tcp_usrreq.c revision 74018
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 74018 2001-03-09 08:16:40Z 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 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; 421 422 if (so->so_state & SS_ISDISCONNECTED) { 423 error = ECONNABORTED; 424 goto out; 425 } 426 COMMON_START(); 427 in_setpeeraddr(so, nam); 428 COMMON_END(PRU_ACCEPT); 429} 430 431#ifdef INET6 432static int 433tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 434{ 435 int s = splnet(); 436 int error = 0; 437 struct inpcb *inp = sotoinpcb(so); 438 struct tcpcb *tp; 439 440 if (so->so_state & SS_ISDISCONNECTED) { 441 error = ECONNABORTED; 442 goto out; 443 } 444 COMMON_START(); 445 in6_mapped_peeraddr(so, nam); 446 COMMON_END(PRU_ACCEPT); 447} 448#endif /* INET6 */ 449/* 450 * Mark the connection as being incapable of further output. 451 */ 452static int 453tcp_usr_shutdown(struct socket *so) 454{ 455 int s = splnet(); 456 int error = 0; 457 struct inpcb *inp = sotoinpcb(so); 458 struct tcpcb *tp; 459 460 COMMON_START(); 461 socantsendmore(so); 462 tp = tcp_usrclosed(tp); 463 if (tp) 464 error = tcp_output(tp); 465 COMMON_END(PRU_SHUTDOWN); 466} 467 468/* 469 * After a receive, possibly send window update to peer. 470 */ 471static int 472tcp_usr_rcvd(struct socket *so, int flags) 473{ 474 int s = splnet(); 475 int error = 0; 476 struct inpcb *inp = sotoinpcb(so); 477 struct tcpcb *tp; 478 479 COMMON_START(); 480 tcp_output(tp); 481 COMMON_END(PRU_RCVD); 482} 483 484/* 485 * Do a send by putting data in output queue and updating urgent 486 * marker if URG set. Possibly send more data. Unlike the other 487 * pru_*() routines, the mbuf chains are our responsibility. We 488 * must either enqueue them or free them. The other pru_* routines 489 * generally are caller-frees. 490 */ 491static int 492tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 493 struct sockaddr *nam, struct mbuf *control, struct proc *p) 494{ 495 int s = splnet(); 496 int error = 0; 497 struct inpcb *inp = sotoinpcb(so); 498 struct tcpcb *tp; 499#ifdef INET6 500 int isipv6; 501#endif 502 TCPDEBUG0; 503 504 if (inp == NULL) { 505 /* 506 * OOPS! we lost a race, the TCP session got reset after 507 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 508 * network interrupt in the non-splnet() section of sosend(). 509 */ 510 if (m) 511 m_freem(m); 512 if (control) 513 m_freem(control); 514 error = ECONNRESET; /* XXX EPIPE? */ 515 tp = NULL; 516 TCPDEBUG1(); 517 goto out; 518 } 519#ifdef INET6 520 isipv6 = nam && nam->sa_family == AF_INET6; 521#endif /* INET6 */ 522 tp = intotcpcb(inp); 523 TCPDEBUG1(); 524 if (control) { 525 /* TCP doesn't do control messages (rights, creds, etc) */ 526 if (control->m_len) { 527 m_freem(control); 528 if (m) 529 m_freem(m); 530 error = EINVAL; 531 goto out; 532 } 533 m_freem(control); /* empty control, just free it */ 534 } 535 if(!(flags & PRUS_OOB)) { 536 sbappend(&so->so_snd, m); 537 if (nam && tp->t_state < TCPS_SYN_SENT) { 538 /* 539 * Do implied connect if not yet connected, 540 * initialize window to default value, and 541 * initialize maxseg/maxopd using peer's cached 542 * MSS. 543 */ 544#ifdef INET6 545 if (isipv6) 546 error = tcp6_connect(tp, nam, p); 547 else 548#endif /* INET6 */ 549 error = tcp_connect(tp, nam, p); 550 if (error) 551 goto out; 552 tp->snd_wnd = TTCP_CLIENT_SND_WND; 553 tcp_mss(tp, -1); 554 } 555 556 if (flags & PRUS_EOF) { 557 /* 558 * Close the send side of the connection after 559 * the data is sent. 560 */ 561 socantsendmore(so); 562 tp = tcp_usrclosed(tp); 563 } 564 if (tp != NULL) { 565 if (flags & PRUS_MORETOCOME) 566 tp->t_flags |= TF_MORETOCOME; 567 error = tcp_output(tp); 568 if (flags & PRUS_MORETOCOME) 569 tp->t_flags &= ~TF_MORETOCOME; 570 } 571 } else { 572 if (sbspace(&so->so_snd) < -512) { 573 m_freem(m); 574 error = ENOBUFS; 575 goto out; 576 } 577 /* 578 * According to RFC961 (Assigned Protocols), 579 * the urgent pointer points to the last octet 580 * of urgent data. We continue, however, 581 * to consider it to indicate the first octet 582 * of data past the urgent section. 