tcp_usrreq.c revision 22719
1/* 2 * Copyright (c) 1982, 1986, 1988, 1993 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 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 34 * $FreeBSD: head/sys/netinet/tcp_usrreq.c 22719 1997-02-14 18:15:53Z wollman $ 35 */ 36 37#include <sys/param.h> 38#include <sys/queue.h> 39#include <sys/systm.h> 40#include <sys/kernel.h> 41#include <sys/sysctl.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/socket.h> 45#include <sys/socketvar.h> 46#include <sys/protosw.h> 47#include <sys/errno.h> 48#include <sys/stat.h> 49 50#include <net/if.h> 51#include <net/route.h> 52 53#include <netinet/in.h> 54#include <netinet/in_systm.h> 55#include <netinet/ip.h> 56#include <netinet/in_pcb.h> 57#include <netinet/in_var.h> 58#include <netinet/ip_var.h> 59#include <netinet/tcp.h> 60#include <netinet/tcp_fsm.h> 61#include <netinet/tcp_seq.h> 62#include <netinet/tcp_timer.h> 63#include <netinet/tcp_var.h> 64#include <netinet/tcpip.h> 65#ifdef TCPDEBUG 66#include <netinet/tcp_debug.h> 67#endif 68 69/* 70 * TCP protocol interface to socket abstraction. 71 */ 72extern char *tcpstates[]; 73 74static int tcp_attach __P((struct socket *)); 75static int tcp_connect __P((struct tcpcb *, struct mbuf *)); 76static struct tcpcb * 77 tcp_disconnect __P((struct tcpcb *)); 78static struct tcpcb * 79 tcp_usrclosed __P((struct tcpcb *)); 80 81#ifdef TCPDEBUG 82#define TCPDEBUG0 int ostate 83#define TCPDEBUG1() ostate = tp ? tp->t_state : 0 84#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ 85 tcp_trace(TA_USER, ostate, tp, 0, req) 86#else 87#define TCPDEBUG0 88#define TCPDEBUG1() 89#define TCPDEBUG2(req) 90#endif 91 92/* 93 * TCP attaches to socket via pru_attach(), reserving space, 94 * and an internet control block. 95 */ 96static int 97tcp_usr_attach(struct socket *so, int proto) 98{ 99 int s = splnet(); 100 int error; 101 struct inpcb *inp = sotoinpcb(so); 102 struct tcpcb *tp = 0; 103 TCPDEBUG0; 104 105 TCPDEBUG1(); 106 if (inp) { 107 error = EISCONN; 108 goto out; 109 } 110 111 error = tcp_attach(so); 112 if (error) 113 goto out; 114 115 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 116 so->so_linger = TCP_LINGERTIME * hz; 117 tp = sototcpcb(so); 118out: 119 TCPDEBUG2(PRU_ATTACH); 120 splx(s); 121 return error; 122} 123 124/* 125 * pru_detach() detaches the TCP protocol from the socket. 126 * If the protocol state is non-embryonic, then can't 127 * do this directly: have to initiate a pru_disconnect(), 128 * which may finish later; embryonic TCB's can just 129 * be discarded here. 130 */ 131static int 132tcp_usr_detach(struct socket *so) 133{ 134 int s = splnet(); 135 int error = 0; 136 struct inpcb *inp = sotoinpcb(so); 137 struct tcpcb *tp; 138 TCPDEBUG0; 139 140 if (inp == 0) { 141 splx(s); 142 return EINVAL; /* XXX */ 143 } 144 tp = intotcpcb(inp); 145 TCPDEBUG1(); 146 if (tp->t_state > TCPS_LISTEN) 147 tp = tcp_disconnect(tp); 148 else 149 tp = tcp_close(tp); 150 151 TCPDEBUG2(PRU_DETACH); 152 splx(s); 153 return error; 154} 155 156#define COMMON_START() TCPDEBUG0; \ 157 do { \ 158 if (inp == 0) { \ 159 splx(s); \ 160 return EINVAL; \ 161 } \ 162 tp = intotcpcb(inp); \ 163 TCPDEBUG1(); \ 164 } while(0) 165 166#define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out 167 168 169/* 170 * Give the socket an address. 