tcp_usrreq.c revision 157993
1/*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. 4 * Copyright (c) 2006 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 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 32 * $FreeBSD: head/sys/netinet/tcp_usrreq.c 157993 2006-04-24 08:20:02Z rwatson $ 33 */ 34 35#include "opt_inet.h" 36#include "opt_inet6.h" 37#include "opt_tcpdebug.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/malloc.h> 42#include <sys/kernel.h> 43#include <sys/sysctl.h> 44#include <sys/mbuf.h> 45#ifdef INET6 46#include <sys/domain.h> 47#endif /* INET6 */ 48#include <sys/socket.h> 49#include <sys/socketvar.h> 50#include <sys/protosw.h> 51#include <sys/proc.h> 52#include <sys/jail.h> 53 54#include <net/if.h> 55#include <net/route.h> 56 57#include <netinet/in.h> 58#include <netinet/in_systm.h> 59#ifdef INET6 60#include <netinet/ip6.h> 61#endif 62#include <netinet/in_pcb.h> 63#ifdef INET6 64#include <netinet6/in6_pcb.h> 65#endif 66#include <netinet/in_var.h> 67#include <netinet/ip_var.h> 68#ifdef INET6 69#include <netinet6/ip6_var.h> 70#include <netinet6/scope6_var.h> 71#endif 72#include <netinet/tcp.h> 73#include <netinet/tcp_fsm.h> 74#include <netinet/tcp_seq.h> 75#include <netinet/tcp_timer.h> 76#include <netinet/tcp_var.h> 77#include <netinet/tcpip.h> 78#ifdef TCPDEBUG 79#include <netinet/tcp_debug.h> 80#endif 81 82/* 83 * TCP protocol interface to socket abstraction. 84 */ 85extern char *tcpstates[]; /* XXX ??? */ 86 87static int tcp_attach(struct socket *); 88static int tcp_connect(struct tcpcb *, struct sockaddr *, 89 struct thread *td); 90#ifdef INET6 91static int tcp6_connect(struct tcpcb *, struct sockaddr *, 92 struct thread *td); 93#endif /* INET6 */ 94static void tcp_disconnect(struct tcpcb *); 95static void tcp_usrclosed(struct tcpcb *); 96static void tcp_fill_info(struct tcpcb *, struct tcp_info *); 97 98#ifdef TCPDEBUG 99#define TCPDEBUG0 int ostate = 0 100#define TCPDEBUG1() ostate = tp ? tp->t_state : 0 101#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ 102 tcp_trace(TA_USER, ostate, tp, 0, 0, req) 103#else 104#define TCPDEBUG0 105#define TCPDEBUG1() 106#define TCPDEBUG2(req) 107#endif 108 109/* 110 * TCP attaches to socket via pru_attach(), reserving space, 111 * and an internet control block. 112 */ 113static int 114tcp_usr_attach(struct socket *so, int proto, struct thread *td) 115{ 116 struct inpcb *inp; 117 struct tcpcb *tp = NULL; 118 int error; 119 TCPDEBUG0; 120 121 inp = sotoinpcb(so); 122 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); 123 INP_INFO_WLOCK(&tcbinfo); 124 TCPDEBUG1(); 125 126 error = tcp_attach(so); 127 if (error) 128 goto out; 129 130 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 131 so->so_linger = TCP_LINGERTIME; 132 133 inp = sotoinpcb(so); 134 tp = intotcpcb(inp); 135out: 136 TCPDEBUG2(PRU_ATTACH); 137 INP_INFO_WUNLOCK(&tcbinfo); 138 return error; 139} 140 141/* 142 * tcp_detach() releases any protocol state that can be reasonably released 143 * when a socket shutdown is requested, and is a shared code path for 144 * tcp_usr_detach() and tcp_usr_abort(), the two socket close entry points. 145 * 146 * Accepts pcbinfo, inpcb locked, will unlock the inpcb (if needed) on 147 * return. 148 */ 149static void 150tcp_detach(struct socket *so, struct inpcb *inp) 151{ 152 struct tcpcb *tp; 153#ifdef INET6 154 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0; 155#endif 156 157 INP_INFO_WLOCK_ASSERT(&tcbinfo); 158 INP_LOCK_ASSERT(inp); 159 160 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp")); 161 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so")); 162 163 if (inp->inp_vflag & INP_TIMEWAIT) { 164 if (inp->inp_vflag & INP_DROPPED) { 165 /* 166 * Connection was in time wait and has been dropped; 167 * the calling path is either via tcp_twclose(), or 168 * as a result of an eventual soclose() after 169 * tcp_twclose() has been called. In either case, 170 * tcp_twclose() has detached the tcptw from the 171 * inpcb, so we just detach and free the inpcb. 172 * 173 * XXXRW: Would it be cleaner to free the tcptw 174 * here? 175 */ 176#ifdef INET6 177 if (isipv6) { 178 in6_pcbdetach(inp); 179 in6_pcbfree(inp); 180 } else { 181#endif 182 in_pcbdetach(inp); 183 in_pcbfree(inp); 184#ifdef INET6 185 } 186#endif 187 } else { 188 /* 189 * Connection is in time wait and has not yet been 190 * dropped; allow the socket to be discarded, but 191 * need to keep inpcb until end of time wait. 192 */ 193#ifdef INET6 194 if (isipv6) 195 in6_pcbdetach(inp); 196 else 197#endif 198 in_pcbdetach(inp); 199 INP_UNLOCK(inp); 200 } 201 } else { 202 /* 203 * If not in timewait, there are two possible paths. First, 204 * the TCP connection is either embryonic or done, in which 205 * case we tear down all state. Second, it may still be 206 * active, in which case we acquire a reference to the socket 207 * and will free it later when TCP is done. 208 */ 209 tp = intotcpcb(inp); 210 if (inp->inp_vflag & INP_DROPPED || 211 tp->t_state < TCPS_SYN_SENT) { 212 tcp_discardcb(tp); 213#ifdef INET6 214 if (isipv6) { 215 in_pcbdetach(inp); 216 in_pcbfree(inp); 217 } else { 218#endif 219 in_pcbdetach(inp); 220 in_pcbfree(inp); 221#ifdef INET6 222 } 223#endif 224 } else { 225 SOCK_LOCK(so); 226 so->so_state |= SS_PROTOREF; 227 SOCK_UNLOCK(so); 228 inp->inp_vflag |= INP_SOCKREF; 229 INP_UNLOCK(inp); 230 } 231 } 232} 233 234/* 235 * pru_detach() detaches the TCP protocol from the socket. 