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 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/limits.h> 43 #include <sys/malloc.h> 44 #include <sys/mbuf.h> 45 #include <sys/domain.h> 46 #include <sys/protosw.h> 47 #include <sys/socket.h> 48 #include <sys/socketvar.h> 49 #include <sys/proc.h> 50 #include <sys/jail.h> 51 #include <sys/kernel.h> 52 #include <sys/sysctl.h> 53 54 #include <vm/uma.h> 55 56 #include <net/if.h> 57 #include <net/if_types.h> 58 #include <net/route.h> 59 60 #include <netinet/in.h> 61 #include <netinet/in_pcb.h> 62 #include <netinet/in_var.h> 63 #include <netinet/ip_var.h> 64 #include <netinet/tcp_var.h> 65 #ifdef INET6 66 #include <netinet/ip6.h> 67 #include <netinet6/ip6_var.h> 68 #endif /* INET6 */ 69 70 #ifdef IPSEC 71 #include <netinet6/ipsec.h> 72 #include <netkey/key.h> 73 #endif /* IPSEC */ 74 75 #ifdef FAST_IPSEC 76 #if defined(IPSEC) || defined(IPSEC_ESP) 77 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!" 78 #endif 79 80 #include <netipsec/ipsec.h> 81 #include <netipsec/key.h> 82 #endif /* FAST_IPSEC */ 83 84 struct in_addr zeroin_addr; 85 86 /* 87 * These configure the range of local port addresses assigned to 88 * "unspecified" outgoing connections/packets/whatever. 89 */ 90 int ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 91 int ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 92 int ipport_firstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 93 int ipport_lastauto = IPPORT_HILASTAUTO; /* 65535 */ 94 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 95 int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 96 97 /* 98 * Reserved ports accessible only to root. There are significant 99 * security considerations that must be accounted for when changing these, 100 * but the security benefits can be great. Please be careful. 101 */ 102 int ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */ 103 int ipport_reservedlow = 0; 104 105 #define RANGECHK(var, min, max) \ 106 if ((var) < (min)) { (var) = (min); } \ 107 else if ((var) > (max)) { (var) = (max); } 108 109 static int 110 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) 111 { 112 int error = sysctl_handle_int(oidp, 113 oidp->oid_arg1, oidp->oid_arg2, req); 114 if (!error) { 115 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); 116 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1); 117 RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX); 118 RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX); 119 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX); 120 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX); 121 } 122 return error; 123 } 124 125 #undef RANGECHK 126 127 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); 128 129 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW, 130 &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", ""); 131 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW, 132 &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", ""); 133 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW, 134 &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", ""); 135 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW, 136 &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", ""); 137 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW, 138 &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", ""); 139 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW, 140 &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", ""); 141 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, 142 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, ""); 143 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, 144 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, ""); 145 146 /* 147 * in_pcb.c: manage the Protocol Control Blocks. 148 * 149 * NOTE: It is assumed that most of these functions will be called at 150 * splnet(). XXX - There are, unfortunately, a few exceptions to this 151 * rule that should be fixed. 152 */ 153 154 /* 155 * Allocate a PCB and associate it with the socket. 