1 /* 2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved. 3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved. 4 * 5 * This code is derived from software contributed to The DragonFly Project 6 * by Jeffrey M. Hsu. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of The DragonFly Project nor the names of its 17 * contributors may be used to endorse or promote products derived 18 * from this software without specific, prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 /* 35 * Copyright (c) 1982, 1986, 1988, 1993 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. All advertising materials mentioning features or use of this software 47 * must display the following acknowledgement: 48 * This product includes software developed by the University of 49 * California, Berkeley and its contributors. 50 * 4. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 67 * $FreeBSD: src/sys/netinet/tcp_usrreq.c,v 1.51.2.17 2002/10/11 11:46:44 ume Exp $ 68 * $DragonFly: src/sys/netinet/tcp_usrreq.c,v 1.51 2008/09/29 20:52:23 dillon Exp $ 69 */ 70 71 #include "opt_ipsec.h" 72 #include "opt_inet6.h" 73 #include "opt_tcpdebug.h" 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/kernel.h> 78 #include <sys/malloc.h> 79 #include <sys/sysctl.h> 80 #include <sys/globaldata.h> 81 #include <sys/thread.h> 82 83 #include <sys/mbuf.h> 84 #ifdef INET6 85 #include <sys/domain.h> 86 #endif /* INET6 */ 87 #include <sys/socket.h> 88 #include <sys/socketvar.h> 89 #include <sys/protosw.h> 90 91 #include <sys/thread2.h> 92 #include <sys/msgport2.h> 93 94 #include <net/if.h> 95 #include <net/netisr.h> 96 #include <net/route.h> 97 98 #include <net/netmsg2.h> 99 100 #include <netinet/in.h> 101 #include <netinet/in_systm.h> 102 #ifdef INET6 103 #include <netinet/ip6.h> 104 #endif 105 #include <netinet/in_pcb.h> 106 #ifdef INET6 107 #include <netinet6/in6_pcb.h> 108 #endif 109 #include <netinet/in_var.h> 110 #include <netinet/ip_var.h> 111 #ifdef INET6 112 #include <netinet6/ip6_var.h> 113 #endif 114 #include <netinet/tcp.h> 115 #include <netinet/tcp_fsm.h> 116 #include <netinet/tcp_seq.h> 117 #include <netinet/tcp_timer.h> 118 #include <netinet/tcp_var.h> 119 #include <netinet/tcpip.h> 120 #ifdef TCPDEBUG 121 #include <netinet/tcp_debug.h> 122 #endif 123 124 #ifdef IPSEC 125 #include <netinet6/ipsec.h> 126 #endif /*IPSEC*/ 127 128 /* 129 * TCP protocol interface to socket abstraction. 130 */ 131 extern char *tcpstates[]; /* XXX ??? */ 132 133 static int tcp_attach (struct socket *, struct pru_attach_info *); 134 static int tcp_connect (struct tcpcb *, struct sockaddr *, 135 struct thread *); 136 #ifdef INET6 137 static int tcp6_connect (struct tcpcb *, struct sockaddr *, 138 struct thread *); 139 #endif /* INET6 */ 140 static struct tcpcb * 141 tcp_disconnect (struct tcpcb *); 142 static struct tcpcb * 143 tcp_usrclosed (struct tcpcb *); 144 145 #ifdef TCPDEBUG 146 #define TCPDEBUG0 int ostate = 0 147 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 148 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ 149 tcp_trace(TA_USER, ostate, tp, 0, 0, req) 150 #else 151 #define TCPDEBUG0 152 #define TCPDEBUG1() 153 #define TCPDEBUG2(req) 154 #endif 155 156 /* 157 * TCP attaches to socket via pru_attach(), reserving space, 158 * and an internet control block. 159 */ 160 static int 161 tcp_usr_attach(struct socket *so, int proto, struct pru_attach_info *ai) 162 { 163 int error; 164 struct inpcb *inp; 165 struct tcpcb *tp = 0; 166 TCPDEBUG0; 167 168 crit_enter(); 169 inp = so->so_pcb; 170 TCPDEBUG1(); 171 if (inp) { 172 error = EISCONN; 173 goto out; 174 } 175 176 error = tcp_attach(so, ai); 177 if (error) 178 goto out; 179 180 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 181 so->so_linger = TCP_LINGERTIME; 182 tp = sototcpcb(so); 183 out: 184 TCPDEBUG2(PRU_ATTACH); 185 crit_exit(); 186 return error; 187 } 188 189 /* 190 * pru_detach() detaches the TCP protocol from the socket. 191 * If the protocol state is non-embryonic, then can't 192 * do this directly: have to initiate a pru_disconnect(), 193 * which may finish later; embryonic TCB's can just 194 * be discarded here. 195 */ 196 static int 197 tcp_usr_detach(struct socket *so) 198 { 199 int error = 0; 200 struct inpcb *inp; 201 struct tcpcb *tp; 202 TCPDEBUG0; 203 204 crit_enter(); 205 inp = so->so_pcb; 206 207 /* 208 * If the inp is already detached it may have been due to an async 209 * close. Just return as if no error occured. 