236 * If the protocol state is non-embryonic, then can't 237 * do this directly: have to initiate a pru_disconnect(), 238 * which may finish later; embryonic TCB's can just 239 * be discarded here. 240 */ 241static void 242tcp_usr_detach(struct socket *so) 243{ 244 struct inpcb *inp; 245 struct tcpcb *tp; 246 TCPDEBUG0; 247 248 inp = sotoinpcb(so); 249 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL")); 250 INP_INFO_WLOCK(&tcbinfo); 251 INP_LOCK(inp); 252 KASSERT(inp->inp_socket != NULL, 253 ("tcp_usr_detach: inp_socket == NULL")); 254 TCPDEBUG1(); 255 256 /* 257 * First, if we still have full TCP state, and we're not dropped, 258 * initiate a disconnect. 259 */ 260 if (!(inp->inp_vflag & INP_TIMEWAIT) && 261 !(inp->inp_vflag & INP_DROPPED)) { 262 tp = intotcpcb(inp); 263 tcp_disconnect(tp); 264 } 265 tcp_detach(so, inp); 266 tp = NULL; 267 TCPDEBUG2(PRU_DETACH); 268 INP_INFO_WUNLOCK(&tcbinfo); 269} 270 271/* 272 * Give the socket an address. 273 */ 274static int 275tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 276{ 277 int error = 0; 278 struct inpcb *inp; 279 struct tcpcb *tp = NULL; 280 struct sockaddr_in *sinp; 281 282 sinp = (struct sockaddr_in *)nam; 283 if (nam->sa_len != sizeof (*sinp)) 284 return (EINVAL); 285 /* 286 * Must check for multicast addresses and disallow binding 287 * to them. 288 */ 289 if (sinp->sin_family == AF_INET && 290 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) 291 return (EAFNOSUPPORT); 292 293 TCPDEBUG0; 294 INP_INFO_WLOCK(&tcbinfo); 295 inp = sotoinpcb(so); 296 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); 297 INP_LOCK(inp); 298 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 299 error = EINVAL; 300 goto out; 301 } 302 tp = intotcpcb(inp); 303 TCPDEBUG1(); 304 error = in_pcbbind(inp, nam, td->td_ucred); 305out: 306 TCPDEBUG2(PRU_BIND); 307 INP_UNLOCK(inp); 308 INP_INFO_WUNLOCK(&tcbinfo); 309 310 return (error); 311} 312 313#ifdef INET6 314static int 315tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 316{ 317 int error = 0; 318 struct inpcb *inp; 319 struct tcpcb *tp = NULL; 320 struct sockaddr_in6 *sin6p; 321 322 sin6p = (struct sockaddr_in6 *)nam; 323 if (nam->sa_len != sizeof (*sin6p)) 324 return (EINVAL); 325 /* 326 * Must check for multicast addresses and disallow binding 327 * to them. 328 */ 329 if (sin6p->sin6_family == AF_INET6 && 330 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) 331 return (EAFNOSUPPORT); 332 333 TCPDEBUG0; 334 INP_INFO_WLOCK(&tcbinfo); 335 inp = sotoinpcb(so); 336 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); 337 INP_LOCK(inp); 338 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 339 error = EINVAL; 340 goto out; 341 } 342 tp = intotcpcb(inp); 343 TCPDEBUG1(); 344 inp->inp_vflag &= ~INP_IPV4; 345 inp->inp_vflag |= INP_IPV6; 346 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 347 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) 348 inp->inp_vflag |= INP_IPV4; 349 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 350 struct sockaddr_in sin; 351 352 in6_sin6_2_sin(&sin, sin6p); 353 inp->inp_vflag |= INP_IPV4; 354 inp->inp_vflag &= ~INP_IPV6; 355 error = in_pcbbind(inp, (struct sockaddr *)&sin, 356 td->td_ucred); 357 goto out; 358 } 359 } 360 error = in6_pcbbind(inp, nam, td->td_ucred); 361out: 362 TCPDEBUG2(PRU_BIND); 363 INP_UNLOCK(inp); 364 INP_INFO_WUNLOCK(&tcbinfo); 365 return (error); 366} 367#endif /* INET6 */ 368 369/* 370 * Prepare to accept connections. 371 */ 372static int 373tcp_usr_listen(struct socket *so, int backlog, struct thread *td) 374{ 375 int error = 0; 376 struct inpcb *inp; 377 struct tcpcb *tp = NULL; 378 379 TCPDEBUG0; 380 INP_INFO_WLOCK(&tcbinfo); 381 inp = sotoinpcb(so); 382 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); 383 INP_LOCK(inp); 384 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 385 error = EINVAL; 386 goto out; 387 } 388 tp = intotcpcb(inp); 389 TCPDEBUG1(); 390 SOCK_LOCK(so); 391 error = solisten_proto_check(so); 392 if (error == 0 && inp->inp_lport == 0) 393 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 394 if (error == 0) { 395 tp->t_state = TCPS_LISTEN; 396 solisten_proto(so, backlog); 397 } 398 SOCK_UNLOCK(so); 399 400out: 401 TCPDEBUG2(PRU_LISTEN); 402 INP_UNLOCK(inp); 403 INP_INFO_WUNLOCK(&tcbinfo); 404 return (error); 405} 406 407#ifdef INET6 408static int 409tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) 410{ 411 int error = 0; 412 struct inpcb *inp; 413 struct tcpcb *tp = NULL; 414 415 TCPDEBUG0; 416 INP_INFO_WLOCK(&tcbinfo); 417 inp = sotoinpcb(so); 418 