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 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University 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 REGENTS 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 REGENTS 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 * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 34 * $FreeBSD$ 35 */ 36 37 #include "opt_ipsec.h" 38 #include "opt_inet6.h" 39 #include "opt_mac.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/limits.h> 44 #include <sys/mac.h> 45 #include <sys/malloc.h> 46 #include <sys/mbuf.h> 47 #include <sys/domain.h> 48 #include <sys/protosw.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/proc.h> 52 #include <sys/jail.h> 53 #include <sys/kernel.h> 54 #include <sys/sysctl.h> 55 56 #include <vm/uma.h> 57 58 #include <net/if.h> 59 #include <net/if_types.h> 60 #include <net/route.h> 61 62 #include <netinet/in.h> 63 #include <netinet/in_pcb.h> 64 #include <netinet/in_var.h> 65 #include <netinet/ip_var.h> 66 #include <netinet/tcp_var.h> 67 #ifdef INET6 68 #include <netinet/ip6.h> 69 #include <netinet6/ip6_var.h> 70 #endif /* INET6 */ 71 72 #ifdef IPSEC 73 #include <netinet6/ipsec.h> 74 #include <netkey/key.h> 75 #endif /* IPSEC */ 76 77 #ifdef FAST_IPSEC 78 #if defined(IPSEC) || defined(IPSEC_ESP) 79 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!" 80 #endif 81 82 #include <netipsec/ipsec.h> 83 #include <netipsec/key.h> 84 #endif /* FAST_IPSEC */ 85 86 struct in_addr zeroin_addr; 87 88 /* 89 * These configure the range of local port addresses assigned to 90 * "unspecified" outgoing connections/packets/whatever. 91 */ 92 int ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 93 int ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 94 int ipport_firstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 95 int ipport_lastauto = IPPORT_HILASTAUTO; /* 65535 */ 96 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 97 int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 98 99 /* 100 * Reserved ports accessible only to root. There are significant 101 * security considerations that must be accounted for when changing these, 102 * but the security benefits can be great. Please be careful. 103 */ 104 int ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 105 int ipport_reservedlow = 0; 106 107 #define RANGECHK(var, min, max) \ 108 if ((var) < (min)) { (var) = (min); } \ 109 else if ((var) > (max)) { (var) = (max); } 110 111 static int 112 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) 113 { 114 int error = sysctl_handle_int(oidp, 115 oidp->oid_arg1, oidp->oid_arg2, req); 116 if (!error) { 117 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); 118 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1); 119 RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX); 120 RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX); 121 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX); 122 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX); 123 } 124 return error; 125 } 126 127 #undef RANGECHK 128 129 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); 130 131 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW, 132 &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", ""); 133 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW, 134 &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", ""); 135 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW, 136 &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", ""); 137 