1 /*- 2 * Copyright (c) 1982, 1986, 1991, 1993, 1995 3 * The Regents of the University of California. 4 * Copyright (c) 2007-2009 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 41 #include <sys/param.h> 42 #include <sys/systm.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/priv.h> 50 #include <sys/proc.h> 51 #include <sys/jail.h> 52 #include <sys/kernel.h> 53 #include <sys/sysctl.h> 54 55 #ifdef DDB 56 #include <ddb/ddb.h> 57 #endif 58 59 #include <vm/uma.h> 60 61 #include <net/if.h> 62 #include <net/if_types.h> 63 #include <net/route.h> 64 #include <net/vnet.h> 65 66 #include <netinet/in.h> 67 #include <netinet/in_pcb.h> 68 #include <netinet/in_var.h> 69 #include <netinet/ip_var.h> 70 #include <netinet/tcp_var.h> 71 #include <netinet/udp.h> 72 #include <netinet/udp_var.h> 73 #ifdef INET6 74 #include <netinet/ip6.h> 75 #include <netinet6/ip6_var.h> 76 #endif /* INET6 */ 77 78 79 #ifdef IPSEC 80 #include <netipsec/ipsec.h> 81 #include <netipsec/key.h> 82 #endif /* IPSEC */ 83 84 #include <security/mac/mac_framework.h> 85 86 /* 87 * These configure the range of local port addresses assigned to 88 * "unspecified" outgoing connections/packets/whatever. 89 */ 90 VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */ 91 VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */ 92 VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */ 93 VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */ 94 VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */ 95 VNET_DEFINE(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 VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */ 103 VNET_DEFINE(int, ipport_reservedlow); 104 105 /* Variables dealing with random ephemeral port allocation. */ 106 VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */ 107 VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */ 108 VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */ 109 VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */ 110 VNET_DEFINE(int, ipport_tcpallocs); 111 static VNET_DEFINE(int, ipport_tcplastcount); 112 113 #define V_ipport_tcplastcount VNET(ipport_tcplastcount) 114 115 #define RANGECHK(var, min, max) \ 116 if ((var) < (min)) { (var) = (min); } \ 117 else if ((var) > (max)) { (var) = (max); } 118 119 static void in_pcbremlists(struct inpcb *inp); 120 121 static int 122 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) 123 { 124 int error; 125 126 #ifdef VIMAGE 127 error = vnet_sysctl_handle_int(oidp, arg1, arg2, req); 128 #else 129 error = sysctl_handle_int(oidp, arg1, arg2, req); 130 #endif 131 if (error == 0) { 132 RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); 133 RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1); 134 RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); 135 RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); 136 RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); 137 RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); 138 } 139 return (error); 140 } 141 142 #undef RANGECHK 143 144 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); 145 146 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, 147 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0, 148 &sysctl_net_ipport_check, "I", ""); 149 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, 150 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0, 151 &sysctl_net_ipport_check, "I", ""); 152 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, first, 153 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0, 154 &sysctl_net_ipport_check, "I", ""); 155 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, last, 156 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0, 157 &sysctl_net_ipport_check, "I", ""); 158 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, 159 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0, 160 &sysctl_net_ipport_check, "I", ""); 161 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, 162 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0, 163 &sysctl_net_ipport_check, "I", ""); 164 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, 165 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, ""); 166 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, 167 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, ""); 168 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW, 169 &VNET_NAME(ipport_randomized), 0, "Enable random port allocation"); 170 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW, 171 &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port " 172 "allocations before switching to a sequental one"); 173 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW, 174 &VNET_NAME(ipport_randomtime), 0, 175 "Minimum time to keep sequental port " 176 "allocation before switching to a random one"); 177 178 /* 179 * in_pcb.c: manage the Protocol Control Blocks. 180 * 181 * NOTE: It is assumed that most of these functions will be called with 182 * the pcbinfo lock held, and often, the inpcb lock held, as these utility 183 * functions often modify hash chains or addresses in pcbs. 184 */ 185 186 /* 187 * Initialize an inpcbinfo -- we should be able to reduce the number of 188 * arguments in time. 189 */ 190 void 191 in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name, 192 struct inpcbhead *listhead, int hash_nelements, int porthash_nelements, 193 char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini, 194 uint32_t inpcbzone_flags) 195 { 196 197 INP_INFO_LOCK_INIT(pcbinfo, name); 198 #ifdef VIMAGE 199 pcbinfo->ipi_vnet = curvnet; 200 #endif 201 pcbinfo->ipi_listhead = listhead; 202 LIST_INIT(pcbinfo->ipi_listhead); 203 pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB, 204 &pcbinfo->ipi_hashmask); 205 pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB, 206 &pcbinfo->ipi_porthashmask); 207 pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb), 208 NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR, 209 inpcbzone_flags); 210 uma_zone_set_max(pcbinfo->ipi_zone, maxsockets); 211 } 212 213 /* 214 * Destroy an inpcbinfo. 215 */ 216 void 217 in_pcbinfo_destroy(struct inpcbinfo *pcbinfo) 218 { 219 220 hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask); 221 hashdestroy(pcbinfo->ipi_porthashbase, M_PCB, 222 pcbinfo->ipi_porthashmask); 223 uma_zdestroy(pcbinfo->ipi_zone); 224 INP_INFO_LOCK_DESTROY(pcbinfo); 225 } 226 227 /* 228 * Allocate a PCB and associate it with the socket. 229 * On success return with the PCB locked. 