tcp_usrreq.c revision 74134
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 74134 2001-03-12 02:57:42Z jlemon $ 35 */ 36 37#include "opt_ipsec.h" 38#include "opt_inet6.h" 39#include "opt_tcpdebug.h" 40 41#include <sys/param.h> 42#include <sys/systm.h> 43#include <sys/kernel.h> 44#include <sys/sysctl.h> 45#include <sys/mbuf.h> 46#ifdef INET6 47#include <sys/domain.h> 48#endif /* INET6 */ 49#include <sys/socket.h> 50#include <sys/socketvar.h> 51#include <sys/protosw.h> 52#include <sys/proc.h> 53#include <sys/jail.h> 54 55#include <net/if.h> 56#include <net/route.h> 57 58#include <netinet/in.h> 59#include <netinet/in_systm.h> 60#ifdef INET6 61#include <netinet/ip6.h> 62#endif 63#include <netinet/in_pcb.h> 64#ifdef INET6 65#include <netinet6/in6_pcb.h> 66#endif 67#include <netinet/in_var.h> 68#include <netinet/ip_var.h> 69#ifdef INET6 70#include <netinet6/ip6_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#ifdef IPSEC 83#include <netinet6/ipsec.h> 84#endif /*IPSEC*/ 85 86/* 87 * TCP protocol interface to socket abstraction. 88 */ 89extern char *tcpstates[]; /* XXX ??? */ 90 91static int tcp_attach __P((struct socket *, struct proc *)); 92static int tcp_connect __P((struct tcpcb *, struct sockaddr *, 93 struct proc *)); 94#ifdef INET6 95static int tcp6_connect __P((struct tcpcb *, struct sockaddr *, 96 struct proc *)); 97#endif /* INET6 */ 98static struct tcpcb * 99 tcp_disconnect __P((struct tcpcb *)); 100static struct tcpcb * 101 tcp_usrclosed __P((struct tcpcb *)); 102 103#ifdef TCPDEBUG 104#define TCPDEBUG0 int ostate = 0 105#define TCPDEBUG1() ostate = tp ? tp->t_state : 0 106#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ 107 tcp_trace(TA_USER, ostate, tp, 0, 0, req) 108#else 109#define TCPDEBUG0 110#define TCPDEBUG1() 111#define TCPDEBUG2(req) 112#endif 113 114/* 115 * TCP attaches to socket via pru_attach(), reserving space, 116 * and an internet control block. 117 */ 118static int 119tcp_usr_attach(struct socket *so, int proto, struct proc *p) 120{ 121 int s = splnet(); 122 int error; 123 struct inpcb *inp = sotoinpcb(so); 124 struct tcpcb *tp = 0; 125 TCPDEBUG0; 126 127 TCPDEBUG1(); 128 if (inp) { 129 error = EISCONN; 130 goto out; 131 } 132 133 error = tcp_attach(so, p); 134 if (error) 135 goto out; 136 137 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 138 so->so_linger = TCP_LINGERTIME; 139 tp = sototcpcb(so); 140out: 141 TCPDEBUG2(PRU_ATTACH); 142 splx(s); 143 return error; 144} 145 146/* 147 * pru_detach() detaches the TCP protocol from the socket. 148 * If the protocol state is non-embryonic, then can't 149 * do this directly: have to initiate a pru_disconnect(), 150 * which may finish later; embryonic TCB's can just 151 * be discarded here. 152 */ 153static int 154tcp_usr_detach(struct socket *so) 155{ 156 int s = splnet(); 157 int error = 0; 158 struct inpcb *inp = sotoinpcb(so); 159 struct tcpcb *tp; 160 TCPDEBUG0; 161 162 if (inp == 0) { 163 splx(s); 164 return EINVAL; /* XXX */ 165 } 166 tp = intotcpcb(inp); 167 TCPDEBUG1(); 168 tp = tcp_disconnect(tp); 169 170 TCPDEBUG2(PRU_DETACH); 171 splx(s); 172 return error; 173} 174 175#define COMMON_START() TCPDEBUG0; \ 176 do { \ 177 if (inp == 0) { \ 178 splx(s); \ 179 return EINVAL; \ 180 } \ 181 tp = intotcpcb(inp); \ 182 TCPDEBUG1(); \ 183 } while(0) 184 185#define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out 186 187 188/* 189 * Give the socket an address. 