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW, 138 &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", ""); 139 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW, 140 &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", ""); 141 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW, 142 &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", ""); 143 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, 144 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, ""); 145 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, 146 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, ""); 147 148 /* 149 * in_pcb.c: manage the Protocol Control Blocks. 150 * 151 * NOTE: It is assumed that most of these functions will be called at 152 * splnet(). XXX - There are, unfortunately, a few exceptions to this 153 * rule that should be fixed. 154 */ 155 156 /* 157 * Allocate a PCB and associate it with the socket. 158 */ 159 int 160 in_pcballoc(so, pcbinfo, td, type) 161 struct socket *so; 162 struct inpcbinfo *pcbinfo; 163 struct thread *td; 164 const char *type; 165 { 166 register struct inpcb *inp; 167 int error; 168 169 INP_INFO_WLOCK_ASSERT(pcbinfo); 170 error = 0; 171 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT | M_ZERO); 172 if (inp == NULL) 173 return (ENOBUFS); 174 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 175 inp->inp_pcbinfo = pcbinfo; 176 inp->inp_socket = so; 177 #ifdef MAC 178 error = mac_init_inpcb(inp, M_NOWAIT); 179 if (error != 0) 180 goto out; 181 mac_create_inpcb_from_socket(so, inp); 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, td) 217 register struct inpcb *inp; 218 struct sockaddr *nam; 219 struct thread *td; 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, td); 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, td) 255 struct inpcb *inp; 256 struct sockaddr *nam; 257 in_addr_t *laddrp; 258 u_short *lportp; 259 struct thread *td; 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(td->td_ucred, 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 td && suser_cred(td->td_ucred, PRISON_ROOT)) 325 return (EACCES); 326 if (td && jailed(td->td_ucred)) 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 && (t->inp_vflag & INP_TIMEWAIT)) { 338 if ((ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 339 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 340 (intotw(t)->tw_so_options & SO_REUSEPORT) == 0) && 341 (so->so_cred->cr_uid != intotw(t)->tw_cred->cr_uid)) 342 return (EADDRINUSE); 343 } else 344 if (t && 345 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 346 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 347 (t->inp_socket->so_options & 348 SO_REUSEPORT) == 0) && 349 (so->so_cred->cr_uid != 350 t->inp_socket->so_cred->cr_uid)) { 351 #if defined(INET6) 352 if (ntohl(sin->sin_addr.s_addr) != 353 INADDR_ANY || 354 ntohl(t->inp_laddr.s_addr) != 355 INADDR_ANY || 356 INP_SOCKAF(so) == 357 INP_SOCKAF(t->inp_socket)) 358 #endif /* defined(INET6) */ 359 return (EADDRINUSE); 360 } 361 } 362 if (prison && 363 prison_ip(td->td_ucred, 0, &sin->sin_addr.s_addr)) 364 return (EADDRNOTAVAIL); 365 t = in_pcblookup_local(pcbinfo, sin->sin_addr, 366 lport, prison ? 