156 */ 157 int 158 in_pcballoc(so, pcbinfo, td) 159 struct socket *so; 160 struct inpcbinfo *pcbinfo; 161 struct thread *td; 162 { 163 register struct inpcb *inp; 164 #if defined(IPSEC) || defined(FAST_IPSEC) 165 int error; 166 #endif 167 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT | M_ZERO); 168 if (inp == NULL) 169 return (ENOBUFS); 170 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 171 inp->inp_pcbinfo = pcbinfo; 172 inp->inp_socket = so; 173 #if defined(IPSEC) || defined(FAST_IPSEC) 174 #ifdef FAST_IPSEC 175 error = ipsec_init_policy(so, &inp->inp_sp); 176 #else 177 error = ipsec_init_pcbpolicy(so, &inp->inp_sp); 178 #endif 179 if (error != 0) { 180 uma_zfree(pcbinfo->ipi_zone, inp); 181 return error; 182 } 183 #endif /*IPSEC*/ 184 #if defined(INET6) 185 if (INP_SOCKAF(so) == AF_INET6) { 186 inp->inp_vflag |= INP_IPV6PROTO; 187 if (ip6_v6only) 188 inp->inp_flags |= IN6P_IPV6_V6ONLY; 189 } 190 #endif 191 LIST_INSERT_HEAD(pcbinfo->listhead, inp, inp_list); 192 pcbinfo->ipi_count++; 193 so->so_pcb = (caddr_t)inp; 194 INP_LOCK_INIT(inp, "inp"); 195 #ifdef INET6 196 if (ip6_auto_flowlabel) 197 inp->inp_flags |= IN6P_AUTOFLOWLABEL; 198 #endif 199 return (0); 200 } 201 202 int 203 in_pcbbind(inp, nam, td) 204 register struct inpcb *inp; 205 struct sockaddr *nam; 206 struct thread *td; 207 { 208 int anonport, error; 209 210 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) 211 return (EINVAL); 212 anonport = inp->inp_lport == 0 && (nam == NULL || 213 ((struct sockaddr_in *)nam)->sin_port == 0); 214 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, 215 &inp->inp_lport, td); 216 if (error) 217 return (error); 218 if (in_pcbinshash(inp) != 0) { 219 inp->inp_laddr.s_addr = INADDR_ANY; 220 inp->inp_lport = 0; 221 return (EAGAIN); 222 } 223 if (anonport) 224 inp->inp_flags |= INP_ANONPORT; 225 return (0); 226 } 227 228 /* 229 * Set up a bind operation on a PCB, performing port allocation 230 * as required, but do not actually modify the PCB. Callers can 231 * either complete the bind by setting inp_laddr/inp_lport and 232 * calling in_pcbinshash(), or they can just use the resulting 233 * port and address to authorise the sending of a once-off packet. 234 * 235 * On error, the values of *laddrp and *lportp are not changed. 236 */ 237 int 238 in_pcbbind_setup(inp, nam, laddrp, lportp, td) 239 struct inpcb *inp; 240 struct sockaddr *nam; 241 in_addr_t *laddrp; 242 u_short *lportp; 243 struct thread *td; 244 { 245 struct socket *so = inp->inp_socket; 246 unsigned short *lastport; 247 struct sockaddr_in *sin; 248 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 249 struct in_addr laddr; 250 u_short lport = 0; 251 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT); 252 int error, prison = 0; 253 254 if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */ 255 return (EADDRNOTAVAIL); 256 laddr.s_addr = *laddrp; 257 if (nam != NULL && laddr.s_addr != INADDR_ANY) 258 return (EINVAL); 259 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 260 wild = 1; 261 if (nam) { 262 sin = (struct sockaddr_in *)nam; 263 if (nam->sa_len != sizeof (*sin)) 264 return (EINVAL); 265 #ifdef notdef 266 /* 267 * We should check the family, but old programs 268 * incorrectly fail to initialize it. 269 */ 270 if (sin->sin_family != AF_INET) 271 return (EAFNOSUPPORT); 272 #endif 273 if (sin->sin_addr.s_addr != INADDR_ANY) 274 if (prison_ip(td->td_ucred, 0, &sin->sin_addr.s_addr)) 275 return(EINVAL); 276 if (sin->sin_port != *lportp) { 277 /* Don't allow the port to change. */ 278 if (*lportp != 0) 279 return (EINVAL); 280 lport = sin->sin_port; 281 } 282 /* NB: lport is left as 0 if the port isn't being changed. */ 283 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 284 /* 285 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 286 * allow complete duplication of binding if 287 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 288 * and a multicast address is bound on both 289 * new and duplicated sockets. 