1 /* 2 * Copyright (c) 1982, 1986, 1991, 1993, 1995 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 34 * $FreeBSD$ 35 */ 36 37 #include "opt_ipsec.h" 38 #include "opt_inet6.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/domain.h> 45 #include <sys/protosw.h> 46 #include <sys/socket.h> 47 #include <sys/socketvar.h> 48 #include <sys/proc.h> 49 #include <sys/jail.h> 50 #include <sys/kernel.h> 51 #include <sys/sysctl.h> 52 53 #include <machine/limits.h> 54 55 #include <vm/vm_zone.h> 56 57 #include <net/if.h> 58 #include <net/if_types.h> 59 #include <net/route.h> 60 61 #include <netinet/in.h> 62 #include <netinet/in_pcb.h> 63 #include <netinet/in_var.h> 64 #include <netinet/ip_var.h> 65 #ifdef INET6 66 #include <netinet/ip6.h> 67 #include <netinet6/ip6_var.h> 68 #endif /* INET6 */ 69 70 #include "faith.h" 71 72 #ifdef IPSEC 73 #include <netinet6/ipsec.h> 74 #include <netkey/key.h> 75 #include <netkey/key_debug.h> 76 #endif /* IPSEC */ 77 78 struct in_addr zeroin_addr; 79 80 static void in_rtchange __P((struct inpcb *, int)); 81 82 /* 83 * These configure the range of local port addresses assigned to 84 * "unspecified" outgoing connections/packets/whatever. 85 */ 86 int ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */ 87 int ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */ 88 int ipport_firstauto = IPPORT_RESERVED; /* 1024 */ 89 int ipport_lastauto = IPPORT_USERRESERVED; /* 5000 */ 90 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */ 91 int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */ 92 93 #define RANGECHK(var, min, max) \ 94 if ((var) < (min)) { (var) = (min); } \ 95 else if ((var) > (max)) { (var) = (max); } 96 97 static int 98 sysctl_net_ipport_check SYSCTL_HANDLER_ARGS 99 { 100 int error = sysctl_handle_int(oidp, 101 oidp->oid_arg1, oidp->oid_arg2, req); 102 if (!error) { 103 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); 104 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1); 105 RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX); 106 RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX); 107 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX); 108 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX); 109 } 110 return error; 111 } 112 113 #undef RANGECHK 114 115 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); 116 117 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW, 118 &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", ""); 119 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW, 120 &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", ""); 121 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW, 122 &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", ""); 123 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW, 124 &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", ""); 125 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW, 126 &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", ""); 127 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW, 128 &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", ""); 129 130 /* 131 * in_pcb.c: manage the Protocol Control Blocks. 132 * 133 * NOTE: It is assumed that most of these functions will be called at 134 * splnet(). XXX - There are, unfortunately, a few exceptions to this 135 * rule that should be fixed. 136 */ 137 138 /* 139 * Allocate a PCB and associate it with the socket. 