1 /*- 2 * Copyright (c) 1982, 1986, 1991, 1993, 1995 3 * The Regents of the University of California. 4 * Copyright (c) 2007 Robert N. M. Watson 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 4. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_ddb.h" 38 #include "opt_ipsec.h" 39 #include "opt_inet6.h" 40 #include "opt_mac.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/domain.h> 47 #include <sys/protosw.h> 48 #include <sys/socket.h> 49 #include <sys/socketvar.h> 50 #include <sys/priv.h> 51 #include <sys/proc.h> 52 #include <sys/jail.h> 53 #include <sys/kernel.h> 54 #include <sys/sysctl.h> 55 #include <sys/vimage.h> 56 57 #ifdef DDB 58 #include <ddb/ddb.h> 59 #endif 60 61 #include <vm/uma.h> 62 63 #include <net/if.h> 64 #include <net/if_types.h> 65 #include <net/route.h> 66 67 #include <netinet/in.h> 68 #include <netinet/in_pcb.h> 69 #include <netinet/in_var.h> 70 #include <netinet/ip_var.h> 71 #include <netinet/tcp_var.h> 72 #include <netinet/udp.h> 73 #include <netinet/udp_var.h> 74 #include <netinet/vinet.h> 75 #ifdef INET6 76 #include <netinet/ip6.h> 77 #include <netinet6/ip6_var.h> 78 #include <netinet6/vinet6.h> 79 #endif /* INET6 */ 80 81 82 #ifdef IPSEC 83 #include <netipsec/ipsec.h> 84 #include <netipsec/key.h> 85 #endif /* IPSEC */ 86 87 #include <security/mac/mac_framework.h> 88 89 #ifdef VIMAGE_GLOBALS 90 /* 91 * These configure the range of local port addresses assigned to 92 * "unspecified" outgoing connections/packets/whatever. 93 */ 94 int ipport_lowfirstauto; 95 int ipport_lowlastauto; 96 int ipport_firstauto; 97 int ipport_lastauto; 98 int ipport_hifirstauto; 99 int ipport_hilastauto; 100 101 /* 102 * Reserved ports accessible only to root. There are significant 103 * security considerations that must be accounted for when changing these, 104 * but the security benefits can be great. Please be careful. 105 */ 106 int ipport_reservedhigh; 107 int ipport_reservedlow; 108 109 /* Variables dealing with random ephemeral port allocation. */ 110 int ipport_randomized; 111 int ipport_randomcps; 112 int ipport_randomtime; 113 int ipport_stoprandom; 114 int ipport_tcpallocs; 115 int ipport_tcplastcount; 116 #endif 117 118 #define RANGECHK(var, min, max) \ 119 if ((var) < (min)) { (var) = (min); } \ 120 else if ((var) > (max)) { (var) = (max); } 121 122 static int 123 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) 124 { 125 INIT_VNET_INET(curvnet); 126 int error; 127 128 error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req); 129 if (error == 0) { 130 RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); 131 RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1); 132 RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); 133 RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); 134 RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); 135 RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); 136 } 137 return (error); 138 } 139 140 #undef RANGECHK 141 142 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); 143 144 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, 145 lowfirst, CTLTYPE_INT|CTLFLAG_RW, ipport_lowfirstauto, 0, 146 &sysctl_net_ipport_check, "I", ""); 147 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, 148 lowlast, CTLTYPE_INT|CTLFLAG_RW, ipport_lowlastauto, 0, 149 &sysctl_net_ipport_check, "I", ""); 150 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, 151 first, CTLTYPE_INT|CTLFLAG_RW, ipport_firstauto, 0, 152 &sysctl_net_ipport_check, "I", ""); 153 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, 154 last, CTLTYPE_INT|CTLFLAG_RW, ipport_lastauto, 0, 155 &sysctl_net_ipport_check, "I", ""); 156 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, 157 hifirst, CTLTYPE_INT|CTLFLAG_RW, ipport_hifirstauto, 0, 158 &sysctl_net_ipport_check, "I", ""); 159 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, 160 hilast, CTLTYPE_INT|CTLFLAG_RW, ipport_hilastauto, 0, 161 &sysctl_net_ipport_check, "I", ""); 162 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, 163 reservedhigh, CTLFLAG_RW|CTLFLAG_SECURE, ipport_reservedhigh, 0, ""); 164 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, reservedlow, 165 CTLFLAG_RW|CTLFLAG_SECURE, ipport_reservedlow, 0, ""); 166 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, randomized, 167 CTLFLAG_RW, ipport_randomized, 0, "Enable random port allocation"); 168 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, randomcps, 169 CTLFLAG_RW, ipport_randomcps, 0, "Maximum number of random port " 170 "allocations before switching to a sequental one"); 171 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, randomtime, 172 CTLFLAG_RW, ipport_randomtime, 0, 173 "Minimum time to keep sequental port " 174 "allocation before switching to a random one"); 175 176 /* 177 * in_pcb.c: manage the Protocol Control Blocks. 178 * 179 * NOTE: It is assumed that most of these functions will be called with 180 * the pcbinfo lock held, and often, the inpcb lock held, as these utility 181 * functions often modify hash chains or addresses in pcbs. 182 */ 183 184 /* 185 * Allocate a PCB and associate it with the socket. 186 * On success return with the PCB locked. 187 */ 188 int 189 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo) 190 { 191 #ifdef INET6 192 INIT_VNET_INET6(curvnet); 193 #endif 194 struct inpcb *inp; 195 int error; 196 197 INP_INFO_WLOCK_ASSERT(pcbinfo); 198 error = 0; 199 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT); 200 if (inp == NULL) 201 return (ENOBUFS); 202 bzero(inp, inp_zero_size); 203 inp->inp_pcbinfo = pcbinfo; 204 inp->inp_socket = so; 205 inp->inp_cred = crhold(so->so_cred); 206 inp->inp_inc.inc_fibnum = so->so_fibnum; 207 #ifdef MAC 208 error = mac_inpcb_init(inp, M_NOWAIT); 209 if (error != 0) 210 goto out; 211 SOCK_LOCK(so); 212 mac_inpcb_create(so, inp); 213 SOCK_UNLOCK(so); 214 #endif 215 216 #ifdef IPSEC 217 error = ipsec_init_policy(so, &inp->inp_sp); 218 if (error != 0) { 219 #ifdef MAC 220 mac_inpcb_destroy(inp); 221 #endif 222 goto out; 223 } 224 #endif /*IPSEC*/ 225 #ifdef INET6 226 if (INP_SOCKAF(so) == AF_INET6) { 227 inp->inp_vflag |= INP_IPV6PROTO; 228 if (V_ip6_v6only) 229 inp->inp_flags |= IN6P_IPV6_V6ONLY; 230 } 231 #endif 232 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list); 233 pcbinfo->ipi_count++; 