230 */ 231 int 232 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo) 233 { 234 struct inpcb *inp; 235 int error; 236 237 INP_INFO_WLOCK_ASSERT(pcbinfo); 238 error = 0; 239 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT); 240 if (inp == NULL) 241 return (ENOBUFS); 242 bzero(inp, inp_zero_size); 243 inp->inp_pcbinfo = pcbinfo; 244 inp->inp_socket = so; 245 inp->inp_cred = crhold(so->so_cred); 246 inp->inp_inc.inc_fibnum = so->so_fibnum; 247 #ifdef MAC 248 error = mac_inpcb_init(inp, M_NOWAIT); 249 if (error != 0) 250 goto out; 251 mac_inpcb_create(so, inp); 252 #endif 253 #ifdef IPSEC 254 error = ipsec_init_policy(so, &inp->inp_sp); 255 if (error != 0) { 256 #ifdef MAC 257 mac_inpcb_destroy(inp); 258 #endif 259 goto out; 260 } 261 #endif /*IPSEC*/ 262 #ifdef INET6 263 if (INP_SOCKAF(so) == AF_INET6) { 264 inp->inp_vflag |= INP_IPV6PROTO; 265 if (V_ip6_v6only) 266 inp->inp_flags |= IN6P_IPV6_V6ONLY; 267 } 268 #endif 269 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list); 270 pcbinfo->ipi_count++; 271 so->so_pcb = (caddr_t)inp; 272 #ifdef INET6 273 if (V_ip6_auto_flowlabel) 274 inp->inp_flags |= IN6P_AUTOFLOWLABEL; 275 #endif 276 INP_WLOCK(inp); 277 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 278 inp->inp_refcount = 1; /* Reference from the inpcbinfo */ 279 #if defined(IPSEC) || defined(MAC) 280 out: 281 if (error != 0) { 282 crfree(inp->inp_cred); 283 uma_zfree(pcbinfo->ipi_zone, inp); 284 } 285 #endif 286 return (error); 287 } 288 289 int 290 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 291 { 292 int anonport, error; 293 294 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 295 INP_WLOCK_ASSERT(inp); 296 297 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) 298 return (EINVAL); 299 anonport = inp->inp_lport == 0 && (nam == NULL || 300 ((struct sockaddr_in *)nam)->sin_port == 0); 301 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, 302 &inp->inp_lport, cred); 303 if (error) 304 return (error); 305 if (in_pcbinshash(inp) != 0) { 306 inp->inp_laddr.s_addr = INADDR_ANY; 307 inp->inp_lport = 0; 308 return (EAGAIN); 309 } 310 if (anonport) 311 inp->inp_flags |= INP_ANONPORT; 312 return (0); 313 } 314 315 /* 316 * Set up a bind operation on a PCB, performing port allocation 317 * as required, but do not actually modify the PCB. Callers can 318 * either complete the bind by setting inp_laddr/inp_lport and 319 * calling in_pcbinshash(), or they can just use the resulting 320 * port and address to authorise the sending of a once-off packet. 321 * 322 * On error, the values of *laddrp and *lportp are not changed. 323 */ 324 int 325 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, 326 u_short *lportp, struct ucred *cred) 327 { 328 struct socket *so = inp->inp_socket; 329 unsigned short *lastport; 330 struct sockaddr_in *sin; 331 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 332 struct in_addr laddr; 333 u_short lport = 0; 334 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT); 335 int error; 336 int dorandom; 337 338 /* 339 * Because no actual state changes occur here, a global write lock on 340 * the pcbinfo isn't required. 341 */ 342 INP_INFO_LOCK_ASSERT(pcbinfo); 343 INP_LOCK_ASSERT(inp); 344 345 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */ 346 return (EADDRNOTAVAIL); 347 laddr.s_addr = *laddrp; 348 if (nam != NULL && laddr.s_addr != INADDR_ANY) 349 return (EINVAL); 350 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 351 wild = INPLOOKUP_WILDCARD; 352 if (nam == NULL) { 353 if ((error = prison_local_ip4(cred, &laddr)) != 0) 354 return (error); 355 } else { 356 sin = (struct sockaddr_in *)nam; 357 if (nam->sa_len != sizeof (*sin)) 358 return (EINVAL); 359 #ifdef notdef 360 /* 361 * We should check the family, but old programs 362 * incorrectly fail to initialize it. 363 */ 364 if (sin->sin_family != AF_INET) 365 return (EAFNOSUPPORT); 366 #endif 367 error = prison_local_ip4(cred, &sin->sin_addr); 368 if (error) 369 return (error); 370 if (sin->sin_port != *lportp) { 371 /* Don't allow the port to change. */ 372 if (*lportp != 0) 373 return (EINVAL); 374 lport = sin->sin_port; 375 } 376 /* NB: lport is left as 0 if the port isn't being changed. */ 377 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 378 /* 379 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 380 * allow complete duplication of binding if 381 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 382 * and a multicast address is bound on both 383 * new and duplicated sockets. 384 */ 385 if (so->so_options & SO_REUSEADDR) 386 reuseport = SO_REUSEADDR|SO_REUSEPORT; 387 } else if (sin->sin_addr.s_addr != INADDR_ANY) { 388 sin->sin_port = 0; /* yech... */ 389 bzero(&sin->sin_zero, sizeof(sin->sin_zero)); 390 /* 391 * Is the address a local IP address? 392 * If INP_BINDANY is set, then the socket may be bound 393 * to any endpoint address, local or not. 394 */ 395 if ((inp->inp_flags & INP_BINDANY) == 0 && 396 ifa_ifwithaddr_check((struct sockaddr *)sin) == 0) 397 return (EADDRNOTAVAIL); 398 } 399 laddr = sin->sin_addr; 400 if (lport) { 401 struct inpcb *t; 402 struct tcptw *tw; 403 404 /* GROSS */ 405 if (ntohs(lport) <= V_ipport_reservedhigh && 406 ntohs(lport) >= V_ipport_reservedlow && 407 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 408 0)) 409 return (EACCES); 410 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) && 411 priv_check_cred(inp->inp_cred, 412 PRIV_NETINET_REUSEPORT, 0) != 0) { 413 t = in_pcblookup_local(pcbinfo, sin->sin_addr, 414 lport, INPLOOKUP_WILDCARD, cred); 415 /* 416 * XXX 417 * This entire block sorely needs a rewrite. 418 */ 419 if (t && 420 ((t->inp_flags & INP_TIMEWAIT) == 0) && 421 (so->so_type != SOCK_STREAM || 422 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && 423 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 424 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 425 (t->inp_socket->so_options & 426 SO_REUSEPORT) == 0) && 427 (inp->inp_cred->cr_uid != 428 t->inp_cred->cr_uid)) 429 return (EADDRINUSE); 430 } 431 t = in_pcblookup_local(pcbinfo, sin->sin_addr, 432 lport, wild, cred); 433 if (t && (t->inp_flags & INP_TIMEWAIT)) { 434 /* 435 * XXXRW: If an incpb has had its timewait 436 * state recycled, we treat the address as 437 * being in use (for now). This is better 438 * than a panic, but not desirable. 439 */ 440 tw = intotw(inp); 441 if (tw == NULL || 442 (reuseport & tw->tw_so_options) == 0) 443 return (EADDRINUSE); 444 } else if (t && 445 (reuseport & t->inp_socket->so_options) == 0) { 446 #ifdef INET6 447 if (ntohl(sin->sin_addr.s_addr) != 448 INADDR_ANY || 449 ntohl(t->inp_laddr.s_addr) != 450 INADDR_ANY || 451 INP_SOCKAF(so) == 452 INP_SOCKAF(t->inp_socket)) 453 #endif 454 return (EADDRINUSE); 455 } 456 } 457 } 458 if (*lportp != 0) 459 lport = *lportp; 460 if (lport == 0) { 461 u_short first, last, aux; 462 int count; 463 464 if (inp->inp_flags & INP_HIGHPORT) { 465 first = V_ipport_hifirstauto; /* sysctl */ 466 last = V_ipport_hilastauto; 467 lastport = &pcbinfo->ipi_lasthi; 468 } else if (inp->inp_flags & INP_LOWPORT) { 469 error = priv_check_cred(cred, 470 PRIV_NETINET_RESERVEDPORT, 0); 471 if (error) 472 return error; 473 first = V_ipport_lowfirstauto; /* 1023 */ 474 last = V_ipport_lowlastauto; /* 600 */ 475 lastport = &pcbinfo->ipi_lastlow; 476 } else { 477 first = V_ipport_firstauto; /* sysctl */ 478 last = V_ipport_lastauto; 479 lastport = &pcbinfo->ipi_lastport; 480 } 481 /* 482 * For UDP, use random port allocation as long as the user 483 * allows it. For TCP (and as of yet unknown) connections, 484 * use random port allocation only if the user allows it AND 485 * ipport_tick() allows it. 486 */ 487 if (V_ipport_randomized && 488 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo)) 489 dorandom = 1; 490 else 491 dorandom = 0; 492 /* 493 * It makes no sense to do random port allocation if 494 * we have the only port available. 