0 : wild); 367 if (t && (t->inp_vflag & INP_TIMEWAIT)) { 368 if ((reuseport & intotw(t)->tw_so_options) == 0) 369 return (EADDRINUSE); 370 } else 371 if (t && 372 (reuseport & t->inp_socket->so_options) == 0) { 373 #if defined(INET6) 374 if (ntohl(sin->sin_addr.s_addr) != 375 INADDR_ANY || 376 ntohl(t->inp_laddr.s_addr) != 377 INADDR_ANY || 378 INP_SOCKAF(so) == 379 INP_SOCKAF(t->inp_socket)) 380 #endif /* defined(INET6) */ 381 return (EADDRINUSE); 382 } 383 } 384 } 385 if (*lportp != 0) 386 lport = *lportp; 387 if (lport == 0) { 388 u_short first, last; 389 int count; 390 391 if (laddr.s_addr != INADDR_ANY) 392 if (prison_ip(td->td_ucred, 0, &laddr.s_addr)) 393 return (EINVAL); 394 395 if (inp->inp_flags & INP_HIGHPORT) { 396 first = ipport_hifirstauto; /* sysctl */ 397 last = ipport_hilastauto; 398 lastport = &pcbinfo->lasthi; 399 } else if (inp->inp_flags & INP_LOWPORT) { 400 if (td && (error = suser_cred(td->td_ucred, 401 PRISON_ROOT)) != 0) 402 return error; 403 first = ipport_lowfirstauto; /* 1023 */ 404 last = ipport_lowlastauto; /* 600 */ 405 lastport = &pcbinfo->lastlow; 406 } else { 407 first = ipport_firstauto; /* sysctl */ 408 last = ipport_lastauto; 409 lastport = &pcbinfo->lastport; 410 } 411 /* 412 * Simple check to ensure all ports are not used up causing 413 * a deadlock here. 414 * 415 * We split the two cases (up and down) so that the direction 416 * is not being tested on each round of the loop. 417 */ 418 if (first > last) { 419 /* 420 * counting down 421 */ 422 count = first - last; 423 424 do { 425 if (count-- < 0) /* completely used? */ 426 return (EADDRNOTAVAIL); 427 --*lastport; 428 if (*lastport > first || *lastport < last) 429 *lastport = first; 430 lport = htons(*lastport); 431 } while (in_pcblookup_local(pcbinfo, laddr, lport, 432 wild)); 433 } else { 434 /* 435 * counting up 436 */ 437 count = last - first; 438 439 do { 440 if (count-- < 0) /* completely used? */ 441 return (EADDRNOTAVAIL); 442 ++*lastport; 443 if (*lastport < first || *lastport > last) 444 *lastport = first; 445 lport = htons(*lastport); 446 } while (in_pcblookup_local(pcbinfo, laddr, lport, 447 wild)); 448 } 449 } 450 if (prison_ip(td->td_ucred, 0, &laddr.s_addr)) 451 return (EINVAL); 452 *laddrp = laddr.s_addr; 453 *lportp = lport; 454 return (0); 455 } 456 457 /* 458 * Connect from a socket to a specified address. 459 * Both address and port must be specified in argument sin. 460 * If don't have a local address for this socket yet, 461 * then pick one. 462 */ 463 int 464 in_pcbconnect(inp, nam, td) 465 register struct inpcb *inp; 466 struct sockaddr *nam; 467 struct thread *td; 468 { 469 u_short lport, fport; 470 in_addr_t laddr, faddr; 471 int anonport, error; 472 473 lport = inp->inp_lport; 474 laddr = inp->inp_laddr.s_addr; 475 anonport = (lport == 0); 476 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, 477 NULL, td); 478 if (error) 479 return (error); 480 481 /* Do the initial binding of the local address if required. */ 482 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { 483 inp->inp_lport = lport; 484 inp->inp_laddr.s_addr = laddr; 485 if (in_pcbinshash(inp) != 0) { 486 inp->inp_laddr.s_addr = INADDR_ANY; 487 inp->inp_lport = 0; 488 return (EAGAIN); 489 } 490 } 491 492 /* Commit the remaining changes. */ 493 inp->inp_lport = lport; 494 inp->inp_laddr.s_addr = laddr; 495 inp->inp_faddr.s_addr = faddr; 496 inp->inp_fport = fport; 497 in_pcbrehash(inp); 498 #ifdef IPSEC 499 if (inp->inp_socket->so_type == SOCK_STREAM) 500 ipsec_pcbconn(inp->inp_sp); 501 #endif 502 if (anonport) 503 inp->inp_flags |= INP_ANONPORT; 504 return (0); 505 } 506 507 /* 508 * Set up for a connect from a socket to the specified address. 509 * On entry, *laddrp and *lportp should contain the current local 510 * address and port for the PCB; these are updated to the values 511 * that should be placed in inp_laddr and inp_lport to complete 512 * the connect. 