tcp_usrreq.c revision 100685
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 100685 2002-07-25 18:10:04Z ume $ 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/malloc.h> 44#include <sys/kernel.h> 45#include <sys/sysctl.h> 46#include <sys/mbuf.h> 47#ifdef INET6 48#include <sys/domain.h> 49#endif /* INET6 */ 50#include <sys/socket.h> 51#include <sys/socketvar.h> 52#include <sys/protosw.h> 53#include <sys/proc.h> 54#include <sys/jail.h> 55 56#include <net/if.h> 57#include <net/route.h> 58 59#include <netinet/in.h> 60#include <netinet/in_systm.h> 61#ifdef INET6 62#include <netinet/ip6.h> 63#endif 64#include <netinet/in_pcb.h> 65#ifdef INET6 66#include <netinet6/in6_pcb.h> 67#endif 68#include <netinet/in_var.h> 69#include <netinet/ip_var.h> 70#ifdef INET6 71#include <netinet6/ip6_var.h> 72#endif 73#include <netinet/tcp.h> 74#include <netinet/tcp_fsm.h> 75#include <netinet/tcp_seq.h> 76#include <netinet/tcp_timer.h> 77#include <netinet/tcp_var.h> 78#include <netinet/tcpip.h> 79#ifdef TCPDEBUG 80#include <netinet/tcp_debug.h> 81#endif 82 83#ifdef IPSEC 84#include <netinet6/ipsec.h> 85#endif /*IPSEC*/ 86 87/* 88 * TCP protocol interface to socket abstraction. 89 */ 90extern char *tcpstates[]; /* XXX ??? */ 91 92static int tcp_attach(struct socket *, struct thread *td); 93static int tcp_connect(struct tcpcb *, struct sockaddr *, 94 struct thread *td); 95#ifdef INET6 96static int tcp6_connect(struct tcpcb *, struct sockaddr *, 97 struct thread *td); 98#endif /* INET6 */ 99static struct tcpcb * 100 tcp_disconnect(struct tcpcb *); 101static struct tcpcb * 102 tcp_usrclosed(struct tcpcb *); 103 104#ifdef TCPDEBUG 105#define TCPDEBUG0 int ostate = 0 106#define TCPDEBUG1() ostate = tp ? tp->t_state : 0 107#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ 108 tcp_trace(TA_USER, ostate, tp, 0, 0, req) 109#else 110#define TCPDEBUG0 111#define TCPDEBUG1() 112#define TCPDEBUG2(req) 113#endif 114 115/* 116 * TCP attaches to socket via pru_attach(), reserving space, 117 * and an internet control block. 118 */ 119static int 120tcp_usr_attach(struct socket *so, int proto, struct thread *td) 121{ 122 int s = splnet(); 123 int error; 124 struct inpcb *inp; 125 struct tcpcb *tp = 0; 126 TCPDEBUG0; 127 128 INP_INFO_WLOCK(&tcbinfo); 129 TCPDEBUG1(); 130 inp = sotoinpcb(so); 131 if (inp) { 132 error = EISCONN; 133 goto out; 134 } 135 136 error = tcp_attach(so, td); 137 if (error) 138 goto out; 139 140 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 141 so->so_linger = TCP_LINGERTIME; 142 143 inp = sotoinpcb(so); 144 tp = intotcpcb(inp); 145out: 146 TCPDEBUG2(PRU_ATTACH); 147 INP_INFO_WUNLOCK(&tcbinfo); 148 splx(s); 149 return error; 150} 151 152/* 153 * pru_detach() detaches the TCP protocol from the socket. 154 * If the protocol state is non-embryonic, then can't 155 * do this directly: have to initiate a pru_disconnect(), 156 * which may finish later; embryonic TCB's can just 157 * be discarded here. 158 */ 159static int 160tcp_usr_detach(struct socket *so) 161{ 162 int s = splnet(); 163 int error = 0; 164 struct inpcb *inp; 165 struct tcpcb *tp; 166 TCPDEBUG0; 167 168 INP_INFO_WLOCK(&tcbinfo); 169 inp = sotoinpcb(so); 170 if (inp == 0) { 171 INP_INFO_WUNLOCK(&tcbinfo); 172 splx(s); 173 return EINVAL; /* XXX */ 174 } 175 INP_LOCK(inp); 176 tp = intotcpcb(inp); 177 TCPDEBUG1(); 178 tp = tcp_disconnect(tp); 179 180 TCPDEBUG2(PRU_DETACH); 181 if (tp) 182 INP_UNLOCK(inp); 183 INP_INFO_WUNLOCK(&tcbinfo); 184 splx(s); 185 return error; 186} 187 188#define INI_NOLOCK 0 189#define INI_READ 1 190#define INI_WRITE 2 191 192#define COMMON_START() \ 193 TCPDEBUG0; \ 194 do { \ 195 if (inirw == INI_READ) \ 196 INP_INFO_RLOCK(&tcbinfo); \ 197 else if (inirw == INI_WRITE) \ 198 INP_INFO_WLOCK(&tcbinfo); \ 199 inp = sotoinpcb(so); \ 200 if (inp == 0) { \ 201 if (inirw == INI_READ) \ 202 INP_INFO_RUNLOCK(&tcbinfo); \ 203 else if (inirw == INI_WRITE) \ 204 INP_INFO_WUNLOCK(&tcbinfo); \ 205 splx(s); \ 206 return EINVAL; \ 207 } \ 208 INP_LOCK(inp); \ 209 if (inirw == INI_READ) \ 210 INP_INFO_RUNLOCK(&tcbinfo); \ 211 tp = intotcpcb(inp); \ 212 TCPDEBUG1(); \ 213} while(0) 214 215#define COMMON_END(req) \ 216out: TCPDEBUG2(req); \ 217 do { \ 218 if (tp) \ 219 INP_UNLOCK(inp); \ 220 if (inirw == INI_WRITE) \ 221 INP_INFO_WUNLOCK(&tcbinfo); \ 222 splx(s); \ 223 return error; \ 224 goto out; \ 225} while(0) 226 227/* 228 * Give the socket an address. 