234 so->so_pcb = (caddr_t)inp; 235 #ifdef INET6 236 if (V_ip6_auto_flowlabel) 237 inp->inp_flags |= IN6P_AUTOFLOWLABEL; 238 #endif 239 INP_WLOCK(inp); 240 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 241 242 #if defined(IPSEC) || defined(MAC) 243 out: 244 if (error != 0) { 245 crfree(inp->inp_cred); 246 uma_zfree(pcbinfo->ipi_zone, inp); 247 } 248 #endif 249 return (error); 250 } 251 252 int 253 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 254 { 255 int anonport, error; 256 257 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 258 INP_WLOCK_ASSERT(inp); 259 260 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) 261 return (EINVAL); 262 anonport = inp->inp_lport == 0 && (nam == NULL || 263 ((struct sockaddr_in *)nam)->sin_port == 0); 264 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, 265 &inp->inp_lport, cred); 266 if (error) 267 return (error); 268 if (in_pcbinshash(inp) != 0) { 269 inp->inp_laddr.s_addr = INADDR_ANY; 270 inp->inp_lport = 0; 271 return (EAGAIN); 272 } 273 if (anonport) 274 inp->inp_flags |= INP_ANONPORT; 275 return (0); 276 } 277 278 /* 279 * Set up a bind operation on a PCB, performing port allocation 280 * as required, but do not actually modify the PCB. Callers can 281 * either complete the bind by setting inp_laddr/inp_lport and 282 * calling in_pcbinshash(), or they can just use the resulting 283 * port and address to authorise the sending of a once-off packet. 284 * 285 * On error, the values of *laddrp and *lportp are not changed. 286 */ 287 int 288 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, 289 u_short *lportp, struct ucred *cred) 290 { 291 INIT_VNET_INET(inp->inp_vnet); 292 struct socket *so = inp->inp_socket; 293 unsigned short *lastport; 294 struct sockaddr_in *sin; 295 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 296 struct in_addr laddr; 297 u_short lport = 0; 298 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT); 299 int error; 300 int dorandom; 301 302 /* 303 * Because no actual state changes occur here, a global write lock on 304 * the pcbinfo isn't required. 305 */ 306 INP_INFO_LOCK_ASSERT(pcbinfo); 307 INP_LOCK_ASSERT(inp); 308 309 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */ 310 return (EADDRNOTAVAIL); 311 laddr.s_addr = *laddrp; 312 if (nam != NULL && laddr.s_addr != INADDR_ANY) 313 return (EINVAL); 314 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 315 wild = INPLOOKUP_WILDCARD; 316 if (nam) { 317 sin = (struct sockaddr_in *)nam; 318 if (nam->sa_len != sizeof (*sin)) 319 return (EINVAL); 320 #ifdef notdef 321 /* 322 * We should check the family, but old programs 323 * incorrectly fail to initialize it. 324 */ 325 if (sin->sin_family != AF_INET) 326 return (EAFNOSUPPORT); 327 #endif 328 if (prison_local_ip4(cred, &sin->sin_addr)) 329 return (EINVAL); 330 if (sin->sin_port != *lportp) { 331 /* Don't allow the port to change. */ 332 if (*lportp != 0) 333 return (EINVAL); 334 lport = sin->sin_port; 335 } 336 /* NB: lport is left as 0 if the port isn't being changed. */ 337 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 338 /* 339 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 340 * allow complete duplication of binding if 341 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 342 * and a multicast address is bound on both 343 * new and duplicated sockets. 344 */ 345 if (so->so_options & SO_REUSEADDR) 346 reuseport = SO_REUSEADDR|SO_REUSEPORT; 347 } else if (sin->sin_addr.s_addr != INADDR_ANY) { 348 sin->sin_port = 0; /* yech... */ 349 bzero(&sin->sin_zero, sizeof(sin->sin_zero)); 350 if (ifa_ifwithaddr((struct sockaddr *)sin) == 0) 351 return (EADDRNOTAVAIL); 352 } 353 laddr = sin->sin_addr; 354 if (lport) { 355 struct inpcb *t; 356 struct tcptw *tw; 357 358 /* GROSS */ 359 if (ntohs(lport) <= V_ipport_reservedhigh && 360 ntohs(lport) >= V_ipport_reservedlow && 361 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 362 0)) 363 return (EACCES); 364 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) && 365 priv_check_cred(inp->inp_cred, 366 PRIV_NETINET_REUSEPORT, 0) != 0) { 367 t = in_pcblookup_local(pcbinfo, sin->sin_addr, 368 lport, INPLOOKUP_WILDCARD, cred); 369 /* 370 * XXX 371 * This entire block sorely needs a rewrite. 372 */ 373 if (t && 374 ((t->inp_vflag & INP_TIMEWAIT) == 0) && 375 (so->so_type != SOCK_STREAM || 376 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && 377 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 378 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 379 (t->inp_socket->so_options & 380 SO_REUSEPORT) == 0) && 381 (inp->inp_cred->cr_uid != 382 t->inp_cred->cr_uid)) 383 return (EADDRINUSE); 384 } 385 if (prison_local_ip4(cred, &sin->sin_addr)) 386 return (EADDRNOTAVAIL); 387 t = in_pcblookup_local(pcbinfo, sin->sin_addr, 388 lport, wild, cred); 389 if (t && (t->inp_vflag & INP_TIMEWAIT)) { 390 /* 391 * XXXRW: If an incpb has had its timewait 392 * state recycled, we treat the address as 393 * being in use (for now). This is better 394 * than a panic, but not desirable. 395 */ 396 tw = intotw(inp); 397 if (tw == NULL || 398 (reuseport & tw->tw_so_options) == 0) 399 return (EADDRINUSE); 400 } else if (t && 401 (reuseport & t->inp_socket->so_options) == 0) { 402 #ifdef INET6 403 if (ntohl(sin->sin_addr.s_addr) != 404 INADDR_ANY || 405 ntohl(t->inp_laddr.s_addr) != 406 INADDR_ANY || 407 INP_SOCKAF(so) == 408 INP_SOCKAF(t->inp_socket)) 409 #endif 410 return (EADDRINUSE); 411 } 412 } 413 } 414 if (*lportp != 0) 415 lport = *lportp; 416 if (lport == 0) { 417 u_short first, last, aux; 418 int count; 419 420 if (prison_local_ip4(cred, &laddr)) 421 return (EINVAL); 422 423 if (inp->inp_flags & INP_HIGHPORT) { 424 first = V_ipport_hifirstauto; /* sysctl */ 425 last = V_ipport_hilastauto; 426 lastport = &pcbinfo->ipi_lasthi; 427 } else if (inp->inp_flags & INP_LOWPORT) { 428 error = priv_check_cred(cred, 429 PRIV_NETINET_RESERVEDPORT, 0); 430 if (error) 431 return error; 432 first = V_ipport_lowfirstauto; /* 1023 */ 433 last = V_ipport_lowlastauto; /* 600 */ 434 lastport = &pcbinfo->ipi_lastlow; 435 } else { 436 first = V_ipport_firstauto; /* sysctl */ 437 last = V_ipport_lastauto; 438 lastport = &pcbinfo->ipi_lastport; 439 } 440 /* 441 * For UDP, use random port allocation as long as the user 442 * allows it. For TCP (and as of yet unknown) connections, 443 * use random port allocation only if the user allows it AND 444 * ipport_tick() allows it. 445 */ 446 if (V_ipport_randomized && 447 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo)) 448 dorandom = 1; 449 else 450 dorandom = 0; 451 /* 452 * It makes no sense to do random port allocation if 453 * we have the only port available. 