190 */ 191static int 192tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p) 193{ 194 int s = splnet(); 195 int error = 0; 196 struct inpcb *inp = sotoinpcb(so); 197 struct tcpcb *tp; 198 struct sockaddr_in *sinp; 199 200 COMMON_START(); 201 202 /* 203 * Must check for multicast addresses and disallow binding 204 * to them. 205 */ 206 sinp = (struct sockaddr_in *)nam; 207 if (sinp->sin_family == AF_INET && 208 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 209 error = EAFNOSUPPORT; 210 goto out; 211 } 212 error = in_pcbbind(inp, nam, p); 213 if (error) 214 goto out; 215 COMMON_END(PRU_BIND); 216 217} 218 219#ifdef INET6 220static int 221tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p) 222{ 223 int s = splnet(); 224 int error = 0; 225 struct inpcb *inp = sotoinpcb(so); 226 struct tcpcb *tp; 227 struct sockaddr_in6 *sin6p; 228 229 COMMON_START(); 230 231 /* 232 * Must check for multicast addresses and disallow binding 233 * to them. 234 */ 235 sin6p = (struct sockaddr_in6 *)nam; 236 if (sin6p->sin6_family == AF_INET6 && 237 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 238 error = EAFNOSUPPORT; 239 goto out; 240 } 241 inp->inp_vflag &= ~INP_IPV4; 242 inp->inp_vflag |= INP_IPV6; 243 if ((inp->inp_flags & IN6P_BINDV6ONLY) == 0) { 244 245 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) 246 inp->inp_vflag |= INP_IPV4; 247 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 248 struct sockaddr_in sin; 249 250 in6_sin6_2_sin(&sin, sin6p); 251 inp->inp_vflag |= INP_IPV4; 252 inp->inp_vflag &= ~INP_IPV6; 253 error = in_pcbbind(inp, (struct sockaddr *)&sin, p); 254 goto out; 255 } 256 } 257 error = in6_pcbbind(inp, nam, p); 258 if (error) 259 goto out; 260 COMMON_END(PRU_BIND); 261} 262#endif /* INET6 */ 263 264/* 265 * Prepare to accept connections. 266 */ 267static int 268tcp_usr_listen(struct socket *so, struct proc *p) 269{ 270 int s = splnet(); 271 int error = 0; 272 struct inpcb *inp = sotoinpcb(so); 273 struct tcpcb *tp; 274 275 COMMON_START(); 276 if (inp->inp_lport == 0) 277 error = in_pcbbind(inp, (struct sockaddr *)0, p); 278 if (error == 0) 279 tp->t_state = TCPS_LISTEN; 280 COMMON_END(PRU_LISTEN); 281} 282 283#ifdef INET6 284static int 285tcp6_usr_listen(struct socket *so, struct proc *p) 286{ 287 int s = splnet(); 288 int error = 0; 289 struct inpcb *inp = sotoinpcb(so); 290 struct tcpcb *tp; 291 292 COMMON_START(); 293 if (inp->inp_lport == 0) { 294 inp->inp_vflag &= ~INP_IPV4; 295 if ((inp->inp_flags & IN6P_BINDV6ONLY) == 0) 296 inp->inp_vflag |= INP_IPV4; 297 error = in6_pcbbind(inp, (struct sockaddr *)0, p); 298 } 299 if (error == 0) 300 tp->t_state = TCPS_LISTEN; 301 COMMON_END(PRU_LISTEN); 302} 303#endif /* INET6 */ 304 305/* 306 * Initiate connection to peer. 307 * Create a template for use in transmissions on this connection. 308 * Enter SYN_SENT state, and mark socket as connecting. 309 * Start keep-alive timer, and seed output sequence space. 310 * Send initial segment on connection. 311 */ 312static int 313tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p) 314{ 315 int s = splnet(); 316 int error = 0; 317 struct inpcb *inp = sotoinpcb(so); 318 struct tcpcb *tp; 319 struct sockaddr_in *sinp; 320 321 COMMON_START(); 322 323 /* 324 * Must disallow TCP ``connections'' to multicast addresses. 325 */ 326 sinp = (struct sockaddr_in *)nam; 327 if (sinp->sin_family == AF_INET 328 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 329 error = EAFNOSUPPORT; 330 goto out; 331 } 332 333 if (p && jailed(p->p_ucred)) 334 prison_remote_ip(p->p_ucred, 0, &sinp->sin_addr.s_addr); 335 336 if ((error = tcp_connect(tp, nam, p)) != 0) 337 goto out; 338 error = tcp_output(tp); 339 COMMON_END(PRU_CONNECT); 340} 341 342#ifdef INET6 343static int 344tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p) 345{ 346 int s = splnet(); 347 int error = 0; 348 struct inpcb *inp = sotoinpcb(so); 349 struct tcpcb *tp; 350 struct sockaddr_in6 *sin6p; 351 352 COMMON_START(); 353 354 /* 355 * Must disallow TCP ``connections'' to multicast addresses. 