140 */ 141 int 142 in_pcballoc(so, pcbinfo, p) 143 struct socket *so; 144 struct inpcbinfo *pcbinfo; 145 struct proc *p; 146 { 147 register struct inpcb *inp; 148 149 inp = zalloci(pcbinfo->ipi_zone); 150 if (inp == NULL) 151 return (ENOBUFS); 152 bzero((caddr_t)inp, sizeof(*inp)); 153 inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 154 inp->inp_pcbinfo = pcbinfo; 155 inp->inp_socket = so; 156 #if defined(INET6) 157 if (ip6_mapped_addr_on) 158 inp->inp_flags &= ~IN6P_BINDV6ONLY; 159 else 160 inp->inp_flags |= IN6P_BINDV6ONLY; 161 #endif 162 LIST_INSERT_HEAD(pcbinfo->listhead, inp, inp_list); 163 pcbinfo->ipi_count++; 164 so->so_pcb = (caddr_t)inp; 165 return (0); 166 } 167 168 int 169 in_pcbbind(inp, nam, p) 170 register struct inpcb *inp; 171 struct sockaddr *nam; 172 struct proc *p; 173 { 174 register struct socket *so = inp->inp_socket; 175 unsigned short *lastport; 176 struct sockaddr_in *sin; 177 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 178 u_short lport = 0; 179 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT); 180 int error, prison = 0; 181 182 if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */ 183 return (EADDRNOTAVAIL); 184 if (inp->inp_lport || inp->inp_laddr.s_addr != INADDR_ANY) 185 return (EINVAL); 186 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 187 wild = 1; 188 if (nam) { 189 sin = (struct sockaddr_in *)nam; 190 if (nam->sa_len != sizeof (*sin)) 191 return (EINVAL); 192 #ifdef notdef 193 /* 194 * We should check the family, but old programs 195 * incorrectly fail to initialize it. 196 */ 197 if (sin->sin_family != AF_INET) 198 return (EAFNOSUPPORT); 199 #endif 200 if (prison_ip(p, 0, &sin->sin_addr.s_addr)) 201 return(EINVAL); 202 lport = sin->sin_port; 203 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 204 /* 205 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 206 * allow complete duplication of binding if 207 * SO_REUSEPORT is set, or if SO_REUSEADDR is set 208 * and a multicast address is bound on both 209 * new and duplicated sockets. 210 */ 211 if (so->so_options & SO_REUSEADDR) 212 reuseport = SO_REUSEADDR|SO_REUSEPORT; 213 } else if (sin->sin_addr.s_addr != INADDR_ANY) { 214 sin->sin_port = 0; /* yech... */ 215 if (ifa_ifwithaddr((struct sockaddr *)sin) == 0) 216 return (EADDRNOTAVAIL); 217 } 218 if (lport) { 219 struct inpcb *t; 220 221 /* GROSS */ 222 if (ntohs(lport) < IPPORT_RESERVED && p && 223 suser_xxx(0, p, PRISON_ROOT)) 224 return (EACCES); 225 if (p && p->p_prison) 226 prison = 1; 227 if (so->so_cred->cr_uid != 0 && 228 !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 229 t = in_pcblookup_local(inp->inp_pcbinfo, 230 sin->sin_addr, lport, 231 prison ? 0 : INPLOOKUP_WILDCARD); 232 if (t && 233 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 234 ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 235 (t->inp_socket->so_options & 236 SO_REUSEPORT) == 0) && 237 (so->so_cred->cr_uid != 238 t->inp_socket->so_cred->cr_uid)) { 239 #if defined(INET6) 240 if ((inp->inp_flags & 241 IN6P_BINDV6ONLY) != 0 || 242 ntohl(sin->sin_addr.s_addr) != 243 INADDR_ANY || 244 ntohl(t->inp_laddr.s_addr) != 245 INADDR_ANY || 246 INP_SOCKAF(so) == 247 INP_SOCKAF(t->inp_socket)) 248 #endif /* defined(INET6) */ 249 return (EADDRINUSE); 250 } 251 } 252 t = in_pcblookup_local(pcbinfo, sin->sin_addr, 253 lport, prison ? 