495 */ 496 if (first == last) 497 dorandom = 0; 498 /* Make sure to not include UDP packets in the count. */ 499 if (pcbinfo != &V_udbinfo) 500 V_ipport_tcpallocs++; 501 /* 502 * Instead of having two loops further down counting up or down 503 * make sure that first is always <= last and go with only one 504 * code path implementing all logic. 505 */ 506 if (first > last) { 507 aux = first; 508 first = last; 509 last = aux; 510 } 511 512 if (dorandom) 513 *lastport = first + 514 (arc4random() % (last - first)); 515 516 count = last - first; 517 518 do { 519 if (count-- < 0) /* completely used? */ 520 return (EADDRNOTAVAIL); 521 ++*lastport; 522 if (*lastport < first || *lastport > last) 523 *lastport = first; 524 lport = htons(*lastport); 525 } while (in_pcblookup_local(pcbinfo, laddr, 526 lport, wild, cred)); 527 } 528 *laddrp = laddr.s_addr; 529 *lportp = lport; 530 return (0); 531 } 532 533 /* 534 * Connect from a socket to a specified address. 535 * Both address and port must be specified in argument sin. 536 * If don't have a local address for this socket yet, 537 * then pick one. 538 */ 539 int 540 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 541 { 542 u_short lport, fport; 543 in_addr_t laddr, faddr; 544 int anonport, error; 545 546 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 547 INP_WLOCK_ASSERT(inp); 548 549 lport = inp->inp_lport; 550 laddr = inp->inp_laddr.s_addr; 551 anonport = (lport == 0); 552 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, 553 NULL, cred); 554 if (error) 555 return (error); 556 557 /* Do the initial binding of the local address if required. */ 558 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { 559 inp->inp_lport = lport; 560 inp->inp_laddr.s_addr = laddr; 561 if (in_pcbinshash(inp) != 0) { 562 inp->inp_laddr.s_addr = INADDR_ANY; 563 inp->inp_lport = 0; 564 return (EAGAIN); 565 } 566 } 567 568 /* Commit the remaining changes. */ 569 inp->inp_lport = lport; 570 inp->inp_laddr.s_addr = laddr; 571 inp->inp_faddr.s_addr = faddr; 572 inp->inp_fport = fport; 573 in_pcbrehash(inp); 574 575 if (anonport) 576 inp->inp_flags |= INP_ANONPORT; 577 return (0); 578 } 579 580 /* 581 * Do proper source address selection on an unbound socket in case 582 * of connect. Take jails into account as well. 583 */ 584 static int 585 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr, 586 struct ucred *cred) 587 { 588 struct ifaddr *ifa; 589 struct sockaddr *sa; 590 struct sockaddr_in *sin; 591 struct route sro; 592 int error; 593 594 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__)); 595 596 /* 597 * Bypass source address selection and use the primary jail IP 598 * if requested. 599 */ 600 if (cred != NULL && !prison_saddrsel_ip4(cred, laddr)) 601 return (0); 602 603 error = 0; 604 bzero(&sro, sizeof(sro)); 605 606 sin = (struct sockaddr_in *)&sro.ro_dst; 607 sin->sin_family = AF_INET; 608 sin->sin_len = sizeof(struct sockaddr_in); 609 sin->sin_addr.s_addr = faddr->s_addr; 610 611 /* 612 * If route is known our src addr is taken from the i/f, 613 * else punt. 614 * 615 * Find out route to destination. 616 */ 617 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) 618 in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum); 619 620 /* 621 * If we found a route, use the address corresponding to 622 * the outgoing interface. 623 * 624 * Otherwise assume faddr is reachable on a directly connected 625 * network and try to find a corresponding interface to take 626 * the source address from. 627 */ 628 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) { 629 struct in_ifaddr *ia; 630 struct ifnet *ifp; 631 632 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin)); 633 if (ia == NULL) 634 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0)); 635 if (ia == NULL) { 636 error = ENETUNREACH; 637 goto done; 638 } 639 640 if (cred == NULL || !prison_flag(cred, PR_IP4)) { 641 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 642 ifa_free(&ia->ia_ifa); 643 goto done; 644 } 645 646 ifp = ia->ia_ifp; 647 ifa_free(&ia->ia_ifa); 648 ia = NULL; 649 IF_ADDR_LOCK(ifp); 650 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 651 652 sa = ifa->ifa_addr; 653 if (sa->sa_family != AF_INET) 654 continue; 655 sin = (struct sockaddr_in *)sa; 656 if (prison_check_ip4(cred, &sin->sin_addr) == 0) { 657 ia = (struct in_ifaddr *)ifa; 658 break; 659 } 660 } 661 if (ia != NULL) { 662 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 663 IF_ADDR_UNLOCK(ifp); 664 goto done; 665 } 666 IF_ADDR_UNLOCK(ifp); 667 668 /* 3. As a last resort return the 'default' jail address. */ 669 error = prison_get_ip4(cred, laddr); 670 goto done; 671 } 672 673 /* 674 * If the outgoing interface on the route found is not 675 * a loopback interface, use the address from that interface. 676 * In case of jails do those three steps: 677 * 1. check if the interface address belongs to the jail. If so use it. 678 * 2. check if we have any address on the outgoing interface 679 * belonging to this jail. If so use it. 680 * 3. as a last resort return the 'default' jail address. 681 */ 682 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) { 683 struct in_ifaddr *ia; 684 struct ifnet *ifp; 685 686 /* If not jailed, use the default returned. */ 687 if (cred == NULL || !prison_flag(cred, PR_IP4)) { 688 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; 689 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 690 goto done; 691 } 692 693 /* Jailed. */ 694 /* 1. Check if the iface address belongs to the jail. */ 695 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr; 696 if (prison_check_ip4(cred, &sin->sin_addr) == 0) { 697 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; 698 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 699 goto done; 700 } 701 702 /* 703 * 2. Check if we have any address on the outgoing interface 704 * belonging to this jail. 705 */ 706 ia = NULL; 707 ifp = sro.ro_rt->rt_ifp; 708 IF_ADDR_LOCK(ifp); 709 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 710 sa = ifa->ifa_addr; 711 if (sa->sa_family != AF_INET) 712 continue; 713 sin = (struct sockaddr_in *)sa; 714 if (prison_check_ip4(cred, &sin->sin_addr) == 0) { 715 ia = (struct in_ifaddr *)ifa; 716 break; 717 } 718 } 719 if (ia != NULL) { 720 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 721 IF_ADDR_UNLOCK(ifp); 722 goto done; 723 } 724 IF_ADDR_UNLOCK(ifp); 725 726 /* 3. As a last resort return the 'default' jail address. */ 727 error = prison_get_ip4(cred, laddr); 728 goto done; 729 } 730 731 /* 732 * The outgoing interface is marked with 'loopback net', so a route 733 * to ourselves is here. 734 * Try to find the interface of the destination address and then 735 * take the address from there. That interface is not necessarily 736 * a loopback interface. 