->sin_port == 0) 546 return (EADDRNOTAVAIL); 547 laddr.s_addr = *laddrp; 548 lport = *lportp; 549 faddr = sin->sin_addr; 550 fport = sin->sin_port; 551 socred = inp->inp_socket->so_cred; 552 if (laddr.s_addr == INADDR_ANY && jailed(socred)) { 553 bzero(&sa, sizeof(sa)); 554 sa.sin_addr.s_addr = htonl(prison_getip(socred)); 555 sa.sin_len = sizeof(sa); 556 sa.sin_family = AF_INET; 557 error = in_pcbbind_setup(inp, (struct sockaddr *)&sa, 558 &laddr.s_addr, &lport, cred); 559 if (error) 560 return (error); 561 } 562 if (!TAILQ_EMPTY(&in_ifaddrhead)) { 563 /* 564 * If the destination address is INADDR_ANY, 565 * use the primary local address. 566 * If the supplied address is INADDR_BROADCAST, 567 * and the primary interface supports broadcast, 568 * choose the broadcast address for that interface. 569 */ 570 if (faddr.s_addr == INADDR_ANY) 571 faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr; 572 else if (faddr.s_addr == (u_long)INADDR_BROADCAST && 573 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags & 574 IFF_BROADCAST)) 575 faddr = satosin(&TAILQ_FIRST( 576 &in_ifaddrhead)->ia_broadaddr)->sin_addr; 577 } 578 if (laddr.s_addr == INADDR_ANY) { 579 struct route sro; 580 581 bzero(&sro, sizeof(sro)); 582 ia = (struct in_ifaddr *)0; 583 /* 584 * If route is known our src addr is taken from the i/f, 585 * else punt. 586 */ 587 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) { 588 /* Find out route to destination */ 589 sro.ro_dst.sa_family = AF_INET; 590 sro.ro_dst.sa_len = sizeof(struct sockaddr_in); 591 ((struct sockaddr_in *)&sro.ro_dst)->sin_addr = faddr; 592 rtalloc_ign(&sro, RTF_CLONING); 593 } 594 /* 595 * If we found a route, use the address 596 * corresponding to the outgoing interface 597 * unless it is the loopback (in case a route 598 * to our address on another net goes to loopback). 599 */ 600 if (sro.ro_rt && !(sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) 601 ia = ifatoia(sro.ro_rt->rt_ifa); 602 if (sro.ro_rt) 603 RTFREE(sro.ro_rt); 604 if (ia == 0) { 605 bzero(&sa, sizeof(sa)); 606 sa.sin_addr = faddr; 607 sa.sin_len = sizeof(sa); 608 sa.sin_family = AF_INET; 609 610 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa))); 611 if (ia == 0) 612 ia = ifatoia(ifa_ifwithnet(sintosa(&sa))); 613 if (ia == 0) 614 return (ENETUNREACH); 615 } 616 /* 617 * If the destination address is multicast and an outgoing 618 * interface has been set as a multicast option, use the 619 * address of that interface as our source address. 620 */ 621 if (IN_MULTICAST(ntohl(faddr.s_addr)) && 622 inp->inp_moptions != NULL) { 623 struct ip_moptions *imo; 624 struct ifnet *ifp; 625 626 imo = inp->inp_moptions; 627 if (imo->imo_multicast_ifp != NULL) { 628 ifp = imo->imo_multicast_ifp; 629 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) 630 if (ia->ia_ifp == ifp) 631 break; 632 if (ia == 0) 633 return (EADDRNOTAVAIL); 634 } 635 } 636 laddr = ia->ia_addr.sin_addr; 637 } 638 639 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport, 640 0, NULL); 641 if (oinp != NULL) { 642 if (oinpp != NULL) 643 *oinpp = oinp; 644 return (EADDRINUSE); 645 } 646 if (lport == 0) { 647 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, 648 cred); 649 if (error) 650 return (error); 651 } 652 *laddrp = laddr.s_addr; 653 *lportp = lport; 654 *faddrp = faddr.s_addr; 655 *fportp = fport; 656 return (0); 657 } 658 659 void 660 in_pcbdisconnect(inp) 661 struct inpcb *inp; 662 { 663 INP_LOCK_ASSERT(inp); 664 665 inp->inp_faddr.s_addr = INADDR_ANY; 666 inp->inp_fport = 0; 667 in_pcbrehash(inp); 668 #ifdef IPSEC 669 ipsec_pcbdisconn(inp->inp_sp); 670 #endif 671 if (inp->inp_socket->so_state & SS_NOFDREF) 672 in_pcbdetach(inp); 673 } 674 675 void 676 in_pcbdetach(inp) 677 struct inpcb *inp; 678 { 679 struct socket *so = inp->inp_socket; 680 struct inpcbinfo *ipi = inp->inp_pcbinfo; 681 682 INP_LOCK_ASSERT(inp); 683 684 #if defined(IPSEC) || defined(FAST_IPSEC) 685 ipsec4_delete_pcbpolicy(inp); 686 #endif /*IPSEC*/ 687 inp->inp_gencnt = ++ipi->ipi_gencnt; 688 in_pcbremlists(inp); 689 if (so) { 690 SOCK_LOCK(so); 691 so->so_pcb = 0; 692 sotryfree(so); 693 } 694 if (inp->inp_options) 695 (void)m_free(inp->inp_options); 696 ip_freemoptions(inp->inp_moptions); 697 inp->inp_vflag = 0; 698 INP_LOCK_DESTROY(inp); 699 #ifdef MAC 700 mac_destroy_inpcb(inp); 701 #endif 702 uma_zfree(ipi->ipi_zone, inp); 703 } 704 705 struct sockaddr * 706 in_sockaddr(port, addr_p) 707 in_port_t port; 708 struct in_addr *addr_p; 709 { 710 struct sockaddr_in *sin; 711 712 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, 713 M_WAITOK | M_ZERO); 714 sin->sin_family = AF_INET; 715 sin->sin_len = sizeof(*sin); 716 sin->sin_addr = *addr_p; 717 sin->sin_port = port; 718 719 return (struct sockaddr *)sin; 720 } 721 722 /* 723 * The wrapper function will pass down the pcbinfo for this function to lock. 724 * The socket must have a valid 725 * (i.e., non-nil) PCB, but it should be impossible to get an invalid one 726 * except through a kernel programming error, so it is acceptable to panic 727 * (or in this case trap) if the PCB is invalid. (Actually, we don't trap 728 * because there actually /is/ a programming error somewhere... XXX) 729 */ 730 int 731 in_setsockaddr(so, nam, pcbinfo) 732 struct socket *so; 733 struct sockaddr **nam; 734 struct inpcbinfo *pcbinfo; 735 { 736 int s; 737 register struct inpcb *inp; 738 struct in_addr addr; 739 in_port_t port; 740 741 s = splnet(); 742 INP_INFO_RLOCK(pcbinfo); 743 inp = sotoinpcb(so); 744 if (!inp) { 745 INP_INFO_RUNLOCK(pcbinfo); 746 splx(s); 747 return ECONNRESET; 748 } 749 INP_LOCK(inp); 750 port = inp->inp_lport; 751 addr = inp->inp_laddr; 752 INP_UNLOCK(inp); 753 INP_INFO_RUNLOCK(pcbinfo); 754 splx(s); 755 756 *nam = in_sockaddr(port, &addr); 757 return 0; 758 } 759 760 /* 761 * The wrapper function will pass down the pcbinfo for this function to lock. 762 */ 763 int 764 in_setpeeraddr(so, nam, pcbinfo) 765 struct socket *so; 766 struct sockaddr **nam; 767 struct inpcbinfo *pcbinfo; 768 { 769 int s; 770 register struct inpcb *inp; 771 struct in_addr addr; 772 in_port_t port; 773 774 s = splnet(); 775 INP_INFO_RLOCK(pcbinfo); 776 inp = sotoinpcb(so); 777 if (!inp) { 778 INP_INFO_RUNLOCK(pcbinfo); 779 splx(s); 780 return ECONNRESET; 781 } 782 INP_LOCK(inp); 783 port = inp->inp_fport; 784 addr = inp->inp_faddr; 785 INP_UNLOCK(inp); 786 INP_INFO_RUNLOCK(pcbinfo); 787 splx(s); 788 789 *nam = in_sockaddr(port, &addr); 790 return 0; 791 } 792 793 void 794 in_pcbnotifyall(pcbinfo, faddr, errno, notify) 