513 * 514 * On success, *faddrp and *fportp will be set to the remote address 515 * and port. These are not updated in the error case. 516 * 517 * If the operation fails because the connection already exists, 518 * *oinpp will be set to the PCB of that connection so that the 519 * caller can decide to override it. In all other cases, *oinpp 520 * is set to NULL. 521 */ 522 int 523 in_pcbconnect_setup(inp, nam, laddrp, lportp, faddrp, fportp, oinpp, td) 524 register struct inpcb *inp; 525 struct sockaddr *nam; 526 in_addr_t *laddrp; 527 u_short *lportp; 528 in_addr_t *faddrp; 529 u_short *fportp; 530 struct inpcb **oinpp; 531 struct thread *td; 532 { 533 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 534 struct in_ifaddr *ia; 535 struct sockaddr_in sa; 536 struct ucred *cred; 537 struct inpcb *oinp; 538 struct in_addr laddr, faddr; 539 u_short lport, fport; 540 int error; 541 542 if (oinpp != NULL) 543 *oinpp = NULL; 544 if (nam->sa_len != sizeof (*sin)) 545 return (EINVAL); 546 if (sin->sin_family != AF_INET) 547 return (EAFNOSUPPORT); 548 if (sin->sin_port == 0) 549 return (EADDRNOTAVAIL); 550 laddr.s_addr = *laddrp; 551 lport = *lportp; 552 faddr = sin->sin_addr; 553 fport = sin->sin_port; 554 cred = inp->inp_socket->so_cred; 555 if (laddr.s_addr == INADDR_ANY && jailed(cred)) { 556 bzero(&sa, sizeof(sa)); 557 sa.sin_addr.s_addr = htonl(prison_getip(cred)); 558 sa.sin_len = sizeof(sa); 559 sa.sin_family = AF_INET; 560 error = in_pcbbind_setup(inp, (struct sockaddr *)&sa, 561 &laddr.s_addr, &lport, td); 562 if (error) 563 return (error); 564 } 565 if (!TAILQ_EMPTY(&in_ifaddrhead)) { 566 /* 567 * If the destination address is INADDR_ANY, 568 * use the primary local address. 569 * If the supplied address is INADDR_BROADCAST, 570 * and the primary interface supports broadcast, 571 * choose the broadcast address for that interface. 572 */ 573 if (faddr.s_addr == INADDR_ANY) 574 faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr; 575 else if (faddr.s_addr == (u_long)INADDR_BROADCAST && 576 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags & 577 IFF_BROADCAST)) 578 faddr = satosin(&TAILQ_FIRST( 579 &in_ifaddrhead)->ia_broadaddr)->sin_addr; 580 } 581 if (laddr.s_addr == INADDR_ANY) { 582 struct route sro; 583 584 bzero(&sro, sizeof(sro)); 585 ia = (struct in_ifaddr *)0; 586 /* 587 * If route is known our src addr is taken from the i/f, 588 * else punt. 589 */ 590 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) { 591 /* Find out route to destination */ 592 sro.ro_dst.sa_family = AF_INET; 593 sro.ro_dst.sa_len = sizeof(struct sockaddr_in); 594 ((struct sockaddr_in *)&sro.ro_dst)->sin_addr = faddr; 595 rtalloc_ign(&sro, RTF_CLONING); 596 } 597 /* 598 * If we found a route, use the address 599 * corresponding to the outgoing interface 600 * unless it is the loopback (in case a route 601 * to our address on another net goes to loopback). 602 */ 603 if (sro.ro_rt && !(sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) 604 ia = ifatoia(sro.ro_rt->rt_ifa); 605 if (sro.ro_rt) 606 RTFREE(sro.ro_rt); 607 if (ia == 0) { 608 bzero(&sa, sizeof(sa)); 609 sa.sin_addr = faddr; 610 sa.sin_len = sizeof(sa); 611 sa.sin_family = AF_INET; 612 613 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa))); 614 if (ia == 0) 615 ia = ifatoia(ifa_ifwithnet(sintosa(&sa))); 616 if (ia == 0) 617 ia = TAILQ_FIRST(&in_ifaddrhead); 618 if (ia == 0) 619 return (EADDRNOTAVAIL); 620 } 621 /* 622 * If the destination address is multicast and an outgoing 623 * interface has been set as a multicast option, use the 624 * address of that interface as our source address. 