737 * In case of jails, check that it is an address of the jail 738 * and if we cannot find, fall back to the 'default' jail address. 739 */ 740 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 741 struct sockaddr_in sain; 742 struct in_ifaddr *ia; 743 744 bzero(&sain, sizeof(struct sockaddr_in)); 745 sain.sin_family = AF_INET; 746 sain.sin_len = sizeof(struct sockaddr_in); 747 sain.sin_addr.s_addr = faddr->s_addr; 748 749 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain))); 750 if (ia == NULL) 751 ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0)); 752 if (ia == NULL) 753 ia = ifatoia(ifa_ifwithaddr(sintosa(&sain))); 754 755 if (cred == NULL || !prison_flag(cred, PR_IP4)) { 756 if (ia == NULL) { 757 error = ENETUNREACH; 758 goto done; 759 } 760 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 761 ifa_free(&ia->ia_ifa); 762 goto done; 763 } 764 765 /* Jailed. */ 766 if (ia != NULL) { 767 struct ifnet *ifp; 768 769 ifp = ia->ia_ifp; 770 ifa_free(&ia->ia_ifa); 771 ia = NULL; 772 IF_ADDR_LOCK(ifp); 773 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 774 775 sa = ifa->ifa_addr; 776 if (sa->sa_family != AF_INET) 777 continue; 778 sin = (struct sockaddr_in *)sa; 779 if (prison_check_ip4(cred, 780 &sin->sin_addr) == 0) { 781 ia = (struct in_ifaddr *)ifa; 782 break; 783 } 784 } 785 if (ia != NULL) { 786 laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 787 IF_ADDR_UNLOCK(ifp); 788 goto done; 789 } 790 IF_ADDR_UNLOCK(ifp); 791 } 792 793 /* 3. As a last resort return the 'default' jail address. */ 794 error = prison_get_ip4(cred, laddr); 795 goto done; 796 } 797 798 done: 799 if (sro.ro_rt != NULL) 800 RTFREE(sro.ro_rt); 801 return (error); 802 } 803 804 /* 805 * Set up for a connect from a socket to the specified address. 806 * On entry, *laddrp and *lportp should contain the current local 807 * address and port for the PCB; these are updated to the values 808 * that should be placed in inp_laddr and inp_lport to complete 809 * the connect. 810 * 811 * On success, *faddrp and *fportp will be set to the remote address 812 * and port. These are not updated in the error case. 813 * 814 * If the operation fails because the connection already exists, 815 * *oinpp will be set to the PCB of that connection so that the 816 * caller can decide to override it. In all other cases, *oinpp 817 * is set to NULL. 818 */ 819 int 820 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam, 821 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp, 822 struct inpcb **oinpp, struct ucred *cred) 823 { 824 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 825 struct in_ifaddr *ia; 826 struct inpcb *oinp; 827 struct in_addr laddr, faddr; 828 u_short lport, fport; 829 int error; 830 831 /* 832 * Because a global state change doesn't actually occur here, a read 833 * lock is sufficient. 834 */ 835 INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo); 836 INP_LOCK_ASSERT(inp); 837 838 if (oinpp != NULL) 839 *oinpp = NULL; 840 if (nam->sa_len != sizeof (*sin)) 841 return (EINVAL); 842 if (sin->sin_family != AF_INET) 843 return (EAFNOSUPPORT); 844 if (sin->sin_port == 0) 845 return (EADDRNOTAVAIL); 846 laddr.s_addr = *laddrp; 847 lport = *lportp; 848 faddr = sin->sin_addr; 849 fport = sin->sin_port; 850 851 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) { 852 /* 853 * If the destination address is INADDR_ANY, 854 * use the primary local address. 855 * If the supplied address is INADDR_BROADCAST, 856 * and the primary interface supports broadcast, 857 * choose the broadcast address for that interface. 858 */ 859 if (faddr.s_addr == INADDR_ANY) { 860 IN_IFADDR_RLOCK(); 861 faddr = 862 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr; 863 IN_IFADDR_RUNLOCK(); 864 if (cred != NULL && 865 (error = prison_get_ip4(cred, &faddr)) != 0) 866 return (error); 867 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) { 868 IN_IFADDR_RLOCK(); 869 if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags & 870 IFF_BROADCAST) 871 faddr = satosin(&TAILQ_FIRST( 872 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr; 873 IN_IFADDR_RUNLOCK(); 874 } 875 } 876 if (laddr.s_addr == INADDR_ANY) { 877 error = in_pcbladdr(inp, &faddr, &laddr, cred); 878 /* 879 * If the destination address is multicast and an outgoing 880 * interface has been set as a multicast option, prefer the 881 * address of that interface as our source address. 882 */ 883 if (IN_MULTICAST(ntohl(faddr.s_addr)) && 884 inp->inp_moptions != NULL) { 885 struct ip_moptions *imo; 886 struct ifnet *ifp; 887 888 imo = inp->inp_moptions; 889 if (imo->imo_multicast_ifp != NULL) { 890 ifp = imo->imo_multicast_ifp; 891 IN_IFADDR_RLOCK(); 892 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 893 if ((ia->ia_ifp == ifp) && 894 (cred == NULL || 895 prison_check_ip4(cred, 896 &ia->ia_addr.sin_addr) == 0)) 897 break; 898 } 899 if (ia == NULL) 900 error = EADDRNOTAVAIL; 901 else { 902 laddr = ia->ia_addr.sin_addr; 903 error = 0; 904 } 905 IN_IFADDR_RUNLOCK(); 906 } 907 } 908 if (error) 909 return (error); 910 } 911 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport, 912 0, NULL); 913 if (oinp != NULL) { 914 if (oinpp != NULL) 915 *oinpp = oinp; 916 return (EADDRINUSE); 917 } 918 if (lport == 0) { 919 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, 920 cred); 921 if (error) 922 return (error); 923 } 924 *laddrp = laddr.s_addr; 925 *lportp = lport; 926 *faddrp = faddr.s_addr; 927 *fportp = fport; 928 return (0); 929 } 930 931 void 932 in_pcbdisconnect(struct inpcb *inp) 933 { 934 935 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 936 INP_WLOCK_ASSERT(inp); 937 938 inp->inp_faddr.s_addr = INADDR_ANY; 939 inp->inp_fport = 0; 940 in_pcbrehash(inp); 941 } 942 943 /* 944 * in_pcbdetach() is responsibe for disassociating a socket from an inpcb. 945 * For most protocols, this will be invoked immediately prior to calling 946 * in_pcbfree(). However, with TCP the inpcb may significantly outlive the 947 * socket, in which case in_pcbfree() is deferred. 