0 : wild); 254 if (t && 255 (reuseport & t->inp_socket->so_options) == 0) { 256 #if defined(INET6) 257 if ((inp->inp_flags & IN6P_BINDV6ONLY) != 0 || 258 ntohl(sin->sin_addr.s_addr) != 259 INADDR_ANY || 260 ntohl(t->inp_laddr.s_addr) != 261 INADDR_ANY || 262 INP_SOCKAF(so) == 263 INP_SOCKAF(t->inp_socket)) 264 #endif /* defined(INET6) */ 265 return (EADDRINUSE); 266 } 267 } 268 inp->inp_laddr = sin->sin_addr; 269 } 270 if (lport == 0) { 271 ushort first, last; 272 int count; 273 274 if (prison_ip(p, 0, &inp->inp_laddr.s_addr )) 275 return (EINVAL); 276 inp->inp_flags |= INP_ANONPORT; 277 278 if (inp->inp_flags & INP_HIGHPORT) { 279 first = ipport_hifirstauto; /* sysctl */ 280 last = ipport_hilastauto; 281 lastport = &pcbinfo->lasthi; 282 } else if (inp->inp_flags & INP_LOWPORT) { 283 if (p && (error = suser_xxx(0, p, PRISON_ROOT))) 284 return error; 285 first = ipport_lowfirstauto; /* 1023 */ 286 last = ipport_lowlastauto; /* 600 */ 287 lastport = &pcbinfo->lastlow; 288 } else { 289 first = ipport_firstauto; /* sysctl */ 290 last = ipport_lastauto; 291 lastport = &pcbinfo->lastport; 292 } 293 /* 294 * Simple check to ensure all ports are not used up causing 295 * a deadlock here. 296 * 297 * We split the two cases (up and down) so that the direction 298 * is not being tested on each round of the loop. 299 */ 300 if (first > last) { 301 /* 302 * counting down 303 */ 304 count = first - last; 305 306 do { 307 if (count-- < 0) { /* completely used? */ 308 /* 309 * Undo any address bind that may have 310 * occurred above. 311 */ 312 inp->inp_laddr.s_addr = INADDR_ANY; 313 return (EAGAIN); 314 } 315 --*lastport; 316 if (*lastport > first || *lastport < last) 317 *lastport = first; 318 lport = htons(*lastport); 319 } while (in_pcblookup_local(pcbinfo, 320 inp->inp_laddr, lport, wild)); 321 } else { 322 /* 323 * counting up 324 */ 325 count = last - first; 326 327 do { 328 if (count-- < 0) { /* completely used? */ 329 /* 330 * Undo any address bind that may have 331 * occurred above. 332 */ 333 inp->inp_laddr.s_addr = INADDR_ANY; 334 return (EAGAIN); 335 } 336 ++*lastport; 337 if (*lastport < first || *lastport > last) 338 *lastport = first; 339 lport = htons(*lastport); 340 } while (in_pcblookup_local(pcbinfo, 341 inp->inp_laddr, lport, wild)); 342 } 343 } 344 inp->inp_lport = lport; 345 if (in_pcbinshash(inp) != 0) { 346 inp->inp_laddr.s_addr = INADDR_ANY; 347 inp->inp_lport = 0; 348 return (EAGAIN); 349 } 350 return (0); 351 } 352 353 /* 354 * Transform old in_pcbconnect() into an inner subroutine for new 355 * in_pcbconnect(): Do some validity-checking on the remote 356 * address (in mbuf 'nam') and then determine local host address 357 * (i.e., which interface) to use to access that remote host. 358 * 359 * This preserves definition of in_pcbconnect(), while supporting a 360 * slightly different version for T/TCP. (This is more than 361 * a bit of a kludge, but cleaning up the internal interfaces would 362 * have forced minor changes in every protocol). 363 */ 364 365 int 366 in_pcbladdr(inp, nam, plocal_sin) 367 register struct inpcb *inp; 368 struct sockaddr *nam; 369 struct sockaddr_in **plocal_sin; 370 { 371 struct in_ifaddr *ia; 372 register struct sockaddr_in *sin = (struct sockaddr_in *)nam; 373 374 if (nam->sa_len != sizeof (*sin)) 375 return (EINVAL); 376 if (sin->sin_family != AF_INET) 377 return (EAFNOSUPPORT); 378 if (sin->sin_port == 0) 379 return (EADDRNOTAVAIL); 380 if (!TAILQ_EMPTY(&in_ifaddrhead)) { 381 /* 382 * If the destination address is INADDR_ANY, 383 * use the primary local address. 384 * If the supplied address is INADDR_BROADCAST, 385 * and the primary interface supports broadcast, 386 * choose the broadcast address for that interface. 