229 */ 230static int 231tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 232{ 233 int s = splnet(); 234 int error = 0; 235 struct inpcb *inp; 236 struct tcpcb *tp; 237 struct sockaddr_in *sinp; 238 const int inirw = INI_READ; 239 240 COMMON_START(); 241 242 /* 243 * Must check for multicast addresses and disallow binding 244 * to them. 245 */ 246 sinp = (struct sockaddr_in *)nam; 247 if (sinp->sin_family == AF_INET && 248 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 249 error = EAFNOSUPPORT; 250 goto out; 251 } 252 error = in_pcbbind(inp, nam, td); 253 if (error) 254 goto out; 255 COMMON_END(PRU_BIND); 256} 257 258#ifdef INET6 259static int 260tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 261{ 262 int s = splnet(); 263 int error = 0; 264 struct inpcb *inp; 265 struct tcpcb *tp; 266 struct sockaddr_in6 *sin6p; 267 const int inirw = INI_READ; 268 269 COMMON_START(); 270 271 /* 272 * Must check for multicast addresses and disallow binding 273 * to them. 274 */ 275 sin6p = (struct sockaddr_in6 *)nam; 276 if (sin6p->sin6_family == AF_INET6 && 277 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 278 error = EAFNOSUPPORT; 279 goto out; 280 } 281 inp->inp_vflag &= ~INP_IPV4; 282 inp->inp_vflag |= INP_IPV6; 283 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 284 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) 285 inp->inp_vflag |= INP_IPV4; 286 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 287 struct sockaddr_in sin; 288 289 in6_sin6_2_sin(&sin, sin6p); 290 inp->inp_vflag |= INP_IPV4; 291 inp->inp_vflag &= ~INP_IPV6; 292 error = in_pcbbind(inp, (struct sockaddr *)&sin, td); 293 goto out; 294 } 295 } 296 error = in6_pcbbind(inp, nam, td); 297 if (error) 298 goto out; 299 COMMON_END(PRU_BIND); 300} 301#endif /* INET6 */ 302 303/* 304 * Prepare to accept connections. 305 */ 306static int 307tcp_usr_listen(struct socket *so, struct thread *td) 308{ 309 int s = splnet(); 310 int error = 0; 311 struct inpcb *inp; 312 struct tcpcb *tp; 313 const int inirw = INI_READ; 314 315 COMMON_START(); 316 if (inp->inp_lport == 0) 317 error = in_pcbbind(inp, (struct sockaddr *)0, td); 318 if (error == 0) 319 tp->t_state = TCPS_LISTEN; 320 COMMON_END(PRU_LISTEN); 321} 322 323#ifdef INET6 324static int 325tcp6_usr_listen(struct socket *so, struct thread *td) 326{ 327 int s = splnet(); 328 int error = 0; 329 struct inpcb *inp; 330 struct tcpcb *tp; 331 const int inirw = INI_READ; 332 333 COMMON_START(); 334 if (inp->inp_lport == 0) { 335 inp->inp_vflag &= ~INP_IPV4; 336 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 337 inp->inp_vflag |= INP_IPV4; 338 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 339 } 340 if (error == 0) 341 tp->t_state = TCPS_LISTEN; 342 COMMON_END(PRU_LISTEN); 343} 344#endif /* INET6 */ 345 346/* 347 * Initiate connection to peer. 348 * Create a template for use in transmissions on this connection. 349 * Enter SYN_SENT state, and mark socket as connecting. 350 * Start keep-alive timer, and seed output sequence space. 351 * Send initial segment on connection. 352 */ 353static int 354tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 355{ 356 int s = splnet(); 357 int error = 0; 358 struct inpcb *inp; 359 struct tcpcb *tp; 360 struct sockaddr_in *sinp; 361 const int inirw = INI_WRITE; 362 363 COMMON_START(); 364 365 /* 366 * Must disallow TCP ``connections'' to multicast addresses. 