356 */ 357 sin6p = (struct sockaddr_in6 *)nam; 358 if (sin6p->sin6_family == AF_INET6 359 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 360 error = EAFNOSUPPORT; 361 goto out; 362 } 363 364 if ((inp->inp_flags & IN6P_BINDV6ONLY) == 0 && 365 IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 366 struct sockaddr_in sin; 367 368 in6_sin6_2_sin(&sin, sin6p); 369 inp->inp_vflag |= INP_IPV4; 370 inp->inp_vflag &= ~INP_IPV6; 371 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, p)) != 0) 372 goto out; 373 error = tcp_output(tp); 374 goto out; 375 } 376 inp->inp_vflag &= ~INP_IPV4; 377 inp->inp_vflag |= INP_IPV6; 378 if ((error = tcp6_connect(tp, nam, p)) != 0) 379 goto out; 380 error = tcp_output(tp); 381 COMMON_END(PRU_CONNECT); 382} 383#endif /* INET6 */ 384 385/* 386 * Initiate disconnect from peer. 387 * If connection never passed embryonic stage, just drop; 388 * else if don't need to let data drain, then can just drop anyways, 389 * else have to begin TCP shutdown process: mark socket disconnecting, 390 * drain unread data, state switch to reflect user close, and 391 * send segment (e.g. FIN) to peer. Socket will be really disconnected 392 * when peer sends FIN and acks ours. 393 * 394 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 395 */ 396static int 397tcp_usr_disconnect(struct socket *so) 398{ 399 int s = splnet(); 400 int error = 0; 401 struct inpcb *inp = sotoinpcb(so); 402 struct tcpcb *tp; 403 404 COMMON_START(); 405 tp = tcp_disconnect(tp); 406 COMMON_END(PRU_DISCONNECT); 407} 408 409/* 410 * Accept a connection. Essentially all the work is 411 * done at higher levels; just return the address 412 * of the peer, storing through addr. 413 */ 414static int 415tcp_usr_accept(struct socket *so, struct sockaddr **nam) 416{ 417 int s = splnet(); 418 int error = 0; 419 struct inpcb *inp = sotoinpcb(so); 420 struct tcpcb *tp = NULL; 421 TCPDEBUG0; 422 423 if (so->so_state & SS_ISDISCONNECTED) { 424 error = ECONNABORTED; 425 goto out; 426 } 427 if (inp == 0) { 428 splx(s); 429 return (EINVAL); 430 } 431 tp = intotcpcb(inp); 432 TCPDEBUG1(); 433 in_setpeeraddr(so, nam); 434 COMMON_END(PRU_ACCEPT); 435} 436 437#ifdef INET6 438static int 439tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 440{ 441 int s = splnet(); 442 int error = 0; 443 struct inpcb *inp = sotoinpcb(so); 444 struct tcpcb *tp = NULL; 445 TCPDEBUG0; 446 447 if (so->so_state & SS_ISDISCONNECTED) { 448 error = ECONNABORTED; 449 goto out; 450 } 451 if (inp == 0) { 452 splx(s); 453 return (EINVAL); 454 } 455 tp = intotcpcb(inp); 456 TCPDEBUG1(); 457 in6_mapped_peeraddr(so, nam); 458 COMMON_END(PRU_ACCEPT); 459} 460#endif /* INET6 */ 461/* 462 * Mark the connection as being incapable of further output. 