583 * Otherwise, snd_up should be one lower. 584 */ 585 sbappend(&so->so_snd, m); 586 if (nam && tp->t_state < TCPS_SYN_SENT) { 587 /* 588 * Do implied connect if not yet connected, 589 * initialize window to default value, and 590 * initialize maxseg/maxopd using peer's cached 591 * MSS. 592 */ 593#ifdef INET6 594 if (isipv6) 595 error = tcp6_connect(tp, nam, p); 596 else 597#endif /* INET6 */ 598 error = tcp_connect(tp, nam, p); 599 if (error) 600 goto out; 601 tp->snd_wnd = TTCP_CLIENT_SND_WND; 602 tcp_mss(tp, -1); 603 } 604 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 605 tp->t_force = 1; 606 error = tcp_output(tp); 607 tp->t_force = 0; 608 } 609 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 610 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 611} 612 613/* 614 * Abort the TCP. 615 */ 616static int 617tcp_usr_abort(struct socket *so) 618{ 619 int s = splnet(); 620 int error = 0; 621 struct inpcb *inp = sotoinpcb(so); 622 struct tcpcb *tp; 623 624 COMMON_START(); 625 tp = tcp_drop(tp, ECONNABORTED); 626 COMMON_END(PRU_ABORT); 627} 628 629/* 630 * Receive out-of-band data. 631 */ 632static int 633tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 634{ 635 int s = splnet(); 636 int error = 0; 637 struct inpcb *inp = sotoinpcb(so); 638 struct tcpcb *tp; 639 640 COMMON_START(); 641 if ((so->so_oobmark == 0 && 642 (so->so_state & SS_RCVATMARK) == 0) || 643 so->so_options & SO_OOBINLINE || 644 tp->t_oobflags & TCPOOB_HADDATA) { 645 error = EINVAL; 646 goto out; 647 } 648 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 649 error = EWOULDBLOCK; 650 goto out; 651 } 652 m->m_len = 1; 653 *mtod(m, caddr_t) = tp->t_iobc; 654 if ((flags & MSG_PEEK) == 0) 655 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 656 COMMON_END(PRU_RCVOOB); 657} 658 659/* xxx - should be const */ 660struct pr_usrreqs tcp_usrreqs = { 661 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 662 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, 663 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, 664 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 665 in_setsockaddr, sosend, soreceive, sopoll 666}; 667 668#ifdef INET6 669struct pr_usrreqs tcp6_usrreqs = { 670 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, 671 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, 672 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, 673 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 674 in6_mapped_sockaddr, sosend, soreceive, sopoll 675}; 676#endif /* INET6 */ 677 678/* 679 * Common subroutine to open a TCP connection to remote host specified 680 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 681 * port number if needed. Call in_pcbladdr to do the routing and to choose 682 * a local host address (interface). If there is an existing incarnation 683 * of the same connection in TIME-WAIT state and if the remote host was 684 * sending CC options and if the connection duration was < MSL, then 685 * truncate the previous TIME-WAIT state and proceed. 686 * Initialize connection parameters and enter SYN-SENT state. 687 */ 688static int 689tcp_connect(tp, nam, p) 690 register struct tcpcb *tp; 691 struct sockaddr *nam; 692 struct proc *p; 693{ 694 struct inpcb *inp = tp->t_inpcb, *oinp; 695 struct socket *so = inp->inp_socket; 696 struct tcpcb *otp; 697 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 698 struct sockaddr_in *ifaddr; 699 struct rmxp_tao *taop; 700 struct rmxp_tao tao_noncached; 701 int error; 702 703 if (inp->inp_lport == 0) { 704 error = in_pcbbind(inp, (struct sockaddr *)0, p); 705 if (error) 706 return error; 707 } 708 709 /* 710 * Cannot simply call in_pcbconnect, because there might be an 711 * earlier incarnation of this same connection still in 712 * TIME_WAIT state, creating an ADDRINUSE error. 