367 */ 368 sinp = (struct sockaddr_in *)nam; 369 if (sinp->sin_family == AF_INET 370 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 371 error = EAFNOSUPPORT; 372 goto out; 373 } 374 375 if (td && jailed(td->td_ucred)) 376 prison_remote_ip(td->td_ucred, 0, &sinp->sin_addr.s_addr); 377 378 if ((error = tcp_connect(tp, nam, td)) != 0) 379 goto out; 380 error = tcp_output(tp); 381 COMMON_END(PRU_CONNECT); 382} 383 384#ifdef INET6 385static int 386tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 387{ 388 int s = splnet(); 389 int error = 0; 390 struct inpcb *inp; 391 struct tcpcb *tp; 392 struct sockaddr_in6 *sin6p; 393 const int inirw = INI_WRITE; 394 395 COMMON_START(); 396 397 /* 398 * Must disallow TCP ``connections'' to multicast addresses. 399 */ 400 sin6p = (struct sockaddr_in6 *)nam; 401 if (sin6p->sin6_family == AF_INET6 402 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 403 error = EAFNOSUPPORT; 404 goto out; 405 } 406 407 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 408 struct sockaddr_in sin; 409 410 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) 411 return(EINVAL); 412 413 in6_sin6_2_sin(&sin, sin6p); 414 inp->inp_vflag |= INP_IPV4; 415 inp->inp_vflag &= ~INP_IPV6; 416 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 417 goto out; 418 error = tcp_output(tp); 419 goto out; 420 } 421 inp->inp_vflag &= ~INP_IPV4; 422 inp->inp_vflag |= INP_IPV6; 423 inp->inp_inc.inc_isipv6 = 1; 424 if ((error = tcp6_connect(tp, nam, td)) != 0) 425 goto out; 426 error = tcp_output(tp); 427 COMMON_END(PRU_CONNECT); 428} 429#endif /* INET6 */ 430 431/* 432 * Initiate disconnect from peer. 433 * If connection never passed embryonic stage, just drop; 434 * else if don't need to let data drain, then can just drop anyways, 435 * else have to begin TCP shutdown process: mark socket disconnecting, 436 * drain unread data, state switch to reflect user close, and 437 * send segment (e.g. FIN) to peer. Socket will be really disconnected 438 * when peer sends FIN and acks ours. 439 * 440 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 441 */ 442static int 443tcp_usr_disconnect(struct socket *so) 444{ 445 int s = splnet(); 446 int error = 0; 447 struct inpcb *inp; 448 struct tcpcb *tp; 449 const int inirw = INI_WRITE; 450 451 COMMON_START(); 452 tp = tcp_disconnect(tp); 453 COMMON_END(PRU_DISCONNECT); 454} 455 456/* 457 * Accept a connection. Essentially all the work is 458 * done at higher levels; just return the address 459 * of the peer, storing through addr. 460 */ 461static int 462tcp_usr_accept(struct socket *so, struct sockaddr **nam) 463{ 464 int s; 465 int error = 0; 466 struct inpcb *inp = NULL; 467 struct tcpcb *tp = NULL; 468 struct sockaddr_in *sin; 469 const int inirw = INI_READ; 470 TCPDEBUG0; 471 472 if (so->so_state & SS_ISDISCONNECTED) { 473 error = ECONNABORTED; 474 goto out; 475 } 476 477 /* 478 * Do the malloc first in case it blocks. 479 */ 480 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, 481 M_WAITOK | M_ZERO); 482 sin->sin_family = AF_INET; 483 sin->sin_len = sizeof(*sin); 484 485 s = splnet(); 486 INP_INFO_RLOCK(&tcbinfo); 487 inp = sotoinpcb(so); 488 if (!inp) { 489 INP_INFO_RUNLOCK(&tcbinfo); 490 splx(s); 491 free(sin, M_SONAME); 492 return (EINVAL); 493 } 494 INP_LOCK(inp); 495 INP_INFO_RUNLOCK(&tcbinfo); 496 tp = intotcpcb(inp); 497 TCPDEBUG1(); 498 499 /* 500 * We inline in_setpeeraddr here, because we have already done 501 * the locking and the malloc. 