463 */ 464static int 465tcp_usr_shutdown(struct socket *so) 466{ 467 int s = splnet(); 468 int error = 0; 469 struct inpcb *inp = sotoinpcb(so); 470 struct tcpcb *tp; 471 472 COMMON_START(); 473 socantsendmore(so); 474 tp = tcp_usrclosed(tp); 475 if (tp) 476 error = tcp_output(tp); 477 COMMON_END(PRU_SHUTDOWN); 478} 479 480/* 481 * After a receive, possibly send window update to peer. 482 */ 483static int 484tcp_usr_rcvd(struct socket *so, int flags) 485{ 486 int s = splnet(); 487 int error = 0; 488 struct inpcb *inp = sotoinpcb(so); 489 struct tcpcb *tp; 490 491 COMMON_START(); 492 tcp_output(tp); 493 COMMON_END(PRU_RCVD); 494} 495 496/* 497 * Do a send by putting data in output queue and updating urgent 498 * marker if URG set. Possibly send more data. Unlike the other 499 * pru_*() routines, the mbuf chains are our responsibility. We 500 * must either enqueue them or free them. The other pru_* routines 501 * generally are caller-frees. 502 */ 503static int 504tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 505 struct sockaddr *nam, struct mbuf *control, struct proc *p) 506{ 507 int s = splnet(); 508 int error = 0; 509 struct inpcb *inp = sotoinpcb(so); 510 struct tcpcb *tp; 511#ifdef INET6 512 int isipv6; 513#endif 514 TCPDEBUG0; 515 516 if (inp == NULL) { 517 /* 518 * OOPS! we lost a race, the TCP session got reset after 519 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 520 * network interrupt in the non-splnet() section of sosend(). 521 */ 522 if (m) 523 m_freem(m); 524 if (control) 525 m_freem(control); 526 error = ECONNRESET; /* XXX EPIPE? */ 527 tp = NULL; 528 TCPDEBUG1(); 529 goto out; 530 } 531#ifdef INET6 532 isipv6 = nam && nam->sa_family == AF_INET6; 533#endif /* INET6 */ 534 tp = intotcpcb(inp); 535 TCPDEBUG1(); 536 if (control) { 537 /* TCP doesn't do control messages (rights, creds, etc) */ 538 if (control->m_len) { 539 m_freem(control); 540 if (m) 541 m_freem(m); 542 error = EINVAL; 543 goto out; 544 } 545 m_freem(control); /* empty control, just free it */ 546 } 547 if(!(flags & PRUS_OOB)) { 548 sbappend(&so->so_snd, m); 549 if (nam && tp->t_state < TCPS_SYN_SENT) { 550 /* 551 * Do implied connect if not yet connected, 552 * initialize window to default value, and 553 * initialize maxseg/maxopd using peer's cached 554 * MSS. 555 */ 556#ifdef INET6 557 if (isipv6) 558 error = tcp6_connect(tp, nam, p); 559 else 560#endif /* INET6 */ 561 error = tcp_connect(tp, nam, p); 562 if (error) 563 goto out; 564 tp->snd_wnd = TTCP_CLIENT_SND_WND; 565 tcp_mss(tp, -1); 566 } 567 568 if (flags & PRUS_EOF) { 569 /* 570 * Close the send side of the connection after 571 * the data is sent. 572 */ 573 socantsendmore(so); 574 tp = tcp_usrclosed(tp); 575 } 576 if (tp != NULL) { 577 if (flags & PRUS_MORETOCOME) 578 tp->t_flags |= TF_MORETOCOME; 579 error = tcp_output(tp); 580 if (flags & PRUS_MORETOCOME) 581 tp->t_flags &= ~TF_MORETOCOME; 582 } 583 } else { 584 if (sbspace(&so->so_snd) < -512) { 585 m_freem(m); 586 error = ENOBUFS; 587 goto out; 588 } 589 /* 590 * According to RFC961 (Assigned Protocols), 591 * the urgent pointer points to the last octet 592 * of urgent data. We continue, however, 593 * to consider it to indicate the first octet 594 * of data past the urgent section. 595 * Otherwise, snd_up should be one lower. 596 */ 597 sbappend(&so->so_snd, m); 598 if (nam && tp->t_state < TCPS_SYN_SENT) { 599 /* 600 * Do implied connect if not yet connected, 601 * initialize window to default value, and 602 * initialize maxseg/maxopd using peer's cached 603 * MSS. 604 */ 605#ifdef INET6 606 if (isipv6) 607 error = tcp6_connect(tp, nam, p); 608 else 609#endif /* INET6 */ 610 error = tcp_connect(tp, nam, p); 611 if (error) 612 goto out; 613 tp->snd_wnd = TTCP_CLIENT_SND_WND; 614 tcp_mss(tp, -1); 615 } 616 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 617 tp->t_force = 1; 618 error = tcp_output(tp); 619 tp->t_force = 0; 620 } 621 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 622 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 