795 struct inpcbinfo *pcbinfo; 796 struct in_addr faddr; 797 int errno; 798 struct inpcb *(*notify)(struct inpcb *, int); 799 { 800 struct inpcb *inp, *ninp; 801 struct inpcbhead *head; 802 int s; 803 804 s = splnet(); 805 INP_INFO_WLOCK(pcbinfo); 806 head = pcbinfo->listhead; 807 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) { 808 INP_LOCK(inp); 809 ninp = LIST_NEXT(inp, inp_list); 810 #ifdef INET6 811 if ((inp->inp_vflag & INP_IPV4) == 0) { 812 INP_UNLOCK(inp); 813 continue; 814 } 815 #endif 816 if (inp->inp_faddr.s_addr != faddr.s_addr || 817 inp->inp_socket == NULL) { 818 INP_UNLOCK(inp); 819 continue; 820 } 821 if ((*notify)(inp, errno)) 822 INP_UNLOCK(inp); 823 } 824 INP_INFO_WUNLOCK(pcbinfo); 825 splx(s); 826 } 827 828 void 829 in_pcbpurgeif0(pcbinfo, ifp) 830 struct inpcbinfo *pcbinfo; 831 struct ifnet *ifp; 832 { 833 struct inpcb *inp; 834 struct ip_moptions *imo; 835 int i, gap; 836 837 /* why no splnet here? XXX */ 838 INP_INFO_RLOCK(pcbinfo); 839 LIST_FOREACH(inp, pcbinfo->listhead, inp_list) { 840 INP_LOCK(inp); 841 imo = inp->inp_moptions; 842 if ((inp->inp_vflag & INP_IPV4) && 843 imo != NULL) { 844 /* 845 * Unselect the outgoing interface if it is being 846 * detached. 847 */ 848 if (imo->imo_multicast_ifp == ifp) 849 imo->imo_multicast_ifp = NULL; 850 851 /* 852 * Drop multicast group membership if we joined 853 * through the interface being detached. 854 */ 855 for (i = 0, gap = 0; i < imo->imo_num_memberships; 856 i++) { 857 if (imo->imo_membership[i]->inm_ifp == ifp) { 858 in_delmulti(imo->imo_membership[i]); 859 gap++; 860 } else if (gap != 0) 861 imo->imo_membership[i - gap] = 862 imo->imo_membership[i]; 863 } 864 imo->imo_num_memberships -= gap; 865 } 866 INP_UNLOCK(inp); 867 } 868 INP_INFO_RUNLOCK(pcbinfo); 869 } 870 871 /* 872 * Lookup a PCB based on the local address and port. 873 */ 874 struct inpcb * 875 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay) 876 struct inpcbinfo *pcbinfo; 877 struct in_addr laddr; 878 u_int lport_arg; 879 int wild_okay; 880 { 881 register struct inpcb *inp; 882 int matchwild = 3, wildcard; 883 u_short lport = lport_arg; 884 885 INP_INFO_WLOCK_ASSERT(pcbinfo); 886 887 if (!wild_okay) { 888 struct inpcbhead *head; 889 /* 890 * Look for an unconnected (wildcard foreign addr) PCB that 891 * matches the local address and port we're looking for. 892 */ 893 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 894 LIST_FOREACH(inp, head, inp_hash) { 895 #ifdef INET6 896 if ((inp->inp_vflag & INP_IPV4) == 0) 897 continue; 898 #endif 899 if (inp->inp_faddr.s_addr == INADDR_ANY && 900 inp->inp_laddr.s_addr == laddr.s_addr && 901 inp->inp_lport == lport) { 902 /* 903 * Found. 904 */ 905 return (inp); 906 } 907 } 908 /* 909 * Not found. 910 */ 911 return (NULL); 912 } else { 913 struct inpcbporthead *porthash; 914 struct inpcbport *phd; 915 struct inpcb *match = NULL; 916 /* 917 * Best fit PCB lookup. 918 * 919 * First see if this local port is in use by looking on the 920 * port hash list. 921 */ 922 retrylookup: 923 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport, 924 pcbinfo->porthashmask)]; 925 LIST_FOREACH(phd, porthash, phd_hash) { 926 if (phd->phd_port == lport) 927 break; 928 } 929 if (phd != NULL) { 930 /* 931 * Port is in use by one or more PCBs. Look for best 932 * fit. 933 */ 934 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 935 wildcard = 0; 936 #ifdef INET6 937 if ((inp->inp_vflag & INP_IPV4) == 0) 938 continue; 939 #endif 940 /* 941 * Clean out old time_wait sockets if they 942 * are clogging up needed local ports. 