502 */ 503 sin->sin_port = inp->inp_fport; 504 sin->sin_addr = inp->inp_faddr; 505 *nam = (struct sockaddr *)sin; 506 507 COMMON_END(PRU_ACCEPT); 508} 509 510#ifdef INET6 511static int 512tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 513{ 514 int s; 515 struct inpcb *inp = NULL; 516 int error = 0; 517 struct tcpcb *tp = NULL; 518 const int inirw = INI_READ; 519 TCPDEBUG0; 520 521 if (so->so_state & SS_ISDISCONNECTED) { 522 error = ECONNABORTED; 523 goto out; 524 } 525 526 s = splnet(); 527 INP_INFO_RLOCK(&tcbinfo); 528 inp = sotoinpcb(so); 529 if (inp == 0) { 530 INP_INFO_RUNLOCK(&tcbinfo); 531 splx(s); 532 return (EINVAL); 533 } 534 INP_LOCK(inp); 535 INP_INFO_RUNLOCK(&tcbinfo); 536 tp = intotcpcb(inp); 537 TCPDEBUG1(); 538 in6_mapped_peeraddr(so, nam); 539 COMMON_END(PRU_ACCEPT); 540} 541#endif /* INET6 */ 542 543/* 544 * This is the wrapper function for in_setsockaddr. We just pass down 545 * the pcbinfo for in_setsockaddr to lock. We don't want to do the locking 546 * here because in_setsockaddr will call malloc and can block. 547 */ 548static int 549tcp_sockaddr(struct socket *so, struct sockaddr **nam) 550{ 551 return (in_setsockaddr(so, nam, &tcbinfo)); 552} 553 554/* 555 * This is the wrapper function for in_setpeeraddr. We just pass down 556 * the pcbinfo for in_setpeeraddr to lock. 557 */ 558static int 559tcp_peeraddr(struct socket *so, struct sockaddr **nam) 560{ 561 return (in_setpeeraddr(so, nam, &tcbinfo)); 562} 563 564/* 565 * Mark the connection as being incapable of further output. 566 */ 567static int 568tcp_usr_shutdown(struct socket *so) 569{ 570 int s = splnet(); 571 int error = 0; 572 struct inpcb *inp; 573 struct tcpcb *tp; 574 const int inirw = INI_WRITE; 575 576 COMMON_START(); 577 socantsendmore(so); 578 tp = tcp_usrclosed(tp); 579 if (tp) 580 error = tcp_output(tp); 581 COMMON_END(PRU_SHUTDOWN); 582} 583 584/* 585 * After a receive, possibly send window update to peer. 586 */ 587static int 588tcp_usr_rcvd(struct socket *so, int flags) 589{ 590 int s = splnet(); 591 int error = 0; 592 struct inpcb *inp; 593 struct tcpcb *tp; 594 const int inirw = INI_READ; 595 596 COMMON_START(); 597 tcp_output(tp); 598 COMMON_END(PRU_RCVD); 599} 600 601/* 602 * Do a send by putting data in output queue and updating urgent 603 * marker if URG set. Possibly send more data. Unlike the other 604 * pru_*() routines, the mbuf chains are our responsibility. We 605 * must either enqueue them or free them. The other pru_* routines 606 * generally are caller-frees. 607 */ 608static int 609tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 610 struct sockaddr *nam, struct mbuf *control, struct thread *td) 611{ 612 int s = splnet(); 613 int error = 0; 614 struct inpcb *inp; 615 struct tcpcb *tp; 616 const int inirw = INI_WRITE; 617#ifdef INET6 618 int isipv6; 619#endif 620 TCPDEBUG0; 621 622 /* 623 * Need write lock here because this function might call 624 * tcp_connect or tcp_usrclosed. 625 * We really want to have to this function upgrade from read lock 626 * to write lock. XXX 627 */ 628 INP_INFO_WLOCK(&tcbinfo); 629 inp = sotoinpcb(so); 630 if (inp == NULL) { 631 /* 632 * OOPS! we lost a race, the TCP session got reset after 633 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 634 * network interrupt in the non-splnet() section of sosend(). 635 */ 636 if (m) 637 m_freem(m); 638 if (control) 639 m_freem(control); 640 error = ECONNRESET; /* XXX EPIPE? */ 641 tp = NULL; 642 TCPDEBUG1(); 643 goto out; 644 } 645 INP_LOCK(inp); 646#ifdef INET6 647 isipv6 = nam && nam->sa_family == AF_INET6; 648#endif /* INET6 */ 649 tp = intotcpcb(inp); 650 TCPDEBUG1(); 651 if (control) { 652 /* TCP doesn't do control messages (rights, creds, etc) */ 653 if (control->m_len) { 654 m_freem(control); 655 if (m) 656 m_freem(m); 657 error = EINVAL; 658 goto out; 659 } 660 m_freem(control); /* empty control, just free it */ 661 } 662 if (!