454 */ 455 if (first == last) 456 dorandom = 0; 457 /* Make sure to not include UDP packets in the count. */ 458 if (pcbinfo != &V_udbinfo) 459 V_ipport_tcpallocs++; 460 /* 461 * Instead of having two loops further down counting up or down 462 * make sure that first is always <= last and go with only one 463 * code path implementing all logic. 464 */ 465 if (first > last) { 466 aux = first; 467 first = last; 468 last = aux; 469 } 470 471 if (dorandom) 472 *lastport = first + 473 (arc4random() % (last - first)); 474 475 count = last - first; 476 477 do { 478 if (count-- < 0) /* completely used? */ 479 return (EADDRNOTAVAIL); 480 ++*lastport; 481 if (*lastport < first || *lastport > last) 482 *lastport = first; 483 lport = htons(*lastport); 484 } while (in_pcblookup_local(pcbinfo, laddr, 485 lport, wild, cred)); 486 } 487 if (prison_local_ip4(cred, &laddr)) 488 return (EINVAL); 489 *laddrp = laddr.s_addr; 490 *lportp = lport; 491 return (0); 492 } 493 494 /* 495 * Connect from a socket to a specified address. 496 * Both address and port must be specified in argument sin. 497 * If don't have a local address for this socket yet, 498 * then pick one. 499 */ 500 int 501 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 502 { 503 u_short lport, fport; 504 in_addr_t laddr, faddr; 505 int anonport, error; 506 507 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 508 INP_WLOCK_ASSERT(inp); 509 510 lport = inp->inp_lport; 511 laddr = inp->inp_laddr.s_addr; 512 anonport = (lport == 0); 513 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, 514 NULL, cred); 515 if (error) 516 return (error); 517 518 /* Do the initial binding of the local address if required. */ 519 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { 520 inp->inp_lport = lport; 521 inp->inp_laddr.s_addr = laddr; 522 if (in_pcbinshash(inp) != 0) { 523 inp->inp_laddr.s_addr = INADDR_ANY; 524 inp->inp_lport = 0; 525 return (EAGAIN); 526 } 527 } 528 529 /* Commit the remaining changes. */ 530 inp->inp_lport = lport; 531 inp->inp_laddr.s_addr = laddr; 532 inp->inp_faddr.s_addr = faddr; 533 inp->inp_fport = fport; 534 in_pcbrehash(inp); 535 536 if (anonport) 537 inp->inp_flags |= INP_ANONPORT; 538 return (0); 539 } 540 541 /* 542 * Do proper source address selection on an unbound socket in case 543 * of connect. Take jails into account as well. 544 */ 545 static int 546 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr, 547 struct ucred *cred) 548 { 549 struct in_ifaddr *ia; 550 struct ifaddr *ifa; 551 struct sockaddr *sa; 552 struct sockaddr_in *sin; 553 struct route sro; 554 int error; 555 556 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__)); 557 558 error = 0; 559 ia = NULL; 560 bzero(&sro, sizeof(sro)); 561 562 sin = (struct sockaddr_in *)&sro.ro_dst; 563 sin->sin_family = AF_INET; 564 sin->sin_len = sizeof(struct sockaddr_in); 565 sin->sin_addr.s_addr = faddr->s_addr; 566 567 /* 568 * If route is known our src addr is taken from the i/f, 569 * else punt. 570 * 571 * Find out route to destination. 572 */ 573 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) 574 in_rtalloc_ign(&sro, RTF_CLONING, inp->inp_inc.inc_fibnum); 575 576 /* 577 * If we found a route, use the address corresponding to 578 * the outgoing interface. 579 * 580 * Otherwise assume faddr is reachable on a directly connected 581 * network and try to find a corresponding interface to take 582 * the source address from. 583 */ 584 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) { 585 struct ifnet *ifp; 586 587 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin)); 588 if (ia == NULL) 589 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin)); 590 if (ia == NULL) { 591 error = ENETUNREACH; 592 goto done; 593 } 594 595 if (cred == NULL || !jailed(cred)) { 596 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 597 goto done; 598 } 599 600 ifp = ia->ia_ifp; 601 ia = NULL; 602 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 603 604 sa = ifa->ifa_addr; 605 if (sa->sa_family != AF_INET) 606 continue; 607 sin = (struct sockaddr_in *)sa; 608 if (prison_check_ip4(cred, &sin->sin_addr)) { 609 ia = (struct in_ifaddr *)ifa; 610 break; 611 } 612 } 613 if (ia != NULL) { 614 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 615 goto done; 616 } 617 618 /* 3. As a last resort return the 'default' jail address. */ 619 if (prison_getip4(cred, laddr) != 0) 620 error = EADDRNOTAVAIL; 621 goto done; 622 } 623 624 /* 625 * If the outgoing interface on the route found is not 626 * a loopback interface, use the address from that interface. 627 * In case of jails do those three steps: 628 * 1. check if the interface address belongs to the jail. If so use it. 629 * 2. check if we have any address on the outgoing interface 630 * belonging to this jail. If so use it. 631 * 3. as a last resort return the 'default' jail address. 632 */ 633 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) { 634 635 /* If not jailed, use the default returned. */ 636 if (cred == NULL || !jailed(cred)) { 637 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; 638 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 639 goto done; 640 } 641 642 /* Jailed. */ 643 /* 1. Check if the iface address belongs to the jail. */ 644 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr; 645 if (prison_check_ip4(cred, &sin->sin_addr)) { 646 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; 647 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 648 goto done; 649 } 650 651 /* 652 * 2. Check if we have any address on the outgoing interface 653 * belonging to this jail. 654 */ 655 TAILQ_FOREACH(ifa, &sro.ro_rt->rt_ifp->if_addrhead, ifa_link) { 656 657 sa = ifa->ifa_addr; 658 if (sa->sa_family != AF_INET) 659 continue; 660 sin = (struct sockaddr_in *)sa; 661 if (prison_check_ip4(cred, &sin->sin_addr)) { 662 ia = (struct in_ifaddr *)ifa; 663 break; 664 } 665 } 666 if (ia != NULL) { 667 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 668 goto done; 669 } 670 671 /* 3. As a last resort return the 'default' jail address. */ 672 if (prison_getip4(cred, laddr) != 0) 673 error = EADDRNOTAVAIL; 674 goto done; 675 } 676 677 /* 678 * The outgoing interface is marked with 'loopback net', so a route 679 * to ourselves is here. 680 * Try to find the interface of the destination address and then 681 * take the address from there. That interface is not necessarily 682 * a loopback interface. 