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); 419 INP_LOCK(inp); 420 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 421 error = EINVAL; 422 goto out; 423 } 424 tp = intotcpcb(inp); 425 TCPDEBUG1(); 426 SOCK_LOCK(so); 427 error = solisten_proto_check(so); 428 if (error == 0 && inp->inp_lport == 0) { 429 inp->inp_vflag &= ~INP_IPV4; 430 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 431 inp->inp_vflag |= INP_IPV4; 432 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 433 } 434 if (error == 0) { 435 tp->t_state = TCPS_LISTEN; 436 solisten_proto(so, backlog); 437 } 438 SOCK_UNLOCK(so); 439 440out: 441 TCPDEBUG2(PRU_LISTEN); 442 INP_UNLOCK(inp); 443 INP_INFO_WUNLOCK(&tcbinfo); 444 return (error); 445} 446#endif /* INET6 */ 447 448/* 449 * Initiate connection to peer. 450 * Create a template for use in transmissions on this connection. 451 * Enter SYN_SENT state, and mark socket as connecting. 452 * Start keep-alive timer, and seed output sequence space. 453 * Send initial segment on connection. 454 */ 455static int 456tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 457{ 458 int error = 0; 459 struct inpcb *inp; 460 struct tcpcb *tp = NULL; 461 struct sockaddr_in *sinp; 462 463 sinp = (struct sockaddr_in *)nam; 464 if (nam->sa_len != sizeof (*sinp)) 465 return (EINVAL); 466 /* 467 * Must disallow TCP ``connections'' to multicast addresses. 468 */ 469 if (sinp->sin_family == AF_INET 470 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) 471 return (EAFNOSUPPORT); 472 if (jailed(td->td_ucred)) 473 prison_remote_ip(td->td_ucred, 0, &sinp->sin_addr.s_addr); 474 475 TCPDEBUG0; 476 INP_INFO_WLOCK(&tcbinfo); 477 inp = sotoinpcb(so); 478 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); 479 INP_LOCK(inp); 480 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 481 error = EINVAL; 482 goto out; 483 } 484 tp = intotcpcb(inp); 485 TCPDEBUG1(); 486 if ((error = tcp_connect(tp, nam, td)) != 0) 487 goto out; 488 error = tcp_output(tp); 489out: 490 TCPDEBUG2(PRU_CONNECT); 491 INP_UNLOCK(inp); 492 INP_INFO_WUNLOCK(&tcbinfo); 493 return (error); 494} 495 496#ifdef INET6 497static int 498tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 499{ 500 int error = 0; 501 struct inpcb *inp; 502 struct tcpcb *tp = NULL; 503 struct sockaddr_in6 *sin6p; 504 505 TCPDEBUG0; 506 507 sin6p = (struct sockaddr_in6 *)nam; 508 if (nam->sa_len != sizeof (*sin6p)) 509 return (EINVAL); 510 /* 511 * Must disallow TCP ``connections'' to multicast addresses. 512 */ 513 if (sin6p->sin6_family == AF_INET6 514 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) 515 return (EAFNOSUPPORT); 516 517 INP_INFO_WLOCK(&tcbinfo); 518 inp = sotoinpcb(so); 519 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); 520 INP_LOCK(inp); 521 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 522 error = EINVAL; 523 goto out; 524 } 525 tp = intotcpcb(inp); 526 TCPDEBUG1(); 527 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 528 struct sockaddr_in sin; 529 530 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 531 error = EINVAL; 532 goto out; 533 } 534 535 in6_sin6_2_sin(&sin, sin6p); 536 inp->inp_vflag |= INP_IPV4; 537 inp->inp_vflag &= ~INP_IPV6; 538 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 539 goto out; 540 error = tcp_output(tp); 541 goto out; 542 } 543 inp->inp_vflag &= ~INP_IPV4; 544 inp->inp_vflag |= INP_IPV6; 545 inp->inp_inc.inc_isipv6 = 1; 546 if ((error = tcp6_connect(tp, nam, td)) != 0) 547 goto out; 548 error = tcp_output(tp); 549 550out: 551 TCPDEBUG2(PRU_CONNECT); 552 INP_UNLOCK(inp); 553 INP_INFO_WUNLOCK(&tcbinfo); 554 return (error); 555} 556#endif /* INET6 */ 557 558/* 559 * Initiate disconnect from peer. 560 * If connection never passed embryonic stage, just drop; 561 * else if don't need to let data drain, then can just drop anyways, 562 * else have to begin TCP shutdown process: mark socket disconnecting, 563 * drain unread data, state switch to reflect user close, and 564 * send segment (e.g. FIN) to peer. Socket will be really disconnected 565 * when peer sends FIN and acks ours. 566 * 567 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 568 */ 569static int 570tcp_usr_disconnect(struct socket *so) 571{ 572 struct inpcb *inp; 573 struct tcpcb *tp = NULL; 574 int error = 0; 575 576 TCPDEBUG0; 577 INP_INFO_WLOCK(&tcbinfo); 578 inp = sotoinpcb(so); 579 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); 580 INP_LOCK(inp); 581 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 582 error = EINVAL; 583 goto out; 584 } 585 tp = intotcpcb(inp); 586 TCPDEBUG1(); 587 tcp_disconnect(tp); 588out: 589 TCPDEBUG2(PRU_DISCONNECT); 590 INP_UNLOCK(inp); 591 INP_INFO_WUNLOCK(&tcbinfo); 592 return (error); 593} 594 595/* 596 * Accept a connection. Essentially all the work is 597 * done at higher levels; just return the address 598 * of the peer, storing through addr. 599 */ 600static int 601tcp_usr_accept(struct socket *so, struct sockaddr **nam) 602{ 603 int error = 0; 604 struct inpcb *inp = NULL; 605 struct tcpcb *tp = NULL; 606 struct in_addr addr; 607 in_port_t port = 0; 608 TCPDEBUG0; 609 610 if (so->so_state & SS_ISDISCONNECTED) 611 return (ECONNABORTED); 612 613 inp = sotoinpcb(so); 614 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); 615 INP_LOCK(inp); 616 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 617 error = ECONNABORTED; 618 goto out; 619 } 620 tp = intotcpcb(inp); 621 TCPDEBUG1(); 622 623 /* 624 * We inline in_setpeeraddr and COMMON_END here, so that we can 625 * copy the data of interest and defer the malloc until after we 626 * release the lock. 627 */ 628 port = inp->inp_fport; 629 addr = inp->inp_faddr; 630 631out: 632 TCPDEBUG2(PRU_ACCEPT); 633 INP_UNLOCK(inp); 634 if (error == 0) 635 *nam = in_sockaddr(port, &addr); 636 return error; 637} 638 639#ifdef INET6 640static int 641tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 642{ 643 struct inpcb *inp = NULL; 644 int error = 0; 645 struct tcpcb *tp = NULL; 646 struct in_addr addr; 647 struct in6_addr addr6; 648 in_port_t port = 0; 649 int v4 = 0; 650 TCPDEBUG0; 651 652 if (so->so_state & SS_ISDISCONNECTED) { 653 error = ECONNABORTED; 654 goto out; 655 } 656 657 inp = sotoinpcb(so); 658 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); 659 INP_LOCK(inp); 660 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 661 error = EINVAL; 662 goto out; 663 } 664 tp = intotcpcb(inp); 665 TCPDEBUG1(); 666 667 /* 668 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can 669 * copy the data of interest and defer the malloc until after we 670 * release the lock. 671 */ 672 if (inp->inp_vflag & INP_IPV4) { 673 v4 = 1; 674 port = inp->inp_fport; 675 addr = inp->inp_faddr; 676 } else { 677 port = inp->inp_fport; 678 addr6 = inp->in6p_faddr; 679 } 680 681out: 682 TCPDEBUG2(PRU_ACCEPT); 683 INP_UNLOCK(inp); 684 if (error == 0) { 685 if (v4) 686 *nam = in6_v4mapsin6_sockaddr(port, &addr); 687 else 688 *nam = in6_sockaddr(port, &addr6); 689 } 690 return error; 691} 692#endif /* INET6 */ 693 694/* 695 * This is the wrapper function for in_setsockaddr. We just pass down 696 * the pcbinfo for in_setsockaddr to lock. We don't want to do the locking 697 * here because in_setsockaddr will call malloc and can block. 698 */ 699static int 700tcp_sockaddr(struct socket *so, struct sockaddr **nam) 701{ 702 return (in_setsockaddr(so, nam, &tcbinfo)); 703} 704 705/* 706 * This is the wrapper function for in_setpeeraddr. We just pass down 707 * the pcbinfo for in_setpeeraddr to lock. 708 */ 709static int 710tcp_peeraddr(struct socket *so, struct sockaddr **nam) 711{ 712 return (in_setpeeraddr(so, nam, &tcbinfo)); 713} 714 715/* 716 * Mark the connection as being incapable of further output. 717 */ 718static int 719tcp_usr_shutdown(struct socket *so) 720{ 721 int error = 0; 722 struct inpcb *inp; 723 struct tcpcb *tp = NULL; 724 725 TCPDEBUG0; 726 INP_INFO_WLOCK(&tcbinfo); 727 inp = sotoinpcb(so); 728 KASSERT(inp != NULL, ("inp == NULL")); 729 INP_LOCK(inp); 730 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 731 error = EINVAL; 732 goto out; 733 } 734 tp = intotcpcb(inp); 735 TCPDEBUG1(); 736 socantsendmore(so); 737 tcp_usrclosed(tp); 738 error = tcp_output(tp); 739 740out: 741 TCPDEBUG2(PRU_SHUTDOWN); 742 INP_UNLOCK(inp); 743 INP_INFO_WUNLOCK(&tcbinfo); 744 745 return (error); 746} 747 748/* 749 * After a receive, possibly send window update to peer. 750 */ 751static int 752tcp_usr_rcvd(struct socket *so, int flags) 753{ 754 struct inpcb *inp; 755 struct tcpcb *tp = NULL; 756 int error = 0; 757 758 TCPDEBUG0; 759 inp = sotoinpcb(so); 760 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); 761 INP_LOCK(inp); 762 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 763 error = EINVAL; 764 goto out; 765 } 766 tp = intotcpcb(inp); 767 TCPDEBUG1(); 768 tcp_output(tp); 769 770out: 771 TCPDEBUG2(PRU_RCVD); 772 INP_UNLOCK(inp); 773 return (error); 774} 775 776/* 777 * Do a send by putting data in output queue and updating urgent 778 * marker if URG set. Possibly send more data. Unlike the other 779 * pru_*() routines, the mbuf chains are our responsibility. We 780 * must either enqueue them or free them. The other pru_* routines 781 * generally are caller-frees. 782 */ 783static int 784tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 785 struct sockaddr *nam, struct mbuf *control, struct thread *td) 786{ 787 int error = 0; 788 struct inpcb *inp; 789 struct tcpcb *tp = NULL; 790 int headlocked = 0; 791#ifdef INET6 792 int isipv6; 793#endif 794 TCPDEBUG0; 795 796 /* 797 * We require the pcbinfo lock in two cases: 798 * 799 * (1) An implied connect is taking place, which can result in 800 * binding IPs and ports and hence modification of the pcb hash 801 * chains. 