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 #include <netinet/udp.h> 63 #include <netinet/udp_var.h> 64 #ifdef INET6 65 #include <netinet/ip6.h> 66 #include <netinet6/ip6_var.h> 67 #endif /* INET6 */ 68 69 #ifdef IPSEC 70 #include <netinet6/ipsec.h> 71 #include <netkey/key.h> 72 #endif /* IPSEC */ 73 74 #ifdef FAST_IPSEC 75 #if defined(IPSEC) || defined(IPSEC_ESP) 76 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!" 77 #endif 78 79 #include <netipsec/ipsec.h> 80 #include <netipsec/key.h> 81 #endif /* FAST_IPSEC */ 82 83 /* 84 * These configure the range of local port addresses assigned to 85 * "unspecified" outgoing connections/packets/whatever. 86 */ 87 int ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 88 int ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 89 int ipport_firstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 90 int ipport_lastauto = IPPORT_HILASTAUTO; /* 65535 */ 91 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 92 int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 93 94 /* 95 * Reserved ports accessible only to root. There are significant 96 * security considerations that must be accounted for when changing these, 97 * but the security benefits can be great. Please be careful. 98 */ 99 int ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 100 int ipport_reservedlow = 0; 101 102 /* Variables dealing with random ephemeral port allocation. */ 103 int ipport_randomized = 1; /* user controlled via sysctl */ 104 int ipport_randomcps = 10; /* user controlled via sysctl */ 105 int ipport_randomtime = 45; /* user controlled via sysctl */ 106 int ipport_stoprandom = 0; /* toggled by ipport_tick */ 107 int ipport_tcpallocs; 108 int ipport_tcplastcount; 109 110 #define RANGECHK(var, min, max) \ 111 if ((var) < (min)) { (var) = (min); } \ 112 else if ((var) > (max)) { (var) = (max); } 113 114 static int 115 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) 116 { 117 int error; 118 119 error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req); 120 if (error == 0) { 121 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); 122 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1); 123 RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); 124 RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); 125 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); 126 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); 127 } 128 return (error); 129 } 130 131 #undef RANGECHK 132 133 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); 134 135 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW, 136 &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", ""); 137 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW, 138 &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", ""); 139 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW, 140 &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", ""); 141 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW, 142 &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", ""); 143 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW, 144 &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", ""); 145 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW, 146 &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", ""); 147 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, 148 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, ""); 149 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, 150 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, ""); 151 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW, 152 &ipport_randomized, 0, "Enable random port allocation"); 153 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW, 154 &ipport_randomcps, 0, "Maximum number of random port " 155 "allocations before switching to a sequental one"); 156 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW, 157 &ipport_randomtime, 0, "Minimum time to keep sequental port " 158 "allocation before switching to a random one"); 159 160 /* 161 * in_pcb.c: manage the Protocol Control Blocks. 