210 */ 211 if (inp == NULL) { 212 crit_exit(); 213 return 0; 214 } 215 216 /* 217 * It's possible for the tcpcb (tp) to disconnect from the inp due 218 * to tcp_drop()->tcp_close() being called. This may occur *after* 219 * the detach message has been queued so we may find a NULL tp here. 220 */ 221 if ((tp = intotcpcb(inp)) != NULL) { 222 TCPDEBUG1(); 223 tp = tcp_disconnect(tp); 224 TCPDEBUG2(PRU_DETACH); 225 } 226 crit_exit(); 227 return error; 228 } 229 230 /* 231 * Note: ignore_error is non-zero for certain disconnection races 232 * which we want to silently allow, otherwise close() may return 233 * an unexpected error. 234 */ 235 #define COMMON_START(so, inp, ignore_error) \ 236 TCPDEBUG0; \ 237 \ 238 crit_enter(); \ 239 inp = so->so_pcb; \ 240 do { \ 241 if (inp == NULL) { \ 242 crit_exit(); \ 243 return (ignore_error ? 0 : EINVAL); \ 244 } \ 245 tp = intotcpcb(inp); \ 246 TCPDEBUG1(); \ 247 } while(0) 248 249 #define COMMON_END(req) out: TCPDEBUG2(req); crit_exit(); return error; goto out 250 251 252 /* 253 * Give the socket an address. 254 */ 255 static int 256 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 257 { 258 int error = 0; 259 struct inpcb *inp; 260 struct tcpcb *tp; 261 struct sockaddr_in *sinp; 262 263 COMMON_START(so, inp, 0); 264 265 /* 266 * Must check for multicast addresses and disallow binding 267 * to them. 268 */ 269 sinp = (struct sockaddr_in *)nam; 270 if (sinp->sin_family == AF_INET && 271 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 272 error = EAFNOSUPPORT; 273 goto out; 274 } 275 error = in_pcbbind(inp, nam, td); 276 if (error) 277 goto out; 278 COMMON_END(PRU_BIND); 279 280 } 281 282 #ifdef INET6 283 static int 284 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 285 { 286 int error = 0; 287 struct inpcb *inp; 288 struct tcpcb *tp; 289 struct sockaddr_in6 *sin6p; 290 291 COMMON_START(so, inp, 0); 292 293 /* 294 * Must check for multicast addresses and disallow binding 295 * to them. 296 */ 297 sin6p = (struct sockaddr_in6 *)nam; 298 if (sin6p->sin6_family == AF_INET6 && 299 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 300 error = EAFNOSUPPORT; 301 goto out; 302 } 303 inp->inp_vflag &= ~INP_IPV4; 304 inp->inp_vflag |= INP_IPV6; 305 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 306 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) 307 inp->inp_vflag |= INP_IPV4; 308 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 309 struct sockaddr_in sin; 310 311 in6_sin6_2_sin(&sin, sin6p); 312 inp->inp_vflag |= INP_IPV4; 313 inp->inp_vflag &= ~INP_IPV6; 314 error = in_pcbbind(inp, (struct sockaddr *)&sin, td); 315 goto out; 316 } 317 } 318 error = in6_pcbbind(inp, nam, td); 319 if (error) 320 goto out; 321 COMMON_END(PRU_BIND); 322 } 323 #endif /* INET6 */ 324 325 #ifdef SMP 326 struct netmsg_inswildcard { 327 struct netmsg nm_netmsg; 328 struct inpcb *nm_inp; 329 struct inpcbinfo *nm_pcbinfo; 330 }; 331 332 static void 333 in_pcbinswildcardhash_handler(struct netmsg *msg0) 334 { 335 struct netmsg_inswildcard *msg = (struct netmsg_inswildcard *)msg0; 336 337 in_pcbinswildcardhash_oncpu(msg->nm_inp, msg->nm_pcbinfo); 338 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0); 339 } 340 #endif 341 342 /* 343 * Prepare to accept connections. 344 */ 345 static int 346 tcp_usr_listen(struct socket *so, struct thread *td) 347 { 348 int error = 0; 349 struct inpcb *inp; 350 struct tcpcb *tp; 351 #ifdef SMP 352 int cpu; 353 #endif 354 355 COMMON_START(so, inp, 0); 356 if (inp->inp_lport == 0) { 357 error = in_pcbbind(inp, NULL, td); 358 if (error != 0) 359 goto out; 360 } 361 362 tp->t_state = TCPS_LISTEN; 363 #ifdef SMP 364 /* 365 * We have to set the flag because we can't have other cpus 366 * messing with our inp's flags. 367 */ 368 inp->inp_flags |= INP_WILDCARD_MP; 369 for (cpu = 0; cpu < ncpus2; cpu++) { 370 struct netmsg_inswildcard *msg; 371 372 if (cpu == mycpu->gd_cpuid) { 373 in_pcbinswildcardhash(inp); 374 continue; 375 } 376 377 msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, 378 M_INTWAIT); 379 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0, 380 in_pcbinswildcardhash_handler); 381 msg->nm_inp = inp; 382 msg->nm_pcbinfo = &tcbinfo[cpu]; 383 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_netmsg.nm_lmsg); 384 } 385 #else 386 in_pcbinswildcardhash(inp); 387 #endif 388 COMMON_END(PRU_LISTEN); 389 } 390 391 #ifdef INET6 392 static int 393 tcp6_usr_listen(struct socket *so, struct thread *td) 394 { 395 int error = 0; 396 struct inpcb *inp; 397 struct tcpcb *tp; 398 #ifdef SMP 399 int cpu; 400 #endif 401 402 COMMON_START(so, inp, 0); 403 if (inp->inp_lport == 0) { 404 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY)) 405 inp->inp_vflag |= INP_IPV4; 406 else 407 inp->inp_vflag &= ~INP_IPV4; 408 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 409 } 410 if (error == 0) 411 tp->t_state = TCPS_LISTEN; 412 #ifdef SMP 413 /* 414 * We have to set the flag because we can't have other cpus 415 * messing with our inp's flags. 