625 */ 626 if (IN_MULTICAST(ntohl(faddr.s_addr)) && 627 inp->inp_moptions != NULL) { 628 struct ip_moptions *imo; 629 struct ifnet *ifp; 630 631 imo = inp->inp_moptions; 632 if (imo->imo_multicast_ifp != NULL) { 633 ifp = imo->imo_multicast_ifp; 634 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) 635 if (ia->ia_ifp == ifp) 636 break; 637 if (ia == 0) 638 return (EADDRNOTAVAIL); 639 } 640 } 641 laddr = ia->ia_addr.sin_addr; 642 } 643 644 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport, 645 0, NULL); 646 if (oinp != NULL) { 647 if (oinpp != NULL) 648 *oinpp = oinp; 649 return (EADDRINUSE); 650 } 651 if (lport == 0) { 652 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, td); 653 if (error) 654 return (error); 655 } 656 *laddrp = laddr.s_addr; 657 *lportp = lport; 658 *faddrp = faddr.s_addr; 659 *fportp = fport; 660 return (0); 661 } 662 663 void 664 in_pcbdisconnect(inp) 665 struct inpcb *inp; 666 { 667 INP_LOCK_ASSERT(inp); 668 669 inp->inp_faddr.s_addr = INADDR_ANY; 670 inp->inp_fport = 0; 671 in_pcbrehash(inp); 672 if (inp->inp_socket->so_state & SS_NOFDREF) 673 in_pcbdetach(inp); 674 #ifdef IPSEC 675 ipsec_pcbdisconn(inp->inp_sp); 676 #endif 677 } 678 679 void 680 in_pcbdetach(inp) 681 struct inpcb *inp; 682 { 683 struct socket *so = inp->inp_socket; 684 struct inpcbinfo *ipi = inp->inp_pcbinfo; 685 686 INP_LOCK_ASSERT(inp); 687 688 #if defined(IPSEC) || defined(FAST_IPSEC) 689 ipsec4_delete_pcbpolicy(inp); 690 #endif /*IPSEC*/ 691 inp->inp_gencnt = ++ipi->ipi_gencnt; 692 in_pcbremlists(inp); 693 if (so) { 694 so->so_pcb = 0; 695 sotryfree(so); 696 } 697 if (inp->inp_options) 698 (void)m_free(inp->inp_options); 699 ip_freemoptions(inp->inp_moptions); 700 inp->inp_vflag = 0; 701 INP_LOCK_DESTROY(inp); 702 #ifdef MAC 703 mac_destroy_inpcb(inp); 704 #endif 705 uma_zfree(ipi->ipi_zone, inp); 706 } 707 708 struct sockaddr * 709 in_sockaddr(port, addr_p) 710 in_port_t port; 711 struct in_addr *addr_p; 712 { 713 struct sockaddr_in *sin; 714 715 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, 716 M_WAITOK | M_ZERO); 717 sin->sin_family = AF_INET; 718 sin->sin_len = sizeof(*sin); 719 sin->sin_addr = *addr_p; 720 sin->sin_port = port; 721 722 return (struct sockaddr *)sin; 723 } 724 725 /* 726 * The wrapper function will pass down the pcbinfo for this function to lock. 727 * The socket must have a valid 728 * (i.e., non-nil) PCB, but it should be impossible to get an invalid one 729 * except through a kernel programming error, so it is acceptable to panic 730 * (or in this case trap) if the PCB is invalid. (Actually, we don't trap 731 * because there actually /is/ a programming error somewhere... XXX) 732 */ 733 int 734 in_setsockaddr(so, nam, pcbinfo) 735 struct socket *so; 736 struct sockaddr **nam; 737 struct inpcbinfo *pcbinfo; 738 { 739 int s; 740 register struct inpcb *inp; 741 struct in_addr addr; 742 in_port_t port; 743 744 s = splnet(); 745 INP_INFO_RLOCK(pcbinfo); 746 inp = sotoinpcb(so); 747 if (!inp) { 748 INP_INFO_RUNLOCK(pcbinfo); 749 splx(s); 750 return ECONNRESET; 751 } 752 INP_LOCK(inp); 753 port = inp->inp_lport; 754 addr = inp->inp_laddr; 755 INP_UNLOCK(inp); 756 INP_INFO_RUNLOCK(pcbinfo); 757 splx(s); 758 759 *nam = in_sockaddr(port, &addr); 760 return 0; 761 } 762 763 /* 764 * The wrapper function will pass down the pcbinfo for this function to lock. 765 */ 766 int 767 in_setpeeraddr(so, nam, pcbinfo) 768 struct socket *so; 769 struct