(flags & PRUS_OOB)) { 663 sbappend(&so->so_snd, m); 664 if (nam && tp->t_state < TCPS_SYN_SENT) { 665 /* 666 * Do implied connect if not yet connected, 667 * initialize window to default value, and 668 * initialize maxseg/maxopd using peer's cached 669 * MSS. 670 */ 671#ifdef INET6 672 if (isipv6) 673 error = tcp6_connect(tp, nam, td); 674 else 675#endif /* INET6 */ 676 error = tcp_connect(tp, nam, td); 677 if (error) 678 goto out; 679 tp->snd_wnd = TTCP_CLIENT_SND_WND; 680 tcp_mss(tp, -1); 681 } 682 683 if (flags & PRUS_EOF) { 684 /* 685 * Close the send side of the connection after 686 * the data is sent. 687 */ 688 socantsendmore(so); 689 tp = tcp_usrclosed(tp); 690 } 691 if (tp != NULL) { 692 if (flags & PRUS_MORETOCOME) 693 tp->t_flags |= TF_MORETOCOME; 694 error = tcp_output(tp); 695 if (flags & PRUS_MORETOCOME) 696 tp->t_flags &= ~TF_MORETOCOME; 697 } 698 } else { 699 if (sbspace(&so->so_snd) < -512) { 700 m_freem(m); 701 error = ENOBUFS; 702 goto out; 703 } 704 /* 705 * According to RFC961 (Assigned Protocols), 706 * the urgent pointer points to the last octet 707 * of urgent data. We continue, however, 708 * to consider it to indicate the first octet 709 * of data past the urgent section. 710 * Otherwise, snd_up should be one lower. 711 */ 712 sbappend(&so->so_snd, m); 713 if (nam && tp->t_state < TCPS_SYN_SENT) { 714 /* 715 * Do implied connect if not yet connected, 716 * initialize window to default value, and 717 * initialize maxseg/maxopd using peer's cached 718 * MSS. 719 */ 720#ifdef INET6 721 if (isipv6) 722 error = tcp6_connect(tp, nam, td); 723 else 724#endif /* INET6 */ 725 error = tcp_connect(tp, nam, td); 726 if (error) 727 goto out; 728 tp->snd_wnd = TTCP_CLIENT_SND_WND; 729 tcp_mss(tp, -1); 730 } 731 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 732 tp->t_force = 1; 733 error = tcp_output(tp); 734 tp->t_force = 0; 735 } 736 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 737 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 738} 739 740/* 741 * Abort the TCP. 742 */ 743static int 744tcp_usr_abort(struct socket *so) 745{ 746 int s = splnet(); 747 int error = 0; 748 struct inpcb *inp; 749 struct tcpcb *tp; 750 const int inirw = INI_WRITE; 751 752 COMMON_START(); 753 tp = tcp_drop(tp, ECONNABORTED); 754 COMMON_END(PRU_ABORT); 755} 756 757/* 758 * Receive out-of-band data. 759 */ 760static int 761tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 762{ 763 int s = splnet(); 764 int error = 0; 765 struct inpcb *inp; 766 struct tcpcb *tp; 767 const int inirw = INI_READ; 768 769 COMMON_START(); 770 if ((so->so_oobmark == 0 && 771 (so->so_state & SS_RCVATMARK) == 0) || 772 so->so_options & SO_OOBINLINE || 773 tp->t_oobflags & TCPOOB_HADDATA) { 774 error = EINVAL; 775 goto out; 776 } 777 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 778 error = EWOULDBLOCK; 779 goto out; 780 } 781 m->m_len = 1; 782 *mtod(m, caddr_t) = tp->t_iobc; 783 if ((flags & MSG_PEEK) == 0) 784 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 785 COMMON_END(PRU_RCVOOB); 786} 787 788/* xxx - should be const */ 789struct pr_usrreqs tcp_usrreqs = { 790 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 791 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, 792 tcp_usr_disconnect, tcp_usr_listen, tcp_peeraddr, tcp_usr_rcvd, 793 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 794 tcp_sockaddr, sosend, soreceive, sopoll 795}; 796 797#ifdef INET6 798struct pr_usrreqs tcp6_usrreqs = { 799 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, 800 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, 801 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, 802 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 803 in6_mapped_sockaddr, sosend, soreceive, sopoll 804}; 805#endif /* INET6 */ 806 807/* 808 * Common subroutine to open a TCP connection to remote host specified 809 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 810 * port number if needed. Call in_pcbladdr to do the routing and to choose 811 * a local host address (interface). If there is an existing incarnation 812 * of the same connection in TIME-WAIT state and if the remote host was 813 * sending CC options and if the connection duration was < MSL, then 814 * truncate the previous TIME-WAIT state and proceed. 