416 */ 417 inp->inp_flags |= INP_WILDCARD_MP; 418 for (cpu = 0; cpu < ncpus2; cpu++) { 419 struct netmsg_inswildcard *msg; 420 421 if (cpu == mycpu->gd_cpuid) { 422 in_pcbinswildcardhash(inp); 423 continue; 424 } 425 426 msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, 427 M_INTWAIT); 428 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0, 429 in_pcbinswildcardhash_handler); 430 msg->nm_inp = inp; 431 msg->nm_pcbinfo = &tcbinfo[cpu]; 432 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_netmsg.nm_lmsg); 433 } 434 #else 435 in_pcbinswildcardhash(inp); 436 #endif 437 COMMON_END(PRU_LISTEN); 438 } 439 #endif /* INET6 */ 440 441 #ifdef SMP 442 static void 443 tcp_output_dispatch(struct netmsg *nmsg) 444 { 445 struct lwkt_msg *msg = &nmsg->nm_lmsg; 446 struct tcpcb *tp = msg->u.ms_resultp; 447 int error; 448 449 error = tcp_output(tp); 450 lwkt_replymsg(msg, error); 451 } 452 #endif 453 454 /* 455 * Initiate connection to peer. 456 * Create a template for use in transmissions on this connection. 457 * Enter SYN_SENT state, and mark socket as connecting. 458 * Start keep-alive timer, and seed output sequence space. 459 * Send initial segment on connection. 460 */ 461 static int 462 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 463 { 464 int error = 0; 465 struct inpcb *inp; 466 struct tcpcb *tp; 467 struct sockaddr_in *sinp; 468 #ifdef SMP 469 lwkt_port_t port; 470 #endif 471 472 COMMON_START(so, inp, 0); 473 474 /* 475 * Must disallow TCP ``connections'' to multicast addresses. 476 */ 477 sinp = (struct sockaddr_in *)nam; 478 if (sinp->sin_family == AF_INET 479 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 480 error = EAFNOSUPPORT; 481 goto out; 482 } 483 484 if (!prison_remote_ip(td, (struct sockaddr*)sinp)) { 485 error = EAFNOSUPPORT; /* IPv6 only jail */ 486 goto out; 487 } 488 489 if ((error = tcp_connect(tp, nam, td)) != 0) 490 goto out; 491 492 #ifdef SMP 493 port = tcp_addrport(inp->inp_faddr.s_addr, inp->inp_fport, 494 inp->inp_laddr.s_addr, inp->inp_lport); 495 if (port != &curthread->td_msgport) { 496 struct netmsg nmsg; 497 struct lwkt_msg *msg; 498 499 netmsg_init(&nmsg, &curthread->td_msgport, 0, 500 tcp_output_dispatch); 501 msg = &nmsg.nm_lmsg; 502 msg->u.ms_resultp = tp; 503 504 error = lwkt_domsg(port, msg, 0); 505 } else 506 #endif 507 error = tcp_output(tp); 508 COMMON_END(PRU_CONNECT); 509 } 510 511 #ifdef INET6 512 static int 513 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 514 { 515 int error = 0; 516 struct inpcb *inp; 517 struct tcpcb *tp; 518 struct sockaddr_in6 *sin6p; 519 520 COMMON_START(so, inp, 0); 521 522 /* 523 * Must disallow TCP ``connections'' to multicast addresses. 524 */ 525 sin6p = (struct sockaddr_in6 *)nam; 526 if (sin6p->sin6_family == AF_INET6 527 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 528 error = EAFNOSUPPORT; 529 goto out; 530 } 531 532 if (!prison_remote_ip(td, nam)) { 533 error = EAFNOSUPPORT; /* IPv4 only jail */ 534 goto out; 535 } 536 537 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 538 struct sockaddr_in sin; 539 540 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 541 error = EINVAL; 542 goto out; 543 } 544 545 in6_sin6_2_sin(&sin, sin6p); 546 inp->inp_vflag |= INP_IPV4; 547 inp->inp_vflag &= ~INP_IPV6; 548 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 549 goto out; 550 error = tcp_output(tp); 551 goto out; 552 } 553 inp->inp_vflag &= ~INP_IPV4; 554 inp->inp_vflag |= INP_IPV6; 555 inp->inp_inc.inc_isipv6 = 1; 556 if ((error = tcp6_connect(tp, nam, td)) != 0) 557 goto out; 558 error = tcp_output(tp); 559 COMMON_END(PRU_CONNECT); 560 } 561 #endif /* INET6 */ 562 563 /* 564 * Initiate disconnect from peer. 565 * If connection never passed embryonic stage, just drop; 566 * else if don't need to let data drain, then can just drop anyways, 