387 */ 388 #define satosin(sa) ((struct sockaddr_in *)(sa)) 389 #define sintosa(sin) ((struct sockaddr *)(sin)) 390 #define ifatoia(ifa) ((struct in_ifaddr *)(ifa)) 391 if (sin->sin_addr.s_addr == INADDR_ANY) 392 sin->sin_addr = IA_SIN(in_ifaddrhead.tqh_first)->sin_addr; 393 else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST && 394 (in_ifaddrhead.tqh_first->ia_ifp->if_flags & IFF_BROADCAST)) 395 sin->sin_addr = satosin(&in_ifaddrhead.tqh_first->ia_broadaddr)->sin_addr; 396 } 397 if (inp->inp_laddr.s_addr == INADDR_ANY) { 398 register struct route *ro; 399 400 ia = (struct in_ifaddr *)0; 401 /* 402 * If route is known or can be allocated now, 403 * our src addr is taken from the i/f, else punt. 404 */ 405 ro = &inp->inp_route; 406 if (ro->ro_rt && 407 (satosin(&ro->ro_dst)->sin_addr.s_addr != 408 sin->sin_addr.s_addr || 409 inp->inp_socket->so_options & SO_DONTROUTE)) { 410 RTFREE(ro->ro_rt); 411 ro->ro_rt = (struct rtentry *)0; 412 } 413 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/ 414 (ro->ro_rt == (struct rtentry *)0 || 415 ro->ro_rt->rt_ifp == (struct ifnet *)0)) { 416 /* No route yet, so try to acquire one */ 417 ro->ro_dst.sa_family = AF_INET; 418 ro->ro_dst.sa_len = sizeof(struct sockaddr_in); 419 ((struct sockaddr_in *) &ro->ro_dst)->sin_addr = 420 sin->sin_addr; 421 rtalloc(ro); 422 } 423 /* 424 * If we found a route, use the address 425 * corresponding to the outgoing interface 426 * unless it is the loopback (in case a route 427 * to our address on another net goes to loopback). 428 */ 429 if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) 430 ia = ifatoia(ro->ro_rt->rt_ifa); 431 if (ia == 0) { 432 u_short fport = sin->sin_port; 433 434 sin->sin_port = 0; 435 ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin))); 436 if (ia == 0) 437 ia = ifatoia(ifa_ifwithnet(sintosa(sin))); 438 sin->sin_port = fport; 439 if (ia == 0) 440 ia = in_ifaddrhead.tqh_first; 441 if (ia == 0) 442 return (EADDRNOTAVAIL); 443 } 444 /* 445 * If the destination address is multicast and an outgoing 446 * interface has been set as a multicast option, use the 447 * address of that interface as our source address. 448 */ 449 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) && 450 inp->inp_moptions != NULL) { 451 struct ip_moptions *imo; 452 struct ifnet *ifp; 453 454 imo = inp->inp_moptions; 455 if (imo->imo_multicast_ifp != NULL) { 456 ifp = imo->imo_multicast_ifp; 457 for (ia = in_ifaddrhead.tqh_first; ia; 458 ia = ia->ia_link.tqe_next) 459 if (ia->ia_ifp == ifp) 460 break; 461 if (ia == 0) 462 return (EADDRNOTAVAIL); 463 } 464 } 465 /* 466 * Don't do pcblookup call here; return interface in plocal_sin 467 * and exit to caller, that will do the lookup. 468 */ 469 *plocal_sin = &ia->ia_addr; 470 471 } 472 return(0); 473 } 474 475 /* 476 * Outer subroutine: 477 * Connect from a socket to a specified address. 478 * Both address and port must be specified in argument sin. 479 * If don't have a local address for this socket yet, 480 * then pick one. 481 */ 482 int 483 in_pcbconnect(inp, nam, p) 484 register struct inpcb *inp; 485 struct sockaddr *nam; 486 struct proc *p; 487 { 488 struct sockaddr_in *ifaddr; 489 register struct sockaddr_in *sin = (struct sockaddr_in *)nam; 490 int error; 491 492 /* 493 * Call inner routine, to assign local interface address. 