948 */ 949 void 950 in_pcbdetach(struct inpcb *inp) 951 { 952 953 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__)); 954 955 inp->inp_socket->so_pcb = NULL; 956 inp->inp_socket = NULL; 957 } 958 959 /* 960 * in_pcbfree_internal() frees an inpcb that has been detached from its 961 * socket, and whose reference count has reached 0. It will also remove the 962 * inpcb from any global lists it might remain on. 963 */ 964 static void 965 in_pcbfree_internal(struct inpcb *inp) 966 { 967 struct inpcbinfo *ipi = inp->inp_pcbinfo; 968 969 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); 970 KASSERT(inp->inp_refcount == 0, ("%s: refcount !0", __func__)); 971 972 INP_INFO_WLOCK_ASSERT(ipi); 973 INP_WLOCK_ASSERT(inp); 974 975 #ifdef IPSEC 976 if (inp->inp_sp != NULL) 977 ipsec_delete_pcbpolicy(inp); 978 #endif /* IPSEC */ 979 inp->inp_gencnt = ++ipi->ipi_gencnt; 980 in_pcbremlists(inp); 981 #ifdef INET6 982 if (inp->inp_vflag & INP_IPV6PROTO) { 983 ip6_freepcbopts(inp->in6p_outputopts); 984 if (inp->in6p_moptions != NULL) 985 ip6_freemoptions(inp->in6p_moptions); 986 } 987 #endif 988 if (inp->inp_options) 989 (void)m_free(inp->inp_options); 990 if (inp->inp_moptions != NULL) 991 inp_freemoptions(inp->inp_moptions); 992 inp->inp_vflag = 0; 993 crfree(inp->inp_cred); 994 995 #ifdef MAC 996 mac_inpcb_destroy(inp); 997 #endif 998 INP_WUNLOCK(inp); 999 uma_zfree(ipi->ipi_zone, inp); 1000 } 1001 1002 /* 1003 * in_pcbref() bumps the reference count on an inpcb in order to maintain 1004 * stability of an inpcb pointer despite the inpcb lock being released. This 1005 * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded, 1006 * but where the inpcb lock is already held. 1007 * 1008 * While the inpcb will not be freed, releasing the inpcb lock means that the 1009 * connection's state may change, so the caller should be careful to 1010 * revalidate any cached state on reacquiring the lock. Drop the reference 1011 * using in_pcbrele(). 1012 */ 1013 void 1014 in_pcbref(struct inpcb *inp) 1015 { 1016 1017 INP_WLOCK_ASSERT(inp); 1018 1019 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); 1020 1021 inp->inp_refcount++; 1022 } 1023 1024 /* 1025 * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to 1026 * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we 1027 * return a flag indicating whether or not the inpcb remains valid. If it is 1028 * valid, we return with the inpcb lock held. 1029 */ 1030 int 1031 in_pcbrele(struct inpcb *inp) 1032 { 1033 #ifdef INVARIANTS 1034 struct inpcbinfo *ipi = inp->inp_pcbinfo; 1035 #endif 1036 1037 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); 1038 1039 INP_INFO_WLOCK_ASSERT(ipi); 1040 INP_WLOCK_ASSERT(inp); 1041 1042 inp->inp_refcount--; 1043 if (inp->inp_refcount > 0) 1044 return (0); 1045 in_pcbfree_internal(inp); 1046 return (1); 1047 } 1048 1049 /* 1050 * Unconditionally schedule an inpcb to be freed by decrementing its 1051 * reference count, which should occur only after the inpcb has been detached 1052 * from its socket. If another thread holds a temporary reference (acquired 1053 * using in_pcbref()) then the free is deferred until that reference is 1054 * released using in_pcbrele(), but the inpcb is still unlocked. 1055 */ 1056 void 1057 in_pcbfree(struct inpcb *inp) 1058 { 1059 #ifdef INVARIANTS 1060 struct inpcbinfo *ipi = inp->inp_pcbinfo; 1061 #endif 1062 1063 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", 1064 __func__)); 1065 1066 INP_INFO_WLOCK_ASSERT(ipi); 1067 INP_WLOCK_ASSERT(inp); 1068 1069 if (!in_pcbrele(inp)) 1070 INP_WUNLOCK(inp); 1071 } 1072 1073 /* 1074 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and 1075 * port reservation, and preventing it from being returned by inpcb lookups. 1076 * 1077 * It is used by TCP to mark an inpcb as unused and avoid future packet 1078 * delivery or event notification when a socket remains open but TCP has 1079 * closed. This might occur as a result of a shutdown()-initiated TCP close 1080 * or a RST on the wire, and allows the port binding to be reused while still 1081 * maintaining the invariant that so_pcb always points to a valid inpcb until 1082 * in_pcbdetach(). 1083 * 1084 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by 1085 * in_pcbnotifyall() and in_pcbpurgeif0()? 1086 */ 1087 void 1088 in_pcbdrop(struct inpcb *inp) 1089 { 1090 1091 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 1092 INP_WLOCK_ASSERT(inp); 1093 1094 inp->inp_flags |= INP_DROPPED; 1095 if (inp->inp_flags & INP_INHASHLIST) { 1096 struct inpcbport *phd = inp->inp_phd; 1097 1098 LIST_REMOVE(inp, inp_hash); 1099 LIST_REMOVE(inp, inp_portlist); 1100 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 1101 LIST_REMOVE(phd, phd_hash); 1102 free(phd, M_PCB); 1103 } 1104 inp->inp_flags &= ~INP_INHASHLIST; 1105 } 1106 } 1107 1108 /* 1109 * Common routines to return the socket addresses associated with inpcbs. 1110 */ 1111 struct sockaddr * 1112 in_sockaddr(in_port_t port, struct in_addr *addr_p) 1113 { 1114 struct sockaddr_in *sin; 1115 1116 sin = malloc(sizeof *sin, M_SONAME, 1117 M_WAITOK | M_ZERO); 1118 sin->sin_family = AF_INET; 1119 sin->sin_len = sizeof(*sin); 1120 sin->sin_addr = *addr_p; 1121 sin->sin_port = port; 1122 1123 return (struct sockaddr *)sin; 1124 } 1125 1126 int 1127 in_getsockaddr(struct socket *so, struct sockaddr **nam) 1128 { 1129 struct inpcb *inp; 1130 struct in_addr addr; 1131 in_port_t port; 1132 1133 inp = sotoinpcb(so); 1134 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL")); 1135 1136 INP_RLOCK(inp); 1137 port = inp->inp_lport; 1138 addr = inp->inp_laddr; 1139 INP_RUNLOCK(inp); 1140 1141 *nam = in_sockaddr(port, &addr); 1142 return 0; 1143 } 1144 1145 int 1146 in_getpeeraddr(struct socket *so, struct sockaddr **nam) 1147 { 1148 struct inpcb *inp; 1149 struct in_addr addr; 1150 in_port_t port; 1151 1152 inp = sotoinpcb(so); 1153 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL")); 1154 1155 INP_RLOCK(inp); 1156 port = inp->inp_fport; 1157 addr = inp->inp_faddr; 1158 INP_RUNLOCK(inp); 1159 1160 *nam = in_sockaddr(port, &addr); 1161 return 0; 1162 } 1163 1164 void 1165 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno, 1166 struct inpcb *(*notify)(struct inpcb *, int)) 1167 { 1168 struct inpcb *inp, *inp_temp; 1169 1170 INP_INFO_WLOCK(pcbinfo); 1171 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { 1172 INP_WLOCK(inp); 1173 #ifdef INET6 1174 if ((inp->inp_vflag & INP_IPV4) == 0) { 1175 INP_WUNLOCK(inp); 1176 continue; 1177 } 1178 #endif 1179 if (inp->inp_faddr.s_addr != faddr.s_addr || 1180 inp->inp_socket == NULL) { 1181 INP_WUNLOCK(inp); 1182 continue; 1183 } 1184 if ((*notify)(inp, errno)) 1185 INP_WUNLOCK(inp); 1186 } 1187 INP_INFO_WUNLOCK(pcbinfo); 1188 } 1189 1190 void 1191 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) 1192 { 1193 struct inpcb *inp; 1194 struct ip_moptions *imo; 1195 int i, gap; 1196 1197 INP_INFO_RLOCK(pcbinfo); 1198 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) { 1199 INP_WLOCK(inp); 1200 imo = inp->inp_moptions; 1201 if ((inp->inp_vflag & INP_IPV4) && 1202 imo != NULL) { 1203 /* 1204 * Unselect the outgoing interface if it is being 1205 * detached. 