567 * else have to begin TCP shutdown process: mark socket disconnecting, 568 * drain unread data, state switch to reflect user close, and 569 * send segment (e.g. FIN) to peer. Socket will be really disconnected 570 * when peer sends FIN and acks ours. 571 * 572 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 573 */ 574 static int 575 tcp_usr_disconnect(struct socket *so) 576 { 577 int error = 0; 578 struct inpcb *inp; 579 struct tcpcb *tp; 580 581 COMMON_START(so, inp, 1); 582 tp = tcp_disconnect(tp); 583 COMMON_END(PRU_DISCONNECT); 584 } 585 586 /* 587 * Accept a connection. Essentially all the work is 588 * done at higher levels; just return the address 589 * of the peer, storing through addr. 590 */ 591 static int 592 tcp_usr_accept(struct socket *so, struct sockaddr **nam) 593 { 594 int error = 0; 595 struct inpcb *inp; 596 struct tcpcb *tp = NULL; 597 TCPDEBUG0; 598 599 crit_enter(); 600 inp = so->so_pcb; 601 if (so->so_state & SS_ISDISCONNECTED) { 602 error = ECONNABORTED; 603 goto out; 604 } 605 if (inp == 0) { 606 crit_exit(); 607 return (EINVAL); 608 } 609 tp = intotcpcb(inp); 610 TCPDEBUG1(); 611 in_setpeeraddr(so, nam); 612 COMMON_END(PRU_ACCEPT); 613 } 614 615 #ifdef INET6 616 static int 617 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 618 { 619 int error = 0; 620 struct inpcb *inp; 621 struct tcpcb *tp = NULL; 622 TCPDEBUG0; 623 624 crit_enter(); 625 inp = so->so_pcb; 626 627 if (so->so_state & SS_ISDISCONNECTED) { 628 error = ECONNABORTED; 629 goto out; 630 } 631 if (inp == 0) { 632 crit_exit(); 633 return (EINVAL); 634 } 635 tp = intotcpcb(inp); 636 TCPDEBUG1(); 637 in6_mapped_peeraddr(so, nam); 638 COMMON_END(PRU_ACCEPT); 639 } 640 #endif /* INET6 */ 641 /* 642 * Mark the connection as being incapable of further output. 643 */ 644 static int 645 tcp_usr_shutdown(struct socket *so) 646 { 647 int error = 0; 648 struct inpcb *inp; 649 struct tcpcb *tp; 650 651 COMMON_START(so, inp, 0); 652 socantsendmore(so); 653 tp = tcp_usrclosed(tp); 654 if (tp) 655 error = tcp_output(tp); 656 COMMON_END(PRU_SHUTDOWN); 657 } 658 659 /* 660 * After a receive, possibly send window update to peer. 661 */ 662 static int 663 tcp_usr_rcvd(struct socket *so, int flags) 664 { 665 int error = 0; 666 struct inpcb *inp; 667 struct tcpcb *tp; 668 669 COMMON_START(so, inp, 0); 670 tcp_output(tp); 671 COMMON_END(PRU_RCVD); 672 } 673 674 /* 675 * Do a send by putting data in output queue and updating urgent 676 * marker if URG set. Possibly send more data. Unlike the other 677 * pru_*() routines, the mbuf chains are our responsibility. We 678 * must either enqueue them or free them. The other pru_* routines 679 * generally are caller-frees. 680 */ 681 static int 682 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 683 struct sockaddr *nam, struct mbuf *control, struct thread *td) 684 { 685 int error = 0; 686 struct inpcb *inp; 687 struct tcpcb *tp; 688 #ifdef INET6 689 int isipv6; 690 #endif 691 TCPDEBUG0; 692 693 crit_enter(); 694 inp = so->so_pcb; 695 696 if (inp == NULL) { 697 /* 698 * OOPS! we lost a race, the TCP session got reset after 699 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 700 * network interrupt in the non-critical section of sosend(). 701 */ 702 if (m) 703 m_freem(m); 704 if (control) 705 m_freem(control); 706 error = ECONNRESET; /* XXX EPIPE? */ 707 tp = NULL; 708 TCPDEBUG1(); 709 goto out; 710 } 711 #ifdef INET6 712 isipv6 = nam && nam->sa_family == AF_INET6; 713 #endif /* INET6 */ 714 tp = intotcpcb(inp); 715 TCPDEBUG1(); 716 if (control) { 717 /* TCP doesn't do control messages (rights, creds, etc) */ 718 if (control->m_len) { 719 m_freem(control); 720 if (m) 721 m_freem(m); 722 error = EINVAL; 723 goto out; 724 } 725 m_freem(control); /* empty control, just free it */ 726 } 727 if(!(flags & PRUS_OOB)) { 728 ssb_appendstream(&so->so_snd, m); 729 if (nam && tp->t_state < TCPS_SYN_SENT) { 730 /* 731 * Do implied connect if not yet connected, 732 * initialize window to default value, and 733 * initialize maxseg/maxopd using peer's cached 734 * MSS. 735 */ 736 #ifdef INET6 737 if (isipv6) 738 error = tcp6_connect(tp, nam, td); 739 else 740 #endif /* INET6 */ 741 error = tcp_connect(tp, nam, td); 742 if (error) 743 goto out; 744 tp->snd_wnd = TTCP_CLIENT_SND_WND; 745 tcp_mss(tp, -1); 746 } 747 748 if (flags & PRUS_EOF) { 749 /* 750 * Close the send side of the connection after 751 * the data is sent. 752 */ 753 socantsendmore(so); 754 tp = tcp_usrclosed(tp); 755 } 756 if (tp != NULL) { 757 if (flags & PRUS_MORETOCOME) 758 tp->t_flags |= TF_MORETOCOME; 759 error = tcp_output(tp); 760 if (flags & PRUS_MORETOCOME) 761 tp->t_flags &= ~TF_MORETOCOME; 762 } 763 } else { 764 if (ssb_space(&so->so_snd) < -512) { 765 m_freem(m); 766 error = ENOBUFS; 767 goto out; 768 } 769 /* 770 * According to RFC961 (Assigned Protocols), 771 * the urgent pointer points to the last octet 772 * of urgent data. We continue, however, 773 * to consider it to indicate the first octet 774 * of data past the urgent section. 