162 * 163 * NOTE: It is assumed that most of these functions will be called at 164 * splnet(). XXX - There are, unfortunately, a few exceptions to this 165 * rule that should be fixed. 166 */ 167 168 /* 169 * Allocate a PCB and associate it with the socket. 170 */ 171 int 172 in_pcballoc(so, pcbinfo, type) 173 struct socket *so; 174 struct inpcbinfo *pcbinfo; 175 const char *type; 176 { 177 register struct inpcb *inp; 178 int error; 179 180 INP_INFO_WLOCK_ASSERT(pcbinfo); 181 error = 0; 182 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT | M_ZERO); 183 if (inp == NULL) 184 return (ENOBUFS); 185 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 186 inp->inp_pcbinfo = pcbinfo; 187 inp->inp_socket = so; 188 #ifdef MAC 189 error = mac_init_inpcb(inp, M_NOWAIT); 190 if (error != 0) 191 goto out; 192 SOCK_LOCK(so); 193 mac_create_inpcb_from_socket(so, inp); 194 SOCK_UNLOCK(so); 195 #endif 196 #if defined(IPSEC) || defined(FAST_IPSEC) 197 #ifdef FAST_IPSEC 198 error = ipsec_init_policy(so, &inp->inp_sp); 199 #else 200 error = ipsec_init_pcbpolicy(so, &inp->inp_sp); 201 #endif 202 if (error != 0) 203 goto out; 204 #endif /*IPSEC*/ 205 #if defined(INET6) 206 if (INP_SOCKAF(so) == AF_INET6) { 207 inp->inp_vflag |= INP_IPV6PROTO; 208 if (ip6_v6only) 209 inp->inp_flags |= IN6P_IPV6_V6ONLY; 210 } 211 #endif 212 LIST_INSERT_HEAD(pcbinfo->listhead, inp, inp_list); 213 pcbinfo->ipi_count++; 214 so->so_pcb = (caddr_t)inp; 215 INP_LOCK_INIT(inp, "inp", type); 216 #ifdef INET6 217 if (ip6_auto_flowlabel) 218 inp->inp_flags |= IN6P_AUTOFLOWLABEL; 219 #endif 220 #if defined(IPSEC) || defined(FAST_IPSEC) || defined(MAC) 221 out: 222 if (error != 0) 223 uma_zfree(pcbinfo->ipi_zone, inp); 224 #endif 225 return (error); 226 } 227 228 int 229 in_pcbbind(inp, nam, cred) 230 register struct inpcb *inp; 231 struct sockaddr *nam; 232 struct ucred *cred; 233 { 234 int anonport, error; 235 236 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 237 INP_LOCK_ASSERT(inp); 238 239 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) 240 return (EINVAL); 241 anonport = inp->inp_lport == 0 && (nam == NULL || 242 ((struct sockaddr_in *)nam)->sin_port == 0); 243 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, 244 &inp->inp_lport, cred); 245 if (error) 246 return (error); 247 if (in_pcbinshash(inp) != 0) { 248 inp->inp_laddr.s_addr = INADDR_ANY; 249 inp->inp_lport = 0; 250 return (EAGAIN); 251 } 252 if (anonport) 253 inp->inp_flags |= INP_ANONPORT; 254 return (0); 255 } 256 257 /* 258 * Set up a bind operation on a PCB, performing port allocation 259 * as required, but do not actually modify the PCB. Callers can 260 * either complete the bind by setting inp_laddr/inp_lport and 261 * calling in_pcbinshash(), or they can just use the resulting 262 * port and address to authorise the sending of a once-off packet. 263 * 264 * On error, the values of *laddrp and *lportp are not changed. 265 */ 266 int 267 in_pcbbind_setup(inp, nam, laddrp, lportp, cred) 268 struct inpcb *inp; 269 struct sockaddr *nam; 270 in_addr_t *laddrp; 271 u_short *lportp; 272 struct ucred *cred; 273 { 274 struct socket *so = inp->inp_socket; 275 unsigned short *lastport; 276 struct sockaddr_in *sin; 277 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 278 struct in_addr laddr; 279 u_short lport = 0; 280 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT); 281 int error, prison = 0; 282 int dorandom; 283 284 INP_INFO_WLOCK_ASSERT(pcbinfo); 285 INP_LOCK_ASSERT(inp); 286 287 if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */ 288 return (EADDRNOTAVAIL); 289 laddr.s_addr = *laddrp; 290 if (nam != NULL && laddr.s_addr != INADDR_ANY) 291 return (EINVAL); 292 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 293 wild = 1; 294 if (nam) { 295 sin = (struct sockaddr_in *)nam; 296 if (nam->sa_len != sizeof (*sin)) 297 return (EINVAL); 