1206 */ 1207 if (imo->imo_multicast_ifp == ifp) 1208 imo->imo_multicast_ifp = NULL; 1209 1210 /* 1211 * Drop multicast group membership if we joined 1212 * through the interface being detached. 1213 */ 1214 for (i = 0, gap = 0; i < imo->imo_num_memberships; 1215 i++) { 1216 if (imo->imo_membership[i]->inm_ifp == ifp) { 1217 in_delmulti(imo->imo_membership[i]); 1218 gap++; 1219 } else if (gap != 0) 1220 imo->imo_membership[i - gap] = 1221 imo->imo_membership[i]; 1222 } 1223 imo->imo_num_memberships -= gap; 1224 } 1225 INP_WUNLOCK(inp); 1226 } 1227 INP_INFO_RUNLOCK(pcbinfo); 1228 } 1229 1230 /* 1231 * Lookup a PCB based on the local address and port. 1232 */ 1233 #define INP_LOOKUP_MAPPED_PCB_COST 3 1234 struct inpcb * 1235 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr, 1236 u_short lport, int wild_okay, struct ucred *cred) 1237 { 1238 struct inpcb *inp; 1239 #ifdef INET6 1240 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST; 1241 #else 1242 int matchwild = 3; 1243 #endif 1244 int wildcard; 1245 1246 INP_INFO_LOCK_ASSERT(pcbinfo); 1247 1248 if (!wild_okay) { 1249 struct inpcbhead *head; 1250 /* 1251 * Look for an unconnected (wildcard foreign addr) PCB that 1252 * matches the local address and port we're looking for. 1253 */ 1254 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 1255 0, pcbinfo->ipi_hashmask)]; 1256 LIST_FOREACH(inp, head, inp_hash) { 1257 #ifdef INET6 1258 /* XXX inp locking */ 1259 if ((inp->inp_vflag & INP_IPV4) == 0) 1260 continue; 1261 #endif 1262 if (inp->inp_faddr.s_addr == INADDR_ANY && 1263 inp->inp_laddr.s_addr == laddr.s_addr && 1264 inp->inp_lport == lport) { 1265 /* 1266 * Found? 1267 */ 1268 if (cred == NULL || 1269 prison_equal_ip4(cred->cr_prison, 1270 inp->inp_cred->cr_prison)) 1271 return (inp); 1272 } 1273 } 1274 /* 1275 * Not found. 1276 */ 1277 return (NULL); 1278 } else { 1279 struct inpcbporthead *porthash; 1280 struct inpcbport *phd; 1281 struct inpcb *match = NULL; 1282 /* 1283 * Best fit PCB lookup. 1284 * 1285 * First see if this local port is in use by looking on the 1286 * port hash list. 1287 */ 1288 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, 1289 pcbinfo->ipi_porthashmask)]; 1290 LIST_FOREACH(phd, porthash, phd_hash) { 1291 if (phd->phd_port == lport) 1292 break; 1293 } 1294 if (phd != NULL) { 1295 /* 1296 * Port is in use by one or more PCBs. Look for best 1297 * fit. 1298 */ 1299 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 1300 wildcard = 0; 1301 if (cred != NULL && 1302 !prison_equal_ip4(inp->inp_cred->cr_prison, 1303 cred->cr_prison)) 1304 continue; 1305 #ifdef INET6 1306 /* XXX inp locking */ 1307 if ((inp->inp_vflag & INP_IPV4) == 0) 1308 continue; 1309 /* 1310 * We never select the PCB that has 1311 * INP_IPV6 flag and is bound to :: if 1312 * we have another PCB which is bound 1313 * to 0.0.0.0. If a PCB has the 1314 * INP_IPV6 flag, then we set its cost 1315 * higher than IPv4 only PCBs. 1316 * 1317 * Note that the case only happens 1318 * when a socket is bound to ::, under 1319 * the condition that the use of the 1320 * mapped address is allowed. 1321 */ 1322 if ((inp->inp_vflag & INP_IPV6) != 0) 1323 wildcard += INP_LOOKUP_MAPPED_PCB_COST; 1324 #endif 1325 if (inp->inp_faddr.s_addr != INADDR_ANY) 1326 wildcard++; 1327 if (inp->inp_laddr.s_addr != INADDR_ANY) { 1328 if (laddr.s_addr == INADDR_ANY) 1329 wildcard++; 1330 else if (inp->inp_laddr.s_addr != laddr.s_addr) 1331 continue; 1332 } else { 1333 if (laddr.s_addr != INADDR_ANY) 1334 wildcard++; 1335 } 1336 if (wildcard < matchwild) { 1337 match = inp; 1338 matchwild = wildcard; 1339 if (matchwild == 0) 1340 break; 1341 } 1342 } 1343 } 1344 return (match); 1345 } 1346 } 1347 #undef INP_LOOKUP_MAPPED_PCB_COST 1348 1349 /* 1350 * Lookup PCB in hash list. 1351 */ 1352 struct inpcb * 1353 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, 1354 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard, 1355 struct ifnet *ifp) 1356 { 1357 struct inpcbhead *head; 1358 struct inpcb *inp, *tmpinp; 1359 u_short fport = fport_arg, lport = lport_arg; 1360 1361 INP_INFO_LOCK_ASSERT(pcbinfo); 1362 1363 /* 1364 * First look for an exact match. 1365 */ 1366 tmpinp = NULL; 1367 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, 1368 pcbinfo->ipi_hashmask)]; 1369 LIST_FOREACH(inp, head, inp_hash) { 1370 #ifdef INET6 1371 /* XXX inp locking */ 1372 if ((inp->inp_vflag & INP_IPV4) == 0) 1373 continue; 1374 #endif 1375 if (inp->inp_faddr.s_addr == faddr.s_addr && 1376 inp->inp_laddr.s_addr == laddr.s_addr && 1377 inp->inp_fport == fport && 1378 inp->inp_lport == lport) { 1379 /* 1380 * XXX We should be able to directly return 1381 * the inp here, without any checks. 1382 * Well unless both bound with SO_REUSEPORT? 1383 */ 1384 if (prison_flag(inp->inp_cred, PR_IP4)) 1385 return (inp); 1386 if (tmpinp == NULL) 1387 tmpinp = inp; 1388 } 1389 } 1390 if (tmpinp != NULL) 1391 return (tmpinp); 1392 1393 /* 1394 * Then look for a wildcard match, if requested. 1395 */ 1396 if (wildcard == INPLOOKUP_WILDCARD) { 1397 struct inpcb *local_wild = NULL, *local_exact = NULL; 1398 #ifdef INET6 1399 struct inpcb *local_wild_mapped = NULL; 1400 #endif 1401 struct inpcb *jail_wild = NULL; 1402 int injail; 1403 1404 /* 1405 * Order of socket selection - we always prefer jails. 1406 * 1. jailed, non-wild. 1407 * 2. jailed, wild. 1408 * 3. non-jailed, non-wild. 1409 * 4. non-jailed, wild. 1410 */ 1411 1412 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 1413 0, pcbinfo->ipi_hashmask)]; 1414 LIST_FOREACH(inp, head, inp_hash) { 1415 #ifdef INET6 1416 /* XXX inp locking */ 1417 if ((inp->inp_vflag & INP_IPV4) == 0) 1418 continue; 1419 #endif 1420 if (inp->inp_faddr.s_addr != INADDR_ANY || 1421 inp->inp_lport != lport) 1422 continue; 1423 1424 /* XXX inp locking */ 1425 if (ifp && ifp->if_type == IFT_FAITH && 1426 (inp->inp_flags & INP_FAITH) == 0) 1427 continue; 1428 1429 injail = prison_flag(inp->inp_cred, PR_IP4); 1430 if (injail) { 1431 if (prison_check_ip4(inp->inp_cred, 1432 &laddr) != 0) 1433 continue; 1434 } else { 1435 if (local_exact != NULL) 1436 continue; 1437 } 1438 1439 if (inp->inp_laddr.s_addr == laddr.s_addr) { 1440 if (injail) 1441 return (inp); 1442 else 1443 local_exact = inp; 1444 } else if (inp->inp_laddr.s_addr == INADDR_ANY) { 1445 #ifdef INET6 1446 /* XXX inp locking, NULL check */ 1447 if (inp->inp_vflag & INP_IPV6PROTO) 1448 local_wild_mapped = inp; 1449 else 1450 #endif /* INET6 */ 1451 if (injail) 1452 jail_wild = inp; 1453 else 1454 local_wild = inp; 1455 } 1456 } /* LIST_FOREACH */ 1457 if (jail_wild != NULL) 1458 return (jail_wild); 1459 if (local_exact != NULL) 1460 return (local_exact); 1461 if (local_wild != NULL) 1462 return (local_wild); 1463 #ifdef INET6 1464 if (local_wild_mapped != NULL) 1465 return (local_wild_mapped); 1466 #endif /* defined(INET6) */ 1467 } /* if (wildcard == INPLOOKUP_WILDCARD) */ 1468 1469 return (NULL); 1470 } 1471 1472 /* 1473 * Insert PCB onto various hash lists. 