775 * Otherwise, snd_up should be one lower. 776 */ 777 ssb_appendstream(&so->so_snd, m); 778 if (nam && tp->t_state < TCPS_SYN_SENT) { 779 /* 780 * Do implied connect if not yet connected, 781 * initialize window to default value, and 782 * initialize maxseg/maxopd using peer's cached 783 * MSS. 784 */ 785 #ifdef INET6 786 if (isipv6) 787 error = tcp6_connect(tp, nam, td); 788 else 789 #endif /* INET6 */ 790 error = tcp_connect(tp, nam, td); 791 if (error) 792 goto out; 793 tp->snd_wnd = TTCP_CLIENT_SND_WND; 794 tcp_mss(tp, -1); 795 } 796 tp->snd_up = tp->snd_una + so->so_snd.ssb_cc; 797 tp->t_flags |= TF_FORCE; 798 error = tcp_output(tp); 799 tp->t_flags &= ~TF_FORCE; 800 } 801 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 802 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 803 } 804 805 /* 806 * Abort the TCP. 807 */ 808 static int 809 tcp_usr_abort(struct socket *so) 810 { 811 int error = 0; 812 struct inpcb *inp; 813 struct tcpcb *tp; 814 815 COMMON_START(so, inp, 1); 816 tp = tcp_drop(tp, ECONNABORTED); 817 COMMON_END(PRU_ABORT); 818 } 819 820 /* 821 * Receive out-of-band data. 822 */ 823 static int 824 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 825 { 826 int error = 0; 827 struct inpcb *inp; 828 struct tcpcb *tp; 829 830 COMMON_START(so, inp, 0); 831 if ((so->so_oobmark == 0 && 832 (so->so_state & SS_RCVATMARK) == 0) || 833 so->so_options & SO_OOBINLINE || 834 tp->t_oobflags & TCPOOB_HADDATA) { 835 error = EINVAL; 836 goto out; 837 } 838 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 839 error = EWOULDBLOCK; 840 goto out; 841 } 842 m->m_len = 1; 843 *mtod(m, caddr_t) = tp->t_iobc; 844 if ((flags & MSG_PEEK) == 0) 845 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 846 COMMON_END(PRU_RCVOOB); 847 } 848 849 /* xxx - should be const */ 850 struct pr_usrreqs tcp_usrreqs = { 851 .pru_abort = tcp_usr_abort, 852 .pru_accept = tcp_usr_accept, 853 .pru_attach = tcp_usr_attach, 854 .pru_bind = tcp_usr_bind, 855 .pru_connect = tcp_usr_connect, 856 .pru_connect2 = pru_connect2_notsupp, 857 .pru_control = in_control, 858 .pru_detach = tcp_usr_detach, 859 .pru_disconnect = tcp_usr_disconnect, 860 .pru_listen = tcp_usr_listen, 861 .pru_peeraddr = in_setpeeraddr, 862 .pru_rcvd = tcp_usr_rcvd, 863 .pru_rcvoob = tcp_usr_rcvoob, 864 .pru_send = tcp_usr_send, 865 .pru_sense = pru_sense_null, 866 .pru_shutdown = tcp_usr_shutdown, 867 .pru_sockaddr = in_setsockaddr, 868 .pru_sosend = sosend, 869 .pru_soreceive = soreceive, 870 .pru_sopoll = sopoll 871 }; 872 873 #ifdef INET6 874 struct pr_usrreqs tcp6_usrreqs = { 875 .pru_abort = tcp_usr_abort, 876 .pru_accept = tcp6_usr_accept, 877 .pru_attach = tcp_usr_attach, 878 .pru_bind = tcp6_usr_bind, 879 .pru_connect = tcp6_usr_connect, 880 .pru_connect2 = pru_connect2_notsupp, 881 .pru_control = in6_control, 882 .pru_detach = tcp_usr_detach, 883 .pru_disconnect = tcp_usr_disconnect, 884 .pru_listen = tcp6_usr_listen, 885 .pru_peeraddr = in6_mapped_peeraddr, 886 .pru_rcvd = tcp_usr_rcvd, 887 .pru_rcvoob = tcp_usr_rcvoob, 888 .pru_send = tcp_usr_send, 889 .pru_sense = pru_sense_null, 890 .pru_shutdown = tcp_usr_shutdown, 891 .pru_sockaddr = in6_mapped_sockaddr, 892 .pru_sosend = sosend, 893 .pru_soreceive = soreceive, 894 .pru_sopoll = sopoll 895 }; 896 #endif /* INET6 */ 897 898 static int 899 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin, 900 struct sockaddr_in *if_sin) 901 { 902 struct inpcb *inp = tp->t_inpcb, *oinp; 903 struct socket *so = inp->inp_socket; 904 struct tcpcb *otp; 905 struct rmxp_tao *taop; 906 struct rmxp_tao tao_noncached; 907 908 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid], 909 sin->sin_addr, sin->sin_port, 910 inp->inp_laddr.s_addr != INADDR_ANY ? 