298 #ifdef notdef 299 /* 300 * We should check the family, but old programs 301 * incorrectly fail to initialize it. 302 */ 303 if (sin->sin_family != AF_INET) 304 return (EAFNOSUPPORT); 305 #endif 306 if (sin->sin_addr.s_addr != INADDR_ANY) 307 if (prison_ip(cred, 0, &sin->sin_addr.s_addr)) 308 return(EINVAL); 309 if (sin->sin_port != *lportp) { 310 /* Don't allow the port to change. */ 311 if (*lportp != 0) 312 return (EINVAL); 313 lport = sin->sin_port; 314 } 315 /* NB: lport is left as 0 if the port isn't being changed. */ 316 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 317 /* 318 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 319 * allow complete duplication of binding if 320 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 321 * and a multicast address is bound on both 322 * new and duplicated sockets. 323 */ 324 if (so->so_options & SO_REUSEADDR) 325 reuseport = SO_REUSEADDR|SO_REUSEPORT; 326 } else if (sin->sin_addr.s_addr != INADDR_ANY) { 327 sin->sin_port = 0; /* yech... */ 328 bzero(&sin->sin_zero, sizeof(sin->sin_zero)); 329 if (ifa_ifwithaddr((struct sockaddr *)sin) == 0) 330 return (EADDRNOTAVAIL); 331 } 332 laddr = sin->sin_addr; 333 if (lport) { 334 struct inpcb *t; 335 /* GROSS */ 336 if (ntohs(lport) <= ipport_reservedhigh && 337 ntohs(lport) >= ipport_reservedlow && 338 suser_cred(cred, SUSER_ALLOWJAIL)) 339 return (EACCES); 340 if (jailed(cred)) 341 prison = 1; 342 if (so->so_cred->cr_uid != 0 && 343 !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 344 t = in_pcblookup_local(inp->inp_pcbinfo, 345 sin->sin_addr, lport, 346 prison ? 0 : INPLOOKUP_WILDCARD); 347 /* 348 * XXX 349 * This entire block sorely needs a rewrite. 350 */ 351 if (t && 352 ((t->inp_vflag & INP_TIMEWAIT) == 0) && 353 (so->so_type != SOCK_STREAM || 354 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && 355 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 356 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 357 (t->inp_socket->so_options & 358 SO_REUSEPORT) == 0) && 359 (so->so_cred->cr_uid != 360 t->inp_socket->so_cred->cr_uid)) 361 return (EADDRINUSE); 362 } 363 if (prison && prison_ip(cred, 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(cred, 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 ((error = suser_cred(cred, SUSER_ALLOWJAIL)) != 0) 401 return error; 402 first = ipport_lowfirstauto; /* 1023 */ 403 last = ipport_lowlastauto; /* 600 */ 404 lastport = &pcbinfo->lastlow; 405 } else { 406 first = ipport_firstauto; /* sysctl */ 407 last = ipport_lastauto; 408 lastport = &pcbinfo->lastport; 409 } 410 /* 411 * For UDP, use random port allocation as long as the user 412 * allows it. For TCP (and as of yet unknown) connections, 413 * use random port allocation only if the user allows it AND 414 * ipport_tick() allows it. 415 */ 416 if (ipport_randomized && 417 (!ipport_stoprandom || pcbinfo == &udbinfo)) 418 dorandom = 1; 419 else 420 dorandom = 0; 421 /* 422 * It makes no sense to do random port allocation if 423 * we have the only port available. 424 */ 425 if (first == last) 426 dorandom = 0; 427 /* Make sure to not include UDP packets in the count. */ 428 if (pcbinfo != &udbinfo) 429 ipport_tcpallocs++; 430 /* 431 * Simple check to ensure all ports are not used up causing 432 * a deadlock here. 433 * 434 * We split the two cases (up and down) so that the direction 435 * is not being tested on each round of the loop. 436 */ 437 if (first > last) { 438 /* 439 * counting down 440 */ 441 if (dorandom) 442 *lastport = first - 443 (arc4random() % (first - last)); 444 count = first - last; 445 446 do { 447 if (count-- < 0) /* completely used? */ 448 return (EADDRNOTAVAIL); 449 --*lastport; 450 if (*lastport > first || *lastport < last) 451 *lastport = first; 452 lport = htons(*lastport); 453 } while (in_pcblookup_local(pcbinfo, laddr, lport, 454 wild)); 455 } else { 456 /* 457 * counting up 458 */ 459 if (dorandom) 460 *lastport = first + 461 (arc4random() % (last - first)); 462 count = last - first; 463 464 do { 465 if (count-- < 0) /* completely used? */ 466 return (EADDRNOTAVAIL); 467 ++*lastport; 468 if (*lastport < first || *lastport > last) 469 *lastport = first; 470 lport = htons(*lastport); 471 } while (in_pcblookup_local(pcbinfo, laddr, lport, 472 wild)); 473 } 474 } 475 if (prison_ip(cred, 0, &laddr.s_addr)) 476 return (EINVAL); 477 *laddrp = laddr.s_addr; 478 *lportp = lport; 479 return (0); 480 } 481 482 /* 483 * Connect from a socket to a specified address. 