290 */ 291 if (so->so_options & SO_REUSEADDR) 292 reuseport = SO_REUSEADDR|SO_REUSEPORT; 293 } else if (sin->sin_addr.s_addr != INADDR_ANY) { 294 sin->sin_port = 0; /* yech... */ 295 bzero(&sin->sin_zero, sizeof(sin->sin_zero)); 296 if (ifa_ifwithaddr((struct sockaddr *)sin) == 0) 297 return (EADDRNOTAVAIL); 298 } 299 laddr = sin->sin_addr; 300 if (lport) { 301 struct inpcb *t; 302 /* GROSS */ 303 if (ntohs(lport) <= ipport_reservedhigh && 304 ntohs(lport) >= ipport_reservedlow && 305 td && suser_cred(td->td_ucred, PRISON_ROOT)) 306 return (EACCES); 307 if (td && jailed(td->td_ucred)) 308 prison = 1; 309 if (so->so_cred->cr_uid != 0 && 310 !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 311 t = in_pcblookup_local(inp->inp_pcbinfo, 312 sin->sin_addr, lport, 313 prison ? 0 : INPLOOKUP_WILDCARD); 314 /* 315 * XXX 316 * This entire block sorely needs a rewrite. 317 */ 318 if (t && (t->inp_vflag & INP_TIMEWAIT)) { 319 if ((ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 320 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 321 (intotw(t)->tw_so_options & SO_REUSEPORT) == 0) && 322 (so->so_cred->cr_uid != intotw(t)->tw_cred->cr_uid)) 323 return (EADDRINUSE); 324 } else 325 if (t && 326 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 327 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 328 (t->inp_socket->so_options & 329 SO_REUSEPORT) == 0) && 330 (so->so_cred->cr_uid != 331 t->inp_socket->so_cred->cr_uid)) { 332 #if defined(INET6) 333 if (ntohl(sin->sin_addr.s_addr) != 334 INADDR_ANY || 335 ntohl(t->inp_laddr.s_addr) != 336 INADDR_ANY || 337 INP_SOCKAF(so) == 338 INP_SOCKAF(t->inp_socket)) 339 #endif /* defined(INET6) */ 340 return (EADDRINUSE); 341 } 342 } 343 if (prison && 344 prison_ip(td->td_ucred, 0, &sin->sin_addr.s_addr)) 345 return (EADDRNOTAVAIL); 346 t = in_pcblookup_local(pcbinfo, sin->sin_addr, 347 lport, prison ? 0 : wild); 348 if (t && (t->inp_vflag & INP_TIMEWAIT)) { 349 if ((reuseport & intotw(t)->tw_so_options) == 0) 350 return (EADDRINUSE); 351 } else 352 if (t && 353 (reuseport & t->inp_socket->so_options) == 0) { 354 #if defined(INET6) 355 if (ntohl(sin->sin_addr.s_addr) != 356 INADDR_ANY || 357 ntohl(t->inp_laddr.s_addr) != 358 INADDR_ANY || 359 INP_SOCKAF(so) == 360 INP_SOCKAF(t->inp_socket)) 361 #endif /* defined(INET6) */ 362 return (EADDRINUSE); 363 } 364 } 365 } 366 if (*lportp != 0) 367 lport = *lportp; 368 if (lport == 0) { 369 u_short first, last; 370 int count; 371 372 if (laddr.s_addr != INADDR_ANY) 373 if (prison_ip(td->td_ucred, 0, &laddr.s_addr)) 374 return (EINVAL); 375 376 if (inp->inp_flags & INP_HIGHPORT) { 377 first = ipport_hifirstauto; /* sysctl */ 378 last = ipport_hilastauto; 379 lastport = &pcbinfo->lasthi; 380 } else if (inp->inp_flags & INP_LOWPORT) { 381 if (td && (error = suser_cred(td->td_ucred, 382 PRISON_ROOT)) != 0) 383 return error; 384 first = ipport_lowfirstauto; /* 1023 */ 385 last = ipport_lowlastauto; /* 600 */ 386 lastport = &pcbinfo->lastlow; 387 } else { 388 first = ipport_firstauto; /* sysctl */ 389 last = ipport_lastauto; 390 lastport = &pcbinfo->lastport; 391 } 392 /* 393 * Simple check to ensure all ports are not used up causing 394 * a deadlock here. 395 * 396 * We split the two cases (up and down) so that the direction 397 * is not being tested on each round of the loop. 398 */ 399 if (first > last) { 400 /* 401 * counting down 402 */ 403 count = first - last; 404 405 do { 406 if (count-- < 0) /* completely used? */ 407 return (EADDRNOTAVAIL); 408 --*lastport; 409 if (*lastport > first || *lastport < last) 410 *lastport = first; 411 lport = htons(*lastport); 412 } while (in_pcblookup_local(pcbinfo, laddr, lport, 413 wild)); 414 } else { 415 /* 416 * counting up 417 */ 418 count = last - first; 419 420 do { 421 if (count-- < 0) /* completely used? */ 422 return (EADDRNOTAVAIL); 423 ++*lastport; 424 if (*lastport < first || *lastport > last) 425 *lastport = first; 426 lport = htons(*lastport); 427 } while (in_pcblookup_local(pcbinfo, laddr, lport, 428 wild)); 429 } 430 } 431 if (prison_ip(td->td_ucred, 0, &laddr.s_addr)) 432 return (EINVAL); 433 *laddrp = laddr.s_addr; 434 *lportp = lport; 435 return (0); 436 } 437 438 /* 439 * Connect from a socket to a specified address. 