943 */ 944 if ((inp->inp_vflag & INP_TIMEWAIT) != 0) { 945 if (tcp_twrecycleable((struct tcptw *)inp->inp_ppcb)) { 946 INP_LOCK(inp); 947 tcp_twclose((struct tcptw *)inp->inp_ppcb, 0); 948 match = NULL; 949 goto retrylookup; 950 } 951 } 952 if (inp->inp_faddr.s_addr != INADDR_ANY) 953 wildcard++; 954 if (inp->inp_laddr.s_addr != INADDR_ANY) { 955 if (laddr.s_addr == INADDR_ANY) 956 wildcard++; 957 else if (inp->inp_laddr.s_addr != laddr.s_addr) 958 continue; 959 } else { 960 if (laddr.s_addr != INADDR_ANY) 961 wildcard++; 962 } 963 if (wildcard < matchwild) { 964 match = inp; 965 matchwild = wildcard; 966 if (matchwild == 0) { 967 break; 968 } 969 } 970 } 971 } 972 return (match); 973 } 974 } 975 976 /* 977 * Lookup PCB in hash list. 978 */ 979 struct inpcb * 980 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, 981 ifp) 982 struct inpcbinfo *pcbinfo; 983 struct in_addr faddr, laddr; 984 u_int fport_arg, lport_arg; 985 int wildcard; 986 struct ifnet *ifp; 987 { 988 struct inpcbhead *head; 989 register struct inpcb *inp; 990 u_short fport = fport_arg, lport = lport_arg; 991 992 INP_INFO_RLOCK_ASSERT(pcbinfo); 993 /* 994 * First look for an exact match. 995 */ 996 head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)]; 997 LIST_FOREACH(inp, head, inp_hash) { 998 #ifdef INET6 999 if ((inp->inp_vflag & INP_IPV4) == 0) 1000 continue; 1001 #endif 1002 if (inp->inp_faddr.s_addr == faddr.s_addr && 1003 inp->inp_laddr.s_addr == laddr.s_addr && 1004 inp->inp_fport == fport && 1005 inp->inp_lport == lport) { 1006 /* 1007 * Found. 1008 */ 1009 return (inp); 1010 } 1011 } 1012 if (wildcard) { 1013 struct inpcb *local_wild = NULL; 1014 #if defined(INET6) 1015 struct inpcb *local_wild_mapped = NULL; 1016 #endif /* defined(INET6) */ 1017 1018 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 1019 LIST_FOREACH(inp, head, inp_hash) { 1020 #ifdef INET6 1021 if ((inp->inp_vflag & INP_IPV4) == 0) 1022 continue; 1023 #endif 1024 if (inp->inp_faddr.s_addr == INADDR_ANY && 1025 inp->inp_lport == lport) { 1026 if (ifp && ifp->if_type == IFT_FAITH && 1027 (inp->inp_flags & INP_FAITH) == 0) 1028 continue; 1029 if (inp->inp_laddr.s_addr == laddr.s_addr) 1030 return (inp); 1031 else if (inp->inp_laddr.s_addr == INADDR_ANY) { 1032 #if defined(INET6) 1033 if (INP_CHECK_SOCKAF(inp->inp_socket, 1034 AF_INET6)) 1035 local_wild_mapped = inp; 1036 else 1037 #endif /* defined(INET6) */ 1038 local_wild = inp; 1039 } 1040 } 1041 } 1042 #if defined(INET6) 1043 if (local_wild == NULL) 1044 return (local_wild_mapped); 1045 #endif /* defined(INET6) */ 1046 return (local_wild); 1047 } 1048 1049 /* 1050 * Not found. 1051 */ 1052 return (NULL); 1053 } 1054 1055 /* 1056 * Insert PCB onto various hash lists. 