484 * Both address and port must be specified in argument sin. 485 * If don't have a local address for this socket yet, 486 * then pick one. 487 */ 488 int 489 in_pcbconnect(inp, nam, cred) 490 register struct inpcb *inp; 491 struct sockaddr *nam; 492 struct ucred *cred; 493 { 494 u_short lport, fport; 495 in_addr_t laddr, faddr; 496 int anonport, error; 497 498 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 499 INP_LOCK_ASSERT(inp); 500 501 lport = inp->inp_lport; 502 laddr = inp->inp_laddr.s_addr; 503 anonport = (lport == 0); 504 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, 505 NULL, cred); 506 if (error) 507 return (error); 508 509 /* Do the initial binding of the local address if required. */ 510 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { 511 inp->inp_lport = lport; 512 inp->inp_laddr.s_addr = laddr; 513 if (in_pcbinshash(inp) != 0) { 514 inp->inp_laddr.s_addr = INADDR_ANY; 515 inp->inp_lport = 0; 516 return (EAGAIN); 517 } 518 } 519 520 /* Commit the remaining changes. */ 521 inp->inp_lport = lport; 522 inp->inp_laddr.s_addr = laddr; 523 inp->inp_faddr.s_addr = faddr; 524 inp->inp_fport = fport; 525 in_pcbrehash(inp); 526 #ifdef IPSEC 527 if (inp->inp_socket->so_type == SOCK_STREAM) 528 ipsec_pcbconn(inp->inp_sp); 529 #endif 530 if (anonport) 531 inp->inp_flags |= INP_ANONPORT; 532 return (0); 533 } 534 535 /* 536 * Set up for a connect from a socket to the specified address. 537 * On entry, *laddrp and *lportp should contain the current local 538 * address and port for the PCB; these are updated to the values 539 * that should be placed in inp_laddr and inp_lport to complete 540 * the connect. 541 * 542 * On success, *faddrp and *fportp will be set to the remote address 543 * and port. These are not updated in the error case. 544 * 545 * If the operation fails because the connection already exists, 546 * *oinpp will be set to the PCB of that connection so that the 547 * caller can decide to override it. In all other cases, *oinpp 548 * is set to NULL. 549 */ 550 int 551 in_pcbconnect_setup(inp, nam, laddrp, lportp, faddrp, fportp, oinpp, cred) 552 register struct inpcb *inp; 553 struct sockaddr *nam; 554 in_addr_t *laddrp; 555 u_short *lportp; 556 in_addr_t *faddrp; 557 u_short *fportp; 558 struct inpcb **oinpp; 559 struct ucred *cred; 560 { 561 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 562 struct in_ifaddr *ia; 563 struct sockaddr_in sa; 564 struct ucred *socred; 565 struct inpcb *oinp; 566 struct in_addr laddr, faddr; 567 u_short lport, fport; 568 int error; 569 570 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 571 INP_LOCK_ASSERT(inp); 572 573 if (oinpp != NULL) 574 *oinpp = NULL; 575 if (nam->sa_len != sizeof (*sin)) 576 return (EINVAL); 577 if (sin->sin_family != AF_INET) 578 return (EAFNOSUPPORT); 579 if (sin->sin_port == 0) 580 return (EADDRNOTAVAIL); 581 laddr.s_addr = *laddrp; 582 lport = *lportp; 583 faddr = sin->sin_addr; 584 fport = sin->sin_port; 585 socred = inp->inp_socket->so_cred; 586 if (laddr.s_addr == INADDR_ANY && jailed(socred)) { 587 bzero(&sa, sizeof(sa)); 588 sa.sin_addr.s_addr = htonl(prison_getip(socred)); 589 sa.sin_len = sizeof(sa); 590 sa.sin_family = AF_INET; 591 error = in_pcbbind_setup(inp, (struct sockaddr *)&sa, 592 &laddr.s_addr, &lport, cred); 593 if (error) 594 return (error); 595 } 596 if (!TAILQ_EMPTY(&in_ifaddrhead)) { 597 /* 598 * If the destination address is INADDR_ANY, 599 * use the primary local address. 