911 inp->inp_laddr : if_sin->sin_addr, 912 inp->inp_lport, 0, NULL); 913 if (oinp != NULL) { 914 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 915 otp->t_state == TCPS_TIME_WAIT && 916 (ticks - otp->t_starttime) < tcp_msl && 917 (otp->t_flags & TF_RCVD_CC)) 918 tcp_close(otp); 919 else 920 return (EADDRINUSE); 921 } 922 if (inp->inp_laddr.s_addr == INADDR_ANY) 923 inp->inp_laddr = if_sin->sin_addr; 924 inp->inp_faddr = sin->sin_addr; 925 inp->inp_fport = sin->sin_port; 926 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid]; 927 in_pcbinsconnhash(inp); 928 929 /* Compute window scaling to request. */ 930 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 931 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat) 932 tp->request_r_scale++; 933 934 soisconnecting(so); 935 tcpstat.tcps_connattempt++; 936 tp->t_state = TCPS_SYN_SENT; 937 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 938 tp->iss = tcp_new_isn(tp); 939 tcp_sendseqinit(tp); 940 941 /* 942 * Generate a CC value for this connection and 943 * check whether CC or CCnew should be used. 944 */ 945 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 946 taop = &tao_noncached; 947 bzero(taop, sizeof *taop); 948 } 949 950 tp->cc_send = CC_INC(tcp_ccgen); 951 if (taop->tao_ccsent != 0 && 952 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 953 taop->tao_ccsent = tp->cc_send; 954 } else { 955 taop->tao_ccsent = 0; 956 tp->t_flags |= TF_SENDCCNEW; 957 } 958 959 return (0); 960 } 961 962 #ifdef SMP 963 964 struct netmsg_tcp_connect { 965 struct netmsg nm_netmsg; 966 struct tcpcb *nm_tp; 967 struct sockaddr_in *nm_sin; 968 struct sockaddr_in *nm_ifsin; 969 }; 970 971 static void 972 tcp_connect_handler(netmsg_t netmsg) 973 { 974 struct netmsg_tcp_connect *msg = (void *)netmsg; 975 int error; 976 977 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin); 978 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, error); 979 } 980 981 #endif 982 983 /* 984 * Common subroutine to open a TCP connection to remote host specified 985 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 986 * port number if needed. Call in_pcbladdr to do the routing and to choose 987 * a local host address (interface). If there is an existing incarnation 988 * of the same connection in TIME-WAIT state and if the remote host was 989 * sending CC options and if the connection duration was < MSL, then 990 * truncate the previous TIME-WAIT state and proceed. 991 * Initialize connection parameters and enter SYN-SENT state. 992 */ 993 static int 994 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 995 { 996 struct inpcb *inp = tp->t_inpcb; 997 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 998 struct sockaddr_in *if_sin; 999 int error; 1000 #ifdef SMP 1001 lwkt_port_t port; 1002 #endif 1003 1004 if (inp->inp_lport == 0) { 1005 error = in_pcbbind(inp, (struct sockaddr *)NULL, td); 1006 if (error) 1007 return (error); 1008 } 1009 1010 /* 1011 * Cannot simply call in_pcbconnect, because there might be an 1012 * earlier incarnation of this same connection still in 1013 * TIME_WAIT state, creating an ADDRINUSE error. 1014 */ 1015 error = in_pcbladdr(inp, nam, &if_sin, td); 1016 if (error) 1017 return (error); 1018 1019 #ifdef SMP 1020 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port, 1021 inp->inp_laddr.s_addr ? 1022 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr, 1023 inp->inp_lport); 1024 1025 if (port != &curthread->td_msgport) { 1026 struct netmsg_tcp_connect msg; 1027 struct route *ro = &inp->inp_route; 1028 1029 /* 1030 * in_pcbladdr() may have allocated a route entry for us 1031 * on the current CPU, but we need a route entry on the 1032 * target CPU, so free it here. 1033 */ 1034 if (ro->ro_rt != NULL) 1035 RTFREE(ro->ro_rt); 1036 bzero(ro, sizeof(*ro)); 1037 1038 netmsg_init(&msg.nm_netmsg, &curthread->td_msgport, 0, 1039 tcp_connect_handler); 1040 msg.nm_tp = tp; 1041 msg.nm_sin = sin; 1042 msg.nm_ifsin = if_sin; 1043 error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, 0); 1044 } else 1045 #endif 1046 error = tcp_connect_oncpu(tp, sin, if_sin); 1047 1048 return (error); 1049 } 1050 1051 #ifdef INET6 1052 static int 1053 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 1054 { 1055 struct inpcb *inp = tp->t_inpcb, *oinp; 1056 struct socket *so = inp->inp_socket; 1057 struct tcpcb *otp; 1058 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 1059 struct in6_addr *addr6; 1060 struct rmxp_tao *taop; 1061 struct rmxp_tao tao_noncached; 1062 int error; 1063 1064 if (inp->inp_lport == 0) { 1065 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 1066 if (error) 1067 return error; 1068 } 1069 1070 /* 1071 * Cannot simply call in_pcbconnect, because there might be an 1072 * earlier incarnation of this same connection still in 1073 * TIME_WAIT state, creating an ADDRINUSE error. 