440 * Both address and port must be specified in argument sin. 441 * If don't have a local address for this socket yet, 442 * then pick one. 443 */ 444 int 445 in_pcbconnect(inp, nam, td) 446 register struct inpcb *inp; 447 struct sockaddr *nam; 448 struct thread *td; 449 { 450 u_short lport, fport; 451 in_addr_t laddr, faddr; 452 int anonport, error; 453 454 lport = inp->inp_lport; 455 laddr = inp->inp_laddr.s_addr; 456 anonport = (lport == 0); 457 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, 458 NULL, td); 459 if (error) 460 return (error); 461 462 /* Do the initial binding of the local address if required. */ 463 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { 464 inp->inp_lport = lport; 465 inp->inp_laddr.s_addr = laddr; 466 if (in_pcbinshash(inp) != 0) { 467 inp->inp_laddr.s_addr = INADDR_ANY; 468 inp->inp_lport = 0; 469 return (EAGAIN); 470 } 471 } 472 473 /* Commit the remaining changes. */ 474 inp->inp_lport = lport; 475 inp->inp_laddr.s_addr = laddr; 476 inp->inp_faddr.s_addr = faddr; 477 inp->inp_fport = fport; 478 in_pcbrehash(inp); 479 #ifdef IPSEC 480 if (inp->inp_socket->so_type == SOCK_STREAM) 481 ipsec_pcbconn(inp->inp_sp); 482 #endif 483 if (anonport) 484 inp->inp_flags |= INP_ANONPORT; 485 return (0); 486 } 487 488 /* 489 * Set up for a connect from a socket to the specified address. 490 * On entry, *laddrp and *lportp should contain the current local 491 * address and port for the PCB; these are updated to the values 492 * that should be placed in inp_laddr and inp_lport to complete 493 * the connect. 494 * 495 * On success, *faddrp and *fportp will be set to the remote address 496 * and port. These are not updated in the error case. 497 * 498 * If the operation fails because the connection already exists, 499 * *oinpp will be set to the PCB of that connection so that the 500 * caller can decide to override it. In all other cases, *oinpp 501 * is set to NULL. 502 */ 503 int 504 in_pcbconnect_setup(inp, nam, laddrp, lportp, faddrp, fportp, oinpp, td) 505 register struct inpcb *inp; 506 struct sockaddr *nam; 507 in_addr_t *laddrp; 508 u_short *lportp; 509 in_addr_t *faddrp; 510 u_short *fportp; 511 struct inpcb **oinpp; 512 struct thread *td; 513 { 514 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 515 struct in_ifaddr *ia; 516 struct sockaddr_in sa; 517 struct ucred *cred; 518 struct inpcb *oinp; 519 struct in_addr laddr, faddr; 520 u_short lport, fport; 521 int error; 522 523 if (oinpp != NULL) 524 *oinpp = NULL; 525 if (nam->sa_len != sizeof (*sin)) 526 return (EINVAL); 527 if (sin->sin_family != AF_INET) 528 return (EAFNOSUPPORT); 529 if (sin->sin_port == 0) 530 return (EADDRNOTAVAIL); 531 laddr.s_addr = *laddrp; 532 lport = *lportp; 533 faddr = sin->sin_addr; 534 fport = sin->sin_port; 535 cred = inp->inp_socket->so_cred; 536 if (laddr.s_addr == INADDR_ANY && jailed(cred)) { 537 bzero(&sa, sizeof(sa)); 538 sa.sin_addr.s_addr = htonl(prison_getip(cred)); 539 sa.sin_len = sizeof(sa); 540 sa.sin_family = AF_INET; 541 error = in_pcbbind_setup(inp, (struct sockaddr *)&sa, 542 &laddr.s_addr, &lport, td); 543 if (error) 544 return (error); 545 } 546 547 if (!TAILQ_EMPTY(&in_ifaddrhead)) { 548 /* 549 * If the destination address is INADDR_ANY, 550 * use the primary local address. 551 * If the supplied address is INADDR_BROADCAST, 552 * and the primary interface supports broadcast, 553 * choose the broadcast address for that interface. 554 */ 555 if (faddr.s_addr == INADDR_ANY) 556 faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr; 557 else if (faddr.s_addr == (u_long)INADDR_BROADCAST && 558 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags & 559 IFF_BROADCAST)) 560 faddr = satosin(&TAILQ_FIRST( 561 &in_ifaddrhead)->ia_broadaddr)->sin_addr; 562 } 563 if (laddr.s_addr == INADDR_ANY) { 564 register struct route *ro; 565 566 ia = (struct in_ifaddr *)0; 567 /* 568 * If route is known or can be allocated now, 569 * our src addr is taken from the i/f, else punt. 570 * Note that we should check the address family of the cached 571 * destination, in case of sharing the cache with IPv6. 