815 * Initialize connection parameters and enter SYN-SENT state. 816 */ 817static int 818tcp_connect(tp, nam, td) 819 register struct tcpcb *tp; 820 struct sockaddr *nam; 821 struct thread *td; 822{ 823 struct inpcb *inp = tp->t_inpcb, *oinp; 824 struct socket *so = inp->inp_socket; 825 struct tcpcb *otp; 826 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 827 struct sockaddr_in *ifaddr; 828 struct rmxp_tao *taop; 829 struct rmxp_tao tao_noncached; 830 int error; 831 832 if (inp->inp_lport == 0) { 833 error = in_pcbbind(inp, (struct sockaddr *)0, td); 834 if (error) 835 return error; 836 } 837 838 /* 839 * Cannot simply call in_pcbconnect, because there might be an 840 * earlier incarnation of this same connection still in 841 * TIME_WAIT state, creating an ADDRINUSE error. 842 */ 843 error = in_pcbladdr(inp, nam, &ifaddr); 844 if (error) 845 return error; 846 oinp = in_pcblookup_hash(inp->inp_pcbinfo, 847 sin->sin_addr, sin->sin_port, 848 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr 849 : ifaddr->sin_addr, 850 inp->inp_lport, 0, NULL); 851 if (oinp) { 852 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 853 otp->t_state == TCPS_TIME_WAIT && 854 (ticks - otp->t_starttime) < tcp_msl && 855 (otp->t_flags & TF_RCVD_CC)) 856 otp = tcp_close(otp); 857 else 858 return EADDRINUSE; 859 } 860 if (inp->inp_laddr.s_addr == INADDR_ANY) 861 inp->inp_laddr = ifaddr->sin_addr; 862 inp->inp_faddr = sin->sin_addr; 863 inp->inp_fport = sin->sin_port; 864 in_pcbrehash(inp); 865 866 /* Compute window scaling to request. */ 867 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 868 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 869 tp->request_r_scale++; 870 871 soisconnecting(so); 872 tcpstat.tcps_connattempt++; 873 tp->t_state = TCPS_SYN_SENT; 874 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 875 tp->iss = tcp_new_isn(tp); 876 tcp_sendseqinit(tp); 877 878 /* 879 * Generate a CC value for this connection and 880 * check whether CC or CCnew should be used. 881 */ 882 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 883 taop = &tao_noncached; 884 bzero(taop, sizeof(*taop)); 885 } 886 887 tp->cc_send = CC_INC(tcp_ccgen); 888 if (taop->tao_ccsent != 0 && 889 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 890 taop->tao_ccsent = tp->cc_send; 891 } else { 892 taop->tao_ccsent = 0; 893 tp->t_flags |= TF_SENDCCNEW; 894 } 895 896 return 0; 897} 898 899#ifdef INET6 900static int 901tcp6_connect(tp, nam, td) 902 register struct tcpcb *tp; 903 struct sockaddr *nam; 904 struct thread *td; 905{ 906 struct inpcb *inp = tp->t_inpcb, *oinp; 907 struct socket *so = inp->inp_socket; 908 struct tcpcb *otp; 909 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 910 struct in6_addr *addr6; 911 struct rmxp_tao *taop; 912 struct rmxp_tao tao_noncached; 913 int error; 914 915 if (inp->inp_lport == 0) { 916 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 917 if (error) 918 return error; 919 } 920 921 /* 922 * Cannot simply call in_pcbconnect, because there might be an 923 * earlier incarnation of this same connection still in 924 * TIME_WAIT state, creating an ADDRINUSE error. 925 */ 926 error = in6_pcbladdr(inp, nam, &addr6); 927 if (error) 928 return error; 929 oinp = in6_pcblookup_hash(inp->inp_pcbinfo, 930 &sin6->sin6_addr, sin6->sin6_port, 931 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) 932 ? addr6 933 : &inp->in6p_laddr, 934 inp->inp_lport, 0, NULL); 935 if (oinp) { 936 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 937 otp->t_state == TCPS_TIME_WAIT && 938 (ticks - otp->t_starttime) < tcp_msl && 939 (otp->t_flags & TF_RCVD_CC)) 940 otp = tcp_close(otp); 941 else 942 return EADDRINUSE; 943 } 944 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 945 inp->in6p_laddr = *addr6; 946 inp->in6p_faddr = sin6->sin6_addr; 947 inp->inp_fport = sin6->sin6_port; 948 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) 