1474 */ 1475 int 1476 in_pcbinshash(struct inpcb *inp) 1477 { 1478 struct inpcbhead *pcbhash; 1479 struct inpcbporthead *pcbporthash; 1480 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1481 struct inpcbport *phd; 1482 u_int32_t hashkey_faddr; 1483 1484 INP_INFO_WLOCK_ASSERT(pcbinfo); 1485 INP_WLOCK_ASSERT(inp); 1486 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0, 1487 ("in_pcbinshash: INP_INHASHLIST")); 1488 1489 #ifdef INET6 1490 if (inp->inp_vflag & INP_IPV6) 1491 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1492 else 1493 #endif /* INET6 */ 1494 hashkey_faddr = inp->inp_faddr.s_addr; 1495 1496 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, 1497 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; 1498 1499 pcbporthash = &pcbinfo->ipi_porthashbase[ 1500 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)]; 1501 1502 /* 1503 * Go through port list and look for a head for this lport. 1504 */ 1505 LIST_FOREACH(phd, pcbporthash, phd_hash) { 1506 if (phd->phd_port == inp->inp_lport) 1507 break; 1508 } 1509 /* 1510 * If none exists, malloc one and tack it on. 1511 */ 1512 if (phd == NULL) { 1513 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT); 1514 if (phd == NULL) { 1515 return (ENOBUFS); /* XXX */ 1516 } 1517 phd->phd_port = inp->inp_lport; 1518 LIST_INIT(&phd->phd_pcblist); 1519 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 1520 } 1521 inp->inp_phd = phd; 1522 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 1523 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 1524 inp->inp_flags |= INP_INHASHLIST; 1525 return (0); 1526 } 1527 1528 /* 1529 * Move PCB to the proper hash bucket when { faddr, fport } have been 1530 * changed. NOTE: This does not handle the case of the lport changing (the 1531 * hashed port list would have to be updated as well), so the lport must 1532 * not change after in_pcbinshash() has been called. 1533 */ 1534 void 1535 in_pcbrehash(struct inpcb *inp) 1536 { 1537 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1538 struct inpcbhead *head; 1539 u_int32_t hashkey_faddr; 1540 1541 INP_INFO_WLOCK_ASSERT(pcbinfo); 1542 INP_WLOCK_ASSERT(inp); 1543 KASSERT(inp->inp_flags & INP_INHASHLIST, 1544 ("in_pcbrehash: !INP_INHASHLIST")); 1545 1546 #ifdef INET6 1547 if (inp->inp_vflag & INP_IPV6) 1548 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 1549 else 1550 #endif /* INET6 */ 1551 hashkey_faddr = inp->inp_faddr.s_addr; 1552 1553 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, 1554 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; 1555 1556 LIST_REMOVE(inp, inp_hash); 1557 LIST_INSERT_HEAD(head, inp, inp_hash); 1558 } 1559 1560 /* 1561 * Remove PCB from various lists. 1562 */ 1563 static void 1564 in_pcbremlists(struct inpcb *inp) 1565 { 1566 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 1567 1568 INP_INFO_WLOCK_ASSERT(pcbinfo); 1569 INP_WLOCK_ASSERT(inp); 1570 1571 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 1572 if (inp->inp_flags & INP_INHASHLIST) { 1573 struct inpcbport *phd = inp->inp_phd; 1574 1575 LIST_REMOVE(inp, inp_hash); 1576 LIST_REMOVE(inp, inp_portlist); 1577 if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 1578 LIST_REMOVE(phd, phd_hash); 1579 free(phd, M_PCB); 1580 } 1581 inp->inp_flags &= ~INP_INHASHLIST; 1582 } 1583 LIST_REMOVE(inp, inp_list); 1584 pcbinfo->ipi_count--; 1585 } 1586 1587 /* 1588 * A set label operation has occurred at the socket layer, propagate the 1589 * label change into the in_pcb for the socket. 1590 */ 1591 void 1592 in_pcbsosetlabel(struct socket *so) 1593 { 1594 #ifdef MAC 1595 struct inpcb *inp; 1596 1597 inp = sotoinpcb(so); 1598 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL")); 1599 1600 INP_WLOCK(inp); 1601 SOCK_LOCK(so); 1602 mac_inpcb_sosetlabel(so, inp); 1603 SOCK_UNLOCK(so); 1604 INP_WUNLOCK(inp); 1605 #endif 1606 } 1607 1608 /* 1609 * ipport_tick runs once per second, determining if random port allocation 1610 * should be continued. If more than ipport_randomcps ports have been 1611 * allocated in the last second, then we return to sequential port 1612 * allocation. We return to random allocation only once we drop below 1613 * ipport_randomcps for at least ipport_randomtime seconds. 1614 */ 1615 void 1616 ipport_tick(void *xtp) 1617 { 1618 VNET_ITERATOR_DECL(vnet_iter); 1619 1620 VNET_LIST_RLOCK_NOSLEEP(); 1621 VNET_FOREACH(vnet_iter) { 1622 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */ 1623 if (V_ipport_tcpallocs <= 1624 V_ipport_tcplastcount + V_ipport_randomcps) { 1625 if (V_ipport_stoprandom > 0) 1626 V_ipport_stoprandom--; 1627 } else 1628 V_ipport_stoprandom = V_ipport_randomtime; 1629 V_ipport_tcplastcount = V_ipport_tcpallocs; 1630 CURVNET_RESTORE(); 1631 } 1632 VNET_LIST_RUNLOCK_NOSLEEP(); 1633 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL); 1634 } 1635 1636 void 1637 inp_wlock(struct inpcb *inp) 1638 { 1639 1640 INP_WLOCK(inp); 1641 } 1642 1643 void 1644 inp_wunlock(struct inpcb *inp) 1645 { 1646 1647 INP_WUNLOCK(inp); 1648 } 1649 1650 void 1651 inp_rlock(struct inpcb *inp) 1652 { 1653 1654 INP_RLOCK(inp); 1655 } 1656 1657 void 1658 inp_runlock(struct inpcb *inp) 1659 { 1660 1661 INP_RUNLOCK(inp); 1662 } 1663 1664 #ifdef INVARIANTS 1665 void 1666 inp_lock_assert(struct inpcb *inp) 1667 { 1668 1669 INP_WLOCK_ASSERT(inp); 1670 } 1671 1672 void 1673 inp_unlock_assert(struct inpcb *inp) 1674 { 1675 1676 INP_UNLOCK_ASSERT(inp); 1677 } 1678 #endif 1679 1680 void 1681 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg) 1682 { 1683 struct inpcb *inp; 1684 1685 INP_INFO_RLOCK(&V_tcbinfo); 1686 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) { 1687 INP_WLOCK(inp); 1688 func(inp, arg); 1689 INP_WUNLOCK(inp); 1690 } 1691 INP_INFO_RUNLOCK(&V_tcbinfo); 1692 } 1693 1694 struct socket * 1695 inp_inpcbtosocket(struct inpcb *inp) 1696 { 1697 1698 INP_WLOCK_ASSERT(inp); 1699 return (inp->inp_socket); 1700 } 1701 1702 struct tcpcb * 1703 inp_inpcbtotcpcb(struct inpcb *inp) 1704 { 1705 1706 INP_WLOCK_ASSERT(inp); 1707 return ((struct tcpcb *)inp->inp_ppcb); 1708 } 1709 1710 int 1711 inp_ip_tos_get(const struct inpcb *inp) 1712 { 1713 1714 return (inp->inp_ip_tos); 1715 } 1716 1717 void 1718 inp_ip_tos_set(struct inpcb *inp, int val) 1719 { 1720 1721 inp->inp_ip_tos = val; 1722 } 1723 1724 void 1725 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, 1726 uint32_t *faddr, uint16_t *fp) 1727 { 1728 1729 INP_LOCK_ASSERT(inp); 1730 *laddr = inp->inp_laddr.s_addr; 1731 *faddr = inp->inp_faddr.s_addr; 1732 *lp = inp->inp_lport; 1733 *fp = inp->inp_fport; 