1057 */ 1058 int 1059 in_pcbinshash(inp) 1060 struct inpcb *inp; 1061 { 1062 struct inpcbhead *pcbhash; 1063 struct inpcbporthead *pcbporthash; 1064 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1065 struct inpcbport *phd; 1066 u_int32_t hashkey_faddr; 1067 1068 INP_INFO_WLOCK_ASSERT(pcbinfo); 1069 #ifdef INET6 1070 if (inp->inp_vflag & INP_IPV6) 1071 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1072 else 1073 #endif /* INET6 */ 1074 hashkey_faddr = inp->inp_faddr.s_addr; 1075 1076 pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1077 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1078 1079 pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport, 1080 pcbinfo->porthashmask)]; 1081 1082 /* 1083 * Go through port list and look for a head for this lport. 1084 */ 1085 LIST_FOREACH(phd, pcbporthash, phd_hash) { 1086 if (phd->phd_port == inp->inp_lport) 1087 break; 1088 } 1089 /* 1090 * If none exists, malloc one and tack it on. 1091 */ 1092 if (phd == NULL) { 1093 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT); 1094 if (phd == NULL) { 1095 return (ENOBUFS); /* XXX */ 1096 } 1097 phd->phd_port = inp->inp_lport; 1098 LIST_INIT(&phd->phd_pcblist); 1099 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 1100 } 1101 inp->inp_phd = phd; 1102 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 1103 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 1104 return (0); 1105 } 1106 1107 /* 1108 * Move PCB to the proper hash bucket when { faddr, fport } have been 1109 * changed. NOTE: This does not handle the case of the lport changing (the 1110 * hashed port list would have to be updated as well), so the lport must 1111 * not change after in_pcbinshash() has been called. 1112 */ 1113 void 1114 in_pcbrehash(inp) 1115 struct inpcb *inp; 1116 { 1117 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1118 struct inpcbhead *head; 1119 u_int32_t hashkey_faddr; 1120 1121 INP_INFO_WLOCK_ASSERT(pcbinfo); 1122 INP_LOCK_ASSERT(inp); 1123 #ifdef INET6 1124 if (inp->inp_vflag & INP_IPV6) 1125 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1126 else 1127 #endif /* INET6 */ 1128 hashkey_faddr = inp->inp_faddr.s_addr; 1129 1130 head = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1131 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1132 1133 LIST_REMOVE(inp, inp_hash); 1134 LIST_INSERT_HEAD(head, inp, inp_hash); 1135 } 1136 1137 /* 1138 * Remove PCB from various lists. 1139 */ 1140 void 1141 in_pcbremlists(inp) 1142 struct inpcb *inp; 1143 { 1144 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1145 1146 INP_INFO_WLOCK_ASSERT(pcbinfo); 1147 INP_LOCK_ASSERT(inp); 1148 1149 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 1150 if (inp->inp_lport) { 1151 struct inpcbport *phd = inp->inp_phd; 1152 1153 LIST_REMOVE(inp, inp_hash); 1154 LIST_REMOVE(inp, inp_portlist); 1155 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 1156 LIST_REMOVE(phd, phd_hash); 1157 free(phd, M_PCB); 1158 } 1159 } 1160 LIST_REMOVE(inp, inp_list); 1161 pcbinfo->ipi_count--; 1162 } 1163 1164 /* 1165 * A set label operation has occurred at the socket layer, propagate the 1166 * label change into the in_pcb for the socket. 1167 */ 1168 void 1169 in_pcbsosetlabel(so) 1170 struct socket *so; 1171 { 1172 #ifdef MAC 1173 struct inpcb *inp; 1174 1175 inp = (struct inpcb *)so->so_pcb; 1176 INP_LOCK(inp); 1177 SOCK_LOCK(so); 1178 mac_inpcb_sosetlabel(so, inp); 1179 SOCK_UNLOCK(so); 1180 INP_UNLOCK(inp); 1181 #endif 1182 } 1183