683 * In case of jails, check that it is an address of the jail 684 * and if we cannot find, fall back to the 'default' jail address. 685 */ 686 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 687 struct sockaddr_in sain; 688 689 bzero(&sain, sizeof(struct sockaddr_in)); 690 sain.sin_family = AF_INET; 691 sain.sin_len = sizeof(struct sockaddr_in); 692 sain.sin_addr.s_addr = faddr->s_addr; 693 694 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain))); 695 if (ia == NULL) 696 ia = ifatoia(ifa_ifwithnet(sintosa(&sain))); 697 698 if (cred == NULL || !jailed(cred)) { 699 if (ia == NULL) { 700 error = ENETUNREACH; 701 goto done; 702 } 703 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 704 goto done; 705 } 706 707 /* Jailed. */ 708 if (ia != NULL) { 709 struct ifnet *ifp; 710 711 ifp = ia->ia_ifp; 712 ia = NULL; 713 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 714 715 sa = ifa->ifa_addr; 716 if (sa->sa_family != AF_INET) 717 continue; 718 sin = (struct sockaddr_in *)sa; 719 if (prison_check_ip4(cred, &sin->sin_addr)) { 720 ia = (struct in_ifaddr *)ifa; 721 break; 722 } 723 } 724 if (ia != NULL) { 725 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 726 goto done; 727 } 728 } 729 730 /* 3. As a last resort return the 'default' jail address. */ 731 if (prison_getip4(cred, laddr) != 0) 732 error = EADDRNOTAVAIL; 733 goto done; 734 } 735 736 done: 737 if (sro.ro_rt != NULL) 738 RTFREE(sro.ro_rt); 739 return (error); 740 } 741 742 /* 743 * Set up for a connect from a socket to the specified address. 744 * On entry, *laddrp and *lportp should contain the current local 745 * address and port for the PCB; these are updated to the values 746 * that should be placed in inp_laddr and inp_lport to complete 747 * the connect. 748 * 749 * On success, *faddrp and *fportp will be set to the remote address 750 * and port. These are not updated in the error case. 751 * 752 * If the operation fails because the connection already exists, 753 * *oinpp will be set to the PCB of that connection so that the 754 * caller can decide to override it. In all other cases, *oinpp 755 * is set to NULL. 756 */ 757 int 758 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam, 759 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp, 760 struct inpcb **oinpp, struct ucred *cred) 761 { 762 INIT_VNET_INET(inp->inp_vnet); 763 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 764 struct in_ifaddr *ia; 765 struct inpcb *oinp; 766 struct in_addr laddr, faddr, jailia; 767 u_short lport, fport; 768 int error; 769 770 /* 771 * Because a global state change doesn't actually occur here, a read 772 * lock is sufficient. 773 */ 774 INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo); 775 INP_LOCK_ASSERT(inp); 776 777 if (oinpp != NULL) 778 *oinpp = NULL; 779 if (nam->sa_len != sizeof (*sin)) 780 return (EINVAL); 781 if (sin->sin_family != AF_INET) 782 return (EAFNOSUPPORT); 783 if (sin->sin_port == 0) 784 return (EADDRNOTAVAIL); 785 laddr.s_addr = *laddrp; 786 lport = *lportp; 787 faddr = sin->sin_addr; 788 fport = sin->sin_port; 789 790 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) { 791 /* 792 * If the destination address is INADDR_ANY, 793 * use the primary local address. 794 * If the supplied address is INADDR_BROADCAST, 795 * and the primary interface supports broadcast, 796 * choose the broadcast address for that interface. 797 */ 798 if (faddr.s_addr == INADDR_ANY) { 799 if (cred != NULL && jailed(cred)) { 800 if (prison_getip4(cred, &jailia) != 0) 801 return (EADDRNOTAVAIL); 802 faddr.s_addr = jailia.s_addr; 803 } else { 804 faddr = 805 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))-> 806 sin_addr; 807 } 808 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST && 809 (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags & 810 IFF_BROADCAST)) 811 faddr = satosin(&TAILQ_FIRST( 812 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr; 813 } 814 if (laddr.s_addr == INADDR_ANY) { 815 error = in_pcbladdr(inp, &faddr, &laddr, cred); 816 if (error) 817 return (error); 818 819 /* 820 * If the destination address is multicast and an outgoing 821 * interface has been set as a multicast option, use the 822 * address of that interface as our source address. 823 */ 824 if (IN_MULTICAST(ntohl(faddr.s_addr)) && 825 inp->inp_moptions != NULL) { 826 struct ip_moptions *imo; 827 struct ifnet *ifp; 828 829 imo = inp->inp_moptions; 830 if (imo->imo_multicast_ifp != NULL) { 831 ifp = imo->imo_multicast_ifp; 832 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) 833 if (ia->ia_ifp == ifp) 834 break; 835 if (ia == NULL) 836 return (EADDRNOTAVAIL); 837 laddr = ia->ia_addr.sin_addr; 838 } 839 } 840 } 841 842 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport, 843 0, NULL); 844 if (oinp != NULL) { 845 if (oinpp != NULL) 846 *oinpp = oinp; 847 return (EADDRINUSE); 848 } 849 if (lport == 0) { 850 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, 851 cred); 852 if (error) 853 return (error); 854 } 855 *laddrp = laddr.s_addr; 856 *lportp = lport; 857 *faddrp = faddr.s_addr; 858 *fportp = fport; 859 return (0); 860 } 861 862 void 863 in_pcbdisconnect(struct inpcb *inp) 864 { 865 866 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 867 INP_WLOCK_ASSERT(inp); 868 869 inp->inp_faddr.s_addr = INADDR_ANY; 870 inp->inp_fport = 0; 871 in_pcbrehash(inp); 872 } 873 874 /* 875 * Historically, in_pcbdetach() included the functionality now found in 876 * in_pcbfree() and in_pcbdrop(). They are now broken out to reflect the 877 * more complex life cycle of TCP. 878 * 879 * in_pcbdetach() is responsibe for disconnecting the socket from an inpcb. 880 * For most protocols, this will be invoked immediately prior to calling 881 * in_pcbfree(). However, for TCP the inpcb may significantly outlive the 882 * socket, in which case in_pcbfree() may be deferred. 