600 * If the supplied address is INADDR_BROADCAST, 601 * and the primary interface supports broadcast, 602 * choose the broadcast address for that interface. 603 */ 604 if (faddr.s_addr == INADDR_ANY) 605 faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr; 606 else if (faddr.s_addr == (u_long)INADDR_BROADCAST && 607 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags & 608 IFF_BROADCAST)) 609 faddr = satosin(&TAILQ_FIRST( 610 &in_ifaddrhead)->ia_broadaddr)->sin_addr; 611 } 612 if (laddr.s_addr == INADDR_ANY) { 613 struct route sro; 614 615 bzero(&sro, sizeof(sro)); 616 ia = (struct in_ifaddr *)0; 617 /* 618 * If route is known our src addr is taken from the i/f, 619 * else punt. 620 */ 621 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) { 622 /* Find out route to destination */ 623 sro.ro_dst.sa_family = AF_INET; 624 sro.ro_dst.sa_len = sizeof(struct sockaddr_in); 625 ((struct sockaddr_in *)&sro.ro_dst)->sin_addr = faddr; 626 rtalloc_ign(&sro, RTF_CLONING); 627 } 628 /* 629 * If we found a route, use the address 630 * corresponding to the outgoing interface. 631 */ 632 if (sro.ro_rt) { 633 ia = ifatoia(sro.ro_rt->rt_ifa); 634 RTFREE(sro.ro_rt); 635 } 636 if (ia == 0) { 637 bzero(&sa, sizeof(sa)); 638 sa.sin_addr = faddr; 639 sa.sin_len = sizeof(sa); 640 sa.sin_family = AF_INET; 641 642 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa))); 643 if (ia == 0) 644 ia = ifatoia(ifa_ifwithnet(sintosa(&sa))); 645 if (ia == 0) 646 return (ENETUNREACH); 647 } 648 /* 649 * If the destination address is multicast and an outgoing 650 * interface has been set as a multicast option, use the 651 * address of that interface as our source address. 652 */ 653 if (IN_MULTICAST(ntohl(faddr.s_addr)) && 654 inp->inp_moptions != NULL) { 655 struct ip_moptions *imo; 656 struct ifnet *ifp; 657 658 imo = inp->inp_moptions; 659 if (imo->imo_multicast_ifp != NULL) { 660 ifp = imo->imo_multicast_ifp; 661 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) 662 if (ia->ia_ifp == ifp) 663 break; 664 if (ia == 0) 665 return (EADDRNOTAVAIL); 666 } 667 } 668 laddr = ia->ia_addr.sin_addr; 669 } 670 671 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport, 672 0, NULL); 673 if (oinp != NULL) { 674 if (oinpp != NULL) 675 *oinpp = oinp; 676 return (EADDRINUSE); 677 } 678 if (lport == 0) { 679 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, 680 cred); 681 if (error) 682 return (error); 683 } 684 *laddrp = laddr.s_addr; 685 *lportp = lport; 686 *faddrp = faddr.s_addr; 687 *fportp = fport; 688 return (0); 689 } 690 691 void 692 in_pcbdisconnect(inp) 693 struct inpcb *inp; 694 { 695 INP_LOCK_ASSERT(inp); 696 697 inp->inp_faddr.s_addr = INADDR_ANY; 698 inp->inp_fport = 0; 699 in_pcbrehash(inp); 700 #ifdef IPSEC 701 ipsec_pcbdisconn(inp->inp_sp); 702 #endif 703 if (inp->inp_socket->so_state & SS_NOFDREF) 704 in_pcbdetach(inp); 705 } 706 707 void 708 in_pcbdetach(inp) 709 struct inpcb *inp; 710 { 711 struct socket *so = inp->inp_socket; 712 struct inpcbinfo *ipi = inp->inp_pcbinfo; 713 714 INP_LOCK_ASSERT(inp); 715 716 #if defined(IPSEC) || defined(FAST_IPSEC) 717 ipsec4_delete_pcbpolicy(inp); 718 #endif /*IPSEC*/ 719 inp->inp_gencnt = ++ipi->ipi_gencnt; 720 in_pcbremlists(inp); 721 if (so) { 722 ACCEPT_LOCK(); 723 SOCK_LOCK(so); 724 so->so_pcb = NULL; 725 sotryfree(so); 726 } 727 if (inp->inp_options) 728 (void)m_free(inp->inp_options); 729 ip_freemoptions(inp->inp_moptions); 730 inp->inp_vflag = 0; 731 INP_LOCK_DESTROY(inp); 732 #ifdef MAC 733 mac_destroy_inpcb(inp); 734 #endif 735 uma_zfree(ipi->ipi_zone, inp); 736 } 737 738 struct sockaddr * 739 in_sockaddr(port, addr_p) 740 in_port_t port; 741 struct in_addr *addr_p; 742 { 743 struct sockaddr_in *sin; 744 745 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, 746 M_WAITOK | M_ZERO); 747 sin->sin_family = AF_INET; 748 sin->sin_len = sizeof(*sin); 749 sin->sin_addr = *addr_p; 750 sin->sin_port = port; 751 752 return (struct sockaddr *)sin; 753 } 754 755 /* 756 * The wrapper function will pass down the pcbinfo for this function to lock. 757 * The socket must have a valid 758 * (i.e., non-nil) PCB, but it should be impossible to get an invalid one 759 * except through a kernel programming error, so it is acceptable to panic 760 * (or in this case trap) if the PCB is invalid. (Actually, we don't trap 761 * because there actually /is/ a programming error somewhere... XXX) 762 */ 763 int 764 in_setsockaddr(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_lport; 784 addr = inp->inp_laddr; 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 /* 794 * The wrapper function will pass down the pcbinfo for this function to lock. 795 */ 796 int 797 in_setpeeraddr(so, nam, pcbinfo) 798 struct socket *so; 799 struct sockaddr **nam; 800 struct inpcbinfo *pcbinfo; 801 { 802 int s; 803 register struct inpcb *inp; 804 struct in_addr addr; 805 in_port_t port; 806 807 s = splnet(); 808 INP_INFO_RLOCK(pcbinfo); 809 inp = sotoinpcb(so); 810 if (!inp) { 811 INP_INFO_RUNLOCK(pcbinfo); 812 splx(s); 813 return ECONNRESET; 814 } 815 INP_LOCK(inp); 816 port = inp->inp_fport; 817 addr = inp->inp_faddr; 818 INP_UNLOCK(inp); 819 INP_INFO_RUNLOCK(pcbinfo); 820 splx(s); 821 822 *nam = in_sockaddr(port, &addr); 823 return 0; 824 } 825 826 void 827 in_pcbnotifyall(pcbinfo, faddr, errno, notify) 828 struct inpcbinfo *pcbinfo; 829 struct in_addr faddr; 830 int errno; 831 struct inpcb *(*notify)(struct inpcb *, int); 832 { 833 struct inpcb *inp, *ninp; 834 struct inpcbhead *head; 835 int s; 836 837 s = splnet(); 838 INP_INFO_WLOCK(pcbinfo); 839 head = pcbinfo->listhead; 840 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) { 841 INP_LOCK(inp); 842 ninp = LIST_NEXT(inp, inp_list); 843 #ifdef INET6 844 if ((inp->inp_vflag & INP_IPV4) == 0) { 845 INP_UNLOCK(inp); 846 continue; 847 } 848 #endif 849 if (inp->inp_faddr.s_addr != faddr.s_addr || 850 inp->inp_socket == NULL) { 851 INP_UNLOCK(inp); 852 continue; 853 } 854 if ((*notify)(inp, errno)) 855 INP_UNLOCK(inp); 856 } 857 INP_INFO_WUNLOCK(pcbinfo); 858 splx(s); 859 } 860 861 void 862 in_pcbpurgeif0(pcbinfo, ifp) 863 struct inpcbinfo *pcbinfo; 864 struct ifnet *ifp; 865 { 866 struct inpcb *inp; 867 struct ip_moptions *imo; 868 int i, gap; 869 870 /* why no splnet here? XXX */ 871 INP_INFO_RLOCK(pcbinfo); 872 LIST_FOREACH(inp, pcbinfo->listhead, inp_list) { 873 INP_LOCK(inp); 874 imo = inp->inp_moptions; 875 if ((inp->inp_vflag & INP_IPV4) && 876 imo != NULL) { 877 /* 878 * Unselect the outgoing interface if it is being 879 * detached. 880 */ 881 if (imo->imo_multicast_ifp == ifp) 882 imo->imo_multicast_ifp = NULL; 883 884 /* 885 * Drop multicast group membership if we joined 886 * through the interface being detached. 887 */ 888 for (i = 0, gap = 0; i < imo->imo_num_memberships; 889 i++) { 890 if (imo->imo_membership[i]->inm_ifp == ifp) { 891 in_delmulti(imo->imo_membership[i]); 892 gap++; 893 } else if (gap != 0) 894 imo->imo_membership[i - gap] = 895 imo->imo_membership[i]; 896 } 897 imo->imo_num_memberships -= gap; 898 } 899 INP_UNLOCK(inp); 900 } 901 INP_INFO_RUNLOCK(pcbinfo); 902 } 903 904 /* 905 * Lookup a PCB based on the local address and port. 906 */ 907 struct inpcb * 908 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay) 909 struct inpcbinfo *pcbinfo; 910 struct in_addr laddr; 911 u_int lport_arg; 912 int wild_okay; 913 { 914 register struct inpcb *inp; 915 int matchwild = 3, wildcard; 916 u_short lport = lport_arg; 917 918 INP_INFO_WLOCK_ASSERT(pcbinfo); 919 920 if (!wild_okay) { 921 struct inpcbhead *head; 922 /* 923 * Look for an unconnected (wildcard foreign addr) PCB that 924 * matches the local address and port we're looking for. 925 */ 926 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 927 LIST_FOREACH(inp, head, inp_hash) { 928 #ifdef INET6 929 if ((inp->inp_vflag & INP_IPV4) == 0) 930 continue; 931 #endif 932 if (inp->inp_faddr.s_addr == INADDR_ANY && 933 inp->inp_laddr.s_addr == laddr.s_addr && 934 inp->inp_lport == lport) { 935 /* 936 * Found. 