572 */ 573 ro = &inp->inp_route; 574 if (ro->ro_rt && 575 (ro->ro_dst.sa_family != AF_INET || 576 satosin(&ro->ro_dst)->sin_addr.s_addr != faddr.s_addr || 577 inp->inp_socket->so_options & SO_DONTROUTE)) { 578 RTFREE(ro->ro_rt); 579 ro->ro_rt = (struct rtentry *)0; 580 } 581 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/ 582 (ro->ro_rt == (struct rtentry *)0 || 583 ro->ro_rt->rt_ifp == (struct ifnet *)0)) { 584 /* No route yet, so try to acquire one */ 585 bzero(&ro->ro_dst, sizeof(struct sockaddr_in)); 586 ro->ro_dst.sa_family = AF_INET; 587 ro->ro_dst.sa_len = sizeof(struct sockaddr_in); 588 ((struct sockaddr_in *)&ro->ro_dst)->sin_addr = faddr; 589 rtalloc(ro); 590 } 591 /* 592 * If we found a route, use the address 593 * corresponding to the outgoing interface 594 * unless it is the loopback (in case a route 595 * to our address on another net goes to loopback). 596 */ 597 if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) 598 ia = ifatoia(ro->ro_rt->rt_ifa); 599 if (ia == 0) { 600 bzero(&sa, sizeof(sa)); 601 sa.sin_addr = faddr; 602 sa.sin_len = sizeof(sa); 603 sa.sin_family = AF_INET; 604 605 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa))); 606 if (ia == 0) 607 ia = ifatoia(ifa_ifwithnet(sintosa(&sa))); 608 if (ia == 0) 609 ia = TAILQ_FIRST(&in_ifaddrhead); 610 if (ia == 0) 611 return (EADDRNOTAVAIL); 612 } 613 /* 614 * If the destination address is multicast and an outgoing 615 * interface has been set as a multicast option, use the 616 * address of that interface as our source address. 617 */ 618 if (IN_MULTICAST(ntohl(faddr.s_addr)) && 619 inp->inp_moptions != NULL) { 620 struct ip_moptions *imo; 621 struct ifnet *ifp; 622 623 imo = inp->inp_moptions; 624 if (imo->imo_multicast_ifp != NULL) { 625 ifp = imo->imo_multicast_ifp; 626 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) 627 if (ia->ia_ifp == ifp) 628 break; 629 if (ia == 0) 630 return (EADDRNOTAVAIL); 631 } 632 } 633 laddr = ia->ia_addr.sin_addr; 634 } 635 636 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport, 637 0, NULL); 638 if (oinp != NULL) { 639 if (oinpp != NULL) 640 *oinpp = oinp; 641 return (EADDRINUSE); 642 } 643 if (lport == 0) { 644 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, td); 645 if (error) 646 return (error); 647 } 648 *laddrp = laddr.s_addr; 649 *lportp = lport; 650 *faddrp = faddr.s_addr; 651 *fportp = fport; 652 return (0); 653 } 654 655 void 656 in_pcbdisconnect(inp) 657 struct inpcb *inp; 658 { 659 660 inp->inp_faddr.s_addr = INADDR_ANY; 661 inp->inp_fport = 0; 662 in_pcbrehash(inp); 663 if (inp->inp_socket->so_state & SS_NOFDREF) 664 in_pcbdetach(inp); 665 #ifdef IPSEC 666 ipsec_pcbdisconn(inp->inp_sp); 667 #endif 668 } 669 670 void 671 in_pcbdetach(inp) 672 struct inpcb *inp; 673 { 674 struct socket *so = inp->inp_socket; 675 struct inpcbinfo *ipi = inp->inp_pcbinfo; 676 677 #if defined(IPSEC) || defined(FAST_IPSEC) 678 ipsec4_delete_pcbpolicy(inp); 679 #endif /*IPSEC*/ 680 inp->inp_gencnt = ++ipi->ipi_gencnt; 681 in_pcbremlists(inp); 682 if (so) { 683 so->so_pcb = 0; 684 sotryfree(so); 685 } 686 if (inp->inp_options) 687 (void)m_free(inp->inp_options); 688 if (inp->inp_route.ro_rt) 689 RTFREE(inp->inp_route.ro_rt); 690 ip_freemoptions(inp->inp_moptions); 691 inp->inp_vflag = 0; 692 INP_LOCK_DESTROY(inp); 693 uma_zfree(ipi->ipi_zone, inp); 694 } 695 696 struct sockaddr * 697 in_sockaddr(port, addr_p) 698 in_port_t port; 699 struct in_addr *addr_p; 700 { 701 struct sockaddr_in *sin; 702 703 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, 704 M_WAITOK | M_ZERO); 705 sin->sin_family = AF_INET; 706 sin->sin_len = sizeof(*sin); 707 sin->sin_addr = *addr_p; 708 sin->sin_port = port; 709 710 return (struct sockaddr *)sin; 711 } 712 713 /* 714 * The wrapper function will pass down the pcbinfo for this function to lock. 715 * The socket must have a valid 716 * (i.e., non-nil) PCB, but it should be impossible to get an invalid one 