883 */ 884 void 885 in_pcbdetach(struct inpcb *inp) 886 { 887 888 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__)); 889 890 inp->inp_socket->so_pcb = NULL; 891 inp->inp_socket = NULL; 892 } 893 894 /* 895 * in_pcbfree() is responsible for freeing an already-detached inpcb, as well 896 * as removing it from any global inpcb lists it might be on. 897 */ 898 void 899 in_pcbfree(struct inpcb *inp) 900 { 901 struct inpcbinfo *ipi = inp->inp_pcbinfo; 902 903 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); 904 905 INP_INFO_WLOCK_ASSERT(ipi); 906 INP_WLOCK_ASSERT(inp); 907 908 #ifdef IPSEC 909 if (inp->inp_sp != NULL) 910 ipsec_delete_pcbpolicy(inp); 911 #endif /* IPSEC */ 912 inp->inp_gencnt = ++ipi->ipi_gencnt; 913 in_pcbremlists(inp); 914 #ifdef INET6 915 if (inp->inp_vflag & INP_IPV6PROTO) { 916 ip6_freepcbopts(inp->in6p_outputopts); 917 ip6_freemoptions(inp->in6p_moptions); 918 } 919 #endif 920 if (inp->inp_options) 921 (void)m_free(inp->inp_options); 922 if (inp->inp_moptions != NULL) 923 inp_freemoptions(inp->inp_moptions); 924 inp->inp_vflag = 0; 925 crfree(inp->inp_cred); 926 927 #ifdef MAC 928 mac_inpcb_destroy(inp); 929 #endif 930 INP_WUNLOCK(inp); 931 uma_zfree(ipi->ipi_zone, inp); 932 } 933 934 /* 935 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and 936 * port reservation, and preventing it from being returned by inpcb lookups. 937 * 938 * It is used by TCP to mark an inpcb as unused and avoid future packet 939 * delivery or event notification when a socket remains open but TCP has 940 * closed. This might occur as a result of a shutdown()-initiated TCP close 941 * or a RST on the wire, and allows the port binding to be reused while still 942 * maintaining the invariant that so_pcb always points to a valid inpcb until 943 * in_pcbdetach(). 944 * 945 * XXXRW: An inp_lport of 0 is used to indicate that the inpcb is not on hash 946 * lists, but can lead to confusing netstat output, as open sockets with 947 * closed TCP connections will no longer appear to have their bound port 948 * number. An explicit flag would be better, as it would allow us to leave 949 * the port number intact after the connection is dropped. 950 * 951 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by 952 * in_pcbnotifyall() and in_pcbpurgeif0()? 953 */ 954 void 955 in_pcbdrop(struct inpcb *inp) 956 { 957 958 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 959 INP_WLOCK_ASSERT(inp); 960 961 inp->inp_vflag |= INP_DROPPED; 962 if (inp->inp_lport) { 963 struct inpcbport *phd = inp->inp_phd; 964 965 LIST_REMOVE(inp, inp_hash); 966 LIST_REMOVE(inp, inp_portlist); 967 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 968 LIST_REMOVE(phd, phd_hash); 969 free(phd, M_PCB); 970 } 971 inp->inp_lport = 0; 972 } 973 } 974 975 /* 976 * Common routines to return the socket addresses associated with inpcbs. 977 */ 978 struct sockaddr * 979 in_sockaddr(in_port_t port, struct in_addr *addr_p) 980 { 981 struct sockaddr_in *sin; 982 983 sin = malloc(sizeof *sin, M_SONAME, 984 M_WAITOK | M_ZERO); 985 sin->sin_family = AF_INET; 986 sin->sin_len = sizeof(*sin); 987 sin->sin_addr = *addr_p; 988 sin->sin_port = port; 989 990 return (struct sockaddr *)sin; 991 } 992 993 int 994 in_getsockaddr(struct socket *so, struct sockaddr **nam) 995 { 996 struct inpcb *inp; 997 struct in_addr addr; 998 in_port_t port; 999 1000 inp = sotoinpcb(so); 1001 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL")); 1002 1003 INP_RLOCK(inp); 1004 port = inp->inp_lport; 1005 addr = inp->inp_laddr; 1006 INP_RUNLOCK(inp); 1007 1008 *nam = in_sockaddr(port, &addr); 1009 return 0; 1010 } 1011 1012 int 1013 in_getpeeraddr(struct socket *so, struct sockaddr **nam) 1014 { 1015 struct inpcb *inp; 1016 struct in_addr addr; 1017 in_port_t port; 1018 1019 inp = sotoinpcb(so); 1020 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL")); 1021 1022 INP_RLOCK(inp); 1023 port = inp->inp_fport; 1024 addr = inp->inp_faddr; 1025 INP_RUNLOCK(inp); 1026 1027 *nam = in_sockaddr(port, &addr); 1028 return 0; 1029 } 1030 1031 void 1032 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno, 1033 struct inpcb *(*notify)(struct inpcb *, int)) 1034 { 1035 struct inpcb *inp, *inp_temp; 1036 1037 INP_INFO_WLOCK(pcbinfo); 1038 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { 1039 INP_WLOCK(inp); 1040 #ifdef INET6 1041 if ((inp->inp_vflag & INP_IPV4) == 0) { 1042 INP_WUNLOCK(inp); 1043 continue; 1044 } 1045 #endif 1046 if (inp->inp_faddr.s_addr != faddr.s_addr || 1047 inp->inp_socket == NULL) { 1048 INP_WUNLOCK(inp); 1049 continue; 1050 } 1051 if ((*notify)(inp, errno)) 1052 INP_WUNLOCK(inp); 1053 } 1054 INP_INFO_WUNLOCK(pcbinfo); 1055 } 1056 1057 void 1058 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) 1059 { 1060 struct inpcb *inp; 1061 struct ip_moptions *imo; 1062 int i, gap; 1063 1064 INP_INFO_RLOCK(pcbinfo); 1065 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) { 1066 INP_WLOCK(inp); 1067 imo = inp->inp_moptions; 1068 if ((inp->inp_vflag & INP_IPV4) && 1069 imo != NULL) { 1070 /* 1071 * Unselect the outgoing interface if it is being 1072 * detached. 1073 */ 1074 if (imo->imo_multicast_ifp == ifp) 1075 imo->imo_multicast_ifp = NULL; 1076 1077 /* 1078 * Drop multicast group membership if we joined 1079 * through the interface being detached. 1080 */ 1081 for (i = 0, gap = 0; i < imo->imo_num_memberships; 1082 i++) { 1083 if (imo->imo_membership[i]->inm_ifp == ifp) { 1084 in_delmulti(imo->imo_membership[i]); 1085 gap++; 1086 } else if (gap != 0) 1087 imo->imo_membership[i - gap] = 1088 imo->imo_membership[i]; 1089 } 1090 imo->imo_num_memberships -= gap; 1091 } 1092 INP_WUNLOCK(inp); 1093 } 1094 INP_INFO_RUNLOCK(pcbinfo); 1095 } 1096 1097 /* 1098 * Lookup a PCB based on the local address and port. 1099 */ 1100 #define INP_LOOKUP_MAPPED_PCB_COST 3 1101 struct inpcb * 1102 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr, 1103 u_short lport, int wild_okay, struct ucred *cred) 1104 { 1105 struct inpcb *inp; 1106 #ifdef INET6 1107 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST; 1108 #else 1109 int matchwild = 3; 1110 #endif 1111 int wildcard; 1112 1113 INP_INFO_LOCK_ASSERT(pcbinfo); 1114 1115 if (!wild_okay) { 1116 struct inpcbhead *head; 1117 /* 1118 * Look for an unconnected (wildcard foreign addr) PCB that 1119 * matches the local address and port we're looking for. 1120 */ 1121 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 1122 0, pcbinfo->ipi_hashmask)]; 1123 LIST_FOREACH(inp, head, inp_hash) { 1124 #ifdef INET6 1125 /* XXX inp locking */ 1126 if ((inp->inp_vflag & INP_IPV4) == 0) 1127 continue; 1128 #endif 1129 if (inp->inp_faddr.s_addr == INADDR_ANY && 1130 inp->inp_laddr.s_addr == laddr.s_addr && 1131 inp->inp_lport == lport) { 1132 /* 1133 * Found? 