802 * 803 * (2) PRUS_EOF is set, resulting in explicit close on the send. 804 */ 805 if ((nam != NULL) || (flags & PRUS_EOF)) { 806 INP_INFO_WLOCK(&tcbinfo); 807 headlocked = 1; 808 } 809 inp = sotoinpcb(so); 810 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); 811 INP_LOCK(inp); 812 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 813 error = EINVAL; 814 goto out; 815 } 816#ifdef INET6 817 isipv6 = nam && nam->sa_family == AF_INET6; 818#endif /* INET6 */ 819 tp = intotcpcb(inp); 820 TCPDEBUG1(); 821 if (control) { 822 /* TCP doesn't do control messages (rights, creds, etc) */ 823 if (control->m_len) { 824 m_freem(control); 825 if (m) 826 m_freem(m); 827 error = EINVAL; 828 goto out; 829 } 830 m_freem(control); /* empty control, just free it */ 831 } 832 if (!(flags & PRUS_OOB)) { 833 sbappendstream(&so->so_snd, m); 834 if (nam && tp->t_state < TCPS_SYN_SENT) { 835 /* 836 * Do implied connect if not yet connected, 837 * initialize window to default value, and 838 * initialize maxseg/maxopd using peer's cached 839 * MSS. 840 */ 841 INP_INFO_WLOCK_ASSERT(&tcbinfo); 842#ifdef INET6 843 if (isipv6) 844 error = tcp6_connect(tp, nam, td); 845 else 846#endif /* INET6 */ 847 error = tcp_connect(tp, nam, td); 848 if (error) 849 goto out; 850 tp->snd_wnd = TTCP_CLIENT_SND_WND; 851 tcp_mss(tp, -1); 852 } 853 if (flags & PRUS_EOF) { 854 /* 855 * Close the send side of the connection after 856 * the data is sent. 857 */ 858 INP_INFO_WLOCK_ASSERT(&tcbinfo); 859 socantsendmore(so); 860 tcp_usrclosed(tp); 861 } 862 if (headlocked) { 863 INP_INFO_WUNLOCK(&tcbinfo); 864 headlocked = 0; 865 } 866 if (tp != NULL) { 867 if (flags & PRUS_MORETOCOME) 868 tp->t_flags |= TF_MORETOCOME; 869 error = tcp_output(tp); 870 if (flags & PRUS_MORETOCOME) 871 tp->t_flags &= ~TF_MORETOCOME; 872 } 873 } else { 874 /* 875 * XXXRW: PRUS_EOF not implemented with PRUS_OOB? 876 */ 877 SOCKBUF_LOCK(&so->so_snd); 878 if (sbspace(&so->so_snd) < -512) { 879 SOCKBUF_UNLOCK(&so->so_snd); 880 m_freem(m); 881 error = ENOBUFS; 882 goto out; 883 } 884 /* 885 * According to RFC961 (Assigned Protocols), 886 * the urgent pointer points to the last octet 887 * of urgent data. We continue, however, 888 * to consider it to indicate the first octet 889 * of data past the urgent section. 890 * Otherwise, snd_up should be one lower. 891 */ 892 sbappendstream_locked(&so->so_snd, m); 893 SOCKBUF_UNLOCK(&so->so_snd); 894 if (nam && tp->t_state < TCPS_SYN_SENT) { 895 /* 896 * Do implied connect if not yet connected, 897 * initialize window to default value, and 898 * initialize maxseg/maxopd using peer's cached 899 * MSS. 900 */ 901 INP_INFO_WLOCK_ASSERT(&tcbinfo); 902#ifdef INET6 903 if (isipv6) 904 error = tcp6_connect(tp, nam, td); 905 else 906#endif /* INET6 */ 907 error = tcp_connect(tp, nam, td); 908 if (error) 909 goto out; 910 tp->snd_wnd = TTCP_CLIENT_SND_WND; 911 tcp_mss(tp, -1); 912 INP_INFO_WUNLOCK(&tcbinfo); 913 headlocked = 0; 914 } else if (nam) { 915 INP_INFO_WUNLOCK(&tcbinfo); 916 headlocked = 0; 917 } 918 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 919 tp->t_flags |= TF_FORCEDATA; 920 error = tcp_output(tp); 921 tp->t_flags &= ~TF_FORCEDATA; 922 } 923out: 924 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB : 925 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 926 INP_UNLOCK(inp); 927 if (headlocked) 928 INP_INFO_WUNLOCK(&tcbinfo); 929 return (error); 930} 931 932/* 933 * Abort the TCP. 934 * 935 * First, drop the connection. Then collect state if possible. 936 */ 937static void 938tcp_usr_abort(struct socket *so) 939{ 940 struct inpcb *inp; 941 struct tcpcb *tp; 942 TCPDEBUG0; 943 944 inp = sotoinpcb(so); 945 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); 946 947 INP_INFO_WLOCK(&tcbinfo); 948 INP_LOCK(inp); 949 KASSERT(inp->inp_socket != NULL, 950 ("tcp_usr_abort: inp_socket == NULL")); 951 TCPDEBUG1(); 952 953 /* 954 * First, if we still have full TCP state, and we're not dropped, 955 * drop. 956 */ 957 if (!(inp->inp_vflag & INP_TIMEWAIT) && 958 !(inp->inp_vflag & INP_DROPPED)) { 959 tp = intotcpcb(inp); 960 tcp_drop(tp, ECONNABORTED); 961 } 962 tcp_detach(so, inp); 963 tp = NULL; 964 TCPDEBUG2(PRU_DETACH); 965 INP_INFO_WUNLOCK(&tcbinfo); 966} 967 968/* 969 * Receive out-of-band data. 970 */ 971static int 972tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 973{ 974 int error = 0; 975 struct inpcb *inp; 976 struct tcpcb *tp = NULL; 977 978 TCPDEBUG0; 979 inp = sotoinpcb(so); 980 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); 981 INP_LOCK(inp); 982 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 983 error = EINVAL; 984 goto out; 985 } 986 tp = intotcpcb(inp); 987 TCPDEBUG1(); 988 if ((so->so_oobmark == 0 && 989 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || 990 so->so_options & SO_OOBINLINE || 991 tp->t_oobflags & TCPOOB_HADDATA) { 992 error = EINVAL; 993 goto out; 994 } 995 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 996 error = EWOULDBLOCK; 997 goto out; 998 } 999 m->m_len = 1; 1000 *mtod(m, caddr_t) = tp->t_iobc; 1001 if ((flags & MSG_PEEK) == 0) 1002 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 1003 1004out: 1005 TCPDEBUG2(PRU_RCVOOB); 1006 INP_UNLOCK(inp); 1007 return (error); 1008} 1009 1010struct pr_usrreqs tcp_usrreqs = { 1011 .pru_abort = tcp_usr_abort, 1012 .pru_accept = tcp_usr_accept, 1013 .pru_attach = tcp_usr_attach, 1014 .pru_bind = tcp_usr_bind, 1015 .pru_connect = tcp_usr_connect, 1016 .pru_control = in_control, 1017 .pru_detach = tcp_usr_detach, 1018 .pru_disconnect = tcp_usr_disconnect, 1019 .pru_listen = tcp_usr_listen, 1020 .pru_peeraddr = tcp_peeraddr, 1021 .pru_rcvd = tcp_usr_rcvd, 1022 .pru_rcvoob = tcp_usr_rcvoob, 1023 .pru_send = tcp_usr_send, 1024 .pru_shutdown = tcp_usr_shutdown, 1025 .pru_sockaddr = tcp_sockaddr, 1026 .pru_sosetlabel = in_pcbsosetlabel 1027}; 1028 1029#ifdef INET6 1030struct pr_usrreqs tcp6_usrreqs = { 1031 .pru_abort = tcp_usr_abort, 1032 .pru_accept = tcp6_usr_accept, 1033 .pru_attach = tcp_usr_attach, 1034 .pru_bind = tcp6_usr_bind, 1035 .pru_connect = tcp6_usr_connect, 1036 .pru_control = in6_control, 1037 .pru_detach = tcp_usr_detach, 1038 .pru_disconnect = tcp_usr_disconnect, 1039 .pru_listen = tcp6_usr_listen, 1040 .pru_peeraddr = in6_mapped_peeraddr, 1041 .pru_rcvd = tcp_usr_rcvd, 1042 .pru_rcvoob = tcp_usr_rcvoob, 1043 .pru_send = tcp_usr_send, 1044 .pru_shutdown = tcp_usr_shutdown, 1045 .pru_sockaddr = in6_mapped_sockaddr, 1046 .pru_sosetlabel = in_pcbsosetlabel 1047}; 1048#endif /* INET6 */ 1049 1050/* 1051 * Common subroutine to open a TCP connection to remote host specified 1052 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 1053 * port number if needed. Call in_pcbconnect_setup to do the routing and 1054 * to choose a local host address (interface). If there is an existing 1055 * incarnation of the same connection in TIME-WAIT state and if the remote 1056 * host was sending CC options and if the connection duration was < MSL, then 1057 * truncate the previous TIME-WAIT state and proceed. 1058 * Initialize connection parameters and enter SYN-SENT state. 1059 */ 1060static int 1061tcp_connect(tp, nam, td) 1062 register struct tcpcb *tp; 1063 struct sockaddr *nam; 1064 struct thread *td; 1065{ 1066 struct inpcb *inp = tp->t_inpcb, *oinp; 1067 struct socket *so = inp->inp_socket; 1068 struct in_addr laddr; 1069 u_short lport; 1070 int error; 1071 1072 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1073 INP_LOCK_ASSERT(inp); 1074 1075 if (inp->inp_lport == 0) { 1076 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 1077 if (error) 1078 return error; 1079 } 1080 1081 /* 1082 * Cannot simply call in_pcbconnect, because there might be an 1083 * earlier incarnation of this same connection still in 1084 * TIME_WAIT state, creating an ADDRINUSE error. 1085 */ 1086 laddr = inp->inp_laddr; 1087 lport = inp->inp_lport; 1088 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport, 1089 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred); 1090 if (error && oinp == NULL) 1091 return error; 1092 if (oinp) 1093 return EADDRINUSE; 1094 inp->inp_laddr = laddr; 1095 in_pcbrehash(inp); 1096 1097 /* Compute window scaling to request. */ 1098 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1099 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 1100 tp->request_r_scale++; 1101 1102 soisconnecting(so); 1103 tcpstat.tcps_connattempt++; 1104 tp->t_state = TCPS_SYN_SENT; 1105 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 1106 tp->iss = tcp_new_isn(tp); 1107 tp->t_bw_rtseq = tp->iss; 1108 tcp_sendseqinit(tp); 1109 1110 return 0; 1111} 1112 1113#ifdef INET6 1114static int 1115tcp6_connect(tp, nam, td) 1116 register struct tcpcb *tp; 1117 struct sockaddr *nam; 1118 struct thread *td; 1119{ 1120 struct inpcb *inp = tp->t_inpcb, *oinp; 1121 struct socket *so = inp->inp_socket; 1122 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 1123 struct in6_addr *addr6; 1124 int error; 1125 1126 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1127 INP_LOCK_ASSERT(inp); 1128 1129 if (inp->inp_lport == 0) { 1130 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 1131 if (error) 1132 return error; 1133 } 1134 1135 /* 1136 * Cannot simply call in_pcbconnect, because there might be an 1137 * earlier incarnation of this same connection still in 1138 * TIME_WAIT state, creating an ADDRINUSE error. 1139 * in6_pcbladdr() also handles scope zone IDs. 1140 */ 1141 error = in6_pcbladdr(inp, nam, &addr6); 1142 if (error) 1143 return error; 1144 oinp = in6_pcblookup_hash(inp->inp_pcbinfo, 1145 &sin6->sin6_addr, sin6->sin6_port, 1146 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) 1147 ? addr6 1148 : &inp->in6p_laddr, 1149 inp->inp_lport, 0, NULL); 1150 if (oinp) 1151 return EADDRINUSE; 1152 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 1153 inp->in6p_laddr = *addr6; 1154 inp->in6p_faddr = sin6->sin6_addr; 1155 inp->inp_fport = sin6->sin6_port; 1156 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ 1157 inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK; 1158 if (inp->in6p_flags & IN6P_AUTOFLOWLABEL) 1159 inp->in6p_flowinfo |= 1160 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); 1161 in_pcbrehash(inp); 1162 1163 /* Compute window scaling to request. */ 