494 */ 495 if ((error = in_pcbladdr(inp, nam, &ifaddr)) != 0) 496 return(error); 497 498 if (in_pcblookup_hash(inp->inp_pcbinfo, sin->sin_addr, sin->sin_port, 499 inp->inp_laddr.s_addr ? inp->inp_laddr : ifaddr->sin_addr, 500 inp->inp_lport, 0, NULL) != NULL) { 501 return (EADDRINUSE); 502 } 503 if (inp->inp_laddr.s_addr == INADDR_ANY) { 504 if (inp->inp_lport == 0) { 505 error = in_pcbbind(inp, (struct sockaddr *)0, p); 506 if (error) 507 return (error); 508 } 509 inp->inp_laddr = ifaddr->sin_addr; 510 } 511 inp->inp_faddr = sin->sin_addr; 512 inp->inp_fport = sin->sin_port; 513 in_pcbrehash(inp); 514 return (0); 515 } 516 517 void 518 in_pcbdisconnect(inp) 519 struct inpcb *inp; 520 { 521 522 inp->inp_faddr.s_addr = INADDR_ANY; 523 inp->inp_fport = 0; 524 in_pcbrehash(inp); 525 if (inp->inp_socket->so_state & SS_NOFDREF) 526 in_pcbdetach(inp); 527 } 528 529 void 530 in_pcbdetach(inp) 531 struct inpcb *inp; 532 { 533 struct socket *so = inp->inp_socket; 534 struct inpcbinfo *ipi = inp->inp_pcbinfo; 535 536 #ifdef IPSEC 537 if (inp->inp_sp != NULL) 538 ipsec4_delete_pcbpolicy(inp); 539 #endif /*IPSEC*/ 540 inp->inp_gencnt = ++ipi->ipi_gencnt; 541 in_pcbremlists(inp); 542 so->so_pcb = 0; 543 sofree(so); 544 if (inp->inp_options) 545 (void)m_free(inp->inp_options); 546 if (inp->inp_route.ro_rt) 547 rtfree(inp->inp_route.ro_rt); 548 ip_freemoptions(inp->inp_moptions); 549 inp->inp_vflag = 0; 550 zfreei(ipi->ipi_zone, inp); 551 } 552 553 /* 554 * The calling convention of in_setsockaddr() and in_setpeeraddr() was 555 * modified to match the pru_sockaddr() and pru_peeraddr() entry points 556 * in struct pr_usrreqs, so that protocols can just reference then directly 557 * without the need for a wrapper function. The socket must have a valid 558 * (i.e., non-nil) PCB, but it should be impossible to get an invalid one 559 * except through a kernel programming error, so it is acceptable to panic 560 * (or in this case trap) if the PCB is invalid. (Actually, we don't trap 561 * because there actually /is/ a programming error somewhere... XXX) 562 */ 563 int 564 in_setsockaddr(so, nam) 565 struct socket *so; 566 struct sockaddr **nam; 567 { 568 int s; 569 register struct inpcb *inp; 570 register struct sockaddr_in *sin; 571 572 /* 573 * Do the malloc first in case it blocks. 574 */ 575 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, M_WAITOK); 576 bzero(sin, sizeof *sin); 577 sin->sin_family = AF_INET; 578 sin->sin_len = sizeof(*sin); 579 580 s = splnet(); 581 inp = sotoinpcb(so); 582 if (!inp) { 583 splx(s); 584 free(sin, M_SONAME); 585 return ECONNRESET; 586 } 587 sin->sin_port = inp->inp_lport; 588 sin->sin_addr = inp->inp_laddr; 589 splx(s); 590 591 *nam = (struct sockaddr *)sin; 592 return 0; 593 } 594 595 int 596 in_setpeeraddr(so, nam) 597 struct socket *so; 598 struct sockaddr **nam; 599 { 600 int s; 601 struct inpcb *inp; 602 register struct sockaddr_in *sin; 603 604 /* 605 * Do the malloc first in case it blocks. 606 */ 607 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, M_WAITOK); 608 bzero(sin, sizeof (*sin)); 609 sin->sin_family = AF_INET; 610 sin->sin_len = sizeof(*sin); 611 612 s = splnet(); 613 inp = sotoinpcb(so); 614 if (!inp) { 615 splx(s); 616 free(sin, M_SONAME); 617 return ECONNRESET; 618 } 619 sin->sin_port = inp->inp_fport; 620 sin->sin_addr = inp->inp_faddr; 621 splx(s); 622 623 *nam = (struct sockaddr *)sin; 624 return 0; 625 } 626 627 /* 628 * Pass some notification to all connections of a protocol 629 * associated with address dst. The local address and/or port numbers 630 * may be specified to limit the search. The "usual action" will be 631 * taken, depending on the ctlinput cmd. The caller must filter any 632 * cmds that are uninteresting (e.g., no error in the map). 