949 inp->in6p_flowinfo = sin6->sin6_flowinfo; 950 in_pcbrehash(inp); 951 952 /* Compute window scaling to request. */ 953 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 954 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 955 tp->request_r_scale++; 956 957 soisconnecting(so); 958 tcpstat.tcps_connattempt++; 959 tp->t_state = TCPS_SYN_SENT; 960 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 961 tp->iss = tcp_new_isn(tp); 962 tcp_sendseqinit(tp); 963 964 /* 965 * Generate a CC value for this connection and 966 * check whether CC or CCnew should be used. 967 */ 968 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 969 taop = &tao_noncached; 970 bzero(taop, sizeof(*taop)); 971 } 972 973 tp->cc_send = CC_INC(tcp_ccgen); 974 if (taop->tao_ccsent != 0 && 975 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 976 taop->tao_ccsent = tp->cc_send; 977 } else { 978 taop->tao_ccsent = 0; 979 tp->t_flags |= TF_SENDCCNEW; 980 } 981 982 return 0; 983} 984#endif /* INET6 */ 985 986/* 987 * The new sockopt interface makes it possible for us to block in the 988 * copyin/out step (if we take a page fault). Taking a page fault at 989 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now 990 * use TSM, there probably isn't any need for this function to run at 991 * splnet() any more. This needs more examination.) 992 */ 993int 994tcp_ctloutput(so, sopt) 995 struct socket *so; 996 struct sockopt *sopt; 997{ 998 int error, opt, optval, s; 999 struct inpcb *inp; 1000 struct tcpcb *tp; 1001 1002 error = 0; 1003 s = splnet(); /* XXX */ 1004 INP_INFO_RLOCK(&tcbinfo); 1005 inp = sotoinpcb(so); 1006 if (inp == NULL) { 1007 INP_INFO_RUNLOCK(&tcbinfo); 1008 splx(s); 1009 return (ECONNRESET); 1010 } 1011 INP_LOCK(inp); 1012 INP_INFO_RUNLOCK(&tcbinfo); 1013 if (sopt->sopt_level != IPPROTO_TCP) { 1014#ifdef INET6 1015 if (INP_CHECK_SOCKAF(so, AF_INET6)) 1016 error = ip6_ctloutput(so, sopt); 1017 else 1018#endif /* INET6 */ 1019 error = ip_ctloutput(so, sopt); 1020 INP_UNLOCK(inp); 1021 splx(s); 1022 return (error); 1023 } 1024 tp = intotcpcb(inp); 1025 1026 switch (sopt->sopt_dir) { 1027 case SOPT_SET: 1028 switch (sopt->sopt_name) { 1029 case TCP_NODELAY: 1030 case TCP_NOOPT: 1031 error = sooptcopyin(sopt, &optval, sizeof optval, 1032 sizeof optval); 1033 if (error) 1034 break; 1035 1036 switch (sopt->sopt_name) { 1037 case TCP_NODELAY: 1038 opt = TF_NODELAY; 1039 break; 1040 case TCP_NOOPT: 1041 opt = TF_NOOPT; 1042 break; 1043 default: 1044 opt = 0; /* dead code to fool gcc */ 1045 break; 1046 } 1047 1048 if (optval) 1049 tp->t_flags |= opt; 1050 else 1051 tp->t_flags &= ~opt; 1052 break; 1053 1054 case TCP_NOPUSH: 1055 error = sooptcopyin(sopt, &optval, sizeof optval, 1056 sizeof optval); 1057 if (error) 1058 break; 1059 1060 if (optval) 1061 tp->t_flags |= TF_NOPUSH; 1062 else { 1063 tp->t_flags &= ~TF_NOPUSH; 1064 error = tcp_output(tp); 1065 } 1066 break; 1067 1068 case TCP_MAXSEG: 1069 error = sooptcopyin(sopt, &optval, sizeof optval, 1070 sizeof optval); 1071 if (error) 1072 break; 1073 1074 if (optval > 0 && optval <= tp->t_maxseg) 1075 tp->t_maxseg = optval; 1076 else 1077 error = EINVAL; 1078 break; 1079 1080 default: 1081 error = ENOPROTOOPT; 1082 break; 1083 } 1084 break; 1085 1086 case SOPT_GET: 1087 switch (sopt->sopt_name) { 1088 case TCP_NODELAY: 1089 optval = tp->t_flags & TF_NODELAY; 1090 break; 1091 case TCP_MAXSEG: 1092 optval = tp->t_maxseg; 1093 break; 1094 case TCP_NOOPT: 1095 optval = tp->t_flags & TF_NOOPT; 1096 break; 1097 case TCP_NOPUSH: 1098 optval = tp->t_flags & TF_NOPUSH; 1099 break; 1100 default: 1101 error = ENOPROTOOPT; 1102 break; 1103 } 1104 if (error == 0) 1105 error = sooptcopyout(sopt, &optval, sizeof optval); 1106 break; 1107 } 1108 INP_UNLOCK(inp); 1109 splx(s); 1110 return (error); 1111} 1112 1113/* 1114 * tcp_sendspace and tcp_recvspace are the default send and receive window 1115 * sizes, respectively. These are obsolescent (this information should 1116 * be set by the route). 