1164 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1165 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 1166 tp->request_r_scale++; 1167 1168 soisconnecting(so); 1169 tcpstat.tcps_connattempt++; 1170 tp->t_state = TCPS_SYN_SENT; 1171 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 1172 tp->iss = tcp_new_isn(tp); 1173 tp->t_bw_rtseq = tp->iss; 1174 tcp_sendseqinit(tp); 1175 1176 return 0; 1177} 1178#endif /* INET6 */ 1179 1180/* 1181 * Export TCP internal state information via a struct tcp_info, based on the 1182 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently 1183 * (TCP state machine, etc). We export all information using FreeBSD-native 1184 * constants -- for example, the numeric values for tcpi_state will differ 1185 * from Linux. 1186 */ 1187static void 1188tcp_fill_info(tp, ti) 1189 struct tcpcb *tp; 1190 struct tcp_info *ti; 1191{ 1192 1193 INP_LOCK_ASSERT(tp->t_inpcb); 1194 bzero(ti, sizeof(*ti)); 1195 1196 ti->tcpi_state = tp->t_state; 1197 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) 1198 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; 1199 if (tp->sack_enable) 1200 ti->tcpi_options |= TCPI_OPT_SACK; 1201 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { 1202 ti->tcpi_options |= TCPI_OPT_WSCALE; 1203 ti->tcpi_snd_wscale = tp->snd_scale; 1204 ti->tcpi_rcv_wscale = tp->rcv_scale; 1205 } 1206 ti->tcpi_snd_ssthresh = tp->snd_ssthresh; 1207 ti->tcpi_snd_cwnd = tp->snd_cwnd; 1208 1209 /* 1210 * FreeBSD-specific extension fields for tcp_info. 1211 */ 1212 ti->tcpi_rcv_space = tp->rcv_wnd; 1213 ti->tcpi_snd_wnd = tp->snd_wnd; 1214 ti->tcpi_snd_bwnd = tp->snd_bwnd; 1215} 1216 1217/* 1218 * The new sockopt interface makes it possible for us to block in the 1219 * copyin/out step (if we take a page fault). Taking a page fault at 1220 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now 1221 * use TSM, there probably isn't any need for this function to run at 1222 * splnet() any more. This needs more examination.) 1223 * 1224 * XXXRW: The locking here is wrong; we may take a page fault while holding 1225 * the inpcb lock. 1226 */ 1227int 1228tcp_ctloutput(so, sopt) 1229 struct socket *so; 1230 struct sockopt *sopt; 1231{ 1232 int error, opt, optval; 1233 struct inpcb *inp; 1234 struct tcpcb *tp; 1235 struct tcp_info ti; 1236 1237 error = 0; 1238 inp = sotoinpcb(so); 1239 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); 1240 INP_LOCK(inp); 1241 if (sopt->sopt_level != IPPROTO_TCP) { 1242 INP_UNLOCK(inp); 1243#ifdef INET6 1244 if (INP_CHECK_SOCKAF(so, AF_INET6)) 1245 error = ip6_ctloutput(so, sopt); 1246 else 1247#endif /* INET6 */ 1248 error = ip_ctloutput(so, sopt); 1249 return (error); 1250 } 1251 if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { 1252 error = ECONNRESET; 1253 goto out; 1254 } 1255 tp = intotcpcb(inp); 1256 1257 switch (sopt->sopt_dir) { 1258 case SOPT_SET: 1259 switch (sopt->sopt_name) { 1260#ifdef TCP_SIGNATURE 1261 case TCP_MD5SIG: 1262 error = sooptcopyin(sopt, &optval, sizeof optval, 1263 sizeof optval); 1264 if (error) 1265 break; 1266 1267 if (optval > 0) 1268 tp->t_flags |= TF_SIGNATURE; 1269 else 1270 tp->t_flags &= ~TF_SIGNATURE; 1271 break; 1272#endif /* TCP_SIGNATURE */ 1273 case TCP_NODELAY: 1274 case TCP_NOOPT: 1275 error = sooptcopyin(sopt, &optval, sizeof optval, 1276 sizeof optval); 1277 if (error) 1278 break; 1279 1280 switch (sopt->sopt_name) { 1281 case TCP_NODELAY: 1282 opt = TF_NODELAY; 1283 break; 1284 case TCP_NOOPT: 1285 opt = TF_NOOPT; 1286 break; 1287 default: 1288 opt = 0; /* dead code to fool gcc */ 1289 break; 1290 } 1291 1292 if (optval) 1293 tp->t_flags |= opt; 1294 else 1295 tp->t_flags &= ~opt; 1296 break; 1297 1298 case TCP_NOPUSH: 1299 error = sooptcopyin(sopt, &optval, sizeof optval, 1300 sizeof optval); 1301 if (error) 1302 break; 1303 1304 if (optval) 1305 tp->t_flags |= TF_NOPUSH; 1306 else { 1307 tp->t_flags &= ~TF_NOPUSH; 1308 error = tcp_output(tp); 1309 } 1310 break; 1311 1312 case TCP_MAXSEG: 1313 error = sooptcopyin(sopt, &optval, sizeof optval, 1314 sizeof optval); 1315 if (error) 1316 break; 1317 1318 if (optval > 0 && optval <= tp->t_maxseg && 1319 optval + 40 >= tcp_minmss) 1320 tp->t_maxseg = optval; 1321 else 1322 error = EINVAL; 1323 break; 1324 1325 case TCP_INFO: 1326 error = EINVAL; 1327 break; 1328 1329 default: 1330 error = ENOPROTOOPT; 1331 break; 1332 } 1333 break; 1334 1335 case SOPT_GET: 1336 switch (sopt->sopt_name) { 1337#ifdef TCP_SIGNATURE 1338 case TCP_MD5SIG: 1339 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; 1340 error = sooptcopyout(sopt, &optval, sizeof optval); 1341 break; 1342#endif 1343 case TCP_NODELAY: 1344 optval = tp->t_flags & TF_NODELAY; 1345 error = sooptcopyout(sopt, &optval, sizeof optval); 1346 break; 1347 case TCP_MAXSEG: 1348 optval = tp->t_maxseg; 1349 error = sooptcopyout(sopt, &optval, sizeof optval); 1350 break; 1351 case TCP_NOOPT: 1352 optval = tp->t_flags & TF_NOOPT; 1353 error = sooptcopyout(sopt, &optval, sizeof optval); 1354 break; 1355 case TCP_NOPUSH: 1356 optval = tp->t_flags & TF_NOPUSH; 1357 error = sooptcopyout(sopt, &optval, sizeof optval); 1358 break; 1359 case TCP_INFO: 1360 tcp_fill_info(tp, &ti); 1361 error = sooptcopyout(sopt, &ti, sizeof ti); 1362 break; 1363 default: 1364 error = ENOPROTOOPT; 1365 break; 1366 } 1367 break; 1368 } 1369out: 1370 INP_UNLOCK(inp); 1371 return (error); 1372} 1373 1374/* 1375 * tcp_sendspace and tcp_recvspace are the default send and receive window 1376 * sizes, respectively. These are obsolescent (this information should 1377 * be set by the route). 