623} 624 625/* 626 * Abort the TCP. 627 */ 628static int 629tcp_usr_abort(struct socket *so) 630{ 631 int s = splnet(); 632 int error = 0; 633 struct inpcb *inp = sotoinpcb(so); 634 struct tcpcb *tp; 635 636 COMMON_START(); 637 tp = tcp_drop(tp, ECONNABORTED); 638 COMMON_END(PRU_ABORT); 639} 640 641/* 642 * Receive out-of-band data. 643 */ 644static int 645tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 646{ 647 int s = splnet(); 648 int error = 0; 649 struct inpcb *inp = sotoinpcb(so); 650 struct tcpcb *tp; 651 652 COMMON_START(); 653 if ((so->so_oobmark == 0 && 654 (so->so_state & SS_RCVATMARK) == 0) || 655 so->so_options & SO_OOBINLINE || 656 tp->t_oobflags & TCPOOB_HADDATA) { 657 error = EINVAL; 658 goto out; 659 } 660 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 661 error = EWOULDBLOCK; 662 goto out; 663 } 664 m->m_len = 1; 665 *mtod(m, caddr_t) = tp->t_iobc; 666 if ((flags & MSG_PEEK) == 0) 667 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 668 COMMON_END(PRU_RCVOOB); 669} 670 671/* xxx - should be const */ 672struct pr_usrreqs tcp_usrreqs = { 673 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 674 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, 675 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, 676 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 677 in_setsockaddr, sosend, soreceive, sopoll 678}; 679 680#ifdef INET6 681struct pr_usrreqs tcp6_usrreqs = { 682 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, 683 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, 684 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, 685 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 686 in6_mapped_sockaddr, sosend, soreceive, sopoll 687}; 688#endif /* INET6 */ 689 690/* 691 * Common subroutine to open a TCP connection to remote host specified 692 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 693 * port number if needed. Call in_pcbladdr to do the routing and to choose 694 * a local host address (interface). If there is an existing incarnation 695 * of the same connection in TIME-WAIT state and if the remote host was 696 * sending CC options and if the connection duration was < MSL, then 697 * truncate the previous TIME-WAIT state and proceed. 698 * Initialize connection parameters and enter SYN-SENT state. 699 */ 700static int 701tcp_connect(tp, nam, p) 702 register struct tcpcb *tp; 703 struct sockaddr *nam; 704 struct proc *p; 705{ 706 struct inpcb *inp = tp->t_inpcb, *oinp; 707 struct socket *so = inp->inp_socket; 708 struct tcpcb *otp; 709 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 710 struct sockaddr_in *ifaddr; 711 struct rmxp_tao *taop; 712 struct rmxp_tao tao_noncached; 713 int error; 714 715 if (inp->inp_lport == 0) { 716 error = in_pcbbind(inp, (struct sockaddr *)0, p); 717 if (error) 718 return error; 719 } 720 721 /* 722 * Cannot simply call in_pcbconnect, because there might be an 723 * earlier incarnation of this same connection still in 724 * TIME_WAIT state, creating an ADDRINUSE error. 