sockaddr **nam; 770 struct inpcbinfo *pcbinfo; 771 { 772 int s; 773 register struct inpcb *inp; 774 struct in_addr addr; 775 in_port_t port; 776 777 s = splnet(); 778 INP_INFO_RLOCK(pcbinfo); 779 inp = sotoinpcb(so); 780 if (!inp) { 781 INP_INFO_RUNLOCK(pcbinfo); 782 splx(s); 783 return ECONNRESET; 784 } 785 INP_LOCK(inp); 786 port = inp->inp_fport; 787 addr = inp->inp_faddr; 788 INP_UNLOCK(inp); 789 INP_INFO_RUNLOCK(pcbinfo); 790 splx(s); 791 792 *nam = in_sockaddr(port, &addr); 793 return 0; 794 } 795 796 void 797 in_pcbnotifyall(pcbinfo, faddr, errno, notify) 798 struct inpcbinfo *pcbinfo; 799 struct in_addr faddr; 800 int errno; 801 struct inpcb *(*notify)(struct inpcb *, int); 802 { 803 struct inpcb *inp, *ninp; 804 struct inpcbhead *head; 805 int s; 806 807 s = splnet(); 808 INP_INFO_WLOCK(pcbinfo); 809 head = pcbinfo->listhead; 810 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) { 811 INP_LOCK(inp); 812 ninp = LIST_NEXT(inp, inp_list); 813 #ifdef INET6 814 if ((inp->inp_vflag & INP_IPV4) == 0) { 815 INP_UNLOCK(inp); 816 continue; 817 } 818 #endif 819 if (inp->inp_faddr.s_addr != faddr.s_addr || 820 inp->inp_socket == NULL) { 821 INP_UNLOCK(inp); 822 continue; 823 } 824 if ((*notify)(inp, errno)) 825 INP_UNLOCK(inp); 826 } 827 INP_INFO_WUNLOCK(pcbinfo); 828 splx(s); 829 } 830 831 void 832 in_pcbpurgeif0(pcbinfo, ifp) 833 struct inpcbinfo *pcbinfo; 834 struct ifnet *ifp; 835 { 836 struct inpcb *inp; 837 struct ip_moptions *imo; 838 int i, gap; 839 840 /* why no splnet here? XXX */ 841 INP_INFO_RLOCK(pcbinfo); 842 LIST_FOREACH(inp, pcbinfo->listhead, inp_list) { 843 INP_LOCK(inp); 844 imo = inp->inp_moptions; 845 if ((inp->inp_vflag & INP_IPV4) && 846 imo != NULL) { 847 /* 848 * Unselect the outgoing interface if it is being 849 * detached. 850 */ 851 if (imo->imo_multicast_ifp == ifp) 852 imo->imo_multicast_ifp = NULL; 853 854 /* 855 * Drop multicast group membership if we joined 856 * through the interface being detached. 857 */ 858 for (i = 0, gap = 0; i < imo->imo_num_memberships; 859 i++) { 860 if (imo->imo_membership[i]->inm_ifp == ifp) { 861 in_delmulti(imo->imo_membership[i]); 862 gap++; 863 } else if (gap != 0) 864 imo->imo_membership[i - gap] = 865 imo->imo_membership[i]; 866 } 867 imo->imo_num_memberships -= gap; 868 } 869 INP_UNLOCK(inp); 870 } 871 INP_INFO_RUNLOCK(pcbinfo); 872 } 873 874 /* 875 * Lookup a PCB based on the local address and port. 876 */ 877 struct inpcb * 878 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay) 879 struct inpcbinfo *pcbinfo; 880 struct in_addr laddr; 881 u_int lport_arg; 882 int wild_okay; 883 { 884 register struct inpcb *inp; 885 int matchwild = 3, wildcard; 886 u_short lport = lport_arg; 887 888 INP_INFO_WLOCK_ASSERT(pcbinfo); 889 890 if (!wild_okay) { 891 struct inpcbhead *head; 892 /* 893 * Look for an unconnected (wildcard foreign addr) PCB that 894 * matches the local address and port we're looking for. 895 */ 896 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 897 LIST_FOREACH(inp, head, inp_hash) { 898 #ifdef INET6 899 if ((inp->inp_vflag & INP_IPV4) == 0) 900 continue; 901 #endif 902 if (inp->inp_faddr.s_addr == INADDR_ANY && 903 inp->inp_laddr.s_addr == laddr.s_addr && 904 inp->inp_lport == lport) { 905 /* 906 * Found. 907 */ 908 return (inp); 909 } 910 } 911 /* 912 * Not found. 913 */ 914 return (NULL); 915 } else { 916 struct inpcbporthead *porthash; 917 struct inpcbport *phd; 918 struct inpcb *match = NULL; 919 /* 920 * Best fit PCB lookup. 