171 */ 172static int 173tcp_usr_bind(struct socket *so, struct mbuf *nam) 174{ 175 int s = splnet(); 176 int error = 0; 177 struct inpcb *inp = sotoinpcb(so); 178 struct tcpcb *tp; 179 struct sockaddr_in *sinp; 180 181 COMMON_START(); 182 183 /* 184 * Must check for multicast addresses and disallow binding 185 * to them. 186 */ 187 sinp = mtod(nam, struct sockaddr_in *); 188 if (sinp->sin_family == AF_INET && 189 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 190 error = EAFNOSUPPORT; 191 goto out; 192 } 193 error = in_pcbbind(inp, nam); 194 if (error) 195 goto out; 196 COMMON_END(PRU_BIND); 197 198} 199 200/* 201 * Prepare to accept connections. 202 */ 203static int 204tcp_usr_listen(struct socket *so) 205{ 206 int s = splnet(); 207 int error = 0; 208 struct inpcb *inp = sotoinpcb(so); 209 struct tcpcb *tp; 210 211 COMMON_START(); 212 if (inp->inp_lport == 0) 213 error = in_pcbbind(inp, NULL); 214 if (error == 0) 215 tp->t_state = TCPS_LISTEN; 216 COMMON_END(PRU_LISTEN); 217} 218 219/* 220 * Initiate connection to peer. 221 * Create a template for use in transmissions on this connection. 222 * Enter SYN_SENT state, and mark socket as connecting. 223 * Start keep-alive timer, and seed output sequence space. 224 * Send initial segment on connection. 225 */ 226static int 227tcp_usr_connect(struct socket *so, struct mbuf *nam) 228{ 229 int s = splnet(); 230 int error = 0; 231 struct inpcb *inp = sotoinpcb(so); 232 struct tcpcb *tp; 233 struct sockaddr_in *sinp; 234 235 COMMON_START(); 236 237 /* 238 * Must disallow TCP ``connections'' to multicast addresses. 239 */ 240 sinp = mtod(nam, struct sockaddr_in *); 241 if (sinp->sin_family == AF_INET 242 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 243 error = EAFNOSUPPORT; 244 goto out; 245 } 246 247 if ((error = tcp_connect(tp, nam)) != 0) 248 goto out; 249 error = tcp_output(tp); 250 COMMON_END(PRU_CONNECT); 251} 252 253/* 254 * Initiate disconnect from peer. 255 * If connection never passed embryonic stage, just drop; 256 * else if don't need to let data drain, then can just drop anyways, 257 * else have to begin TCP shutdown process: mark socket disconnecting, 258 * drain unread data, state switch to reflect user close, and 259 * send segment (e.g. FIN) to peer. Socket will be really disconnected 260 * when peer sends FIN and acks ours. 261 * 262 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 263 */ 264static int 265tcp_usr_disconnect(struct socket *so) 266{ 267 int s = splnet(); 268 int error = 0; 269 struct inpcb *inp = sotoinpcb(so); 270 struct tcpcb *tp; 271 272 COMMON_START(); 273 tp = tcp_disconnect(tp); 274 COMMON_END(PRU_DISCONNECT); 275} 276 277/* 278 * Accept a connection. Essentially all the work is 279 * done at higher levels; just return the address 280 * of the peer, storing through addr. 281 */ 282static int 283tcp_usr_accept(struct socket *so, struct mbuf *nam) 284{ 285 int s = splnet(); 286 int error = 0; 287 struct inpcb *inp = sotoinpcb(so); 288 struct tcpcb *tp; 289 290 COMMON_START(); 291 in_setpeeraddr(inp, nam); 292 COMMON_END(PRU_ACCEPT); 293} 294 295/* 296 * Mark the connection as being incapable of further output. 