1074 */ 1075 error = in6_pcbladdr(inp, nam, &addr6, td); 1076 if (error) 1077 return error; 1078 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo, 1079 &sin6->sin6_addr, sin6->sin6_port, 1080 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? 1081 addr6 : &inp->in6p_laddr, 1082 inp->inp_lport, 0, NULL); 1083 if (oinp) { 1084 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 1085 otp->t_state == TCPS_TIME_WAIT && 1086 (ticks - otp->t_starttime) < tcp_msl && 1087 (otp->t_flags & TF_RCVD_CC)) 1088 otp = tcp_close(otp); 1089 else 1090 return (EADDRINUSE); 1091 } 1092 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 1093 inp->in6p_laddr = *addr6; 1094 inp->in6p_faddr = sin6->sin6_addr; 1095 inp->inp_fport = sin6->sin6_port; 1096 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0) 1097 inp->in6p_flowinfo = sin6->sin6_flowinfo; 1098 in_pcbinsconnhash(inp); 1099 1100 /* Compute window scaling to request. */ 1101 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1102 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat) 1103 tp->request_r_scale++; 1104 1105 soisconnecting(so); 1106 tcpstat.tcps_connattempt++; 1107 tp->t_state = TCPS_SYN_SENT; 1108 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 1109 tp->iss = tcp_new_isn(tp); 1110 tcp_sendseqinit(tp); 1111 1112 /* 1113 * Generate a CC value for this connection and 1114 * check whether CC or CCnew should be used. 1115 */ 1116 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 1117 taop = &tao_noncached; 1118 bzero(taop, sizeof *taop); 1119 } 1120 1121 tp->cc_send = CC_INC(tcp_ccgen); 1122 if (taop->tao_ccsent != 0 && 1123 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 1124 taop->tao_ccsent = tp->cc_send; 1125 } else { 1126 taop->tao_ccsent = 0; 1127 tp->t_flags |= TF_SENDCCNEW; 1128 } 1129 1130 return (0); 1131 } 1132 #endif /* INET6 */ 1133 1134 /* 1135 * The new sockopt interface makes it possible for us to block in the 1136 * copyin/out step (if we take a page fault). Taking a page fault while 1137 * in a critical section is probably a Bad Thing. (Since sockets and pcbs 1138 * both now use TSM, there probably isn't any need for this function to 1139 * run in a critical section any more. This needs more examination.) 1140 */ 1141 int 1142 tcp_ctloutput(struct socket *so, struct sockopt *sopt) 1143 { 1144 int error, opt, optval; 1145 struct inpcb *inp; 1146 struct tcpcb *tp; 1147 1148 error = 0; 1149 crit_enter(); /* XXX */ 1150 inp = so->so_pcb; 1151 if (inp == NULL) { 1152 crit_exit(); 1153 return (ECONNRESET); 1154 } 1155 if (sopt->sopt_level != IPPROTO_TCP) { 1156 #ifdef INET6 1157 if (INP_CHECK_SOCKAF(so, AF_INET6)) 1158 error = ip6_ctloutput(so, sopt); 1159 else 1160 #endif /* INET6 */ 1161 error = ip_ctloutput(so, sopt); 1162 crit_exit(); 1163 return (error); 1164 } 1165 tp = intotcpcb(inp); 1166 1167 switch (sopt->sopt_dir) { 1168 case SOPT_SET: 1169 error = soopt_to_kbuf(sopt, &optval, sizeof optval, 1170 sizeof optval); 1171 if (error) 1172 break; 1173 switch (sopt->sopt_name) { 1174 case TCP_NODELAY: 1175 case TCP_NOOPT: 1176 switch (sopt->sopt_name) { 1177 case TCP_NODELAY: 1178 opt = TF_NODELAY; 1179 break; 1180 case TCP_NOOPT: 1181 opt = TF_NOOPT; 1182 break; 1183 default: 1184 opt = 0; /* dead code to fool gcc */ 1185 break; 1186 } 1187 1188 if (optval) 1189 tp->t_flags |= opt; 1190 else 1191 tp->t_flags &= ~opt; 1192 break; 1193 1194 case TCP_NOPUSH: 1195 if (optval) 1196 tp->t_flags |= TF_NOPUSH; 1197 else { 1198 tp->t_flags &= ~TF_NOPUSH; 1199 error = tcp_output(tp); 1200 } 1201 break; 1202 1203 case TCP_MAXSEG: 1204 if (optval > 0 && optval <= tp->t_maxseg) 1205 tp->t_maxseg = optval; 1206 else 1207 error = EINVAL; 1208 break; 1209 1210 default: 1211 error = ENOPROTOOPT; 1212 break; 1213 } 1214 break; 1215 1216 case SOPT_GET: 1217 switch (sopt->sopt_name) { 1218 case TCP_NODELAY: 1219 optval = tp->t_flags & TF_NODELAY; 1220 break; 1221 case TCP_MAXSEG: 1222 optval = tp->t_maxseg; 1223 break; 1224 case TCP_NOOPT: 1225 optval = tp->t_flags & TF_NOOPT; 1226 break; 1227 case TCP_NOPUSH: 1228 optval = tp->t_flags & TF_NOPUSH; 1229 break; 1230 default: 1231 error = ENOPROTOOPT; 1232 break; 1233 } 1234 if (error == 0) 1235 soopt_from_kbuf(sopt, &optval, sizeof optval); 1236 break; 1237 } 1238 crit_exit(); 1239 return (error); 1240 } 1241 1242 /* 1243 * tcp_sendspace and tcp_recvspace are the default send and receive window 1244 * sizes, respectively. These are obsolescent (this information should 1245 * be set by the route). 