713 */ 714 error = in_pcbladdr(inp, nam, &ifaddr); 715 if (error) 716 return error; 717 oinp = in_pcblookup_hash(inp->inp_pcbinfo, 718 sin->sin_addr, sin->sin_port, 719 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr 720 : ifaddr->sin_addr, 721 inp->inp_lport, 0, NULL); 722 if (oinp) { 723 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 724 otp->t_state == TCPS_TIME_WAIT && 725 (ticks - otp->t_starttime) < tcp_msl && 726 (otp->t_flags & TF_RCVD_CC)) 727 otp = tcp_close(otp); 728 else 729 return EADDRINUSE; 730 } 731 if (inp->inp_laddr.s_addr == INADDR_ANY) 732 inp->inp_laddr = ifaddr->sin_addr; 733 inp->inp_faddr = sin->sin_addr; 734 inp->inp_fport = sin->sin_port; 735 in_pcbrehash(inp); 736 737 tp->t_template = tcp_template(tp); 738 if (tp->t_template == 0) { 739 in_pcbdisconnect(inp); 740 return ENOBUFS; 741 } 742 743 /* Compute window scaling to request. */ 744 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 745 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 746 tp->request_r_scale++; 747 748 soisconnecting(so); 749 tcpstat.tcps_connattempt++; 750 tp->t_state = TCPS_SYN_SENT; 751 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 752 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; 753 tcp_sendseqinit(tp); 754 755 /* 756 * Generate a CC value for this connection and 757 * check whether CC or CCnew should be used. 758 */ 759 if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { 760 taop = &tao_noncached; 761 bzero(taop, sizeof(*taop)); 762 } 763 764 tp->cc_send = CC_INC(tcp_ccgen); 765 if (taop->tao_ccsent != 0 && 766 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 767 taop->tao_ccsent = tp->cc_send; 768 } else { 769 taop->tao_ccsent = 0; 770 tp->t_flags |= TF_SENDCCNEW; 771 } 772 773 return 0; 774} 775 776#ifdef INET6 777static int 778tcp6_connect(tp, nam, p) 779 register struct tcpcb *tp; 780 struct sockaddr *nam; 781 struct proc *p; 782{ 783 struct inpcb *inp = tp->t_inpcb, *oinp; 784 struct socket *so = inp->inp_socket; 785 struct tcpcb *otp; 786 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 787 struct in6_addr *addr6; 788 struct rmxp_tao *taop; 789 struct rmxp_tao tao_noncached; 790 int error; 791 792 if (inp->inp_lport == 0) { 793 error = in6_pcbbind(inp, (struct sockaddr *)0, p); 794 if (error) 795 return error; 796 } 797 798 /* 799 * Cannot simply call in_pcbconnect, because there might be an 800 * earlier incarnation of this same connection still in 801 * TIME_WAIT state, creating an ADDRINUSE error. 802 */ 803 error = in6_pcbladdr(inp, nam, &addr6); 804 if (error) 805 return error; 806 oinp = in6_pcblookup_hash(inp->inp_pcbinfo, 807 &sin6->sin6_addr, sin6->sin6_port, 808 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) 809 ? addr6 810 : &inp->in6p_laddr, 811 inp->inp_lport, 0, NULL); 812 if (oinp) { 813 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 814 otp->t_state == TCPS_TIME_WAIT && 815 (ticks - otp->t_starttime) < tcp_msl && 816 (otp->t_flags & TF_RCVD_CC)) 817 otp = tcp_close(otp); 818 else 819 return EADDRINUSE; 820 } 821 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 822 inp->in6p_laddr = *addr6; 823 inp->in6p_faddr = sin6->sin6_addr; 824 inp->inp_fport = sin6->sin6_port; 825 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) 826 inp->in6p_flowinfo = sin6->sin6_flowinfo; 827 in_pcbrehash(inp); 828 829 tp->t_template = tcp_template(tp); 830 if (tp->t_template == 0) { 831 in6_pcbdisconnect(inp); 832 return ENOBUFS; 833 } 834 835 /* Compute window scaling to request. */ 836 