297 */ 298static int 299tcp_usr_shutdown(struct socket *so) 300{ 301 int s = splnet(); 302 int error = 0; 303 struct inpcb *inp = sotoinpcb(so); 304 struct tcpcb *tp; 305 306 COMMON_START(); 307 socantsendmore(so); 308 tp = tcp_usrclosed(tp); 309 if (tp) 310 error = tcp_output(tp); 311 COMMON_END(PRU_SHUTDOWN); 312} 313 314/* 315 * After a receive, possibly send window update to peer. 316 */ 317static int 318tcp_usr_rcvd(struct socket *so, int flags) 319{ 320 int s = splnet(); 321 int error = 0; 322 struct inpcb *inp = sotoinpcb(so); 323 struct tcpcb *tp; 324 325 COMMON_START(); 326 tcp_output(tp); 327 COMMON_END(PRU_RCVD); 328} 329 330/* 331 * Do a send by putting data in output queue and updating urgent 332 * marker if URG set. Possibly send more data. 333 */ 334static int 335tcp_usr_send(struct socket *so, int flags, struct mbuf *m, struct mbuf *nam, 336 struct mbuf *control) 337{ 338 int s = splnet(); 339 int error = 0; 340 struct inpcb *inp = sotoinpcb(so); 341 struct tcpcb *tp; 342 343 COMMON_START(); 344 if (control && control->m_len) { 345 m_freem(control); /* XXX shouldn't caller do this??? */ 346 if (m) 347 m_freem(m); 348 return EINVAL; 349 } 350 351 if(!(flags & PRUS_OOB)) { 352 sbappend(&so->so_snd, m); 353 if (nam && tp->t_state < TCPS_SYN_SENT) { 354 /* 355 * Do implied connect if not yet connected, 356 * initialize window to default value, and 357 * initialize maxseg/maxopd using peer's cached 358 * MSS. 359 */ 360 error = tcp_connect(tp, nam); 361 if (error) 362 goto out; 363 tp->snd_wnd = TTCP_CLIENT_SND_WND; 364 tcp_mss(tp, -1); 365 } 366 367 if (flags & PRUS_EOF) { 368 /* 369 * Close the send side of the connection after 370 * the data is sent. 371 */ 372 socantsendmore(so); 373 tp = tcp_usrclosed(tp); 374 } 375 if (tp != NULL) 376 error = tcp_output(tp); 377 } else { 378 if (sbspace(&so->so_snd) < -512) { 379 m_freem(m); 380 error = ENOBUFS; 381 goto out; 382 } 383 /* 384 * According to RFC961 (Assigned Protocols), 385 * the urgent pointer points to the last octet 386 * of urgent data. We continue, however, 387 * to consider it to indicate the first octet 388 * of data past the urgent section. 389 * Otherwise, snd_up should be one lower. 390 */ 391 sbappend(&so->so_snd, m); 392 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 393 tp->t_force = 1; 394 error = tcp_output(tp); 395 tp->t_force = 0; 396 } 397 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 398 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 399} 400 401/* 402 * Abort the TCP. 403 */ 404static int 405tcp_usr_abort(struct socket *so) 406{ 407 int s = splnet(); 408 int error = 0; 409 struct inpcb *inp = sotoinpcb(so); 410 struct tcpcb *tp; 411 412 COMMON_START(); 413 tp = tcp_drop(tp, ECONNABORTED); 414 COMMON_END(PRU_ABORT); 415} 416 417/* 418 * Fill in st_bklsize for fstat() operations on a socket. 419 */ 420static int 421tcp_usr_sense(struct socket *so, struct stat *sb) 422{ 423 int s = splnet(); 424 425 sb->st_blksize = so->so_snd.sb_hiwat; 426 splx(s); 427 return 0; 428} 429 430/* 431 * Receive out-of-band data. 432 */ 433static int 434tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 435{ 436 int s = splnet(); 437 int error = 0; 438 struct inpcb *inp = sotoinpcb(so); 439 struct tcpcb *tp; 440 441 COMMON_START(); 442 if ((so->so_oobmark == 0 && 443 (so->so_state & SS_RCVATMARK) == 0) || 444 so->so_options & SO_OOBINLINE || 445 tp->t_oobflags & TCPOOB_HADDATA) { 446 error = EINVAL; 447 goto out; 448 } 449 