937 */ 938 return (inp); 939 } 940 } 941 /* 942 * Not found. 943 */ 944 return (NULL); 945 } else { 946 struct inpcbporthead *porthash; 947 struct inpcbport *phd; 948 struct inpcb *match = NULL; 949 /* 950 * Best fit PCB lookup. 951 * 952 * First see if this local port is in use by looking on the 953 * port hash list. 954 */ 955 retrylookup: 956 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport, 957 pcbinfo->porthashmask)]; 958 LIST_FOREACH(phd, porthash, phd_hash) { 959 if (phd->phd_port == lport) 960 break; 961 } 962 if (phd != NULL) { 963 /* 964 * Port is in use by one or more PCBs. Look for best 965 * fit. 966 */ 967 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 968 wildcard = 0; 969 #ifdef INET6 970 if ((inp->inp_vflag & INP_IPV4) == 0) 971 continue; 972 #endif 973 /* 974 * Clean out old time_wait sockets if they 975 * are clogging up needed local ports. 976 */ 977 if ((inp->inp_vflag & INP_TIMEWAIT) != 0) { 978 if (tcp_twrecycleable((struct tcptw *)inp->inp_ppcb)) { 979 INP_LOCK(inp); 980 tcp_twclose((struct tcptw *)inp->inp_ppcb, 0); 981 match = NULL; 982 goto retrylookup; 983 } 984 } 985 if (inp->inp_faddr.s_addr != INADDR_ANY) 986 wildcard++; 987 if (inp->inp_laddr.s_addr != INADDR_ANY) { 988 if (laddr.s_addr == INADDR_ANY) 989 wildcard++; 990 else if (inp->inp_laddr.s_addr != laddr.s_addr) 991 continue; 992 } else { 993 if (laddr.s_addr != INADDR_ANY) 994 wildcard++; 995 } 996 if (wildcard < matchwild) { 997 match = inp; 998 matchwild = wildcard; 999 if (matchwild == 0) { 1000 break; 1001 } 1002 } 1003 } 1004 } 1005 return (match); 1006 } 1007 } 1008 1009 /* 1010 * Lookup PCB in hash list. 1011 */ 1012 struct inpcb * 1013 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, 1014 ifp) 1015 struct inpcbinfo *pcbinfo; 1016 struct in_addr faddr, laddr; 1017 u_int fport_arg, lport_arg; 1018 int wildcard; 1019 struct ifnet *ifp; 1020 { 1021 struct inpcbhead *head; 1022 register struct inpcb *inp; 1023 u_short fport = fport_arg, lport = lport_arg; 1024 1025 INP_INFO_RLOCK_ASSERT(pcbinfo); 1026 /* 1027 * First look for an exact match. 1028 */ 1029 head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)]; 1030 LIST_FOREACH(inp, head, inp_hash) { 1031 #ifdef INET6 1032 if ((inp->inp_vflag & INP_IPV4) == 0) 1033 continue; 1034 #endif 1035 if (inp->inp_faddr.s_addr == faddr.s_addr && 1036 inp->inp_laddr.s_addr == laddr.s_addr && 1037 inp->inp_fport == fport && 1038 inp->inp_lport == lport) { 1039 /* 1040 * Found. 1041 */ 1042 return (inp); 1043 } 1044 } 1045 if (wildcard) { 1046 struct inpcb *local_wild = NULL; 1047 #if defined(INET6) 1048 struct inpcb *local_wild_mapped = NULL; 1049 #endif /* defined(INET6) */ 1050 1051 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 1052 LIST_FOREACH(inp, head, inp_hash) { 1053 #ifdef INET6 1054 if ((inp->inp_vflag & INP_IPV4) == 0) 1055 continue; 1056 #endif 1057 if (inp->inp_faddr.s_addr == INADDR_ANY && 1058 inp->inp_lport == lport) { 1059 if (ifp && ifp->if_type == IFT_FAITH && 1060 (inp->inp_flags & INP_FAITH) == 0) 1061 continue; 1062 if (inp->inp_laddr.s_addr == laddr.s_addr) 1063 return (inp); 1064 else if (inp->inp_laddr.s_addr == INADDR_ANY) { 1065 #if defined(INET6) 1066 if (INP_CHECK_SOCKAF(inp->inp_socket, 1067 AF_INET6)) 1068 local_wild_mapped = inp; 1069 else 1070 #endif /* defined(INET6) */ 1071 local_wild = inp; 1072 } 1073 } 1074 } 1075 #if defined(INET6) 1076 if (local_wild == NULL) 1077 return (local_wild_mapped); 1078 #endif /* defined(INET6) */ 1079 return (local_wild); 1080 } 1081 1082 /* 1083 * Not found. 1084 */ 1085 return (NULL); 1086 } 1087 1088 /* 1089 * Insert PCB onto various hash lists. 