1134 */ 1135 if (cred == NULL || 1136 inp->inp_cred->cr_prison == cred->cr_prison) 1137 return (inp); 1138 } 1139 } 1140 /* 1141 * Not found. 1142 */ 1143 return (NULL); 1144 } else { 1145 struct inpcbporthead *porthash; 1146 struct inpcbport *phd; 1147 struct inpcb *match = NULL; 1148 /* 1149 * Best fit PCB lookup. 1150 * 1151 * First see if this local port is in use by looking on the 1152 * port hash list. 1153 */ 1154 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, 1155 pcbinfo->ipi_porthashmask)]; 1156 LIST_FOREACH(phd, porthash, phd_hash) { 1157 if (phd->phd_port == lport) 1158 break; 1159 } 1160 if (phd != NULL) { 1161 /* 1162 * Port is in use by one or more PCBs. Look for best 1163 * fit. 1164 */ 1165 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 1166 wildcard = 0; 1167 if (cred != NULL && 1168 inp->inp_cred->cr_prison != cred->cr_prison) 1169 continue; 1170 #ifdef INET6 1171 /* XXX inp locking */ 1172 if ((inp->inp_vflag & INP_IPV4) == 0) 1173 continue; 1174 /* 1175 * We never select the PCB that has 1176 * INP_IPV6 flag and is bound to :: if 1177 * we have another PCB which is bound 1178 * to 0.0.0.0. If a PCB has the 1179 * INP_IPV6 flag, then we set its cost 1180 * higher than IPv4 only PCBs. 1181 * 1182 * Note that the case only happens 1183 * when a socket is bound to ::, under 1184 * the condition that the use of the 1185 * mapped address is allowed. 1186 */ 1187 if ((inp->inp_vflag & INP_IPV6) != 0) 1188 wildcard += INP_LOOKUP_MAPPED_PCB_COST; 1189 #endif 1190 if (inp->inp_faddr.s_addr != INADDR_ANY) 1191 wildcard++; 1192 if (inp->inp_laddr.s_addr != INADDR_ANY) { 1193 if (laddr.s_addr == INADDR_ANY) 1194 wildcard++; 1195 else if (inp->inp_laddr.s_addr != laddr.s_addr) 1196 continue; 1197 } else { 1198 if (laddr.s_addr != INADDR_ANY) 1199 wildcard++; 1200 } 1201 if (wildcard < matchwild) { 1202 match = inp; 1203 matchwild = wildcard; 1204 if (matchwild == 0) 1205 break; 1206 } 1207 } 1208 } 1209 return (match); 1210 } 1211 } 1212 #undef INP_LOOKUP_MAPPED_PCB_COST 1213 1214 /* 1215 * Lookup PCB in hash list. 1216 */ 1217 struct inpcb * 1218 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, 1219 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard, 1220 struct ifnet *ifp) 1221 { 1222 struct inpcbhead *head; 1223 struct inpcb *inp, *tmpinp; 1224 u_short fport = fport_arg, lport = lport_arg; 1225 1226 INP_INFO_LOCK_ASSERT(pcbinfo); 1227 1228 /* 1229 * First look for an exact match. 1230 */ 1231 tmpinp = NULL; 1232 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, 1233 pcbinfo->ipi_hashmask)]; 1234 LIST_FOREACH(inp, head, inp_hash) { 1235 #ifdef INET6 1236 /* XXX inp locking */ 1237 if ((inp->inp_vflag & INP_IPV4) == 0) 1238 continue; 1239 #endif 1240 if (inp->inp_faddr.s_addr == faddr.s_addr && 1241 inp->inp_laddr.s_addr == laddr.s_addr && 1242 inp->inp_fport == fport && 1243 inp->inp_lport == lport) { 1244 /* 1245 * XXX We should be able to directly return 1246 * the inp here, without any checks. 1247 * Well unless both bound with SO_REUSEPORT? 1248 */ 1249 if (jailed(inp->inp_cred)) 1250 return (inp); 1251 if (tmpinp == NULL) 1252 tmpinp = inp; 1253 } 1254 } 1255 if (tmpinp != NULL) 1256 return (tmpinp); 1257 1258 /* 1259 * Then look for a wildcard match, if requested. 1260 */ 1261 if (wildcard == INPLOOKUP_WILDCARD) { 1262 struct inpcb *local_wild = NULL, *local_exact = NULL; 1263 #ifdef INET6 1264 struct inpcb *local_wild_mapped = NULL; 1265 #endif 1266 struct inpcb *jail_wild = NULL; 1267 int injail; 1268 1269 /* 1270 * Order of socket selection - we always prefer jails. 1271 * 1. jailed, non-wild. 1272 * 2. jailed, wild. 1273 * 3. non-jailed, non-wild. 1274 * 4. non-jailed, wild. 1275 */ 1276 1277 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 1278 0, pcbinfo->ipi_hashmask)]; 1279 LIST_FOREACH(inp, head, inp_hash) { 1280 #ifdef INET6 1281 /* XXX inp locking */ 1282 if ((inp->inp_vflag & INP_IPV4) == 0) 1283 continue; 1284 #endif 1285 if (inp->inp_faddr.s_addr != INADDR_ANY || 1286 inp->inp_lport != lport) 1287 continue; 1288 1289 /* XXX inp locking */ 1290 if (ifp && ifp->if_type == IFT_FAITH && 1291 (inp->inp_flags & INP_FAITH) == 0) 1292 continue; 1293 1294 injail = jailed(inp->inp_cred); 1295 if (injail) { 1296 if (!prison_check_ip4(inp->inp_cred, &laddr)) 1297 continue; 1298 } else { 1299 if (local_exact != NULL) 1300 continue; 1301 } 1302 1303 if (inp->inp_laddr.s_addr == laddr.s_addr) { 1304 if (injail) 1305 return (inp); 1306 else 1307 local_exact = inp; 1308 } else if (inp->inp_laddr.s_addr == INADDR_ANY) { 1309 #ifdef INET6 1310 /* XXX inp locking, NULL check */ 1311 if (inp->inp_vflag & INP_IPV6PROTO) 1312 local_wild_mapped = inp; 1313 else 1314 #endif /* INET6 */ 1315 if (injail) 1316 jail_wild = inp; 1317 else 1318 local_wild = inp; 1319 } 1320 } /* LIST_FOREACH */ 1321 if (jail_wild != NULL) 1322 return (jail_wild); 1323 if (local_exact != NULL) 1324 return (local_exact); 1325 if (local_wild != NULL) 1326 return (local_wild); 1327 #ifdef INET6 1328 if (local_wild_mapped != NULL) 1329 return (local_wild_mapped); 1330 #endif /* defined(INET6) */ 1331 } /* if (wildcard == INPLOOKUP_WILDCARD) */ 1332 1333 return (NULL); 1334 } 1335 1336 /* 1337 * Insert PCB onto various hash lists. 1338 */ 1339 int 1340 in_pcbinshash(struct inpcb *inp) 1341 { 1342 struct inpcbhead *pcbhash; 1343 struct inpcbporthead *pcbporthash; 1344 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1345 struct inpcbport *phd; 1346 u_int32_t hashkey_faddr; 1347 1348 INP_INFO_WLOCK_ASSERT(pcbinfo); 1349 INP_WLOCK_ASSERT(inp); 1350 1351 #ifdef INET6 1352 if (inp->inp_vflag & INP_IPV6) 1353 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1354 else 1355 #endif /* INET6 */ 1356 hashkey_faddr = inp->inp_faddr.s_addr; 1357 1358 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, 1359 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; 1360 1361 pcbporthash = &pcbinfo->ipi_porthashbase[ 1362 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)]; 1363 1364 /* 1365 * Go through port list and look for a head for this lport. 