717 * except through a kernel programming error, so it is acceptable to panic 718 * (or in this case trap) if the PCB is invalid. (Actually, we don't trap 719 * because there actually /is/ a programming error somewhere... XXX) 720 */ 721 int 722 in_setsockaddr(so, nam, pcbinfo) 723 struct socket *so; 724 struct sockaddr **nam; 725 struct inpcbinfo *pcbinfo; 726 { 727 int s; 728 register struct inpcb *inp; 729 struct in_addr addr; 730 in_port_t port; 731 732 s = splnet(); 733 INP_INFO_RLOCK(pcbinfo); 734 inp = sotoinpcb(so); 735 if (!inp) { 736 INP_INFO_RUNLOCK(pcbinfo); 737 splx(s); 738 return ECONNRESET; 739 } 740 INP_LOCK(inp); 741 port = inp->inp_lport; 742 addr = inp->inp_laddr; 743 INP_UNLOCK(inp); 744 INP_INFO_RUNLOCK(pcbinfo); 745 splx(s); 746 747 *nam = in_sockaddr(port, &addr); 748 return 0; 749 } 750 751 /* 752 * The wrapper function will pass down the pcbinfo for this function to lock. 753 */ 754 int 755 in_setpeeraddr(so, nam, pcbinfo) 756 struct socket *so; 757 struct sockaddr **nam; 758 struct inpcbinfo *pcbinfo; 759 { 760 int s; 761 register struct inpcb *inp; 762 struct in_addr addr; 763 in_port_t port; 764 765 s = splnet(); 766 INP_INFO_RLOCK(pcbinfo); 767 inp = sotoinpcb(so); 768 if (!inp) { 769 INP_INFO_RUNLOCK(pcbinfo); 770 splx(s); 771 return ECONNRESET; 772 } 773 INP_LOCK(inp); 774 port = inp->inp_fport; 775 addr = inp->inp_faddr; 776 INP_UNLOCK(inp); 777 INP_INFO_RUNLOCK(pcbinfo); 778 splx(s); 779 780 *nam = in_sockaddr(port, &addr); 781 return 0; 782 } 783 784 void 785 in_pcbnotifyall(pcbinfo, faddr, errno, notify) 786 struct inpcbinfo *pcbinfo; 787 struct in_addr faddr; 788 int errno; 789 struct inpcb *(*notify)(struct inpcb *, int); 790 { 791 struct inpcb *inp, *ninp; 792 struct inpcbhead *head; 793 int s; 794 795 s = splnet(); 796 INP_INFO_WLOCK(pcbinfo); 797 head = pcbinfo->listhead; 798 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) { 799 INP_LOCK(inp); 800 ninp = LIST_NEXT(inp, inp_list); 801 #ifdef INET6 802 if ((inp->inp_vflag & INP_IPV4) == 0) { 803 INP_UNLOCK(inp); 804 continue; 805 } 806 #endif 807 if (inp->inp_faddr.s_addr != faddr.s_addr || 808 inp->inp_socket == NULL) { 809 INP_UNLOCK(inp); 810 continue; 811 } 812 if ((*notify)(inp, errno)) 813 INP_UNLOCK(inp); 814 } 815 INP_INFO_WUNLOCK(pcbinfo); 816 splx(s); 817 } 818 819 void 820 in_pcbpurgeif0(pcbinfo, ifp) 821 struct inpcbinfo *pcbinfo; 822 struct ifnet *ifp; 823 { 824 struct inpcb *inp; 825 struct ip_moptions *imo; 826 int i, gap; 827 828 /* why no splnet here? XXX */ 829 INP_INFO_RLOCK(pcbinfo); 830 LIST_FOREACH(inp, pcbinfo->listhead, inp_list) { 831 INP_LOCK(inp); 832 imo = inp->inp_moptions; 833 if ((inp->inp_vflag & INP_IPV4) && 834 imo != NULL) { 835 /* 836 * Unselect the outgoing interface if it is being 837 * detached. 838 */ 839 if (imo->imo_multicast_ifp == ifp) 840 imo->imo_multicast_ifp = NULL; 841 842 /* 843 * Drop multicast group membership if we joined 844 * through the interface being detached. 845 */ 846 for (i = 0, gap = 0; i < imo->imo_num_memberships; 847 i++) { 848 if (imo->imo_membership[i]->inm_ifp == ifp) { 849 in_delmulti(imo->imo_membership[i]); 850 gap++; 851 } else if (gap != 0) 852 imo->imo_membership[i - gap] = 853 imo->imo_membership[i]; 854 } 855 imo->imo_num_memberships -= gap; 856 } 857 INP_UNLOCK(inp); 858 } 859 INP_INFO_RUNLOCK(pcbinfo); 860 } 861 862 /* 863 * Check for alternatives when higher level complains 864 * about service problems. For now, invalidate cached 865 * routing information. If the route was created dynamically 866 * (by a redirect), time to try a default gateway again. 867 */ 868 void 869 in_losing(inp) 870 struct inpcb *inp; 871 { 872 register struct rtentry *rt; 873 struct rt_addrinfo info; 874 875 if ((rt = inp->inp_route.ro_rt)) { 876 RT_LOCK(rt); 877 inp->inp_route.ro_rt = NULL; 878 bzero((caddr_t)&info, sizeof(info)); 879 info.rti_flags = rt->rt_flags; 880 info.rti_info[RTAX_DST] = rt_key(rt); 881 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 882 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 883 rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0); 884 if (rt->rt_flags & RTF_DYNAMIC) 885 rtexpunge(rt); 886 RTFREE_LOCKED(rt); 887 /* 888 * A new route can be allocated 889 * the next time output is attempted. 