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 450 error = EWOULDBLOCK; 451 goto out; 452 } 453 m->m_len = 1; 454 *mtod(m, caddr_t) = tp->t_iobc; 455 if ((flags & MSG_PEEK) == 0) 456 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 457 COMMON_END(PRU_RCVOOB); 458} 459 460static int 461tcp_usr_sockaddr(struct socket *so, struct mbuf *nam) 462{ 463 int s = splnet(); 464 int error = 0; 465 struct inpcb *inp = sotoinpcb(so); 466 struct tcpcb *tp; 467 468 COMMON_START(); 469 in_setsockaddr(inp, nam); 470 COMMON_END(PRU_SOCKADDR); 471} 472 473static int 474tcp_usr_peeraddr(struct socket *so, struct mbuf *nam) 475{ 476 int s = splnet(); 477 int error = 0; 478 struct inpcb *inp = sotoinpcb(so); 479 struct tcpcb *tp; 480 481 COMMON_START(); 482 in_setpeeraddr(inp, nam); 483 COMMON_END(PRU_PEERADDR); 484} 485 486/* 487 * XXX - this should just be a call to in_control, but we need to get 488 * the types worked out. 489 */ 490static int 491tcp_usr_control(struct socket *so, int cmd, caddr_t arg, struct ifnet *ifp) 492{ 493 return in_control(so, cmd, arg, ifp); 494} 495 496/* xxx - should be const */ 497struct pr_usrreqs tcp_usrreqs = { 498 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 499 tcp_usr_connect, pru_connect2_notsupp, tcp_usr_control, tcp_usr_detach, 500 tcp_usr_disconnect, tcp_usr_listen, tcp_usr_peeraddr, tcp_usr_rcvd, 501 tcp_usr_rcvoob, tcp_usr_send, tcp_usr_sense, tcp_usr_shutdown, 502 tcp_usr_sockaddr 503}; 504 505/* 506 * Common subroutine to open a TCP connection to remote host specified 507 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 508 * port number if needed. Call in_pcbladdr to do the routing and to choose 509 * a local host address (interface). If there is an existing incarnation 510 * of the same connection in TIME-WAIT state and if the remote host was 511 * sending CC options and if the connection duration was < MSL, then 512 * truncate the previous TIME-WAIT state and proceed. 513 * Initialize connection parameters and enter SYN-SENT state. 514 */ 515static int 516tcp_connect(tp, nam) 517 register struct tcpcb *tp; 518 struct mbuf *nam; 519{ 520 struct inpcb *inp = tp->t_inpcb, *oinp; 521 struct socket *so = inp->inp_socket; 522 struct tcpcb *otp; 523 struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *); 524 struct sockaddr_in *ifaddr; 525 int error; 526 struct rmxp_tao *taop; 527 struct rmxp_tao tao_noncached; 528 529 if (inp->inp_lport == 0) { 530 error = in_pcbbind(inp, NULL); 531 if (error) 532 return error; 533 } 534 535 /* 536 * Cannot simply call in_pcbconnect, because there might be an 537 * earlier incarnation of this same connection still in 538 * TIME_WAIT state, creating an ADDRINUSE error. 539 */ 540 error = in_pcbladdr(inp, nam, &ifaddr); 541 if (error) 542 return error; 543 oinp = in_pcblookuphash(inp->inp_pcbinfo, 544 sin->sin_addr, sin->sin_port, 545 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr 546 : ifaddr->sin_addr, 547 inp->inp_lport, 0); 548 if (oinp) { 549 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 550 otp->t_state == TCPS_TIME_WAIT && 551 otp->t_duration < TCPTV_MSL && 552 (otp->t_flags & TF_RCVD_CC)) 553 otp = tcp_close(otp); 554 else 555 return EADDRINUSE; 556 } 557 if (inp->inp_laddr.s_addr == INADDR_ANY) 558 inp->inp_laddr = ifaddr->sin_addr; 559 inp->inp_faddr = sin->sin_addr; 560 inp->inp_fport = sin->sin_port; 561 in_pcbrehash(inp); 562 563 tp->t_template = tcp_template(tp); 564 if (tp->t_template == 0) { 565 in_pcbdisconnect(inp); 566 return ENOBUFS; 567 } 568 569 /* Compute window scaling to request. */ 570 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 571 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 572 tp->request_r_scale++; 573 574 soisconnecting(so); 575 tcpstat.tcps_connattempt++; 576 tp->t_state = TCPS_SYN_SENT; 577 tp->t_timer[TCPT_KEEP] = tcp_keepinit; 578 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; 579 tcp_sendseqinit(tp); 580 581 /* 582 * Generate a CC value for this connection and 583 * check whether CC or CCnew should be used. 584 */ 585 if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { 586 taop = &tao_noncached; 587 bzero(taop, sizeof(*taop)); 588 } 589 590 tp->cc_send = CC_INC(tcp_ccgen); 591 if (taop->tao_ccsent != 0 && 592 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 593 taop->tao_ccsent = tp->cc_send; 594 } else { 595 taop->tao_ccsent = 0; 596 tp->t_flags |= TF_SENDCCNEW; 597 } 598 599 return 0; 600} 601 602int 603tcp_ctloutput(op, so, level, optname, mp) 604 int op; 605 struct socket *so; 606 int level, optname; 607 struct mbuf **mp; 608{ 609 int error = 0, s; 610 struct inpcb *inp; 611 register struct tcpcb *tp; 612 register struct mbuf *m; 613 register int i; 614 615 s = splnet(); 616 inp = sotoinpcb(so); 617 if (inp == NULL) { 618 splx(s); 619 if (op == PRCO_SETOPT && *mp) 620 (void) m_free(*mp); 621 return (ECONNRESET); 622 } 623 if (level != IPPROTO_TCP) { 624 error = ip_ctloutput(op, so, level, optname, mp); 625 splx(s); 626 return (error); 627 } 628 tp = intotcpcb(inp); 629 630 switch (op) { 631 632 case PRCO_SETOPT: 633 m = *mp; 634 switch (optname) { 635 636 case TCP_NODELAY: 637 if (m == NULL || m->m_len < sizeof (int)) 638 error = EINVAL; 639 else if (*mtod(m, int *)) 640 tp->t_flags |= TF_NODELAY; 641 else 642 tp->t_flags &= ~TF_NODELAY; 643 break; 644 645 case TCP_MAXSEG: 646 if (m && (i = *mtod(m, int *)) > 0 && i <= tp->t_maxseg) 647 tp->t_maxseg = i; 648 else 649 error = EINVAL; 650 break; 651 652 case TCP_NOOPT: 653 if (m == NULL || m->m_len < sizeof (int)) 654 error = EINVAL; 655 else if (*mtod(m, int *)) 656 tp->t_flags |= TF_NOOPT; 657 else 658 tp->t_flags &= ~TF_NOOPT; 659 break; 660 661 case TCP_NOPUSH: 662 if (m == NULL || m->m_len < sizeof (int)) 663 error = EINVAL; 664 else if (*mtod(m, int *)) 665 tp->t_flags |= TF_NOPUSH; 666 else 667 tp->t_flags &= ~TF_NOPUSH; 668 break; 669 670 default: 671 error = ENOPROTOOPT; 672 break; 673 } 674 if (m) 675 (void) m_free(m); 676 break; 677 678 case PRCO_GETOPT: 679 *mp = m = m_get(M_WAIT, MT_SOOPTS); 680 m->m_len = sizeof(int); 681 682 switch (optname) { 683 case TCP_NODELAY: 684 *mtod(m, int *) = tp->t_flags & TF_NODELAY; 685 break; 686 case TCP_MAXSEG: 687 *mtod(m, int *) = tp->t_maxseg; 688 break; 689 case TCP_NOOPT: 690 *mtod(m, int *) = tp->t_flags & TF_NOOPT; 691 break; 692 case TCP_NOPUSH: 693 *mtod(m, int *) = tp->t_flags & TF_NOPUSH; 694 break; 695 default: 696 error = ENOPROTOOPT; 697 break; 698 } 699 break; 700 } 701 splx(s); 702 return (error); 703} 704 705/* 706 * tcp_sendspace and tcp_recvspace are the default send and receive window 707 * sizes, respectively. These are obsolescent (this information should 708 * be set by the route). 