1366 */ 1367 LIST_FOREACH(phd, pcbporthash, phd_hash) { 1368 if (phd->phd_port == inp->inp_lport) 1369 break; 1370 } 1371 /* 1372 * If none exists, malloc one and tack it on. 1373 */ 1374 if (phd == NULL) { 1375 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT); 1376 if (phd == NULL) { 1377 return (ENOBUFS); /* XXX */ 1378 } 1379 phd->phd_port = inp->inp_lport; 1380 LIST_INIT(&phd->phd_pcblist); 1381 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 1382 } 1383 inp->inp_phd = phd; 1384 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 1385 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 1386 return (0); 1387 } 1388 1389 /* 1390 * Move PCB to the proper hash bucket when { faddr, fport } have been 1391 * changed. NOTE: This does not handle the case of the lport changing (the 1392 * hashed port list would have to be updated as well), so the lport must 1393 * not change after in_pcbinshash() has been called. 1394 */ 1395 void 1396 in_pcbrehash(struct inpcb *inp) 1397 { 1398 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1399 struct inpcbhead *head; 1400 u_int32_t hashkey_faddr; 1401 1402 INP_INFO_WLOCK_ASSERT(pcbinfo); 1403 INP_WLOCK_ASSERT(inp); 1404 1405 #ifdef INET6 1406 if (inp->inp_vflag & INP_IPV6) 1407 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1408 else 1409 #endif /* INET6 */ 1410 hashkey_faddr = inp->inp_faddr.s_addr; 1411 1412 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, 1413 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; 1414 1415 LIST_REMOVE(inp, inp_hash); 1416 LIST_INSERT_HEAD(head, inp, inp_hash); 1417 } 1418 1419 /* 1420 * Remove PCB from various lists. 1421 */ 1422 void 1423 in_pcbremlists(struct inpcb *inp) 1424 { 1425 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1426 1427 INP_INFO_WLOCK_ASSERT(pcbinfo); 1428 INP_WLOCK_ASSERT(inp); 1429 1430 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 1431 if (inp->inp_lport) { 1432 struct inpcbport *phd = inp->inp_phd; 1433 1434 LIST_REMOVE(inp, inp_hash); 1435 LIST_REMOVE(inp, inp_portlist); 1436 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 1437 LIST_REMOVE(phd, phd_hash); 1438 free(phd, M_PCB); 1439 } 1440 } 1441 LIST_REMOVE(inp, inp_list); 1442 pcbinfo->ipi_count--; 1443 } 1444 1445 /* 1446 * A set label operation has occurred at the socket layer, propagate the 1447 * label change into the in_pcb for the socket. 1448 */ 1449 void 1450 in_pcbsosetlabel(struct socket *so) 1451 { 1452 #ifdef MAC 1453 struct inpcb *inp; 1454 1455 inp = sotoinpcb(so); 1456 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL")); 1457 1458 INP_WLOCK(inp); 1459 SOCK_LOCK(so); 1460 mac_inpcb_sosetlabel(so, inp); 1461 SOCK_UNLOCK(so); 1462 INP_WUNLOCK(inp); 1463 #endif 1464 } 1465 1466 /* 1467 * ipport_tick runs once per second, determining if random port allocation 1468 * should be continued. If more than ipport_randomcps ports have been 1469 * allocated in the last second, then we return to sequential port 1470 * allocation. We return to random allocation only once we drop below 1471 * ipport_randomcps for at least ipport_randomtime seconds. 1472 */ 1473 void 1474 ipport_tick(void *xtp) 1475 { 1476 VNET_ITERATOR_DECL(vnet_iter); 1477 1478 VNET_LIST_RLOCK(); 1479 VNET_FOREACH(vnet_iter) { 1480 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */ 1481 INIT_VNET_INET(vnet_iter); 1482 if (V_ipport_tcpallocs <= 1483 V_ipport_tcplastcount + V_ipport_randomcps) { 1484 if (V_ipport_stoprandom > 0) 1485 V_ipport_stoprandom--; 1486 } else 1487 V_ipport_stoprandom = V_ipport_randomtime; 1488 V_ipport_tcplastcount = V_ipport_tcpallocs; 1489 CURVNET_RESTORE(); 1490 } 1491 VNET_LIST_RUNLOCK(); 1492 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL); 1493 } 1494 1495 void 1496 inp_wlock(struct inpcb *inp) 1497 { 1498 1499 INP_WLOCK(inp); 1500 } 1501 1502 void 1503 inp_wunlock(struct inpcb *inp) 1504 { 1505 1506 INP_WUNLOCK(inp); 1507 } 1508 1509 void 1510 inp_rlock(struct inpcb *inp) 1511 { 1512 1513 INP_RLOCK(inp); 1514 } 1515 1516 void 1517 inp_runlock(struct inpcb *inp) 1518 { 1519 1520 INP_RUNLOCK(inp); 1521 } 1522 1523 #ifdef INVARIANTS 1524 void 1525 inp_lock_assert(struct inpcb *inp) 1526 { 1527 1528 INP_WLOCK_ASSERT(inp); 1529 } 1530 1531 void 1532 inp_unlock_assert(struct inpcb *inp) 1533 { 1534 1535 INP_UNLOCK_ASSERT(inp); 1536 } 1537 #endif 1538 1539 void 1540 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg) 1541 { 1542 INIT_VNET_INET(curvnet); 1543 struct inpcb *inp; 1544 1545 INP_INFO_RLOCK(&V_tcbinfo); 1546 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) { 1547 INP_WLOCK(inp); 1548 func(inp, arg); 1549 INP_WUNLOCK(inp); 1550 } 1551 INP_INFO_RUNLOCK(&V_tcbinfo); 1552 } 1553 1554 struct socket * 1555 inp_inpcbtosocket(struct inpcb *inp) 1556 { 1557 1558 INP_WLOCK_ASSERT(inp); 1559 return (inp->inp_socket); 1560 } 1561 1562 struct tcpcb * 1563 inp_inpcbtotcpcb(struct inpcb *inp) 1564 { 1565 1566 INP_WLOCK_ASSERT(inp); 1567 return ((struct tcpcb *)inp->inp_ppcb); 1568 } 1569 1570 int 1571 inp_ip_tos_get(const struct inpcb *inp) 1572 { 1573 1574 return (inp->inp_ip_tos); 1575 } 1576 1577 void 1578 inp_ip_tos_set(struct inpcb *inp, int val) 1579 { 1580 1581 inp->inp_ip_tos = val; 1582 } 1583 1584 void 1585 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, 1586 uint32_t *faddr, uint16_t *fp) 1587 { 1588 1589 INP_LOCK_ASSERT(inp); 1590 *laddr = inp->inp_laddr.s_addr; 1591 *faddr = inp->inp_faddr.s_addr; 1592 *lp = inp->inp_lport; 1593 *fp = inp->inp_fport; 1594 } 1595 1596 struct inpcb * 1597 so_sotoinpcb(struct socket *so) 1598 { 1599 1600 return (sotoinpcb(so)); 1601 } 1602 1603 struct tcpcb * 1604 so_sototcpcb(struct socket *so) 1605 { 1606 1607 return (sototcpcb(so)); 1608 } 1609 1610 #ifdef DDB 1611 static void 1612 db_print_indent(int indent) 1613 { 1614 int i; 1615 1616 for (i = 0; i < indent; i++) 1617 db_printf(" "); 1618 } 1619 1620 static void 1621 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent) 1622 { 1623 char faddr_str[48], laddr_str[48]; 1624 1625 