633 * Call the protocol specific routine (if any) to report 634 * any errors for each matching socket. 635 */ 636 void 637 in_pcbnotify(head, dst, fport_arg, laddr, lport_arg, cmd, notify) 638 struct inpcbhead *head; 639 struct sockaddr *dst; 640 u_int fport_arg, lport_arg; 641 struct in_addr laddr; 642 int cmd; 643 void (*notify) __P((struct inpcb *, int)); 644 { 645 register struct inpcb *inp, *oinp; 646 struct in_addr faddr; 647 u_short fport = fport_arg, lport = lport_arg; 648 int errno, s; 649 650 if ((unsigned)cmd > PRC_NCMDS || dst->sa_family != AF_INET) 651 return; 652 faddr = ((struct sockaddr_in *)dst)->sin_addr; 653 if (faddr.s_addr == INADDR_ANY) 654 return; 655 656 /* 657 * Redirects go to all references to the destination, 658 * and use in_rtchange to invalidate the route cache. 659 * Dead host indications: notify all references to the destination. 660 * Otherwise, if we have knowledge of the local port and address, 661 * deliver only to that socket. 662 */ 663 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) { 664 fport = 0; 665 lport = 0; 666 laddr.s_addr = 0; 667 if (cmd != PRC_HOSTDEAD) 668 notify = in_rtchange; 669 } 670 errno = inetctlerrmap[cmd]; 671 s = splnet(); 672 for (inp = head->lh_first; inp != NULL;) { 673 #ifdef INET6 674 if ((inp->inp_vflag & INP_IPV4) == 0) { 675 inp = LIST_NEXT(inp, inp_list); 676 continue; 677 } 678 #endif 679 if (inp->inp_faddr.s_addr != faddr.s_addr || 680 inp->inp_socket == 0 || 681 (lport && inp->inp_lport != lport) || 682 (laddr.s_addr && inp->inp_laddr.s_addr != laddr.s_addr) || 683 (fport && inp->inp_fport != fport)) { 684 inp = inp->inp_list.le_next; 685 continue; 686 } 687 oinp = inp; 688 inp = inp->inp_list.le_next; 689 if (notify) 690 (*notify)(oinp, errno); 691 } 692 splx(s); 693 } 694 695 /* 696 * Check for alternatives when higher level complains 697 * about service problems. For now, invalidate cached 698 * routing information. If the route was created dynamically 699 * (by a redirect), time to try a default gateway again. 700 */ 701 void 702 in_losing(inp) 703 struct inpcb *inp; 704 { 705 register struct rtentry *rt; 706 struct rt_addrinfo info; 707 708 if ((rt = inp->inp_route.ro_rt)) { 709 inp->inp_route.ro_rt = 0; 710 bzero((caddr_t)&info, sizeof(info)); 711 info.rti_info[RTAX_DST] = 712 (struct sockaddr *)&inp->inp_route.ro_dst; 713 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 714 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 715 rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0); 716 if (rt->rt_flags & RTF_DYNAMIC) 717 (void) rtrequest(RTM_DELETE, rt_key(rt), 718 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 719 (struct rtentry **)0); 720 else 721 /* 722 * A new route can be allocated 723 * the next time output is attempted. 724 */ 725 rtfree(rt); 726 } 727 } 728 729 /* 730 * After a routing change, flush old routing 731 * and allocate a (hopefully) better one. 732 */ 733 static void 734 in_rtchange(inp, errno) 735 register struct inpcb *inp; 736 int errno; 737 { 738 if (inp->inp_route.ro_rt) { 739 rtfree(inp->inp_route.ro_rt); 740 inp->inp_route.ro_rt = 0; 741 /* 742 * A new route can be allocated the next time 743 * output is attempted. 