921 * 922 * First see if this local port is in use by looking on the 923 * port hash list. 924 */ 925 retrylookup: 926 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport, 927 pcbinfo->porthashmask)]; 928 LIST_FOREACH(phd, porthash, phd_hash) { 929 if (phd->phd_port == lport) 930 break; 931 } 932 if (phd != NULL) { 933 /* 934 * Port is in use by one or more PCBs. Look for best 935 * fit. 936 */ 937 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 938 wildcard = 0; 939 #ifdef INET6 940 if ((inp->inp_vflag & INP_IPV4) == 0) 941 continue; 942 #endif 943 /* 944 * Clean out old time_wait sockets if they 945 * are clogging up needed local ports. 946 */ 947 if ((inp->inp_vflag & INP_TIMEWAIT) != 0) { 948 if (tcp_twrecycleable((struct tcptw *)inp->inp_ppcb)) { 949 INP_LOCK(inp); 950 tcp_twclose((struct tcptw *)inp->inp_ppcb, 0); 951 match = NULL; 952 goto retrylookup; 953 } 954 } 955 if (inp->inp_faddr.s_addr != INADDR_ANY) 956 wildcard++; 957 if (inp->inp_laddr.s_addr != INADDR_ANY) { 958 if (laddr.s_addr == INADDR_ANY) 959 wildcard++; 960 else if (inp->inp_laddr.s_addr != laddr.s_addr) 961 continue; 962 } else { 963 if (laddr.s_addr != INADDR_ANY) 964 wildcard++; 965 } 966 if (wildcard < matchwild) { 967 match = inp; 968 matchwild = wildcard; 969 if (matchwild == 0) { 970 break; 971 } 972 } 973 } 974 } 975 return (match); 976 } 977 } 978 979 /* 980 * Lookup PCB in hash list. 981 */ 982 struct inpcb * 983 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, 984 ifp) 985 struct inpcbinfo *pcbinfo; 986 struct in_addr faddr, laddr; 987 u_int fport_arg, lport_arg; 988 int wildcard; 989 struct ifnet *ifp; 990 { 991 struct inpcbhead *head; 992 register struct inpcb *inp; 993 u_short fport = fport_arg, lport = lport_arg; 994 995 INP_INFO_RLOCK_ASSERT(pcbinfo); 996 /* 997 * First look for an exact match. 998 */ 999 head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)]; 1000 LIST_FOREACH(inp, head, inp_hash) { 1001 #ifdef INET6 1002 if ((inp->inp_vflag & INP_IPV4) == 0) 1003 continue; 1004 #endif 1005 if (inp->inp_faddr.s_addr == faddr.s_addr && 1006 inp->inp_laddr.s_addr == laddr.s_addr && 1007 inp->inp_fport == fport && 1008 inp->inp_lport == lport) { 1009 /* 1010 * Found. 1011 */ 1012 return (inp); 1013 } 1014 } 1015 if (wildcard) { 1016 struct inpcb *local_wild = NULL; 1017 #if defined(INET6) 1018 struct inpcb *local_wild_mapped = NULL; 1019 #endif /* defined(INET6) */ 1020 1021 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 1022 LIST_FOREACH(inp, head, inp_hash) { 1023 #ifdef INET6 1024 if ((inp->inp_vflag & INP_IPV4) == 0) 1025 continue; 1026 #endif 1027 if (inp->inp_faddr.s_addr == INADDR_ANY && 1028 inp->inp_lport == lport) { 1029 if (ifp && ifp->if_type == IFT_FAITH && 1030 (inp->inp_flags & INP_FAITH) == 0) 1031 continue; 1032 if (inp->inp_laddr.s_addr == laddr.s_addr) 1033 return (inp); 1034 else if (inp->inp_laddr.s_addr == INADDR_ANY) { 1035 #if defined(INET6) 1036 if (INP_CHECK_SOCKAF(inp->inp_socket, 1037 AF_INET6)) 1038 local_wild_mapped = inp; 1039 else 1040 #endif /* defined(INET6) */ 1041 local_wild = inp; 1042 } 1043 } 1044 } 1045 #if defined(INET6) 1046 if (local_wild == NULL) 1047 return (local_wild_mapped); 1048 #endif /* defined(INET6) */ 1049 return (local_wild); 1050 } 1051 1052 /* 1053 * Not found. 1054 */ 1055 return (NULL); 1056 } 1057 1058 /* 1059 * Insert PCB onto various hash lists. 