725 */ 726 error = in_pcbladdr(inp, nam, &ifaddr); 727 if (error) 728 return error; 729 oinp = in_pcblookup_hash(inp->inp_pcbinfo, 730 sin->sin_addr, sin->sin_port, 731 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr 732 : ifaddr->sin_addr, 733 inp->inp_lport, 0, NULL); 734 if (oinp) { 735 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 736 otp->t_state == TCPS_TIME_WAIT && 737 (ticks - otp->t_starttime) < tcp_msl && 738 (otp->t_flags & TF_RCVD_CC)) 739 otp = tcp_close(otp); 740 else 741 return EADDRINUSE; 742 } 743 if (inp->inp_laddr.s_addr == INADDR_ANY) 744 inp->inp_laddr = ifaddr->sin_addr; 745 inp->inp_faddr = sin->sin_addr; 746 inp->inp_fport = sin->sin_port; 747 in_pcbrehash(inp); 748 749 tp->t_template = tcp_template(tp); 750 if (tp->t_template == 0) { 751 in_pcbdisconnect(inp); 752 return ENOBUFS; 753 } 754 755 /* Compute window scaling to request. */ 756 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 757 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 758 tp->request_r_scale++; 759 760 soisconnecting(so); 761 tcpstat.tcps_connattempt++; 762 tp->t_state = TCPS_SYN_SENT; 763 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 764 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; 765 tcp_sendseqinit(tp); 766 767 /* 768 * Generate a CC value for this connection and 769 * check whether CC or CCnew should be used. 770 */ 771 if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { 772 taop = &tao_noncached; 773 bzero(taop, sizeof(*taop)); 774 } 775 776 tp->cc_send = CC_INC(tcp_ccgen); 777 if (taop->tao_ccsent != 0 && 778 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 779 taop->tao_ccsent = tp->cc_send; 780 } else { 781 taop->tao_ccsent = 0; 782 tp->t_flags |= TF_SENDCCNEW; 783 } 784 785 return 0; 786} 787 788#ifdef INET6 789static int 790tcp6_connect(tp, nam, p) 791 register struct tcpcb *tp; 792 struct sockaddr *nam; 793 struct proc *p; 794{ 795 struct inpcb *inp = tp->t_inpcb, *oinp; 796 struct socket *so = inp->inp_socket; 797 struct tcpcb *otp; 798 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 799 struct in6_addr *addr6; 800 struct rmxp_tao *taop; 801 struct rmxp_tao tao_noncached; 802 int error; 803 804 if (inp->inp_lport == 0) { 805 error = in6_pcbbind(inp, (struct sockaddr *)0, p); 806 if (error) 807 return error; 808 } 809 810 /* 811 * Cannot simply call in_pcbconnect, because there might be an 812 * earlier incarnation of this same connection still in 813 * TIME_WAIT state, creating an ADDRINUSE error. 814 */ 815 error = in6_pcbladdr(inp, nam, &addr6); 816 if (error) 817 return error; 818 oinp = in6_pcblookup_hash(inp->inp_pcbinfo, 819 &sin6->sin6_addr, sin6->sin6_port, 820 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) 821 ? addr6 822 : &inp->in6p_laddr, 823 inp->inp_lport, 0, NULL); 824 if (oinp) { 825 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 826 otp->t_state == TCPS_TIME_WAIT && 827 (ticks - otp->t_starttime) < tcp_msl && 828 (otp->t_flags & TF_RCVD_CC)) 829 otp = tcp_close(otp); 830 else 831 return EADDRINUSE; 832 } 833 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 834 inp->in6p_laddr = *addr6; 835 inp->in6p_faddr = sin6->sin6_addr; 836 inp->inp_fport = sin6->sin6_port; 837 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) 838 inp->in6p_flowinfo = sin6->sin6_flowinfo; 839 in_pcbrehash(inp); 840 841 tp->t_template = tcp_template(tp); 842 if (tp->t_template == 0) { 843 in6_pcbdisconnect(inp); 844 return ENOBUFS; 845 } 846 847 /* Compute window scaling to request. */ 848 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 849 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 850 tp->request_r_scale++; 851 852 soisconnecting(so); 853 tcpstat.tcps_connattempt++; 854 tp->t_state = TCPS_SYN_SENT; 855 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 856 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; 857 tcp_sendseqinit(tp); 858 859 /* 860 * Generate a CC value for this connection and 861 * check whether CC or CCnew should be used. 862 */ 863 if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { 864 taop = &tao_noncached; 865 bzero(taop, sizeof(*taop)); 866 } 867 868 tp->cc_send = CC_INC(tcp_ccgen); 869 if (taop->tao_ccsent != 0 && 870 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 871 taop->tao_ccsent = tp->cc_send; 872 } else { 873 taop->tao_ccsent = 0; 874 tp->t_flags |= TF_SENDCCNEW; 875 } 876 877 return 0; 878} 879#endif /* INET6 */ 880 881/* 882 * The new sockopt interface makes it possible for us to block in the 883 * copyin/out step (if we take a page fault). Taking a page fault at 884 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now 885 * use TSM, there probably isn't any need for this function to run at 886 * splnet() any more. This needs more examination.) 887 */ 888int 889tcp_ctloutput(so, sopt) 890 struct socket *so; 891 struct sockopt *sopt; 892{ 893 int error, opt, optval, s; 894 struct inpcb *inp; 895 struct tcpcb *tp; 896 897 error = 0; 898 s = splnet(); /* XXX */ 899 inp = sotoinpcb(so); 900 if (inp == NULL) { 901 splx(s); 902 return (ECONNRESET); 903 } 904 if (sopt->sopt_level != IPPROTO_TCP) { 905#ifdef INET6 906 if (INP_CHECK_SOCKAF(so, AF_INET6)) 907 error = ip6_ctloutput(so, sopt); 908 else 909#endif /* INET6 */ 910 error = ip_ctloutput(so, sopt); 911 splx(s); 912 return (error); 913 } 914 tp = intotcpcb(inp); 915 916 switch (sopt->sopt_dir) { 917 case SOPT_SET: 918 switch (sopt->sopt_name) { 919 case TCP_NODELAY: 920 case TCP_NOOPT: 921 error = sooptcopyin(sopt, &optval, sizeof optval, 922 sizeof optval); 923 if (error) 924 break; 925 926 switch (sopt->sopt_name) { 927 case TCP_NODELAY: 928 opt = TF_NODELAY; 929 break; 930 case TCP_NOOPT: 931 opt = TF_NOOPT; 932 break; 933 default: 934 opt = 0; /* dead code to fool gcc */ 935 break; 936 } 937 938 if (optval) 939 tp->t_flags |= opt; 940 else 941 tp->t_flags &= ~opt; 942 break; 943 944 case TCP_NOPUSH: 945 error = sooptcopyin(sopt, &optval, sizeof optval, 946 sizeof optval); 947 if (error) 948 break; 949 950 if (optval) 951 tp->t_flags |= TF_NOPUSH; 952 else { 953 tp->t_flags &= ~TF_NOPUSH; 954 error = tcp_output(tp); 955 } 956 break; 957 958 case TCP_MAXSEG: 959 error = sooptcopyin(sopt, &optval, sizeof optval, 960 sizeof optval); 961 if (error) 962 break; 963 964 if (optval > 0 && optval <= tp->t_maxseg) 965 tp->t_maxseg = optval; 966 else 967 error = EINVAL; 968 break; 969 970 default: 971 error = ENOPROTOOPT; 972 break; 973 } 974 break; 975 976 case SOPT_GET: 977 switch (sopt->sopt_name) { 978 case TCP_NODELAY: 979 optval = tp->t_flags & TF_NODELAY; 980 break; 981 case TCP_MAXSEG: 982 optval = tp->t_maxseg; 983 break; 984 case TCP_NOOPT: 985 optval = tp->t_flags & TF_NOOPT; 986 break; 987 case TCP_NOPUSH: 988 optval = tp->t_flags & TF_NOPUSH; 989 break; 990 default: 991 error = ENOPROTOOPT; 992 break; 993 } 994 if (error == 0) 995 error = sooptcopyout(sopt, &optval, sizeof optval); 996 break; 997 } 998 splx(s); 999 return (error); 1000} 1001 1002/* 1003 * tcp_sendspace and tcp_recvspace are the default send and receive window 1004 * sizes, respectively. These are obsolescent (this information should 1005 * be set by the route). 1006 */ 1007u_long tcp_sendspace = 1024*16; 1008SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1009 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1010u_long tcp_recvspace = 1024*16; 1011SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1012 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1013 1014/* 1015 * Attach TCP protocol to socket, allocating 1016 * internet protocol control block, tcp control block, 1017 * bufer space, and entering LISTEN state if to accept connections. 