744 */ 745 } 746 } 747 748 /* 749 * Lookup a PCB based on the local address and port. 750 */ 751 struct inpcb * 752 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay) 753 struct inpcbinfo *pcbinfo; 754 struct in_addr laddr; 755 u_int lport_arg; 756 int wild_okay; 757 { 758 register struct inpcb *inp; 759 int matchwild = 3, wildcard; 760 u_short lport = lport_arg; 761 762 if (!wild_okay) { 763 struct inpcbhead *head; 764 /* 765 * Look for an unconnected (wildcard foreign addr) PCB that 766 * matches the local address and port we're looking for. 767 */ 768 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 769 for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) { 770 #ifdef INET6 771 if ((inp->inp_vflag & INP_IPV4) == 0) 772 continue; 773 #endif 774 if (inp->inp_faddr.s_addr == INADDR_ANY && 775 inp->inp_laddr.s_addr == laddr.s_addr && 776 inp->inp_lport == lport) { 777 /* 778 * Found. 779 */ 780 return (inp); 781 } 782 } 783 /* 784 * Not found. 785 */ 786 return (NULL); 787 } else { 788 struct inpcbporthead *porthash; 789 struct inpcbport *phd; 790 struct inpcb *match = NULL; 791 /* 792 * Best fit PCB lookup. 793 * 794 * First see if this local port is in use by looking on the 795 * port hash list. 796 */ 797 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport, 798 pcbinfo->porthashmask)]; 799 for (phd = porthash->lh_first; phd != NULL; phd = phd->phd_hash.le_next) { 800 if (phd->phd_port == lport) 801 break; 802 } 803 if (phd != NULL) { 804 /* 805 * Port is in use by one or more PCBs. Look for best 806 * fit. 807 */ 808 for (inp = phd->phd_pcblist.lh_first; inp != NULL; 809 inp = inp->inp_portlist.le_next) { 810 wildcard = 0; 811 #ifdef INET6 812 if ((inp->inp_vflag & INP_IPV4) == 0) 813 continue; 814 #endif 815 if (inp->inp_faddr.s_addr != INADDR_ANY) 816 wildcard++; 817 if (inp->inp_laddr.s_addr != INADDR_ANY) { 818 if (laddr.s_addr == INADDR_ANY) 819 wildcard++; 820 else if (inp->inp_laddr.s_addr != laddr.s_addr) 821 continue; 822 } else { 823 if (laddr.s_addr != INADDR_ANY) 824 wildcard++; 825 } 826 if (wildcard < matchwild) { 827 match = inp; 828 matchwild = wildcard; 829 if (matchwild == 0) { 830 break; 831 } 832 } 833 } 834 } 835 return (match); 836 } 837 } 838 839 /* 840 * Lookup PCB in hash list. 841 */ 842 struct inpcb * 843 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, 844 ifp) 845 struct inpcbinfo *pcbinfo; 846 struct in_addr faddr, laddr; 847 u_int fport_arg, lport_arg; 848 int wildcard; 849 struct ifnet *ifp; 850 { 851 struct inpcbhead *head; 852 register struct inpcb *inp; 853 u_short fport = fport_arg, lport = lport_arg; 854 855 /* 856 * First look for an exact match. 857 */ 858 head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)]; 859 for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) { 860 #ifdef INET6 861 if ((inp->inp_vflag & INP_IPV4) == 0) 862 continue; 863 #endif 864 if (inp->inp_faddr.s_addr == faddr.s_addr && 865 inp->inp_laddr.s_addr == laddr.s_addr && 866 inp->inp_fport == fport && 867 inp->inp_lport == lport) { 868 /* 869 * Found. 870 */ 871 return (inp); 872 } 873 } 874 if (wildcard) { 875 struct inpcb *local_wild = NULL; 876 #if defined(INET6) 877 struct inpcb *local_wild_mapped = NULL; 878 #endif /* defined(INET6) */ 879 880 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)]; 881 for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) { 882 #ifdef INET6 883 if ((inp->inp_vflag & INP_IPV4) == 0) 884 continue; 885 #endif 886 if (inp->inp_faddr.s_addr == INADDR_ANY && 887 inp->inp_lport == lport) { 888 #if defined(NFAITH) && NFAITH > 0 889 if (ifp && ifp->if_type == IFT_FAITH && 890 (inp->inp_flags & INP_FAITH) == 0) 891 continue; 892 #endif 893 if (inp->inp_laddr.s_addr == laddr.s_addr) 894 return (inp); 895 else if (inp->inp_laddr.s_addr == INADDR_ANY) { 896 #if defined(INET6) 897 if (INP_CHECK_SOCKAF(inp->inp_socket, 898 AF_INET6)) 899 local_wild_mapped = inp; 900 else 901 #endif /* defined(INET6) */ 902 local_wild = inp; 903 } 904 } 905 } 906 #if defined(INET6) 907 if (local_wild == NULL) 908 return (local_wild_mapped); 909 #endif /* defined(INET6) */ 910 return (local_wild); 911 } 912 913 /* 914 * Not found. 