1246 * 1247 * Use a default that does not require tcp window scaling to be turned 1248 * on. Individual programs or the administrator can increase the default. 1249 */ 1250 u_long tcp_sendspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ 1251 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1252 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1253 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ 1254 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1255 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1256 1257 /* 1258 * Attach TCP protocol to socket, allocating 1259 * internet protocol control block, tcp control block, 1260 * bufer space, and entering LISTEN state if to accept connections. 1261 */ 1262 static int 1263 tcp_attach(struct socket *so, struct pru_attach_info *ai) 1264 { 1265 struct tcpcb *tp; 1266 struct inpcb *inp; 1267 int error; 1268 int cpu; 1269 #ifdef INET6 1270 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0; 1271 #endif 1272 1273 if (so->so_snd.ssb_hiwat == 0 || so->so_rcv.ssb_hiwat == 0) { 1274 error = soreserve(so, tcp_sendspace, tcp_recvspace, 1275 ai->sb_rlimit); 1276 if (error) 1277 return (error); 1278 } 1279 cpu = mycpu->gd_cpuid; 1280 error = in_pcballoc(so, &tcbinfo[cpu]); 1281 if (error) 1282 return (error); 1283 inp = so->so_pcb; 1284 #ifdef INET6 1285 if (isipv6) { 1286 inp->inp_vflag |= INP_IPV6; 1287 inp->in6p_hops = -1; /* use kernel default */ 1288 } 1289 else 1290 #endif 1291 inp->inp_vflag |= INP_IPV4; 1292 tp = tcp_newtcpcb(inp); 1293 if (tp == 0) { 1294 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1295 1296 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1297 #ifdef INET6 1298 if (isipv6) 1299 in6_pcbdetach(inp); 1300 else 1301 #endif 1302 in_pcbdetach(inp); 1303 so->so_state |= nofd; 1304 return (ENOBUFS); 1305 } 1306 tp->t_state = TCPS_CLOSED; 1307 return (0); 1308 } 1309 1310 /* 1311 * Initiate (or continue) disconnect. 1312 * If embryonic state, just send reset (once). 1313 * If in ``let data drain'' option and linger null, just drop. 1314 * Otherwise (hard), mark socket disconnecting and drop 1315 * current input data; switch states based on user close, and 1316 * send segment to peer (with FIN). 1317 */ 1318 static struct tcpcb * 1319 tcp_disconnect(struct tcpcb *tp) 1320 { 1321 struct socket *so = tp->t_inpcb->inp_socket; 1322 1323 if (tp->t_state < TCPS_ESTABLISHED) 1324 tp = tcp_close(tp); 1325 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1326 tp = tcp_drop(tp, 0); 1327 else { 1328 soisdisconnecting(so); 1329 sbflush(&so->so_rcv.sb); 1330 tp = tcp_usrclosed(tp); 1331 if (tp) 1332 tcp_output(tp); 1333 } 1334 return (tp); 1335 } 1336 1337 /* 1338 * User issued close, and wish to trail through shutdown states: 1339 * if never received SYN, just forget it. If got a SYN from peer, 1340 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1341 * If already got a FIN from peer, then almost done; go to LAST_ACK 1342 * state. In all other cases, have already sent FIN to peer (e.g. 1343 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1344 * for peer to send FIN or not respond to keep-alives, etc. 1345 * We can let the user exit from the close as soon as the FIN is acked. 1346 */ 1347 static struct tcpcb * 1348 tcp_usrclosed(struct tcpcb *tp) 1349 { 1350 1351 switch (tp->t_state) { 1352 1353 case TCPS_CLOSED: 1354 case TCPS_LISTEN: 1355 tp->t_state = TCPS_CLOSED; 1356 tp = tcp_close(tp); 1357 break; 1358 1359 case TCPS_SYN_SENT: 1360 case TCPS_SYN_RECEIVED: 1361 tp->t_flags |= TF_NEEDFIN; 1362 break; 1363 1364 case TCPS_ESTABLISHED: 1365 tp->t_state = TCPS_FIN_WAIT_1; 1366 break; 1367 1368 case TCPS_CLOSE_WAIT: 1369 tp->t_state = TCPS_LAST_ACK; 1370 break; 1371 } 1372 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1373 soisdisconnected(tp->t_inpcb->inp_socket); 1374 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1375 if (tp->t_state == TCPS_FIN_WAIT_2) 1376 callout_reset(tp->tt_2msl, tcp_maxidle, 1377 tcp_timer_2msl, tp); 1378 } 1379 return (tp); 1380 } 1381