709 */ 710u_long tcp_sendspace = 1024*16; 711SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, 712 CTLFLAG_RW, &tcp_sendspace , 0, ""); 713u_long tcp_recvspace = 1024*16; 714SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, 715 CTLFLAG_RW, &tcp_recvspace , 0, ""); 716 717/* 718 * Attach TCP protocol to socket, allocating 719 * internet protocol control block, tcp control block, 720 * bufer space, and entering LISTEN state if to accept connections. 721 */ 722static int 723tcp_attach(so) 724 struct socket *so; 725{ 726 register struct tcpcb *tp; 727 struct inpcb *inp; 728 int error; 729 730 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 731 error = soreserve(so, tcp_sendspace, tcp_recvspace); 732 if (error) 733 return (error); 734 } 735 error = in_pcballoc(so, &tcbinfo); 736 if (error) 737 return (error); 738 inp = sotoinpcb(so); 739 tp = tcp_newtcpcb(inp); 740 if (tp == 0) { 741 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 742 743 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 744 in_pcbdetach(inp); 745 so->so_state |= nofd; 746 return (ENOBUFS); 747 } 748 tp->t_state = TCPS_CLOSED; 749 return (0); 750} 751 752/* 753 * Initiate (or continue) disconnect. 754 * If embryonic state, just send reset (once). 755 * If in ``let data drain'' option and linger null, just drop. 756 * Otherwise (hard), mark socket disconnecting and drop 757 * current input data; switch states based on user close, and 758 * send segment to peer (with FIN). 759 */ 760static struct tcpcb * 761tcp_disconnect(tp) 762 register struct tcpcb *tp; 763{ 764 struct socket *so = tp->t_inpcb->inp_socket; 765 766 if (tp->t_state < TCPS_ESTABLISHED) 767 tp = tcp_close(tp); 768 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 769 tp = tcp_drop(tp, 0); 770 else { 771 soisdisconnecting(so); 772 sbflush(&so->so_rcv); 773 tp = tcp_usrclosed(tp); 774 if (tp) 775 (void) tcp_output(tp); 776 } 777 return (tp); 778} 779 780/* 781 * User issued close, and wish to trail through shutdown states: 782 * if never received SYN, just forget it. If got a SYN from peer, 783 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 784 * If already got a FIN from peer, then almost done; go to LAST_ACK 785 * state. In all other cases, have already sent FIN to peer (e.g. 786 * after PRU_SHUTDOWN), and just have to play tedious game waiting 787 * for peer to send FIN or not respond to keep-alives, etc. 788 * We can let the user exit from the close as soon as the FIN is acked. 789 */ 790static struct tcpcb * 791tcp_usrclosed(tp) 792 register struct tcpcb *tp; 793{ 794 795 switch (tp->t_state) { 796 797 case TCPS_CLOSED: 798 case TCPS_LISTEN: 799 tp->t_state = TCPS_CLOSED; 800 tp = tcp_close(tp); 801 break; 802 803 case TCPS_SYN_SENT: 804 case TCPS_SYN_RECEIVED: 805 tp->t_flags |= TF_NEEDFIN; 806 break; 807 808 case TCPS_ESTABLISHED: 809 tp->t_state = TCPS_FIN_WAIT_1; 810 break; 811 812 case TCPS_CLOSE_WAIT: 813 tp->t_state = TCPS_LAST_ACK; 814 break; 815 } 816 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 817 soisdisconnected(tp->t_inpcb->inp_socket); 818 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 819 if (tp->t_state == TCPS_FIN_WAIT_2) 820 tp->t_timer[TCPT_2MSL] = tcp_maxidle; 821 } 822 return (tp); 823} 824 825