1090 */ 1091 int 1092 in_pcbinshash(inp) 1093 struct inpcb *inp; 1094 { 1095 struct inpcbhead *pcbhash; 1096 struct inpcbporthead *pcbporthash; 1097 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1098 struct inpcbport *phd; 1099 u_int32_t hashkey_faddr; 1100 1101 INP_INFO_WLOCK_ASSERT(pcbinfo); 1102 #ifdef INET6 1103 if (inp->inp_vflag & INP_IPV6) 1104 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1105 else 1106 #endif /* INET6 */ 1107 hashkey_faddr = inp->inp_faddr.s_addr; 1108 1109 pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1110 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1111 1112 pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport, 1113 pcbinfo->porthashmask)]; 1114 1115 /* 1116 * Go through port list and look for a head for this lport. 1117 */ 1118 LIST_FOREACH(phd, pcbporthash, phd_hash) { 1119 if (phd->phd_port == inp->inp_lport) 1120 break; 1121 } 1122 /* 1123 * If none exists, malloc one and tack it on. 1124 */ 1125 if (phd == NULL) { 1126 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT); 1127 if (phd == NULL) { 1128 return (ENOBUFS); /* XXX */ 1129 } 1130 phd->phd_port = inp->inp_lport; 1131 LIST_INIT(&phd->phd_pcblist); 1132 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 1133 } 1134 inp->inp_phd = phd; 1135 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 1136 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 1137 return (0); 1138 } 1139 1140 /* 1141 * Move PCB to the proper hash bucket when { faddr, fport } have been 1142 * changed. NOTE: This does not handle the case of the lport changing (the 1143 * hashed port list would have to be updated as well), so the lport must 1144 * not change after in_pcbinshash() has been called. 1145 */ 1146 void 1147 in_pcbrehash(inp) 1148 struct inpcb *inp; 1149 { 1150 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1151 struct inpcbhead *head; 1152 u_int32_t hashkey_faddr; 1153 1154 INP_INFO_WLOCK_ASSERT(pcbinfo); 1155 INP_LOCK_ASSERT(inp); 1156 #ifdef INET6 1157 if (inp->inp_vflag & INP_IPV6) 1158 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1159 else 1160 #endif /* INET6 */ 1161 hashkey_faddr = inp->inp_faddr.s_addr; 1162 1163 head = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1164 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1165 1166 LIST_REMOVE(inp, inp_hash); 1167 LIST_INSERT_HEAD(head, inp, inp_hash); 1168 } 1169 1170 /* 1171 * Remove PCB from various lists. 1172 */ 1173 void 1174 in_pcbremlists(inp) 1175 struct inpcb *inp; 1176 { 1177 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1178 1179 INP_INFO_WLOCK_ASSERT(pcbinfo); 1180 INP_LOCK_ASSERT(inp); 1181 1182 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 1183 if (inp->inp_lport) { 1184 struct inpcbport *phd = inp->inp_phd; 1185 1186 LIST_REMOVE(inp, inp_hash); 1187 LIST_REMOVE(inp, inp_portlist); 1188 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 1189 LIST_REMOVE(phd, phd_hash); 1190 free(phd, M_PCB); 1191 } 1192 } 1193 LIST_REMOVE(inp, inp_list); 1194 pcbinfo->ipi_count--; 1195 } 1196 1197 /* 1198 * A set label operation has occurred at the socket layer, propagate the 1199 * label change into the in_pcb for the socket. 1200 */ 1201 void 1202 in_pcbsosetlabel(so) 1203 struct socket *so; 1204 { 1205 #ifdef MAC 1206 struct inpcb *inp; 1207 1208 inp = (struct inpcb *)so->so_pcb; 1209 INP_LOCK(inp); 1210 SOCK_LOCK(so); 1211 mac_inpcb_sosetlabel(so, inp); 1212 SOCK_UNLOCK(so); 1213 INP_UNLOCK(inp); 1214 #endif 1215 } 1216 1217 /* 1218 * ipport_tick runs once per second, determining if random port 1219 * allocation should be continued. If more than ipport_randomcps 1220 * ports have been allocated in the last second, then we return to 1221 * sequential port allocation. We return to random allocation only 1222 * once we drop below ipport_randomcps for at least ipport_randomtime 1223 * seconds. 1224 */ 1225 1226 void 1227 ipport_tick(xtp) 1228 void *xtp; 1229 { 1230 if (ipport_tcpallocs > ipport_tcplastcount + ipport_randomcps) { 1231 ipport_stoprandom = ipport_randomtime; 1232 } else { 1233 if (ipport_stoprandom > 0) 1234 ipport_stoprandom--; 1235 } 1236 ipport_tcplastcount = ipport_tcpallocs; 1237 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL); 1238 } 1239