915 */ 916 return (NULL); 917 } 918 919 /* 920 * Insert PCB onto various hash lists. 921 */ 922 int 923 in_pcbinshash(inp) 924 struct inpcb *inp; 925 { 926 struct inpcbhead *pcbhash; 927 struct inpcbporthead *pcbporthash; 928 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 929 struct inpcbport *phd; 930 u_int32_t hashkey_faddr; 931 932 #ifdef INET6 933 if (inp->inp_vflag & INP_IPV6) 934 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 935 else 936 #endif /* INET6 */ 937 hashkey_faddr = inp->inp_faddr.s_addr; 938 939 pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 940 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)]; 941 942 pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport, 943 pcbinfo->porthashmask)]; 944 945 /* 946 * Go through port list and look for a head for this lport. 947 */ 948 for (phd = pcbporthash->lh_first; phd != NULL; phd = phd->phd_hash.le_next) { 949 if (phd->phd_port == inp->inp_lport) 950 break; 951 } 952 /* 953 * If none exists, malloc one and tack it on. 954 */ 955 if (phd == NULL) { 956 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT); 957 if (phd == NULL) { 958 return (ENOBUFS); /* XXX */ 959 } 960 phd->phd_port = inp->inp_lport; 961 LIST_INIT(&phd->phd_pcblist); 962 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 963 } 964 inp->inp_phd = phd; 965 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 966 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 967 return (0); 968 } 969 970 /* 971 * Move PCB to the proper hash bucket when { faddr, fport } have been 972 * changed. NOTE: This does not handle the case of the lport changing (the 973 * hashed port list would have to be updated as well), so the lport must 974 * not change after in_pcbinshash() has been called. 975 */ 976 void 977 in_pcbrehash(inp) 978 struct inpcb *inp; 979 { 980 struct inpcbhead *head; 981 u_int32_t hashkey_faddr; 982 983 #ifdef INET6 984 if (inp->inp_vflag & INP_IPV6) 985 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */; 986 else 987 #endif /* INET6 */ 988 hashkey_faddr = inp->inp_faddr.s_addr; 989 990 head = &inp->inp_pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr, 991 inp->inp_lport, inp->inp_fport, inp->inp_pcbinfo->hashmask)]; 992 993 LIST_REMOVE(inp, inp_hash); 994 LIST_INSERT_HEAD(head, inp, inp_hash); 995 } 996 997 /* 998 * Remove PCB from various lists. 999 */ 1000 void 1001 in_pcbremlists(inp) 1002 struct inpcb *inp; 1003 { 1004 inp->inp_gencnt = ++inp->inp_pcbinfo->ipi_gencnt; 1005 if (inp->inp_lport) { 1006 struct inpcbport *phd = inp->inp_phd; 1007 1008 LIST_REMOVE(inp, inp_hash); 1009 LIST_REMOVE(inp, inp_portlist); 1010 if (phd->phd_pcblist.lh_first == NULL) { 1011 LIST_REMOVE(phd, phd_hash); 1012 free(phd, M_PCB); 1013 } 1014 } 1015 LIST_REMOVE(inp, inp_list); 1016 inp->inp_pcbinfo->ipi_count--; 1017 } 1018 1019 int 1020 prison_xinpcb(struct proc *p, struct inpcb *inp) 1021 { 1022 if (!p->p_prison) 1023 return (0); 1024 if (ntohl(inp->inp_laddr.s_addr) == p->p_prison->pr_ip) 1025 return (0); 1026 return (1); 1027 } 1028