1378 */ 1379u_long tcp_sendspace = 1024*32; 1380SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1381 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1382u_long tcp_recvspace = 1024*64; 1383SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1384 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1385 1386/* 1387 * Attach TCP protocol to socket, allocating 1388 * internet protocol control block, tcp control block, 1389 * bufer space, and entering LISTEN state if to accept connections. 1390 */ 1391static int 1392tcp_attach(so) 1393 struct socket *so; 1394{ 1395 register struct tcpcb *tp; 1396 struct inpcb *inp; 1397 int error; 1398#ifdef INET6 1399 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0; 1400#endif 1401 1402 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1403 1404 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1405 error = soreserve(so, tcp_sendspace, tcp_recvspace); 1406 if (error) 1407 return (error); 1408 } 1409 error = in_pcballoc(so, &tcbinfo, "tcpinp"); 1410 if (error) 1411 return (error); 1412 inp = sotoinpcb(so); 1413#ifdef INET6 1414 if (isipv6) { 1415 inp->inp_vflag |= INP_IPV6; 1416 inp->in6p_hops = -1; /* use kernel default */ 1417 } 1418 else 1419#endif 1420 inp->inp_vflag |= INP_IPV4; 1421 tp = tcp_newtcpcb(inp); 1422 if (tp == NULL) { 1423 INP_LOCK(inp); 1424#ifdef INET6 1425 if (isipv6) { 1426 in6_pcbdetach(inp); 1427 in6_pcbfree(inp); 1428 } else { 1429#endif 1430 in_pcbdetach(inp); 1431 in_pcbfree(inp); 1432#ifdef INET6 1433 } 1434#endif 1435 return (ENOBUFS); 1436 } 1437 tp->t_state = TCPS_CLOSED; 1438 return (0); 1439} 1440 1441/* 1442 * Initiate (or continue) disconnect. 1443 * If embryonic state, just send reset (once). 1444 * If in ``let data drain'' option and linger null, just drop. 1445 * Otherwise (hard), mark socket disconnecting and drop 1446 * current input data; switch states based on user close, and 1447 * send segment to peer (with FIN). 1448 */ 1449static void 1450tcp_disconnect(tp) 1451 register struct tcpcb *tp; 1452{ 1453 struct inpcb *inp = tp->t_inpcb; 1454 struct socket *so = inp->inp_socket; 1455 1456 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1457 INP_LOCK_ASSERT(inp); 1458 1459 /* 1460 * Neither tcp_close() nor tcp_drop() should return NULL, as the 1461 * socket is still open. 1462 */ 1463 if (tp->t_state < TCPS_ESTABLISHED) { 1464 tp = tcp_close(tp); 1465 KASSERT(tp != NULL, 1466 ("tcp_disconnect: tcp_close() returned NULL")); 1467 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { 1468 tp = tcp_drop(tp, 0); 1469 KASSERT(tp != NULL, 1470 ("tcp_disconnect: tcp_drop() returned NULL")); 1471 } else { 1472 soisdisconnecting(so); 1473 sbflush(&so->so_rcv); 1474 tcp_usrclosed(tp); 1475 if (!(inp->inp_vflag & INP_DROPPED)) 1476 tcp_output(tp); 1477 } 1478} 1479 1480/* 1481 * User issued close, and wish to trail through shutdown states: 1482 * if never received SYN, just forget it. If got a SYN from peer, 1483 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1484 * If already got a FIN from peer, then almost done; go to LAST_ACK 1485 * state. In all other cases, have already sent FIN to peer (e.g. 1486 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1487 * for peer to send FIN or not respond to keep-alives, etc. 1488 * We can let the user exit from the close as soon as the FIN is acked. 1489 */ 1490static void 1491tcp_usrclosed(tp) 1492 register struct tcpcb *tp; 1493{ 1494 1495 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1496 INP_LOCK_ASSERT(tp->t_inpcb); 1497 1498 switch (tp->t_state) { 1499 1500 case TCPS_CLOSED: 1501 case TCPS_LISTEN: 1502 tp->t_state = TCPS_CLOSED; 1503 tp = tcp_close(tp); 1504 /* 1505 * tcp_close() should never return NULL here as the socket is 1506 * still open. 1507 */ 1508 KASSERT(tp != NULL, 1509 ("tcp_usrclosed: tcp_close() returned NULL")); 1510 break; 1511 1512 case TCPS_SYN_SENT: 1513 case TCPS_SYN_RECEIVED: 1514 tp->t_flags |= TF_NEEDFIN; 1515 break; 1516 1517 case TCPS_ESTABLISHED: 1518 tp->t_state = TCPS_FIN_WAIT_1; 1519 break; 1520 1521 case TCPS_CLOSE_WAIT: 1522 tp->t_state = TCPS_LAST_ACK; 1523 break; 1524 } 1525 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1526 soisdisconnected(tp->t_inpcb->inp_socket); 1527 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1528 if (tp->t_state == TCPS_FIN_WAIT_2) 1529 callout_reset(tp->tt_2msl, tcp_maxidle, 1530 tcp_timer_2msl, tp); 1531 } 1532} 1533