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 837 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 838 tp->request_r_scale++; 839 840 soisconnecting(so); 841 tcpstat.tcps_connattempt++; 842 tp->t_state = TCPS_SYN_SENT; 843 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 844 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; 845 tcp_sendseqinit(tp); 846 847 /* 848 * Generate a CC value for this connection and 849 * check whether CC or CCnew should be used. 850 */ 851 if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) { 852 taop = &tao_noncached; 853 bzero(taop, sizeof(*taop)); 854 } 855 856 tp->cc_send = CC_INC(tcp_ccgen); 857 if (taop->tao_ccsent != 0 && 858 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 859 taop->tao_ccsent = tp->cc_send; 860 } else { 861 taop->tao_ccsent = 0; 862 tp->t_flags |= TF_SENDCCNEW; 863 } 864 865 return 0; 866} 867#endif /* INET6 */ 868 869/* 870 * The new sockopt interface makes it possible for us to block in the 871 * copyin/out step (if we take a page fault). Taking a page fault at 872 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now 873 * use TSM, there probably isn't any need for this function to run at 874 * splnet() any more. This needs more examination.) 875 */ 876int 877tcp_ctloutput(so, sopt) 878 struct socket *so; 879 struct sockopt *sopt; 880{ 881 int error, opt, optval, s; 882 struct inpcb *inp; 883 struct tcpcb *tp; 884 885 error = 0; 886 s = splnet(); /* XXX */ 887 inp = sotoinpcb(so); 888 if (inp == NULL) { 889 splx(s); 890 return (ECONNRESET); 891 } 892 if (sopt->sopt_level != IPPROTO_TCP) { 893#ifdef INET6 894 if (INP_CHECK_SOCKAF(so, AF_INET6)) 895 error = ip6_ctloutput(so, sopt); 896 else 897#endif /* INET6 */ 898 error = ip_ctloutput(so, sopt); 899 splx(s); 900 return (error); 901 } 902 tp = intotcpcb(inp); 903 904 switch (sopt->sopt_dir) { 905 case SOPT_SET: 906 switch (sopt->sopt_name) { 907 case TCP_NODELAY: 908 case TCP_NOOPT: 909 error = sooptcopyin(sopt, &optval, sizeof optval, 910 sizeof optval); 911 if (error) 912 break; 913 914 switch (sopt->sopt_name) { 915 case TCP_NODELAY: 916 opt = TF_NODELAY; 917 break; 918 case TCP_NOOPT: 919 opt = TF_NOOPT; 920 break; 921 default: 922 opt = 0; /* dead code to fool gcc */ 923 break; 924 } 925 926 if (optval) 927 tp->t_flags |= opt; 928 else 929 tp->t_flags &= ~opt; 930 break; 931 932 case TCP_NOPUSH: 933 error = sooptcopyin(sopt, &optval, sizeof optval, 934 sizeof optval); 935 if (error) 936 break; 937 938 if (optval) 939 tp->t_flags |= TF_NOPUSH; 940 else { 941 tp->t_flags &= ~TF_NOPUSH; 942 error = tcp_output(tp); 943 } 944 break; 945 946 case TCP_MAXSEG: 947 error = sooptcopyin(sopt, &optval, sizeof optval, 948 sizeof optval); 949 if (error) 950 break; 951 952 if (optval > 0 && optval <= tp->t_maxseg) 953 tp->t_maxseg = optval; 954 else 955 error = EINVAL; 956 break; 957 958 default: 959 error = ENOPROTOOPT; 960 break; 961 } 962 break; 963 964 case SOPT_GET: 965 switch (sopt->sopt_name) { 966 case TCP_NODELAY: 967 optval = tp->t_flags & TF_NODELAY; 968 break; 969 case TCP_MAXSEG: 970 optval = tp->t_maxseg; 971 break; 972 case TCP_NOOPT: 973 optval = tp->t_flags & TF_NOOPT; 974 break; 975 case TCP_NOPUSH: 976 optval = tp->t_flags & TF_NOPUSH; 977 break; 978 default: 979 error = ENOPROTOOPT; 980 break; 981 } 982 if (error == 0) 983 error = sooptcopyout(sopt, &optval, sizeof optval); 984 break; 985 } 986 splx(s); 987 return (error); 988} 989 990/* 991 * tcp_sendspace and tcp_recvspace are the default send and receive window 992 * sizes, respectively. These are obsolescent (this information should 993 * be set by the route). 