890 */ 891 } 892 } 893 894 /* 895 * After a routing change, flush old routing 896 * and allocate a (hopefully) better one. 897 */ 898 struct inpcb * 899 in_rtchange(inp, errno) 900 register struct inpcb *inp; 901 int errno; 902 { 903 if (inp->inp_route.ro_rt) { 904 RTFREE(inp->inp_route.ro_rt); 905 inp->inp_route.ro_rt = 0; 906 /* 907 * A new route can be allocated the next time 908 * output is attempted. 909 */ 910 } 911 return inp; 912 } 913 914 /* 915 * Lookup a PCB based on the local address and port. 916 */ 917 struct inpcb * 918 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay) 919 struct inpcbinfo *pcbinfo; 920 struct in_addr laddr; 921 u_int lport_arg; 922 int wild_okay; 923 { 924 register struct inpcb *inp; 925 int matchwild = 3, wildcard; 926 u_short lport = lport_arg; 927 928 if (!wild_okay) { 929 struct inpcbhead *head; 930 /* 931 * Look for an unconnected (wildcard foreign addr) PCB that 932 * matches the local address and port we're looking for. 933 */ 934 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 935 LIST_FOREACH(inp, head, inp_hash) { 936 #ifdef INET6 937 if ((inp->inp_vflag & INP_IPV4) == 0) 938 continue; 939 #endif 940 if (inp->inp_faddr.s_addr == INADDR_ANY && 941 inp->inp_laddr.s_addr == laddr.s_addr && 942 inp->inp_lport == lport) { 943 /* 944 * Found. 945 */ 946 return (inp); 947 } 948 } 949 /* 950 * Not found. 951 */ 952 return (NULL); 953 } else { 954 struct inpcbporthead *porthash; 955 struct inpcbport *phd; 956 struct inpcb *match = NULL; 957 /* 958 * Best fit PCB lookup. 959 * 960 * First see if this local port is in use by looking on the 961 * port hash list. 962 */ 963 retrylookup: 964 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport, 965 pcbinfo->porthashmask)]; 966 LIST_FOREACH(phd, porthash, phd_hash) { 967 if (phd->phd_port == lport) 968 break; 969 } 970 if (phd != NULL) { 971 /* 972 * Port is in use by one or more PCBs. Look for best 973 * fit. 974 */ 975 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 976 wildcard = 0; 977 #ifdef INET6 978 if ((inp->inp_vflag & INP_IPV4) == 0) 979 continue; 980 #endif 981 /* 982 * Clean out old time_wait sockets if they 983 * are clogging up needed local ports. 984 */ 985 if ((inp->inp_vflag & INP_TIMEWAIT) != 0) { 986 if (tcp_twrecycleable((struct tcptw *)inp->inp_ppcb)) { 987 tcp_twclose((struct tcptw *)inp->inp_ppcb, 0); 988 match = NULL; 989 goto retrylookup; 990 } 991 } 992 if (inp->inp_faddr.s_addr != INADDR_ANY) 993 wildcard++; 994 if (inp->inp_laddr.s_addr != INADDR_ANY) { 995 if (laddr.s_addr == INADDR_ANY) 996 wildcard++; 997 else if (inp->inp_laddr.s_addr != laddr.s_addr) 998 continue; 999 } else { 1000 if (laddr.s_addr != INADDR_ANY) 1001 wildcard++; 1002 } 1003 if (wildcard < matchwild) { 1004 match = inp; 1005 matchwild = wildcard; 1006 if (matchwild == 0) { 1007 break; 1008 } 1009 } 1010 } 1011 } 1012 return (match); 1013 } 1014 } 1015 1016 /* 1017 * Lookup PCB in hash list. 1018 */ 1019 struct inpcb * 1020 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, 1021 ifp) 1022 struct inpcbinfo *pcbinfo; 1023 struct in_addr faddr, laddr; 1024 u_int fport_arg, lport_arg; 1025 int wildcard; 1026 struct ifnet *ifp; 1027 { 1028 struct inpcbhead *head; 1029 register struct inpcb *inp; 1030 u_short fport = fport_arg, lport = lport_arg; 1031 1032 /* 1033 * First look for an exact match. 1034 */ 1035 head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)]; 1036 LIST_FOREACH(inp, head, inp_hash) { 1037 #ifdef INET6 1038 if ((inp->inp_vflag & INP_IPV4) == 0) 1039 continue; 1040 #endif 1041 if (inp->inp_faddr.s_addr == faddr.s_addr && 1042 inp->inp_laddr.s_addr == laddr.s_addr && 1043 inp->inp_fport == fport && 1044 inp->inp_lport == lport) { 1045 /* 1046 * Found. 