db_print_indent(indent); 1626 db_printf("%s at %p\n", name, inc); 1627 1628 indent += 2; 1629 1630 #ifdef INET6 1631 if (inc->inc_flags == 1) { 1632 /* IPv6. */ 1633 ip6_sprintf(laddr_str, &inc->inc6_laddr); 1634 ip6_sprintf(faddr_str, &inc->inc6_faddr); 1635 } else { 1636 #endif 1637 /* IPv4. */ 1638 inet_ntoa_r(inc->inc_laddr, laddr_str); 1639 inet_ntoa_r(inc->inc_faddr, faddr_str); 1640 #ifdef INET6 1641 } 1642 #endif 1643 db_print_indent(indent); 1644 db_printf("inc_laddr %s inc_lport %u\n", laddr_str, 1645 ntohs(inc->inc_lport)); 1646 db_print_indent(indent); 1647 db_printf("inc_faddr %s inc_fport %u\n", faddr_str, 1648 ntohs(inc->inc_fport)); 1649 } 1650 1651 static void 1652 db_print_inpflags(int inp_flags) 1653 { 1654 int comma; 1655 1656 comma = 0; 1657 if (inp_flags & INP_RECVOPTS) { 1658 db_printf("%sINP_RECVOPTS", comma ? ", " : ""); 1659 comma = 1; 1660 } 1661 if (inp_flags & INP_RECVRETOPTS) { 1662 db_printf("%sINP_RECVRETOPTS", comma ? ", " : ""); 1663 comma = 1; 1664 } 1665 if (inp_flags & INP_RECVDSTADDR) { 1666 db_printf("%sINP_RECVDSTADDR", comma ? ", " : ""); 1667 comma = 1; 1668 } 1669 if (inp_flags & INP_HDRINCL) { 1670 db_printf("%sINP_HDRINCL", comma ? ", " : ""); 1671 comma = 1; 1672 } 1673 if (inp_flags & INP_HIGHPORT) { 1674 db_printf("%sINP_HIGHPORT", comma ? ", " : ""); 1675 comma = 1; 1676 } 1677 if (inp_flags & INP_LOWPORT) { 1678 db_printf("%sINP_LOWPORT", comma ? ", " : ""); 1679 comma = 1; 1680 } 1681 if (inp_flags & INP_ANONPORT) { 1682 db_printf("%sINP_ANONPORT", comma ? ", " : ""); 1683 comma = 1; 1684 } 1685 if (inp_flags & INP_RECVIF) { 1686 db_printf("%sINP_RECVIF", comma ? ", " : ""); 1687 comma = 1; 1688 } 1689 if (inp_flags & INP_MTUDISC) { 1690 db_printf("%sINP_MTUDISC", comma ? ", " : ""); 1691 comma = 1; 1692 } 1693 if (inp_flags & INP_FAITH) { 1694 db_printf("%sINP_FAITH", comma ? ", " : ""); 1695 comma = 1; 1696 } 1697 if (inp_flags & INP_RECVTTL) { 1698 db_printf("%sINP_RECVTTL", comma ? ", " : ""); 1699 comma = 1; 1700 } 1701 if (inp_flags & INP_DONTFRAG) { 1702 db_printf("%sINP_DONTFRAG", comma ? ", " : ""); 1703 comma = 1; 1704 } 1705 if (inp_flags & IN6P_IPV6_V6ONLY) { 1706 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : ""); 1707 comma = 1; 1708 } 1709 if (inp_flags & IN6P_PKTINFO) { 1710 db_printf("%sIN6P_PKTINFO", comma ? ", " : ""); 1711 comma = 1; 1712 } 1713 if (inp_flags & IN6P_HOPLIMIT) { 1714 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : ""); 1715 comma = 1; 1716 } 1717 if (inp_flags & IN6P_HOPOPTS) { 1718 db_printf("%sIN6P_HOPOPTS", comma ? ", " : ""); 1719 comma = 1; 1720 } 1721 if (inp_flags & IN6P_DSTOPTS) { 1722 db_printf("%sIN6P_DSTOPTS", comma ? ", " : ""); 1723 comma = 1; 1724 } 1725 if (inp_flags & IN6P_RTHDR) { 1726 db_printf("%sIN6P_RTHDR", comma ? ", " : ""); 1727 comma = 1; 1728 } 1729 if (inp_flags & IN6P_RTHDRDSTOPTS) { 1730 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : ""); 1731 comma = 1; 1732 } 1733 if (inp_flags & IN6P_TCLASS) { 1734 db_printf("%sIN6P_TCLASS", comma ? ", " : ""); 1735 comma = 1; 1736 } 1737 if (inp_flags & IN6P_AUTOFLOWLABEL) { 1738 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : ""); 1739 comma = 1; 1740 } 1741 if (inp_flags & IN6P_RFC2292) { 1742 db_printf("%sIN6P_RFC2292", comma ? ", " : ""); 1743 comma = 1; 1744 } 1745 if (inp_flags & IN6P_MTU) { 1746 db_printf("IN6P_MTU%s", comma ? ", " : ""); 1747 comma = 1; 1748 } 1749 } 1750 1751 static void 1752 db_print_inpvflag(u_char inp_vflag) 1753 { 1754 int comma; 1755 1756 comma = 0; 1757 if (inp_vflag & INP_IPV4) { 1758 db_printf("%sINP_IPV4", comma ? ", " : ""); 1759 comma = 1; 1760 } 1761 if (inp_vflag & INP_IPV6) { 1762 db_printf("%sINP_IPV6", comma ? ", " : ""); 1763 comma = 1; 1764 } 1765 if (inp_vflag & INP_IPV6PROTO) { 1766 db_printf("%sINP_IPV6PROTO", comma ? ", " : ""); 1767 comma = 1; 1768 } 1769 if (inp_vflag & INP_TIMEWAIT) { 1770 db_printf("%sINP_TIMEWAIT", comma ? ", " : ""); 1771 comma = 1; 1772 } 1773 if (inp_vflag & INP_ONESBCAST) { 1774 db_printf("%sINP_ONESBCAST", comma ? ", " : ""); 1775 comma = 1; 1776 } 1777 if (inp_vflag & INP_DROPPED) { 1778 db_printf("%sINP_DROPPED", comma ? ", " : ""); 1779 comma = 1; 1780 } 1781 if (inp_vflag & INP_SOCKREF) { 1782 db_printf("%sINP_SOCKREF", comma ? ", " : ""); 1783 comma = 1; 1784 } 1785 } 1786 1787 void 1788 db_print_inpcb(struct inpcb *inp, const char *name, int indent) 1789 { 1790 1791 db_print_indent(indent); 1792 db_printf("%s at %p\n", name, inp); 1793 1794 indent += 2; 1795 1796 db_print_indent(indent); 1797 db_printf("inp_flow: 0x%x\n", inp->inp_flow); 1798 1799 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent); 1800 1801 db_print_indent(indent); 1802 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n", 1803 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket); 1804 1805 db_print_indent(indent); 1806 db_printf("inp_label: %p inp_flags: 0x%x (", 1807 inp->inp_label, inp->inp_flags); 1808 db_print_inpflags(inp->inp_flags); 1809 db_printf(")\n"); 1810 1811 db_print_indent(indent); 1812 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp, 1813 inp->inp_vflag); 1814 db_print_inpvflag(inp->inp_vflag); 1815 db_printf(")\n"); 1816 1817 db_print_indent(indent); 1818 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n", 1819 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl); 1820 1821 db_print_indent(indent); 1822 #ifdef INET6 1823 if (inp->inp_vflag & INP_IPV6) { 1824 db_printf("in6p_options: %p in6p_outputopts: %p " 1825 "in6p_moptions: %p\n", inp->in6p_options, 1826 inp->in6p_outputopts, inp->in6p_moptions); 1827 db_printf("in6p_icmp6filt: %p in6p_cksum %d " 1828 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum, 1829 inp->in6p_hops); 1830 } else 1831 #endif 1832 { 1833 db_printf("inp_ip_tos: %d inp_ip_options: %p " 1834 "inp_ip_moptions: %p\n", inp->inp_ip_tos, 1835 inp->inp_options, inp->inp_moptions); 1836 } 1837 1838 db_print_indent(indent); 1839 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd, 1840 (uintmax_t)inp->inp_gencnt); 1841 } 1842 1843 DB_SHOW_COMMAND(inpcb, db_show_inpcb) 1844 { 1845 struct inpcb *inp; 1846 1847 if (!have_addr) { 1848 db_printf("usage: show inpcb <addr>\n"); 1849 return; 1850 } 1851 inp = (struct inpcb *)addr; 1852 1853 db_print_inpcb(inp, "inpcb", 0); 1854 } 1855 #endif 1856