1117 */ 1118u_long tcp_sendspace = 1024*32; 1119SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1120 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1121u_long tcp_recvspace = 1024*64; 1122SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1123 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1124 1125/* 1126 * Attach TCP protocol to socket, allocating 1127 * internet protocol control block, tcp control block, 1128 * bufer space, and entering LISTEN state if to accept connections. 1129 */ 1130static int 1131tcp_attach(so, td) 1132 struct socket *so; 1133 struct thread *td; 1134{ 1135 register struct tcpcb *tp; 1136 struct inpcb *inp; 1137 int error; 1138#ifdef INET6 1139 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; 1140#endif 1141 1142 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1143 error = soreserve(so, tcp_sendspace, tcp_recvspace); 1144 if (error) 1145 return (error); 1146 } 1147 error = in_pcballoc(so, &tcbinfo, td); 1148 if (error) 1149 return (error); 1150 inp = sotoinpcb(so); 1151#ifdef INET6 1152 if (isipv6) { 1153 inp->inp_vflag |= INP_IPV6; 1154 inp->in6p_hops = -1; /* use kernel default */ 1155 } 1156 else 1157#endif 1158 inp->inp_vflag |= INP_IPV4; 1159 tp = tcp_newtcpcb(inp); 1160 if (tp == 0) { 1161 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1162 1163 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1164#ifdef INET6 1165 if (isipv6) 1166 in6_pcbdetach(inp); 1167 else 1168#endif 1169 in_pcbdetach(inp); 1170 so->so_state |= nofd; 1171 return (ENOBUFS); 1172 } 1173 tp->t_state = TCPS_CLOSED; 1174 return (0); 1175} 1176 1177/* 1178 * Initiate (or continue) disconnect. 1179 * If embryonic state, just send reset (once). 1180 * If in ``let data drain'' option and linger null, just drop. 1181 * Otherwise (hard), mark socket disconnecting and drop 1182 * current input data; switch states based on user close, and 1183 * send segment to peer (with FIN). 1184 */ 1185static struct tcpcb * 1186tcp_disconnect(tp) 1187 register struct tcpcb *tp; 1188{ 1189 struct socket *so = tp->t_inpcb->inp_socket; 1190 1191 if (tp->t_state < TCPS_ESTABLISHED) 1192 tp = tcp_close(tp); 1193 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1194 tp = tcp_drop(tp, 0); 1195 else { 1196 soisdisconnecting(so); 1197 sbflush(&so->so_rcv); 1198 tp = tcp_usrclosed(tp); 1199 if (tp) 1200 (void) tcp_output(tp); 1201 } 1202 return (tp); 1203} 1204 1205/* 1206 * User issued close, and wish to trail through shutdown states: 1207 * if never received SYN, just forget it. If got a SYN from peer, 1208 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1209 * If already got a FIN from peer, then almost done; go to LAST_ACK 1210 * state. In all other cases, have already sent FIN to peer (e.g. 1211 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1212 * for peer to send FIN or not respond to keep-alives, etc. 1213 * We can let the user exit from the close as soon as the FIN is acked. 1214 */ 1215static struct tcpcb * 1216tcp_usrclosed(tp) 1217 register struct tcpcb *tp; 1218{ 1219 1220 switch (tp->t_state) { 1221 1222 case TCPS_CLOSED: 1223 case TCPS_LISTEN: 1224 tp->t_state = TCPS_CLOSED; 1225 tp = tcp_close(tp); 1226 break; 1227 1228 case TCPS_SYN_SENT: 1229 case TCPS_SYN_RECEIVED: 1230 tp->t_flags |= TF_NEEDFIN; 1231 break; 1232 1233 case TCPS_ESTABLISHED: 1234 tp->t_state = TCPS_FIN_WAIT_1; 1235 break; 1236 1237 case TCPS_CLOSE_WAIT: 1238 tp->t_state = TCPS_LAST_ACK; 1239 break; 1240 } 1241 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1242 soisdisconnected(tp->t_inpcb->inp_socket); 1243 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1244 if (tp->t_state == TCPS_FIN_WAIT_2) 1245 callout_reset(tp->tt_2msl, tcp_maxidle, 1246 tcp_timer_2msl, tp); 1247 } 1248 return (tp); 1249} 1250 1251