1734 } 1735 1736 struct inpcb * 1737 so_sotoinpcb(struct socket *so) 1738 { 1739 1740 return (sotoinpcb(so)); 1741 } 1742 1743 struct tcpcb * 1744 so_sototcpcb(struct socket *so) 1745 { 1746 1747 return (sototcpcb(so)); 1748 } 1749 1750 #ifdef DDB 1751 static void 1752 db_print_indent(int indent) 1753 { 1754 int i; 1755 1756 for (i = 0; i < indent; i++) 1757 db_printf(" "); 1758 } 1759 1760 static void 1761 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent) 1762 { 1763 char faddr_str[48], laddr_str[48]; 1764 1765 db_print_indent(indent); 1766 db_printf("%s at %p\n", name, inc); 1767 1768 indent += 2; 1769 1770 #ifdef INET6 1771 if (inc->inc_flags & INC_ISIPV6) { 1772 /* IPv6. */ 1773 ip6_sprintf(laddr_str, &inc->inc6_laddr); 1774 ip6_sprintf(faddr_str, &inc->inc6_faddr); 1775 } else { 1776 #endif 1777 /* IPv4. */ 1778 inet_ntoa_r(inc->inc_laddr, laddr_str); 1779 inet_ntoa_r(inc->inc_faddr, faddr_str); 1780 #ifdef INET6 1781 } 1782 #endif 1783 db_print_indent(indent); 1784 db_printf("inc_laddr %s inc_lport %u\n", laddr_str, 1785 ntohs(inc->inc_lport)); 1786 db_print_indent(indent); 1787 db_printf("inc_faddr %s inc_fport %u\n", faddr_str, 1788 ntohs(inc->inc_fport)); 1789 } 1790 1791 static void 1792 db_print_inpflags(int inp_flags) 1793 { 1794 int comma; 1795 1796 comma = 0; 1797 if (inp_flags & INP_RECVOPTS) { 1798 db_printf("%sINP_RECVOPTS", comma ? ", " : ""); 1799 comma = 1; 1800 } 1801 if (inp_flags & INP_RECVRETOPTS) { 1802 db_printf("%sINP_RECVRETOPTS", comma ? ", " : ""); 1803 comma = 1; 1804 } 1805 if (inp_flags & INP_RECVDSTADDR) { 1806 db_printf("%sINP_RECVDSTADDR", comma ? ", " : ""); 1807 comma = 1; 1808 } 1809 if (inp_flags & INP_HDRINCL) { 1810 db_printf("%sINP_HDRINCL", comma ? ", " : ""); 1811 comma = 1; 1812 } 1813 if (inp_flags & INP_HIGHPORT) { 1814 db_printf("%sINP_HIGHPORT", comma ? ", " : ""); 1815 comma = 1; 1816 } 1817 if (inp_flags & INP_LOWPORT) { 1818 db_printf("%sINP_LOWPORT", comma ? ", " : ""); 1819 comma = 1; 1820 } 1821 if (inp_flags & INP_ANONPORT) { 1822 db_printf("%sINP_ANONPORT", comma ? ", " : ""); 1823 comma = 1; 1824 } 1825 if (inp_flags & INP_RECVIF) { 1826 db_printf("%sINP_RECVIF", comma ? ", " : ""); 1827 comma = 1; 1828 } 1829 if (inp_flags & INP_MTUDISC) { 1830 db_printf("%sINP_MTUDISC", comma ? ", " : ""); 1831 comma = 1; 1832 } 1833 if (inp_flags & INP_FAITH) { 1834 db_printf("%sINP_FAITH", comma ? ", " : ""); 1835 comma = 1; 1836 } 1837 if (inp_flags & INP_RECVTTL) { 1838 db_printf("%sINP_RECVTTL", comma ? ", " : ""); 1839 comma = 1; 1840 } 1841 if (inp_flags & INP_DONTFRAG) { 1842 db_printf("%sINP_DONTFRAG", comma ? ", " : ""); 1843 comma = 1; 1844 } 1845 if (inp_flags & IN6P_IPV6_V6ONLY) { 1846 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : ""); 1847 comma = 1; 1848 } 1849 if (inp_flags & IN6P_PKTINFO) { 1850 db_printf("%sIN6P_PKTINFO", comma ? ", " : ""); 1851 comma = 1; 1852 } 1853 if (inp_flags & IN6P_HOPLIMIT) { 1854 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : ""); 1855 comma = 1; 1856 } 1857 if (inp_flags & IN6P_HOPOPTS) { 1858 db_printf("%sIN6P_HOPOPTS", comma ? ", " : ""); 1859 comma = 1; 1860 } 1861 if (inp_flags & IN6P_DSTOPTS) { 1862 db_printf("%sIN6P_DSTOPTS", comma ? ", " : ""); 1863 comma = 1; 1864 } 1865 if (inp_flags & IN6P_RTHDR) { 1866 db_printf("%sIN6P_RTHDR", comma ? ", " : ""); 1867 comma = 1; 1868 } 1869 if (inp_flags & IN6P_RTHDRDSTOPTS) { 1870 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : ""); 1871 comma = 1; 1872 } 1873 if (inp_flags & IN6P_TCLASS) { 1874 db_printf("%sIN6P_TCLASS", comma ? ", " : ""); 1875 comma = 1; 1876 } 1877 if (inp_flags & IN6P_AUTOFLOWLABEL) { 1878 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : ""); 1879 comma = 1; 1880 } 1881 if (inp_flags & INP_TIMEWAIT) { 1882 db_printf("%sINP_TIMEWAIT", comma ? ", " : ""); 1883 comma = 1; 1884 } 1885 if (inp_flags & INP_ONESBCAST) { 1886 db_printf("%sINP_ONESBCAST", comma ? ", " : ""); 1887 comma = 1; 1888 } 1889 if (inp_flags & INP_DROPPED) { 1890 db_printf("%sINP_DROPPED", comma ? ", " : ""); 1891 comma = 1; 1892 } 1893 if (inp_flags & INP_SOCKREF) { 1894 db_printf("%sINP_SOCKREF", comma ? ", " : ""); 1895 comma = 1; 1896 } 1897 if (inp_flags & IN6P_RFC2292) { 1898 db_printf("%sIN6P_RFC2292", comma ? ", " : ""); 1899 comma = 1; 1900 } 1901 if (inp_flags & IN6P_MTU) { 1902 db_printf("IN6P_MTU%s", comma ? ", " : ""); 1903 comma = 1; 1904 } 1905 } 1906 1907 static void 1908 db_print_inpvflag(u_char inp_vflag) 1909 { 1910 int comma; 1911 1912 comma = 0; 1913 if (inp_vflag & INP_IPV4) { 1914 db_printf("%sINP_IPV4", comma ? ", " : ""); 1915 comma = 1; 1916 } 1917 if (inp_vflag & INP_IPV6) { 1918 db_printf("%sINP_IPV6", comma ? ", " : ""); 1919 comma = 1; 1920 } 1921 if (inp_vflag & INP_IPV6PROTO) { 1922 db_printf("%sINP_IPV6PROTO", comma ? ", " : ""); 1923 comma = 1; 1924 } 1925 } 1926 1927 static void 1928 db_print_inpcb(struct inpcb *inp, const char *name, int indent) 1929 { 1930 1931 db_print_indent(indent); 1932 db_printf("%s at %p\n", name, inp); 1933 1934 indent += 2; 1935 1936 db_print_indent(indent); 1937 db_printf("inp_flow: 0x%x\n", inp->inp_flow); 1938 1939 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent); 1940 1941 db_print_indent(indent); 1942 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n", 1943 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket); 1944 1945 db_print_indent(indent); 1946 db_printf("inp_label: %p inp_flags: 0x%x (", 1947 inp->inp_label, inp->inp_flags); 1948 db_print_inpflags(inp->inp_flags); 1949 db_printf(")\n"); 1950 1951 db_print_indent(indent); 1952 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp, 1953 inp->inp_vflag); 1954 db_print_inpvflag(inp->inp_vflag); 1955 db_printf(")\n"); 1956 1957 db_print_indent(indent); 1958 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n", 1959 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl); 1960 1961 db_print_indent(indent); 1962 #ifdef INET6 1963 if (inp->inp_vflag & INP_IPV6) { 1964 db_printf("in6p_options: %p in6p_outputopts: %p " 1965 "in6p_moptions: %p\n", inp->in6p_options, 1966 inp->in6p_outputopts, inp->in6p_moptions); 1967 db_printf("in6p_icmp6filt: %p in6p_cksum %d " 1968 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum, 1969 inp->in6p_hops); 1970 } else 1971 #endif 1972 { 1973 db_printf("inp_ip_tos: %d inp_ip_options: %p " 1974 "inp_ip_moptions: %p\n", inp->inp_ip_tos, 1975 inp->inp_options, inp->inp_moptions); 1976 } 1977 1978 db_print_indent(indent); 1979 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd, 1980 (uintmax_t)inp->inp_gencnt); 1981 } 1982 1983 DB_SHOW_COMMAND(inpcb, db_show_inpcb) 1984 { 1985 struct inpcb *inp; 1986 1987 if (!have_addr) { 1988 db_printf("usage: show inpcb <addr>\n"); 1989 return; 1990 } 1991 inp = (struct inpcb *)addr; 1992 1993 db_print_inpcb(inp, "inpcb", 0); 1994 } 1995 #endif 1996