994 */ 995u_long tcp_sendspace = 1024*16; 996SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 997 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 998u_long tcp_recvspace = 1024*16; 999SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1000 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1001 1002/* 1003 * Attach TCP protocol to socket, allocating 1004 * internet protocol control block, tcp control block, 1005 * bufer space, and entering LISTEN state if to accept connections. 1006 */ 1007static int 1008tcp_attach(so, p) 1009 struct socket *so; 1010 struct proc *p; 1011{ 1012 register struct tcpcb *tp; 1013 struct inpcb *inp; 1014 int error; 1015#ifdef INET6 1016 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; 1017#endif 1018 1019 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1020 error = soreserve(so, tcp_sendspace, tcp_recvspace); 1021 if (error) 1022 return (error); 1023 } 1024 error = in_pcballoc(so, &tcbinfo, p); 1025 if (error) 1026 return (error); 1027 inp = sotoinpcb(so); 1028#ifdef IPSEC 1029 error = ipsec_init_policy(so, &inp->inp_sp); 1030 if (error) { 1031#ifdef INET6 1032 if (isipv6) 1033 in6_pcbdetach(inp); 1034 else 1035#endif 1036 in_pcbdetach(inp); 1037 return (error); 1038 } 1039#endif /*IPSEC*/ 1040#ifdef INET6 1041 if (isipv6) { 1042 inp->inp_vflag |= INP_IPV6; 1043 inp->in6p_hops = -1; /* use kernel default */ 1044 } 1045 else 1046#endif 1047 inp->inp_vflag |= INP_IPV4; 1048 tp = tcp_newtcpcb(inp); 1049 if (tp == 0) { 1050 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1051 1052 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1053#ifdef INET6 1054 if (isipv6) 1055 in6_pcbdetach(inp); 1056 else 1057#endif 1058 in_pcbdetach(inp); 1059 so->so_state |= nofd; 1060 return (ENOBUFS); 1061 } 1062 tp->t_state = TCPS_CLOSED; 1063 return (0); 1064} 1065 1066/* 1067 * Initiate (or continue) disconnect. 1068 * If embryonic state, just send reset (once). 1069 * If in ``let data drain'' option and linger null, just drop. 1070 * Otherwise (hard), mark socket disconnecting and drop 1071 * current input data; switch states based on user close, and 1072 * send segment to peer (with FIN). 1073 */ 1074static struct tcpcb * 1075tcp_disconnect(tp) 1076 register struct tcpcb *tp; 1077{ 1078 struct socket *so = tp->t_inpcb->inp_socket; 1079 1080 if (tp->t_state < TCPS_ESTABLISHED) 1081 tp = tcp_close(tp); 1082 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1083 tp = tcp_drop(tp, 0); 1084 else { 1085 soisdisconnecting(so); 1086 sbflush(&so->so_rcv); 1087 tp = tcp_usrclosed(tp); 1088 if (tp) 1089 (void) tcp_output(tp); 1090 } 1091 return (tp); 1092} 1093 1094/* 1095 * User issued close, and wish to trail through shutdown states: 1096 * if never received SYN, just forget it. If got a SYN from peer, 1097 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1098 * If already got a FIN from peer, then almost done; go to LAST_ACK 1099 * state. In all other cases, have already sent FIN to peer (e.g. 1100 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1101 * for peer to send FIN or not respond to keep-alives, etc. 1102 * We can let the user exit from the close as soon as the FIN is acked. 1103 */ 1104static struct tcpcb * 1105tcp_usrclosed(tp) 1106 register struct tcpcb *tp; 1107{ 1108 1109 switch (tp->t_state) { 1110 1111 case TCPS_CLOSED: 1112 case TCPS_LISTEN: 1113 tp->t_state = TCPS_CLOSED; 1114 tp = tcp_close(tp); 1115 break; 1116 1117 case TCPS_SYN_SENT: 1118 case TCPS_SYN_RECEIVED: 1119 tp->t_flags |= TF_NEEDFIN; 1120 break; 1121 1122 case TCPS_ESTABLISHED: 1123 tp->t_state = TCPS_FIN_WAIT_1; 1124 break; 1125 1126 case TCPS_CLOSE_WAIT: 1127 tp->t_state = TCPS_LAST_ACK; 1128 break; 1129 } 1130 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1131 soisdisconnected(tp->t_inpcb->inp_socket); 1132 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1133 if (tp->t_state == TCPS_FIN_WAIT_2) 1134 callout_reset(tp->tt_2msl, tcp_maxidle, 1135 tcp_timer_2msl, tp); 1136 } 1137 return (tp); 1138} 1139 1140