1047 */ 1048 return (inp); 1049 } 1050 } 1051 if (wildcard) { 1052 struct inpcb *local_wild = NULL; 1053 #if defined(INET6) 1054 struct inpcb *local_wild_mapped = NULL; 1055 #endif /* defined(INET6) */ 1056 1057 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 1058 LIST_FOREACH(inp, head, inp_hash) { 1059 #ifdef INET6 1060 if ((inp->inp_vflag & INP_IPV4) == 0) 1061 continue; 1062 #endif 1063 if (inp->inp_faddr.s_addr == INADDR_ANY && 1064 inp->inp_lport == lport) { 1065 if (ifp && ifp->if_type == IFT_FAITH && 1066 (inp->inp_flags & INP_FAITH) == 0) 1067 continue; 1068 if (inp->inp_laddr.s_addr == laddr.s_addr) 1069 return (inp); 1070 else if (inp->inp_laddr.s_addr == INADDR_ANY) { 1071 #if defined(INET6) 1072 if (INP_CHECK_SOCKAF(inp->inp_socket, 1073 AF_INET6)) 1074 local_wild_mapped = inp; 1075 else 1076 #endif /* defined(INET6) */ 1077 local_wild = inp; 1078 } 1079 } 1080 } 1081 #if defined(INET6) 1082 if (local_wild == NULL) 1083 return (local_wild_mapped); 1084 #endif /* defined(INET6) */ 1085 return (local_wild); 1086 } 1087 1088 /* 1089 * Not found. 1090 */ 1091 return (NULL); 1092 } 1093 1094 /* 1095 * Insert PCB onto various hash lists. 1096 */ 1097 int 1098 in_pcbinshash(inp) 1099 struct inpcb *inp; 1100 { 1101 struct inpcbhead *pcbhash; 1102 struct inpcbporthead *pcbporthash; 1103 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1104 struct inpcbport *phd; 1105 u_int32_t hashkey_faddr; 1106 1107 #ifdef INET6 1108 if (inp->inp_vflag & INP_IPV6) 1109 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1110 else 1111 #endif /* INET6 */ 1112 hashkey_faddr = inp->inp_faddr.s_addr; 1113 1114 pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1115 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 1116 1117 pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport, 1118 pcbinfo->porthashmask)]; 1119 1120 /* 1121 * Go through port list and look for a head for this lport. 1122 */ 1123 LIST_FOREACH(phd, pcbporthash, phd_hash) { 1124 if (phd->phd_port == inp->inp_lport) 1125 break; 1126 } 1127 /* 1128 * If none exists, malloc one and tack it on. 1129 */ 1130 if (phd == NULL) { 1131 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT); 1132 if (phd == NULL) { 1133 return (ENOBUFS); /* XXX */ 1134 } 1135 phd->phd_port = inp->inp_lport; 1136 LIST_INIT(&phd->phd_pcblist); 1137 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 1138 } 1139 inp->inp_phd = phd; 1140 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 1141 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 1142 return (0); 1143 } 1144 1145 /* 1146 * Move PCB to the proper hash bucket when { faddr, fport } have been 1147 * changed. NOTE: This does not handle the case of the lport changing (the 1148 * hashed port list would have to be updated as well), so the lport must 1149 * not change after in_pcbinshash() has been called. 1150 */ 1151 void 1152 in_pcbrehash(inp) 1153 struct inpcb *inp; 1154 { 1155 struct inpcbhead *head; 1156 u_int32_t hashkey_faddr; 1157 1158 #ifdef INET6 1159 if (inp->inp_vflag & INP_IPV6) 1160 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1161 else 1162 #endif /* INET6 */ 1163 hashkey_faddr = inp->inp_faddr.s_addr; 1164 1165 head = &inp->inp_pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 1166 inp->inp_lport, inp->inp_fport, inp->inp_pcbinfo->hashmask)]; 1167 1168 LIST_REMOVE(inp, inp_hash); 1169 LIST_INSERT_HEAD(head, inp, inp_hash); 1170 } 1171 1172 /* 1173 * Remove PCB from various lists. 1174 */ 1175 void 1176 in_pcbremlists(inp) 1177 struct inpcb *inp; 1178 { 1179 inp->inp_gencnt = ++inp->inp_pcbinfo->ipi_gencnt; 1180 if (inp->inp_lport) { 1181 struct inpcbport *phd = inp->inp_phd; 1182 1183 LIST_REMOVE(inp, inp_hash); 1184 LIST_REMOVE(inp, inp_portlist); 1185 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 1186 LIST_REMOVE(phd, phd_hash); 1187 free(phd, M_PCB); 1188 } 1189 } 1190 LIST_REMOVE(inp, inp_list); 1191 inp->inp_pcbinfo->ipi_count--; 1192 } 1193 1194 int 1195 prison_xinpcb(struct thread *td, struct inpcb *inp) 1196 { 1197 if (!jailed(td->td_ucred)) 1198 return (0); 1199 if (ntohl(inp->inp_laddr.s_addr) == prison_getip(td->td_ucred)) 1200 return (0); 1201 return (1); 1202 } 1203