1060 */ 1061 int 1062 in_pcbinshash(inp) 1063 struct inpcb *inp; 1064 { 1065 struct inpcbhead *pcbhash; 1066 struct inpcbporthead *pcbporthash; 1067 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1068 struct inpcbport *phd; 1069 u_int32_t hashkey_faddr; 1070 1071 INP_INFO_WLOCK_ASSERT(pcbinfo); 1072 #ifdef INET6 1073 if (inp->inp_vflag & INP_IPV6) 1074 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1075 else 1076 #endif /* INET6 */ 1077 hashkey_faddr = inp->inp_faddr.s_addr; 1078 1079 pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1080 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1081 1082 pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport, 1083 pcbinfo->porthashmask)]; 1084 1085 /* 1086 * Go through port list and look for a head for this lport. 1087 */ 1088 LIST_FOREACH(phd, pcbporthash, phd_hash) { 1089 if (phd->phd_port == inp->inp_lport) 1090 break; 1091 } 1092 /* 1093 * If none exists, malloc one and tack it on. 1094 */ 1095 if (phd == NULL) { 1096 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT); 1097 if (phd == NULL) { 1098 return (ENOBUFS); /* XXX */ 1099 } 1100 phd->phd_port = inp->inp_lport; 1101 LIST_INIT(&phd->phd_pcblist); 1102 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 1103 } 1104 inp->inp_phd = phd; 1105 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 1106 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 1107 return (0); 1108 } 1109 1110 /* 1111 * Move PCB to the proper hash bucket when { faddr, fport } have been 1112 * changed. NOTE: This does not handle the case of the lport changing (the 1113 * hashed port list would have to be updated as well), so the lport must 1114 * not change after in_pcbinshash() has been called. 1115 */ 1116 void 1117 in_pcbrehash(inp) 1118 struct inpcb *inp; 1119 { 1120 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1121 struct inpcbhead *head; 1122 u_int32_t hashkey_faddr; 1123 1124 INP_INFO_WLOCK_ASSERT(pcbinfo); 1125 /* XXX? INP_LOCK_ASSERT(inp); */ 1126 #ifdef INET6 1127 if (inp->inp_vflag & INP_IPV6) 1128 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1129 else 1130 #endif /* INET6 */ 1131 hashkey_faddr = inp->inp_faddr.s_addr; 1132 1133 head = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1134 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1135 1136 LIST_REMOVE(inp, inp_hash); 1137 LIST_INSERT_HEAD(head, inp, inp_hash); 1138 } 1139 1140 /* 1141 * Remove PCB from various lists. 1142 */ 1143 void 1144 in_pcbremlists(inp) 1145 struct inpcb *inp; 1146 { 1147 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1148 1149 INP_INFO_WLOCK_ASSERT(pcbinfo); 1150 INP_LOCK_ASSERT(inp); 1151 1152 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 1153 if (inp->inp_lport) { 1154 struct inpcbport *phd = inp->inp_phd; 1155 1156 LIST_REMOVE(inp, inp_hash); 1157 LIST_REMOVE(inp, inp_portlist); 1158 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 1159 LIST_REMOVE(phd, phd_hash); 1160 free(phd, M_PCB); 1161 } 1162 } 1163 LIST_REMOVE(inp, inp_list); 1164 pcbinfo->ipi_count--; 1165 } 1166 1167 /* 1168 * A set label operation has occurred at the socket layer, propagate the 1169 * label change into the in_pcb for the socket. 1170 */ 1171 void 1172 in_pcbsosetlabel(so) 1173 struct socket *so; 1174 { 1175 #ifdef MAC 1176 struct inpcb *inp; 1177 1178 /* XXX: Will assert socket lock when we have them. */ 1179 inp = (struct inpcb *)so->so_pcb; 1180 INP_LOCK(inp); 1181 mac_inpcb_sosetlabel(so, inp); 1182 INP_UNLOCK(inp); 1183 #endif 1184 } 1185 1186 int 1187 prison_xinpcb(struct thread *td, struct inpcb *inp) 1188 { 1189 if (!jailed(td->td_ucred)) 1190 return (0); 1191 if (ntohl(inp->inp_laddr.s_addr) == prison_getip(td->td_ucred)) 1192 return (0); 1193 return (1); 1194 } 1195