1018 */ 1019static int 1020tcp_attach(so, p) 1021 struct socket *so; 1022 struct proc *p; 1023{ 1024 register struct tcpcb *tp; 1025 struct inpcb *inp; 1026 int error; 1027#ifdef INET6 1028 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; 1029#endif 1030 1031 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1032 error = soreserve(so, tcp_sendspace, tcp_recvspace); 1033 if (error) 1034 return (error); 1035 } 1036 error = in_pcballoc(so, &tcbinfo, p); 1037 if (error) 1038 return (error); 1039 inp = sotoinpcb(so); 1040#ifdef IPSEC 1041 error = ipsec_init_policy(so, &inp->inp_sp); 1042 if (error) { 1043#ifdef INET6 1044 if (isipv6) 1045 in6_pcbdetach(inp); 1046 else 1047#endif 1048 in_pcbdetach(inp); 1049 return (error); 1050 } 1051#endif /*IPSEC*/ 1052#ifdef INET6 1053 if (isipv6) { 1054 inp->inp_vflag |= INP_IPV6; 1055 inp->in6p_hops = -1; /* use kernel default */ 1056 } 1057 else 1058#endif 1059 inp->inp_vflag |= INP_IPV4; 1060 tp = tcp_newtcpcb(inp); 1061 if (tp == 0) { 1062 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1063 1064 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1065#ifdef INET6 1066 if (isipv6) 1067 in6_pcbdetach(inp); 1068 else 1069#endif 1070 in_pcbdetach(inp); 1071 so->so_state |= nofd; 1072 return (ENOBUFS); 1073 } 1074 tp->t_state = TCPS_CLOSED; 1075 return (0); 1076} 1077 1078/* 1079 * Initiate (or continue) disconnect. 1080 * If embryonic state, just send reset (once). 1081 * If in ``let data drain'' option and linger null, just drop. 1082 * Otherwise (hard), mark socket disconnecting and drop 1083 * current input data; switch states based on user close, and 1084 * send segment to peer (with FIN). 1085 */ 1086static struct tcpcb * 1087tcp_disconnect(tp) 1088 register struct tcpcb *tp; 1089{ 1090 struct socket *so = tp->t_inpcb->inp_socket; 1091 1092 if (tp->t_state < TCPS_ESTABLISHED) 1093 tp = tcp_close(tp); 1094 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1095 tp = tcp_drop(tp, 0); 1096 else { 1097 soisdisconnecting(so); 1098 sbflush(&so->so_rcv); 1099 tp = tcp_usrclosed(tp); 1100 if (tp) 1101 (void) tcp_output(tp); 1102 } 1103 return (tp); 1104} 1105 1106/* 1107 * User issued close, and wish to trail through shutdown states: 1108 * if never received SYN, just forget it. If got a SYN from peer, 1109 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1110 * If already got a FIN from peer, then almost done; go to LAST_ACK 1111 * state. In all other cases, have already sent FIN to peer (e.g. 1112 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1113 * for peer to send FIN or not respond to keep-alives, etc. 1114 * We can let the user exit from the close as soon as the FIN is acked. 1115 */ 1116static struct tcpcb * 1117tcp_usrclosed(tp) 1118 register struct tcpcb *tp; 1119{ 1120 1121 switch (tp->t_state) { 1122 1123 case TCPS_CLOSED: 1124 case TCPS_LISTEN: 1125 tp->t_state = TCPS_CLOSED; 1126 tp = tcp_close(tp); 1127 break; 1128 1129 case TCPS_SYN_SENT: 1130 case TCPS_SYN_RECEIVED: 1131 tp->t_flags |= TF_NEEDFIN; 1132 break; 1133 1134 case TCPS_ESTABLISHED: 1135 tp->t_state = TCPS_FIN_WAIT_1; 1136 break; 1137 1138 case TCPS_CLOSE_WAIT: 1139 tp->t_state = TCPS_LAST_ACK; 1140 break; 1141 } 1142 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1143 soisdisconnected(tp->t_inpcb->inp_socket); 1144 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1145 if (tp->t_state == TCPS_FIN_WAIT_2) 1146 callout_reset(tp->tt_2msl, tcp_maxidle, 1147 tcp_timer_2msl, tp); 1148 } 1149 return (tp); 1150} 1151 1152